Pre-Industrial Charcoal Kiln Sites in Wallonia, Belgium: Spatial Distribution, Effects on Soil Properties and Long-Term Fate of Charcoal in Soil

Pre-industrial charcoal kiln sites in Wallonia, Belgium: spatial distribution, effects on soil properties and long-term fate of charcoal in soil.

Brieuc Hardy

Thèse présentée en vue de l’obtention du grade de Docteur en Sciences agronomiques et Ingénierie biologique

Mars 2017

Faculté des bioingénieurs Université catholique de Louvain

Jury composition

Supervisors Pr. Pierre Defourny, Earth and Life Institute, Université catholique de Louvain, Belgium. Pr. J.-T. Cornélis, Department BIOSystem Engineering, University of Liege – Gembloux Agro-Bio Tech.

Members of the examination board Pr. Bruno Delvaux, Earth and Life Institute, Université catholique de Louvain, Belgium (Jury president) Pr. Bas Van Wesemael, Earth and Life Institute, Université catholique de Louvain, Belgium (Jury secretary) Pr. Joseph Dufey, Earth and Life Institute, Université catholique de Louvain, Belgium. Pr. Erik Smolders, Division Soil and Water Management, Katholieke Universiteit Leuven, Belgium Pr. Jens Leifeld, Institute of Environmental Geosciences, University of Basel, Switzerland

Remerciements

Mes premières pensées vont à Joseph Dufey, qui m’a transmis sa passion pour la science du sol, les sorties de terrain et les fameuses « aires de faulde », ces tâches noires qui parsèment les labourés de l’entre-Sambre-et-Meuse et dont il s’est longtemps questionné sur l’origine alors qu’il parcourait les champs de la ferme familiale étant gamin… En m’offrant l’opportunité de poursuivre mon mémoire de fin d’étude en thèse de doctorat, il m’a également offert mon premier travail : assistant au laboratoire de Sciences du Sol de l’UCL. Une sacrée aventure… Mon travail de doctorat a commencé par d’épiques campagnes de terrain aux 4 coins du territoire wallon, par tous les temps et en toute saison, avec comme inconditionnels Joseph et sa Toyota au grand coffre dans lequel toutes les tarières, sondes et autres bêches trouvent place… à côté du frigobox. Même s’il s’est détaché de ses responsabilités administratives dès le début de ses « vacances éternelles » fin d’année 2013, Joseph est resté disponible jusqu’au bout pour me partager sa connaissance, son expérience et son amitié. Travailler avec Joseph est un privilège car il est apprécié de tous pour sa rigueur professionnelle, son humour, son humilité et le respect qu’il porte aux autres. Je remercie mes deux promoteurs, Jean-Thomas Cornélis et Pierre Defourny, pour leur accompagnement au cours de ces six années. Leurs rôles furent complémentaires : Jean-Thomas, très accessible et toujours joignable pour discuter d’un problème, échanger des idées ou m’aider à garder le cap quand la motivation n’était pas au zénith. Pierre, clairvoyant et doté d’une vision d’échelle exceptionnelle qui m’a grandement aidé à structurer mon travail et à faire les bons choix à certains moments charnière du projet. Merci à mes lecteurs, Bas Van Wesemael, Erik Smolders et Jens Leifeld ainsi qu’au président de mon jury, Bruno Delvaux. Si j’ai eu envie de persévérer dans l’environnement de mon mémoire de fin d’étude, c’est essentiellement parce que j’ai tout de suite apprécié la bonne humeur et le côté humain des collègues du labo SOLS. Merci à la grande famille SOLS de représenter un si bon terreau à l’épanouissement. Merci à Françoise, à la barre pour toutes les tâches organisationnelles et administratives. Merci à Anne et Claudine pour leur bonne humeur et les innombrables analyses de laboratoire qui auraient été impossible à réaliser sans elles. Travailler avec elles aux labos didactiques du cours de sciences du sol a été un véritable plaisir. Merci à Patrick et André, qui ont été d’une grande

3 aide pour le travail de terrain et le traitement des échantillons en début de thèse. Merci à David, Hugues et Benoît pour les nombreuses discussions et conseils avisés sur les thèmes du biochar, de la pédogenèse ou des statistiques. Merci à Bruno, Philippe, Sophie et Pierre qui furent des sources d‘inspiration à de nombreux égards. Merci à Aubry, Marie-Liesse, Marie, Yolanda, Rawaa, Zimin, Mathilde, Charles, Inga, François, Carlos, Gabriella, Michelle, Clayria, Elena et Koffi, mes compagnons d’armes, pour les rires et tous les moments partagés au labo et en dehors (combien de litres de café écoulés pendant les pauses ?? Je vous renvoie à la thèse de Marie-Liesse Vermeire pour avoir le chiffre exact. A noter que le café n’est jamais aussi bon que quand c’est Claudine qui l’a préparé). Merci aussi d’avoir soigné mon équilibre alimentaire : impossible d’être carencé en quoique ce soit quand on alterne entre la gastronomie chinoise et libanaise, en passant par des saveurs espagnoles, russes ou équatoriennes. Outre mes collègues directs, ce document n’aurait pu aboutir sans l’aide et le travail d’un grand nombre de personnes qui m’ont fait bénéficier de leurs connaissances, de leurs compétences ou qui m’ont consacré du temps au cours de ces six années. Merci à Richard Lambert et au personnel du centre Agri- environnemental de Michamps pour les premières caractérisations des sols d’étude, à Marco Bravin pour les nombreuses analyses C et N, à Pierre Eloy pour les analyses XPS, à Julien Radoux pour les conseils en télédétection et géomatique, à Koen Deforce pour les identifications de charbons de bois, à Heike Knicker pour les analyses NMR, à Ludivine Van den Biggelaar pour les analyses FTIR, à Nils Borchard pour les échantillons caractérisés au BPCA, à Laurence Ryelandt pour les analyses SEM-EDX, à Steven Sleutel et Caroline Churchland pour les analyses PLFA, à Julien Minet pour ses conseils pour l’utilisation du GPS différentiel, à François Thonon et Nathalie Sondag pour le prêt de leur GPS, à Catherine Rasse, et Alain Guillet pour leurs conseils en statistiques, à Thierry Kervyn pour les données et les informations relatives aux cartes anciennes, à Alain Plante pour les essais d’analyses TG- DSC-EGA, à Margaret Oliver qui m’a aidé à améliorer mon anglais scientifique, à Georges Rousseau qui a régulièrement participé aux campagnes de terrain, aux agents du Département de la Nature et des Forêts (DNF) pour leur contribution au repérage des aires de faulde, aux nombreux agriculteurs qui nous ont laissé échantillonner leurs terres. Une pensée particulière à Vincent Brahy et Patrick Engels du Service Public de Wallonie qui ont été nos principaux répondants lors de la convention qui a couvert les frais de fonctionnement durant mes deux premières années de

4 recherche. Ils nous ont toujours soutenus et continuent à manifester de l’intérêt pour nos travaux. Cette convention a ouvert de nombreuses possibilités pour la suite du projet. Un tout grand merci à Sébastien Françoisse, François-Xavier Henrard et Martin Berwart qui ont réalisé leur travail de fin d’étude sur les aires de faulde et qui ont par ce biais contribué à cette thèse. Sébastien a réalisé un superbe travail de détection des aires de faulde par télédétection qui a permis d’estimer leur nombre en Wallonie. François-Xavier a contribué à l’échantillonnage et aux analyses des charbons de bois prélevés le long d’une séquence de mise en culture. Martin s’est penché sur la problématique de la dynamique de l’azote dans les aires de faulde. Leur travail a généré une partie des données utilisées dans ce document, et les nombreuses interactions que nous avons eues ont égayé mes journées de travail. Je salue Alain Goy et les Bons Cousins de la forêt de Chaux qui nous ont offert un accueil chaleureux et initié à la pratique du charbonnage. La veille de la meule, au milieu des bois et de nuit restera un souvenir intemporel. I also warmly thank Jens Leifeld and Robin Giger from the Agroscope of Zürich for their indispensable help for the numerous DSC and elemental analyzes that were made. Many thanks to the team of the Institute for Sustainability Sciences of the Agroscope of Zürich (Switzerland) for the kind welcome that I received during my two stays at the Agroscope. Particularly, Roman Hüppi and his roommate, Florian Eichenberger, made my stay delighting by offering me bed and board, friendship and visits of Zürich and surroundings. Some great ping-pong games were played at night. We also went tourskiing in the Swiss Alps for one day. Great memories. I really hope to see you sometimes in Belgium for a visit. Merci à Thierry Marique pour la mise en page du document préliminaire, ce qui m’a probablement sauvé du burnout au stade critique de la fin de rédaction. Merci aux membres de la Gibloux beach, Lulu, Karlotta, Brioche, Mat-Mat- la-menace, Alex, Phao, Val et Céline, qui m’ont accompagné durant ces trois dernières années. La cohabitation avec vous a été un plaisir permanent. Merci pour votre amitié et tous ces grands moments partagés. Merci pour les petits plats, apéros, les pauses café-thé, les soirées garage, les jeux de société, les dames blanches, les rallyes chambres, les écognoles, le dentifrice canin au foie de veau, … et d’avoir supporté mes humeurs pendant l’épreuve de la rédaction.

Merci à mes frères et soeur d’avoir toujours été là. Chloé, Clément et Lionel. Merci à Céline d’avoir si bien suivi notre chantier dans les moments chauds durant lesquels j’avais peu de temps et d’esprit à y consacrer. Céline, mon entrepreneuse. Carrelage, parquets, cuisine, peinture… Rien ne l’arrête (à condition que maman biche, son meilleur lieutenant, sois présente… Biche, merci de nous aider sans arrêt pour l’intendance et les travaux)! Merci à mes parents Bénédicte et Jean-Paul, principal sponsor de mes études de bioingénieur qui m’ont donné accès à ce doctorat. L’éducation est la clé des grands enjeux de société d’aujourd’hui et de demain.

List of publications1

Hardy, B., Dufey, J., 2012a. Estimation des besoins en charbon de bois et en superficie forestière pour la sidérurgie wallonne préindustrielle (1750- 1830) Première partie : les besoins en charbon de bois. Revue forestière française 477–487. Hardy, B., Dufey, J., 2012b. Estimation des besoins en charbon de bois et en superficie forestière pour la sidérurgie wallonne préindustrielle (1750- 1830) Deuxième partie : les besoins en superficie forestière. Revue forestière française 799–806. Hardy, B., Dufey, J.E., 2015a. La forêt wallonne, composante vitale de la sidérurgie préindustrielle. Forêt.nature 135, 12–20. Hardy, B., Dufey, J.E., 2015b. Les aires de faulde en forêt wallonne: Repérage, morphologie et distribution spatiale. Forêt.nature 135, 21– 32. Hardy, B., Cornelis, J.-T., Houben, D., Lambert, R., Dufey, J.E., 2016. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium. European Journal of Soil Science 67, 206– 216. doi:10.1111/ejss.12324 Hardy, B., Cornelis, J.-T., Houben, D., Leifeld, J., Lambert, R., Dufey, J., 2017. Evaluation of the long-term effect of biochar on properties of temperate agricultural soil at pre-industrial charcoal kiln sites in Wallonia, Belgium. European Journal of Soil Science 68, 80–89. doi:10.1111/ejss.12395 Hardy, B., Leifeld, J., Knicker, H., Dufey, J.E., Deforce, K., Cornélis, J.-T. Long-term changes of chemical properties of preindustrial charcoal particles aged in forest and agricultural temperate soil. Organic Geochemistry. In press. doi: 10.1016/j.orggeochem.2017.02.008

1 Several additional chapters of the thesis should be published in a near future. Publications “in preparation” are indicated in footnotes at the start of each chapter. 7

List of abbreviations

ANOVA – Analysis of variance AR – Arenosol BC – Black Carbon BPCA – Benzene polycarboxylic acids BS – base saturation 13C-NMR-CPMAS - 13C- nuclear magnetic resonance-cross polarization magic angle spinning CEC – Cation exchange capacity CM – Cambisol DEM – Digital elevation model DN – Digital numbers DSC – Differential Scanning calorimetry DOC – Dissolved organic carbon EC – Electrical conductivity EDX – Energy dispersive X-ray spectroscopy EGA – Evolved gas analysis FTIR – Fourier transform infrared spectroscopy GC-MS – Gas chromatography-mass spectroscopy GCP – Ground control points GNSS – Global navigation satellite system ICP-AES – Inductively coupled plasma-atomic emission spectroscopy ICP-MS - Inductively coupled plasma-mass spectrometry LiDAR – Light Detection and Ranging

Lrad – Satellite radiance LV – Luvisol MS – Multispectral NEXAFS – synchrotron-based near-edge X-Ray absorption fine structure NIR – Near infrared NMR – Nuclear magnetic resonance OC – Organic carbon OM – Organic matter

PAN – panchromatique PCA – Principal component analysis PCR – Principal component regression PLFA – Phospholipid fatty acids PZ – Podzol RMSE – Root mean square error RPC – Rationale polynomial coefficient ρtop – « top of atmosphere » reflectance SEM – Scanning electron microscopy SOC – Soil organic carbon SOM – Soil organic matter TG – thermogravimetry TOC – Total organic carbon USDA – United States Department of Agriculture VIS – visible WEOC – water extractable organic carbon WRB – World Reference Base XPS – X-ray photoelectron spectroscopy XRD – X-Ray diffraction

Table of content Jury composition ...... 1 Remerciements ...... 3 List of publications ...... 7 List of abbreviations ...... 9 General summary ...... 19 Chapter 1. General introduction ...... 25 1.1. Background ...... 25 1.1.1. Functions and ecosystem services of soil organic matter ...... 25 1.1.2. Black carbon: definition, occurrence and environmental significance in soil ...... 26 1.1.3. The BC continuum: a challenge for quantification...... 31 1.1.4. Degradation, transformation and movement of BC ...... 32 1.1.5. Quality of BC: the role of feedstock and conditions of production 33 1.1.6. Environmental drivers of BC storage and degradation in soil ...... 38 1.1.7. Soil amelioration with BC ...... 42 1.1.8. Biochar for environmental management ...... 46 1.1.9. Gap of knowledge ...... 49 1.1.10. Pre-industrial charcoal kiln sites of Wallonia: a model to evaluate the long-term effects of biochar on properties of temperate soils ...... 49 1.2. Aims and objectives ...... 53 1.3. Thesis outline ...... 53 1.3.1. The extent of pre-industrial charcoal production in Wallonia...... 53 1.3.2. Identification and quantification of charcoal-C in the soil of pre- industrial charcoal kiln sites ...... 55 1.3.3. The effect of pre-industrial charcoal production on soil properties...... 55 1.3.4. Stability and dynamics of charcoal in the soil of pre-industrial charcoal kiln sites...... 56 Chapter 2. The extent of pre-industrial charcoal production in Wallonia (Belgium): a historical approach ...... 57 Summary ...... 57 2.1. Introduction ...... 58 2.2. Methodology ...... 59 2.3. Results and discussion ...... 60 2.3.1. The number of blast furnaces in Wallonia and the production of pig iron ...... 60 2.3.2. The demand for charcoal of one average blast furnace ...... 61

2.3.3. The wood for charcoal production ...... 62 2.3.4. The forest area needed for an average blast furnace ...... 63 2.3.5. Extrapolation to Wallonia ...... 63 2.4. Conclusion ...... 64 Chapter 3. Magnitude of pre-industrial charcoal production in Wallonia (Belgium) revealed by LiDAR data and high resolution aerial photographs .... 67 Summary ...... 67 3.1. Introduction ...... 68 3.2. Materials and methods ...... 71 3.2.1. Field database ...... 71 3.2.2. LiDAR data ...... 72 3.2.3. Definition of the potential area for charcoal production ...... 73 3.2.4. Detection of charcoal kiln sites by remote sensing ...... 74 3.2.5. Sampling in forested and deforested areas ...... 77 3.2.6. False positive and false negative detection with LiDAR data ...... 78 3.3. Results and discussion ...... 79 3.3.1. False positive and false negative detection ...... 79 3.3.2. Sites distribution and density...... 81 3.3.3. Distribution, density and covering of charcoal kiln sites in Wallonia ...... 83 3.3.4. Sites diameter and cover ...... 87 3.3.5. Estimate of the number of sites in Wallonia ...... 90 3.3.6. Sources of uncertainty in the Estimates ...... 91 3.3.6.1. Definition of the area of sampling ...... 91 3.3.6.2. Detection ...... 92 3.4. Conclusion ...... 93 Acknowledgements ...... 94 Chapter 4. Sampling campaigns and description of soils and charcoals. ... 95 Summary ...... 95 4.1. Introduction ...... 95 4.2. Sampling campaigns ...... 95 4.2.1. Forest soil (Chapters 5, 6, 7 and 9) ...... 96 4.2.2. Cropland soil (Chapters 5, 6, 8 and 9) ...... 98 4.2.3. Currently active kiln site (Chapter 5, 6, 7 and 10) ...... 100 4.2.4. Charcoals (Chapters 5 and 10) ...... 101 Acknowledgements ...... 105

Chapter 5. Characterization and quantification of charcoal in the soil of pre- industrial charcoal kiln sites by differential scanning calorimetry...... 107 Summary ...... 107 5.1. Introduction ...... 108 5.2. Material and methods ...... 111 5.2.1. Soil samples...... 111 5.2.1.1. 13C nuclear magnetic resonance-cross polarization magic angle spinning (13C NMR-CPMAS) ...... 112 5.2.2. Charcoals ...... 112 5.2.2.1. Elemental composition (C, H, O, N) ...... 113 5.2.2.2. XPS ...... 113 5.2.3. Differential scanning calorimetry analysis ...... 113 5.2.4. Standard addition experiment ...... 114 5.2.5. Sample saturation with Ca2+ ...... 115 5.3. Results ...... 116 5.3.1. Thermal analysis of organo-mineral soils ...... 116 5.3.2. 13C NMR analysis ...... 119 5.3.3. Thermal analysis of charcoals ...... 120 5.3.4. Buffering of forest soils at neutral pH and saturation with Ca2+ .. 122 5.3.5. Quantification of charcoal-C with DSC and comparison with BPCA ...... 123 5.4. Discussion ...... 127 5.4.1. Thermal signature of charcoals ...... 127 5.4.2. Thermal analysis of organo-mineral soils ...... 130 5.4.3. Quantification of charcoal-C by DSC: advantages and limitations...... 131 5.5. Conclusion ...... 134 Acknowledgements ...... 135 Chapter 6. The resistance of pre-industrial charcoal residues to the “Walkley-Black” oxidation...... 137 Summary ...... 137 6.1. Introduction ...... 138 6.2. Material and methods ...... 140 6.2.1. Soil samples...... 140 6.2.2. Total and oxidizable organic carbon content ...... 140 6.2.3. Charcoal-C content ...... 141 6.3. Results ...... 142 6.3.1. Forest soils...... 142 6.3.2. Cropland soils ...... 146

6.3.3. Currently active kiln site ...... 146 6.3.4. Recovery of charcoal ...... 146 6.4. Discussion ...... 147 6.4.1. Recovery of uncharred SOC ...... 147 6.4.2. Recovery of charcoal-C ...... 148 6.4.3. Implications for global SOC budgets ...... 150 6.5. Conclusion ...... 151 Acknowledgements ...... 151 Chapter 7. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium...... 153 Summary ...... 153 7.1. Introduction ...... 154 7.2. Materials and methods ...... 155 7.2.1. Soil samples...... 155 7.2.2. Chemical analyses ...... 156 7.2.3. Data analysis ...... 158 7.3. Results ...... 158 7.3.1. Currently active charcoal kiln site ...... 158 7.3.2. Pre-industrial charcoal kiln sites of Wallonia ...... 159 7.4. Discussion ...... 165 7.4.1. Organic carbon content and quality ...... 165 7.4.2. Cation exchange capacity ...... 166 7.4.3. Soil acidity...... 168 7.4.4. Plant nutrients ...... 170 7.5. Conclusion ...... 173 Acknowledgements ...... 174 Chapter 8. Pre-industrial charcoal kiln sites in Wallonia (Belgium) to evaluate the long-term effect of biochar on temperate cropland soil properties...... 175 Summary ...... 175 8.1. Introduction ...... 176 8.2. Material and methods ...... 177 8.2.1. Soil samples...... 177 8.2.2. Physico-chemical analyzes ...... 178 8.2.3. Quantification of charcoal-C ...... 179 8.2.4. Data analysis ...... 179 8.3. Results ...... 179

8.4. Discussion ...... 188 8.4.1. Charcoal-C, uncharred SOC and N ...... 188 8.4.2. Soil acidity...... 189 8.4.3. Cation exchange capacity (CEC) ...... 190 8.4.4. Exchangeable cations ...... 190 8.4.5. Availability of P, Cu and Zn...... 191 8.5. Conclusion ...... 192 Acknowledgements ...... 193 Chapter 9. The long-term effect of charcoal accumulation at pre-industrial kiln sites on microbial activity, biomass and community structure of forest and cropland temperate soils...... 195 Summary ...... 195 9.1. Introduction ...... 196 9.2. Material and methods ...... 199 9.2.1. Soil samples...... 199 9.2.2. Physico-chemical properties ...... 199 9.2.3. Quantification of charcoal-C content ...... 200 9.2.4. Incubation experiment ...... 201 9.2.5. Microbial biomass and community structure ...... 202 9.2.6. Statistics and data analysis ...... 203 9.2.6.1. Soil properties ...... 203

9.2.6.2. CO2 mineralization rates ...... 203 9.2.6.3. Principal components analysis ...... 203 9.3. Results ...... 204 9.3.1. Soil properties ...... 204 9.3.2. Mineralization rates ...... 206 9.3.3. Microbial biomass and community structure ...... 208 9.4. Discussion ...... 214 9.4.1. The effect of charcoal on CO2 mineralization, soil microbial abundance and activity ...... 214 9.4.2. The effect of charcoal on microbial community structure ...... 217 9.5. Conclusion ...... 220 Acknowledgements ...... 220 Chapter 10. Long-term changes of chemical properties of preindustrial charcoal particles aged in forest and agricultural temperate soil...... 221 Summary ...... 221 10.1. Introduction ...... 222 10.2. Material and methods ...... 224 10.2.1. Charcoals ...... 224 15

10.2.2. Elemental composition, loss on ignition ...... 225 10.2.3. XPS ...... 226 10.2.4. FTIR ...... 227 10.2.5. 13C NMR–CPMAS ...... 227 10.2.6. DSC ...... 227 10.2.7. Dichromate oxidation ...... 228 10.3. Results ...... 228 10.3.1. XPS ...... 228 10.3.2. FTIR ...... 230 10.3.3. Carbonate ...... 231 10.3.4. Organic composition ...... 232 10.3.5. 13C NMR-CPMAS ...... 233 10.3.6. DSC ...... 234 10.3.7. Dichromate oxidation ...... 235 10.4. Discussion ...... 236 10.4.1. Organic composition of charcoal ...... 236 10.4.2. Inorganic composition of charcoal and organo-mineral association ...... 238 10.4.3. Stability of charcoal...... 242 10.5. Conclusion ...... 244 Acknowledgements ...... 245 Chapter 11. Evaluation of carbon stocks at pre-industrial charcoal kiln sites by remote sensing along a chronosequence of a land use change from forest to cropland...... 247 Summary ...... 247 11.1. Introduction ...... 248 11.2. Material and methods ...... 250 11.2.1. Satellite imagery ...... 250 11.2.2. Soil sampling and determination of SOC content ...... 252 11.2.3. Bulk density...... 253 11.2.4. Procedure of estimation of carbon stocks ...... 254 11.2.5. Chronosequence of cultivation ...... 255 11.3. Results ...... 256 11.3.1. The effect of SOC on soil reflectance ...... 256 11.3.2. Prediction of SOC content with soil reflectance ...... 258 11.3.3. Bulk density...... 261 11.3.4. Carbon storage in pre-industrial charcoal kiln sites along a chronosequence of cropping history ...... 261 11.4. Discussion ...... 262 11.4.1. Estimations of SOC stocks by remote sensing ...... 262 16

11.4.1.1. Variability of soil surface ...... 262 11.4.1.2. Instrumental and methological issues ...... 264 11.4.2. Evolution of SOC stocks at charcoal kiln site over time of cultivation ...... 267 11.5. Conclusion ...... 268 Acknowledgements ...... 269 Chapter 12. General conclusions and perspectives...... 271 12.1. Magnitude of pre-industrial charcoal production in Wallonia ...... 271 12.2. Identification and quantification of charcoal in the soil of pre-industrial charcoal kiln sites ...... 273 12.3. The effect of pre-industrial charcoal kiln sites on soil properties, under contrasting soil conditions ...... 275 12.4. Evaluation of the stability and dynamics of charcoal in the soil of pre- industrial charcoal kiln sites, with respect to land use...... 278 12.5. Perspectives ...... 280 References ...... 282 Appendices ...... 313

Summary

General summary

Black carbon (BC), the solid residue of the incomplete combustion of biomass and fossil fuels, is pervasive in soil and contributes to critical functions such as long-term soil carbon sequestration and retention of plant available nutrients. Therefore, soil amendment with artificial BC (biochar) is more and more regarded as a means to mitigate greenhouse gas emissions while improving sustainably soil fertility. Nevertheless, the long-term fate of BC in soil is still poorly understood because changes are slow and complex, with a multiplicity of drivers. Residence time of BC in soil exceeds by several orders of magnitude the lifetime of most laboratory and field trials, which prevents from implementing long-term experiments. In Wallonia, Belgium, intensive in situ charcoal production that was closely linked to preindustrial smelting and steel-making affected a main part of the area that was forested in the late 18th century. Under forest, pre-industrial charcoal kiln sites are slightly heightened domes of about 10 m in diameter, with the topsoil largely enriched with charcoal residues. On bare agricultural soil, they appear as circular or elliptical black spots up to 40 m in diameter, diluted by repeated tillage over time. These sites offer a unique opportunity to evaluate the long-term effect of charcoal enrichment on the properties of temperate soils, over a range of soil conditions. According to this main objective, four specific objectives were addressed: (i) to assess the magnitude of pre-industrial charcoal production in Wallonia; (ii) to develop a procedure to identify and quantify charcoal in the soil of pre-industrial charcoal kiln sites; (iii) to assess the effect of pre-industrial charcoal kiln sites on soil properties, under contrasting soil conditions; and (iv) to evaluate the long-term changes of chemical properties and stability of charcoal particles, with respect to land use To assess the magnitude of pre-industrial charcoal production in Wallonia, we calculated the demand of the smelting industry for charcoal from historical records of the production of pig iron of blast furnaces that were in activity in the late 18th century, when the charcoal-based smelting reached a peak in Wallonia. About three tons of charcoal were necessary to produce and refine one ton of pig iron, and an average blast furnace produced between 500 and 550 tons of pig iron annually. This corresponds to an annual consumption of 1.600 tons of charcoal, or 20.000 steres of wood. Therefore, 1.500.00 steres of wood were needed annually to meet the demand for charcoal of the 73 blast

Summary furnaces that were active in Wallonia in 1790. Given that a coppice forest of 20 years supplies 80 to 100 steres of wood per ha, the forest area equivalent to the demand for charcoal of an average blast furnace was 222 ha annually, or 4.444 ha for a forest rotation of 20 years. In total, 325.000 ha of forest were required to meet the demand of the pre-industrial steel industry for charcoal, which represents 75 % of the forested area on the map of Ferraris (1770– 1778). This highlights the intense pressure exerted on the Walloon forest in the late 18th century through the production of charcoal, which explains for the ubiquity of pre-industrial charcoal kiln sites in areas that were forested at that time. To estimate the number of pre-industrial charcoal kiln sites at the scale of Wallonia, we used remote sensing data to detect the sites in sampling areas generated randomly in the forest mapped by Ferraris (1770–1778). When the sample was located in cropland, the sites were detected on bare soil thanks to four sets of orthoimages, based on the black color of the charcoal-rich soil. In forested areas, charcoal kiln sites were detected based on characteristic relief features appearing on a high resolution digital elevation model (DEM) derived from light detection and ranging (LiDAR) data. Ancient hardwood forest bodies that had been converted to softwood forest were removed from the sampling area to decrease the rate of false negative detection caused by the strong interception of the LiDAR signal by the canopy of some coniferous plantations. In total, we identified charcoal kiln sites in 93.9 % of the sampling plots, with a median site density of 1.2 sites per ha. We estimated a total number of sites between 400 000 and 450.000 for Wallonia. For soil analysis, we sampled 40 pre-industrial charcoal kiln sites from forest and cropland on four different soil types (Arenosols, Cambisols, Luvisols and Podzols). Soil properties were compared to that of adjacent reference soils, unaffected by charcoal production. To trace the evolution of soil properties over time, soil characteristics of pre-industrial charcoal kiln sites were compared to that of a currently active kiln site. Differential scanning calorimetry (DSC) was used to identify, characterize and quantify charcoal in the soil of pre-industrial charcoal kiln sites. Among existing methods to discriminate between BC and uncharred SOC, dynamic thermal analysis has the advantage to be rapid and inexpensive, to require little equipment and sample preparation and to provide a high density of information on the thermal continuum of SOM. Regardless of the dataset, total heat released during analysis was strongly correlated (r > 0.98) to soil organic carbon (SOC) content, which supports the view that the combustion of SOM 20

Summary and charcoal is the main driver of heat fluxes measured by DSC in the soils of this study, with a secondary effect of soil minerals. Our data showed that oxidation through aging decreases the thermal stability of charcoal, and that soil conditions can modify its thermal stability, particularly the content of Ca. Overall, charcoal is more thermally stable than uncharred SOM, which results from the binding energy of C=C bonds from aromatic clusters of charcoal larger than that of C-C, C-H and C-O bonds that predominate in uncharred SOM. Nevertheless, the ranges of thermal stability of charcoal and uncharred SOM overlap, which stresses the issue of BC quantification in soil. Standard additions of charcoal to an organo-mineral soil initially free of charcoal were used to calibrate the relationship between charcoal-C content and an index based on peak height of exotherms from the combustion of charcoal relative to that from the combustion of uncharred SOM. This index was used to estimate the content of charcoal-C in the soil of pre-industrial charcoal kiln sites. The content of charcoal and uncharred SOM was related to the content of plant available nutrients, acidity, cation exchange capacity and microbial activity, biomass and community structure of soil. The charcoal-rich topsoil of pre-industrial kiln sites has a larger C:N ratio, C:P ratio and cation exchange capacity (CEC) per unit of organic carbon than the reference soil. The CEC of fresh charcoal is initially poor, but surface oxidation over time creates abundant carboxyl groups that are partly to completely deprotonated at soil pH, which explains for the large CEC of aged charcoal. In forest, the largest CECs per unit of organic carbon were observed on soil with coarser textures. We estimated from cropland soils that the -1 average CEC per unit of C is 414 cmolc kg for charcoal, which is about twice -1 the CEC per unit of C of 213 cmolc kg estimated for uncharred SOM. The CEC of pre-industrial charcoal at kiln site remains considerably smaller than that estimated for BC in millennial Amazonian terra preta soils, which suggests that the CEC of charcoal might continue to increase by progressive oxidation over longer periods of time. In cropland, we measured a small increase in nitrates in the kiln soil that might relate to greater mineralization and nitrification of organic N, as the kiln soil contain significantly more uncharred SOM than adjacent reference soils. On acidic forest soils, base saturation increases in the subsoil of pre-industrial charcoal kiln sites, which reflects the past liming effect of ash produced by wood pyrolysis. The topsoil is reacidified but soil pH remains, however, slightly higher than that of adjacent reference soil. In contrast, soil pH of kiln sites located on carbonate-rich forest soil and cropland soil that are frequently

Summary limed is slightly decreased. This results possibly from an excess of variable charges from acidic functions on the surface of charcoal, so that lime requirement increases in the kiln soil to reach the same base saturation as in adjacent reference soil. Larger nitrification rates, as suggested by the enrichment of nitrates, might also decrease the pH of the kiln soil. Under forest, the kiln soil is greatly depleted in exchangeable K+ and available P regardless of soil type. The small concentration of exchangeable K+ is attributed to the small affinity of the exchange complex of charcoal for K+. Consequently, K+ is preferentially lixiviated from the topsoil of kiln sites in contrast to exchangeable Ca2+ that forms strong complexes with carboxyl groups at the surface of charcoal. The low concentration of available P is attributed to a decrease in P availability with time and to the fact that aged charcoal does not take part in the biological cycling of nutrients because of its low degradability. This underlines that charcoal does not support the same ecological functions as uncharred organic matter. Charcoal can also form strong complexes with Cu, which reduces its availability. Charcoal had a limited effect on microbial community structure, particularly in cropland where soil conditions are modified by the application of organic and inorganic fertilizers. We highlighted that the content of uncharred SOM and pH explain a main part of the variability of CO2 emissions from soil, regardless of the presence of charcoal. This supports the view that on the long- term, when the labile fraction of charcoal has been completely degraded, the effect of charcoal on microbial properties is indirect, mainly related to a modification of soil conditions such as acidity and, possibly, the availability of nutrients. To assess the effect of cultivation on the long-term dynamics of BC in soil, we investigated chemical properties of charcoal particles extracted from soil along a chronosequence of land use change from forest to cropland, up to 200 years of cultivation. Cultivation increased association of charcoal with soil minerals, favored by deprotonation of carboxylic acids under liming. The O- rich, less thermally stable fraction of charcoal decreased with time of cultivation, leading to the relative increase of the thermally most stable fraction of charcoal. This might result from the preferential loss of the O-rich fraction or the slowdown of charcoal from oxidation by association with minerals. The resistance of charcoal to dichromate oxidation decreased with both ageing and cultivation, related to the H:C of charcoal. Our results highlighted that land use significantly affects the properties of BC through the

Summary modification of soil conditions, which might influence the kinetics of BC loss from soil. Overall, we highlighted that the features of charcoal and its effect on soil properties evolve over time and depend on soil conditions. As BC is very persistent in soil, it is crucial to better assess how plant growth and crop yields respond to biochar on the long-term before allowing its large-scale application.

Chapter 1. Introduction

Chapter 1. General introduction

1.1. Background

Functions and ecosystem services of soil organic matter

Soil organic matter (SOM) plays a central role in chemical, physical and biological properties that are essential to some of the most important functions and ecosystem services of soil, such as biomass production, nutrient cycling, water purification and carbon sequestration (Baveye et al., 2016). As prime ecosystem service, soils support a main fraction of the global primary production and nearly all terrestrial food production. SOM influences soil fertility in multiple ways. It contributes largely to the structural stability of soil by interacting with minerals to form stable aggregates that promote soil aeration and drainage, increase water holding capacity and improve the resistance of soil to erosion and compaction. SOM is a reservoir of energy and nutrients for heterotrophic organisms and therefore supports a vast array of biologically mediated processes. By decomposition, SOM recycles macro- and micronutrients indispensable for plant growth, with a relative accumulation of most-limiting nutrients in the topsoil (Jobbagy and Jackson, 2001). SOM also favors the retention of nutrients in the rhizosphere by preventing them from lixiviation. Indeed, SOM has a large cation exchange capacity (CEC) related to the variable charge of organic functional groups such as carboxyl that provide a negative charge to SOM in the range of pH of soil. Accordingly, the content of SOM is considered as the most important indicator of soil health and quality in long-term agricultural soil systems (Reeves, 1997). Composed of 40–60 % of carbon (C), SOM is the largest terrestrial reservoir of organic C as it contains more C than the atmosphere and the biosphere combined (Houghton, 2007). Risks associated to the rapid change in climate and global temperature warming are among the main current environmental concerns of human societies, and reinforcement of atmospheric radiative forcing by anthropogenic greenhouse gas emissions since the pre-industrial era has been identified as the most important contribution to global warming (IPCC, 2014). Therefore, the function of soil to act as a large, dynamic reservoir of SOC that can be a sink or source of greenhouse gases has received much recent attention (e.g. Davidson et al., 2006; Schmidt et al., 2011; Todd- Brown et al., 2014). Batjes (2016) provided the most recent estimate of global SOC stock, based on an updated harmonized dataset of soil properties derived 25

Chapter 1. Introduction from 21.000 soil profiles for the world, and soil mapping units delimited by crossing the Harmonized World Soil Database and the Köppen-Geiger climate zones (Peel et al., 2007). As a result, he found that soil stores 2060 PgC (Pg = Petagram = 1015 g) in the uppermost two meters (Batjes, 2016), which accords with former estimates of about 1500 PgC for the first meter of soil (Batjes, 1996; Hiederer and Köchy, 2012; Jobbagy and Jackson, 2000). Decomposition of SOM (including litter) accounts for ~58 PgC yr-1 emitted from soil in the form of CO2 and CH4 (Houghton, 2007), which is 6.2 times more than the contribution of fossil fuel combustion (9.3 PgC yr-1) to global emissions. The consumption of atmospheric CO2 by photosynthesis responsible for the return of C to the SOM pool through biomass residues and the leaching of dissolved organic carbon from soils to oceans accounts for ~59 PgC yr-1 (Houghton, 2007), which compensates the emissions. Therefore, direct (e.g. land use changes, reduced tillage) or indirect (global warming) anthropogenic perturbations of the input and output of SOM are important drivers of atmospheric concentrations of greenhouse gases that influence climate.

Black carbon: definition, occurrence and environmental significance in soil

Black carbon (BC), also referred to as ‘pyrogenic carbon’, is an important component of the soil organic carbon (SOC) pool because it is pervasive in soil and has particular properties, such as its highly refractory character. Based on an inventory of literature, Reisser et al. (2016) estimated that BC accounts for 13.7 % of SOC content on average and up to 60 %, making it one of the largest group of compounds in SOM. BC is the solid, thermally altered residue of the incomplete combustion of biomass and fossil fuels. It comprises a wide range of products from slightly charred biomass to highly refractory condensates such as soot, produced by natural or human-induced fire and pyrolysis (Goldberg, 1985; Schmidt and Noack, 2000a). Despite the diversity of forms and quality of BC materials, the highly aromatic structure of BC (Figure 1.1) is a characteristic feature that makes it different from any other organic compounds if we exclude organic matter that has been transformed by geogenic processes such as coal and graphite. BC varies in size from large macroparticles to individual molecules and is omnipresent in aerosols, waters, sediments and soil (Bird et al., 2015; Kuhlbusch, 1998; Preston and Schmidt, 2006; Schmidt and Noack, 2000a).

Chapter 1. Introduction

For the sake of clarity, we defined four terms that refer to thermally altered biomass (black carbon, pyrogenic carbon, charcoal and biochar) and that will be used abundantly throughout the manuscript in the Toolbox 1. Toolbox 1: Black carbon, pyrogenic carbon, charcoal and biochar

Black carbon or pyrogenic carbon, is the solid, thermally altered residue of the incomplete combustion or pyrolysis of biomass and fossil fuels. It comprises a wide range of products from slightly charred biomass to highly refractory condensates such as soot, produced by natural or human-induced fire and pyrolysis. Despite the diversity of forms and quality of BC materials, the highly aromatic structure of BC is a characteristic feature that makes it different from any other organic compounds if we exclude organic matter that has been transformed by geogenic processes such as coal and graphite.

The oldest BC particles identified in geologic records date from 420 Myr (Cressler, 2001), shortly after the development of the earliest vascular plants, when the concentration of atmospheric O2 has become sufficient to support fire. Since then, the occurrence of BC residues in archeological records attest that history of fire (and associated production of BC) has been governed by the main controls of fire activity like atmospheric O2 concentration, climatic conditions, vegetation, catastrophic events and human activity (Scott and Glasspool, 2006). Wildfire is the largest contributor to the global production of BC, with about 3 % of the carbon of the burnt ecosystem left behind (Forbes et al., 2006). Global production of BC by fire was estimated at 50–270 Tg yr- 1 (Kuhlbusch and Crutzen, 1996), which accords with more recent estimations (Forbes et al., 2006; Kuhlbusch, 1998).

Chapter 1. Introduction

Figure 1.1. Basic structural units and two principal structures of black carbon (BC), by Schmidt and Noack (2000); a) chemical structure of BC as formed in the laboratory (Sergides et al., 1987); b) basic planar structure of micrographitic crystallites of three to four layers in BC (Heidenreich et al., 1968); c) Schematic draw of the chemical structure of BC, with randomly oriented micrographitic crystallites in an amorphous matrix of aromatic C; d) Onion-type soot particle with several condensation seeds (Heidenreich et al., 1968). The different shapes of BC particles result from different hybridization states of C atoms. Planar and curved structures correspond to sp² and mixed sp²-sp³ hybridization states, respectively. All BC is produced in terrestrial environments; therefore, a major fraction of BC ends-up in soil, which must be the largest terrestrial reservoir of BC (Forbes et al., 2006). Masiello and Druffel (1998) estimated with radiocarbon measurements that BC found at two deep ocean sites was 2.400 and 13.900 yr older than uncharred SOC deposited concurrently. This supports the idea that BC had been stored in an intermediate reservoir, and soil is very likely to be this reservoir. BC has received increasing interest in recent years for its implication in a range of biogeochemical processes (Schmidt and Noack, 2000a), particularly for its potential to act as an important sink in the global carbon cycle because of the refractory nature of a fraction of BC (Czimczik

Chapter 1. Introduction and Masiello, 2007; Knicker, 2011a; Masiello, 2004; Preston and Schmidt, 2006; Schmidt and Noack, 2000a). Nevertheless, the cycle of black carbon is incompletely understood, and the current estimates of the stocks and fluxes between the atmospheric, oceanic and terrestrial reservoirs (Figure 1.2) are poorly constrained (Bird et al., 2015; Forbes et al., 2006). Based on the Harmonized World Soil Database (HWSD) estimates of SOC stocks (Hiederer and Köchy, 2012) and estimates of the contribution of BC to SOC stocks for the main IPCC climate regions of the world, Bird et al. (2015) calculated that soil stores 54–109 Pg in the uppermost meter. Reisser et al. (2016) build up a soil BC database including more than 560 measurements from 55 published studies. Used in combination with other datasets, they estimated from their database that BC stocks are around 200 Pg in the two uppermost meters of soil. Despite the fact that these recent estimates provide a plausible order of magnitude, the contribution of BC to the soil carbon pool is currently very uncertain because (i) data at global scale are scarce and of limited spatial coverage, (ii) BC content varies with depth and (iii) BC measurements with different techniques are inconsistent with each other (Bird et al., 2015; Forbes et al., 2006).

Figure 1.2. The pyrogenic carbon cycle by Bird et al. (2015). Estimates of the major sources (TgC yr-1), reservoirs (PgC or TgC) and fluxes (TgC yr-1) are represented. Numbers in parentheses indicate the total range of estimates.

Chapter 1. Introduction

The BC continuum: a challenge for quantification.

The discrepancy between estimates of BC content by different quantification procedures has for main cause that BC includes a wide range of materials from slightly charred plant biomass to highly condensed soot that cover a wide continuum of composition and properties (Figure 1.3) (Goldberg, 1985; Hedges et al., 2000; Masiello, 2004). BC has a broad significance in soil, sediment and atmospheric science, ecology, archeology and geology. Consequently, many procedures of BC quantification were set up at different times by scientists from different disciplines, to respond to different questions. Hence, existing quantification procedures target distinct components of the BC continuum (Hammes et al., 2007) and rely on operational definitions that can either overestimate or underestimate the amount of BC as defined earlier (i.e. the solid residues of the incomplete combustion of biomass) (Masiello, 2004). Another issue originates from the fact that the elemental, physical, chemical and spectroscopic properties of slightly charred biomass overlap with that of uncharred biomass, which hinders the separation.

Figure 1.3. The concept of black carbon (BC) continuum as illustrated by Hammes et al. (2007), and elemental H/C and O/C ratios according to the degree of alteration of biomass. Techniques of identification and quantification of BC are either destructive or non-destructive and can be divided into four classes. These include (i) physical separation of BC particles based on a difference of density or size; (ii) separation by preferential oxidation of uncharred SOC by chemical, thermal or radiative (UV photooxidation) treatments applied alone or in combination, followed by quantification of BC in the residue; (iii) identification of BC by spectroscopic methods such as 13C nuclear magnetic resonance (13C NMR), or Fourier-transform infrared spectroscopy (FTIR), based on the chemical signature of BC; and (iv) molecular marker techniques that consist in decomposing a sample either chemically (benzene polycarboxylic acids (BPCA) method; Brodowski et al., 2005) or thermally (Hydrogen Pyrolysis; McBeath et al., 2015) and measuring the type and abundance of BC markers that are liberated, by gas or liquid chromatography. Schmidt et al., (2001) 31

Chapter 1. Introduction reported variations in the recovery rates of BC up to > 2 orders of magnitude for one individual sample analyzed by four different quantification techniques. Accordingly, the choice of an adapted procedure requires judgment in line with the objectives of the analysis and the form of BC in the sample. In this goal, Hammes et al. (2007) made an intercomparison study of several of the most commonly used methods of BC quantification in soil and sediments, based on 12 reference materials of BC and potentially interfering substances. This provides an overview of the advantages, inconveniences and the fraction of the BC continuum targeted by each procedure.

Degradation, transformation and movement of BC

BC has long been regarded as a poorly reactive, (bio-)chemically inert component of the carbon cycle because of its presence in the sedimentary records back to the Devonian, in glacial and lacustrine sediments of the late Quaternary and in terrestrial environments over thousands to millions of years (Schmidt and Noack, 2000a). For that reason, BC is used as a tracer of high significance in ecology, archeology and geology to unravel Earth’s fire history (Bird and Cali, 1998) and palaeoenvironments (Scott and Damblon, 2010). However, the content of BC stored in soil is low regarding annual production rates from forest wildfires, which provides evidence for important losses of BC from soil (Masiello, 2004; Schmidt, 2004; Schmidt and Noack, 2000a). The concept of BC continuum was key to reconcile the occurrence of extremely old BC particles and evidences of rapid losses of some BC, as it supports the view that BC represents a range of materials with a range of degradation potentials by a range of mechanisms (Bird et al., 2015). Among processes involved in BC loss, microbial decomposition has been proved to play a role (Baldock and Smernik, 2002; Hamer et al., 2004; Wengel et al., 2006), possibly by way of co-metabolism with labile organic molecules (Hamer et al., 2004; Kuzyakov et al., 2009). Large amounts of terrestrial BC can also be transported from soil through erosion (Rumpel et al., 2006) or dissolution (Hockaday et al., 2007; Jaffé et al., 2013) and build-up C stocks of riverine and oceanic waters and sediments. The residence time of BC in soil is the result of a combination of factors including the initial properties of the feedstock (grain size, chemical composition, ash content), the conditions of production (temperature, heating rate, time of heating, pressure, humidity, oxygen supply) and the strength of the drivers of BC degradation in environmental conditions (temperature, humidity, pH, texture, mineralogy, microbial activity, land use, …). Hence,

Chapter 1. Introduction

BC materials have contrasting degradation potentials by nature, but the kinetics of degradation will depend on environmental conditions where BC is deposited (Bird et al., 2015). A minor, labile fraction of BC is readily available for microbial decomposition in the very short-term (Kuzyakov et al., 2009; Sagrilo et al., 2014). This fraction is made of products derived from the incomplete transformation of cellulose and proteins such as anhydrosugars or methoxylated phenols and tend to decrease with temperature of pyrolysis (Fabbri et al., 2012). Therefore, a two pool approach with a labile BC pool with a short turnover and a stable BC pool with a long turnover is generally used to model the degradation of BC (Foereid et al., 2011). The stable fraction of BC is mainly aromatic, which is a key feature for the identification of BC with spectroscopic techniques such as 13C NMR (Czimczik et al., 2003; Haumaier and Zech, 1995; Knicker, 2011a; Laird et al., 2008; Solomon et al., 2007b). Nevertheless, transmission electron micrographs of modern and fossil charcoal samples provided evidence for the presence of disorganized domains in the matrix of a same BC particle but also organized domains composed of graphite-like microcrystallites (Cohen-Ofri et al., 2006). This illustrates the fact that BC can have distinct degrees of cristallinity and aromatic condensation (referring to the size and organization of aromatic clusters), which explains for the stability of BC in the environment better than the raw content of aromatics (McBeath and Smernik, 2009; Wiedemeier et al., 2015).

Quality of BC: the role of feedstock and conditions of production

Among the different type of materials included in the definition of BC, charcoal is intentionally produced by pyrolysis in controlled conditions to be used as a fuel. Charcoal has been produced for millennia for various domestic or industrial uses. It is the most valued reductant of the metallurgical industry because of it contains less S, Hg and N than fossil fuels and is characterized by a high specific surface area and reactivity (Antal and Grønli, 2003). Because of the economic importance of charcoal and other products of wood distillation, the process of pyrolysis has been extensively studied to optimize yields and design pyrolysis product with desirable properties, such as activated charcoals. The correct understanding of chemical transformations occurring during pyrolysis and the influence of pyrolysis conditions on the properties of BC are key to understand the reactivity of BC in the environment. The main constituents of vascular plants are cellulose, hemicellulose and lignin. The first two components are polysaccharides whereas lignin is a complex, heterogeneous biopolymer with no defined primary structure, 33

Chapter 1. Introduction commonly composed of three molignol monomers (p-coumaryl alcohol, coniferyl alcohol and synapil alcohol) (Adler, 1977). During pyrolysis, the molecular framework of the sugar moieties composing biomass is grossly rearranged to form aromatic structures (Antal and Grønli, 2003). In the first stage of the reaction, cellulose is thermally activated and transformed by dehydration, depolymerisation (cracking of the glucosidic bonds) and ring opening into anhydrocellulose, levoglucosan and tarry vapors that contain a complex soup of organic compounds (Figure 1.4). These are further rearranged into an aromatic structure by reforming and condensation reactions (Antal and Grønli, 2003; Liao et al., 2004). Furans, pyranones and benzofurans are among the small aromatic units formed in the first stages of the process of aromatization (Knicker, 2011a).

Figure 1.4. Mechanism of pyrolysis of cellulose of Brodio-Shafizadeh, modified by Liao et al. (2004) The same mechanisms are expected to occur during the pyrolysis of hemicellulose. The pathways of transformation of lignin by pyrolysis are more complex because of the high degree of heterogeneity of lignin. Transformation involves reactions of dehydration and depolymerization, as well as reactions of demethoxylation, the formation of biphenyls and cleavage of aryl ethers for a complete loss of O-alkyl C structure at 450 °C (Knicker, 2010). Yang et al. (2007) used thermogravimetry and differential scanning calorimetry (TG- DSC) coupled to evolved gas analysis (EGA) to unravel the changes occurring during pyrolysis by following mass loss, energetic flux and gas release from cellulose, hemicellulose and lignin powders. Materials were pyrolysed under -1 a N2 and a heating rate of 10 °C min , up to 900 °C. They found that the

Chapter 1. Introduction transformation of hemicellulose and cellulose occur quickly, with the main weight loss at 220–315 °C for hemicellulose and at 315–400 °C for cellulose. Lignin ignites rapidly, but decomposition occurs on a wide range of temperatures. Among the three compounds, lignin contributes the most to the formation of BC as 60 % of the initial weight remains at 700 °C against ~15 % for hemicellulose (Yang et al., 2007). In contrast, loss weight of cellulose was 95 % at 350 °C and almost complete at 400 °C, which indicates that its contribution to BC formation is negligible for temperatures > 400°C in the conditions of reaction (Yang et al., 2007). These results underline that both the composition of the biomass (proportion of cellulose, hemicelluloses and lignin) and the conditions of reaction are of prime importance for the final properties of BC.

Figure 1.5. Effects of temperature and heating rate on (a) yields and (b) elemental contents of C, H and O of beech chars (data from Schenkel (1999), redrawn by Antal and Grønli (2003)). The solid and dashed lines represent heating rates of 2 and 10 °C min-1, respectively. Pressure, heating rates, oxygen content and maximum temperature of pyrolysis are the factors that influence the yield and quality of BC the most. Pressure accelerates the reaction and improves the yield of charcoal production by decreasing the loss of biomass in the form of volatile tars. These tarry vapors condensate at the surface of charcoal to form secondary charcoal and release H2O, CO, CO2, CH4 and H2 (Antal and Grønli, 2003). Under high pressure, charcoal yields approach the theoretical maximum determined by thermochemical equilibrium calculations (Antal and Grønli, 2003). Maximum pyrolysis temperature is certainly the most important driver of quality and properties of BC. Antal and Grønli (2003) reviewed the literature on the technology of charcoal production and provided a complete summary of the

Chapter 1. Introduction effect of increasing pyrolysis temperature on basic properties of charcoal. Overall charcoal yield decreases with increasing temperature of pyrolysis (Figure 1.5a), concomitantly to the continued loss of O and H atoms (Figure 1.5b), corresponding to the purification of the carbonaceous backbone from O- and H-rich functional groups during the process of aromatization and the progressive growth of aromatic clusters (Budai et al., 2014; Wiedemeier et al., 2015). Accordingly, the mass fraction of C increases (Figure 1.5b), as well as the content of fixed C (defined as the content of C remaining after heating the sample at 950 °C for 6 minutes in a closed vessel; in contrast, volatile C is defined as the content of C that was lost during heating). A rapid heating rate tends to increase losses by gasification at the expense of charcoal yield, and can decrease the mechanic resistance of charcoal (Antal and Grønli, 2003). Whereas charcoal yields were the main driver of early research on charcoal, there is more and more evidence that the degree of aromatic condensation of BC is critical for the intrinsic stability of a BC particle deposited in the environment. Pyrolysis starts at ~280 °C and wood components have entirely lost their initial feature at 400–450 °C (Keiluweit et al., 2010), with a complete transformation into aromatics (Keiluweit et al., 2010; McBeath et al., 2015). Nevertheless, aromatic C in low temperature chars mainly occurs in an amorphous, disorganized form whereas organized domains of micrographitic sheets (Cohen-Ofri et al., 2006) grow at higher temperature (Keiluweit et al., 2010; McBeath et al., 2015; McBeath and Smernik, 2009). The average size of aromatic clusters that tends to increase with temperature of pyrolysis is commonly referred to as the degree of aromatic condensation of BC (Figure 1.6). At comparable aromaticity, resistance to thermal degradation depends mainly on the degree of aromatic condensation of char (Harvey et al., 2012; Leifeld, 2007), which supports the idea that aromatic condensation reflects the stability of BC exposed to environmental conditions better than the aromaticity itself.

Chapter 1. Introduction

Figure 1.6. Structure of black carbon relative to maximum temperature of pyrolysis (source: Chia et al., 2015); a) highly disordered aromatic C in amorphous mass; b) growing sheets of micrographitic crystallites; c) Structure becomes graphitic with order in the third direction. Keiluweit et al. (2010) used a variety of spectroscopic techniques (Fourier transform infrared spectroscopy (FTIR), X-ray diffraction (XRD) and synchrotron-based near-edge X-Ray absorption fine structure (NEXAFS) to highlight physical and chemical transitions and progressive changes occurring over pyrolysis of wood and grass biomass, from 100 to 700 °C. From their measurements and previous literature, they proposed to consider four different categories of chars according to the degree of transformation of original biomass into amorphous aromatics or turbostratic microcristallites, largely governed by the maximal temperature of pyrolysis. To explain the variability of the intrinsic resistance of BC to degradation, Bird et al. (2015) proposed to split the BC continuum into three different levels of contrasting (bio- )degradability: (i) a minor, labile fraction of BC made of products derived from the incomplete transformation of cellulose and proteins, such as anhydrosugars or methoxylated phenols (Fabbri et al., 2012); (ii) a semi-labile aromatic pool composed of amorphous, disorganized aromatic clusters of small size (< 7) and (iii) stable polycyclic aromatic C pool made of aromatic clusters of a large size. The approach of Bird et al. (2015) relies on the observation of organized and disorganized domains visible at nanometric scale by transmission electron micrograph (Cohen-Ofri et al., 2006), supporting the idea that micrographitic structures crystalize in the matrix of 37

Chapter 1. Introduction amorphous aromatics when temperature of pyrolysis is sufficient, and grow up with increasing temperature. They advocate that 13C NMR allows to detect (and estimate the content of) the whole aromaticity (e.g. McBeath et al., 2011) and that hydrogen pyrolysis, in line with diamagnetic ring current NMR results (McBeath and Smernik, 2009), is a reliable indicator of the stable polycyclic aromatic C content by selecting clusters made of 7 or more aromatic rings (McBeath et al., 2015; Wurster et al., 2013), assumed to confer to BC its refractory character. Threshold of maximum aromaticity is reached from 450 °C whereas the main shift in resistance to hydrogen pyrolysis occurs between 450 and 600-700 °C (McBeath et al., 2015). The quality of initial feedstock, by its organo-chemical composition and ash content, can also influence the yield and quality of BC and shift the temperature of chemical transitions and progressive changes occurring over pyrolysis (Antal and Grønli, 2003; Keiluweit et al., 2010; Schmidt and Noack, 2000a), which is expected to affect the degradability of BC (Hamer et al., 2004).

Environmental drivers of BC storage and degradation in soil

Recent experimental evidences have demonstrated the crucial role of environmental drivers on the persistence of organic matter in soil, supporting the view that the residence time of SOC is governed predominantly by ecosystem properties rather than molecular structure (Schmidt et al., 2011). This statement might also apply for chars to some extent, despite the refractory character of the more condensed fraction of BC. Immediately after pyrolysis, BC is highly reactive because it has a high specific surface area and because many chemical bonds are left dangling (Antal and Grønli, 2003). Speaking of charcoal, Antal and Grønli (2003) warn that significant amounts of oxygen and moisture can be chemisorbed during storage. Accordingly, BC has a strong affinity for organic compounds (Pignatello et al., 2006), but the affinity decreases with aging due to saturation of reactive sites (Cheng et al., 2014). In regards to oxido-reduction reactions in soil, SOM acts as a reactor that supplies electrons to a number of more oxidized species present in soil through decomposition (Chestworth, 2004). Despite being more refractory than other types of SOM, BC is a meta-stable material that is thermodynamically unstable under the oxidative conditions of most surface soils (Joseph et al., 2010). Aging of BC mainly consists of oxidation reactions of exposed C rings with a high density of π electrons and free radicals (Joseph et al., 2010), which creates a high density of oxygenated functional groups at the surface of BC over time (Cheng et al., 2006). Characterization of functional groups of BC

Chapter 1. Introduction particles aged in soil for a centennial to millennial period of time, by Boehm titration (Cheng et al., 2006) or synchrotron-based near-edge X-ray absorption fine structure (NEXAFS) spectroscopy (Lehmann et al., 2005), has shown a dominance of carboxylic and, to a lesser extent, phenolic groups that decrease from the center to the exterior of the particle (Lehmann et al., 2005). The drivers of oxidation can be biotic (Baldock and Smernik, 2002; Hamer et al., 2004; Wengel et al., 2006) but are mainly abiotic, at least shortly after introduction of BC to soil (Cheng et al., 2006). Adsorption of dissolved organic molecules can contribute to the increase of O content at the surface of

BC (Lehmann et al., 2005). In absence of O2, secondary electron acceptors (e.g. Mn (hydr-)oxides) might contribute to oxidize aromatic C (Chestworth, 2004; Nguyen and Lehmann, 2009). Aging causes dramatic changes in surface properties of BC. Through oxidation processes, the surface of BC becomes more and more polar, which decreases dramatically its hydrophobicity (Criscuoli et al., 2014; Knicker, 2011a) and might promote further physical, chemical and microbial weathering (Hammes and Schmidt, 2009). Cheng et al. (2008) investigated the surface positive and negative charge of BC over a wide range of pH by the “indifferent” ion adsorption method (Uehara and Gillman, 1981), using a 0.01 M KCl solution. They compared surface charge of freshly produced BC to that of BC aged in laboratory conditions during twelve months at 30 and 70 °C and centennial BC particles from the remnants of historical blast furnaces to unravel the evolution of surface charge over aging (Figure 1.7; Cheng et al., 2008). They found that fresh BC has a significant anion exchange capacity (positive charge) and a point of zero net charge (PZNC; the pH of the intercept point between negative and positive charge curves) of ~7, which corresponds to a very low CEC (negative charge) compared to other types of SOM. However, after twelve months of incubation, the positive charge of BC has almost disappeared, and negative charge has increased dramatically by progressive oxidation (Figure 1.7). Positive charge of centennial BC is negligible regardless of pH, and negative charge at pH 7 is about 10 fold larger than for BC incubated at 70 °C for one year, which indicates that oxidation progresses continuously over time in soil (Cheng et al., 2008a).

Chapter 1. Introduction

Figure 1.7. Surface positive charge (triangles) and surface negative charge (circles) against pH for a) fresh laboratory-produced black carbon (BC); b) BC incubated for 12 months at 30 °C; c) BC incubated for 12 months at 70 °C; d) Centennial charcoal particles from the remnants of historical blast furnace in Port Henry, New York. Note the different scale of the y axes between new and centennial BC. The figure was adapted from Cheng et al. (2008). Once BC has been released in the environment, environmental conditions can influence the kinetics of BC transformation in soil and thereby affect its residence time in soil. Climatic conditions can play an important role on the kinetics of BC alteration. An increase of the temperature of incubation has been shown to accelerate the oxidation of BC in laboratory conditions (Cheng et al., 2008a; Cheng and Lehmann, 2009; Nguyen et al., 2010), regardless of the temperature of production of BC (Nguyen et al., 2010). Accordingly, BC storage at the sites of historic charcoal blast furnace across a climatic gradient in North America decreased with increasing mean annual temperature, which supports the idea that the effect of temperature on the kinetics of BC transformation is applicable in field conditions (Cheng et al., 2008c). Like temperature, moisture is another crucial driver of microbial activity and a carrier of dissolved reactants that can have an influence on the kinetics of BC alteration. In an incubation experiment, Nguyen and Lehmann (2009) studied the effect of a range of water regimes (saturated, unsaturated and alternating saturated-unsaturated conditions) on the evolution of the properties of corn

Chapter 1. Introduction and oak BC produced at 350 and 600 °C. The faster changes (increase of carboxyl and hydroxyl groups, loss of aliphatic C) were obtained under unsaturated and alternating water regimes. Soil conditions have also been identified as important drivers of BC storage in soil, particularly clay content and pH. Despite the fact that the persistence of BC in soil is generally explained by the refractory character of BC, stabilization by association with minerals might be one of the most important mechanism of BC storage in soils affected by wildfires (Czimczik and Masiello, 2007; Reisser et al., 2016). There is some evidence that BC interacts with soil minerals (Brodowski et al., 2005a; Glaser et al., 2000; Nguyen et al., 2008) through mechanisms that might be might be similar to organo-mineral associations of uncharred SOM (Czimczik and Masiello, 2007). Based on a global inventory made of 55 studies, Reisser et al. (2016) found a significant effect of clay on BC content (expressed as a fraction of total SOC content), with clay-rich soils (> 50 % of clay) containing on average > twice more BC than coarse-textured soils (< 5 % of clay). By oxidation in soil, the surface of BC particles develops carboxyl and phenol functionalities that interact with mineral surfaces through complexation with metallic cations from the surface of minerals. Stabilization can be more effective when multiple carboxyl groups occur, thereby, the high density of functional groups at the surface of BC related to its large specific surface area can make organo-mineral stabilization of BC even stronger than that of uncharred SOC (Czimczik and Masiello, 2007). Abundant associations with minerals in combination with a refractory aromatic core might make BC storage very effective in clay-rich soils. In their inventory, Reisser et al. (2016) observed that soils with high pH values (pH > 7) had the largest BC content whereas acidic soil (pH < 5) had the lowest BC content. High pH goes together with high concentrations of alkaline and alkaline earth cations, including Ca2+. Therefore, stabilization of BC by interaction with minerals through Ca-bridging has been proposed to explain the high concentrations of BC found in many Mollisols and Chernozerms (Czimczik and Masiello, 2007). Nevertheless, these soils with a relatively high primary production and subject to a high fire frequency might also be subject to large inputs of BC. Any other climatic, pedogenic or human factor that can influence the kinetics of physical, microbial or chemical alteration of BC is likely to play a role in the persistence of BC. For instance, mechanical disturbance by freeze-thaw and wet-dry cycles or tillage might unprotect charcoal pieces by the disruption of soil aggregates (Kuzyakov et al., 2009) or directly accelerate the

Chapter 1. Introduction mechanical breakdown and consequent (bio-)chemical alteration of BC particles (Nguyen et al., 2008). By a better soil aeration, tillage might also promote the abiotic oxidation of charcoal, which is the first step in the weathering of BC (Cheng et al., 2006; Lehmann et al., 2009). The drivers and mechanisms of BC storage or the kinetics of BC degradation and losses from soil are still poorly understood and require further research for a better understanding of the role of BC in soil and in the global C cycle.

Soil amelioration with BC

Observation of charred plant residues and charcoal in free and mineral occluded forms and identification of highly aromatic humic acids in a range of soils such as European, Russian and American Chernozerms and Japanese Andisols that were subject to continuous burning during genesis support the view that charred organic materials are the source of refractory humic acids and the black color of a variety of soils (Schmidt and Noack, 2000a). It appears that the presence of BC generally affects positively soil physico-chemical properties, which is in agreement with the several traditional agricultural systems that use(d) fire and fire residues as a component of land management (Wiedner and Glaser, 2015). Slash-and-burn is commonly used to clear land and fertilize soil in shifting cultivation systems in many regions of the world. There are many indications that slash-and-burn agriculture occurred in Western Europe since the Neolithic, mainly during the Late Neolithic period (6.400–4.200 yr BP) (Wiedner and Glaser, 2015), ‘Hormigueros’ are another example of the use of fire in traditional agriculture in Spain that lasted up to the 1960s (Olarieta et al., 2011). These are small structure of about 0.5 m³ where dry woody vegetation was piled and burned under a soil cover (Olarieta et al., 2011). The resulting material was spread on the field as a fertilizer and soil conditioner, containing available nutrients such as exchangeable K and available P in ash released by the partial combustion of biomass. The temperature reached in the soil cover also had desirable weed-killing and disinfectant properties. A similar land management practice is described by Hoyois (1949) for the Belgian Ardennes, referred to as “écobuage” in french, and similar structures are still part of shifting cultivation systems in India and Bhutan (Olarieta et al., 2011). Nevertheless, it was established that the input of BC by both slash-and-burn and hormigueros was small (Glaser et al., 2001; Olarieta et al., 2011), because the main fraction of biomass is combusted during the process. Therefore, these practices result in a limited enrichment of BC. In contrast, Neolithic man-made soils with high amounts of BC were

Chapter 1. Introduction found in Europe on the northern isles of Scotland and in the southern part of Lower Saxony of Germany, but the intentional application of BC to soil for agricultural purposes has not been demonstrated (Wiedner and Glaser, 2015).

Figure 1.8. Left: Typical Ferralsol profile. The shallow surface horizon is light brown colored and roots are concentrated at the soil surface. Bellow this horizon follows a thin transition horizon to a subsoil horizon which can be several meters thick. The texture of these soils is loamy or sandy and the structure is dominated by stable micro aggregates (pseudo-sand). Right: typical terra preta profile. The topsoil horizons are dark grey or black colored and can reach a depth of more than 1 m. Potsherds, small bone and charcoal particles are characteristic for this horizon. Roots reach deeper down in higher density than in Ferralsols and signs of bioturbation and aggregates of biogenic origin can be found frequently. Below follow transition horizons which are lighter colored and typically show patches of different brown, grey and black colors with clear signs of mixing of topsoil and subsoil material. The subsoil horizons are identical to the subsoil horizons of adjacent soils. Terra pretas typically have the same texture like surrounding soils (source: Glaser and Birk, 2012; Glaser et al., 2001). Amazonian ‘terra preta’ (literally ‘dark earths’ in Portuguese) are probably the best example of soil amelioration with BC (Figure 1.8). Terra preta are highly fertile, BC-rich Anthrosols occurring in patches of about 20 ha on average and up to 350 ha in the immediate vicinity of highly weathered tropical soils of the Amazonian basin that become infertile rapidly after deforestation for cropping (Glaser et al., 2002). Genesis of Amazonian terra preta was related to the activities of pre-Colombian settlements and involved the application to 43

Chapter 1. Introduction soil of a number of inorganic (ash, bones…) and organic (biomass waste, manure, excrement, urine …) wastes, including fire residues. These inputs have been metabolized and humified, with fungi playing a dominant role in the transformation of BC (Glaser and Birk, 2012). Comparable Anthrosols were also identified in other regions of South America (Ecuador, Peru) and Africa (Benin, Liberia, South Africa) (Glaser et al., 2002). As a result of these various anthropogenic inputs, terra preta soils contain 3– 4 times more SOC than adjacent soils, and up to 70 times more BC (Glaser et al., 2002). The enhanced fertility of these soils is related to high concentrations of nutrients such as N, P, Ca, Mg, Mn and Zn (Glaser and Woods, 2004). An important fraction of N (~70 %) is in the form of heterocyclic N in the structure of BC, whereas remaining N occurs in the form of amino acids (18– 25 %), amino sugars (4–7 %) and inorganic N (1–2 %), which accords with the composition found in surrounding non terra preta sites (Glaser and Woods, 2004). P is derived mainly from excrements and bone residues, and Ca, Mg, Zn, Mn from waste products of vegetative or animal origin. Enrichment in characteristic chemical elements (Cu, Mn, Zn) is attributed to residues of palms that were used to cover houses by pre-Colombian populations (Glaser and Woods, 2004). The introduction of fire residues is a crucial factor in the genesis of terra preta soil because it contains BC and ash. On the one hand, the increase in soil pH to neutral or slightly acidic values due to the liming effect of ash suppresses aluminum activity (and toxicity for plants and microbes) and enhances the availability of nutrients (Glaser and Woods, 2004). On the other hand, BC provides charged surfaces that are scarce in highly weathered tropical soils, thereby enhancing dramatically the capacity of retention of nutrients (Figure 1.9) (Glaser and Woods, 2004). Thanks to its persistence in the system, BC is considered as the main driver of the long- lasting fertility of terra preta soils. Liang et al. (2006) have attributed the large CEC per unit of C of terra preta soils to the great charge density of these soils related to the large specific surface area of BC and the presence of oxygenated functions such as carboxyl groups created by oxidation of BC or sorption of uncharred SOC molecules to the surface of BC particles. Resulting CECs per unit of C are up to three times more elevated in terra preta soils than in adjacent reference soils (Figure 1.9; Glaser and Birk, 2012).

Chapter 1. Introduction

Figure 1.9. Cation exchange capacity (CEC) of BC-rich terra preta soils against soil organic carbon (SOC) content. The line drawn on the graph represent the typical relationship between CEC on SOC of BC-poor natural soils adjacent to terra preta (redrawn after Sombroek et al. (1993) by (Glaser and Birk, 2012). Using quantitative 13C NMR spectroscopy, Mao et al. (2012) found that SOM of terra preta soils consists predominantly of ~6 fused aromatic rings substituted by carboxyl groups that considerably increase the CEC of soil (Figure 1.10). Similar organic compounds were identified in highly productive, grassland-derived soils in the US (Chernozerms) that were affected by pre-settlement fires (Mao et al., 2012). These highly aromatic humic acids produced by humification of BC in soil represent from 40 to 50 % of SOC in these Chernozerms, which is much more abundant than previously thought (Mao et al., 2012). Such aromatics might be a common product of decomposition of BC and an interesting signature to identify soils that were affected by fire residues.

Chapter 1. Introduction

Figure 1.10. Model of a typical stable and fertile aromatic cluster in Terra Preta and temperate grassland soils, derived from 13C NMR spectroscopy and H-C measurements (source: Mao et al., 2012). Even though the intentional character of the genesis of terra preta is still debated, Amazonian terra preta soils are considered as a model for sustainable agriculture in the humid tropics (Glaser et al., 2001). The recent discovery of millennial anthropogenic dark earths in Northern Germany related to comparable genesis and having properties similar to that of Amazonian terra preta supports the view that such model of sustainable agriculture might be applicable to temperate regions (Wiedner et al., 2015). Cumulic Anthrosols (also named ‘terra preta Australis’ by authors) developed on the residues of anthropogenic oven mounds in Australia, dating from 650 and 1609 years BP, are in agreement with this idea and indicate that the introduction of BC to soil can enhance the potential of soil for carbon sequestration on the mid- to long- term (Downie et al., 2011). Therefore, soil amendment with artificial BC (or “biochar”) might be an efficient mean to sequester carbon in soil on the long- term and improve sustainably soil fertility, like in terra preta soils.

Biochar for environmental management

Biochar is defined as the solid residues of the incomplete combustion of biomass or pyrolysis, intentionally produced to be amended to soil for environmental and agronomic benefits. Technically, biochar is a type of BC but it differs in essence from charcoal, which is aimed to be burned as a domestic or industrial fuel, and residues from wildfire or biomass and fossil fuel combustion that end-up unintentionally in the environment. The (re- )discovery of Amazonian terra preta since the end of the 1990s has triggered interest of many soil scientists for biochar, and research on biochar has increased exponentially in the last decade (Lehmann and Joseph, 2015).

Chapter 1. Introduction

Figure 1.11. a) The natural cycle of carbon in the soil-plant system and b) the cycle modified by the carbon-negative biochar technology (source: Lehmann, 2007b). However, there is evidence that the particular effects of charcoal on soil properties and plant growth had already drawn the attention of researchers a long time ago (Tryon, 1948). The concept of the biochar technology relies on a win-win strategy that consists of producing renewable energy by biomass pyrolysis and amending soil with biochar, a by-product of the reaction, to improve soil quality (Figure 1.11) (Lehmann, 2007a, 2007b). Three main environmental benefits are expected from biochar application to soil. First, soil amendment with biochar is carbon-negative thanks to the stability of pyrolysed materials against decay higher than that of the initial feedstock. Therefore, it can contribute to mitigate climate changes (Laird, 2008). Woolf et al. (2010) have estimated that biochar production from waste feedstock has the potential to introduce 110–220 TgC yr-1 at global scale, which is of the same order of magnitude as the current contribution of wildfire. Second, biochar has the potential to promote food security by the restoration or enhancement of soil fertility, as supported by terra preta soils. Third, biochar can reduce environmental pollution and improve water quality by its adsorbent properties, which were shown to decrease nutrient losses by

Chapter 1. Introduction leaching and gaseous emissions under certain conditions (Cayuela et al., 2014). In developing countries where the access to mineral fertilizers is limited, slash-and-char is proposed as an alternative to slash-and-burn to improve the standards of living of farmer by increasing their income via the production and sale of charcoal and by increasing yields through soil amelioration by amending charcoal residues to soil (Lehmann et al., 2006). The mitigation of climate change is a positive externality of slash-and-char, because a very small fraction of initial biomass is converted into BC by burning in slash-and-burn shifting cultivation systems. Additionally, pyrolysis residues are expected to have a longer residence time in soil than fire residues, as the latter are produced in presence of O2 and are subject to rapid heating rates, which can decrease the physical stability of BC (Antal and Grønli, 2003). Despite the potential of biochar for soil amelioration, short-term field and laboratory trials have resulted in very heterogeneous crop responses (Crane- Droesch et al., 2013). Meta-analyses have shown that biochar application to soil increases slightly the average crop production (Biederman and Harpole, 2013; Jeffery et al., 2011), however, the effect varies greatly depending on soil conditions (Biederman and Harpole, 2013; Jeffery et al., 2011) and quality of the biochar (Biederman and Harpole, 2013; Jeffery et al., 2011; Manyà, 2012). Although biochar application to soil may decrease crop productivity in some circumstances, acidic to neutral soil with coarse to medium texture generally benefits from biochar amendment (Jeffery et al., 2011). This supports the idea that biochar’s liming effect and water-holding capacity are the main factors in improving soil fertility. Yields of crops grown on soil with small organic carbon (OC) content and cation exchange capacity (CEC) also respond positively to amendment with biochar (Crane-Droesch et al., 2013). It is also clear that the application rate is a crucial factor to consider for optimizing the benefits from biochar application in terms of crop production and carbon sequestration. The effect of biochar on crop yields is very difficult to predict because it depends on complex interactions between biochar quality and application rate, environmental conditions and the agricultural context. Therefore, there is a need to design biochar with favorable agronomic and environmental properties to ensure that the goals targeted by biochar application to soil are met (Abiven et al., 2014).

Chapter 1. Introduction

Gap of knowledge

Despite of increasing research on the black carbon cycle partly driven by the recent idea of managing the BC cycle through the application of biochar to soil for agronomic and environmental benefits, many questions remain. The mechanisms of storage and the rates of BC transformation and transport that govern the longevity of BC in soil are still poorly understood, mainly because changes are slow and complex, with a multiplicity of drivers. Residence time of BC in soil exceeds by several orders of magnitude the lifetime of most laboratory and field experiment, which makes implementation of long-term experiments difficult, if not impossible. Consequently, most estimates of stocks and properties of BC as well as fluxes between the different reservoirs rely on historical BC deposits, which are poorly constrained because the initial input is unknown and because the quantification of BC is challenging. Limitation of the risks (interaction with herbicides and pesticides, introduction of heavy metals and polycyclic aromatic hydrocarbons (PAH) to soil, unbalanced addition of nutrients…) and optimization of environmental and agronomic benefits related to biochar application to soil require more insight into the interrelationship between biochar quality, application rates, soil and climatic conditions, and evolution of the properties of BC over time. Amending arable lands with biochar will have implications for centuries and therefore some guarantee of success is needed first.

Pre-industrial charcoal kiln sites of Wallonia: a model to evaluate the long-term effects of biochar on properties of temperate soils

Since time immemorial, charcoal has been a valued fuel for domestic and industrial purposes and might have been the first synthetic material produced by man, as suggested by prehistoric paintings made of charcoal (Antal and Grønli, 2003). Particularly, charcoal has long been the favourite fuel of the metallurgical industry because of its high calorific power and its reductant properties (Antal and Grønli, 2003). In 2009, the FAO reported a global annual charcoal production of 47 million m³. Africa accounts for 63 % of total production, Latin America and the Caribbean for 18.7 % and Asia for 15.7 % (Steierer, 2011). In developing countries, charcoal is mainly produced in situ with traditional earthmound kilns (Schenkel et al., 1998; Steierer, 2011).

Chapter 1. Introduction

Figure 1.12. Schematic draw of the different steps of traditional charcoal production by the earthmound kiln technique (Source: Panckoucke, 1783). Topsoil layers were removed and the soil was leveled. Wood was stacked into a mound around a vertical post and covered with vegetation residues and soil. The central post was removed and a fire was ignited in the central chimney. Pyrolysis lasted up to two weeks for the largest structures (100–150 steres of wood). Similar traditional charcoal kilns (also referred to as charcoal hearths or meilers in the literature) were commonly used for charcoal production in most Western European countries and the Mid-Atlantic states of the United States (Mikan and Abrams, 1995), where charcoal was the main pre-industrial fuel used for smelting and steel-making before the introduction of coke. Wallonia, Southern Belgium, had much charcoal-based smelting in the late 18th century as about 75 smelting operations were active at that time, with a majority of blast furnaces where the iron ore was smelted into pig iron (Hansotte, 1980). Consequently, thousands of hectares of forest were required to supply charcoal to the Walloon steel industry (Hoyois, 1953). Charcoal production declined from 1830 as charcoal was gradually substituted by coke in the metallurgical industry, and it was completely abandoned by 1860 (Evrard, 1956). The process of charcoal production with traditional earthmound kilns is subject to variations (site preparation, shape and volume of the mound, diameter and length of the logs …). The technique presented in the following lines relies mainly on the method described in the book “traité pratique de carbonisation” from Lepoivre and Septembre (1941). This description accords with that of other regional authors and thereby is assumed to be representative of the conditions of pre-industrial charcoal production in the late 18th century in Wallonia. An important loss of weight and volume occurs during pyrolysis, which makes the transport of charcoal much more convenient than that of initial wood.

Chapter 1. Introduction

Therefore, charcoal was produced in situ to minimize the transport of wood. (Hoyois, 1953; Lepoivre and Septembre, 1941). The forest litter and humus- rich surface soil layers were removed (Gebhardt, 2007) and a terrace was levelled to set up the kiln on a flat surface (Figure 1.12). Only the wood from about 3 to 10 cm in diameter was used for charcoal production, because large wood was directly exploited as lumber or fuel wood and because poor yields were obtained from small branches. Wood was stored for a few months between harvest and charcoal production to decrease moisture content to about 20 %, which increased the yields of pyrolysis. Logs from 0.7 to 1 m in length were stacked vertically around a central post into a circular mound (Figure 1.12). The mound was isolated from oxygen in the air with a cover of vegetation residues (leaves, litter ferns, grasses, branches ...) and a layer of 8– 10 cm of soil.

Figure 1.13. Cross section of a traditional earthmound kiln before pyrolysis (source: Lepoivre and Septembre, 1941). The arrows indicate the pathway of gases through the structure. Pyrolysis starts at the center of the mound, beside the central chimney where a fire is lit. The front of pyrolysis moves down progressively by opening and closing blowholes from the top to the bottom of the mound. Blowholes were opened at the bottom of the structure. The post was then removed to create a central chimney where a fire was lit (Schenkel et al., 1998). The warm gases free of oxygen passing through the mound activated wood pyrolysis that reached a maximum temperature of 400–450 °C. Blowholes were opened successively from the top to the bottom of the mound to conduct heat fluxes and pyrolysis through the whole structure (Figure 1.13). The whole process lasted about 50 hours for a small mound of 10 steres (a stere is defined as one cubic metre of stacked woods of one meter in length) and up to 2 weeks for the larger mounds that could contain up to 100–150

Chapter 1. Introduction steres of wood (Lepoivre and Septembre, 1941). At the end of the process, all openings were closed and the mound was left to cool down for several days before collecting charcoal. Final volume was about 30–40 % of initial volume and mass yield about 20 % (Schenkel et al., 1998). When a next episode of charcoal production occurred in the same area, the site of the previous kilns were used preferentially because the preparation of the platform required less effort as the area was already flat and plant colonization was delayed by the lethal thermal action of pyrolysis (Mikan and Abrams, 1995). Moreover, insulating properties of soil were improved by heating, which increased the yields of pyrolysis up to 10 % after three occurrences (Lepoivre and Septembre, 1941), and the soil-charcoal mixture was an ideal substrate readily available to cover the next mound. In contrast to pyrolysis in controlled conditions, in situ pyrolysis with a traditional mound kiln was not completely free of O2 as a small input of air is necessary for the circulation of gases through the structure. This is an important point to consider for the final quality of charcoal that can differ from biochar produced in controlled conditions in the same range of temperature.

Figure 1.14 Soil pit across a pre-industrial charcoal kiln site on a forest Luvisol in Louvain-la-Neuve. The depth of the charcoal-rich topsoil horizon is 40 cm on average whereas that of organo-mineral horizon of adjacent soil unaffected by charcoal production is 10 cm.

Chapter 1. Introduction

Today, charcoal kiln sites of Wallonia can be identified on forest floors as domes a few decimetres in height and about 10 m in diameter. They are especially easy to detect on slopes, where the flat bases of the kilns are clearly visible. The very black topsoil consists of thermally altered (organo-)mineral soil and vegetation remains mixed with charcoal residues that were left on the site after pyrolysis (Figure 1.14). Enrichment of the topsoil by charcoal, site preparation and thermal action of wood pyrolysis are major forms of disturbance that affected the soil at a kiln site.

1.2. Aims and objectives

The aim of this thesis was to evaluate the long-term effect of pre-industrial charcoal kiln sites on properties of temperate soil, which provides a unique opportunity to understand long-term implications of a biochar soil amendment in temperate regions. In particular, we investigated how charcoal enrichment affected soil properties in contrasting soil conditions, and how soil conditions influenced the long-term evolution of chemical properties, stability and residence time of charcoal in soil. Accordingly, the four following specific objectives were addressed: (i) to assess the magnitude of pre-industrial charcoal production in Wallonia; (ii) to develop a protocol to identify and quantify charcoal in the soil of pre-industrial charcoal kiln sites; (iii) to assess the effect of pre-industrial charcoal kiln sites on soil properties, under contrasting soil conditions; and (iv) to evaluate the stability and dynamics of charcoal in the soil of pre-industrial charcoal kiln sites, with respect to land use.

1.3. Thesis outline

The overall research strategy is illustrated in a schematic draw interconnecting the different chapters of the thesis (Figure 1.15). Each chapter aims to respond to one of the four specific objectives presented earlier. We provide a short outline of the development of the chapters with respect to these specific objectives.

The extent of pre-industrial charcoal production in Wallonia.

In Chapter 2 and 3, the magnitude of pre-industrial charcoal production and the spatial distribution of charcoal kiln sites at the scale of Wallonia was investigated. As pre-industrial charcoal was produced mainly to supply fuel to the smelting industry, we estimated the equivalent area of forest necessary to meet the demand for charcoal of blast furnaces that were active in the late

Chapter 1. Introduction

18th century, when charcoal-based smelting reached a peak in Wallonia (Chapter 2). We also estimated the number and distribution of pre-industrial charcoal kiln sites based on their detection by remote sensing (Chapter 3). A LiDAR-derived digital elevation model (DEM) was interpreted to detect the sites based on their topographic impact in forested areas, whereas orthoimages were used to detect the sites on bare soil, in agricultural areas. The output of these chapters allowed identifying kiln sites in contrasting conditions for chemical analysis of soil and charcoal, and gave the opportunity to extrapolate site-specific results to regional scale. A short chapter (Chapter 4) at the end of this section is dedicated to the sampling of soils and charcoals that are analyzed in the following Chapters.

Figure 1.15. Overall research strategy corresponding to the thesis structure.

Chapter 1. Introduction

Identification and quantification of charcoal-C in the soil of pre-industrial charcoal kiln sites

In Chapters 5 and 6, one thermal and one chemical method were tested to identify, characterize and quantify charcoal in the soil of pre-industrial charcoal kiln sites, assuming that pre-industrial charcoal residues are more resistant to thermal and chemical oxidation than uncharred SOM. In Chapter 5, we explored the potential of differential scanning calorimetry (DSC) as a tool to discriminate between charcoal and uncharred SOM in soil. A methodology of charcoal-C quantification in soil based on DSC was proposed, and compared to the BC content estimated by benzene polycarboxylic acids (BPCA) molecular markers method, widely used for the quantification of BC in soil and sediments. In Chapter 6, two dichromate-based SOC quantification methods were tested to discriminate between uncharred organic matter and charcoal residues in the topsoil of pre-industrial charcoal kiln sites. Results were compared to charcoal-C and uncharred SOC contents estimated by DSC (Chapter 5).

The effect of pre-industrial charcoal production on soil properties.

In Chapters 7, 8 and 9, the effect of pre-industrial charcoal kiln sites on chemical and biological soil properties was studied by comparison with adjacent reference soils, unaffected by charcoal production. Our investigations relied on the three following assumptions: (i) charcoal production modified physico-chemical soil properties, the balance of nutrients and biological activity and community structure at kiln site due to the introduction of charcoal residues, wood ash and the thermal effect of pyrolysis, (ii) properties of the charcoal-rich soil evolve over time due to charcoal ageing and nutrient leaching, and will be modified after two centuries (iii) the long-term evolution of soil properties can be influenced by soil conditions and will depend on soil type or land use. In Chapter 7, we determined chemical properties of 20 charcoal kiln sites from forest, distributed on four different soil types. Data were analyzed in relation to depth and soil conditions. To address the evolution of soil properties over time, we compared the properties of pre-industrial kiln sites to that of a currently active kiln site located close to Dole (France), where traditional charcoal production in a mound kiln has been practiced for cultural and tourist purposes since the early 1990s. In Chapter 8, we investigated the effect of pre-industrial kiln sites on physico-chemical properties of cropland soil. Soil amendment with biochar

Chapter 1. Introduction targets mainly cropland soils, therefore, we drew an analogy between the effect of pre-industrial charcoal accumulation at kiln site and the long-term effect of biochar on soil properties in a temperate region. In Chapter 9, we studied the effect of long-term charcoal accumulation on microbial activity, biomass and community structure for a selection of forest and cropland sites on Luvisols. Carbon mineralization rates and biological properties were determined and related to soil physico-chemical properties and the contents of charcoal-C and uncharred SOC in soil.

Stability and dynamics of charcoal in the soil of pre-industrial charcoal kiln sites.

In chapters 10 and 11, we investigated the long term dynamics of charcoal in soil, with a special focus on the effect of land-use change from forest to cropland on the chemical properties, the stability and the stocks of charcoal in soil. We assumed that a land use change from forest to cropland might accelerate the physical, chemical and biological alteration of charcoal by way of the mechanical action of tillage and the improved soil fertility related to the application of organic and inorganic fertilizers. We developed two different approaches, the first based on the chemical analysis of charcoal particles extracted from soil and the second based on the estimation of carbon stocks at the site of pre-industrial charcoal kilns by remote sensing. In Chapter 10, we measured the chemical properties and stability of charcoal particles extracted from pre-industrial kiln sites along a chronosequence of land-use change from forest to cropland, up to 200 years of cultivation. In Chapter 11, we used remote sensing data (VIS-NIR) to predict SOC content based on the reflectance of bare cropland soil on high resolution satellite imagery (Ikonos II). We calculated the stocks of charcoal-C in fields with contrasting cultivation history to investigate whether cultivation influence the residence time of charcoal in soil or not. In the conclusion and perspectives, we reflect on the future of biochar amendments and the functions of BC in soil in light of the results of this thesis.

Chapter 2. The extent of pre-industrial charcoal production

Chapter 2. The extent of pre-industrial charcoal production in Wallonia (Belgium): a historical approach2

Summary

Until the early 19th century, charcoal was the only combustible used by the steel industry. We aimed to assess the extent of pre-industrial charcoal production when charcoal-based smelting reached a peak in Wallonia, in the late 18th century. Therefore, we calculated the demand for charcoal of the steel industry based on the amount of pig iron that was produced by blast furnaces that were active at that time. About three tons of charcoal were needed to produce one ton of pig iron and refine it into an end product. The annual production of pig iron of an average blast furnace was between 500 and 550 tons, which corresponds to an annual consumption of 1.600 tons of charcoal, or 20.000 steres of wood. Therefore, 1.500.000 steres of wood were needed annually to meet the demand for charcoal of the 73 blast furnaces that were active in Wallonia in 1790. Given that a coppice forest of 20 years supplies 80 to 100 steres of wood per ha, the forest area equivalent to the demand for charcoal of an average blast furnace was 222 ha annually, or 4.444 ha for a forest rotation of 20 years. In total, 325.000 ha of forest were required to meet the demand of the pre-industrial steel sector for charcoal, which represents 75 % of the forested area on the map of Ferraris (1770–1778). This highlights the intense pressure exerted on the Walloon forest in the late 18th century through

2 Adapted from:

Hardy, B., Dufey, J., 2012b. Estimation des besoins en charbon de bois et en superficie forestière pour la sidérurgie wallonne préindustrielle (1750-1830) Deuxième partie : les besoins en superficie forestière. Revue Forestière française, 6, 799–806.

Chapter 2. The extent of pre-industrial charcoal production the production of charcoal, which explains for the ubiquity of pre-industrial charcoal kiln sites in areas that were forest at that time.

2.1. Introduction

In some regions of Wallonia, black spots from 20 to 40 m in diameter are frequently observed on bare cropland soil (Figure 2.1). Such soil artefacts are located exclusively in areas that were formerly forested on the map of Ferraris (1770–1778), and deforested for cultivation. These are pre-industrial charcoal kiln sites, where charcoal was produced by the traditional “mound kiln” method that is described in detail by Schenkel et al., (1998). On forest soil, pre-industrial charcoal kiln sites generally have the shape of domes of about 10 m in diameter. On a slope, the platform left at the site of a mound kiln is particularly visible (Ludemann, 2010; Stolz et al., 2012). The very black topsoil of charcoal kiln sites consists of thermally altered (organo-) mineral soil and vegetation remains mixed with charcoal residues that were left on the site after pyrolysis.

Figure 2.1. Aerial photograph (a) and ground view (b) of pre-industrial charcoal kiln sites on bare cropland soil. Charcoal kiln relics are omnipresent in the forest area mapped by Ferraris (1777). The observation of aerial photographs and satellite imagery revealed that the presence of pre-industrial kiln sites is almost systematic in areas that were deforested for cultivation since 1777, recently mapped by Kervyn et al., (2014). In forest, we reached the same conclusion, as more than 250 sites were detected thanks to their topographical impact during field campaigns. Charcoal kiln sites were found in most visited places where forest was delineated on Ferraris’s map.

Chapter 2. The extent of pre-industrial charcoal production

It is clear that the history of the Walloon forest was deeply affected by charcoal production, particularly in the 18th century (Figure 2.2; Tallier, 2004). Until the early 19th century, charcoal was the only pre-industrial combustible used to smelt the iron ore and refine pig iron in Wallonia. Coke, the refined product of coal, progressively replaced charcoal in the early 19th century, concomitantly to the importation of the Cockerill technology from England and railway development that made cost-effective the transport of coal out of the mining areas (Goblet d’Alviella, 1927; Woronoff, 1994). It is likely that charcoal was also used in traditional bricks kilns in the loessic belt of Wallonia, where bricks is the main construction material, whereas iron ore was abundant in several regions of the south of Wallonia. Some charcoal might also have been used for domestic heating and cooking.

Figure 2.2. Evolution of the industrial use of wood over time in Wallonia (adapted from Tallier (2004)). In this chapter, we aimed to assess the extent of pre-industrial charcoal production when charcoal-based smelting reached a peak in Wallonia, in the late 18th century. This text presents the main outputs of a more detailed study that was published in the Revue Forestière Française (Hardy and Dufey, 2012a, 2012b).

2.2. Methodology

Charcoal was produced by directly in the forest where the trees were cut, to limit the transport of this heavy material (only 20 % of the weight of the initial biomass remains after pyrolysis; Schenkel et al., 1998). Very few documents report figures for charcoal production, whereas the production of pig iron in blast furnaces is better documented. Therefore, we evaluated the production 59

Chapter 2. The extent of pre-industrial charcoal production of pig iron by blast furnaces that were active on the territory when charcoal- based smelting reached a peak in Wallonia, and we calculated the equivalent amount of charcoal that was needed in the different steps of the process in the transformation of iron ore to obtain an end-product. Previous studies have relied on a similar methodology in a comparable goal (e.g. Noirot (1843); Woronoff (1990)). Once the demand for charcoal was determined, we estimated the equivalent forest area required for its production, which involved an estimation of the yield of the transformation of wood into charcoal by pyrolysis with traditional mound kilns, and the equivalent forest area able to supply this amount of wood within the period of time required to grow trees of an adequate diameter for charcoal production (about 10 cm for the largest logs; Lepoivre (1940)). The estimated equivalent forest area was then compared to the forest area mapped by Ferraris in 1770–1778, which is assumed to be representative of the forest resources available in Wallonia at the peak of charcoal production.

2.3. Results and discussion

The number of blast furnaces in Wallonia and the production of pig iron

In Wallonia, a number of vestiges attest the importance of pre-industrial smelting activities. Evrard (1956) inventoried ancient forges in Wallonia and the Grand Duchy of Luxembourg, whose remains are often limited to ponds that were dug to supply the hydraulic power needed by blast furnaces (Figure 2.3). Forges that were inventoried (Figure 2.3) did not all include a blast furnace. Blast furnaces were mainly found in the provinces of Namur, Hainaut and Luxembourg. Most operations in the surroundings of Charleroi and Liège were dedicated to the production of end products out of pig iron supplied by other provinces. Many blast furnaces were located in the “Entre-Sambre-et- Meuse”, which is, according to Heuschling and Van der Maelen (1838), the territory that contains the largest resource of iron ore of all Europe. Iron ore was also abundant in the south of the province of Luxembourg, where a high density of smelters was also recorded. The presence of iron ore but also that of forest and rivers, indispensable resources for the steel industry, dictated the location of blast furnaces (Figure 2.3).

Chapter 2. The extent of pre-industrial charcoal production

Figure 2.3. Ancient forges inventoried by Evrard (1956) and forest surface mapped by Ferraris in 1770–1778 (Kervyn, 2014) A detailed historical survey of Hansotte (1980) revealed that about 90 blast furnaces were active in the Austrian Netherlands and the region of Liège and Stavelot in 1790. Looking only at the current boundaries of Wallonia, 73 blast furnaces were in activity in 1790. This author reports annual productions of around 400 to 650 tons of pig iron per blast furnace in the 1760–1790 period. By comparison, the average production of pig iron calculated from the data of Evrard (1956) is 528 tons per year (with a standard deviation of 120 t), which matches perfectly with the estimation of Hansotte (1980).

The demand for charcoal of one average blast furnace

The conversion of the amount of pig iron produced by one average blast furnace into an equivalent consumption of charcoal relies on a large survey in 1811 of the former “Département des Forêts” reported by Wagner (1921), which included the Grand Duchy of Luxembourg and a main part of the province of Luxembourg. In reports from the steel industry, combustible is expressed in stere of wood (a stere is defined as one cubic meter of woods of one meter in length) or in kg of charcoal. A yield of 80 kg of charcoal for one stere of hardwood is generally assumed in these statistics. Several independent references, including detailed experimental values measured at that time in the department of Ardennes and Meuse, agree with this conversion rate of wood into charcoal by pyrolysis in traditional mound kilns. This corresponds to a 61

Chapter 2. The extent of pre-industrial charcoal production mass yield of about 20 %, which is similar to the yield obtained today with traditional mound kilns in developing countries, where charcoal is still produced by this technique (Schenkel et al., 1998). Based on data from the survey of the former “Département des forêts” (Wagner, 1921), we calculated that 24.8 steres of wood, which corresponds to 1.98 tons of charcoal, were consumed on average to produce one ton of pig iron. To obtain an end product, pig iron needed to be refined and hammered, which required 13.2 steres of wood, or 1.05 tons of charcoal per ton of pig iron. Nevertheless, some blast furnaces produced cast iron rather than pig iron, which was not refined but directly transformed into andirons, pots, cannons or cannonballs. The production of cast iron was, however, more charcoal- consuming than the production of pig iron, and required an amount of charcoal similar to that needed to smelt and refine pig iron. Transformations subsequent to refining to obtain end-products were neglected in the calculation because they represent only about one to two percent of the combustible used for the entire process. As a result, we estimated that 1.600 tons of charcoal, or 20.000 steres of wood, were needed annually to smelt iron ore and refine pig iron for an average blast furnace in 1790. At the scale of Wallonia, this corresponds to about 1.500.000 steres of wood supplied by the forest to the 73 blast furnaces active in the late 18th century.

The wood for charcoal production

The preferred wood for the production of charcoal was from 3 to 10 cm in diameter, generally obtained from a short rotation coppice clearcut every 20 years (Goblet d’Alviella, 1927; Lepoivre, 1940). Hardwood charcoal was particularly adapted for the smelting of iron ore whereas softwood charcoal was preferred to refine pig iron into wrought iron. In many regions, coppice- with-standards forest evolved progressively to pure coppice forest to meet the growing demand for charcoal of the steel industry. Consequently, the income from a coppice became larger than that of old-growth forest. Standards also suffered from the expansion of the clearings to support the activities of charcoal-makers and from wildfires that they caused unintendedly. Determination of the forest area equivalent to the amount wood consumed by blast furnaces required an estimation of the primary productivity of a coppice of 20 years. A productivity ranging from 80 to 100 steres per ha is a reasonable estimation (Francoeur et al., 1824; Institut pour la Forêt, 2012). This corresponds to a production capacity of 6 to 8 tons of charcoal per ha every 20 years. 62

Chapter 2. The extent of pre-industrial charcoal production

The forest area needed for an average blast furnace

All information necessary to estimate the forest surface used by an average blast furnace for the smelting of iron ore and refining of pig iron can then be combined. Assuming 3.03 tons of charcoal consumed per ton of iron produced, 0.08 ton of charcoal obtained from one stere of wood, and a production of 90 steres of wood for a coppice of 20 years, a blast furnace with an average annual production of 528 tons of cast iron required a forest area of 222 ha to meet annual needs in charcoal. As regeneration of the coppice took 20 years, the overall forest area exploited by an average blast furnace was 4.444 ha. Some authors reported a territorial influence of the pre-industrial steel industry comparable to our estimation (Gaudin, 1996; Goblet d’Alviella, 1927). Nevertheless, most of them do not provide details of the calculation and do not cite references for the estimation of required parameters. Other studies (that generally do not cite references) report numbers smaller than ours. For instance, Delvaux (1998) estimated that 1440 ha of forest were needed to supply one blast furnace, based on an annual production of 280 tons of pig iron. Even though the source of the latter value is not given, this might correspond to the production of a furnace from England in 1720, reported by Feltz and Incourt (1995). This production capacity is clearly underestimated for blast furnaces of Wallonia in the late 18th century. Moreover, Delvaux (1998) considers that 1.8 tons of charcoal is needed per ton of iron produced. This number is plausible for the production of pig iron alone, but neglects subsequent steps in the production of iron that must be taken into account for an accurate estimation of the territorial influence of blast furnaces. Despite the uncertainties on each parameter involved in our calculation and discussed earlier in this work, this proposed estimation relies on experimental data from the literature, which is an obvious quality of this survey.

Extrapolation to Wallonia

Assuming that 73 blast furnaces were in activity in Wallonia in 1790, 324.412 ha of forest were needed to supply combustible to the steel industry. There might be a margin of error in the estimation, as mentioned earlier in the text, nonetheless, such a large surface corresponds to an important fraction of the area that was forested in Wallonia in the late 18th century. The most reliable source of information to estimate the total area forested at that time is the map or Ferraris (1770–1778). Recently, Kervyn et al. (2014) georeferenced a digital version of this map and derived the forested areas in vector format, 63

Chapter 2. The extent of pre-industrial charcoal production which allows quantifying the forest area in the late 18th century. A surface of about 60.000 ha of the current territory of Wallonia was not mapped by Ferraris. This includes the Duchy of Bouillon that was not part of the Austrian Netherlands. By attributing to these unmapped areas a similar afforestation rate to that of the natural region it belongs to, the total surface of the forest of Wallonia in 1770–1778 was estimated to 432.000 ha. According to our estimation, the forested area attributed to the production of charcoal in the late 18th would therefore correspond to 75 % of the total forest. This highlights the intense pressure of the steel industry on the Walloon forest in the late 18th century through the production of charcoal, which explains for the omnipresence of pre-industrial charcoal kiln sites in the former forest of Ferraris. However, many areas that are currently forested in Wallonia do not contain pre-industrial charcoal kiln sites, because they were afforested after the period of in situ charcoal production. There might be a few exceptions because the production of charcoal resumed at small scale to supply fuel to vehicles running on producer gas, until after the Second World War. This contribution is, however, negligible compared to the production of charcoal for the pre- industrial steel sector.

2.4. Conclusion

Our historical investigation demonstrated how important wood resources have been for the development of the pre-industrial steel industry in Wallonia and, consequently, how they contributed to the global leadership of the Walloon steel sector that was maintained and even strengthened when coal mining substituted to traditional charcoal production. Adding to the demand of the steel industry for wood other demands for industrial and domestic purposes, we understand how strong the pressure on the Walloon forest was in the late 18th century. Several authors have described the decay of the Walloon forest in that period and conflicts that have occurred because wood demand exceeded the supply (Dorban, 1988; Hoyois, 1953). Nevertheless, others have nuanced this point, arguing that charcoal production for the steel industry was an unprecedented outlet for wood resources, which made the wealth of forest landowners (Belhoste, 1990). With the decline of the charcoal-based smelting in the first half of the 19th century, the pressure on wood resources dropped in a short period of time.

Chapter 2. The extent of pre-industrial charcoal production

The charcoal-based steel industry also deeply affected forest ecosystems by favoring short rotation coppicing, at the expense of old growth forest that was restored progressively by after. New demands for wood emerged in the 19th century, particularly in the mining and papermaking sectors, which lead to the afforestation of important areas in the Ardennes, mainly with softwood, and the conversion of hardwood forest into softwood forest (Kervyn et al., 2014). Recently, light detection and ranging (LiDAR) data were acquired over the whole territory of Wallonia. This airborne survey providing digital elevation model of a very high spatial resolution, is explored in Chapter 3 as a source to detect today charcoal kiln sites under forest, to confirm the large extent of pre- industrial charcoal production highlighted in this chapter by our historical approach.

Chapter 3. Detection of the sites by remote sensing

Chapter 3. Magnitude of pre-industrial charcoal production in Wallonia (Belgium) revealed by LiDAR data and high resolution aerial photographs

Summary

In this chapter, we used remote sensing to estimate the distribution, density and cover of charcoal kiln sites related to pre-industrial smelting in the late 18th century in Wallonia. We defined the potential area for charcoal production as the forest area mapped by Ferraris (1770–1778) restricted to locations where soil conditions were adapted for in situ charcoal production. Two hundred forty five circular sampling plots 150 m in radius were generated randomly in this area and explored for the presence of pre-industrial charcoal kiln sites. When the sample was located in cropland, the sites were detected on bare soil thanks to four sets of orthoimages, based on the black color of the charcoal-rich soil. In forested areas, charcoal kiln sites were detected based on characteristic relief features appearing on a high resolution digital elevation model (DEM) derived from light detection and ranging (LiDAR) data. Ancient hardwood forest bodies that had been converted to softwood forest were removed from the sampling area to decrease the rate of false negative detection caused by the strong interception of the LiDAR signal by the canopy of some coniferous plantations. By comparison with field detection, we obtained a rate of false negative detection (proportion of sites undetected with LiDAR) of 29.3 % that decreased to 20.5 % after exclusion of areas invaded by brambles in the northern part of Wallonia, dominated by field crops. We estimated a rate of false positive detection (proportion of reliefs erroneously detected as kiln sites on LiDAR) of 15.7 %. In total, we identified charcoal kiln sites in 93.9 % of the sampling plots, with a median site density of 1.2 sites per ha. Regional distribution of the density of sites accords with the distribution of blast furnaces that were active in the late 18th century, which were located mainly in the “Entre-Sambre-et-Meuse” and the South of the province of Luxembourg. These results strongly supports the idea that a major part of wood resources were allocated to charcoal production in the late 18th century, and that charcoal was mainly used as a combustible by the pre- industrial steel industry. The average outer diameter of forest sites is 10.1 m, which accords with the typical range of diameters recorded in the black forest,

Chapter 3. Detection of the sites by remote sensing the Vosges and neighboring regions in western central Europe. We estimated a total number of sites of about 450.000 in Wallonia, covering a cumulated area of approximately 4.000 ha in total. They offer a unique opportunity to study the long-term effect of charcoal on the properties of soil in various conditions, which is the main topic of this dissertation, developed in the following chapters.

3.1. Introduction

Pre-industrial charcoal production occurred in many regions of western Europe, mainly to supply fuel to the smelting industry (Ludemann, 2012). Wallonia, southern region of Belgium, is often considered as the cradle of the pre-industrial steel industry because early blast furnaces originate from the surroundings of Liège (Gillard, 1971; Houbrechts and Petit, 2004). Until the early 19th century, charcoal was the only fuel used to smelt and refine iron in Wallonia. The charcoal-based steel industry reached a peak in the late 18th century. Hardy and Dufey (2012a, 2012b) estimated that the equivalent of 75 % of the forested area was needed to supply charcoal to the 73 blast furnaces that were active in Wallonia at that time and to refine the pig iron that they produced (Chapter 2). Consequently, the pressure on wood resources to meet the demand of blast furnaces for charcoal was huge and might have been one of the main causes that conducted the steel sector to turn to coke as a fuel in the early 19th century. Concomitantly, the Cockerill technology was imported from England and transport revolution lowered the expense of the export of coal from mining areas (Hardy and Dufey, 2012a). Charcoal was produced in situ with traditional mound kilns. The process of charcoal making by the mound kiln technique is well described by Schenkel et al. (1998). After the production of charcoal, the soil is highly disturbed at kiln site (Mikan and Abrams, 1995), mainly because of site preparation (Gebhardt, 2007), thermal action of pyrolysis and large inputs of charcoal residues. At kiln site, charcoal enrichment darkens the topsoil (Hardy et al., 2016) and markedly increases carbon stocks compared to adjacent reference soils (Borchard et al., 2014; Criscuoli et al., 2014). Soil relief is always impacted by charcoal production, which is key for the detection of kiln sites. On a flat area, abandoned charcoal kiln sites generally appear as slightly heightened domes around 10 m in diameter (Figure 3.1a) (Hardy et al., 2016). Sometimes, particularly when stone load is important, the platform is surrounded by a circular bulge delimiting the site (Figure 3.1b) (Hardy and

Chapter 3. Detection of the sites by remote sensing

Dufey, 2015). On a slope, mound kilns were established on small terraces that are clearly visible (Figure 3.1c) (Ludemann, 2010; Stolz et al., 2012).

Figure 3.1 Schematic drawing of the cross-sectional view of three typical charcoal kiln site patterns; (a) on a flat area with low stone load; (b) on a flat area with important stone load; (c) on a slope; Redrawn after Hardy and Dufey (2015) Nowadays, the light detection and ranging (LiDAR) technology, also called laser airborne scanning, is more and more used to acquire altimetric data of high spatial resolution, even in presence of a vegetation cover (Hesse, 2010; Wehr and Lohr, 1999). The high resolution digital elevation models (DEM) derived from LiDAR data are powerful tools for the detection of archeological features (Devereux et al., 2008; Hesse, 2010; Kokalj et al., 2011). A few recent studies successfully used LiDAR-derived DEM for the detection of pre- industrial charcoal kiln sites thanks to their relief (Bollandsås et al., 2012; Deforce et al., 2013; Hesse, 2010; Ludemann, 2012, 2011; Raab et al., 2014; Schneider et al., 2015), which advantageously complete field work and allows a rapid exploration of large areas of land (Ludemann, 2012; Raab et al., 2014). The regularity in the shape and size of the charcoal kiln sites coupled to their high occurrence (up to 5 sites per ha according to Dussart and Wilmet, 1970) and constant distribution in the landscape make their footprint very typical so that they can hardly be confounded with other circular objects (Hesse, 2010)

Chapter 3. Detection of the sites by remote sensing such as shell holes, wind falls, tumuli or other burial mounds, which generally differ in size, shape or distribution in the landscape (Figure 3.2).

Figure 3.2. Footprint of shell holes (a), burial mounds (b) and charcoal kiln sites (c) on the LiDAR-derived digital elevation model (DEM). To detect charcoal kiln sites by photo-interpretation, the LiDAR-derived DEM needs adequate transformation to make the relief of the ground appear, previously to manual (Raab et al., 2014; Risbøl et al., 2013) or (semi- )automated detection (Hesse, 2010; Schneider et al., 2015). The performance of detection is highly variable depending on field conditions such as the slope (Raab et al., 2014), but also the size of the sites (Raab et al., 2014), their shape (Schneider et al., 2015) or the visualization algorithm implemented and its parameters (Bennett et al., 2012). Dense low vegetation can interfere with the signal of the ground, decreasing the signal-to-noise ratio and making the detection of kiln sites complicated, if not impossible (Hesse, 2010; Raab et al., 2014). The Public Service of Wallonia recently acquired a complete cover of the region with LiDAR data. Derived products that include a 1m spatial resolution DEM were made public in February 2015. In this chapter, we aimed to use

Chapter 3. Detection of the sites by remote sensing

LiDAR and VIS-NIR remote sensing data to estimate the distribution, density and cover of charcoal kiln sites in Wallonia related to pre-industrial smelting in the late 18th century. We aimed to validate the large extent of charcoal production estimated indirectly by an historical approach (Chapter 2) by direct detection of the sites with remote sensing data. Moreover, an overview of the distribution of pre-industrial charcoal kiln sites in Wallonia and a robust estimate of the total number of sites provides the opportunity (i) to identify charcoal kiln sites under a vast array of soil conditions for the analysis of soil properties (ii) to estimate of the contribution of pre-industrial charcoal kiln sites to soil carbon storage at regional scale. To reach these goals, we first delimited the potential area for charcoal production based on the forested area on the map of Ferraris (created when charcoal-based smelting reached a peak in Wallonia) restricted to locations that were adapted for charcoal production. Second, we randomly generated sampling plots in this area in which we enumerated the sites, by interpretation of the LiDAR-derived DEM in forest and by photo-interpretation of orthoimages on bare cropland soil. The results were then extrapolated to the potential area for charcoal production. The performance of kiln sites detection with LiDAR data was established by comparison with field detection, and the limits of the approach were discussed.

3.2. Materials and methods

Field database

One hundred and ninety height forest charcoal kiln sites were detected on the field between September 2010 and January 2015 in 26 different forested areas all around Wallonia, previously to the publication of the LiDAR data. On the field, irregularities in the topography were first identified and then validated or invalidated as charcoal kiln sites by soil augering, on the basis of soil color (kiln topsoil is very black compared to the adjacent reference soils) and the presence of charcoal residues. Each site was located with GPS coordinates (GPS map 62, Garmin). The diameter of the sites and the thickness of the charcoal-enriched topsoil were measured for 132 of them, which have a median diameter of 10.0 m and a median topsoil of 33.8 cm in thickness, largely enriched with charcoal residues (Figure 3.3).

Chapter 3. Detection of the sites by remote sensing

Figure 3.3. Frequency histogram of the diameter of 132 charcoal kiln sites of Wallonia (a) and of the thickness of their charcoal-enriched topsoil (b).

LiDAR data

LiDAR data were acquired between the 12 December 2012 and the 9 March 2014, with an average pulse density of 1.51 pulse/m² for all Wallonia. A 1 m spatial resolution DEM was derived from the data. The missing values were interpolated by Voronoi polygonation. The LiDAR-derived DEM was released online by the Public Service of Wallonia in February 2015. To make the ground relief appear for the detection of the charcoal kiln sites, two different algorithms, slope and hillshade, were applied to the DEM thanks to the software ArcMap 10.4 (© ArcGIS). The slope algorithm produces a raster that gives a slope value for each pixel, therefore, a kiln site appears either as a platform surrounded by local steep slope values or as a platform cutting the slope, depending on the topography (Figure 3.4a). We applied a vertical slope exaggeration factor of 5, which is optimal for the detection of the kiln sites (David Novak, personal communication). The hillshade algorithm provides an illustrative representation of the topography so that it is widely used for archeological interpretation (Kokalj et al., 2011). Ground relief is illuminated with an imaginary light source at constant zenithal and azimuthal angles. Pixels exposed to the light beam appear illuminated whereas pixels with a lesser exposition are in shade (Kokalj et al., 2011) (Figure 3.4b). Quality of detection of features vary to a large extent depending on the angle of incidence of the light beam (Devereux et al., 2008), the contrast of the image and the visualization scale (Raab et al., 2014). We chose a 315° azimuth angle and a 45° zenith angle for the light source and visualized the data at a scale ranging

Chapter 3. Detection of the sites by remote sensing between 1/1.500 and 1/3.000. At regional scale, (semi-)automated detection of the charcoal kiln sites is difficult to implement due to the variability in the pattern of the kiln sites. Therefore, manual detection of the sites was preferred, based on the joint visualization of the slope and hillshade transformed DEM.

Figure 3.4. Illustration of the transformations of the LiDAR-derived digital elevation model by the algorithms slope (a) and hillshade (b) that were visualized jointly for the detection of the charcoal kiln sites in forest. Each “button” on the images is one charcoal kiln site.

Definition of the potential area for charcoal production

Traditional charcoal production always occurred under forest, to avoid the transport of wood that is much heavier than charcoal (Lepoivre and Septembre, 1941). Accordingly, the presence of charcoal kiln relics attests that the area was forested at the time of charcoal production (Foard, 2001). In Wallonia, the peak of charcoal-based smelting occurred in the late 18th century, shortly before charcoal was replaced by coal as an industrial combustible in the early 19th century (Heuschling and Van der Maelen, 1838). Therefore, we assumed that the forested area mapped by Ferraris in 1770-1778 73

Chapter 3. Detection of the sites by remote sensing reliably delimits the area potentially affected by traditional charcoal production in the late 18th century. Nevertheless, this area must be restricted to locations where field conditions were adapted to charcoal production. Steep slopes, waterlogged environments and clay-rich soils were not suitable for an efficient wood pyrolysis (Lepoivre and Septembre, 1941). To fix threshold values for conditions (slope, waterlogging and texture) that prevented in situ charcoal production, we referred to field observations. No charcoal kiln site was observed on slopes > 30 % and on permanently waterlogged soils. Some charcoal kiln sites were observed on clay-rich soils, but they were systematically absent from superficial calcareous outcrops, very shallow and rich in secondary clay minerals. We validated these observations by a preliminary exploration of the LiDAR-derived DEM in several large forest areas covering a diversity of soil conditions and where charcoal kiln sites were present. The LiDAR-derived DEM was crossed with the digital soil map of Wallonia to investigate the presence of charcoal kiln sites in relation to soil properties such as stone load, waterlogging and texture. To test the interaction between the steepness of slope and the presence of charcoal kiln sites, slopes were extracted from the LiDAR-derived DEM resampled on a 10 x 10 m grid to smooth the microtopography. On this basis, we defined the potential area for charcoal production as the forested area in 1770–1778 on slopes < 30 %, not permanently waterlogged or located on calcareous outcrops.

Detection of charcoal kiln sites by remote sensing

Since 1770-1778, the Walloon forest was subject to mutations, including deforestation, conversion of hardwood forest to softwood forest and net afforestation (Kervyn et al., 2014) (Figure 3.5). The deforested areas were mainly converted into cropland, grassland or urban areas. In grassland and particularly in cropland, soil tillage erased the relief of the sites partly to completely, which made impossible the detection of the sites on the LiDAR- derived DEM. Nevertheless, former charcoal kiln sites are clearly visible on bare cropland soils as circular to elliptical black spots of a few decameters in diameter, due to charcoal enrichment darkening the soil. Thus, high-resolution aerial photographs or satellite imagery is an alternative for detection on bare cropland soil (Figure 3.6).

Chapter 3. Detection of the sites by remote sensing

Figure 3.5. Mutations of the Walloon forest since 1770–1778; Redrawn from Kervyn et al. (2014)

Figure 3.6. Charcoal kiln sites (black spots) on bare cropland soils of Wallonia on four sets of orthoimages (© Service Public de Wallonie). Each orthoimage shows the same location with a different combination of bare fields, in which the detection of kiln sites is possible.

In 1770–1778, there was exclusively hardwood forest in Wallonia. Since then, part of the former forest was converted into mixed or softwood forest (Figure 75

Chapter 3. Detection of the sites by remote sensing

3.5). Most of the new forest is also coniferous, mainly Picea abies and Abies nordmanniana. In contrast to hardwood forest, charcoal kiln site detection with the LiDAR-derived DEM is often not reliable in softwood forest for two reasons. First, the vegetation cover in coniferous plantations is generally very dense, which can sharply decrease the transmission of the LiDAR signal across the canopy and result in large interpolated areas on the LiDAR-derived DEM (Figure 3.7). Second, forestry operations are very frequent in coniferous plantations, including wood transportation or tree stump extraction. As a result, the relief of the sites is often disturbed or even erased, which also makes the detection impossible on LiDAR data. Therefore, we excluded softwood and mixed forest from the sampling area to limit the number of false negative detections.

Figure 3.7. Area of softwood forest surrounded by hardwood forest. (a) Orthoimage of the Public Service of Wallonia; (b) Ground relief on the LiDAR-derived digital elevation model (DEM) visualized by the hillshade algorithm; Large zones of the DEM are interpolated because the dense cover of coniferous trees lowers or even annihilates the transmission of the laser signal. In contrast, hardwood forests are generally old-growth forests that have rarely been subject to anthropogenic disturbances that destroyed the relief of the ground. Moreover, a deciduous canopy is generally less dense than that of a coniferous forest and never completely stops the transmission of the laser signal. The areas where charcoal kiln site detection by LiDAR or VIS-NIR remote sensing is valid, according to soil occupation, are summarized on Figure 3.8.

Chapter 3. Detection of the sites by remote sensing

Figure 3.8. Summary of the areas where charcoal kiln site detection by LiDAR or visible and near infrared (VIS-NIR) remote sensing is valid, according to soil occupation.

Sampling in forested and deforested areas

Sampling was restricted to the fraction of the potential area for charcoal production where detection by remote sensing is valid (Figure 3.8). Forest and cropland were sampled separately. In forest, four hundred fifty points were generated randomly. A circular buffer of 150 m in radius was created for each point. The area of sampling was delimited by intersecting the buffers with the surface area defined as valid for the detection of charcoal kiln sites. Only buffers containing > 75 % of the initial area were retained, which corresponds to 204 samples. In sampling areas, the charcoal kiln sites were detected manually based on the joint visualization of the slope and hillshade transformed LiDAR-derived DEM (Figure 3.4). Thirty samples were rejected manually because they contained unexpected interpolated data or prominent striations giving evidence of past soil tillage, generally located in peri-urban areas. The center of each relief interpreted as a kiln site was precisely located and the inner diameter of the sites, defined as the flat part of the platform, was measured. The kiln sites were enumerated and reported to the exact area of each buffer to calculate the density of sites per unit of area.

In deforested areas, one hundred points were generated randomly. Depending on the geographical region, cropland and grassland do not occupy the same proportion of the area. There are almost exclusively field crops to the North

Chapter 3. Detection of the sites by remote sensing of the Sambre-and-Meuse line3, whereas grassland represents an important fraction of agricultural area to the South of the Sambre-and-Meuse line. To avoid the introduction of a bias by sampling more often in regions where field crops are dominant, we sampled the field closest to each sampling point showing bare soil on at least one orthoimage. Photo-interpretation relied on four sets of orthoimages of the Service Public of Wallonia, acquired respectively in 1994–2000, 2006–2007, 2009–2010 and 2012–2013 (Figure 3.6). Samples falling in large artificialized areas were rejected. Fields containing the points were used as sampling units. In total, 72 fields were sampled. Black spots with an equivalent diameter larger than 10 meters were enumerated. The density of sites was calculated based on the area of the field. The sites were not measured because their sizes vary to a large extent depending on the tillage history of the plot that caused the lateral dilution of the sites and made their limits very diffuse.

False positive and false negative detection with LiDAR data

The rate of false negative detection (omission error) was established thanks to the field database established previously to the publication of the LiDAR data (see section 2.1 of this chapter). The 26 forest places where charcoal kiln relics were located on the field were posteriorly explored with the LiDAR-derived DEM for the detection of all features considered as charcoal kiln sites. The rate of false negative detection corresponds to the proportion of charcoal kiln sites that were detected on the field but undetected on the DEM. The rate of false positive detection (commission error) was calculated based on the detection of 57 relief features interpreted as charcoal kiln sites on the LiDAR-derived DEM located in four forested areas covering a wide range of soil and topographic conditions. These relief features were then validated or invalidated as charcoal kiln sites on the field by soil augering. The rate of false positive detection corresponds to the proportion of relief features that were incorrectly interpreted on the DEM as charcoal kiln sites.

3 The Sambre-and-Meuse line is a natural limit made by two large rivers, the Sambre and the Meuse, crossing Wallonia from West–South-West to North-East. This line is clearly visible on Figure 3.11 and Figure 3.13. The Sambre-and-Meuse line corresponds to the South limit of the loess belt of Belgium. 78

Chapter 3. Detection of the sites by remote sensing

3.3. Results and discussion

False positive and false negative detection

Among the 198 kiln sites inventoried in the field database, 140 were correctly identified as pre-industrial charcoal kiln sites by interpretation of the LiDAR- derived DEM, which corresponds to 29.3 % of false negative detection overall (Table 3.1). Table 3.1. Rates of false negative detection of charcoal kiln site in forest with LiDAR data. Number of these Rate of false Number of sites sites detected on negative detection detected on the field LiDAR (%)

Total 198 140 29.3

North to the Sambre-et-Meuse 52 24 53.8 Line South to the Sambre-et-Meuse 146 116 20.5 Line However, if we split the database into two subsets located respectively to the North and to the South of the Sambre-et-Meuse line, we obtain contrasting results, with a rate of false negative increasing to 53.8 % for the northern region (n=52) and dropping to 20.5 % for the southern region (n=146). The low rate of detection to the North of the Sambre-et-Meuse line is attributed, in part at least, to the presence of a dense cover of low vegetation in many investigated areas. Most of the soils of this region are Luvisols developed in quaternary loess and have a high agricultural potential. The very few areas that remained forested are subject to much anthropogenic disturbance, including nutrient inputs from agriculture that favor the propagation of ruderal species (Jacquemin et al., 2014). Consequently, many of these small forest areas are colonized by brambles (Figure 3.9a), which are generally absent from the large forested areas to the South of the Sambre-et-Meuse line, less impacted by agriculture (Figure 3.9c). The presence of dense low vegetation interferes with the LiDAR signal of the ground (Hesse, 2010; Raab et al., 2014) and, therefore, makes the detection of charcoal kiln sites difficult on the LiDAR-derived DEM (Figure 3.9b). The forested area North to the Sambre-et-Meuse line corresponds to less than 5 % of the forest mapped by Ferraris. Therefore, we expect very few samples to have been subject to low detection rates caused by the presence of brambles. 79

Chapter 3. Detection of the sites by remote sensing

The main part of forested areas stands to the South of the Sambre-et-Meuse line, where the best negative detection rates were obtained, with about 80 % of sites correctly detected by interpretation of the LiDAR-derived DEM (Table 3.1).

Figure 3.9. Dense low vegetation interferes with the signal of the ground, and therefore with the detection of charcoal kiln sites. (a) Small forest area located to the North of the Sambre-et-Meuse line, invaded by brambles and (b) corresponding relief of the ground surface on the LiDAR data; (c) Large forest area located to the South of the Sambre-et-Meuse line with limited low vegetation and (d) corresponding relief of the ground surface on the LiDAR data. For the determination of the rate of false positive detection, all sites were sampled to the South of the Sambre-et-Meuse line (Table 3.2). We obtain a rate of false positive detection of 15.7 %, which is slightly smaller than the rate of false negative detection in this part of Wallonia. False positive detections were due to anthropogenic or natural reliefs that were confounded with kiln sites, or to wood residues piled after exploitation of a forest stand. Table 3.2. Rates of false positive detection of charcoal kiln sites in forest with LiDAR data.

Number of sites detected Number of sites confirmed on Rate of false positive on LiDAR the field detection (%)

57 48 15.7

In a study in the southern black forest in Germany, Ludemann (2012) investigated the validity of charcoal kiln sites detection with LiDAR data in a one km² test area. As a result, 104 structure identified by LiDAR were

Chapter 3. Detection of the sites by remote sensing confirmed as charcoal kiln sites on the field, whereas 20 structures were other relief features. By a systematic field investigation, 49 charcoal kiln sites undetected on LiDAR were inventoried. This corresponds to a rate of false negative detection of 32.0 % and a rate of false positive detection of 16.1 %. Both rates are close to the values that we obtained in this study (29.3 and 15.7 %, respectively). Nevertheless, the false negative detection rate calculated based on the data of Ludemann (2012) is substantially larger than that obtained for data from the South of the Sambre-and-Meuse line (20.5 %), after removing points where the LiDAR signal was noisy. An aspect that might explain a better detection of the sites with LiDAR in this study with respect to that of Ludemann (2012) is the joint visualization of the data with the “slope” and “hillshade” algorithms, whereas Ludemann (2012) used hillshade only, with a unique angle of illumination. The “slope” algorithm was very reliable for the detection of the sites, particularly to identify the terraces left on slopes. Hillshade was used mainly to confirm the presence of sites in relatively flat areas, when the slope is less than a few percent (160 structures out of 1313 sites in total). By comparison of various detection techniques, Benett et al. (2012) obtained a maximum rate of detection rate of archeological features of 77 % for one individual technique, which supports the view that the combination of visualization algorithms can increase the detection rate (Bennett et al., 2012; Raab et al., 2014; Schneider et al., 2015). Ludemann (2012) also studied the hillshade footprint of 2448 sites previously recorded on the field. In total, 1994 sites were visible on the hillshade, corresponding to a rate of false negative detection of 18.6 %, which accords with our results. Nevertheless, he also indicated that 668 out of the 1994 sites were not easily visible; therefore, part of these sites might have been omitted in a detection exercise without prior knowledge of the location of the sites. It would be of interest to repeat the identification with a combination of detection techniques to determine whether the detection rate is improved or not.

Sites distribution and density

Total forested area on the map of Ferraris is 415.315 ha (Kervyn et al., 2014), including 290.611 ha that remained forested and 123.704 ha that were deforested (Table 3.3). On the current territory of Wallonia, 59.021 ha were not mapped by Ferraris.

Chapter 3. Detection of the sites by remote sensing

Table 3.3. Evolution of the forest mapped by Ferraris (1770–1778)

Area (ha) Fraction of total (%)

Forested area on Ferraris’s map 415.315 100

Area that remained forested 291.611 70

Deforested area 123.704 30

Area unmapped by Ferraris 59.021

In the 291.611 ha that remained forested since the late 18th century, 98.525 ha were removed from the sampling area in forest because they have been converted to softwood or mixed forest. In the 193.085 ha of deciduous forest, 160.493 ha delimited the sampling area, after exclusion of places where soil conditions were unsuitable for charcoal production. Slope is the main factor that restricted sampling area to 83 % of total deciduous forest (Table 3.4). Table 3.4. Definition of potential area of charcoal production according to land use.

Soil occupation Total area Potential area for charcoal production

ha ha %

Area that remained forested 291.611 249.370 86

Old deciduous forest 193.085 160.493 83

Old softwood or mixed forest 98.525 88.877 90

Deforested area 123.704 - -

Cropland 41.485 41.183 99

Grassland 34.769 33.817 97

Others 47.450 - -

In the 123.704 ha that were deforested, 41.485 ha stand as cropland and 34.769 ha as grassland. These areas were deforested because of their high agronomic value related to deep, well-drained soil with no or low stone load in areas with no or gentle relief. Such conditions were also adapted for in situ pyrolysis, which explains for the high percentage of this area potentially affected by charcoal production (99 and 97 %, respectively). Other land uses account for

Chapter 3. Detection of the sites by remote sensing

47.450 ha, corresponding mainly to artificialized areas (Table 3.4). We were not able to calculate the fraction of this area suitable for charcoal production because soil properties like drainage and texture are not given on the digital soil map of Wallonia for artificial soil. Therefore, we assumed that they had a potential for charcoal production similar to other deforested areas at the time of charcoal production. Among deforested areas, cropland is the only land use that allows site detection with orthoimages. 173 plots were sampled in the 160.493 ha of deciduous forest defined as suitable for charcoal production, which corresponds to an area of 1.060 ha investigated for the detection of charcoal kiln sites with the LiDAR-derived DEM. 164 of the plots contained charcoal kiln sites (94.3 % of samples) (Table 3.5). In addition, 72 fields were sampled in the 41.183 ha of deforested area suitable for charcoal production, and 66 contained charcoal kiln sites (91.7 % of samples). Overall, 93.9 % of samples contained kiln sites. In total, we enumerated 2.042 kiln sites, including 1.313 in forest and 729 in cropland. The ubiquity of charcoal kiln relics in the forest of Ferraris highlights the large extent of charcoal production in the late 18th century in Wallonia. Table 3.5. Statistics of sampling in forested and deforested areas

Forested Deforested

Area suitable for sampling (ha) 160.493 41.183

Number of samples 173 72

Area of sampling (ha) 1.060 615

Samples with sites detected 164 66 Proportion of samples containing kiln sites 94.3 91.7 (%) Number of kiln sites detected 1.313 729

Distribution, density and covering of charcoal kiln sites in Wallonia

Samples from forested and deforested areas were aggregated to investigate the regional distribution of charcoal kiln sites in Wallonia. The density of the sites ranges from 0 to 5.7 sites per ha, with a mean density of 1.35 and a median density of 1.2 (Figure 3.10a). The distribution is slightly skewed towards high values, with four samples containing more than four sites per ha.

Chapter 3. Detection of the sites by remote sensing

Figure 3.10. Distribution of density (a), inner diameter (b) and cover (c) of pre- industrial charcoal kiln sites in Wallonia. Data of site density includes samples from forested and deforested areas (n=246) whereas that of inner diameter and cover is calculated based on forested samples only (n=164). Site cover was calculated based on the Estimate of outer diameter of the sites (inner diameter multiplied by 1.35). On the boxplots, the horizontal black line indicates the median of the distribution and the open diamond indicates the mean. The box spans from the first to the third quartile of the distribution. At regional scale, we observe that eight out of the 15 samples that do not contain kiln sites are located in the northwest part of Wallonia, in provinces of Hainaut and Brabant (Figure 3.11). Three more samples are located in the east of Wallonia, in the province of Liège. In contrast, in the “Entre-Sambre- et-Meuse” and in the southern part of the province of Luxembourg, charcoal kiln sites were identified in every sample, and densities of > 2 sites per ha were often recorded (Figure 3.11). In particular, the South of the province of Luxembourg contains a median density of 2.3 sites per ha, which is significantly larger than the overall site density of 1.2 calculated for Wallonia (P<0.001; Wilcoxon non parametric test).

Chapter 3. Detection of the sites by remote sensing

Figure 3.11. Location of the sampling points for the detection of charcoal kiln sites. Samples from forest and deforested areas were gathered (n=246). The size of the circles indicates the density of charcoal kiln sites (number of sites per ha) in sampling areas. The distribution of the density of charcoal kiln sites of the sampling points is presented with respect to the location of ancient Forges inventoried by Evrard (1956) (Figure 3.11). In the late 18th century, the region of Wallonia located to the North of the Sambre-et-Meuse line was already greatly deforested as a consequence of the great agronomic potential of the soils in this region. The limited amount of wood resources and, most likely, the absence of iron ore explain why no blast furnace stands in this area. The pressure on wood resources for charcoal production appears to have been small in the northeast of Wallonia (province of Liège), too. Smelting operations in the surroundings of Liège were mainly dedicated to the transformation of pig iron into end products (Hansotte, 1980), which required less charcoal than smelting the iron ore to produce pig iron. The scarcity of iron ore is another factor that might have limited the implantation of blast furnaces in this province. Nevertheless, very few sampling points occur in this area despite the presence of relatively large forest bodies on the map of Ferraris (Figure 3.11). This is because an important fraction of these areas were converted to softwood forest after the

Chapter 3. Detection of the sites by remote sensing period of charcoal production, which justified their exclusion from the sampling area. The few forested areas of Wallonia that were not used for charcoal production were probably too far from smelting operations for a profitable production of charcoal because of the cost of transport (transport revolution occurred in the 19th century), and have apparently not been allocated to the production of charcoal for any other domestic or industrial use. In contrast, all sampling points contain charcoal kiln sites in the direct vicinity of blast furnaces of the “Entre-Sambre-et-Meuse” and of the South of the province of Luxembourg (Figure 3.11). This observation highlights that charcoal production was mainly related to pre-industrial steel industry in Wallonia, as reported by many historical records (Hardy and Dufey, 2012a). Several sources attest the close link between pre-industrial smelting and charcoal production in other places of Western Europe. Important productions of charcoal related to pre-industrial metallurgy were also reported in the Rockingham forest, England (Foard, 2001), in the Vosges, the Swabian Alps and neighboring regions (Ludemann, 2010) and in the Netherlands (Groenewoudt, 2005). In the Jänswalder Heide royal forest of Lower Lusatia in Germany, Raab et al. (2014) detected on a LiDAR-derived DEM 5500 relief features attributed to charcoal kiln sites on an area of 32 km² (or 3.200 ha). This corresponds to a site density of 1.7 sites per ha, which exceeds the average site density estimated for Wallonia, but which is less than the mean site density observed in the South of the province of Luxembourg (2.3 sites per ha). Lower site densities were reported for an area of 1.600 km² (160.000 ha) in the South of the black forest in Germany (Ludemann, 2012). Ludemann (2012) detected charcoal kiln sites manually with LiDAR data, in sampling plots of one km². More than 95 % of samples contained less than 0.5 sites per ha, and samples never exceeded 1.5 sites per ha. Even though the density of sites cannot be directly interpreted in terms of intensity of charcoal production, high site density and regular distribution of the sites in the landscape suggests that clear-cutting of the forest has occurred at least once in the past for charcoal production. Only the demand of an industry can justify such a massive production of charcoal in areas remote from a big city, and smelting industry appears to have been by far the main consumer of charcoal in Europe in pre-industrial times.

Chapter 3. Detection of the sites by remote sensing

Sites diameter and cover

In forest, inner diameters of sites range from 4.8 to 10.7 m and are normally distributed, with a mean value of 7.51 m (Figure 3.10b). Despite local variations in the size of the sites, no clear pattern of distribution appeared at regional scale. By comparison of outer diameter measured on the field with inner diameter measured on the LiDAR-derived DEM for the sites of the field database, we calculated that inner diameters must be multiplied by 1.35 on average to equal outer diameters. After multiplication by 1.35, we obtain an Estimate of outer diameters, which are of 10.1 m on average. Unsurprisingly, this value is similar to average outer diameters of the field database (10.0 m). It also accords perfectly with the characteristic diameters of 8–12 m of kiln sites from the low mountain regions of the Vosges, the Black forest and the Swabian Alps (Ludemann, 2010). In contrast, Raab et al. (2014) reported site inner diameters ranging from 3 to 29 m with average diameters 12.4 m for 757 excavated kiln sites in the Jänswalder Heide royal forest of Lower Lusatia in Germany. A comparable range of values was reported for the Rockingham forest of England (Foard, 2001). This spectrum of values is very large compared to our values, and mean value is significantly larger. In their study, Raab et al. (2014) demonstrated that pine trees (Pinus sylvestris) between 40 and 70 years old had been used for charcoal production in lower Lusatia. This contrast with clearcutting of short-rotation coppice forest reported in historical documents for charcoal production in Wallonia (Goblet d’Alviella, 1927; Lepoivre and Septembre, 1941). Exploitation of old growth forest for charcoal production might have resulted in a larger amount of wood per unit of area and therefore justify the construction of larger structures, made possible by the flat topography in this area (Raab et al. 2014). They also identified various structures < 6 m in diameter, which is another discrepancy with our results. In this regard, Raab et al. (2014) have relied on a large field database of about 800 excavated kiln sites previously described by Rösler et al. (2012). Consequently, they were able to identify structures of a small size, possibly dating from another period of charcoal production. Comparable structures have not been taken into account in our sampling. Small black spots were sometimes observed in the immediate vicinity of larger structures in deforested areas of the bottom of the province of Hainaut (Figure 3.12). In the fields containing various black spots of different sizes, reddish spots are also often observed (Figure 3.12). They might correspond to places where the iron ore was crushed before smelting in a low furnace, in times preceding the invention of blast furnace (personal

Chapter 3. Detection of the sites by remote sensing communication of Christophe Colliou, archeologist). This supports the idea that smaller structures are not related to the peak of pre-industrial charcoal production of the late 18th century and justifies that we did not take them into account in our survey.

Figure 3.12. Field containing small and large black spots that might result from two or more periods of charcoal production. Reddish spot might correspond to the place where iron ore was crushed before smelting in a low furnace, in times preceding the invention of blast furnace (personal communication of Christophe Colliou, archeologist).

Figure 3.13. Map of the cover of charcoal kiln sites (m²/ha) in samples from forest (n=164). By crossing the density of sites in forest with site area calculated from outer diameters, we obtained site cover per unit of area (m²/ha). Assuming that all 88

Chapter 3. Detection of the sites by remote sensing sites were used for charcoal production at the same time, site cover is indicator of the pressure of charcoal production on wood resources. Site cover ranges from 7.1 to 475.5 m²/ha, with a mean value of 119.6 m²/ha and a median value of 104.3 m²/ha, which corresponds to slightly more than one percent of the forest sampling area. In cropland, site cover was not measured but is probably two to three times larger than in forest, because the sites are diluted by repeated tillage over time. Like for site densities, the distribution is skewed, with some spreading in the range of high values (Figure 3.10c). At regional scale, we observe that site cover is particularly high in the South of the province of Luxembourg, like it was observed for the density of sites (Figure 3.13). It is difficult to establish the relationship between the size of one charcoal kiln site and the amount of wood that was pyrolysed, because the height and the shape of the mound as well as the fraction of the kiln site that was occupied by the mound are uncertain. Nevertheless, there should be a close link between the volume of charred wood and the area of the charcoal kiln sites. Several sources accord on the fact that the typical volume of a pre-industrial Walloon mound kiln was about 50 steres (Lepoivre and Septembre, 1941). If we assume that this typical volume corresponds to a kiln site of about 10 m in outer diameter (median value of the field database), the pyrolysis of 50 steres of wood would require ~78.5 m² of land (the area of a circular platform of 10 m in diameter). According to this ratio, the median site cover of ~105 m² ha-1 obtained experimentally corresponds to the pyrolysis of ~67 steres of wood per ha. 67 steres of wood correspond to ~74 % of the production of a coppice of 20 years in Wallonia, which produces ~90 steres of wood on average (Francoeur et al., 1824; Institut pour la Forêt, 2012). At the scale of Wallonia, this percentage is reduced to 67 % if we take into account that a fraction of the forest area on the map of Ferraris was unsuitable for charcoal production but contributed nonetheless to the production of wood. If we consider the mean site cover (119.6 m² ha-1) rather than the median, this last value increases to 76 %. Values obtained from this rough conversion of the site cover into a volume of wood used for charcoal production accord with the results of the historical approach. Indeed, it was estimated from the production of pig iron that the equivalent of 75 % of the forest area mapped by Ferraris was necessary to meet the demand of the metallurgical industry for charcoal in the late 18th century (Chapter 2). Given the high proportion of available wood resources that were needed to supply fuel to the smelting industry, it is obvious that in

Chapter 3. Detection of the sites by remote sensing regions where charcoal production was the most intense, like in the South of the province of Luxembourg, all wood resources, or close to, were allocated to the production of charcoal in the late 18th century.

Estimate of the number of sites in Wallonia

To provide an Estimate of the number of charcoal kiln sites in Wallonia, we multiplied the density of sites by the potential area for charcoal production. We log-transformed the data of site density to make the distribution symmetrical and calculated the mean and a 95 % confidence interval of the log-transformed data. Calculations were made on the basis of the geometrical mean and confidence intervals, after back transformation of the results. The contribution of areas that have been artificialized since 1770-1778 and areas that have not been mapped by Ferraris (because they were not part of the Austrian) can either be neglected or integrated in the calculation. To address this uncertainty, we present four different scenarios for regional estimates (Table 3.6). Table 3.6. Number of charcoal kiln sites in Wallonia estimated according to four scenarios for the area of extrapolation. Results are calculated according to the geometrical mean of the density of sites per unit of area, and 95 % confidence intervals (IC 95 %) on the geometrical mean.

Scenario - IC 95 % Mean + IC 95 % Artificialized areas included, unmapped areas excluded 394.489 432.469 474.106 Artificialized and unmapped areas excluded 345.279 378.522 414.964 Unmapped and artificialized areas included 410.052 449.530 492.809 Unmapped areas included, artificialized areas excluded 360.842 395.582 433.668 In the first two scenarios, we did not take areas unmapped by Ferraris into account. We first calculated a number of sites based on an area that includes artificialized area. We obtained a number of sites of 432.469 (Table 3.6). If we completely exclude artificialized zones from the area of charcoal production, assuming that the majority of these sites were destroyed, we obtain a number of sites of 378.522, which is the less optimistic scenario. We repeated both scenarios after adding a plausible Estimate for the areas unmapped by Ferraris, considering a similar afforestation rate to that of the natural region it belongs to. As a result, we obtained a number of 449.530 sites for the most optimistic scenario (unmapped and artificialized areas included) and a number of 395.582 sites if we exclude artificialized areas. The 95 % confidence interval represents approximatively 10 % of the values estimated based on the geometrical mean of the density of sites. Regardless of the scenario, the area covered by charcoal kiln sites in Wallonia is about 90

Chapter 3. Detection of the sites by remote sensing

40.000.000 m² (on the basis of site cover estimated for forest sites), which corresponds to a cumulated area of 4.000 ha or 40 km². The scenario including both artificialized and unmapped areas is probably the most representative of the number of sites that were present in the early 19th century, shortly after the peak of charcoal production in Wallonia. On these 449.530 sites, we estimated that 395.582 sites are located in current forest or agricultural areas. These are a potential study site to determine the long-term effect of pre-industrial charcoal production on soil properties in various soil conditions.

Sources of uncertainty in the Estimates

The methodology developed in this study aimed to control the main sources of false negative detection, such as the presence of a dense vegetation cover, to estimate the number of charcoal kiln sites in Wallonia. Nevertheless, given the large extent of the study area, several elements might have affected regional estimates. Here, we discuss these sources of uncertainty that were divided into two main classes, related (i) to the definition of the area of sampling and (ii) to the detection of the sites, respectively.

3.3.6.1. Definition of the area of sampling

The first and maybe the largest source of uncertainty in the extrapolation of results at regional scale is the bias in the location of sampling points, dictated by soil occupation. We have no information for hardwood forest converted to softwood or mixed forest because we excluded them from the sampling area (to limit the rate of false negative detection). Nevertheless, current mixed and softwood forest represent more than 20 % of the forested area mapped by Ferraris. Their exclusion from sampling wouldn’t be an issue if these forests were distributed homogeneously across Wallonia, but that is not the case. An important fraction of these forests stand in the eastern part of Wallonia, where very few sampling points were generated (Figure 3.11). Given the relative distance of these forests from the vicinity of blast furnaces, site density might have been overestimated in this region. Artificialized areas are another challenge for the extrapolation at regional scale. The choice to take into account in the calculation the sites that were excavated or covered with concrete depends on the objective to achieve. Obviously, the contribution of artificialized areas must be integrated if the

Chapter 3. Detection of the sites by remote sensing focus is on wood resources allocated to charcoal production in the late 18th century. We also lack information for the forested area unmapped by Ferraris, especially the former Duchy of Bouillon that contains large bodies of hardwood forest that have certainly been subject to charcoal production. We tried to limit the underestimation by considering an afforestation rate in these unmapped area similar to that of the natural region they belong to, but these unmapped area remain an important source of uncertainty.

3.3.6.2. Detection

Even though we excluded softwood forest from the sampling area, some samples in hardwood forest might have been subject to unsuspected causes of false negative detection, such as topsoil disturbance from forestry operations or natural erosion. False positive detections might also have affected considerably the results for some sampling points. Among circular objects that can be confounded with charcoal kiln sites on the LiDAR-derived DEM, we have large tree trunks, tree falls, burial mounds, backfilled materials, piles of branches after wood harvest, … Nevertheless, we expect that false negative detections somehow compensate false positive detections at regional scale, as we estimated rates of the same order of magnitude for false positive and false negative detections. It is also important to mention that the rates of false negative and false positive detection calculated in this work might vary to some extent depending on the subjectivity of the experimenter and his practice in the identification of the sites. Other issues might have affected the detection of the sites. For instance, the map of Ferraris can be subject to local distortions or geographical shifts up to 100 m (personal communication of Thierry Kervyn, who georeferenced the map of Ferraris). This is a possible source of underestimation of the number of sites in samples located at the edge of a forest area on the map of Ferraris. The smaller the forest area is, the more this effect becomes likely. Therefore, some samples in the small forest bodies from the North of the Sambre-and- Meuse line might have been affected by geographical shifts. Another possible source of underestimation of the sites in cropland is the attenuation of the black color of the sites over time because of the dilution of the site by repeated tillage and, possibly, mineralization of charcoal accentuated by cultivation. Very attenuated sites were observed in the loessic belt of Wallonia, in areas that were deforested about two centuries ago. In such conditions, detection of the site relies greatly on the intrinsic quality of the orthoimage (quality is 92

Chapter 3. Detection of the sites by remote sensing poorer for the set of image from 1994–2000 compared to more recent covers) and surface condition of the field when the photograph was taken. Dussart and Wilmet (1970) reported densities of charcoal kiln sites up to 5 sites per ha in the bottom of the province of Hainaut, identified by photo- interpretation of aerial photographs. We also recorded high site densities in this area. Nevertheless, the detection of sites is sometimes ambiguous, because small black spots are often visible between the large spots (Figure 3.12). These small spots might result from an episode of charcoal production that occurred before the period of interest in this study. Therefore, inclusion of spots of small size in the enumeration of charcoal kiln sites is a subjective choice that can greatly influence the estimation of site densities. Last but not least, the quality of the DEM may vary to some extent from one location to another depending on the plan of airborne scanning and resulting pulse density, as well as atmospheric conditions during the flight, or post-processing of data to derive the DEM. Investigation of these technical aspects was very limited in this survey and could be better documented. Despite these sources of uncertainty, regional estimates made in this chapter based on direct detection of the sites by remote sensing accord with the results of the historical survey developed in Chapter 2, which supports the view that LiDAR is a reliable tool for the detection of charcoal kiln site under forest when detection constraints are controlled.

3.4. Conclusion

In this study, we assessed the distribution, density and cover of pre-industrial charcoal kiln sites in Wallonia by detection of the sites with remote sensing data. Depending on soil occupation, LiDAR and orthoimages were used to detect charcoal kiln sites in the potential area for charcoal production, defined as the forest area on the map of Ferraris restricted to locations where conditions were adapted for charcoal production with traditional mound kilns. LiDAR appeared to be a reliable technique for the detection of charcoal kiln sites under forest. Nevertheless, the dense canopy of some coniferous plantations and the dense herbaceous cover in forests invaded by brambles were shown to increase the rate of false negative detection. We detected charcoal kiln sites in 93.9 % of sampling areas, which strongly supports the idea that a main part of wood resources was needed to meet the demand for charcoal in the late 18th century. Site density is particularly high given the large extent of the study area, with median values of 1.2 sites per ha. The largest density of sites was recorded in the South of the province of 93

Chapter 3. Detection of the sites by remote sensing

Luxembourg (average density of sites of 2.3 sites per ha) and in the Entre- Sambre-et-Meuse, where most blast furnaces were sited. The few samples that were free of charcoal kiln sites stood relatively far away from the location of blast furnaces that were active in the late 18th century, probably because the transport of charcoal over long distances was not cost-effective. This result illustrates the close link between pre-industrial charcoal production and metallurgical activities, and the dependence of pre-industrial blast furnaces on forest resources, as stated in many historical documents. Under forest, the average outer diameter of sites is 10.1 m, which accords with the typical range of diameters recorded in the black forest, the Vosges and neighboring regions in western central Europe. By crossing densities with sites diameter, we calculated a median cover of 104.3 m²/ha, which corresponds to about 1% of the sampling area in forest. From the sampling of forest and cropland areas, we estimated that about 450.000 sites were present in the early 19th century at the scale of the current territory of Wallonia. Today, about 400.000 of these sites stand in forested or agricultural areas. They offer a unique opportunity to study the long-term effect of charcoal on the properties of soil in various conditions, which is developed in the following chapters as the main topic of the dissertation.

Acknowledgements

We are grateful to Sébastien Françoisse, who contributed greatly to this study as a master thesis student. Sebastien developed the methodology of sampling, detected the sites by interpretation of the LiDAR-derived DEM and for forest samples and documented the rate of false positive detection of the sites with LiDAR data. He also made a very good review of literature that represented a solid basis for the writing of this chapter

Chapter 4. Sampling

Chapter 4. Sampling campaigns and description of soils and charcoals.

Summary

In this methodological chapter, the information about sampling campaigns was gathered, and soils and charcoals analyzed in the next chapters (5 to 10) are described. To assess the effect of pre-industrial charcoal production on soil properties, we realized three sampling campaigns of sites under different soil conditions. First, we sampled nineteen charcoal kiln sites under forest by soil horizon, on four different soil types (Arenosols, Cambisols, Luvisols, and Podzols). Second, we sampled the topsoil and the subsoil of 17 sites that had been deforested for cultivation since the time of charcoal production. To address the change of soil properties at kiln site occurring over time, we also sampled the site of a currently active charcoal kiln located close to Dole (France). The soil of kiln site was systematically compared to that of adjacent reference soil, unaffected by charcoal production. To test the effect of cropping on chemical properties and stability of charcoal, we extracted charcoal particles from soil along a chronosequence of land-use change from forest to cropland, up to 200 years of cultivation. History of deforestation and cropping of the fields was traced with the help of historical maps. A selection of charcoal fragments (n=995) were inspected with an incident light microscope with dark field illumination to determine tree species on the basis of characteristic features of wood anatomy. Oak (Quercus sp.) and hornbeam (Carpinus betulus) were dominant in the assemblages.

4.1. Introduction

We describe here soil and charcoal samples that were used in the following chapters of this dissertation. Several sampling campaigns were realized to meet different goals. Some sets of samples were acquired to reach specific objectives of a single chapter whereas others were used in several chapters. We provide an overview of the material that will be used across the document.

4.2. Sampling campaigns

All pre-industrial charcoal kiln sites were sampled in Wallonia (Southern Belgium), a region that covers 16 844 km², with altitudes of between ~100 and 694 m above sea level. The climate is oceanic and cold temperate, with

Chapter 4. Sampling mean annual temperature between 6.4 and 9.5 °C and rainfall from 750 to 1400 mm.

Forest soil (Chapters 5, 6, 7 and 9)

In 2012, the forested area of Wallonia was approximately 545.000 ha, which included about 300.000 ha that were already forested on Ferraris’s map of the Austrian Netherlands of 1770–1778 (Kervyn et al., 2014), when charcoal production reached a peak (Hardy and Dufey, 2012b). We restricted our research of charcoal kiln sites to locations that were mapped as forested by Ferraris (see Chapters 2 and 3). We cross-referenced this with the digital soil map of Wallonia to cover a diverse range of soil types. Among 198 charcoal kiln sites inventoried on the field, we selected 19 sites on four different soil types derived from the weathering of loess, sand, shale, schist, sandstone, limestone, marl and dolomite rock. Regardless of soil conditions, the soil of pre-industrial charcoal kiln sites contrast with adjacent soil unaffected by charcoal production by its very dark, deep topsoil horizon enriched with charcoal residues (Figure 4.1).

Figure 4.1. Soil profile of a pre-industrial charcoal kiln site (a) compared to the natural reference soil and (b) on a dystric Arenosol (AR2).

Table 4.1. Description of study sites, according to the World Reference Base (WRB). Soil texture was determined from textural analysis of the reference topsoil. The soil parent material was determined from the geological map of Wallonia combined with field observations.

Chapter 4. Sampling

The main characteristics of the reference soils, which were unaffected by charcoal production, are listed in Table 4.1. They cover a wide range of textural classes, from sand to clay loam. Their soil types are Arenosols (AR), Cambisols (CM), Luvisols (LV) or Podzols (PZ) according to the WRB 2014 classification (IUSS Working Group WRB, 2014). We differentiated further between acidic (CMA) and calcareous Cambisols (CMC). All soils are acidic to very acidic, except for the three calcareous Cambisols that have pHs values close to neutral in the subsoil. Between November 2011 and May 2012, the soil at each site was sampled with an auger of 8-cm diameter at four points 1 m from the center of the kiln and equidistant from each other. The samples from the same horizon at each point were bulked to form a single composite sample. The reference soil was sampled following the same protocol, but at four points about 5 m from the perimeter of the charcoal kiln site. In total, 101 kiln and 86 reference soil samples were analyzed. There are more kiln than reference soil samples because we often split the kiln topsoil into two or three sub-horizons because of their heterogeneity. In addition, a bleached eluvial horizon, which was absent from the reference soil, was observed under the topsoil of the charcoal kiln sites CMA1 and CMA2.

Cropland soil (Chapters 5, 6, 8 and 9)

We identified former forested areas in Wallonia that had been cleared after the period of charcoal production based on the map of the mutations of the forest of Wallonia since 1770-1778 (Kervyn et al., 2014). In these areas, the altitude ranges between 75 and 300 m above sea level. The climate is oceanic cold temperate, with mean annual temperatures between 8.9 and 10.4 °C and mean annual rainfall of between 859 and 1179 mm. According to the digital soil map of Wallonia, soils that were deforested after 1770-1778 were mainly poorly drained silt loam Luvisols.

Chapter 4. Sampling

Table 4.2. The GPS coordinates, WRB soil type, mean annual temperature (MAT) and precipitation (P) of the study sites.

ID Latitude N /° Longitude E °/ WRB Soil Type MAT /° P /mm 1 50.36608 4.63910 Haplic Luvisol 9.6 937 2 50.37697 4.65450 Haplic Luvisol 9.6 937 3 50.34300 5.07875 Haplic Luvisol 9.6 937 4 49.61652 5.58478 Haplic Luvisol 8.9 1179 5 50.28002 4.73444 Haplic Luvisol 9.6 937 6 50.21583 4.68852 Haplic Luvisol 9.6 937 7 50.53488 4.76052 Haplic Luvisol 10.0 864 8 50.53679 4.76418 Haplic Luvisol 10.0 864 9 50.53432 4.77233 Haplic Luvisol 10.0 864 10 49.96352 4.23288 Eutric Cambisol 9.2 1145 11 49.97483 4.22683 Haplic Luvisol 9.2 1145 12 49.98057 4.28413 Haplic Luvisol 9.2 1145 13 50.47178 5.20392 Colluvic Regosol 9.6 937 14 50.45758 5.17905 Haplic Luvisol 9.6 937 15 50.47093 5.16752 Haplic Luvisol 9.6 937 16 50.55988 4.40287 Haplic Luvisol 10.0 864 17 50.58967 4.50737 Haplic Luvisol 10.0 864

We identified fields containing former charcoal kiln sites by photo- interpretation of high spatial resolution aerial photographs and satellite imagery (Figure 4.2b). On bare cropland soil, the kiln sites are charcoal-rich circular or elliptical spots of 15 to 40 m in diameter that have been diluted by repeated tillage over time (Figure 4.2). We selected seventeen fields randomly, and in each we sampled the topsoil (0–25 cm) and subsoil (35–50 cm) of one charcoal kiln site with a gouge auger. Each sample was bulked from 20 cores taken within a radius of 2 m from the center of the kiln site. We sampled the adjacent soil similarly, from opposite sides of the kiln (10 cores from each side), as reference samples of soil unaffected by charcoal production. Soil types of the reference soil include Haplic Luvisols (15 sites), Eutric Cambisol (1 site) or Colluvic Regosol (1 site) according to the WRB 2014 classification (IUSS Working Group WRB, 2014; Table 4.2). The particle-size analysis of topsoil (USDA texture; IUSS Working Group WRB, 2014) indicated that the texture is silt loam at 15 sites and loam at two sites.

Chapter 4. Sampling

Figure 4.2. (a) Ground view and (b) aerial photograph of pre-industrial charcoal kiln sites on bare cropland soil.

Currently active kiln site (Chapter 5, 6, 7 and 10)

To understand better the evolution of soil properties over time, we also sampled the site of a currently active charcoal kiln in August 2012 close to Dole, France, where the old-fashioned wood charring in an earth mound kiln has been practiced for cultural and tourist purposes since the early 1990s (Figure 4.3). Dole’s climate is very similar to Wallonia’s climate, with mean annual temperature and precipitations of 10.5 °C and 800 mm, respectively. The kiln site is on a silt loam Luvisol developed on lacustrine and alluvial quaternary deposits. Charcoal is produced once or twice a year, always at the same place, generally during the spring and summer. We sampled the soil at the site by soil horizon, following the same protocol as that for the forest charcoal kilns sites of Wallonia.

100

Chapter 4. Sampling

Figure 4.3. Charcoal earthmound kiln of a volume of 10–12 steres, ready for pyrolysis. August 2012 in the Forêt de Chaux, Dole (France).

Charcoals (Chapters 5 and 10)

To test the influence of time and cultivation on the chemical properties and stability of charcoal, particles were extracted from the soil of pre-industrial charcoal kiln sites sampled along a chronosequence of land-use change from forest to cropland, up to 200 years of cropping. We selected fields affected by charcoal production exclusively on Luvisols developed in quaternary loess, a quite homogeneous substrate, to better discriminate between the effect of cultivation and that of other environmental factors on properties of charcoal. For the fields, we identified different episodes of land use change from forest to cropland thanks to a number of historical maps (Figure 4.4). For early deforestation, we referred to the map of Ferraris (1770–1778), that of Vandermaelen (1846–1854), that of the “dépôt de la guerre” (1871–1875) and those of the “Institut de Cartographie Militaire” that include several mapping campaigns between 1878 and 1940. For more recent cultivation time steps, we referred to several maps of the National Geographic institute, completed by interviews of farmers. In total, eight plots with six different cultivation times were chosen, at three different locations. For time zero of cultivation, we chose a plot that remained in grassland after deforestation. Historical maps allowed us to define a minimal and maximal age of land use change from 101

Chapter 4. Sampling forest to cropland, but not an accurate time of deforestation. Therefore, for further data analysis, we attributed to each plot the mean value of each range of possible cultivation time (Table 4.3).

Figure 4.4. Sequence of historical maps showing the evolution of the forested area for the surroundings of Beuzet, in the commune of Gembloux (Belgium). a) Extract of the Ferraris’s map of the Austrian Netherlands (1770–1778); b) Extract of the map of Vandermaelen (1846–1854); c) Extract of the map of the Dépôt de la Guerre (1871– 1875); d) Extract of the map of the military cartographic institute (1901). Charcoal production virtually ceased in the early 19th century, when coke replaced charcoal as an industrial fuel in iron metallurgy, and had stopped by 1860 (Evrard, 1956). Therefore, we assume that our sites were active in the same time period and thereby that land use is the main discriminatory variable that might have affected the evolution of soil and charcoal properties at the kiln sites over time. More precise radiocarbon dating of the kilns is complicated because of the ‘Suess-effect’, which makes difficult to assign a unique date for ages younger than 1650 AD (Reimer, 2013). Moreover, as one

102

Chapter 4. Sampling kiln site was generally re-used periodically, the age of one charcoal particle might not be representative of other charcoal residues remaining at kiln site from other episodes of charcoal production. Table 4.3. Coordinates, land use and years of cultivation at time of sampling for fields of the chronosequence of cultivation.

Lat. N Long.E Time of Land use Dominant feedstocka (°) (°) cultivation(yr) Quercus sp., Carpinus betulus L., Forest - - 0 Corylus avellana L., Betula sp. Carpinus betulus L., Quercus sp., Fagus Grassland 49.996 4.209 0 sylvatica L., Betula sp. Carpinus betulus L., Quercus sp., Fagus Cropland 49.997 4.213 2 sylvatica L. Carpinus betulus L., Quercus sp., Betula Cropland 50.002 4.207 30 sp. Quercus sp., Carpinus betulus L., Fagus Cropland 4.995 4.201 30 sylvatica L., Betula sp. Quercus sp., Corylus avellana L., Cropland 50.534 4.761 126 Carpinus betulus L. Quercus sp., Corylus avellana L., Alnus Cropland 50.541 4.764 152 sp. Cropland 50.521 4.755 200 Carpinus betulus L., Corylus avellana L. Carpinus betulus L., Corylus avellana Cropland 50.588 4.509 200 L., Betula sp. a Refers to wood species identified from microscopic observation of randomly chosen charcoal particles. Despite the uncertainty on the exact moment when the shift from forest to cropland occurred, the total organic carbon (TOC) content of the soil adjacent to the kiln sites, unaffected by charcoal production, decreased progressively with mean cultivation time as estimated from historical maps. This supports the idea that we correctly identified a chronosequence of cultivation (Figure 4.5), as conversion of land from forest to cropland decreases soil carbon stocks by the decrease of SOM inputs because of harvest and the faster decomposition of labile OM (Goidts and van Wesemael, 2007; Guo and Gifford, 2002; Nye and Greenland, 1964; Solomon et al., 2007a). For each field, we sampled the topsoil (0–30 cm) of four charcoal kiln sites with a gouge auger 3 cm in diameter. Each sample was bulked from 20 cores from the center of the kiln site.

103

Chapter 4. Sampling

Figure 4.5. TOC content of soil adjacent to kiln sites vs. time of cultivation. TOC content decreases gradually with cultivation time estimated based on historical maps, which supports that we correctly identified a chronosequence of cultivation. For comparison with charcoal aged in forest soil never subject to cultivation since charcoal production, we refer to the four sites on Luvisols sampled in forest and described previously (Table 4.1), which are the forest equivalent of the sites deforested for cultivation. We also refer to the topsoil of a currently active charcoal kiln close to Dole (France) and characterized previously (Hardy et al., 2016), as a reference subjected to limited ageing. Charcoal particles > 1 mm were separated from soil by way of wet sieving. The residue, containing charcoal particles, was rinsed abundantly with demineralized water and air dried. Charcoal pieces were separated from inorganic material by flotation in water, and then rinsed again several times with demineralized water in a 500 ml beaker, until the water was clear. Plant residues were removed manually from charcoal pieces. Between 50 and > 400 charcoal pieces were collected for each site. From the 18th century, the forest grown for charcoal production was a coppiced woodland of short rotation that was clear-cut every 20 years or less (Goblet d’Alviella, 1927). In the environmental context of the sites, the potential natural forest is dominated by oak (Quercus robur L.), with mainly hornbeam (Carpinus betulus L.) in the understorey (Bohn et al., 2003). To verify the assumption that the forest species were same as for the feedstock used for charcoal production, randomly selected charcoal fragments (n = 995) were identified for at least one kiln site from each plot. The pieces were broken in transversal, radial and tangential planes and viewed with an incident light microscope with dark field illumination at 50 to 500 x magnification (Figure 4.6). Tree species were assigned on the basis of characteristic features of wood anatomy (Schweingruber, 1990) and of a reference collection of modern

104

Chapter 4. Sampling charcoal samples. Except for the modern kiln dominated by beech (Fagus sylvatica), the charcoal assemblages of all the samples were generally dominated by oak (Quercus sp.) and hornbeam (Carpinus betulus) (Table 4.3). Beech (Fagus sylvatica), hazel (Corylus avellana) and birch (Betula sp.) also occurred frequently, whereas maple (Acer sp.), alder (Alnus sp.), dogwood (Cornus sp.), common ash (Fraxinus excelsior), apple subfamily (Maloideae) and willow (Salix sp.) were minor components in the assemblages. The results support that the idea that all pre-industrial kilns originated from exploitation of the same woodland type, with some local variations.

Figure 4.6. Charcoal pieces observed with an incident light microscope with dark field illumination at magnification 200 x. a) Birch (Betula sp.), radial plane; b) Beech (Fagus sylvatica), transversal plane; c) Oak (Quercus sp.), transversal plane; d) Hornbeam (Carpinus betulus), tangential plane.

Acknowledgements

We are extremely grateful to Koen Deforce from the Directorate Earth and History of life for his time and expertise for the anthracological identification of the charcoal remains.

105

Chapter 5. Thermal analysis

Chapter 5. Characterization and quantification of charcoal in the soil of pre- industrial charcoal kiln sites by differential scanning calorimetry.4

Summary

Among the various existing methods to discriminate between black carbon (BC) and uncharred organic carbon in soil, thermal methods have the advantage to be rapid, inexpensive and to require little equipment and sample preparation. Dynamic thermal analysis is particularly interesting because it provides a high density of information on the complete thermal continuum of soil organic matter. In this chapter, we investigated the potential of differential scanning calorimetry (DSC) to identify, characterize and quantify charcoal in the soil of pre-industrial charcoal kiln sites. We analyzed various soil samples from kiln sites and reference soils from a diversity of forest and cropland soils from Wallonia. We also analyzed a set of pre-industrial kiln soils from Germany that were previously characterized with the benzene polycarboxylic acids (BPCA) method (Borchard et al., 2014), to compare thermal characteristics of soil to the BC content estimated by a widely used procedure. We also analyzed a selection of charcoals extracted from the soil of pre- industrial kiln sites and related thermal characteristics to organic and inorganic composition of charcoal. Regardless of the dataset, total heat released during analysis was strongly correlated (r > 0.98) to soil organic carbon (SOC) content, which supports the view that the combustion of SOM and charcoal is the main driver of most heat fluxes recorded by DSC in our soils, with a small effect of minerals. Pre-industrial charcoals and uncharred SOM have a distinct thermal signature that allows their discrimination. Charcoal is more thermally stable overall, which might result from the binding energy of C=C larger than that of C-C, C-H or C-O bonds. Nevertheless, thermal decomposition of charcoal spans over a wide range of temperatures

4 Hardy, B., Leifeld, J., Borchard, N. Characterization and quantification of charcoal in the soil of pre-industrial charcoal kiln sites by differential scanning calorimetry. In preparation

107

Chapter 5. Thermal analysis that overlaps with the thermal signature of uncharred SOM, which stresses the challenge of BC quantification in soil. Our data also showed that aging decreases thermal stability of charcoal. We found a strong positive correlation between the area of the less thermally stable peak of charcoal and O content, which indicates that oxygenation of charcoal particles over time decreases the thermal stability. Soil conditions were also shown to alter the thermal signature of charcoal to some extent, particularly the concentration of Ca in soil. To quantify charcoal-C in base-rich soils, which had relatively constant properties, we proposed an index based on peak height of exotherms attributed to the combustion of charcoal relative to that of uncharred SOC. As a result, the amount of charcoal-C estimated by DSC was linearly related to the content of BPCA-C in soil, but it exceeded BPCA-C content by 5 times. The low recovery of charcoal-C by the BPCA procedure results from a complete destruction of the less condensed fraction of charcoal and, possibly, by the incomplete recovery of the most refractory fraction. Despite the fact that our approach to quantify charcoal-C cannot be generalized to all types of BC and soil conditions, it has the advantage to consider all forms of C in charcoal. Overall, we encourage soil scientist to add thermal analysis to the panel of tools used to characterized SOM in soil because it provides rapidly and at low cost a high density of information on the stability of SOM for the complete continuum of organic materials present in soil.

5.1. Introduction

Black carbon (BC) is the solid residue of incomplete combustion of biomass and fossil fuel, which comprises a wide range of thermally altered materials from slightly charred biomass to highly recalcitrant condensates such as soot, produced by natural or human induced fires or pyrolysis (Goldberg, 1985; Schmidt and Noack, 2000a). A main fraction of terrestrial BC is stored in soil (Forbes et al., 2006; Preston and Schmidt, 2006), where it has a long residence time relative to uncharred organic matter. The increased resistance of BC to (a)biotic degradation has been related to its fused aromatic ring structure (e.g. Solomon et al., 2007). Recently, BC has received much interest from soil scientists because it plays an important role for ecosystems services such as terrestrial carbon storage (Czimczik and Masiello, 2007; Knicker, 2011a; Masiello, 2004; Preston and Schmidt, 2006; Schmidt and Noack, 2000a) and sustainable soil fertility (Glaser et al., 2002, 2001; Glaser and Birk, 2012). Nevertheless, the accurate quantification of BC in environmental matrices is a crucial issue that makes the role of BC unclear in geochemical processes. According to the general definition reported earlier, BC comprises a wide 108

Chapter 5. Thermal analysis range of materials with no clear-cut boundaries, which hinders quantification (Schmidt et al., 2001). The various forms of BC cover a large molecular continuum (Hammes et al., 2007; Masiello, 2004) that reflects contrasting conditions of production (Keiluweit et al., 2010; Wiedemeier et al., 2015). Moreover, properties of chars produced at relatively low temperature overlap with that of uncharred organic compounds naturally present in soil. Consequently, quantification of BC relies on operational definitions depending on specific objectives defined by researchers from very different fields in atmospheric, soil, sediment and paleoenvironmental sciences (Hammes et al., 2007; Schmidt et al., 2001). The various methods record systematic differences because they recover a different fraction of the BC continuum (Hammes et al., 2007). As a result, Schmidt et al. (2001) reported that variation in the BC content estimated by four different methods for one individual sample was up to more than two orders of magnitude. Five classes of techniques of identification and quantification of BC in an environmental matrix exist: physical, thermal, chemical, spectroscopic and molecular markers techniques (Bird et al., 2015; Hammes et al., 2007). The principle of thermal and (thermo-)chemical separation techniques relies on the exposition of the sample to an oxidative treatment in standard conditions, and surviving C is determined by mass loss or elemental analysis, and operationally defined as BC. 13C NMR is sometimes used to identify BC in the residue based on its aromatic signature. A method that is widely used by soil scientists consists in the quantification of benzene polycarboxylic acid (BPCA) markers liberated by digestion of BC in an acid medium (Brodowski et al., 2005b; Glaser et al., 1998). This procedure has two main advantages: (i) it relies on the chemical decomposition of BC into chemical markers that are specific to BC and (ii) it provides information on the degree of crystallinity of BC according to the number of carboxyl groups on the edge of benzene of the BPCA markers, which increases with the degree of condensation (Glaser et al., 1998). Nevertheless, the quantitative character of this analysis is 1 questionable. BPCA-C was shown to correspond to a maximum of of total 2.27 C in charcoal in optimal conditions of recovery (Glaser et al., 1998). Glaser et al. (1998) speculated that the less condensed fraction of chars might be completely decomposed by the strong oxidative treatment, since no or only traces of BPCA with less than three carboxyls were detected (Glaser et al., 1998). It is also supposed that the most refractory, graphitized forms of BC might resist to the digestion treatment preceding BPCA analysis (Brodowski et al., 2005b).

109

Chapter 5. Thermal analysis

In this work, we looked for a method to quantify charcoal-C in the soil of pre- industrial charcoal kiln sites of Wallonia. Charcoal was produced at relatively low temperature (400–450 °C; Emrich, 1985) at kiln sites and therefore might contain a fraction of aliphatic-C and amorphous aromatic clusters of small size that would not be recovered as BC by the majority of existing BC quantification procedures. Moreover, physical, chemical and biological alterations occurring over time in soil are likely to have decreased the stability of charcoal (Ascough et al., 2011), creating H- and O-rich C functionalities (Cheng et al., 2008a; Lehmann et al., 2005) that have a small resistance to oxidation. Consequently, C atoms from these functional groups are not accounted for BC by classic quantification techniques; nonetheless, they play a major role in soil amelioration with charcoal. The best example is the huge cation exchange capacity (CEC) of aged chars that promotes the retention of nutrients in terra preta soils (Glaser and Birk, 2012; Smith, 1980; Sombroek et al., 1993). This large CEC is attributed mainly to O-rich carboxylate groups created by oxidation over time (Liang et al., 2006; Mao et al., 2012). Therefore, it is necessary to include C atoms from these functional groups in the quantification of charcoal-C to reliably estimate the CEC of pre-industrial charcoals and to accurately determine the content of charcoal-C in soil. Among the panel of methods used for BC quantification, thermal methods are convenient because they are rapid, reproducible, inexpensive and require little sample preparation (Plante et al., 2009). Moreover, thermal resistance has been related to the biological availability of chars (Harvey et al., 2012; Plante et al., 2011). Static thermal methods rely on cut-off temperature values to distinguish between BC and non BC components (e.g. Gustafsson et al., 2001). Nevertheless, widely used chemo-thermal oxidation at 375°C is calibrated for soot (Elmquist et al., 2004) and was shown to recover only from 0 to 44 % of C in chars, with no survival for those produced at < 850 °C (Nguyen et al., 2004). In contrast to static methods, dynamic thermal analysis like thermogravimetry (TG) and differential scanning calorimetry (DSC) has the potential to provide information for the entire continuum of materials that compose soil organic matter (SOM) by scanning a sample over a large range of temperatures (Leifeld, 2007; Plante et al., 2009). Leifeld (2007) highlighted that BC has a specific thermal signature that allows an unambiguous discrimination with uncharred SOM, as BC is systematically more thermally stable. Among materials potentially interfering with the signature of BC, only bitumous coal had a thermal stability comparable to that of chars. Kerré et al. (2016) successfully used DSC to quantify charcoal in the soil of pre-industrial

110

Chapter 5. Thermal analysis charcoal kiln sites based on the height of exothermic peaks > 400 °C, attributed to the combustion of charcoal. Nevertheless, the accuracy of their DSC method of BC quantification was decreased by the overlap of exotherms related to the combustion of uncharred SOM in the range of temperature of combustion of pre-industrial charcoals. In this study, we aimed (i) to explore the potential of DSC as a tool to identify and characterize aged charcoal in the soil of pre-industrial charcoal kiln sites of Wallonia; (ii) to establish a method suitable for the quantification of charcoal-C in the topsoil of pre-industrial charcoal kiln sites, recovering all C forms from charcoal and not only the most refractory C fraction; and (iii) to apply our quantification method to a selection of BPCA-characterized pre- industrial kiln soils (Borchard et al., 2014) to compare results with that obtained by a method that is routinely used for characterization and quantification of BC in soil.

5.2. Material and methods

Soil samples

Organo-mineral samples of kiln and adjacent reference soils from forest and cropland were analyzed. Detailed information on these sites is given in Chapter 4. To verify that reference soils were not contaminated by charcoal residues from the kiln site, we added to the sample list three samples from the topsoil of cultivated Haplic Luvisols from fields that were not affected by charcoal production. To investigate the effect of the clay-containing mineralogical background on the thermal signature of soil, we also selected five subsoil samples from relatively clay-rich Argic horizons from forested and cultivated Luvisols that had low soil organic carbon (SOC) content.

Subsoil samples were treated with 6 % H2O2 at 70 °C during 30 days to oxidize SOM with a limited effect on soil mineralogy. The content of SOC that survived oxidation ranged from 0.50 to 0.91 g kg-1 according to elemental analysis (elemental analyzer, HEKAtech). For comparison of DSC results with BPCA biomarkers, we also analyzed 40 kiln and reference forest soil samples from 10 different sites, sampled at two different depths (0–5, 5–20 cm). These had been previously BPCA- characterized (Borchard et al., 2014) following the procedure of Brodowski et al. (2005). Five sites were located in the Siegerland region and five in the Eifel region of Germany. In the Siegerland, soils were Leptic Cambisols (IUSS Working Group WRB, 2014) developed on acidic rock, whereas soils from

111

Chapter 5. Thermal analysis the Eifel region were Haplic Luvisols, Mollic Leptosols and Leptic Cambisols (IUSS Working Group WRB, 2014) formed from the weathering of calcareous rock (Borchard et al., 2014). Soils from Siegerland were very acidic, with median pH values of 3.9 whereas soils from Eifel were base-rich and had a median pH value of 5.5 (Borchard et al. 2014). In both regions, the climate is cold temperate with mean annual temperatures of 8.9 and 7.7 °C and mean annual precipitations of 946 and 717 mm in the Siegerland and the Eifel, respectively. Borchard et al. (2014) indicated that the activity of charcoal production in both study areas had stopped by the end of the 19th century, which supports the idea that the sites were abandoned more than one century ago. Before DSC analysis, the < 2 mm fraction of each soil was ground to powder with an oscillating rings crusher. The total content of C was determined by dry combustion, and corrected for inorganic C to obtain organic carbon content, which includes uncharred SOC and charcoal-C. Previously to DSC -1 analysis, samples exceeding 60 g kg of SOC were diluted with Al2O3 and homogenized in a ball mill. In total, more than 150 soil samples were analyzed.

5.2.1.1. 13C nuclear magnetic resonance-cross polarization magic angle spinning (13C NMR-CPMAS)

To relate the thermal signature of SOM to the proportion of the different types of C, we selected two pairs of kiln and adjacent reference soil, one from forest and one from cropland soil, for solid-state 13C NMR-CPMAS analysis. Spectra were obtained with a Bruker Avance III HD 400 MHz Wideboard (Bruker) at a frequency of 100.65 MHz using zirconium rotors of 4 mm outer diameter with KEL-F-caps, according to (Knicker, 2011b). The cross polarization magic angle spinning (CPMAS) technique was applied during magic-angle spinning of the rotor at 14 kHz. A ramped 1H-pulse was used to circumvent spin modulation of Hartmann-Hahn conditions. A contact time of 1 ms and a 90° 1H-pulse width of 2.5 µs were used for all spectra. Pulse delay between single scans were 300 ms. The 13C-chemical shifts were calibrated to 0 ppm with tetramethylsilane and to 176.04 ppm with glycine.

Charcoals

Several charcoals from pre-industrial kiln sites were analyzed in an attempt to relate thermal characteristics to organic and inorganic composition of charcoal, which can help to set up a reliable procedure of quantification of

112

Chapter 5. Thermal analysis charcoal-C. Soil sampling, separation of charcoal particles from soil and identification of wood species are described in more details in Chapter 4. To compare thermal characteristics of aged charcoal to that of fresh charcoal, we also analyzed a birch charcoal that was produced in a traditional mound kiln in august 2012. Previously to analysis, charcoal particles were ground to a powder with an agate pestle and mortar. In total, 28 charcoal samples were analyzed.

5.2.2.1. Elemental composition (C, H, O, N)

Elemental composition (C, H, N and O) of charcoals was determined with a Euro EA elemental analyzer (HEKAtech). C, H and N were measured via dry combustion and O was analyzed using pyrolysis at 1000 °C. The content of carbonates was estimated by mass loss between 550 and 1000 °C, and the contents of inorganic C and O from carbonates were calculated on the basis of atomic weight. Organic C and O were calculated from the difference between elemental and inorganic content.

5.2.2.2. XPS

To quantify atomic content of major elements (C, O, N, Si, Al, Fe and Ca), charcoal powders were analyzed with a SSX 100/206 photoelectron spectrometer (Surface Science Instruments) equipped with a monochromatized micro focused Al X-ray source powered at 20 mA and 10 kV. Powder was fixed on a stainless steel multi-specimen holder with double sided insulating tape. The analysis chamber was around 10-6 Pa, and the angle between the surface normal and the axis of the analyzer lens was 55°. The pass energy was 150 eV and the area analyzed was ca. 1.4 mm2. In these conditions, the full width at half maximum (FWHM) of the Au 4f7/2 peak of a clean Au standard sample is about 1.6 eV. A flood gun at 8 eV and a Ni grid placed 3 mm above the sample surface were used for charge stabilization (Bryson, 1987). The C-(C, H) component of the C1s peak of carbon was fixed at 284.8 eV to calibrate the binding energy scale. Data were analyzed with CasaXPS (Casa Software Ltd). Atomic fractions were calculated using peak areas after a non-linear background subtraction (Shirley, 1972), based on experimental sensitivity factors and transmission factors provided by the manufacturer.

Differential scanning calorimetry analysis

Soils and charcoals were analyzed by heat flux DSC with a DSC 100 (TA Instruments), which measures a temperature difference between the sample

113

Chapter 5. Thermal analysis and an empty reference beside of it, subjected to a similar heating program. A heat flow rate is calculated based on the voltage signal corresponding to the difference in temperature between the sample and the empty reference (Plante et al., 2009). Between 15 and 25 mg of soil ground to powder were weighed into an aluminum pan and scanned under a flow of 50 ml min-1 synthetic air from room temperature to 600 °C, at a heating rate of 10 °C min-1 (Leifeld, 2007).

Figure 5.1. Measurement of characteristics from DSC thermograms (green line). After identification of peak maxima, peak height (W g-1) was measured as the maximum deviation from a linear baseline drawn between 150 and 600 °C, as illustrated for the main peak of the thermogram (0.9542 W g-1). Total heat release (527.6 J g-1) corresponds to the surface of the graph delimited by the linear baseline. Peak area (J g-1) was obtained by vertical drop between two adjacent peaks based on the position of the minima between two peaks. The black lines indicate the maximum rate of reaction associated to each maximum. After subtraction of a linear baseline drawn between 150 °C and 600 °C, peak temperatures (°C), peak heights (W g-1), temperature of 50 % heat release (T50) and total heat of reaction (J g-1) were measured for each DSC thermogram with the Universal Analysis 2000 software (TA Instruments). Peak area (J g-1) was also measured for charcoals, by vertical drop after identification of a minima between two peaks (Figure 5.1) From repeated measurements (n=6) on one sample, we estimated a 95 % confidence interval for each DSC characteristic that was systematically < 2 % of the measured value.

Standard addition experiment

To calibrate the relationship between DSC characteristics and the content of charcoal-C in soil, we made standard additions of a pre-industrial charcoal to 114

Chapter 5. Thermal analysis an organo-mineral soil assumed to be initially free of charcoal. Previously, we looked carefully at the DSC thermogram of individual samples to choose a reference soil that had a representative pattern of heat release, and that did not seem to be affected by the presence of BC, unless no soil can be guaranteed free of BC (Leifeld, 2007). From the preliminary observation of the thermograms of soils and charcoals, we observed two different patterns of heat release for charcoal: (i) charcoals from acidic forest soils that have one main exotherm at about 400 °C and (ii) charcoals from neutral to alkaline soils (cropland soils, carbonate-rich forest soils) with multiple peaks. Therefore, one standard addition experiment was made for each type of charcoal. For both categories of charcoal, we chose a pre-industrial charcoal with average thermal properties, among investigated charcoals. The soil, diluted twice with -1 Al2O3, contained 13.5 g kg of SOC. Increasing amounts of charcoal were added to aliquots of soil to obtain 2.2, 4.5, 11.2, 22.5 and 45 g kg-1 of charcoal- C in the soil-charcoal mixtures.

Sample saturation with Ca2+

In contrast to spectroscopic techniques, thermal analysis of organic compounds is known to be little specific, with heavy dependence on experimental conditions (Fernández et al., 2010). Preliminary results revealed a strong influence of soil conditions on the pattern of heat release of pre- industrial charcoals. Comparison of DSC characteristics to the atomic composition of charcoal highlighted that Ca might catalyze the oxidation of the thermally less stable peak of charcoal. Between all soils of the dataset, the main factor affecting the chemical environment is soil occupation: forest and cropland soils differ strongly in base saturation and pH due to the use of liming amendments. In an attempt to standardize soil conditions and related thermal characteristics between contrasting soils, a selection of kiln and reference acidic forest soils (n=8) were buffered at pH 7 by equilibration with 1M

CH3COONH4 (a solution naturally buffered at pH 7) by two cycles of agitation-centrifugation in a batch experiment with a soil:solution ratio of 1:25. Then, the soil was rinsed with demineralized water to eliminate the excess of salt (two cycles of agitation-centrifugation with a soil:solution ratio 2+ 1:50). Then, the soil was saturated with Ca by equilibration with 1M CaCl2 (two cycles of agitation-centrifugation with a ratio 1:25). The soil was rinsed again with demineralized water and air dried before DSC analysis.

115

Chapter 5. Thermal analysis

5.3. Results

Thermal analysis of organo-mineral soils

Independently from soil conditions, Reference soils have a characteristic DSC pattern, with a main maximum between 300 and 330 °C, at 310.3±10.0 (mean±s.d.) °C for reference soils from forest (Figure 5.2) and at 319.2±3.7°C for reference soils from cropland (Figure 5.3). This peak is asymmetrical and spreads systematically towards higher temperatures. A smaller peak is sometimes visible in the range of 400–450 °C, particularly in cropland reference soils.

Figure 5.2. Differential scanning calorimetry thermograms of a selection of forest soils from pre-industrial charcoal kiln sites (black curves) and adjacent reference soils (grey curves), from Wallonia (a, b, c, e) and Germany (d, f). Four sites are located on (very) acidic soil (a, b, c, d) and two sites are located on calcareous soil (e, f). The pattern of heat release of cropland soils from a field that was not affected by charcoal production (Figure 5.4a) has a signature very comparable to that

116

Chapter 5. Thermal analysis of cropland reference soils adjacent to kiln sites. Soils of pre-industrial charcoal kiln sites have a more variable signature. In addition to the signal in the 300–330 °C range, they have from one to three additional exotherms of higher thermal stability (Figure 5.2 and Figure 5.3). We observe two main types of DSC signature for kiln soils. In the first category, we have (very) acidic forest soils (Figure 5.2a-d). These have a characteristic main exotherm at 391.8±14.7 °C, and a small peak of higher thermal stability at around 494.8±19.2 °C that is not always clearly visible. The second category comprises cropland soils (Figure 5.3) and calcareous forest soils (Figure 5.2e, f). The main difference with the thermal pattern of forest soils from the first group is the presence of multiple peaks, with an exotherm at 374.7±6.3°C and another at 422.6 ±2.7 °C in place of a unique peak at 400 °C. A striking difference between the two groups of soils is that the first category comprises only very acidic soils whereas the second one contains only soils with an acidity close to neutral due to liming (cropland soils) or to the presence of carbonates from the parent rock (calcareous forest soils). Regardless of the presence of charcoal, most thermograms have a small, sharp endotherm at about 575 °C.

Figure 5.3. Differential scanning calorimetry thermograms of a selection of cropland soils from pre-industrial charcoal kiln sites of Wallonia and adjacent reference soils.

The deep Argic horizons that were H2O2 treated and therefore contained almost no residual organic carbon recorded very limited heat fluxes by DSC 117

Chapter 5. Thermal analysis analysis (Figure 5.4b). In contrast to organo-mineral soils, no heat was released, and a small endotherm was even recorded between 350 and 600 °C, in addition to the same sharp endotherm at ~575°C as observed in most soils.

Figure 5.4. Differential scanning calorimetry thermograms of a) the soil from a field crop on Haplic Luvisol, unaffected by charcoal production and b) of a deep Argic horizon, H2O2 treated.

Figure 5.5. Total heat released between 150 and 600 °C from kiln and reference soils; a) forest soils of this study; b) cropland soils of this study; c) forest soils from Borchard et al. (2014). This result suggests that combustion of SOC from uncharred organic matter or charcoal is the main source of heat fluxes recorded from organo-mineral soils, and that soils minerals, in our soils at least, have minor effects on the

118

Chapter 5. Thermal analysis

DSC signature. Accordingly, Regression of total heat of reaction against SOC content gives a very high determination coefficient (R² ≥ 0.97), regardless of the dataset (Figure 5.5), which confirms the close link between heat released during DSC analysis and the combustion of organic materials. Small intercepts of the linear regressions might express some influence of the mineralogy on heat fluxes.

13C NMR analysis

On 13C NMR-CPMAS spectra (Figure 5.6), dominant carbon chemical bonds were attributed to typical chemical shift regions according to Knicker (2011a). The 0-45 ppm region corresponds to alkyl-C, the 45-110 ppm region to O- and N-alkyl, the 110-160 ppm region to aryl-C and the 160-220 region to carboxyl-, carbonyl- and amide-C. Organic matter of reference soils is dominated by alkyl-C and O- and N-alkyl. In contrast, intensity of the signal in the region corresponding to aryl-C increases strongly in the kiln soil, particularly in forest (Figure 5.6a). Along with the increase in aromaticity, two spinning side bands appear in regions from -50 to 0 ppm and from 225 to 300 ppm. The proportion of aryl-C is smaller in the charcoal kiln site in cropland soil but remains very high (Figure 5.6b).

Figure 5.6. 13C-NMR-CPMAS spectra of charcoal kiln and adjacent reference soil from forest (a) and from cropland (b). Main carbon chemical bonds were attributed to chemical shift region according to Knicker et al. (2011a). Asterisks indicate spinning side bands of aryl-C. Relative to adjacent reference soil, charcoal kiln soil is much enriched in aryl-C.

119

Chapter 5. Thermal analysis

Thermal analysis of charcoals

By superimposition of the DSC thermogram of charcoal particles with that of the kiln soil from which particles were extracted, it appears clearly that peaks recorded at temperature higher than 350 °C are related to the combustion of charcoal (Figure 5.7a).

Figure 5.7. Differential scanning calorimetry thermograms of several charcoals; a) Thermogram of charcoal particles superimposed to the thermogram of the soil of the pre-industrial charcoal kiln site from which particles were extracted, and to the thermogram of adjacent reference soil; b) Thermograms of a selection of charcoals from kiln sites with contrasting history of soil occupation; c) Thermogram of a birch charcoal that was not aged in soil, produced in august 2012 in a traditional mound kiln.

120

Chapter 5. Thermal analysis

Nevertheless, as suggested by the variability of the thermal signature of kiln soils, the pattern of heat release of pre-industrial charcoals is much variable. Particularly, the area and the temperature of the less thermally stable peak of charcoal (peak 1) changes a lot from one charcoal to another (Figure 5.7b). Characteristics of the most thermally stable peak (peak 3) also vary to some extent.

Figure 5.8. (a) Area of the less thermally stable peak (peak 1) against the content of organic O in charcoal, and (b) area of the most thermally stable peak (peak 3) against the content of organic O in charcoal. By comparison of DSC characteristics to organic and inorganic composition of charcoals, we found a strong positive correlation (r=0.91) between the area of the less thermally stable peak and O content, and a strong negative correlation (r=–0.91) between the area of the thermally most stable peak and O content of charcoal (Figure 5.8). We also recorded a high correlation (r=– 0.93) between the temperature of the less thermally stable peak and the concentration of Ca in charcoal (Figure 5.9).

121

Chapter 5. Thermal analysis

Figure 5.9. Temperature of the less thermally stable peak (peak 1) against Ca content of charcoal. The pattern of heat release of pre-industrial charcoals was also compared to that of a fresh charcoal that was never aged in soil, produced in a traditional mound kiln in Dole (France), in august 2012. It is interesting to note that fresh charcoal has a thermal signature very different from that of pre-industrial charcoals (Figure 5.7c). Fresh charcoal has a main peak at 477 °C, with a shoulder at 329 °C, which is lower in temperature than the temperature of the lowest maxima recorded in pre-industrial charcoals (> 355 °C). Overall, the temperature of 50 % heat release of the fresh charcoal is 438°C, which is sensitively higher than that of pre-industrial charcoals that ranges between 388 and 418 °C.

Buffering of forest soils at neutral pH and saturation with Ca2+

A selection of pre-industrial kiln soils from forest, very acidic, were saturated with Ca2+ after buffering the soil at pH 7 (Figure 5.10b). Results seem to confirm the relationship between the thermal stability of the less thermally stable peak of charcoal and the presence of Ca, as previously observed (Figure 5.9). Ca-saturated soil has two discernible maxima at 390 and 419 °C (Figure 5.10b), whereas untreated soil had only one identifiable maximum at 405 °C (Figure 5.10a). As a result, the pattern of heat release of the Ca-saturated soil looks more like that of a cropland soil, or that of a base-rich Cambisol than before saturation with Ca.

122

Chapter 5. Thermal analysis

Figure 5.10. Thermal signature of the soil of a charcoal kiln site from a very acidic forest soil; a) Untreated sample; b) Sample buffered at pH 7 and saturated with Ca.

Quantification of charcoal-C with DSC and comparison with BPCA

For standard additions of a pre-industrial charcoal to a soil initially free of charcoal (Figure 5.11), the total heat release increases linearly with the amount of charcoal-C added to soil (data not shown), which accords with the strong linear relationship between SOC content and total heat release recorded by DSC (Figure 5.5). Similarly, the height of the peaks related to the combustion of charcoal (p1, p2 and p3; (Figure 5.11) increases linearly with the concentration of charcoal-C, with very high coefficients of determination (R² > 0.99; data not shown) for the relationship between peak height and charcoal- C. This was already observed by Leifeld (2007) and is illustrated here for the standard additions of a pre-industrial charcoal from cropland to an organo- mineral soil initially free of charcoal (Figure 5.11).

123

Chapter 5. Thermal analysis

Figure 5.11. Differential scanning calorimetry thermograms of an organo-mineral soil with standard additions of a pre-industrial charcoal from cropland. The three peaks (p1, p2, p3) attributed to the combustion of charcoal increase linearly with charcoal- C. Even though charcoal has a specific thermal signature, it impossible to attribute a temperature range specific to the combustion of charcoal-C to quantitatively discriminate between charcoal and uncharred SOM. Indeed, thermal decomposition of both sources of organic carbon spans over a wide temperature range that overlaps with each other, which stresses the issue of black carbon quantification in soil. Therefore, we formulated an index based on the relative height of peaks to estimate the fraction of charcoal-C and uncharred SOC in soil. For cropland and carbonate-rich soils, we used the sum of heights of the three peaks from the combustion of charcoal (p1+p2+p3) as an indicator of charcoal-C, whereas the height of the peak attributed to the combustion of uncharred SOM (p0; Figure 5.11), multiplied by three (3p0), was chosen as an indicator of uncharred SOM (we obtained a linear relationship by using a multiplicative coefficient of 3 for p0). The relationship between the fraction of charcoal-C in soil (Charcoal-C/SOC) and a peak index (p1+p2+p3)/(p1+p2+p3+3p0) was calibrated based on the standard addition experiment (Figure 5.12).

124

Chapter 5. Thermal analysis

Figure 5.12. Relationship between the fraction of charcoal-C in soil (Charcoal- C/SOC) and peak index (p1+p2+p3)/(p1+p2+p3+3p0) based on the height of differential scanning calorimetry peaks, calibrated based on the standard addition experiment. Cumulated height of peaks attributed to charcoal (p1+p2+p3) is a substitute for charcoal-C, whereas (p1+p2+p3+3p0) is an indicator of total SOC content (R²=0.98). A comparable approach was developed for charcoal from acidic forest soils (data not shown). As these charcoals from acidic soils have only one individual peak clearly identifiable, the index relied only on one peak (p1) for charcoal and the peak for uncharred SOM (p0) and was (p1)/(p1+p0). The relationship between this index and the concentration of charcoal-C was also calibrated based on a standard addition experiment. To test the influence of the shape of the thermogram of charcoal on the estimation, we mathematically simulated soil-charcoal mixtures (n = 18) over a wide range of charcoal-C concentrations based on the DSC pattern of 9 pre- industrial charcoals, extracted from different kiln sites. Simulated mixtures were obtained by summing the thermogram of pure charcoals, at different doses, to the thermogram of a charcoal-free soil. All chosen charcoals had three maxima clearly identifiable by DSC. By comparison of predicted values with calculated values, we obtained a root mean square error (RMSE) of 1.39 % of the amount of charcoal-C added. Thermal characteristics of soils from the study of Borchard et al. (2014) were compared to the BPCA-C content that was previously determined. We found a strong positive Pearson correlation (r=0.935) between the temperature of 50 % heat release and the total amount of BPCA-C in soil, which suggests that thermal stability and chemical signature of charcoal are related (Figure 5.13).

125

Chapter 5. Thermal analysis

Figure 5.13. Temperature of 50 % heat release of bulk soils from the study of Borchard et al. (2014) against the fraction of C from BPCA (BPCA-C/SOC).

Figure 5.14. Fraction of charcoal-C (charcoal-C/SOC) estimated by differential scanning calorimetry against the fraction of BPCA-C (BPCA-C/SOC) of bulk soils from the study of Borchard et al. (2014). We also estimated the content of charcoal-C in the soils from the study of Borchard et al. (2014) thanks to the index previously calibrated (Figure 5.14). As a result, we found a strong linear relationship between charcoal-C content estimated by DSC and the content of BPCA-C recovered from the soil, expressed as a fraction of TOC content (Figure 5.14a) or in absolute terms (Figure 5.14b). For the regression line between charcoal-C and BPCA-C, determination coefficients of 0.89 (Figure 5.14a) and 0.96 (Figure 5.14b) were obtained. According to the slope of the regression line, the content of charcoal- C estimated by DSC overestimates the amount of BPCA-C by more than five times (slopes of 5.6 and 6.0, respectively).

126

Chapter 5. Thermal analysis

5.4. Discussion

Thermal signature of charcoals

Properties and composition of charcoal are known to depend largely on conditions of production such as temperature and heating rate that control the degree of aromaticity and crystallinity of chars (Keiluweit et al., 2010). Accordingly, the high thermal stability of BC has been attributed to its polycondensed aromatic structure (De la Rosa et al., 2008). A positive correlation between the content of aromatic-C estimated by 13C NMR spectroscopy and the proportion of thermally refractory SOM was reported (Harvey, 2012; Leifeld, 2007). The thermal stability of aromatic compounds have larger than that of aliphatic compounds or O- and H-rich C functionalities (Leifeld 2007) is in line with the binding energy of C=C bonds (520 kJ/mol) higher than that of C-C, C-O or C-H bonds (350-412 kJ/mol) (Plante et al., 2009). Nevertheless, the degree of crystallinity might govern thermal resistance beyond aromaticity. Leifeld (2007) showed that hexane soot, charred wood and charred rice straw had the same aromaticity but different thermal stabilities. At comparable aromaticity, thermal resistance depends mainly on the degree of aromatic condensation of char (Harvey, 2012; Leifeld, 2007). Leifeld (2007) proposed temperature of the thermally most stable peak as the most reliable feature to compare thermal stability of charcoal to that of other organic materials. He showed that thermal stability of pine wood charred under N2 increases with charring temperature, and recorded a complete loss of thermally labile compounds at 400 °C, in line with the process of aromatization of charcoal that occurs in the 280–400 °C range of temperature (Bird et al., 2015). Leifeld (2007) also found that the most stable peak of charcoals and charred plant biomass had a temperature > 500 °C. In contrast, fresh charcoal of this study had a main maximum at 477 °C and exhibited a signal in the low range of temperature (< 350 °C). These discrepancies might be explained by (i) differences in the quality of charcoal and (ii) differences in the experimental parameters of the DSC analysis. Leifeld (2007) analyzed his samples with heating rate of 20°C min-1 whereas we used a heating rate of 10°C min-1. Temperature of heat release is decreased by slower heating rates (Fernández et al., 2010; Leifeld, 2007), which can explain for the smaller temperature of the most stable peak measured for our fresh charcoal. On the other hand, the survival of thermally labile compounds for charcoal produced in a traditional mound kiln suggests that temperature of charring was < 400°C.

127

Chapter 5. Thermal analysis

This is not surprising, as wood pyrolysis by the traditional mound kiln method is expected to reach a maximum of 400–450°C (Emrich, 1985), and that maximum temperature of pyrolysis varies within the mound according to the distance from the chimney. In contrast, no signal was recorded in the lower range of temperature (< 350 °C) for pre-industrial charcoals. The presence of aliphatic C compounds, such as proteins and sugars, were detected in chars produced at low temperature and related to the biological accessibility of chars (Fabbri et al., 2012). The amount of labile compounds decreased at higher temperature of pyrolysis (Fabbri et al., 2012). The disappearance of the thermally labile fraction for pre-industrial charcoals aged in soil suggests that this fraction was biologically reactive, probably made of uncharred organic molecules. Therefore, it was more subject to (a)biotic decomposition than the thermally stable fraction of charcoal and was completely degraded since the time of charcoal production. Accordingly, the presence of a labile fraction in engineered biochars was found to contribute to early emissions of CO2 after introduction to soil (Sagrilo et al., 2014). Intriguingly, the thermal signature of fresh and aged charcoals is very different. Aged charcoals are thermally less stable than fresh charcoals and generally have multiple exotherms in the 360–525 °C range. The main process in aging of charcoal is oxygenation starting from the surface and propagating to the core of the particle (Lehmann et al., 2005). By relating thermal characteristics with organic composition of charcoal, we highlighted that the O-rich fraction of charcoal had a specific thermal signature, with a smaller thermal stability than the O-poor fraction (Figure 5.8). This is in line with the smaller binding energy of C-O bond compared to C=C of aromatic clusters in charcoal, as thermal reactivity increases with smaller binding energy (Plante et al., 2009). As a result of oxygenation, the overall thermal stability of aged charcoals is smaller than that of fresh charcoal. We also found that the temperature of combustion of the O-rich C functionalities was strongly negatively correlated to the content of Ca in charcoal (Figure 5.9). The most abundant O-rich functional groups in aged charcoals are carboxyl groups (Lehmann et al., 2005; Mao et al., 2012). These are known to have a strong affinity with Ca2+ (Kalinichev and Kirkpatrick, 2007). The presence of Ca adsorbed to (poly-)carboxylate groups of charcoal might catalyze thermal decomposition by decreasing the binding energy of C-O bonds. Comparably, the presence of Al and Fe in the form of trivalent cations complexed to humic compounds of Podzols was shown to alter thermal stability of SOM (Schnitzer

128

Chapter 5. Thermal analysis et al., 1964). This highlights the importance of soil conditions on the thermal signature of SOM. Transmission electron micrographs of modern and fossil charcoals have provided evidence of organized and disorganized domains in the matrix of charcoal (Cohen-Ofri et al., 2006). The degree of organization, or aromatic condensation (McBeath and Smernik, 2009; Wiedemeier et al., 2015) refers to the size and arrangement of aromatic clusters in BC, which increases with temperature of pyrolysis > 350 °C, once the aromatization of BC is complete or close to (Keiluweit et al., 2010; McBeath et al., 2015). Bird et al. (2015) proposed the existence of three pools of different resistance (labile C, semi- labile aromatic C and stable aromatic polycyclic C) to explain the reactivity of BC. According to the conceptual model of Bird et al. (2015), labile C corresponds to the fraction of BC that is composed of minor pyrolysis products such as anhydrosugars and methoxylated phenols, mineralizable on the (very) short-term; semi-labile aromatic C corresponds to aromatic C with a low degree of aromatic condensation; and stable aromatic polycyclic C to polycyclic aromatic clusters with a ring size > 7. This view of the continuum of stability of BC is reconcilable with our DSC measurements. Labile BC is mainly composed of aliphatic C and therefore might correspond to the fraction of charcoal degraded in the low range of temperature (300–350 °C), like uncharred SOC. Semi-labile aromatic C would thus correspond to the fraction of intermediate thermal resistance, decomposed in the range 350–450 °C, constituting the main fraction of charcoal produced in a mound kiln. Stable polycyclic aromatic C might correspond to the fraction of BC that decomposes at ~500 °C. This fraction is relatively small for our pre-industrial charcoal produced at a low temperature (400–450 °C) (Emrich, 1985). Nevertheless, this view of the thermal continuum of pre-industrial charcoals is not in agreement with the fact that fresh chars produced at low temperature (and thereby containing mainly amorphous aromatic C) decompose completely at ~500°C (Leifeld, 2007). Therefore, we speculate that the thermal resistance of charcoal might reflect the binding energy or reductant capacity of C atoms, driven by both the chemistry and the degree of alteration of charcoal. According to this view, the peak of highest temperature might correspond to the aromatic fraction of BC that has not been yet altered through aging. The negative correlation between the area of the peak at ~500 °C and O content of charcoal supports this assumption, since aging mainly consist in the reactions of exposed C rings with a high density of π electrons and free radicals with oxidizing agents such as O2 (Joseph et al., 2010). The fraction of

129

Chapter 5. Thermal analysis

BC that has survived alteration over centennial timescales is most likely landlocked in aromatic clusters with a high number of rings, which offers a protection against reactions of aging. Therefore, we speculate that DSC of charcoal particles differentiates between (i) C that is directly bonded to O (first peak, less thermally stable); (ii) unaltered aromatic C in clusters of large size (third peak, high thermal resistance) and (iii) aromatic C in clusters of small size that have been altered but that are not directly bonded to O. The process of aging of BC in the environment is still poorly addressed and is of prime importance to unravel the role that BC plays in geochemical cycles. In that sense, dynamic thermal methods have the potential to offer rapid, inexpensive continuous information on the complete BC continuum related to the binding energy of C that can bring precious information on the degree of alteration of BC aged in soil.

Thermal analysis of organo-mineral soils

For the soils of this study, the strong correlation between total heat release measured by DSC and SOC content confirm that SOM governs most of heat fluxes from untreated organo-mineral soils. This supports the view that, for the soils of this study, minerals have a small impact on the thermal signature of untreated bulk soil compared to the combustion of SOM. Consequently, exotherms can be directly related to reactions of combustion of organic components of soil. Moreover, thermal analysis offers a high density of information and a complete view of the continuum of SOM, which is a major advantage compared to chemical or static thermal methods that use cut-off values to operationally define pools of SOM. Thermal analysis of charcoal-rich kiln soils, adjacent reference soils and individual charcoals has highlighted that both charcoal and uncharred SOM are composed of a continuum of materials. These two continuums largely overlap, which stresses the issue of BC quantification in soil: the choice of a cut-off value to discriminate quantitatively between charcoal-C and uncharred SOC is impossible. Nevertheless, thermal analysis has also shown that charcoal and uncharred SOM have a distinct thermal signature that was successfully used to identify BC in soil. Reference soils are dominated by SOM constituents thermally labile that degrade at around 300–330 °C. Spreading of the thermogram towards higher temperatures suggests that more stable compounds are also present in smaller amount. The degradation of SOM in the 300–350 °C range has been related to the decomposition of aliphatic C molecules, such as carbohydrates (Dell’Abate et al., 2002; Kucerík

130

Chapter 5. Thermal analysis et al., 2004) and in the range of 400–450°C to the combustion of aromatic C (Satoh, 1984). Accordingly, Li et al. (2002) reported that thermal stability of lignin is higher than that of cellulose but depends on its structure. Therefore, the pattern of heat release of reference soils suggests that they are composed mainly of aliphatic-C from cellulose- and hemi-cellulose-derived SOM, mixed to a small amount of aromatic-C compounds from lignin-derived SOM that decompose at higher temperature. The dominance of alkyl-C shown by 13C-NMR spectra of reference soil (Figure 5.6) is in agreement with this interpretation of the spectra. Small differences in the shape of the thermograms from cropland and forest reference soils probably result from a difference in the composition of SOM, possibly related to contrasting quality of inputs of organic matter. Reference cropland soils systematically showed a small peak at ~400 °C, which corresponds to the temperature of the main peak attributed to charcoal in kiln soil. By inspection of bulk soils, black particles looking like charcoal were found in each of them. These particles might result, in part at least, from contamination of charcoal from the kiln site. Nevertheless, the same signal was observed for the cultivated soils sampled in fields unaffected by charcoal production. We attribute the presence of fire residues in these soils to burning that generally followed deforestation when land was converted from forest to agricultural land in the past (Hoyois, 1953). Moreover, plant residues left on the field after harvest were commonly burnt until the 1980s (personal communication of Joseph Dufey).

Quantification of charcoal-C by DSC: advantages and limitations.

Base-rich pre-industrial charcoal kiln soils of this study have a thermal signature comparable to that of BC-rich Chernozerm and Vertisol analyzed by Leifeld (2007). This supports the view that aged BC has typical thermal features in given soil conditions. These characteristic features were exploited here to quantify charcoal-C in the soil of pre-industrial kiln sites. For cropland soils, the content of charcoal-C in the topsoil of the kiln site is strongly correlated (r=0.98) to the excess of OC (∆OC) that accumulated in the kiln soil relative to adjacent reference soil (see Chapter 8, Figure 8.4a). The slope of the relationship between the two variables is ~0.80, which indicates that ~80 % of ∆OC is charcoal-C. This is in agreement with the assumption that the increase of OC at kiln sites is mainly related to charcoal enrichment. We also estimated that the content of uncharred SOC at kiln site is about 1.2 times that of adjacent reference soils (see Chapter 8, Figure 8.4b), which is in agreement with estimates made by Kerré et al. (2016). They measured an 131

Chapter 5. Thermal analysis increase of uncharred SOC at kiln site between 1.0 and 1.4 times the amount measured in adjacent reference soils reference soil, with some variability depending on the method of quantification of charcoal-C (they used three different methods of quantification: dichromate oxidation, differential scanning calorimetry and chemo-thermal oxidation). Enrichment of uncharred SOC in presence of BC was also highlighted in other studies (e.g. Liang et al., 2010; Hernandez-Soriano et al., 2015). For BPCA-characterized forest soils from the study of Borchard et al. (2014), the amount of charcoal-C estimated by DSC was strongly linearly related to the concentration of BPCA-C (Figure 5.14). Nonetheless, the content of charcoal-C obtained by DSC overestimated the content of BPCA-C by more than five times. Accordingly, Brodowski et al. (2005) underestimated total C content of charcoal by BPCA analysis by a factor up to 4.5 times. Several aspects can contribute to the strong overestimation of the content of BPCA-C by our method. First, the low recovery of BPCA-C might result from an inaccessibility of the most refractory forms of C in charcoal in the BPCA quantification procedure (Brodowski et al., 2005b). Second, the strong acid treatment might completely destruct the less condensed fraction of chars, composed of non-aromatic C or amorphous aromatic clusters, which explains for the absence of BPCA of less than three carboxyl groups (Glaser et al., 1998). Third, our calibration might be subject to some bias, particularly because of the variability of the shape of soil thermograms according to mineralogy and SOM quality. We were unable to test the effect of the variability of the thermogram of uncharred SOM on the accuracy of the estimation because of the difficulty to find soils completely free of BC. This point should be addressed in the future by adding known amounts of charcoal to variety of soils initially free of charcoal to determine how much the shape of the thermogram of the soil free of charcoal influence the result of the estimation. On the other hand, the shape of the thermogram of pre-industrial charcoals did not seem to have much effect on the estimates. We obtained very accurate estimates of charcoal-C content with our peak index for soil-charcoal mixtures numerically simulated from the thermogram of nine different pre- industrial charcoals with different shapes. Despite the fact that our BC estimates might be subject to some bias, our comparison with BPCA-C illustrates how much classic methods of BC quantification may underestimate the concentration of BC in soil. Our estimates of charcoal-C content supports the view that BPCA-C represents less than one fifth of charcoal-C content in the soil of pre-industrial charcoal

132

Chapter 5. Thermal analysis kiln sites, whereas a multiplicative factor of 2.27 is sometimes used to convert the content of BPCA-C into BC content in soil. This correction factor represents the highest recovery rate obtained in optimal conditions of extraction for one type of BC (Glaser et al., 1998). Our result support the idea that this correction factor of 2.27 is largely underestimated for the soils of Borchard et al. (2014), which questions the accuracy of the BPCA method for the absolute quantification of BC in soil. The fact that DSC offers a complete view of the continuum of SOM is a main advantage to identify, characterize and quantify BC in soil. Nevertheless, the variability of the thermal signature of both BC and mineral soil hinders the generalization of a quantification approach based on height and position of peaks. For the soils of this study, the mineral background had little effect on the shape of exotherms from the combustion of SOM and BC but for other soil types, soil minerals might interfere strongly. In addition to the inversion of quartz-α to quartz-β at 573 °C (visible on most soil thermograms of Figure 5.2 and Figure 5.3), gibbsite, kaolinite and halloysite generate endotherms between 300 and 550 °C (Tan et al., 1986). These minerals, present in relatively small amount in temperate soils of this study, are expected to be present in large amount in some tropical soils (Uehara and Gillman, 1981), and will most probably interfere with the characterization of SOM by DSC.

Nevertheless, the measurement of CO2 emissions from combustion of SOM by evolved gas analysis (EGA) is a solution to get rid of interferences from soil minerals (Peltre et al., 2013). Comparably, the thermal signature of SOM varies to a large extent depending on its composition (Satoh, 1984) and, possibly, association with soil minerals. The type and composition of BC will of course have an impact as well. This has been illustrated by the major differences existing between the thermal signature of fresh and aged charcoals of this study. Leifeld (2007) identified bitumous coal as the most problematic material interfering with thermal identification of BC in soil because it has a maximum peak temperature comparable to that of the charcoals he analyzed. However, the thermal signature of aged charcoals is much more complex than that a fresh charcoal as they have several peaks of lower thermal stability. Consequently, aged charcoals have a thermal signature very similar to that of lignite coal (Leifeld, 2007) or lignin-rich organic matter, such as wood biomass and peat (Plante et al., 2009; Purmalis et al., 2011). The similarity between thermal properties of these materials and that of charcoal might come from a comparable thermal stability of aromatic compounds of lignin and amorphous aromatic clusters in low temperature chars. Interestingly, Purmalis

133

Chapter 5. Thermal analysis et al. (2011) showed that thermal stability of peat increased with age and degree of decomposition that raises with depth, to reach a much greater stability than that of our pre-industrial charcoal after > 8000 years (sampled at > 4 m depth). This result strongly supports the idea that the most thermally stable compounds of peat accumulate preferentially over time by decomposition of the less stable compounds, with thermal stability being related to the binding energy of C bonds, and thereby governing the biological accessibility. An analogy might be drawn with the thermal stability of chars that increase with the degree of crystallinity (Harvey et al., 2012; Leifeld, 2007).

5.5. Conclusion

The main advantage of dynamic thermal analysis to characterize SOM comes from the fact that it provides a complete view of the continuum of organic materials present in soil. DSC analysis of the soil of pre-industrial charcoal kiln sites and adjacent charcoal-unaffected soils has stressed the complexity of BC quantification by highlighting that thermal properties of charcoal and uncharred SOM overlap to a large extent, invalidating the use of cut-off values for an accurate discrimination between aged charcoal and uncharred SOM. Thermal analysis by DSC turned out to be a precious tool to identify and characterize charcoal in the soil of pre-industrial charcoal kiln sites, though. Aged charcoal was shown to have a characteristic thermal signature and overall remains more thermally resistant than uncharred SOM despite the decrease in thermal resistance that occurs by aging in soil. In the particular conditions of this study, the thermal pattern of heat release has been successfully used to quantify charcoal-C in base-rich cropland and forest soils. We found a strong correlation between thermal stability and the content of BPCA-C in soil, which supports the view that thermal resistance of charcoal is related to its fused aromatic ring structure, and overall that thermal stability of SOC is governed by binding energy of chemical bonds. Despite the successful use of DSC to quantify charcoal-C in the soils of this study, our approach based on peak height and position has no general character because of the high variability in the pattern of heat release of BC and SOM depending on composition. Additionally, we showed that the thermal signature of an individual charcoal relies on soil conditions such as the presence of Ca. Therefore, the definition of a procedure to quantify charcoal-C in soil based on thermal analysis can be calibrated successfully only for a certain type of BC material in a certain range of soil conditions. Overall, we encourage soil scientist to add thermal analysis to the panel of analysis used to characterized 134

Chapter 5. Thermal analysis

SOM in soil (Plante et al., 2009) because it provides rapidly and at low cost a high density of information on the stability of SOM (which is related to the binding energy of C) for the complete continuum of organic materials present in soil.

Acknowledgements

We thank Heike Knicker for 13C-NMR-CPMAS analyzes and data treatment. We are grateful to Nils Borchard currently working at the Centre for International Forestry Research (CIFOR) of Indonesia to have provided the BPCA-characterized soils for comparison with thermal analysis, and for his permission to use some of his personal data in this work. Nils is associated to the manuscript in preparation based on results of this chapter, with the hope that this will get published in a near future. We also thank Alain Plante from the University of Pennsylvania for thermogravimetric (TG), differential scanning calorimetry (DSC) and evolved gas analysis (EGA) preliminary tests that he made on a selection of samples. I warmly thank Jens Leifeld and Robin Giger from the Agroscope of Zürich for their indispensable help for the numerous DSC and elemental analyzes that were made. Many thanks the team of the Institute for Sustainability Sciences of the Agroscope of Zürich (Switzerland) for the kind welcome that I received during my two stays at the Agroscope. Particularly, Roman Hüppi and his roommate, Florian Eichenberger, made my stay delighting by offering me bed and board, friendship and visits of Zürich and surroundings. Some great ping-pong games were played at night.

135

Chapter 6. The “Walkley-Black” oxidation

Chapter 6. The resistance of pre-industrial charcoal residues to the “Walkley-Black” oxidation.5

Summary

The “Walkley-Black” oxidation with dichromate is a routine protocol for the estimation of soil organic carbon (SOC) content. Dichromate oxidation is also used to quantify black carbon (BC) in soil and sediments based on the larger resistance of BC to dichromate oxidation than uncharred SOC. The aim of this study was (i) to evaluate whether dichromate-based SOC quantification methods discriminate between uncharred organic matter and charcoal residues in the topsoil of pre-industrial charcoal kiln sites and (ii) to test the influence of aging on the recovery of charcoal by dichromate oxidation. We selected 40 topsoil samples from 20 pre-industrial charcoal kiln sites of Wallonia, Belgium, and adjacent reference soils unaffected by charcoal production. Samples included soils from forest and cropland, with a diversity of soil types and textural classes. The soil of a currently active charcoal kiln site was used as a reference subject to limited aging. SOC content of all samples was estimated by the original Walkley-Black procedure (results were multiplied by the traditional correction factor of 1.32), and by an adapted method that includes external heating of the digestion mixture to make complete the oxidation of uncharred SOC and get rid of a correction factor. Results were compared to the amount of total organic carbon (TOC) content, measured by dry combustion, and to the content of charcoal-C in soil, estimated by differential scanning calorimetry. About 65 % of charcoal-C was recovered by the Walkley-Black procedure from pre-industrial sites, against 23.6 % in the soil of the currently active kiln site, which indicates that the resistance of charcoal to dichromate oxidation decreases with aging. The recovery of charcoal increased to 90 % after boiling of the digestion mixture, highlighting that heat catalyzes the oxidation of charcoal. The substantial oxidation of charcoal by dichromate and the variability of recovery according to the degree of alteration of charcoal and conditions of reaction support the idea that the quantification of BC based on its chemical resistance is challenging and can

5 Hardy, B., Dufey, J. The resistance of pre-industrial charcoal residues aged in forest and cropland soil to the "Walkley-Black" oxidation. Submitted to Geoderma 137

Chapter 6. The “Walkley-Black” oxidation be subject to important biases if calibration is not adapted to the quality of BC (influenced by age, conditions of production, grain size …) in the sample of interest. Because the recovery of BC by the Walkley-Black method is incomplete, the presence of large amounts of BC in soils frequently affected by fires might be a significant cause of underestimation of SOC in regional and global databases.

6.1. Introduction

Wet oxidation of soil organic carbon (SOC) with a known excess of dichromate in a sulfuric acid medium, and determination of SOC content by back titration of residual dichromate with ferrous sulfate, is a routine protocol in soil science, commonly referred to as the “Walkley-Black” oxidation (Walkley and Black, 1934). Originally, this analysis was developed to meet the need for a rapid, inexpensive and simple method to estimate soil organic carbon (SOC) content (Walkley and Black, 1934). The convenience of the method is, however, at the expense of accuracy, because the oxidation of SOC is inherently incomplete, with a percentage of recovery that may vary to some extent depending on the conditions of reaction and the composition of soil organic matter (Walkley, 1947). To overcome incomplete oxidation, the results are generally multiplied by an empirical factor of 1.32. This corresponds to a recovery rate of 76.0±5.6 (mean±s.d.) % of the total SOC, calculated from a set of 20 soils from England (Walkley & Black, 1934). Adapted protocols that include external heating of the digestion mixture were developed to decrease the variability of the recovery by making the oxidation complete, or close to (e.g. Anne, 1945). Nevertheless, none of these protocols has been as successful as the Walkley-Black procedure, perhaps because the heating step diminishes the convenience of the method. The presence of inorganic soil components (e.g. Cl-, FeII) that react either with dichromate or with carbon during digestion are sources of error in the estimation of SOC content (Walkley, 1947) because the determination is based on the back titration of dichromate rather than on direct measurement of CO2 emitted during reaction. In carbonate-rich soils, however, the approach by back titration has the advantage to take only organic C into account and not inorganic C from carbonates. Oxidation stage of SOC may also have an influence on the estimation because calculation of the SOC content assumes that C0 is completely oxidized to CIV (Walkley, 1947). An average oxidation stage of zero for SOC is generally acceptable (Tiessen and Moir, 1993), but it may vary to some extent from one soil organic matter organic compound to another (Masiello et al., 2008). 138

Chapter 6. The “Walkley-Black” oxidation

Dichromate oxidation is also commonly used to quantify black carbon (BC) in soil and sediments because BC is supposed to be more resistant to chemical oxidation than uncharred SOC, related to its fused aromatic ring structure. By definition, BC includes all forms of solid carbonaceous residues from biomass burning or charring and fossil fuel combustion (Schmidt & Noack, 2000), which corresponds to a broad molecular continuum with no clear-cut chemical boundaries (Hammes et al., 2007). Despite its relative resistance, BC is not totally inert to dichromate oxidation, even in its most refractory form (Masiello et al., 2002). Resistance to dichromate oxidation depends on the initial feedstock (Ascough et al., 2011; Bird and Gröcke, 1997; Hammes et al., 2007; Knicker et al., 2007; Skjemstad and Taylor, 2008), the conditions of production (Ascough et al., 2011; Bird and Gröcke, 1997; Naisse et al., 2013) and the size of BC particles (Skjemstad and Taylor, 2008). Many charcoals contain fractions of distinct oxidative resistance (Bird and Gröcke, 1997), which stresses the complexity of BC quantification. Therefore, quantification protocols rely on the difference in the kinetics of degradation between BC and other pools of carbon when soil or sediment is incubated in 2M sulfuric acid with an excess of dichromate. Some authors derive the fraction of BC in the sample from multiple components exponential models fitted on mass or carbon loss over digestion time (Bird and Gröcke, 1997; Hammes et al., 2007; Lim and Cachier, 1996; Masiello et al., 2002; Wolbach and Anders, 1989), whereas others define BC as the amount of carbon remaining after a defined digestion time, which is a conservative, operational definition of BC (Hammes et al., 2007; Knicker et al., 2008; Rumpel et al., 2006; Song et al., 2002). Most protocols include pretreatment of the sample with HCl to remove carbonates and with a mixture of concentrated HF and HCl to liberate carbonaceous material possibly trapped in the sheets of phyllosilicates (Lim and Cachier, 1996). The Walkley-Black oxidation, much easier to implement, was also proposed to quantify charcoal in soil. Based on standard additions of charcoal to soil, Kurth et al., (2006) reported that about 80 % of charcoal was resistant to the Walkley-Black oxidation, which is a relatively effective discrimination between BC and uncharred SOC. In this work, the aim was to assess whether dichromate oxidation by the Walkley-Black procedure, which is a routine protocol used worldwide for the determination of soil organic carbon, discriminates between charcoal and uncharred soil organic matter in the topsoil of pre-industrial charcoal kiln sites, largely enriched in charcoal residues. In these sites, charcoal particles were subject to > 150 years of physical, chemical and biological weathering,

139

Chapter 6. The “Walkley-Black” oxidation which might have decreased their resistance to chemical oxidation with dichromate (Ascough et al., 2011). To test whether the resistance of charcoal to dichromate oxidation changes with aging in soil, we compared the recovery in the soil of pre-industrial charcoal kiln sites to that in the soil of a currently active charcoal kiln site, where charcoal has been subject to limited aging.

6.2. Material and methods

Soil samples

Experiment was conducted on 40 topsoil samples from 20 pre-industrial charcoal kiln sites of Wallonia (Belgium), largely enriched with charcoal residues, and from adjacent reference soils, unaffected by charcoal production. Ten sites were located in forest and 10 sites in cropland. Cropland soils were previously sited under forest and deforested for cultivation in the late 19th century. Soils were sampled between November 2011 and May 2012. According to the WRB 2014 classification (IUSS Working Group WRB, 2014), soil types of reference soils in forest are Dystric Cambisols (5 sites), Eutric or Calcaric Cambisols (3 sites), Albic Luvisol (1 site) and Albic Podzol (1 site) (Table 1). In cropland, reference soils are Haplic Luvisols (8 sites), Eutric Cambisol (1 site) and Colluvic Regosol (1 site) (Table 1). To compare pre-industrial sites with one that had been exposed to limited aging, we also analyzed the topsoil of the site of one currently active charcoal kiln located close to Dole, France, sampled in August 2012 and previously characterized (Hardy et al., 2016). This site is used once or twice a year for charcoal production since the early 1990s. It is located on a silt loam Haplic Luvisol developed on lacustrine and alluvial quaternary deposits (Hardy et al., 2016). All soil samples were air dried and sieved to 2 mm. Before analysis, the < 2 mm fraction was ground to powder with a crusher RS 200 (Retsch) with oscillating rings in tungsten.

Total and oxidizable organic carbon content

Total carbon content was measured by dry combustion with a vario MAX elemental analyzer (Elementar). Inorganic carbon content was measured by the modified-pressure calcimeter method (Sherrod et al., 2002). Total organic carbon (TOC) content was calculated by difference between total and inorganic carbon. The oxidizable SOC content was determined by wet oxidation in an acid dichromate solution according to the original Walkley- Black procedure described by Walkley (1947), with no external source of heat.

140

Chapter 6. The “Walkley-Black” oxidation

Briefly, from 0.1 to 2 g of soil were weighed to contain between 10 and 25 mg of TOC and mixed to 10 ml of a 0.167 M K2Cr2O7 solution in a 500 ml

Erlenmeyer. 20 ml of concentrated H2SO4 were added and the digestion mixture was shaken for one minute. After 30 min, 200 ml of demineralized water was poured, and the excess of dichromate was titrated with 0.25 M

FeSO4 in presence of concentrated H3PO4, BaCl2 and diphenylamine as an indicator. The results were multiplied by the empirical factor of 1.32 to overcome incomplete oxidation of SOC (Walkley & Black, 1934). Wet oxidation was repeated by Anne’s method, an adapted protocol that includes boiling of the digestion mixture for 5 min to make the oxidation of uncharred SOC complete and get rid of the correction factor of 1.32 (Anne, 1945). This author recommends the use of 15 ml of H2SO4 cc. for 10 ml of K2Cr2O7 8 % (0.27 M).

Charcoal-C content

Between 15 and 25 mg of each soil were scanned from room temperature to 600 °C at heating rate of 10 °C min-1 with a differential scanning calorimeter DSC100 (TA Instruments) under a flow of 50 ml min-1 synthetic air (Leifeld, 2007). Soil samples containing more than 60 g kg-1 SOC were previously diluted with Al2O3. Thermograms of soil samples from charcoal kiln sites were interpreted to estimate the proportion of charcoal-C and uncharred SOC in cropland soils based on the relative height of exotherm peaks attributed to the combustion of uncharred organic matter and charcoal (Hardy et al., 2017; Kerré et al., 2016; Leifeld, 2007). Exotherms above 360 °C were attributed to charcoal, according to the signature of a selection of charcoal particles extracted from soil and the signature of reference soils, assumed to have been unaffected by charcoal production (Hardy et al., 2017). The relationship between the fraction of charcoal-C in soil and the relative height of peaks was calibrated thanks to standard additions of charcoal to an organo-mineral soil initially free of charcoal (Hardy et al., 2017). We made two independent calibrations, one for forest and one for cropland soils because charcoal particles aged in forest and cropland have a distinct pattern of heat release. Uncharred SOC content was calculated as the difference between TOC content and uncharred SOC content. Reference soils were assumed to contain a negligible amount of charcoal.

141

Chapter 6. The “Walkley-Black” oxidation

6.3. Results

Forest soils

In forest, TOC content measured by dry combustion ranges from 17.8 to 92.9 g kg-1 in reference soils, whereas in the soil of charcoal kiln sites, it ranges from 38.1 to 187.6 g kg-1 (Table 6.1a). The soil of pre-industrial charcoal kiln sites contains from 33.8 to 173.0 g kg-1 of charcoal-C (Table 6.1), which corresponds to 77.8±12.4 % of TOC content. In reference soils, estimates of SOC content by wet oxidative methods with (Anne’s method) and without (Walkley-Black’s method) boiling accord with TOC content measured by dry combustion (Fig. 1a, b). Consequently, the two oxidative methods are also in agreement with each other. In contrast, the Walkley-Black oxidation underestimates TOC content in the soil of pre- industrial charcoal kiln sites (Fig. 1a). SOC recovery from the Walkley-Black procedure accounts for 72.2±6.4 % of TOC content (P<0.001), despite the strong correlation between the two estimators (r = 0.986). Boiling of the digestion mixture for five minutes (Anne’s method) increases the recovery to 92.5±4.9 % of TOC content (Fig. 1b). However, the underestimation of TOC content remains significant (P=0.003). In agreement with the increase in recovery obtained under boiling, SOC content obtained by Anne’s method overestimates SOC content obtained by the Walkley-Black procedure, by 21.9±6.7 % (Fig. 1c).

142

Figure 6.1. Comparison of the soil organic carbon content measured by the original Walkley-Black procedure (SOC W&B), by Anne’s method (SOC Anne) that includes boiling during digestion and by dry combustion (TOC) for kiln (K) and reference (R) soil samples from forest and cropland. (a) SOC W&B against TOC in forest; (b) SOC Anne against TOC in forest; (c) SOC Anne against SOC W&B in forest; (d) SOC W&B against TOC in cropland; (e) SOC Anne against TOC in cropland; (f) SOC Anne against SOC W&B in cropland.

143

Table 6.1. a) Description of soil samples from charcoal sites (K) and adjacent reference (R) soils for the 10 sites from forest (F). Data includes World Reference Base (WRB) soil class and texture, depth of sampling, total organic carbon (TOC) content, soil organic carbon (SOC) contents estimated by the Walkley–Black procedure (SOCW&B) and by Anne’s method (SOCAnne) and the contents of charcoal-C and uncharred SOC (SOCuncharred).

144

Table 6.2. b) Description of soil samples from charcoal sites (K) and adjacent reference (R) soils for the 10 sites from cropland (C) and for the currently active kiln site (AK).

145

Chapter 6. The “Walkley-Black” oxidation

Cropland soils

In cropland, TOC content measured by dry combustion ranges from 9.0 to 26.4 g kg-1 in reference soils whereas it ranges from 14.3 to 64.4 g kg-1 in the soil of pre-industrial charcoal kiln sites (Table 6.2b). The soil of pre-industrial charcoal kiln sites contains from 1.8 to 33.1 g kg-1 of charcoal-C (Table 6.2b), which corresponds to 36.1±10.6 % of TOC content. Recovery of SOC by dichromate-based quantification methods follows the same trend as in forest soils. In reference soil, SOC content estimated by the two dichromate-based wet oxidative methods agrees with TOC content (Fig 1d, e). In the soil of pre-industrial charcoal kiln sites, SOC content estimated by the Walkley-Black procedure accounts for 81.9±5.2 % of TOC content, which is a significant underestimation (P<0.001). Boiling of the digestion mixture for five minutes by Anne’s method increased the recovery to 91.7±3.4 % of TOC content (Fig. 1b), which remains nonetheless a significant underestimation of TOC content (P=0.002).

Currently active kiln site

The soil of the currently active kiln site contains 83.1 g kg-1 of TOC, including 73.6 g kg-1 of charcoal-C (Table 6.2b). Estimates of SOC content by wet oxidative methods were 26.9 g kg-1 by the Walkley-Black procedure and 74.3 g kg-1 by Anne’s method (Table 6.2b). This corresponds to 32.4 and 89.4 % of TOC content, respectively.

Recovery of charcoal

Regardless of land use, SOC content estimated by the two dichromate-based oxidative methods systematically exceeds the content of uncharred SOC of the soil of pre-industrial charcoal kiln sites (Table 6.2a, b), which indicates that charcoal is partially oxidized by dichromate. Results from reference soils support the view that uncharred SOC is completely recovered by the two dichromate-based SOC quantification procedures. Therefore, the excess of SOC recovered by dichromate oxidation in the soil of pre-industrial charcoal kiln sites (calculated by difference with the content of uncharred SOC) was plotted against the content of charcoal-C to evaluate the fraction of charcoal that was oxidized by dichromate (Fig. 2a, b). The best fitting line of the regression between the excess of SOC and the content of charcoal-C has a slope of 0.65 for the Walkley-Black oxidation (Fig. 2a) and of 0.90 for Anne’s method (Fig. 2b), which points out that about 65 and 90 % of the initial amount

146

Chapter 6. The “Walkley-Black” oxidation of charcoal-C was oxidized by dichromate, respectively. For the currently active kiln site, we obtain by the same approach that 23.6 % of charcoal-C was recovered by the Walkley-Black procedure. The recovery increases to 88.0 % by Anne’s method, after boiling of the digestion mixture for five minutes.

Figure 6.2 a) difference between soil organic carbon (SOC) content estimated by the Walkley-Black procedure and uncharred SOC content (SOCW&B - SOCUncharred) against the content of charcoal-C in the soil of pre-industrial charcoal kiln sites; b) Difference between soil organic carbon (SOC) content estimated by Anne’s method and uncharred SOC content (SOCAnne - SOCUncharred) against the content of charcoal- C in the soil of pre-industrial charcoal kiln sites. Circles are for forest soils and triangles are for cropland soils.

6.4. Discussion

Recovery of uncharred SOC

For both forest and cropland soils, the correction factor of 1.32 inherent to the Walkley-Black procedure to overcome incomplete oxidation of SOC was adequate for reference soils (containing a negligible fraction of charcoal) despite the diversity of soil types, textural classes and land uses. This result is in disagreement with that of Lettens et al., (2007), who found that the correction factor of 1.32 was systematically underestimated for 475 silt loam and sandy soils from forest, grassland and cropland from four sites in Belgium. They concluded that the correction factor of 1.32 was inappropriate and incriminated variables such as soil type, texture, land use or even climate to explain the variability of the recovery of SOC by the Walkley-Black oxidation. The samples of this study also come from Belgium and include soils with textural classes, sampling depths and land uses similar to that of Lettens et al. (2007). Therefore, the fact that the correction factor of 1.32 was adapted

147

Chapter 6. The “Walkley-Black” oxidation for the samples of this study supports the idea that analytical considerations of the Walkley-Black procedure might explain most of the variability of the recovery measured by Lettens et al. (2007), with a secondary importance of soil conditions. For instance, temperature of digestion is critical for the kinetics of oxidation of SOC. Walkley (1947) showed that the recovery of SOC was maximal for a ratio H2SO4:H2O of 2:1, because it maximizes the raise of temperature of the digestion mixture caused by the exothermal dissolution of H2SO4 when it is poured in the aqueous solution of dichromate. Consequently, the size of the vessel chosen for the digestion is of importance, as large vessels will promote heat loss and attenuate the raise of temperature (Walkley, 1947). Evidently, the concentration and volume of reagents, particularly that of H2SO4, is of prime importance for the recovery as they influence the oxidative strength of the system (Walkley, 1947). The amount of SOC subject to digestion is also critical for the recovery of SOC and must be strictly between 10 and 25 mg (Walkley, 1947). We observed during preliminary tests that SOC was systematically underestimated when the sample contained more than 25 mg of SOC. This point is particularly thorny, as the person that analyzes the soil sample has generally no a priori knowledge of the SOC content since the analysis aims to determine it. Therefore, an overweight of soil might be a frequent source of underestimation of SOC by the Walkley-Black procedure, particularly for soils that contain a large amount of SOC, like forest topsoils. Therefore, it is indispensable to strictly stick to the original procedure of Walkley (1947) for an adequate use of the empirical factor of 1.32. For large databases, the validity of the factor should be tested (for example, by comparison with dry combustion) for a representative selection of samples and corrected if necessary. Nevertheless, we highly recommend to rely on dry combustion when an accurate quantification of SOC is required.

Recovery of charcoal-C

The underestimation of TOC content by the two wet oxidative methods in the soil of pre-industrial charcoal kiln sites is attributable to the presence of charcoal residues, more resistant to dichromate oxidation than uncharred organic matter (Bird and Gröcke, 1997; Hammes et al., 2007; Kurth et al., 2006; Lim and Cachier, 1996; Walkley, 1947; Wolbach and Anders, 1989). Nevertheless, we estimated that the Walkley-Black method recovers about 65 % of charcoal-C (Figure 2a), which is a substantial oxidation of charcoal. The recovery increased to 90 % with the adapted procedure of Anne (1945) that

148

Chapter 6. The “Walkley-Black” oxidation involves an external source of heat (Figure 2b), which provides evidence that heat catalyzes the oxidation of charcoal. Our results indicate that the two dichromate-based methods of SOC quantification discriminate poorly between uncharred soil organic matter and pre-industrial charcoal, since a main fraction of charcoal is oxidized. Dichromate oxidation, however, has been proved to be useful for quantifying BC in soil and sediments (Bird and Gröcke, 1997; Lim and Cachier, 1996; Masiello et al., 2002). The main differences between BC-quantification protocols (Bird and Gröcke, 1997; Lim and Cachier, 1996; Masiello et al., 2002) compared to the Walkley-Black procedure are (i) the lower activity of protons (H2SO4 2M is generally used) and (ii) the smaller temperature of incubation (generally < 80 °C), which is by far below the temperature of 110-120 °C reached by dissolution of 20 ml of

H2SO4 in 10 ml of aqueous solution of dichromate in the Walkley-Black procedure (Walkley, 1947). As a result, the oxidative power of the system is smaller in these BC-quantification protocols, which allows discriminating better between uncharred SOC and BC based on a difference in the kinetics of oxidation. For the topsoil of the currently active charcoal kiln site, we estimated that only 23.6 % of charcoal-C was recovered by the Walkley-Black procedure. In comparison, Kurth et al. (2006) obtained a recovery rate of 20 % for standard additions of fresh charcoal to a mineral soil, and Walkley (1947) reported a recovery of 11 % for a pure wood charcoal. These low recoveries contrast with the recovery of about 65 % for pre-industrial charcoals, which suggests that aging in soil decreases the resistance of charcoal to dichromate oxidation, possibly by a decrease of hydrophobicity (Criscuoli et al., 2014; Knicker et al., 2008) and chemical stability related to oxidation of particles over time. It accords with the results of Ascough et al. (2011), who showed that charcoals subject to prolonged environmental exposure are less resistant to dichromate oxidation than unaltered charcoals. Charcoal aging mainly consists in abiotic (Cheng et al., 2006) or biotic oxidation (Baldock and Smernik, 2002; Hamer et al., 2004; Wengel et al., 2006) that starts at the surface of particles (Lehmann et al., 2005). It creates a high density of oxygenated functions, mainly carboxylic and phenolic groups (Lehmann et al., 2005). This polarization of the surface of charcoal decreases dramatically its hydrophobicity (Criscuoli et al., 2014; Knicker, 2011a) and might promote further physical, chemical and microbial weathering (Hammes and Schmidt, 2009). Therefore, the relatively small oxidative resistance of charcoal aged in soil for > 150 years is most likely related to physical, chemical and biological

149

Chapter 6. The “Walkley-Black” oxidation alteration of charcoal through aging (Ascough et al., 2011). The fact that a small fraction of charcoal remains after dichromate oxidation under boiling accords with the presence of a fraction of graphite-like, condensed polycyclic aromatic clusters that is more resistant to oxidation than the main fraction of low temperature chars made of poorly condensed aromatic C (Bird et al., 2015). Therefore, the Walkley-Black oxidation might discriminate better between uncharred SOC and BC compounds that have a higher degree of aromatic condensation, like soots.

Implications for global SOC budgets

Nowadays, robust dry combustion is more and more used for SOC quantification at the expense of the Walkley-Black procedure (rightly, given the number of uncertainties underlying the estimation of SOC by dichromate oxidation). Nevertheless, the latter method remains in use worldwide, for two main reasons: (i) It is a rapid, inexpensive analysis that requires little equipment and provides an estimation of SOC content even in presence of carbonates; (ii) Estimates of SOC content of many regional and national soil databases were determined, partly to completely, by the Walkley-Black procedure. These databases are indispensable to estimate SOC stocks at regional (e.g. Tarnocai et al., 2009; Viscarra Rossel et al., 2014; Zimov et al., 2006) or global scale (e.g. Batjes, 2016, 2014), or to follow changes of SOC stocks over time (e.g. Lettens et al., 2005; Meersmans et al., 2009; Minasny et al., 2011). The variability of the recovery of SOC content by the method of Walkley-Black and the resulting inaccuracy of SOC estimates is cited as a common issue in large scale studies of SOC stocks (Batjes, 2014). The incomplete recovery of BC, which can contribute to an important fraction of SOC in soils frequently affected by fire (Reisser et al., 2016), is possibly a common source of underestimation of SOC estimated by the Walkley-Black procedure. Based on a recent inventory of literature, Reisser et al. (2016) estimated that BC accounts for 13.7 % of SOC content on average and up to 60 %, which corresponds to about 200 Pg of BC in the uppermost two meters of soil at global scale (Reisser et al., 2016). Since only about 65 % of charcoal- C of the soil of preindustrial charcoal kiln sites was recovered by the Walkley- Black procedure, our results support the idea that the presence of BC in soil might be an important source of underestimation of SOC content by the Walkley-Black procedure in regional, national and global databases.

150

Chapter 6. The “Walkley-Black” oxidation

6.5. Conclusion

Despite the relative resistance of charcoal-C to dichromate oxidation, we showed that about 65 % of charcoal-C of the soil of pre-industrial charcoal kiln sites was recovered by the Walkley-Black procedure. This recovery rate is much higher than for fresh charcoal, which indicates that aging decreases the resistance of charcoal to dichromate oxidation. Recovery of charcoal increased to about 90 % by boiling the digestion mixture for five minutes, which highlights that heat controls the kinetics of chemical oxidation of charcoal. The substantial oxidation of charcoal by dichromate and the variability of recovery according to the degree of alteration of charcoal and conditions of reaction support the idea that the quantification of BC based on its chemical resistance is challenging and can be subject to important biases if calibration is not adapted to the quality of BC (influenced by age, conditions of production, grain size …) in the sample of interest. Because the recovery of charcoal by dichromate oxidation is incomplete, the presence of important amounts of BC in soils frequently affected by fires might be a significant source of underestimation of SOC stocks in regional and global databases that relied on the Walkley-Black procedure for the estimation of SOC.

Acknowledgements

We are grateful to Claudine Givron and Anne Iserentant for their participation in laboratory work. Funds were provided by the General Directorate for Agriculture, Natural Resources and Environment – Public Service of Wallonia and the Fonds Spéciaux de Recherche (FSR) - Université catholique de Louvain.

151

Chapter 7. Forest soil

Chapter 7. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium.6

Summary

In Wallonia, Belgium, intensive in situ charcoal production that was closely linked to pre-industrial smelting and steel-making affected a large part of the forested area in the late 18th century. Charcoal kiln relics can be detected under forest as domes of about 10 m in diameter, with the topsoil greatly enriched with charcoal residues. We sampled 19 charcoal kiln sites and the adjacent reference soil by soil horizon on four different soil types (Arenosols, Luvisols, Cambisols and Podzols). Data were analyzed with linear mixed models to assess the effect of the charcoal kiln site on soil properties in relation to depth and soil conditions. We also addressed the evolution of soil properties over time by a comparison of the soil characteristics at a currently active kiln site. The charcoal-rich topsoil has a larger C:N ratio and cation exchange capacity (CEC) per unit of organic carbon than the reference soil. The largest CECs per unit of carbon were observed on soil with coarser textures. On acidic soil, the increase in base saturation in the subsoil reflects the past liming effect of ash produced by wood charring, whereas the topsoil is re-acidified. The acidity of carbonate-rich Cambisols, however, is not reduced. Regardless of soil type, the kiln topsoil is greatly depleted in exchangeable K+ and available P, which may be attributed (i) to the small affinity of the exchange complex of charcoal for K+ and a decrease in P availability with time, respectively; (ii) to the fact that aged charcoal does not take part in the biological cycling of nutrients because of its low degradability. This highlights that charcoal does not support the same ecological functions as uncharred organic matter. Therefore, we recommend to better evaluating long-term implications of soil amendment with biochar before generalizing its large scale application to cropland soils.

6 Adapted from: Hardy, B., Cornelis, J.-T., Houben, D., Lambert, R. & Dufey, J.E. 2016. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium. European Journal of Soil Science, 67, 206–216. 153

Chapter 7. Forest soil

7.1. Introduction

Black carbon (BC), the solid residue of incomplete combustion of biomass, is present in most soil after natural or human-induced fires (Schmidt & Noack, 2000). Black carbon fulfils many environmental functions in soil, such as long-term carbon storage (Schmidt & Noack, 2000; Knicker, 2011) and enhancement of fertility in anthropogenic dark earth (Glaser & Birk, 2012; Wiedner et al., 2015). Therefore, artificial biochars have been investigated increasingly as potential soil amendments over the last few decades (Lehmann, 2007). Among human activities responsible for the introduction of BC in soil, the contribution of in situ charcoal production might have been underestimated. In 2009, the FAO reported a global annual charcoal production of 47 million m³. Africa accounts for 63 % of total production, Latin America and the Caribbean for 18.7 % and Asia for 15.7 % (Steierer, 2011). In developing countries, charcoal is mainly produced in situ with traditional earth mound kilns (Schenkel et al., 1998; Steierer, 2011) similar to those that were used for traditional production of charcoal in Wallonia and traditional charcoal kilns. The very black topsoil of kiln sites consists of thermally altered (organo-) mineral soil and vegetation remains mixed with charcoal residues that were left on the site after pyrolysis. Enrichment of the topsoil by charcoal, site preparation and thermal action of wood pyrolysis are major forms of disturbance that affected the soil at a kiln site. Although these relics of charcoal production are ubiquitous in the former forested areas of Wallonia (Hardy and Dufey, 2012b), there appear to have been no reports of their effect on soil properties. There are few studies on the effect of abandoned charcoal kilns on soil properties in the international literature. They generally concur that, shortly after charring, charcoal- enriched topsoil has a pH close to neutral, which is generally higher than the pH of adjacent charcoal-unaffected soil (Gómez-Luna et al., 2009; Nigussie and Kissi, 2011; Ogundele et al., 2011; Oguntunde et al., 2004). The increase in pH goes together with larger concentrations of exchangeable Ca2+, K+, Mg2+ and available P. These effects tend to attenuate with time, but some remain perceptible after one or two centuries. On initially acidic temperate soil, pH values of the aged charcoal-enriched soil become more acidic but generally remain above the pH values of the reference soil (Borchard et al., 2014; Criscuoli et al., 2014; Mikan and Abrams, 1995; Young et al., 1996), together with larger contents of exchangeable Ca. The contents of exchangeable K+ and available P clearly decrease with time, but differences between the published studies suggest that persistence of the initial enrichment might be site specific.

154

Chapter 7. Forest soil

The cation exchange capacity (CEC) of the charcoal-enriched soil, however, seems to increase with time. Larger potential and effective CEC were measured more than one century after kiln abandonment (Heitkötter and Marschner, 2015; Mikan and Abrams, 1995; Young et al., 1996), whereas soon after charring the CEC does not differ from that of the reference soil (Nigussie and Kissi, 2011; Ogundele et al., 2011; Oguntunde et al., 2004). In this chapter, our main objective was to characterize the effect of charcoal kilns abandoned more than 150 years ago (Evrard, 1956) on the properties of a range of forest soils in Wallonia. Earlier research examined the soil properties of charcoal kiln sites of various ages at a few or single sites, and the results appear to relate to site-specific soil and climate conditions, and are therefore difficult to generalize (Criscuoli et al., 2014). We also aimed to distinguish potential effects of soil type on soil properties at the charcoal kiln site. This had been investigated before only by Borchard et al. (2014), who measured the effects of abandoned charcoal kiln sites on two distinct soil parent materials (acidic and calcareous). Furthermore, we aimed to trace the evolution of the soil properties after charring by comparing abandoned charcoal kiln sites in Wallonia with a currently active kiln site in a comparable environment. We examined soil properties down the soil profile because the decrease in topsoil nutrients over time might result from leaching to the subsoil. Two studies only seem to have measured properties below the charcoal-enriched topsoil of abandoned kilns in the USA (Mikan and Abrams, 1995; Young et al., 1996). Neither of these focused on soil properties, but primarily on the effect of the kiln on the age, structure and composition of the forest.

7.2. Materials and methods

Soil samples

In this chapter, we analyzed soil samples from the set of 19 forest sites on four different soil types (Arenosols, Cambisols, Luvisols and Podzols), sampled by soil horizons (Chapter 4, Table 4.1). To document the change of soil properties > 150 years after the last production of charcoal, soil properties at pre- industrial charcoal kiln sites were compared to that of the currently active charcoal kiln site from Dole (France).

155

Chapter 7. Forest soil

Chemical analyses

All samples were dried at 40 °C, gently ground and sieved to 2 mm. The < 2 mm fraction was then analyzed. The pH was measured in water (pHH2O) and in a 1M KCl solution (pHKCl) with a 1:5 soil:solution ratio. Soil texture of the natural topsoil was measured by Robinson’s pipette sedimentation method (Day, 1965) after the destruction of carbonates with 2.5 M HCl and organic matter with 30 % H2O2 (AFNOR X31–107). Total carbon (C) and nitrogen (N) contents were analyzed by flash combustion with a Vario MAX (Elementar) elemental analyzer. For calcareous Cambisols, the inorganic carbon content was measured by the modified-pressure calcimeter method (Sherrod et al., 2002) on finely ground subsamples (< 200 µm). For the other soil samples, the inorganic carbon content was considered negligible because their pH was acidic to very acidic and their parent material initially contained little or no inorganic C. The organic carbon content (OC) was calculated as the difference between total and inorganic carbon contents. Potential cation exchange capacity (CEC) was measured by percolating soil columns with 1M

CH3COONH4 (ammonium acetate), a solution buffered at pH 7 (Metson, + 1956). Ammonium (NH4 ) was desorbed with a 1.33 M KCl solution and measured by colorimetry (ISO7150/1). Extractable calcium (Ca2+), magnesium (Mg2+), potassium (K+), sodium (Na+) and available phosphorus (PAv) were extracted with an ammonium acetate 0.5 M EDTA 0.02 M solution at pH 4.65 with a 1:5 soil:solution ratio (Lakanen & Erviö, 1971) and measured by spectrophotometry. In samples free from inorganic carbon, agreement between the concentrations of extracted Ca2+, Mg2+, K+ and Na+ and the exchangeable cations (Cottenie et al., 1975) obtained by the 1M ammonium acetate leaching procedure was verified for a selection of soil samples (Figure 7.1). In the samples containing inorganic C, exchangeable cations were extracted by percolation of 1M CH3COONH4 on soil columns (Metson, 1956) and measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES). Base saturation (BS) was calculated as the ratio between the sum of exchangeable Ca2+, Mg2+, K+ and Na+ and the CEC. The percentage of dry matter (105 °C) was determined for each sample to express the result per mass of oven-dry soil.

156

Chapter 7. Forest soil

Figure 7.1 Comparison of the amount of Ca, Mg and K extracted from soil samples by 1M ammonium acetate at pH 7 (Metson, 1956) and by ammonium acetate 0.5M – EDTA 0.02M at pH 4.65 (Lakanen & Erviö, 1971). The points located on the dotted line (1:1 line) correspond to identical values for both methods: (a) 23 soil samples with contrasting characteristics (organo−mineral and mineral soil, charcoal-enriched or charcoal-free, coming from forest and cropland), but containing negligible amounts of inorganic C and (b) Sixteen forest soil samples containing up to large amounts of inorganic C. For the sites from Luvisols (LV; Chapter 4, Table 4.1), we measured the concentrations of total, organic and inorganic P and the concentration of water extractable organic carbon (WEOC) of the topsoil of kiln and adjacent reference soil. We determined organic P by the difference between the P content extracted with 0.5M H2SO4 before and after calcination of the soil at 550 °C (Van Ranst et al., 1999). To measure the total concentrations of P, subsamples of soil were ground mechanically to a powder and were digested with a mixture of three concentrated acids (HF, HNO3 and HClO4), dissolved in aqua regia and analyzed by ICP-AES (Lambrechts et al., 2011). Inorganic P was calculated by difference between total and organic P. All results are expressed per mass of dry soil (105 °C). The content of WEOC was measured on 1g of soil in 25 ml of deionized water. The mixture was stirred for two hours at room temperature (20±1 °C), centrifuged for 20 min at 3000 g and filtered at 0.45 µm. The WEOC content was measured with a liquid phase 157

Chapter 7. Forest soil

Total Organic Carbon Analyzer (Shimadzu) after acidification of the extract with HCl to remove the inorganic C.

Data analysis

The dataset is given in Appendix 1. For statistical analysis, the data were transformed to common logarithms, log10, to reduce the skewness of the distributions and make them close to normal. It improved normality and homogeneity of the variance of the residuals of the models that were applied to test the effect of the charcoal kiln on soil properties. Data were fitted with linear mixed models in R 3.2.2 (R Core Team, 2012) with the package ‘lme4’ (Bates et al., 2014). Charcoal kiln site, depth and the interaction ‘charcoal kiln × depth’ were introduced as fixed effects. We considered two levels of soil depth only, topsoil (organo-mineral) and subsoil samples. The site was introduced as a random effect, including a random intercept and random slopes for the effect of the charcoal kiln site. Normality and homogeneity of the variance of the residuals of the models was verified visually by plotting the residuals against the fitted values. For the fixed effects, P-values were obtained by the analysis of variance (ANOVA) with a Satterthwaite approximation of the denominator degree of freedom of the F-test with the package ‘lmerTest’ (Kuznetsova et al., 2015), based on SAS proc mixed theory. For the random effect, P-values were obtained with a likelihood test ratio, which tests whether the effect of the charcoal kiln site on soil properties depends on the site or not.

7.3. Results

Currently active charcoal kiln site

In the topsoil of the currently active charcoal kiln, the OC content and, to a larger extent, the C:N ratio are greater than in the reference soil (Table 7.1). In contrast, the CEC is smaller. The reference soil is acidic at all depths, with a desaturated exchange complex, whereas the pH is considerably higher in the topsoil of the kiln, with pH values close to neutral and a base saturation of 100 % in the soil mixed with charcoal residues used to cover the mound during pyrolysis. It corresponds to large concentrations of exchangeable Ca2+, Mg2+, K+ and Na+. In addition, available P concentration is more than four times greater than in the reference topsoil. In the subsoil, the differences between the kiln site and reference soil disappear (OC, C:N ratio, CEC and available 2+ 2+ + P) or attenuate with depth (pHH2O, pHKCl, BS, Ca , Mg and K ).

158

Chapter 7. Forest soil

Pre-industrial charcoal kiln sites of Wallonia

The statistical analysis of the fixed and random effects of the linear mixed models is summarized in Table 7.2. According to the models, depth explains a large part of the variance in the data regardless of the variable. This is reasonable because the properties have vertical gradients in the soil profile resulting from pedogenesis. Here, we are more interested in properties affected either by the charcoal kiln site or by the interaction ‘charcoal kiln × depth’, or by both (Table 7.2). The charcoal kiln site has a significant effect 2+ 2+ + on pHKCl, pHH2O, and exchangeable Ca , Mg and Na with no significant relation with depth. The values of each of these properties at the charcoal kiln site are larger than at the reference soil, regardless of soil depth (Figure 7.2).

The pHKCl and the pHH2O both remain (very) acidic on average in the kiln soil, but they are from 0.2 to 0.4 units higher than in the reference soil. 2+ -1 Exchangeable Ca concentration increases from 0.93 to 2.48 cmolc.kg in the -1 topsoil of the kiln and from 0.52 to 1.38 cmolc.kg in the subsoil. The concentration in exchangeable Mg2+ also increases but to a lesser extent, from -1 - 0.27 to 0.32 cmolc.kg in the topsoil of the kiln and from 0.17 to 0.29 cmolc.kg 1 in the subsoil. In spite of the significant increase, the concentration in + -1 exchangeable Na is very small, less than 0.06 cmolc.kg regardless of soil depth.

159

Table 7.1. Evolution of soil properties with depth for the active charcoal kiln site (K) in Dole (France) and the adjacent reference soil (R).

2+ 2+ + + OC/ CEC/ BS/ Ca / Mg / K / Na / PAv/

Horizon Depth/ cm pHH2O pHKCl C:N ratio -1 -1 -1 -1 -1 -1 % cmolc kg % cmolc kg cmolc kg cmolc kg cmolc kg mg.kg 7 A1 < 0 7.1 6.6 8.30 26.0 8.7 100 11.56 1.04 1.44 0.10 84.7 A2 0–25 6.0 5.0 4.11 29.8 9.6 60.0 4.69 0.77 0.90 0.04 17.9 K E1 25–36 5.6 4.4 1.49 16.8 6.9 39.2 1.96 0.52 0.77 0.04 9.5 E2 36–71 4.9 3.9 0.53 11.1 6.8 14.4 0.56 0.28 0.35 0.05 3.3 Bt 71–88 4.8 3.7 0.26 7.1 8.5 11.1 0.33 0.43 0.16 0.05 1.8 Ah 0–8 4.8 3.8 4.29 15.2 13.3 13.7 1.35 0.34 0.45 0.04 17.9 R E1 8–30 4.7 3.9 1.62 14.6 6.6 7.4 0.27 0.13 0.15 0.04 9.1 E2 30–73 4.6 3.8 0.68 12.0 7.8 5.5 0.19 0.15 0.09 0.04 3.9 Bt 73–93 4.7 3.7 0.25 6.8 9.8 7.8 0.25 0.35 0.10 0.05 2.6

7 Mound covering material 160

Figure 7.2 Adjusted means for the topsoil and subsoil of the charcoal kiln site (K) and the reference soils (R). Means and standard errors (error bars) were calculated on log10-transformed data and back-transformed.

161

Table 7.2. Statistical analysis of the fixed (charcoal kiln, depth, charcoal kiln × depth) and random (site) effects of the linear mixed models. Fixed effects were tested with ANOVA by a Satterthwaite approximation of the denominator degree of freedom of the F-test. The random effect was tested with a likelihood test ratio. Significant P-values (P<0.05) are underlined.

pHH2O pHKCl OC C:N ratio Charcoal kiln F(1, 20.1)=9.6; P=0.0056 F(1, 21.4)=7.57; P=0.012 F(1, 92.6)=1.79; P=0.18 F(1, 170.2)=44.8; P<0.0001 Depth F(1, 147.7)=76.2; P<0.0001 F(1, 148.9)=88.3; P<0.0001 F(1, 168.4)=453.8; P<0.0001 F(1, 173.7)=91.5; P<0.0001 Charcoal kiln × depth F(1, 149.5)=1.7; P=0.20 F(1, 151.5)=0.945; P=0.33 F(1, 166.9)=8.04; P=0.005 F(1, 169.8)=9.3; P=0.0026 Site χ²(2)=35.9; P<0.0001 χ²(2)=16.1; P=0.0003 χ²(2)=1.07; P=0.6 χ²(2)=0.00105; P=1

CEC Available P Base saturation Exchangeable Ca2+ Charcoal kiln F(1, 117.8)=18.3; P=3.9e-05 F(1, 25.4)=59.1; P<0.0001 F(1, 20.9)=8.9; P=0.007 F(1, 21.5)=29.2; P<0.0001 Depth F(1, 167.7)=214.6; P<0.0001 F(1, 154.2)=99.8; P<0.0001 F(1, 148.1)=41.7; P<0.0001 F(1, 148.4)=29.2; P<0.0001 Charcoal kiln × depth F(1, 167.3)= 38.3; P<0.0001 F(1, 159.2)=17.8; P<0.0001 F(1, 150.1)=30.2; P<0.0001 F(1, 150.9)=0.0003; P=0.98 Site χ²(2)=0.921; P=0.6 χ²(2)=0.117; P=0.9 χ²(2)=19; P<0.0001 χ²(2)=15.2; P<0.0001

Exchangeable Mg2+ Exchangeable K+ Exchangeable Na+ Charcoal kiln F(1, 20.8)=4.3; P=0.05 F(1, 25.7)=15.1; P=0.0006 F(1, 21.5)=4.54; P=0.04 Depth F(1, 147.8)=6.9; P=0.01 F(1, 147.8)=59.8; P<0.0001 F(1, 146.8)=59.8; P=0.0004 Charcoal kiln × depth F(1, 150.4)=2.9; P=0.09 F(1, 150.8)=5.5; P=0.02 F(1, 148.4)=5.5; P=0.19 Site χ²(2)=14.9; P<0.0001 χ²(2)=3.2; P=0.2 χ²(2)=33.5; P<0.0001

162

Chapter 7. Forest soil

For OC content, the interaction ‘charcoal kiln × soil depth’ only is significant (Table 7.2). It highlights a contrasting effect of the charcoal kiln depending on depth: in the topsoil, the OC content is larger than in the reference soil (6.17 % against 3.99 %) and in the subsoil it is slightly smaller than in the reference soil (0.50 against 0.58 %) (Figure 7.2). Overall, the effect of the charcoal kiln site on OC content is not significant because its effects in the top- and sub-soil counteract each other. Note that our model disregards the thickness of the soil horizons. The differences in OC content might not be interpreted in terms of OC stocks because the topsoil of the kiln is, on average, much thicker than the reference topsoil. Therefore, the thickness of the topsoil of the charcoal kiln sites was analyzed jointly with OC content, regardless of their reference soils. Both properties vary greatly, which relates to soil texture to some extent (Figure 7.3). The thickness of the topsoil tends to decrease with increasing clay content (r = –0.70), whereas the correlation between OC and clay is positive (r = 0.72).

Figure 7.3. (a) Thickness and (b) organic carbon (OC) content of the topsoil of the charcoal kiln sites of Wallonia plotted against the clay content on Arenosols (AR), acidic Cambisols (CMA), calcareous Cambisols (CMC), Luvisols (LV) and Podzols (PZ).

163

Chapter 7. Forest soil

The C:N ratio, CEC, available P concentration, base saturation and exchangeable K+ are affected significantly by both the charcoal kiln site and ‘charcoal kiln × soil depth’ (Table 7.2). The effect of the kiln is significantly different in the topsoil from the subsoil, but the average effect is still significant. In the topsoil, the C:N ratio (27.6 against 14.5) and the CEC (26.0 -1 against 11.9 cmolc.kg ) are much larger for the kiln than reference soil, but these differences disappear in the subsoil (Figure 7.2). Conversely, available -1 + - P (8.9 against 28.8 mg.kg ) and exchangeable K (0.09 against 0.16 cmolc.kg 1) concentrations are much smaller in the topsoil of the kiln than reference topsoil, but the differences attenuate in the subsoil. Base saturation of the topsoil is similar (around 12.5 %) for the charcoal kiln and reference soil, but base saturation of the kiln subsoil increases to 34.7 % on average, whereas it does not increase in the reference subsoil.

Figure 7.4. Boxplots of pH in water, pH in KCl and base saturation (BS) of top- and sub-soil of the charcoal kiln sites of Wallonia (K) and of the reference soils (R). Calcareous Cambisols (CMC) are shown separately from the other soil profiles (acidic soil) to illustrate their antagonist effect on soil acidity.

164

Chapter 7. Forest soil

The effect of the charcoal kiln is site-dependent for the six properties related to soil acidity and exchangeable cations: pHKCl, pHH2O, base saturation and the concentrations of exchangeable Ca2+, Mg2+ and Na+ (Table 7.2). On these variables, the charcoal kiln has no effect or even an antagonist effect in calcareous Cambisols compared to other soil types (Figure 7.4). No other clear effects of soil type were detected. Other non-systematic variation might result from differences in pedogenic factors such as parent material, climate and vegetation or factors related to charcoal production such as age of the site, type of wood that was pyrolysed or practices of the colliers.

7.4. Discussion

Organic carbon content and quality

Charcoal production greatly disturbed the organic matter content and quality at kiln sites because of site preparation, wood pyrolysis and accumulation of charcoal residues. Borchard et al. (2014) showed that the topsoil of charcoal kiln sites in Germany was enriched in BC more than 60 years after charcoal production. In our study, the large C:N ratio in the topsoil of the charcoal kiln sites expresses a difference in soil organic matter quality that is related to charcoal enrichment. It accords with the large C:N ratio of most charred materials, the chemical composition of which is closely linked to that of the parent feedstock. In the topsoil, charcoal residues were mixed with natural organic matter (Borchard et al., 2014) and probably with partly to completely charred soil organic matter and plant material that were part of the mound cover and subject to heating during pyrolysis. The positive correlation between OC and clay content in the topsoil of the kiln (Figure 7.3b) suggests that BC behaves somehow like uncharred SOM and that organo-mineral associations might play an important role in the stabilization BC, although the survival of BC in soil is generally incriminated mainly to inherent chemical recalcitrance. Nguyen et al. (2008) showed that the contents of Si, Al and Fe increased considerably on the surface of BC particles during the first 30 years after deposition, jointly with O-rich functional groups. It suggests that the increase in polarity of the surface of charcoal through oxygenation during aging (Lehmann et al., 2005) favors organo-mineral associations. In contrast, the thickness of the charcoal-rich topsoil is negatively correlated to the clay content (Figure 7.3a). The rarity of charcoal kiln sites on soil rich in clay confirms that these were unsuitable for efficient wood pyrolysis

165

Chapter 7. Forest soil

(Lepoivre and Septembre, 1941). Consequently, the infrequent use of sites on the fine-textured soil of this study, mainly Cambisols, might explain their relatively thin topsoil. It is also likely that deep, unconsolidated substrates with small stone and clay contents, such as sand, provided abundant material for covering the mound in contrast to shallow soil with large stone or clay contents. It accords with the thick topsoil and small OC concentration of the kiln sites on the sandy Arenosols and Podzols, possibly because of the dilution of charcoal residues into a larger soil volume. Sandy textured soil might also favor the vertical migration of charcoal with the soil solution (Hilscher and Knicker, 2011) because of its larger macroporosity than that of finer textured soil, which might also explain the loose, relatively OC-poor topsoil of the sandy kiln soil.

Cation exchange capacity

The large CEC in the topsoil of the charcoal kilns is strongly correlated to the OC content (Figure 7.5). The CEC was plotted against OC after grouping the sites by soil type, which also segregates them effectively by texture except for the Cambisols that range from silty clay loam to sandy loam. We grouped the sandy loam Cambisols with Podzols and Arenosols because all have a coarse texture. For Luvisols, samples from the argic horizon were excluded because their clay content was larger than that of the overlying soil horizons, which results in an increase in CEC that is not linked to OC content. For each soil group, regression lines were drawn to predict the CEC per unit of OC for the kiln and the reference soil. The slope of the regression line is steeper than that of the reference soil regardless of soil group (Figure 7.5), which indicates that aged charcoal is an important contributor to CEC, with a negative charge per unit of OC greater than that of natural soil organic matter.

166

Chapter 7. Forest soil

Figure 7.5. Relations between cation exchange capacity (CEC) and organic carbon (OC) for kiln (K, filled symbols) and reference (R, open symbols) soil samples by soil texture groups: (a) Arenosols (AR), Podzols (PZ) and sandy loam acidic Cambisols (CMA-SL), (b) fine-textured acidic Cambisols (CMA) and calcareous Cambisols (CMC) and (c) Luvisols. For Luvisols, the argic horizons were excluded. This result is consistent with the elevated CEC in the BC-rich terra preta soil of Amazonia (Smith, 1980; Sombroek et al., 1993) where CEC per unit of OC can be up to three times greater than that in adjacent soil (Glaser and Birk, 2012). In contrast, the CEC of the charcoal-enriched topsoil of the currently

167

Chapter 7. Forest soil active kiln site in our study is not enhanced in proportion to OC content (Table 7.1). This corresponds with the small CEC values reported for other charcoal kiln sites shortly after the last charring (Nigussie and Kissi, 2011; Ogundele et al., 2011; Oguntunde et al., 2004). Fresh charcoal has a small negative charge (Cheng et al., 2006), but the oxidation of charcoal particles and interactions with natural soil organic matter over time creates a high density of carboxylic and, to a lesser extent, phenolic groups on the surface of charcoal particles (Lehmann et al., 2005). This leads to an increase in the CEC of BC with aging (Liang et al., 2006; Cheng et al., 2006). The large specific surface area of charcoal might also contribute to the more elevated CEC per unit of carbon relative to humic soil organic matter (Liang et al., 2006), although the specific surface area of charcoal decreases with time because of the blocking of pores by interactions with minerals (Czimczik and Masiello, 2007) and organic matter (Pignatello et al., 2006). Our data also show that soil with coarser textures (Podzols, Arenosols and sandy loam Cambisols) have a larger CEC per unit of OC than the Luvisols and finer textured Cambisols (Figure 7.5). The large negative charge of BC particles after aging is attributed mainly to abiotic oxidation (Cheng et al., 2006), therefore, the larger CEC might correspond to better aeration and oxygenation of charcoal particles.

Soil acidity

All studies report pH values close to neutral in charcoal-enriched soil soon after wood charring (Nigussie and Kissi, 2011; Ogundele et al., 2011; Oguntunde et al., 2004), which our data support for the currently active kiln site (Table 7.1). The liming effect derives from the acid neutralizing capacity of the oxides, hydroxides and carbonates of alkali and alkaline earth metals such as Ca2+, Mg2+ and K+ in the ash produced by partial combustion of the charred biomass (Demeyer et al., 2001). More than 150 years after abandonment, the kiln sites of Wallonia on Arenosols, Luvisols, acidic Cambisols and Podzols have an acidic pH, although pH values are still significantly higher than those of the reference samples. In contrast, on calcareous Cambisols that have slightly acidic to neutral pH values, the charcoal kiln site does not decrease soil acidity and might even increase it slightly. The evolution of base saturation with depth in acidic soil elucidates the re-acidification dynamics that have occurred since the last wood charring episode (Figure 7.6). The base saturation of the topsoil of the kiln is similar to that of the reference topsoil, whereas it is larger in the kiln subsoil. Under the oceanic temperate climate of Wallonia, natural soil acidification over more than 150 years since the last pyrolysis has led to the vertical leaching of cations 168

Chapter 7. Forest soil from the topsoil to the subsoil where the increase in base saturation accords with higher pH values. This is related to 150 years of rainfall, corresponding to about 120 m of water, including about 40 m that were leached through the soil and contributed to the vertical leaching of nutrients. The leaching of ‘base’ cations to the subsoil since abandonment of the kiln is illustrated by comparison of the evolution of base saturation with depth in the currently active kiln to that of a pre-industrial charcoal kiln site in comparable soil conditions (Figure 7.6).

Figure 7.6. Base saturation (BS) plotted against soil depth for the currently active charcoal kiln site (K t=0), the charcoal kiln site > 150 years old LV4 (K t >150) and the adjacent reference soil (R). The currently active charcoal kiln (K t=0) represents the initial state of the base saturation soon after charring. Base saturation is 100 % in the topsoil and decreases continuously to values of about 10 % at the bottom of the soil profile, comparable to that of the reference soil. More than 150 years after abandonment of the kiln site (K t >150), the topsoil is completely desaturated. In contrast, base saturation increases continuously with depth to reach 63 % at the bottom of the profile. The leaching of the ‘base’ cations from the topsoil to the subsoil over the natural re-acidification of the soil explains the switch of base saturation from the top to the bottom of the profile. The buffering effect of the functional groups present at the surface of charcoal particles in the topsoil might explain the persistence of a small increase in pH. The soil acidification kinetics after charcoal enrichment can be expected to vary depending to a large extent on the liming potential of charcoal (Demeyer et al., 2001), but also on soil and climatic conditions. In the calcareous Cambisols, natural carbonates control the pH, which explains why charcoal enrichment has no long-lasting effect on the acidity.

169

Chapter 7. Forest soil

Plant nutrients

The ash fraction of charcoal acts as a source of macro and micronutrients in the short term (Demeyer et al., 2001), which is consistent with the large concentrations of exchangeable Ca2+, Mg2+, K+ and Na+ in the topsoil of the charcoal kiln sites soon after pyrolysis (Nigussie and Kissi, 2011; Ogundele et al., 2011; Oguntunde et al., 2004; Vazquez-Marrufo et al., 2003).

Figure 7.7. Boxplots of the ratio between exchangeable Ca2+, Mg2+ and K+ and the cation exchange capacity (Ca/CEC, Mg/CEC, K/CEC) of top- and sub-soil of the charcoal kiln sites of Wallonia (K) and the reference soils (R). Calcareous Cambisols were excluded from the data. In Amazonian dark earths, BC seems to prevent nutrients from leaching because of its large CEC (Glaser et al., 2001; Lehmann et al., 2003; Liang et al., 2006). However, in Wallonia, the charcoal-enriched topsoil of kiln sites is clearly depleted in K+ compared to the reference soil in spite of their larger CEC. This effect is clear in the base saturations (Figure 7.7), which also show

170

Chapter 7. Forest soil that Mg2+ saturation is less in the charcoal-enriched topsoil whereas that of Ca2+ is no different from the reference topsoil. This suggests a stronger affinity of the strongly carboxylic CEC of charcoal (Cheng et al., 2006) for Ca2+ than for K+ and, to a lesser extent for Mg2+. It accords with the large complexing power of the carboxylate groups of natural organic matter for Ca2+ (Kalinichev and Kirkpatrick, 2007). Consequently, exchangeable K+ and Mg2+ were preferentially leached to the subsoil during the re-acidification process. Accordingly, Lehmann et al. (2003) showed that the growth of K-demanding crops was not improved on terra preta soil. Our data also indicate that the input of available nutrients by ash dissipates from charcoal-enriched topsoil in approximately two centuries in temperate soil. This result suggests that the improved nutrient status of Amazonian dark earths that persists over millennia and has generally been explained by better retention of nutrients by charcoal might also depend on other factors, possibly climate (water balance) and soil type or the organic and inorganic waste from the genesis of terra preta soil (Glaser et al., 2001) that might supplement nutrients in the long term. Soluble orthophosphate is present in the ash (Certini, 2005), therefore available P increases at kiln sites after wood pyrolysis (Chidumayo, 1994; Nigussie and Kissi, 2011; Ogundele et al., 2011). Our data for the currently active kiln site confirm this effect (Table 7.1). In contrast, the topsoil of the abandoned charcoal kiln sites of Wallonia appears to have much smaller concentrations of available P than the reference soils. Some studies also reported small concentrations of available P in the charcoal-enriched soil of abandoned charcoal kiln sites (Gómez-Luna et al., 2009; Vazquez-Marrufo et al., 2003; Young et al., 1996), whereas others reported slightly increased values relative to the reference soils (Borchard et al., 2014; Criscuoli et al., 2014). These discrepancies might result from differences in extraction protocol; four different extracting solutions were used in the five studies considered. Nevertheless, the availability of P seems to decrease over time because available P concentration is less in the soil of an old charcoal kiln than shortly after charcoal production, whereas the increase in total P content remains (Criscuoli et al., 2014). This might be linked to a decrease in pH from close to neutral at the time of charcoal production after a century of soil re- acidification. At low pH, P can precipitate in markedly insoluble Al- and Fe- phosphates or be adsorbed at the surface of poorly crystalline Al and Fe oxides and become less soluble (Sanyal and De Datta, 1991). In addition, the organic functional groups at the surface of charcoal might play the role of ligand for

171

Chapter 7. Forest soil trivalent cations such as Fe3+ and Al3+ whose solubility increases with decreasing pH. Thereby, charcoal might contribute to fix P by cationic bridges or co-precipitation on its surface with Fe and Al. We have discussed the role of lixiviation through re-acidification to explain the decrease of exchangeable cations and the shift in pH from neutral to low values to elucidate the decrease of P availability in the topsoil of pre-industrial charcoal kiln sites. Nevertheless, the low concentration of available nutrients also reflects that aged charcoal, because of its low degradability, does not participate to the recycling of nutrients like uncharred organic matter do. Nutrients most-limiting for plant growth accumulate in the topsoil as a result of biological cycling (Jobbagy and Jackson, 2001). They accumulate in plant and animal biomass and return to soil through the decay of soil organic matter. In contrast, in a biochar system, available nutrients from ash are readily available for plant uptake. Hence, pyrolysis accelerates the return of nutrients to soil by the partial combustion of biomass. Once nutrients have been lixiviated from soil or adsorbed to minerals after > 150 years, poor concentrations of available nutrients remain: aged charcoal plays a limited role in recycling nutrients because it decomposes very slowly. This has important implications, particularly for P availability. We have plotted the content of available P (Figure 7.8a) and two indicators of P availability (Figure 7.8b) against the content of water extractable organic carbon in the topsoil of pre-industrial charcoal kiln sites and adjacent reference soils for the four sites on Luvisols. The graphs show that the low concentration and availability of P in kiln soils goes along with a low concentration of WEOC, which contrasts with the high P availability and WEOC concentration of adjacent reference soils. The abundance of organic ligands in the soil solution can influence P availability because they compete with (di)hydrogenophosphates for the anions adsorption sites on soil colloids (Regelink et al., 2015; Uehara and Gillman, 1981). The correlation between WEOC concentration and available P may also result from dissolved organic P in the extract (Qualls et al., 2000). These results support the idea that the low availability of P in the topsoil of pre-industrial kiln sites is related to the low concentration of dissolved organic matter because of a poor degradability of charcoal. In that sense, it is worth to remind that charcoal cannot support all ecological functions of uncharred organic matter.

172

Chapter 7. Forest soil

Figure 7.8. Relationship between water extractable organic carbon (WEOC) content and (a) available P (Pav) concentration, (b) the ratio between available P and total P (Pav/Ptot) and (c) the ration between available P and inorganic P (Pav/Pinorg) in the topsoil of pre-industrial charcoal kiln sites (black circles) and adjacent reference soils (open circles).

7.5. Conclusion

Charcoal enrichment increases the OC content and, to a larger extent, the C:N ratio of topsoil of the charcoal kiln sites. Shortly after pyrolysis, the charcoal- enriched soil has a small CEC, but it increases considerably over time by oxygenation of the surface of charcoal particles. Therefore, after > 150 years, the CEC per unit of OC of the charcoal-enriched soil is much larger than that of the natural soil. The largest CEC per unit of OC was on the sandy textured soils. On initially acidic soils, the liming effect of wood ash attenuates over time through soil re-acidification, leading to leaching of the ‘base’ cations to the subsoil. The pH of acidic soil remains slightly higher in the kiln soil after >150 years, whereas the acidity of naturally carbonate-rich Cambisols is not reduced. The topsoil of the charcoal kiln sites of Wallonia is much depleted in exchangeable K+ and available P, which contrasts markedly with the increase that is observed shortly after pyrolysis. It is attributed to the weak affinity of K+ for the functional groups at the surface of charcoal, and to a decrease in P availability with time. Therefore, we recommend further investigation of the long-term effects of biochar on the dynamics of plant nutrients.

173

Chapter 7. Forest soil

Acknowledgements

We are grateful to Claudine Givron, Anne Iserentant, Patrick Populaire, André Lannoye and the technical staff of the Centre Agri-Environnemental de Michamps de Michamps for their help with field and laboratory work. We acknowledge Benoît Pereira and Catherine Rasse from the SMCS of the Université catholique de Louvain for their statistical support. We thank the reviewers and editors for their constructive feedback that improved greatly the quality of the manuscript. Finally, we thank the Direction Générale Opérationelle de l’Agriculture, des Ressources Naturelles et de l’Environnement of the Service Public de Wallonie for funding.

174

Chapter 8. Cropland soil

Chapter 8. Pre-industrial charcoal kiln sites in Wallonia (Belgium) to evaluate the long-term effect of biochar on temperate cropland soil properties.8

Summary

Research on biochar has increased, but its long-term effect on the fertility of temperate agricultural soil remains unclear. In Wallonia (Belgium), pre- industrial charcoal production affected former forested areas that were cleared for cultivation in the 19th century. The sites of traditional charcoal kilns, largely enriched in charcoal residues, are similar to soil amended with hardwood biochar more than 150 years ago. We sampled 17 charcoal kiln sites to characterize their effect on soil properties compared to adjacent reference soils. Charcoal-C content was estimated by differential scanning calorimetry. The kiln soil contains from 1.8 to 33.1 g kg-1 of charcoal-C, which markedly increases organic C:N and C:P ratios. It also contains slightly more uncharred SOC than reference soil, which accords with larger total N content. We measured a small increase in nitrates in the kiln soil that might relate to greater mineralization and nitrification of organic N. Frequent application of lime raised the pH to values close to neutral, which offset the residual effect of charcoal production on soil acidity. A cation exchange capacity (CEC) of 414 -1 cmolc kg was estimated for charcoal-C, whereas that of uncharred SOC is -1 213 cmolc kg . The CEC of charcoal remains considerably smaller than that estimated for black carbon in millennial Amazonian terra preta soil, which supports the view that the CEC of charcoal might continue to increase through the progressive oxidation of charcoal over longer periods of time. Despite the large CEC, exchangeable K+ content remained unchanged in the kiln soil, in contrast to exchangeable Ca2+ and Mg2+ that are increased strongly. Charcoal enrichment has little effect on available, inorganic and total P, but it can form strong complexes with Cu, which reduces its availability. Biochar is very persistent in soil, therefore, long-term implications should not be overlooked.

8 Adapted from: Hardy, B., Cornelis, J.-T., Houben, D., Lambert, R., Dufey, J.E., 2016. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium. European Journal of Soil Science 67, 206–216. doi:10.1111/ejss.12324 175

Chapter 8. Cropland soil

8.1. Introduction

Biochar is the solid residue of the pyrolysis of biomass produced intentionally to amend soil (Brown, 2009). In the context of mitigating the effects of climate change, biochar can increase soil carbon sequestration (Lehmann, 2007b). Like other types of black carbon, biochar lasts for longer in soil than uncharred organic matter because it has a fused aromatic ring structure (Solomon et al., 2007b). As biochar also contributes to the sustainable fertility of anthropogenic dark earths (Downie et al., 2011; Glaser and Birk, 2012), its effect on soil properties has been investigated increasingly in the last few decades. Recently, meta-analyses have shown that biochar application to soil increases slightly the average crop production (Biederman and Harpole, 2013; Jeffery et al., 2011), however, the effect varies greatly depending on soil conditions (Biederman and Harpole, 2013; Jeffery et al., 2011) and quality of the biochar (Biederman and Harpole, 2013; Jeffery et al., 2011; Manyà, 2012). Although biochar application to soil may decrease crop productivity in some circumstances, acidic to neutral soil with coarse to medium texture generally benefits from biochar amendment (Jeffery et al., 2011). This supports the idea that biochar’s liming effect and water-holding capacity are the main factors in improving soil fertility. Yields of crops grown on soil with small organic carbon (OC) content and cation exchange capacity (CEC) also respond positively to amendment with biochar (Crane-Droesch et al., 2013). Although there is a better understanding of the short-term effects of biochar application to soil, the mid- to long-term effects are still largely unknown because of the lack of long-term trials. Biochar is very persistent in soil (Singh et al., 2012), and its properties change over time (Cheng et al., 2006; Lehmann et al., 2005). Therefore, its long-term effect on soil properties needs further investigation. Research on historical charcoal deposits in anthropogenic dark earths (Glaser and Birk, 2012; Lehmann et al., 2003; Solomon et al., 2007b), pre-industrial charcoal kilns (Borchard et al., 2014; Criscuoli et al., 2014; Hardy et al., 2016; Mikan and Abrams, 1995) or anthropogenic oven mounds (Downie et al., 2011) has partly filled this gap in knowledge. Nevertheless, data on the long-term effect of biochar on the fertility of soil under intensive agriculture are limited, even though applications of biochar are mainly to cropland soil. Cultivation might affect biochar and its effect on soil properties by the mechanical action of tillage and the use of fertilizers. Therefore, the long-term effect of biochar on cropland soil fertility needs to be clarified. In that goal, pre-industrial charcoal kiln sites that were deforested for cultivation can play a major role. These were largely enriched with charcoal > 150 years

176

Chapter 8. Cropland soil ago, which make them comparable to soil amended with hardwood biochar in the past. To our knowledge, soil properties of pre-industrial charcoal kilns sites were studied exclusively in forest, probably because their occurrence in agricultural soil is rare. These sites, which are soil patches contaminated with charcoal because of in situ fuel production in the past, contrast with Amazonian terra preta that were enriched purposely with fire residues but also organic and inorganic waste to enhance soil fertility of strongly weathered tropical soils (Glaser and Birk, 2012). Pre-industrial charcoal kiln sites also differ from biochar application to soil because in situ pyrolysis disturbed the soil not only by the accumulation of charcoal residues, but also by site preparation and heating (Mikan & Abrams, 1995; Hardy et al., 2016). Nevertheless, in cropland soil, repeated tillage has obliterated the relief of the sites and reduced the effects of charcoal production on topsoil structure. As we will show in this chapter, no clear effect on texture (Figure 8.1) and clay mineralogy (Appendix 2) were detected, which makes charcoal enrichment the most persistent consequence of charcoal production. Therefore, former charcoal kiln sites in cropland soil of Wallonia are natural models to evaluate the long-term effect (>150 years) of hardwood biochar on cropland soil properties in a temperate climate.

8.2. Material and methods

Soil samples

In this chapter, we analyzed almost exclusively topsoil (0–25 cm) and subsoil (35–50 cm) samples of the 17 charcoal kiln sites and adjacent reference soils from cropland, mainly sited on silt loam Luvisols developed from the weathering of quaternary loess (detailed description in Chapter 4). Only the determination inorganic N required a specific sampling, because soils must be kept cold during storage and analyzed as soon as possible after sampling to avoid the that a significant amount of organic N mineralize during storage. Inorganic N was measured on fresh samples from four plots sampled on 7 and 21 November 2013, after the harvest of sugar beet, wheat, chicory and colza. In each plot, we sampled the topsoil (0–30 cm) and the subsoil (30–60 cm) of 4 charcoal kiln sites and their reference soils according to the sampling procedure of the public administration of Wallonia for the control of potentially leachable N content in agricultural land (Vandenberghe et al., 2013).

177

Chapter 8. Cropland soil

Physico-chemical analyzes

Samples from the 17 sites were dried at 40 °C, gently ground and sieved to 2 mm. The < 2 mm fraction was analyzed. Particle-size distribution of the topsoil samples was measured by Robinson’s pipette sedimentation method (Day, 1965) after destruction of carbonates with 2.5 M HCl and organic matter with 30 % H2O2 according to the norm AFNOR X31-107. To test the validity of the analysis in presence of a large amount of charcoal, we measured residual carbon concentration in the granulometric fractions of a selection of samples (n=3). Silt and clay didn’t contain more carbon in kiln soil than in reference soil. Nevertheless, more carbon was measured in the sand fraction, which results from large charcoal particles that survived oxidation with H2O2. Soil pH was measured in demineralized water (pH-H2O) and in a 1M KCl solution (pH-KCl) with a 1:5 soil:solution mass ratio. Elemental C and N contents were measured by dry combustion (vario MAX, Elementar). The inorganic C content was measured by the modified-pressure calcimeter method on finely ground subsamples (< 200 µm) (Sherrod et al., 2002). Inorganic C content was null or below the detection limit (< 0.2 g kg-1). Therefore, total C was considered to correspond exclusively to organic C (OC) and includes charcoal-C. We measured the potential CEC by percolation of 1M ammonium acetate (naturally buffered at pH 7) on soil columns (Metson, 1956). Ammonium was desorbed with a 1.33 M KCl solution and measured by colorimetry (ISO7150/1). Available Ca, Mg, K, Na, P, Cu and Zn were extracted with a 0.5 M ammonium acetate–0.02 M ethylenediaminetetraacetic acid (EDTA) solution at pH 4.65 with a 1:5 soil:solution mass ratio (Lakanen and Erviö, 1971) and measured by inductively coupled plasma-atomic emission spectroscopy (ICP-AES, Thermo). In the absence of carbonates, the extracted Ca, Mg, K and Na correspond to exchangeable cations (Cottenie et al., 1975; Hardy et al., 2016). We calculated the base saturation of the soil as the ratio between the sum of exchangeable Ca2+, Mg2+, K+ and Na+ and the CEC. We determined organic P by the difference between the P content extracted with

0.5M H2SO4 before and after calcination of the soil at 550 °C (Van Ranst et al., 1999). To measure the total concentrations of P, Cu and Zn, subsamples of soil were ground mechanically to a powder and were digested with a mixture of three concentrated acids (HF, HNO3 and HClO4), dissolved in aqua regia and analyzed by ICP-AES (Lambrechts et al., 2011). Inorganic P was calculated by difference between total and organic P. All results are expressed per mass of dry soil (105 °C).

178

Chapter 8. Cropland soil

For the analysis of inorganic N, Immediately after sampling, the soil samples were carried in an icebox and stored into a fridge at 4°C overnight to limit the mineralization of organic N. The day after, the wet soil was sieved to 1 cm. - + Nitrate (N-NO3 ) and ammonium (N-NH4 ) ions were extracted from 40 g of soil with a 0.5 M KCl solution (1:5 soil:solution mass ratio) and measured by colorimetry (ISO 14256-2:2005). The humidity of the soil was measured by gravimetric analysis on 150 g of soil dried at 105°C for 72 hours. The results are presented on a dry matter basis.

Quantification of charcoal-C

Data analysis

Differences between the properties of kiln and reference soils were tested with two-sided paired t-tests in R 3.2.2 (R Core Team, 2012). Data from the topsoil and from the subsoil were analyzed separately. For each variable, the statistical distribution of the differences between pairs (a pair is a kiln soil and the adjacent reference soil) was examined with histograms. No clear deviation from normality was detected. To estimate the contribution of charcoal and uncharred SOC to the CEC of the soil of pre-industrial kiln sites, we fitted a linear model to predict CEC of soil based on clay, charcoal-C and uncharred SOC contents, which are expected to be the main factors that control CEC of our soil. We fixed the CEC of clay -1 to 42.6 cmolc kg based on data from charcoal-free forest Luvisols (Hardy et al., 2016), which are the forest equivalent of the cropland soil of this study. Data were analyzed in R 3.2.2 (R Core Team, 2012).

8.3. Results

More than 150 years after the last production of charcoal, charcoal kiln site has no clear effect on soil texture (Figure 8.1). Clay content is slightly smaller 179

Chapter 8. Cropland soil in kiln soil than in reference soil whereas sand content is slightly larger, but differences are not significant.

Figure 8.1. Textural analysis of cropland topsoil of the kiln plotted against that of the reference soil: (a) clay, (b) silt and (c) sand.

180

Chapter 8. Cropland soil

Figure 8.2. Topsoil (0–25 cm) properties of charcoal kiln sites (K) plotted against that of reference soil (R). The P-values refer to two-sided paired t-tests. The mean relative variation caused by the charcoal kiln as a percentage of the value in the reference soil is indicated on the graph; (a) organic carbon (OC); (b) total nitrogen (N); (c) C:N ratio; (d) C:P ratio (ratio between OC and organic phosphorus); (e) pH measured in water (pH-H2O); (f) pH measured in KCl (pH-KCl); (g) base saturation (BS); (h) cation exchange capacity (CEC); (i) exchangeable calcium (Ca2+); (j) exchangeable magnesium (Mg2+); (k) exchangeable potassium (K+) and (l) exchangeable sodium (Na+).

181

Chapter 8. Cropland soil

Figure 8.3. Topsoil (0–25 cm) properties of charcoal kiln sites (K) plotted against that of reference soil (R). The P-values refer to two-sided paired t-tests. The mean relative variation caused by the charcoal kiln as a percentage of the value in the reference soil is indicated on the graph; (a) total phosphorus (P tot.); (b) organic phosphorus (P org.); (c) inorganic phosphorus (P inorg); (d) available phosphorus (P av.); (e) total copper (Cu tot.); (f) available copper (Cu av.); (g) total zinc (Zn tot.); (h) available zinc (Zn av.). In the topsoil, total OC content (corresponding to the sum of charcoal-C and uncharred SOC) ranges from 10.1 to 26.4 g kg-1 in reference soil and from 14.3 to 64.4 g kg-1 in kiln soil (Figure 8.2a). Although reference soils are supposed to be unaffected by charcoal production, small amounts of charcoal- C were detected in the topsoil of most of them, from 0 to 2.6 g kg-1. This corresponds to 5.3 % of total OC on average. By comparison, topsoil of pre- industrial charcoal kiln sites is enriched strongly with charcoal. It contains from 1.8 to 33.1 g kg-1 of charcoal-C, which corresponds to 39.2 % of total OC on average and up to 51.4 %. The content of charcoal-C in the topsoil of

182

Chapter 8. Cropland soil the kiln site is strongly correlated to the excess of OC (∆OC) that accumulated in the kiln soil relative to adjacent reference soil (Figure 8.4a). Accordingly, there is a strong correlation (r = 0.93) between uncharred SOC content of reference topsoil and that of kiln topsoil (Figure 8.4b). Nevertheless, the latter is +22.8 (14.2) % larger on average (P < 0.001).

Figure 8.4. (a) Charcoal-C content in the topsoil of the kiln sites plotted against the excess of OC in the kiln soil (∆OC), measured by difference between OC content in the kiln and in adjacent reference soil and (b) Uncharred soil organic carbon (SOC) content in kiln soil plotted against that in the adjacent reference soil. Together with the overall enrichment in total OC content, total N content (Figure 8.2b), C:N ratio (Figure 8.2c) and CEC (Figure 8.2h) are considerably larger in the topsoil of the kiln site than in the reference soil. The raise in CEC is strongly related to the raise in OC content (Figure 8.5a), mainly attributed to the enrichment in charcoal-C (Figure 8.5b). The best fitting linear model to -1 predict soil CEC calculated a CEC of 213 cmolc kg for uncharred SOC and -1 of 414 cmolc kg for charcoal-C (Table 8.1).

183

Chapter 8. Cropland soil

Figure 8.5. Cation exchange capacity (CEC) of kiln (K, filled symbols) and reference (R, open symbols) topsoil plotted against: (a) organic carbon (OC) and (b) charcoal- C. Together with large CEC, concentrations of exchangeable Ca2+, Mg2+ and Na+ are also larger in the kiln soil than in the reference soil, whereas that of exchangeable K+ is not different (Figure 8.2i-l). In spite of the larger concentrations of exchangeable cations, base saturation is significantly smaller in the kiln topsoil (Figure 8.2g). In particular, charcoal-rich soil is less saturated in K+ than reference topsoil (–32.6 %) (Figure 8.6c). The saturation of Mg2+ (Figure 8.6b) and Na+ (Figure 8.6d) are also clearly less, but to a smaller extent. Table 8.1. Parameters of the best fitting model to predict cation exchange capacity (CEC) of soil based on charcoal-C and uncharred soil organic carbon (SOC) contents. -1 The CEC of clay was fixed at 42.6 cmolc kg . Coefficients / Std. error t-value P-value (> t) -1 cmolc kg charcoal-C 414 4.4 9.2 < 0.001 uncharred SOC 213 2.5 8.6 < 0.001

184

Chapter 8. Cropland soil

Figure 8.6. Ratios between exchangeable Ca2+, Mg2+, K+ and Na+ and cation exchange capacity (Ca/CEC, Mg/CEC, K/CEC and Na/CEC) in the topsoil (0–25 cm) of charcoal kiln site (K) plotted against that in the reference (R) topsoil. The P-values refer to two-sided paired t-tests. The mean relative variation caused by the charcoal kiln as a percentage of the value in the reference soil is indicated on the graph. (a) Ca/CEC, (b) Mg/CEC, (c) K/CEC and (d) Na/CEC.

Although the pH-H2O is close to neutral in both the kiln and the reference soil (Figure 8.2e), it is slightly lower at the kiln sites, which accords with the smaller base saturation. Total and organic P concentrations are slightly larger in the kiln soil, in contrast to inorganic and available P content which are not affected significantly (Figure 8.3a-d). The kiln soil is enriched in total Cu and Zn (Figure 8.3e, g). Larger plant available Zn concentration is in agreement with it (Figure 8.3h), in contrast to available Cu that is clearly less in the kiln soil (Figure 8.3f). We calculated a decrease in Cu availability of –20.7 (9.8) % (P<0.001) in the kiln relative to reference soil. In the four plots sampled in - November 2013, the N-NO3 concentration measured at kiln site was larger + (+21.6 %; P<0.001) (Figure 8.7). The N-NH4 concentration was below the limit of detection in both the kiln and the reference soil.

185

Chapter 8. Cropland soil

- Figure 8.7. Concentration of N-NO3 in the kiln soil plotted against that in the reference soil of four cropland plots sampled in the intercropping period before winter; The crops preceding the sampling were colza (Brassica napus L., sugar beet (Beta vulgaris L.), chicory (Cichorium intybus L.) and wheat (Triticum aestivum L.); (a) topsoil (0–30 cm) and (b) subsoil (30–60 cm). Most of the effects observed in the topsoil of the kiln sites disappear or attenuate in the subsoil (Figure 8.8). Nevertheless, the C:N ratio, the CEC and the concentration of exchangeable Na+ remain significantly larger in the subsoil of the kiln site than in that of the reference, and both pH-H2O and pH- KCl remain significantly smaller (Figure 8.8). Remarkably, the concentration of exchangeable K+ is 8.8 % larger in the subsoil of the kiln sites than in the reference subsoil (P = 0.02; Figure 8.8k), whereas no difference was recorded in the topsoil.

186

Chapter 8. Cropland soil

Figure 8.8. Subsoil properties of the kiln site plotted against that of the reference soil. The P-values refer to two-sided paired t-tests. The mean relative variation caused by the charcoal kiln as a percentage of the value in the reference soil is indicated on the graph; (a) organic carbon (OC); (b) total nitrogen (N); (c) organic C:N ratio (C:N); (d) Cation exchange capacity (CEC); (e) pH measured in water (pH-H2O); (f) pH measured in KCl (pH-KCl); (g) Available phosphorus (P av.); (h) base saturation (BS); (i) exchangeable calcium (Ca2+); (j) exchangeable magnesium (Mg2+); (k) exchangeable potassium (K+); (l) exchangeable sodium (Na+);

187

Chapter 8. Cropland soil

8.4. Discussion

Charcoal-C, uncharred SOC and N

Kiln and reference soil from the same field have been subject to identical soil and climatic conditions and management practices since charcoal was last produced. Therefore, the content of uncharred SOC of the kiln soil is closely related to that of adjacent reference soil, which explains that the excess of OC recorded in the kiln relative to adjacent reference soil corresponds mainly to charcoal-C. Nevertheless, we measured a significant increase of uncharred SOC in the kiln soil, like Kerré et al. (2016). A greater accumulation of uncharred organic matter related to charcoal or biochar enrichment was also reported by earlier studies (e.g. Liang et al., 2010; Hernandez-Soriano et al., 2015), possibly because of adsorption of dissolved organic molecules on to the surface of charcoal (Pignatello et al., 2006) or from a greater biomass productivity in presence of charcoal (Jeffery et al., 2011), which might have increased the return of organic matter to soil. The content of uncharred SOC at kiln site was about 1.2 times that of adjacent reference soils, which accords with the results of Kerré et al. (2016), who measured a content of uncharred SOC at kiln site ranging from 1.0 to 1.4 times that of adjacent reference soils. We attribute the presence of small amounts of charcoal-C in reference soil, confirmed by microscopic inspection, to contamination from the kiln site, or from burning that generally followed deforestation when land was converted from forest to agricultural land in the past (Hoyois, 1953). The small but systematic increase in total N content of the kiln soil is most likely related to the larger amount of uncharred organic matter, even though ‘black N’ in the molecular structure of charcoal (Knicker, 2010) may - contribute to N storage. The larger content of N-NO3 in the kiln soil might result, in all or in part, from more mineralization and nitrification of organic N during the intercropping period before samples were taken, when plant uptake is not active. A better drainage due to the increase of macroporosity in presence of charcoal (Kerré, unpublished data) and a faster soil warming due to the black color of charcoal, decreasing the albedo, might contribute to increase nitrification rates. Other possible explanations, such as smaller losses - of N-NO3 by leaching (Yao et al., 2012) or denitrification in presence of charcoal, are less likely. Regardless of the process that underlies the enrichment in nitrate, this result supports a positive effect of charcoal on the long-term availability of N in soil, whereas short-term mineralization of the N-poor, labile fraction of hardwood charcoal might cause N immobilization

188

Chapter 8. Cropland soil

(Ameloot et al., 2015). The large C:N ratio in the kiln soil emphasizes the presence of charcoal, because the composition of char is closely related to that of the parent feedstock.

Soil acidity

In acidic forest soil, cations provided by wood ash are leached to the subsoil through the natural soil acidification process, which desaturates the topsoil of the kiln more than 150 years after abandonment of the site (Hardy et al., 2016). Accordingly, soil pH decreases strongly, but remains, however, slightly higher than in the adjacent reference soil (Criscuoli et al., 2014; Hardy et al., 2016; Mikan and Abrams, 1995). At low pH values, the presence of carboxylate (the conjugate base of carboxylic acid) on the surface of charcoal (Lehmann et al., 2005) buffer acidity through association with protons (H+), which contributes to maintain a slight increase in pH on the long-term in forest soils. In the cropland soil of this study, frequent applications of liming amendments have decreased the acidity of both kiln and reference soil to pH values close to neutral and base saturation close to 100 %. These amendments have masked the residual liming effect of ash. The kiln soil is even slightly more acidic than the reference soil, with significantly smaller pH-H2O and base saturation. This results possibly from an excess of variable charges from acidic functions on the surface of charcoal, so that more lime is required in the kiln soil to reach the same base saturation as adjacent reference soil. Larger nitrification rates, - as suggested by the enrichment of N-NO3 , might also decrease the pH of the kiln soil. This result indicates that the benefit of hardwood biochar related to liming from ash is short-lived and becomes less important in soil that is regularly limed. By comparison, the abandoned charcoal kilns of Wallonia have no effect on the acidity of calcareous Cambisols in forest because natural carbonates buffer soil pH at values close to neutral (Hardy et al., 2016). Hardwood biochar contains generally less than 15 % of ash (Camps-Arbestain et al., 2015). If we consider that wood ash has a CaCO3 equivalence of 500 g kg-1 on average (Ohno, 1992), soil amendment with 40 t ha-1 of biochar -1 corresponds to an application of CaCO3 of about 3 t ha or less, which corresponds roughly to the amount of lime needed during a three year rotation in cropland soil of Wallonia. This highlights that the persistence of the liming effect of ash is short compared to that of the charcoal itself, which persists on centennial timescales (Singh et al., 2012). 189

Chapter 8. Cropland soil

Cation exchange capacity (CEC)

- Fresh biochars have a small CEC that generally ranges from 2 to 60 cmolc kg 1 -1 (Ippolito et al., 2015). In contrast, we calculated a CEC of 414 cmolc kg per unit of charcoal-C in the topsoil of pre-industrial charcoal kiln sites, which -1 exceeds by about twice the CEC of 213 cmolc kg obtained for uncharred SOC. This result accords with the large CEC per unit of OC in terra preta soils, up to three times larger than that in adjacent soils (Glaser and Birk, 2012; Sombroek et al., 1993). The large CEC of aged charcoal has been attributed to the large negative charge on the surface of charcoal particles (Cheng et al., 2008a) related to carboxylic and, to a lesser extent, phenolic groups formed predominantly by abiotic oxidation over time (Lehmann et al., 2005), and also to the adsorption of natural organic matter on the surface of charcoal particles (Cheng et al., 2006; Pignatello et al., 2006). These weak acidic functional groups are partly deprotonated in the range of pH of soil, which enhances cation exchange capacity. The large specific surface area of charcoal might explain the density of charge per unit of OC larger than that of uncharred organic matter (Liang et al., 2006). Even though the CEC per unit of C attributed to charcoal in centennial charcoal kiln sites of Wallonia is much more elevated that the average CEC of uncharred SOM, CEC values are considerably smaller than that of BC in millennial terra preta soils. CECs of about 800–900 cmolc kg-1 can be estimated graphically with the data of Sombroek et al. (1993) plotted by Glaser and Birk (2012), and CECs above 1000 cmolc kg-1 were estimated for BC by Mao et al. (2012). This supports the view that the CEC of charcoal might continue to increase through the progressive degradation of charcoal over longer periods of time in pre- industrial charcoal kiln sites of Wallonia. Therefore, the effect of BC on CEC of soil and of the SOM pool will depend strongly on the degree of oxidation and degradation of BC particles, as well as the CEC of uncharred SOM that can vary to a large extent according to soil and climatic conditions (Helling et al., 1964).

Exchangeable cations

In cropland soil, the increase in exchangeable Ca2+, Mg2+ and Na+ concentrations in the topsoil of the kilns accords with the increase in CEC. These cations were provided mostly by amendments with lime and fertilizers from farming that started after abandonment of the kiln sites rather than by ash from charcoal, because the topsoil of the kiln is clearly desaturated after > 150 years of natural acidification in forest (Hardy et al., 2016). In contrast,

190

Chapter 8. Cropland soil the fact that the concentration of exchangeable K+ remained unchanged in the topsoil of the kiln but has slightly increased in the kiln subsoil supports that exchangeable K+ is preferentially leached from the charcoal-rich soil. Compared to other exchangeable cations, particularly Ca2+, exchangeable K+ has a weak affinity for the carboxylic groups of aged charcoal (Hardy et al. 2016). Retention of exchangeable K+ is promoted by specific retention sites of clay minerals. Terra preta soils are generally enriched largely with Ca2+ and other nutrients, but not K+ (Falcão et al., 2009; Lehmann et al., 2003), which accords with our result. The small affinity of charcoal for K+ has been observed in long-term surveys only because directly after biochar application to soil, available K+ content increases systematically (Biederman and Harpole, 2013) by dissolution of K oxides from ash. The presence of aged charcoal in Amazonian terra preta enhances the CEC markedly and plays a major role in preventing cations from leaching and in maintaining soil fertility (Glaser and Birk, 2012). This is important because terra preta soils mainly occur on strongly weathered tropical Ferralsols (Glaser and Birk, 2012) that have a weak permanent negative charge (Uehara and Gillman, 1981). However, the small adsorption of K+ on aged charcoal coupled with the relative enrichment by other cations such as Ca2+ and Mg2+ suggests that K (induced) deficiency might appear in the long-term in biochar- rich soil with little K supply (Falcão et al., 2009). Induced deficiency in K is not likely in temperate soil because most have a permanent negative charge related to 2:1 clay minerals, which promotes K+ adsorption on specific retention sites and supplies bioavailable K through weathering of these 2:1 minerals.

Availability of P, Cu and Zn

Wood ash may contain available P in the form of soluble orthophosphate (Certini, 2005), and can also enhance the availability of P in soil by increasing soil pH. Nevertheless, neither available nor inorganic P concentrations are significantly affected by charcoal enrichment in pre-industrial kiln sites of this study. This supports the view that the input of P from ash is relatively small compared to the content of inorganic P naturally present in soil or provided by fertilizers. Moreover, as charcoal has a limited effect on soil acidity after > 150 years in cropland soil, its effect on P solubility is limited. Nevertheless, total and organic P concentrations are significantly larger in the kiln than in the reference soil. This probably relates to organic P in charcoal particles that is unavailable to plants. The large organic C:P ratio of the kiln soil (~150)

191

Chapter 8. Cropland soil confirms that hardwood charcoal is relatively poor in P, whereas biochar from P-rich feedstock releases considerable amounts of bio-available P (Novak et al., 2014). The limited effect of charcoal enrichment on P concentration in our soils suggests that the large, long-lasting stock of available P measured in most terra preta soils (Glaser and Birk, 2012) is related to enrichment by P-rich (in- )organic waste such as bones, rather than to introduction of fire residues or biochar. Kiln sites were used repeatedly for charcoal production every 20 years on average (Hardy and Dufey, 2012a), therefore, the soil was enriched in trace metals such as Cu and Zn contained in the biomass that was accumulated and pyrolysed at kiln site. The decrease in availability of Cu in presence of charcoal indicates the strong complexing effect of surface functional groups of charcoal for Cu, which can be still accentuated by aging (Cheng et al., 2014). It is not known whether a large accumulation of aged charcoal could cause plants to be deficient in Cu; this question should be addressed in future research on biochar.

8.5. Conclusion

After > 150 years, the topsoil of pre-industrial charcoal kiln sites is still enriched largely with charcoal-C, up to 33.1 g kg-1. It also contains slightly more uncharred SOC than adjacent reference soil. We measured an increase - in N-NO3 in the kiln soil, which might result from more mineralization and nitrification related to a surplus of organic N and modification of soil physical properties (better drainage, smaller albedo). Our results also showed that aged -1 charcoal has a large CEC, estimated to be 414 cmolc kg per unit of C, which is about twice larger than the CEC of uncharred SOM. Hence, charcoal affects the balance of some nutrients in soil. It promotes the retention of exchangeable Ca2+ and Mg2+, which are strongly increased in the kiln soil, but not that of exchangeable K+ that remains unchanged because of a poor affinity for the exchange complex of charcoal. Charcoal forms strong complexes with Cu, which reduces its availability by 20.7 % in kiln soil. To our knowledge, the effect of pre-industrial charcoal kilns on plant growth and crop yield has not been reported in the literature so far. Therefore, further investigation of the effect of these sites on plant growth would be of great interest to assess the long-term benefits and issues associated with the introduction of biochar to temperate soil under intensive cultivation.

192

Chapter 8. Cropland soil

Acknowledgements

We thank Claudine Givron, Anne Iserentant, Patrick Populaire, André Lannoye, Martin Berwart, Marc De Toffoli and Olivier Imbrecht from the Earth and Life Institute of the Université catholique de Louvain and the staff of the Centre Agri-environnemental de Michamps for their involvement in field and laboratory work. Funds were provided by the Direction Générale Opérationnelle de l'Agriculture, des Ressources Naturelles et de l'Environnement - Service Public de Wallonie and the FSR (Fonds Spéciaux de Recherche) - Université catholique de Louvain.

193

Chapter 9. Biological soil properties

Chapter 9. The long-term effect of charcoal accumulation at pre-industrial kiln sites on microbial activity, biomass and community structure of forest and cropland temperate soils.9

Summary

Soil amendment with biochar can modify the dynamics of decomposition of soil organic carbon (SOC) as well as soil microbial abundance and community structure. Nevertheless, the long-term evolution of these effects are unknown and of crucial importance because biochar persists in soil for centuries. We selected nine charcoal kiln sites from forest (4 sites) and cropland (5 sites) and determined microbial properties of their topsoil, largely enriched with charcoal for > 150 years. Adjacent soils were used as references unaffected by charcoal production. Soils were incubated in controlled conditions and emissions of CO2 were measured during 140 days. At day 70, when a steady- state was reached, an aliquot was sampled from each soil to determine microbial abundance and community structure by phospholipid fatty acid (PLFA) analysis. Microbial properties were related to physico-chemical soil properties and the content of charcoal-C and uncharred SOC in soil, estimated by differential scanning calorimetry. Total emissions of CO2 were strongly related to the total PLFA content in soil, with the metabolic quotient of forest soils slightly lower than that of cropland soils. The content of uncharred SOC and pH explained a main part of the variability in CO2 emissions, regardless of the presence of charcoal. Land use had a strong influence on microbial community structure, mainly due to the shift from very acidic soil conditions in forest to acidity close to neutral in cropland. In contrast, charcoal had a limited effect on microbial community structure. A small decrease of the proportion of fungi 18:1 in the kiln soil from forest was the only clear difference with reference soils. Our results support the idea that on the long- term, when the labile fraction of charcoal has been degraded, the effect of charcoal on soil CO2 emissions is indirect, mainly related to a modification of the drivers of soil microbial respiration such as acidity and availability of

9 Hardy, B., Cornelis, J.T., Sleutel, S. & Dufey, J., The long-term effect of charcoal accumulation at pre-industrial kiln sites on microbial activity, biomass and community structure of forest and cropland temperate soils. In preparation 195

Chapter 9. Biological soil properties nutrients. Soil conditions are modified in cropland due to the application of organic and inorganic fertilizers, which attenuates the residual effect of charcoal on soil fertility and therefore its influence on microbial community structure. Our study emphasized that the effect of charcoal on microbial soil properties depends on soil conditions and that long-term implications of charcoal enrichment differ from short-term effects reported in the literature.

9.1. Introduction

Biochar application to soil is a carbon negative technology to tackle climate change by decreasing greenhouse gas emissions (Lehmann et al., 2006). The potential of biochar to increase soil carbon sequestration is related to its larger resistance to decomposition than uncharred soil organic matter (SOM). The scientific community generally agrees on the fact that the low degradability of biochar, like other types of black carbon, derives mainly from intrinsic chemical recalcitrance related to a fused aromatic ring structure (Brodowski et al., 2005a; Glaser et al., 2000; Haumaier and Zech, 1995; Knicker, 2011a; Solomon et al., 2007b), which makes it different from other aromatic compounds naturally present in soil, such as lignin. Despite its low degradability, the introduction of biochar to soil often results in an increase in CO2 emissions in the short-term (Sagrilo et al., 2014). A positive priming of biochar on the decomposition of native SOM is one of the factors that can increase mineralization rates (Maestrini et al., 2014). Such effect is undesirable because it counteracts the benefits from biochar application on soil carbon storage. Nonetheless, a positive priming has only been observed shortly after addition of fresh biochar to soil and doesn’t seem to last over long periods of time (Hamer et al., 2004; Wardle et al., 2008;

Zimmerman et al., 2011). The abiotic release of CO2 can also contribute significantly to early emissions after soil amendment with biochar (Bruun et al., 2014; Jones et al., 2011), deriving from the acid-base reaction of carbonates contained in ash when biochar is added to an acidic soil.

Nevertheless, the main source of the increase in CO2 emission from a biochar amended soil seems to be the biochar itself (e.g. Cross and Sohi, 2011; Hilscher and Knicker, 2011). In a meta-analysis of the results of 46 studies published between 2009 and 2014, Sagrilo et al., (2014) showed that large additions of biochar to soil increased considerably CO2 emissions, whereas a low input of biochar relative to native soil organic carbon (SOC) content did not affect emissions significantly. They also highlighted that biochars produced at temperatures < 350 °C were more subject to increase net CO2

196

Chapter 9. Biological soil properties emissions than biochars produced at higher temperature. These results strongly support the view that biochar contains a labile fraction that is degraded on the short-term after application to soil. Fabbri et al. (2012) related the mineralization rates of 20 biochars from different feedstocks to their chemical composition, determined by pyrolysis coupled to gas chromatography-mass spectrometry. They found that biochars with higher concentrations of proteins and cellulose-derived pyrolysis products, particularly sugars, were associated with the largest mineralization rates. In contrast, biochars produced at higher temperature resulted in lower CO2 emissions (Fabbri et al., 2012). This is probably related to a degree of aromaticity and aromatic condensation increasing with temperature of pyrolysis (Keiluweit et al., 2010; Wiedemeier et al., 2015), which decreases the proportion of the labile fraction of biochar. Overall, the net increase in

CO2 release following application of biochar to soil appears to be a (very) short-term effect, observed only for experiments lasting less than 200 days.

For incubations of a duration > 200 days, the average emission of CO2 was not affected significantly after addition of biochar, or was even slightly decreased for large rates of application (Sagrilo et al., 2014). To explain this result, Sagrilo et al. (2014) proposed that a major part of the labile fraction of biochar might have been consumed over 200 days, or that a change in surface properties of biochar might have favored the sorption of CO2 on the surface of biochar. To our point, another possible explanation is that N deficiency can occur during incubation after application of biochar. In general, biochar has a high C:N ratio, and wood ash contains little N. Consequently, the mineralization of the N-poor labile fraction of charcoal can cause N immobilization (Ameloot et al., 2015). In their survey, Sagrilo et al. (2014) showed that soils with a C:N ratio < 10 were much more subject to an increase in CO2 emissions after addition of biochar than soils with a C:N ratio > 10. They also showed that soils with N fertilization background had much more emissions of CO2 after a large application of biochar than non-fertilized soils. Therefore, we could suggest from their data that N availability might limit mineralization rates. Biochar application to soil generally affects soil biota, which is closely linked to soil CO2 emissions that result mainly from aerobic respiration of soil microorganisms. Biochar can have contrasting effects on soil biology depending on initial soil properties, the quality and quantity of biochar and the group of microorganisms (Lehmann et al., 2011). The main drivers expected to affect soil microbial biomass, activity and community structure after soil

197

Chapter 9. Biological soil properties amendment with biochar are (i) an alteration of the source of organic C available to soil biota, (ii) a modification of the availability of nutrients (Blackwell et al., 2010; Steiner et al., 2009), (iii) a shift in pH (Rousk et al., 2010), (iv) a switch in soil physical properties, such as porosity, bulk density (Major et al., 2010) and tensile strength (Chan et al., 2007), (v) an interaction with organic compounds (Smernik, 2009) such as enzymes, toxins (Kasozi et al., 2010) or signaling compounds and (vi) the introduction of a porous material with a large specific surface area providing a support for microbial adhesion and a diversity of microhabitats (Lehmann et al., 2011). In return, a modification of soil biota might affect the stability of biochar by, among other mechanisms, co-metabolism with uncharred soil organic matter (Hamer et al., 2004; Kuzyakov et al., 2009). Despite the increasing amount of data available on the effect of biochar on soil biota and greenhouse gas emissions, crucial questions remain unanswered. Most data originate from short-term experiments in laboratory conditions (Sagrilo et al., 2014). Therefore, insights in the long-term effects of biochar are lacking (Maestrini et al., 2014; Sagrilo et al., 2014) because biochar persists in soil for centuries (Singh et al., 2012). Long-term implications of a biochar soil amendment are very likely to differ greatly from short-term effects (Joseph et al., 2010). For instance, on long timescales after addition of biochar to soil, a decrease of metabolic quotient (defined as microbial activity reported to soil biomass) or even a lower absolute amount of respired C was observed in black carbon (BC) rich terra preta soils (Jin, 2010; Liang et al., 2010). Nonetheless, it is difficult to extrapolate the effects observed in Amazonian terra preta to the presence of biochar only, because several types of organic and inorganic household wastes were involved in the genesis of terra preta (Glaser, 2007), and not only charcoal. Moreover, it can hardly be speculated that terra preta have been subject to a history of soil management similar to adjacent soils, because they have an improved productivity. In this chapter, we aimed to assess long-term implications of charcoal enrichment at pre-industrial charcoal kiln sites on microbial biomass, activity and community structure, and relate it to physico-chemical properties of soil, with a special focus on the relationship between soil biological properties and the contents of charcoal-C and uncharred SOC. In this goal, we selected charcoal kiln sites on Luvisols from the loess belt of Belgium previously characterized (Chapters 7, 8). Soil conditions may strongly interact with the impact of biochar on soil biological properties (e.g. Blackwell et al., 2010),

198

Chapter 9. Biological soil properties therefore, we selected sites from forest and from cropland to investigate the effect of charcoal enrichment under two contrasting land uses. Soils were incubated in controlled conditions and CO2 emissions were measured during 140 days. When emissions had reached a steady state, an aliquot of incubated soils was sampled to analyze microbial biomass and community structure based on phospholipid fatty acids (PLFA) biomarkers. Soil microbial properties were analyzed according to soil conditions and to the amount of charcoal-C and of uncharred SOC in soil, which were determined by differential scanning calorimetry (DSC).

9.2. Material and methods

Soil samples

We selected organo-mineral horizons of Haplic or Albic Luvisols (IUSS Working Group WRB, 2014) from the loess belt of Belgium. Selected sites include four sites from forest and five sites from cropland. According to the USDA classification, texture was defined as silt or silt loam (IUSS Working Group WRB, 2014). We chose sites on Luvisols because this soil type is the better represented in our database and because loess is probably the most homogeneous soil parent material at the scale of Wallonia. Thereby, we assume that land-use is the main factor that differentiated cropland from forest sites. Forest soils are all very acidic, even in presence of charcoal (pH around 4), which differs sharply from cropland soil that are frequently limed (pH around 7). For each site, organo-mineral (A) horizon of the charcoal kiln site was compared to the A horizon of the directly adjacent reference soil. It is important to note that A horizons of cropland soil samples have the same depth for the kiln and the reference soil, corresponding to the depth of the plow layer (0–25 cm), whereas A horizon of kiln sites from forest was sampled at a depth different from that of adjacent reference soil. This is because soils were sampled by horizon and that charcoal production largely disturbed the topsoil at kiln site, which increased the depth of the organo-mineral soil to 40–50 cm on average whereas that of adjacent soils is only 5–10 cm deep. This has important implications for interpreting the results, because natural SOM and most nutrients limiting plant growth accumulate in surface soil as a result of biological cycling (Jobbagy and Jackson, 2001).

Physico-chemical properties

Soil pH was measured in water (pH-H2O) and in a 1M KCl solution (pH-KCl) with a 1:5 soil:solution mass ratio. Elemental C and N contents were measured 199

Chapter 9. Biological soil properties by dry combustion (vario MAX, Elementar). The inorganic C content was measured by the modified-pressure calcimeter method on finely ground subsamples (< 200 µm) (Sherrod et al., 2002). Inorganic C content was null or below the detection limit (< 0.2 g kg-1). Therefore, total C was considered to correspond exclusively to organic C (OC) and includes charcoal-C. We measured the potential CEC by percolation of 1M ammonium acetate (naturally buffered at pH 7) on soil columns (Metson, 1956). Ammonium was desorbed with a 1.33 M KCl solution and measured by colorimetry (ISO7150/1). Available Ca, Mg, K, Na and P were extracted with a 0.5 M ammonium acetate–0.02 M EDTA solution at pH 4.65, with a 1:5 soil:solution mass ratio (Lakanen and Erviö, 1971), and measured by inductively coupled plasma-atomic emission spectroscopy (ICP-AES, Thermo). We calculated the base saturation of the soil as the ratio between the sum of exchangeable Ca2+, Mg2+, K+ and Na+ and the CEC.

Quantification of charcoal-C content

Differential scanning calorimetry (DSC) was used to determine the contents of charcoal-C and uncharred SOC in soil. The methodology of charcoal-C quantification is detailed in Chapter 5. Briefly, between 15 and 25 mg of soil ground to powder were scanned with a DSC 100 (TA Instruments) under a flow of 50 ml min-1 synthetic air from room temperature to 600 °C, at a heating rate of 10 °C min-1 (Leifeld, 2007). The fraction of charcoal-C content was determined based on the height of the three peaks attributed to the combustion of charcoal relative to the height of the peak attributed to the combustion of uncharred organic matter (Figures 5.9 and 5.10 and related text, Chapter 5). Previously to analysis, forest soils were buffered at pH 7 by equilibration with 1 M ammonium acetate (a solution naturally buffered at pH 7) and then 2+ saturated with Ca by agitation in a solution of 1M CaCl2. This pretreatment aimed to deprotonate most carboxylic acids present at the surface of charcoal, and to saturate carboxylate anions with Ca2+. As explained in detail in Chapter 5, the presence of Ca decreases the thermal stability of the O-rich fraction of charcoal, which enhances peak separation and thereby prevents peaks from overlapping, which would bias the quantification (Chapter 5). Agricultural soil samples were scanned without preliminary preparation because their pH- 2+ H2O was close to neutral, and because they are already saturated with Ca as they are limed frequently.

200

Chapter 9. Biological soil properties

Incubation experiment

For each sample, 120 g of dry soil sieved at 2 mm were weighed in steel cylinders whose bottom were closed by a porous nylon membrane, and were saturated with demineralized water. They were left in a pressure pan until they reached a pF of 2.5 (pF is a logarithmic expression of the suction component of the matric potential of soil, in cm), which corresponds to field capacity for an undisturbed soil. Equilibration lasted two weeks. Each rewetted soil was then split into three subsamples of 40 g (therefore we had triplicates) that were incubated in hermetic jars of 500 ml for 140 days in a climatic room, at a constant temperature of 20 ± 1 °C. To follow CO2 emissions over time, an open recipient with 25 ml of 0.5 M NaOH was placed in the center of each jar, to trap CO2 (one molecule of CO2 reacts with two molecules of NaOH to form

Na2CO3 and H2O). Electrical conductivity (EC) of the NaOH solution decreases linearly with the amount of CO2 that is consumed, therefore, EC was measured to determine the amount of CO2 trapped (Rodella and Saboya, 1999), which was calculated according to equation (1).

퐶1−퐶푥 퐶푂 푎푏푠표푟푏푒푑 = 푉 ∗ 푀 ∗ 44 ∗ 0.5 ∗ (1) 2 퐶1−퐶2 With:

- CO2 absorbed, the mass of CO2 that was trapped in the NaOH solution (mg) - V, the volume of the solution (ml) - M, the initial molarity of the NaOH solution (mol/l)

- 44, the molar mass of CO2 (g/mol)

- 0.5, the number of moles of CO2 reacting with one mole of NaOH - C1, the electrical conductivity of a 0.5 M NaOH solution at 20 °C (mol/l)

- C2, the electrical conductivity of a 0.25 M Na2CO3 solution at 20 °C (in case all NaOH has reacted) (mol/l) - Cx, the electrical conductivity of the solution to analyze (mol/l) Over the incubation period, electrical conductivity was measured after 3, 7, 12, 19, 26, 33, 40, 47, 54, 63, 70, 77, 92, 99, 117, 126 and 140 days, and the amount of CO2 emitted was calculated for each time step according to equation (1). For each measurement, the jar was open to measure the EC of the NaOH solution, which allowed the renewal of air. We calculated that O2 consumption between two measurements never exceeded 10 % of the total content of O2 in the jar, which guarantees that O2 was not deficient for

201

Chapter 9. Biological soil properties microbial respiration. We stopped the analysis at 140 days because the pattern of CO2 emissions of each soil had reached a constant rate. At day 70, we sampled an aliquot from each jar for PLFA analysis.

Microbial biomass and community structure

PLFA analysis is relatively expensive and time consuming. Therefore, aliquots from triplicates of the same soil were pooled together to limit the number of PLFA measurements to one for each soil. Directly after sampling from the incubation jars, soils were freeze-dried and stored at -80°C. During transport, freeze-dried samples were kept cold in dry ice. PLFA were extracted in the Department of Soil Management of Ghent University, according to the procedure described in detail by Sleutel et al. (2012), with the exception that we introduced a known amount of 1,2-dihenarachidoyl-sn-glycero-3- phosphocholine (C21:0 PC; Avanti Polar Lipids Inc. Alabaster, AL, USA), a PLFA standard absent from soil, to test whether the presence of charcoal decreases the PLFA extraction efficiency, as it was observed for fresh biochars (Gomez et al., 2014). We added to each sample 30 µg of C 21:0 PC before the start of the extraction (Gomez et al., 2014). We assume that charcoal interacts similarly with C21:0 PC and with PLFA naturally present in soil. Briefly, four grams of freeze-dried soil were mixed to 3.6 ml phosphate buffer (pH 7.0), 4 ml chloroform and 8 ml methanol. After centrifugation, phospholipids in solution in the supernatant were separated from neutral and glyco- lipids by sequential elution of chloroform and acetone on silica columns (Chromabond, Macherey-Nagel GmbH, Düren, Germany). Phospholipids were recuperated with methanol and saponified to obtain fatty acids. These were dried, dissolved in a methanol:toluene mixture and transformed into fatty acid methyl esters by methylation with 0.2 M methanolic KOH. The concentration of PLFA biomarkers was determined by gas chromatography-mass spectroscopy (GC-MS) with a Thermo Focus GC combined with a Thermo DSQ quadrupole MS (Interscience BVBA) in electron ionization mode. Soil PLFAs provide quantitative information on microbial biomass and community structure. The amount of Gram-positive (G+) bacteria was calculated as the sum of i15:0, a15:0, i16:0, i17:0 and a17:0. We considered that the sum of fatty acids 16:1u7c, 18:1u7c and cy17:0 corresponds to Gram-negative (G-) bacteria. To obtain the total bacterial community, PLFAs 15:0, 17:0 and cy19:0 were added to the sum of G+ and G- bacteria. The sum of 10Me16:0 and 10Me18:0 was used to estimate the proportion of actinomycetes. The fatty acid 18:2u6,9c was considered to be

202

Chapter 9. Biological soil properties representative of fungi, and 20:4u6,9,12,15c and 20:5u3,6,9,12,15 of protozoa (Sleutel et al., 2012). We also used fatty acids 18:1w9c, 18:2c9,1 and 18:3c9,12,15 as indicators of fungi 18:1, fungi 18:2 and fungi 18:3, respectively.

Statistics and data analysis 9.2.6.1. Soil properties

Data were analyzed by analysis of variance (ANOVA) to test the effect of land use (forest or cropland), kiln site (kiln or reference soil) and the interaction ‘land use x kiln site’ on soil properties. Data were analyzed in R 3.3.1 (R Core Team, 2012).

9.2.6.2. CO2 mineralization rates

A double exponential model generally fits well experimental data of carbon loss from a soil incubated in controlled conditions, and was often used to model CO2 fluxes from biochar, or from a biochar amended soil (e.g. Cheng et al., 2008; Hilscher and Knicker, 2011; Singh et al., 2012). Such model assumes the presence of two pools of carbon in soil, one fast-cycling and one slow-cycling pool. Fitting the model to experimental data allows determining the size and the decay rate of the two pools. Double exponential model can be written according to equation (2): 푆푡 = 푋1 푒푥푝−푘1푡 + X2 푒푥푝−푘2푡 (2) With : - St, the OC stock (g kg-1) at time t (yr) of incubation - X1, the fast-cycling pool of OC (g kg-1) - K1, the decay rate of X1 (yr-1) - X2, the slow-cycling pool of OC (g kg-1) - K2, the decay rate of X2 (yr-1) Carbon loss over time of incubation was calculated by difference between the initial SOC content and cumulated amount of CO2 emitted. Data were fitted in Matlab R2016a (®MathWorks) with a two pools model described by equation (2). Estimate of the mean residence time (MRT) of both fast-cycling and one slow-cycling pools is provided by 1/K (the reverse of the decay rate).

9.2.6.3. Principal components analysis

203

Chapter 9. Biological soil properties

To examine how the composition of the microbial community varies within the dataset, PLFA biomarkers were analyzed by principal component analysis (PCA) in R 3.3.1 (R Core Team, 2012). To highlight the effect of land use (forest or cropland) and kiln site (kiln or reference soil) on microbial community structure, these were plotted on the map of individuals as qualitative supplementary variables. We also used pH-H2O (proxy for land use) and charcoal-C content (proxy for kiln site) as quantitative supplementary variables, plotted on the map of variables.

9.3. Results

Soil properties

We tested the effect of land use, kiln site and interaction ‘land use x kiln site’ on soil physico-chemical properties by ANOVA (Table 9.1). Land-use affects significantly (P < 0.05) all soil properties that were investigated except exchangeable Mg2+. Cropland soils contain less total OC, charcoal-C and uncharred SOC than forest soils. They also have a smaller C:N ratio and CEC. They have acidity close to neutral, which contrasts with forest soils that are very acidic. Accordingly, cropland soils are saturated with exchangeable ‘base’ cations (Ca2+, Mg2+, K+) whereas forest soils are very desaturated. The presence of kiln site significantly affects the contents of total OC, charcoal-C and uncharred SOC, the C:N ratio and the CEC. A significant interaction with land use was obtained for charcoal-C and uncharred SOC contents, the C:N ratio and the concentration of available P, which indicates a contrasting effect of kiln site depending on land use.

204

Table 9.1. Soil properties of kiln (K) and reference (R) soils from cropland and forest. Analysis of variance (ANOVA) was performed on the dataset to test the effect of land use (LU), kiln site (K/R) and their interaction (LU x K/R). P-values <0.05 are highlighted in grey.

205

Chapter 9. Biological soil properties

Regardless of land use, the contents of total OC and charcoal-C are larger at kiln site than in adjacent soils. However, the content of charcoal-C is much larger in forest. The C:N ratio and the CEC are systematically increased as well. In forest, uncharred SOC content is smaller in kiln soil than in reference soil, whereas it is slightly larger at kiln site in cropland. The effect of kiln site on the concentration of available P also strongly depends on land use. In cropland, kiln and reference soils have a similar concentration whereas in forest soils, the concentration of available P is strongly reduced at kiln site.

Mineralization rates

Figure 9.1. Mean cumulative emissions of CO2 over time from incubated kiln (black symbols) and reference (grey symbols) soils. Error bars correspond to one standard deviation. a), b), c), d) and e) are cropland sites; f), g), h) and i) are forest sites.

In absolute terms, reference soils from cropland emit less CO2 than reference soils from forest (Figure 9.1). Nevertheless, they emit about twice more CO2 per unit of total OC than forest soils (Figure 9.2). In cropland, the absolute amount of CO2 emitted from kiln soil is quite similar to that of adjacent reference soils (Figure 9.1a to e). In contrast, CO2 emissions from forest kiln soils are much smaller than that from adjacent reference soils (Figure 9.1f to 206

Chapter 9. Biological soil properties

i). If CO2 emissions are expressed per unit of C, they are systematically smaller in kiln soil than in adjacent reference soil, regardless of land-use (Figure 9.2).

Figure 9.2. Mean cumulative emissions of CO2 per unit of SOC (including charcoal- C) over time from incubated kiln (black symbols) and reference (grey symbols) soils. Error bars correspond to one standard deviation. a), b), c), d) and e) are cropland sites; f), g), h) and i) are forest sites.

Data of CO2 emissions were fitted with double exponential models, which consider the contribution of two pools of C to CO2 emissions, one fast-cycling and one slow-cycling (Table 9.2). The models fitted the data particularly well (R² > 0.999). We fitted C loss from the total OC content, including uncharred SOC and charcoal-C. In cropland, the model estimated a fast-cycling pool of C having a similar size and half-life in the kiln and the reference soil (Table 9.2). In contrast, the size of the fast-cycling pool in forest is about twice smaller in the kiln soil than in the reference soil, but they have a comparable turnover (Table 9.2). The size of the slow-cycling pool corresponds to the difference between total OC content that was initially present in soil and C content of the fast-cycling pool, which represents a major part of total OC content. In cropland, the slow-cycling pool of the kiln soil has a half-life of 32

207

Chapter 9. Biological soil properties years on average, which is more than twice larger than the estimated half-life of that of reference soils (Table 9.2). In forest, half-life of the slow-cycling pool of kiln soils is 88.4 years, which is about twice that of adjacent reference soil.

Table 9.2. Parameters of the double exponential models fitted to CO2 emissions from kiln (K) and reference (R) soils from cropland and forest during the 140 days incubation experiment. A first model was fitted to data of C loss from the total initial organic carbon pool, and a second model was fitted to data of C loss from the uncharred organic carbon pool only, assuming that the contribution of charcoal-C to CO2 emissions is negligible. Analysis of variance (ANOVA) was performed on the dataset to test the effect of land use (LU), kiln site (K/R) and their interaction (LU x K/R).

Microbial biomass and community structure

Recovery from C21:0 PC from the soil was low (a few percent only, data not shown), which suggests that the digestion of this standard PLFA during extraction was largely incomplete. Nevertheless, no difference in recovery

208

Chapter 9. Biological soil properties was observed between kiln and reference soil, which supports the idea that the presence of charcoal did not interfere with the recovery of PLFA naturally present in soil. The content of total PLFA in soil is strongly affected by land-use, kiln site and the interaction ‘land-use x kiln site’ (Table 9.3). Overall, the content of PLFA in cropland is low relative to that of forest soils. In cropland, no significant difference between kiln and reference soils was recorded, whereas the content of PLFA in forest kiln soil is much smaller than that of adjacent reference soils. This contrasting effect of kiln site depending on land use explains the significant interaction ‘land use x kiln site’. Community structure is also strongly affected by land use. Excepted fungi 18:1, fungi 18:2 and protozoa that contribute identically to soil microbial biomass in forest and cropland, all PLFA biomarkers (expressed as a percentage of total PLFA) are significantly affected by land use (Table 9.3). The content of gram positive (G+) bacteria, the content of actinomycetes and the ratio bacteria:fungi (B:F) are larger in forest soil, whereas the content of gram negative bacteria (G-), fungi 18:3 and AMF are larger in cropland. In contrast to land use, kiln site has a limited effect on community structure. The only biomarkers that might be slightly affected by kiln site are actinomycetes and fungi 18:1, unless the P-values are > 0.05. The distribution of the proportion of PLFA biomarkers between the different modalities was also illustrated by a principal component analysis (Figure 9.3), Figure 9.4). The first component of the PCA explains 60.52 % of total variance in the dataset and discriminates well between forest and cropland soils (Figure

9.3a, b). Accordingly, the pH-H2O is positively correlated to this first component (Figure 9.3a, b). The second axis explains 16.66 % of total variance, and the third component explains 11.70 %. The third component is the one that discriminates the best between kiln and reference soils (Figure 9.3b). Consistently, the content of charcoal-C is negatively correlated to the third component (Figure 9.3b). The first and second components do not seem to depend on the presence of kiln site at all (Figure 9.3a).

209

Table 9.3. Soil properties of kiln (K) and reference (R) soils from cropland and forest. Analysis of variance (ANOVA) was performed on the dataset to test the effect of land use (LU), kiln site (K/R) and their interaction (LU x K/R) on the data.

210

Chapter 9. Biological soil properties

Figure 9.3 Maps of variables of principal component analysis of phospholipid fatty acids (PLFA) biomarkers, expressed in percent of total PLFA content. pH and charcoal-C content were plotted as quantitative supplementary variables as proxies for land-use and kiln site, respectively. a) Second (Dim. 2) against first (Dim. 1) dimension; b) Third (Dim. 3) against first (Dim. 1) dimension.

211

Chapter 9. Biological soil properties

Figure 9.4. Maps of individuals of principal component analysis of phospholipid fatty acids (PLFA) biomarkers, expressed in percent of total PLFA content. Land-use (forest, F and cropland, C) and kiln site (Kiln, K and reference, R) were used as qualitative supplementary variables. a) Second (Dim. 2) against first (Dim. 1) dimension; b) Third (Dim. 3) against first (Dim. 1) dimension. On the map of individuals of the third principal component plotted against the first principal component (Figure 9.4b), it appears that some distance exists between microbial community structure of kiln and reference soils from forest in the third dimension of the PCA (all kiln soils have more negative values than reference soils in the third dimension), but distance between kiln and reference soil disappears in cropland. 212

Chapter 9. Biological soil properties

The relationship between daily emissions of CO2 and PLFA content per unit of C at day 70 of the incubation was analyzed, which provides information on the microbial metabolic quotient of soil (microbial activity reported to soil biomass). Interestingly, respiration rate appears to be linearly related to the content of PLFA in soil (Figure 9.5), particularly in cropland (r = 0.97) but also in forest. Overall, soils from cropland have more PLFA per unit of C than soils from forest, which accords with larger emissions of CO2 per unit of C. If we fit separately data from forest and cropland with a linear regression, the slope of the best fitting line is less steep for forest soils, which suggests that the metabolic quotient of microbial biomass is slightly lower in forest (P < 0.05).

Figure 9.5. Daily release of CO2 per unit of carbon against the content of PLFA per unit of carbon at day 70 of incubation for kiln (K, black symbols) and reference (R, grey symbols) soils from cropland (circles) and forest (triangles). Regression lines refer to the two different land uses. We also found a close relationship between the amount of PLFA per unit of carbon and the C:N ratio of SOM (Figure 9.6), which is an indicator of microbial availability of plant residues and SOC. The graph highlights that forest soils have a larger C:N ratio on average than soils from cropland, which corroborates with the smaller content of PLFA per unit of OC. We also showed earlier that the C:N ratio was strongly increased in presence of charcoal, regardless of land use.

213

Chapter 9. Biological soil properties

Figure 9.6. Total PLFA content per unit of carbon against C:N ratio for kiln (K, black symbols) and reference (R, grey symbols) soils from cropland (circles) and forest (triangles).

9.4. Discussion

The effect of charcoal on CO2 mineralization, soil microbial abundance and activity

The small rate of respiration per unit of total OC from kiln soils (Figure 9.2) highlights the resistance of charcoal to biotic decomposition. A similar decrease of the mineralization potential of SOC was observed in soils historically enriched with BC, and the poor degradability of BC was attributed to its aromatic structure (Cheng et al., 2008c; Liang et al., 2010, 2008). The poor availability of polycyclic aromatic C for soil biota has been attributed to the large activation energy required for the decomposition of C=C bonds (Plante et al., 2009). In agreement with the slow decomposition of aromatic moieties from charcoal, a decrease of the turnover of the slow-cycling pool of the double exponential model was recorded. Estimates of the residence time of the slow-cycling pool were about twice larger on average than that of adjacent reference soils. Despite being longer than that of reference soils, average half-lives of 32.0±13.8 and 88.4±10.5 years for the stable pool of SOC of charcoal-rich kiln soil seem to be largely underestimated with respect to the age of the sites (> 150 years). The main cause of underestimation of the turnover of slow- cycling pool of OC in kiln soil is most probably the inadequacy of the double exponential model approach to capture the turnover of charcoal-C, because emissions of CO2 from charcoal are too small on the timescale of the experiment (Cheng et al., 2008c; Kuzyakov et al., 2009; Liang et al., 2008). One the one hand, the size of the fast-cycling pool of C as estimated by the

214

Chapter 9. Biological soil properties model is related to the presence of very labile microbial biomass that is consumed in the very short-term when a dry soil is rewetted, which explains the boom of CO2 emission that is generally observed in the first weeks of an incubation experiment. In our study, we probably missed most this boom during the step of equilibration of soil humidity at pF 2.5, which lasted two weeks. On the other hand, the decay of the slow-cycling pool as defined by the double exponential is probably mainly related to the decomposition of more stable uncharred SOM. It is likely that charcoal contributes to increase the size of the pool without contributing significantly to CO2 emissions, which hinders the correct estimation of the turnover of both uncharred SOC (that is probably faster than that of the slow-cycling pool) and charcoal-C (that is probably slower than that of the slow-cycling pool). To support the point that the quality of the fit of the double exponential model does not imply a correct estimate of the turnover of charcoal and uncharred SOC in soil, data were fitted with a polynomial function of second order. The model resulted in coefficients of determination as high as that of the double exponential model (R² > 0.99) even though coefficients of such model have absolutely no meaning in term of carbon pool. This discrepancy is in agreement with the presence of a stability continuum in charcoal that needs to be explained by at least three pools of contrasting reactivity (labile non aromatic C, semi-labile aromatic C and stable polycyclic aromatic C; Bird et al., 2015). It is clear that the inclusion of a third and fourth pools of C to predict their residence time would over-parameterize the model and lead to poorly constrained predictions of long-term decay rates for the less reactive pools (Bird et al., 2015).

A research perspective to tackle the question of the origin of CO2 emissions 14 (uncharred SOM vs charcoal) is to measure the C signature of CO2 emitted, based on the assumption that the 14C/12C ratio would be decreased if > 150 years charcoal contributes significantly to CO2 emissions. Nevertheless, the isotopic signature of CO2 was not investigated in the present study. To overcome the lack of isotopic data to trace the origin of CO2 in our incubation experiment, we have tried to express the emissions of CO2 per unit of uncharred SOC, making the assumption that if charcoal contributes to CO2 emissions from the kiln soil or interacts with the decomposition of uncharred

SOM, the rate of CO2 emissions per unit of uncharred SOC will be modified compared to adjacent reference soils. As a result, no systematic effect of charcoal on CO2 emissions per unit of uncharred SOC was recorded, with values sometimes similar, sometimes lower and sometimes higher in the kiln soil than in adjacent reference soil. CO2 emissions per unit of uncharred SOC

215

Chapter 9. Biological soil properties were then related to soil properties. As a result, we found a strong correlation with soil pH-H2O (Figure 9.7). The best correlation was obtained for the day 7 of incubation (r = 0.94), but correlation coefficient remained strong (r ≥ 0.91) regardless of the time of incubation. This suggests that the content of uncharred SOC and soil acidity explain a main part of the variance of CO2 emissions for the soils of this study and supports the view that charcoal-C was a negligible source of CO2 from soil in our incubation experiment. One important implication of this finding is the fact that on the long-term, when the labile fraction of charcoal is completely decomposed, the presence of charcoal mainly affects CO2 emissions indirectly by modifying the drivers of microbial respiration, such as acidity and, possibly, the availability of nutrients. These results support the idea the “semi-labile”, poorly condensed aromatic C that constitutes the main fraction of low temperature chars (Bird et al., 2015) has a much slower turnover in soil than uncharred SOC, which is encouraging for the potential of low temperature biochar to sequester C in soil on the mid- to long-term.

Figure 9.7. Total emissions of CO2 per unit of uncharred SOC after 7 days of incubation against soil pH measured in water (pH-H2O) for kiln (K, black symbols) and reference (R, grey symbols) soils from cropland (circles) and forest (triangles). The positive effect of an increase in pH on the rate of decomposition of SOC was reported in previous studies (Aciego Pietri and Brookes, 2008; Waschkies and Hüttl, 1999). Consistently, cropland soils of this study (that have pH-H2O of about 6–8.5) emit more CO2 per unit of C than forest soils (that have pH-

H2O of about 3.5–5), which highlights the importance of trophic conditions on soil respiration. An increase in soil pH was also shown to increase bacterial abundance in an arable soil, whereas fungal abundance was not affected by a change of pH (Rousk et al., 2010). As a close link exists between total microbial biomass and respiration rate (Figure 9.5), the increase in CO2

216

Chapter 9. Biological soil properties

emissions with pH-H2O is most likely related to a larger amount of total microbial biomass per unit of uncharred SOC. The slightly lower metabolic quotient in forest soils compared to cropland soil (Figure 9.5) suggests that acidity might negatively affect microbial activity as well as microbial abundance. Beside the positive effect of a pH increase on microbial respiration, Kerré et al. (2017) measured a smaller mineralization of fresh SOM (traced by 13C isotopic signature) when added to the soil of a pre-industrial charcoal kiln sites. They related it to an increase of the sorption of dissolved organic carbon (DOC) that they measured in another experiment, with a preferential adsorption of the DOC rich in aromatics. This result suggests that beyond the role of pH, direct interactions of charcoal with the labile, soluble fraction of SOM can influence the mineralization kinetics of SOM. This might contribute to explain the enrichment of uncharred SOC that was measured in the topsoil of pre-industrial charcoal kiln site in cropland (Chapter 8).

The effect of charcoal on microbial community structure

Charcoal did not affect the recovery of PLFA from soil, which was verified by the introduction of a known amount of a standard PLFA (C 21:0 PC) at the start of the extraction. Even though the recovery of this standard PLFA was low, the close correspondence between total PLFA content in soil and CO2 emissions regardless of charcoal enrichment (Figure 9.5) further supports the view that charcoal didn’t interfere with the recovery of PLFA from soil microbial biomass. This result contrasts with observations of Gomez et al. (2014), who observed a strong decrease in the recovery of PLFA with increasing additions of biochar. We explain this discrepancy by a decrease of hydrophobicity of charcoal (and associated affinity for organic molecules) by oxidation over time (Criscuoli et al., 2014). Moreover, most sites at the surface of charcoal reactive towards dissolved organic molecules might have been saturated over long periods of time in an organo-mineral soil. Therefore, we attribute the absence of interference of aged charcoal with the recovery of PLFA to modifications of surface properties over time. We have discussed earlier that the total content of PLFA in soil explains a main part of the variance in CO2 emissions. Accordingly, both total PLFA content and mineralization rates are strongly influenced by land use. Forest soils contain more microbial biomass and emit more CO2 than cropland soils, but we also showed that microbial abundance per unit of SOC was smaller. The PCA on microbial biomarkers also highlighted that land use rules a large 217

Chapter 9. Biological soil properties part of the variance of the microbial community structure (Figure 9.3). The content of gram positive (G+) bacteria, the content of actinomycetes and the ratio bacteria:fungi (B:F) are larger in forest soil, whereas the content of gram negative bacteria (G-), fungi 18:3 and AMF are larger in cropland. Soil pH, which was shown to have a strong influence on the kinetics of decomposition of uncharred SOM by promoting microbial growth (Aciego Pietri and Brookes, 2008; Waschkies and Hüttl, 1999), could also explain a major part of the variability of the community structure, with a major influence on bacterial biomass and diversity, more sensible to acidity than fungi whose growth is optimal on a wider range of pH (Rousk et al., 2010). Another factor than can be related to the switch in community structure induced by a land use change from forest to cropland is the change in organic C and N inputs due to a shift from forest plant species to field crops (Kardol and Wardle, 2010). Tillage is also suspected to affect the B:F ratio in soil, unless there is no clear evidence of it (Strickland and Rousk, 2010), whereas the abundance of fungi was shown to relate mainly on soil moisture (Frey et al., 1999). In contrast to land use, charcoal-enrichment at kiln site has a limited effect on soil microbial community structure. Nevertheless, the third principal component of the PCA was negatively correlated to charcoal-C content and discriminated quite clearly between kiln and reference soils from forest (Figure 9.4b). This seems to be related to a low proportion of fungi 18:1 in kiln soil, which is negatively correlated to charcoal-C content. Nevertheless, differences on microbial community structure between kiln and reference soils attenuate in cropland. This smaller effect of charcoal enrichment might be caused by the use of fertilizers and liming amendments that standardize soil conditions between kiln and reference soil. Therefore, the residual effect of charcoal accumulation on pH and available nutrients is erased, which attenuates the effect of charcoal on microbial community structure. In agreement with this assumption, the increase in microbial reproduction rate and root colonization by AMF caused by biochar introduction to soil was reduced when mineral fertilizers were applied (Blackwell et al., 2010; Steiner et al., 2009). This supports the idea that the long-term effect of charcoal on biomass community structure mainly relies on indirect effect on soil trophic conditions (pH, availability of nutrients) rather than on a modification of C input for microbes, given the high resistance to oxidation of the most stable fraction of charcoal that survives to oxidation on centennial or millennial timescales.

218

Chapter 9. Biological soil properties

Another factor that might promote the contrast between microbial community structure of kiln and reference soils in forest is the difference in sampling depth. Indeed, forest soils were sampled by horizon. Charcoal production largely increased the depth of the topsoil horizon, from 5–10 cm deep in natural conditions to about 45 cm deep at kiln site. This explains for the low content of uncharred SOC in the soil of the forest kiln sites, as natural SOM that accumulates in the topsoil has been diluted into a thicker soil layer. Depth of sampling can also sharply affect the availability of nutrients limiting microbial growth (like P and K), because most of them accumulate in surface soil as a result of biological cycling (Jobbagy and Jackson, 2001). This difference in sampling depth might have exaggerated the differences between microbial community structure of kiln and reference soil from forest, even though they are small. It is also worth to mention that our analysis of microbial abundance and community structure was limited to the main groups of microorganisms, using PLFA biomarkers as proxies of these groups. We might have revealed more differences between kiln and reference soil by a more detailed taxonomic analysis. Detailed microbial analyzes of terra preta soils have highlighted an important shift in microbial taxonomy from adjacent soils (Grossman et al., 2010; Jin, 2010; Kim et al., 2007), with BC enrichment identified as the main source of variance (Grossman et al., 2010). Nevertheless, soil nutrient availability was sharply modified in terra preta soils by various organic and inorganic inputs other than charcoal, which might have had a much stronger influence on microbial community structure than BC itself.

The pH-H2O was found to be strongly related to the decomposition rate of uncharred SOM, regardless of land use and the presence of charcoal. It is also interesting to note that the C:N ratio is a good indicator of the total microbial PLFA content per unit of carbon, regardless of soil conditions (Figure 9.6). We attribute this to the fact that C:N ratio of soil provides information on the overall stability of soil organic matter. Consequently, this index integrates at the same time the proportion of uncharred SOC and charcoal in soil, and the kinetics of decomposition of the fraction available to microbial decomposition, depending on soil drivers such as acidity and availability of nutrients.

219

Chapter 9. Biological soil properties

9.5. Conclusion

In this study, we related microbial properties of soils historically enriched with charcoal to the content of charcoal-C and uncharred SOC and to physico- chemical properties of soil. Our results highlighted that double exponential models are inadequate to capture the long-term dynamics of charcoal-C in soil at the time scale of our experiment probably because aged charcoal is a negligible source of CO2 emissions from soil. The model provided an estimation of the residence time of the stable pool of SOC that is unrealistic for charcoal. The content of uncharred SOC and pH-H2O explained a large part of the variability of CO2 emissions rates, regardless of the presence of charcoal. We also found a strong correlation between the content of total

PLFA and CO2 emissions, which supports the view that an increase in decomposition is mainly related to a larger microbial abundance rather than an increase of microbial activity. Overall, land use explains a large part of the variance in the microbial community structure. Some differences between kiln and reference soils were detected in forest, but they disappeared in cropland. Our results support the idea that on the long-term, when the labile fraction of charcoal has been completely degraded, the effect of charcoal on CO2 emissions is mainly related to a modification of drivers of soil respiration such as acidity and availability of nutrients. Soil conditions modified in cropland due to the application of organic and inorganic fertilizers, which attenuates the residual effect of charcoal accumulation on soil fertility and therefore its influence on microbial community structure.

Acknowledgements

We thank Jens Leifeld from the Institute for Sustainability Sciences of the Agroscope of Zürich (Switzerland) for his help for DSC analysis. The PLFAs were extracted in the Department of Soil Management of Ghent University, with the help of Steven Sleutel. Funds were provided by the General Directorate for Agriculture, Natural Resources and Environment – Public Service of Wallonia the FSR (Fonds Spéciaux de Recherche) of the Université catholique de Louvain.

220

Chapter 10. Properties of charcoal particles

Chapter 10. Long-term changes of chemical properties of preindustrial charcoal particles aged in forest and agricultural temperate soil.10

Summary

Black carbon (BC) plays an important role in terrestrial carbon storage. Nevertheless, the effect of cultivation on long term dynamics of BC in soil has been poorly addressed. To fill this gap, we studied the chemical properties of charcoal particles extracted from preindustrial kilns in Wallonia, Belgium, along a chronosequence of land use change from forest to agricultural soil, up to 200 years of cultivation. Preindustrial charcoal samples were compared with charcoal subjected to limited ageing in a currently active kiln. Cultivation increased association of charcoal with soil minerals, which is favored by deprotonation of carboxylic acids under liming, thereby enhancing the reactivity of charcoal towards mineral surfaces. The large specific surface area of charcoal, related to its porosity, promotes the precipitation of 2:1 phyllosilicates and CaCO3. Both ageing and cultivation decreased the resistance of charcoal to dichromate oxidation, related to an increase of the H/C of charcoal. Differential scanning calorimetry revealed the presence of three fractions of distinct thermal resistance. Saturation of carboxylate groups with Ca2+ under liming decreased thermal resistance of the O-rich, less thermally stable fraction of charcoal. This fraction decreased over time of cultivation, leading to the relative increase of the thermally most stable fraction of charcoal. This might result from the preferential loss of the O-rich fraction or the slowdown of charcoal from oxidation by association with minerals. Our results highlighted that land use significantly affects the properties of BC through the modification of soil conditions, which might influence the kinetics of BC loss from soil.

10 Hardy, B., Leifeld, J., Knicker, H., Dufey, J.E., Deforce, K., Cornélis, J.-T. Long- term changes of chemical properties of preindustrial charcoal particles aged in forest and agricultural temperate soil. Organic Geochemistry. In press. doi: 10.1016/j.orggeochem.2017.02.008 221

Chapter 10. Properties of charcoal particles

10.1. Introduction

Black carbon (BC), the solid residue of thermal decomposition of biomass, is ubiquitous in soil and sediments (Schmidt and Noack, 2000a). According to recent estimates, soil stores ~200 Pg of BC in the uppermost two meters (Reisser et al., 2016), which corresponds to ~10 % of global soil organic carbon (SOC) stocks calculated by Batjes (2016). BC promotes the long- lasting fertility of Amazonian dark earths (Glaser and Birk, 2012) and contributes to long-term soil carbon (C) storage because it is more resistant to degradation than uncharred soil organic matter (SOM). Hence, soil amendment with biochar (BC produced intentionally to be applied to soil) is increasingly being regarded as a feasible alternative to tackling greenhouse gas emissions by increasing soil carbon sequestration, while enhancing sustainably soil fertility (Lehmann, 2007b). Despite the general agreement that BC lasts for longer in soil than uncharred organic matter, the stability and persistence of BC in soil is still debated. On the one hand, the presence of BC in geological records since the Devonian and of millennial BC in a range of soils at global scale provides evidence that some BC persists for a very long time in the environment (Schmidt and Noack, 2000a). On the other hand, the content of BC stored in soil is low regarding annual production rates from wildfires, which demonstrates that large amounts of BC are lost from soil (Masiello, 2004; Schmidt, 2004; Schmidt and Noack, 2000a), possibly by microbial decomposition (Baldock and Smernik, 2002; Hamer et al., 2004; Wengel et al., 2006), erosion (Rumpel et al., 2006) or dissolution and transport with water fluxes (Hockaday et al., 2007; Jaffé et al., 2013). The longevity of BC in soil seems to depend on both intrinsic quality and environmental conditions where it is deposited (Bird et al., 2015). Feedstock and conditions of production control the degradation potential of BC. Maximum temperature of pyrolysis was shown to play a major role, as the degree of condensation of aromatic clusters responsible for the stability of BC (Bird et al., 2015) increases with temperature (Keiluweit et al., 2010; Wiedemeier et al., 2015). Once deposited in the environment, BC is subject to a range of reactions. Ageing of BC mainly consists of oxidation of exposed C rings with a high density of π electrons and free radicals (Joseph et al., 2010), which creates a high density of O-rich functional groups at the surface of BC. Oxidation starts at the surface of BC and propagates to the core of particles over time (Lehmann et al., 2005), promoting further physical, chemical and microbial

222

Chapter 10. Properties of charcoal particles degradation (Hammes and Schmidt, 2009). The drivers of oxidation can be biotic (Hamer et al., 2004; Wengel et al., 2006) but are mainly abiotic, at least shortly after introduction to soil (Cheng et al., 2006). The adsorption of dissolved organic molecules can contribute to the increase of O content at the surface of BC (Lehmann et al., 2005). Climatic factors such as water regime (Nguyen and Lehmann, 2009) and temperature were shown to influence the kinetics of alteration and decomposition of BC (Cheng et al., 2008c; Nguyen et al., 2010). Soil conditions are also expected to have an influence. For instance, preferential accumulations of BC were observed in clay-rich soils (Reisser et al., 2016), which suggests that association with minerals might be an important mechanism of stabilization of BC in soil (Czimczik and Masiello, 2007). In the last decade, much effort was made to characterize the short term effects of biochar of various types after addition to soil, to design biochar with a high stability and favorable agronomic properties (Abiven et al., 2014). Nevertheless, properties are known to change over time (Joseph et al., 2010) and mid-term to long term dynamics in soil have been poorly addressed due to the lack of long term experiments. Amazonian dark earths, which triggered interest in biochar, is probably the best documented long term case study of BC in soil. Nevertheless, results from terra preta are difficult to extrapolate to other soil, climatic and agronomic contexts. In particular, the dynamics of BC ageing under intensive cultivation have been disregarded. As soil amendment with biochar targets mainly agricultural soils, a better assessment of the effect of cultivation on BC properties is crucial for predicting the long term dynamics of biochar and BC in soil. Recently, Hardy et al. (2017) characterized the effect of preindustrial charcoal kiln sites on the chemical properties of agricultural soils from Wallonia, Belgium. On bare soil, these sites appear as black spots a few decameters in diameter, with the topsoil largely enriched with charcoal residues Hardy et al. (2017). Traditional charcoal production occurred in the forest, but part of the land was cleared for cultivation from the early 19th century. The presence of charcoal kiln sites in agricultural soils provide the opportunity to investigate how cultivation affects the dynamics of charcoal ageing, by comparison with sites from the same episode of charcoal production that remained forested since the time of charcoal production. Nguyen et al. (2008) studied the evolution of properties of BC particles over time in agricultural soils in Kenya cleared from forest by fire. They highlighted that association of charcoal to soil minerals increases over time. 223

Chapter 10. Properties of charcoal particles

They also showed that the content of BC in soil decreased sharply during the first 20 years after clearing but remained similar by after. As the decrease in BC loss correlated with the increase of association with soil minerals, they suggested that organo-mineral association might be an important mechanism in the stabilization of BC. Nevertheless, Nguyen et al. (2008) studied BC dynamics over time in agricultural soil, which does not allow distinguishing between the effect of time and of cultivation on BC degradation. By selecting kiln sites abandoned at the same time and converted to agricultural land at different times, we aimed to discriminate between the effect of time and of cultivation on the organic and inorganic composition and stability of charcoal particles aged in soil. We assumed that cultivation might accelerate the physical, chemical and biological weathering of charcoal by way of the mechanical action of tillage and the improved soil fertility related to liming and the use of organic and inorganic fertilizers. Properties of preindustrial charcoals were also compared to that of charcoal from a currently active kiln site, used as a reference with limited ageing.

10.2. Material and methods

Charcoals

Sampling, separation from soil and identification of wood species of charcoal pieces of the series of charcoals that was analyzed in this chapter are detailed in Chapter 4. The chronosequence included charcoals from the currently active kiln site in Dole, and pre-industrial charcoals from kiln sites of Wallonia from a chronosequence of land use change from forest to cropland, up to 200 years of cropping. Assemblages of charcoals were dominated by oak (Quercus sp.) and hornbeam (Carpinus betulus L.). One limitation of this uncontrolled experiment is the fact that we can only speculate that macroparticles of charcoal that remain in soil after > 150 years are representative of the long-term fate of charcoal residues that were initially left after charcoal production. For example, some types of charcoal remains from softwood species (e.g. Betula sp.), possibly more easily degradable than hardwood species (e.g. Carpinus betulus, Fagus sylvatica) might have been lost preferentially, responsible for the relative increase of less degradable particles. Nevertheless, the composition of the assemblages identified from microscopic observation of charcoal pieces (Chapter 4) seems to be representative of the natural forest growing in the study areas, which supports the view that the comparison of charcoal particles from pre-industrial kiln sites

224

Chapter 10. Properties of charcoal particles and from the site of a currently active kiln is reliable to address the long-term changes of charcoal particles in soil. Previously to chemical analyses, charcoal particles were ground to a powder with an agate pestle and mortar.

Elemental composition, loss on ignition

Elemental composition (C, H, N and O) of charcoal powder was determined with a Euro EA elemental analyzer (HEKAtech). C, H and N were measured via dry combustion and O was analyzed using pyrolysis at 1000 °C. Whereas N and H derive exclusively from OM, C and O originate from OM and carbonate. To differentiate between organic and inorganic C and O, carbonate content was determined by loss on ignition between 550 and 1000 °C; ca. 100 mg charcoal powder was weighed and dried at 105 °C. It was then heated overnight at 550 °C to oxidize all the OM, and heated again overnight at 1000 °C to remove carbonate. OM (Δ550–105 °C) and carbonate (Δ1000–550 °C) were determined gravimetrically. Inorganic C and O from carbonate were calculated on the basis of atomic weight. Organic C and O were calculated from the difference between elemental and inorganic content. OM content, estimated by summing organic C, H, N and O content (namely after retrieving the contribution of carbonate), was consistent with OM content estimated from loss on ignition (Δ550–105 °C) (Figure 10.1).

225

Chapter 10. Properties of charcoal particles

Figure 10.1. a) Organic matter (OM) content in charcoal estimated by loss on ignition (105–550°C) against OM content estimated by elemental analysis (C+O+H+N); b) Organic matter (OM) content in charcoal estimated by loss on ignition (105–550°C) against OM content estimated by elemental analysis after removing the contribution of inorganic C and O from carbonates (Corg+Oorg+H+N).

XPS

To quantify total atomic content of major elements (C, O, N, Si, Al, Fe and Ca), charcoal powder was analyzed with a SSX 100/206 photoelectron spectrometer (Surface Science Instruments) equipped with a monochromatized micro focused Al X-ray source powered at 20 mA and 10 kV. Powder was fixed on a stainless steel multi-specimen holder with double sided insulating tape. The analysis chamber was around 10-6 Pa, and the angle between the surface normal and the axis of the analyzer lens was 55°. The pass energy was 150 eV and the area analyzed was ca. 1.4 mm2. In these conditions, the full width at half maximum (FWHM) of the Au 4f7/2 peak of a clean Au standard sample is about 1.6 eV. A flood gun at 8 eV and a Ni grid placed 3 mm above the sample surface were used for charge stabilization (Bryson, 1987). The C-(C, H) component of the C1s peak of carbon was fixed at 284.8

226

Chapter 10. Properties of charcoal particles eV to calibrate the binding energy scale. Data were analyzed with CasaXPS (Casa Software Ltd, UK). Atomic fractions were calculated using peak areas after a non-linear background subtraction (Shirley, 1972), based on experimental sensitivity factors and transmission factors provided by the manufacturer.

FTIR

FTIR spectra of charcoal powder were obtained with a Bruker Equinox 55 FTIR spectrometer (Bruker), in transmission mode, with single-bounce diamond attenuated total reflection (ATR) equipment and a Trans DTGS detector. One hundred scans were taken for both background and samples, from 4000 to 400 cm-1, with a resolution of 2 cm-1. Spectra were normalized on the basis of maximal deviation. Attribution of organic and inorganic components to peaks was based on Lehmann and Solomon (2010) and Biester et al. (2014).

13C NMR–CPMAS

Solid state 13C NMR spectra were obtained with a Bruker Avance III HD 400 MHz Wideboard spectrometer (Bruker) at 100.65 MHz using Zr rotors of 4 mm OD with KEL-F-caps. CPMAS was applied during magic-angle spinning of the rotor at 14 kHz. A ramped 1H pulse was used to circumvent spin modulation of Hartmann-Hahn conditions. A contact time of 1 ms and a 90 ° 1H-pulse width of 2.5 µs were used. The chemical shifts were calibrated to 0 ppm with tetramethylsilane and to 176.04 ppm with glycine. Pulse delay between single scans were 300 ms. Fractions of carbon chemical bonds were estimated by integrating signal intensity of the spectra in different chemical shift regions (Knicker, 2011b), with MESTRE NOVA software. The region from 0 to 45 ppm was attributed to alkyl-C, from 45 to 110 ppm to O- and N- alkyl and from 160 to 225 ppm to carbonyl-, carboxyl- and amide-C. For quantification of aryl-C, the spinning side bands (-50 to 0 ppm and 225 to 300 ppm) were added the signal ranging from 110 to 160 ppm.

DSC

Samples were scanned with a differential scanning calorimeter (DSC100, TA Instruments) after heat flow calibration with sapphire, and temperature and heat calibration with the melting of In (Danley, 2003). Before analysis, all samples were diluted 20x with Al2O3 and homogenized in a ball mill. Between 15 and 25 mg charcoal powder were weighed into an open Al pan. An empty

227

Chapter 10. Properties of charcoal particles pan served as reference. Samples were heated under a flow of 50 ml/min synthetic air from room temperature to 600 °C at 10 °C/min (Leifeld, 2007). Peak temperature (°C), height (W/g) and area (J/g), and total heat of reaction (J/g) was determined from DSC thermograms with Universal Analysis 2000 software (TA Instruments). After identification of a maximum, peak height was measured as the maximum deviation from a linear baseline drawn between 150 and 600 °C. Peak area was obtained by vertical drop between two adjacent peaks based on the position of the minima.

Dichromate oxidation

The oxidizable C content of charcoal was determined via oxidation with

K2Cr2O7 according to the procedure described by Walkley (1947). In brief, 10 ml 0.167 M K2Cr2O7 was poured in a 500 ml Erlenmeyer flask containing between 25 and 50 mg charcoal powder; 20 ml conc. H2SO4 were added and the mixture was shaken for 1 min. After 30 min, reaction was stopped by 2- adding 200 ml demineralized water. Excess of Cr2O7 was titrated with 0.25

M FeSO4 in the presence of concentrated H3PO4, BaCl2 and diphenylamine as an indicator.

10.3. Results

Raw data of chemical properties of charcoal particles are presented in appendix 3.

XPS

Even though XPS measures the binding energy of core electrons from the surface (< 10 nm) of a sample, we assume that atomic composition established from the method is representative of the bulk properties of a sample, as charcoal samples finely ground to powder were analyzed (Nguyen et al., 2008). Charcoal contained from 45.1 to 77.2% C, 17.4 to 42.2% O and 1.34 to 1.93% N. It also contained 1.46 to 6.18% Si, 0.86 to 3.33% Al, 0.23 to 1.75% Ca and 0.16 to 1.22 % Fe. Small amounts of K and Mg were detected in some samples. Other elements expected to take part in the composition of charcoal, such as P, S or Na, were not present in sufficient concentration to obtain a signal.

228

Chapter 10. Properties of charcoal particles

Figure 10.2. Mean (n=3) atomic composition of charcoal, quantified by X–ray photoelectron spectroscopy (XPS), plotted against estimated cultivation time. Error bars correspond to one standard deviation. On figures, “ageing” indicates switch in content from modern charcoal to pre-industrial charcoals aged in forest soil and “cultivation” from forest pre-industrial charcoals to pre-industrial charcoals under cultivation since 200 years. P-values were obtained by ANOVA and refer to relationship between variable and time of cultivation (excluding charcoal from the currently active kiln). We investigated the evolution of elements present in significant amount in charcoal vs. land-use history (Figure 10.2). Total carbon content decreased largely with both ageing and cultivation time, in contrast to total oxygen,

229

Chapter 10. Properties of charcoal particles which increased sharply (Figure 10.2). Total N did not appear to be affected by ageing or cultivation. Si, Al and Fe all increased sharply with ageing and, overall, cultivation time, whereas Ca decreased with ageing in an acidic forest soil but re-increased sharply with cultivation (Figure 10.2).

FTIR

The spectra all showed a main peak around 1550 cm-1, which corresponds to aromatic C=C vibration and stretching (Figure 10.3). A smaller signal around 1700 cm-1 corresponds to C=O stretching of carboxylic, quinone, amide, ketone, ester or aldehyde (Lehmann and Solomon, 2010). This peak seemed to decrease gradually from forest charcoal to charcoal with the longest cultivation time, like another weak signal around 1230 cm-1 corresponding to C-O stretch or to O-H bending of COOH groups (Figure 10.3).

Figure 10.3. Mean (n=3) Fourier transform infrared spectroscopy (FTIR) spectra of charcoal, presented for each time step of cultivation. Attribution of organic and inorganic components to peaks is based on publications of Lehmann & Solomon (2010). pH value under each curve correspond to the mean pH value of soils from which charcoal was extracted. In contrast, the signal around 1350 cm-1, attributed to COO-, was weaker for charcoal from forest soil than from grassland or agricultural soil. A double peak in the 1100–950 cm-1 region, corresponding to O-Si-O stretching, clearly increased over time of cultivation, as well as the peaks in the 550–400 cm-1 region, attributed to O-Si-O stretching or Fe-O bending (Figure 10.3). The signal from inorganic soil components overlapped with peaks of small

230

Chapter 10. Properties of charcoal particles intensity corresponding to cellulose, carbohydrate and polysaccharide (at around 1200, 1030 and 890 cm-1; Biester et al., 2014).

Carbonate

Carbonate content varies from 0.22 to 5.73%. If data from the active kiln site were removed, carbonate content correlated closely with the pH of the soil from which it was extracted (r = 0.944; Figure 10.4a). Ca content, determined from XPS, appeared to relate closely to carbonate content (r = 0.894; Figure

10.4b), supporting the idea that carbonate is in the form of CaCO3.

Figure 10.4. (a) Carbonate content of charcoal vs. soil pH (measured in demineralized water); (b) Atomic Ca concentration of charcoal, measured with XPS), vs. carbonate content of charcoal.

231

Chapter 10. Properties of charcoal particles

Organic composition

Data were plotted in a Van Krevelen diagram (Figure 10.5) for atomic H/C ratio vs. atomic O/C ratio for the organic fraction (after removal of inorganic C and O content from carbonate). In the diagram, our charcoals are compared with literature data for biomass heated in the absence of air at various temperature values up to 800 °C, resulting in a range of products from uncharred biomass to biochar with a high degree of aromaticity and aromatic condensation (Keiluweit et al., 2010; Budai et al., 2014). Whereas charcoal from the active kiln site had H/C 0.30 and O/C 0.11 (Figure 10.5), pre- industrial charcoal samples had H/C ranging from 0.57 to 0.77 and O/C ranging from 0.34 to 0.43. H/C and O/C for pre-industrial charcoals correlated positively (r = 0.58).

Figure 10.5. Charcoals plotted in a Van Krevelen diagram (organic H/C vs. organic O/C) and compared with literature data from uncharred biomass to biochar produced at ca. 800 °C (Keiluweit et al., 2010; Budai et al., 2014). Organic C and O content, as well as organic H/C and O/C ratios were also plotted vs. time of cultivation (Figure 10.6). Organic C followed the same trend as for total C from XPS, which is a decrease with ageing and time of cultivation. Organic O, however, evolved differently from total O with time of cultivation: it increased largely with ageing but decreased clearly with time of cultivation (Figure 10.6). Both organic H/C and O/C increased with ageing (Figure 10.6), as observed on the Van Krevelen diagram (Figure 10.5). Their evolution with time of cultivation was not clear, however. If we excluded forest charcoal samples, both organic

232

Chapter 10. Properties of charcoal particles

H/C and O/C appeared to decrease slightly vs. time of cultivation (r = -0.44 and r =-0.57, respectively).

Figure 10.6. Elemental properties of the organic fraction of charcoal vs. time of cultivation. P-Values were obtained from ANOVA and refer to relationship between variable and time of cultivation (excluding charcoal from the active kiln); (a) organic C; (b) organic O; (c) organic H/C; (d) Organic O/C.

13C NMR-CPMAS

The proportion of carbon bonds in charcoal was estimated for a selection of pre-industrial charcoals (one per plot) from integration of chemical shift regions in the 13C NMR-CPMAS spectra (Figure 10.7), according to Knicker (2011). All pre-industrial charcoals had a comparable signature, comprising mainly aromatic C ranging from 73.4 to 79.5% (Figure 10.7). They also contained 4.35 to 8.70% alkyl-C, 7.8 to 11.3% O-alkyl, N-alkyl or amide-C and 6.6 to 8.0% carboxyl- or carbonyl-C. All signatures were comparable, with no clear effect of cultivation time.

233

Chapter 10. Properties of charcoal particles

Figure 10.7. (a) 13C NMR–CPMAS spectra of three charcoal samples aged in contrasting soil conditions. Main carbon chemical bonds were attributed to chemical shift region according to Knicker (2011). Asterisks indicate spinning side bands of aryl-C. (b) Proportion of carbon bonds estimated by way of integration of chemical shift regions in 13C NMR–CPMAS spectra (Knicker, 2011) for charcoal aged under contrasting land uses.

DSC

DSC thermograms showed three easily detectable local maxima - peak 1, peak 2 and peak 3 - corresponding to three levels of increasing thermal resistance (Figure 10.8a).

234

Chapter 10. Properties of charcoal particles

Figure 10.8. (a) DSC thermograms of three charcoal samples with contrasting land use history; (b) Area of peak 1 vs. time of cultivation; (c) Area of peak 3 vs. time of cultivation. The temperatures for peak 1 ranged from 360 to 416 °C, that of peak 2 from 418 to 430 °C and that of peak 3 from 493 to 519 °C. We investigated the evolution of both peak temperature and peak area with time of cultivation. The temperature of peak 1 correlated negatively with cultivation time (r = -0.74), whereas temperatures for peak 2 and 3 were poorly positively correlated with time of cultivation (r = 0.39 and r = 0.42, respectively). Comparable with peak temperature, the area of peak 1 is negatively correlated with time of cultivation (r = -0.82; Figure 10.8b). The area of peak 2 correlated very poorly with time of cultivation (r = 0.28) whereas that for peak 3 correlated positively with time of cultivation (r = 0.86; Figure 10.8c).

Dichromate oxidation

Charcoal from the modern kiln was quite resistant to dichromate oxidation, as only 6.6% of OC was oxidized. In contrast, pre-industrial charcoal samples were less resistant to oxidation. For charcoal aged in forest soil, recovery of OC ranged from 7.32 to 20.9 %, whereas oxidation with dichromate recovered

235

Chapter 10. Properties of charcoal particles from 23.9 to 58.6% of OC for charcoal aged in arable soil. Overall, we found a positive relationship between the susceptibility of charcoal to dichromate oxidation and H/C ratio (Figure 10.9).

Figure 10.9. Fraction of organic C oxidized with dichromate (CW&B/Corg) vs. H/C for organic fraction of charcoal.

10.4. Discussion

Organic composition of charcoal

During pyrolysis of biomass, elemental O/C and H/C values decrease, related to the transformation of cellulose, hemicelluloses and lignin to aromatic clusters. The loss of O and H increases gradually with temperature of pyrolysis, according to the degree of aromaticity and aromatic condensation of char (Wiedemeier et al., 2015). Literature data for biochar produced at increasing temperature were plotted in a Van Krevelen diagram to illustrate the transformation (Figure 10.5; Keiluweit et al., 2010; Budai et al., 2014). Charcoal extracted from the topsoil of the active kiln has a signature close to that of fresh biochar, with slightly larger O/C ratio (Figure 10.5). We make the assumption that this site is representative of the initial conditions six months after abandonment of the charcoal kiln site. The larger O content might result from an atmosphere not completely depleted in O during traditional charcoal production with a mound kiln, as conditions of pyrolysis with a traditional mound kiln are more difficult to control than laboratory conditions and varies to some extent within the mound, according to the distance from the central chimney. It might also originate from limited but significant ageing in soil, as charcoal pieces were extracted from soil six months after the last pyrolysis.

236

Chapter 10. Properties of charcoal particles

In contrast to fresh biochar, preindustrial charcoal samples have large H/C and O/C values, ranging respectively from 0.57 to 0.77 and from 0.34 to 0.43 (Figure 10.5). Ageing of charcoal in soil results mainly from oxidation starting from the surface and propagating to the core of the particles (Lehmann et al., 2005). Oxidation creates phenol, carbonyl and carboxyl functionalities at the edge of aromatic rings, responsible for the increase in cation exchange capacity of charcoal over time (Cheng et al., 2008a). Accordingly, both O/C and H/C ratios of charcoal increase through ageing (Cheng et al., 2008a; Calvelo Pereira et al., 2014), somehow counterbalancing the transformations that occur through pyrolysis (aromatization and loss of O-rich functional groups). Weathering is not exactly the reverse process of pyrolysis, however. During pyrolysis, biochar loses on average 2.5 H for one O when the temperature increases (Keiluweit et al., 2010; Budai et al., 2014; cf. Figure 10.5). Through weathering, we evaluated from our data that charcoal gains on average 1.36 H atoms for one O. This may result from hydroxylation and carboxylation at edges of the graphene sheets in charcoal (Lehmann et al., 2005), but also from contamination by organic acids from soil solution (Pignatello et al., 2006) or microbial films in the porosity of charcoal (Lehmann et al., 2011). Nguyen et al. (2008) estimated from XPS an organic O/C ratio of ca. 0.35 for fire–derived pyrogenic carbon residue in cultivated soils cleared by slash and burn. In their study, ageing did not affect bulk O/C of charcoal over one century. Compared with the value of 0.11 for charcoal from the active kiln, an O/C value of 0.35 is very large. This is attributed to the O-rich conditions of production of fire residue contrasting with the O-depleted conditions of pyrolysis. As oxidation of charcoal particle is a preliminary step for microbial decomposition (Hammes and Schmidt, 2009), a smaller O/C value supports a greater stability of charcoal produced by pyrolysis than for fire–derived pyrogenic OM. Nevertheless, after > 150 yr, preindustrial charcoals have O/C values comparable with that, constant over time, of fire–derived pyrogenic OM (Nguyen et al. 2008). This suggests that oxidation of charcoal increases with time to reach a constant value. The slow (a)biotic decomposition of the O–rich fraction of charcoal (Hamer et al., 2004; Kuzyakov et al., 2009), more microbiologically reactive than condensed aromatic rings (Hammes and Schmidt, 2009), might balance oxidation over time and explain why oxidation reaches a steady state after a long time in soil. In contrast to ageing, there is no clear effect of cultivation on the atomic composition of charcoal. The decrease in organic C content and O content

237

Chapter 10. Properties of charcoal particles with time of cultivation (Figure 10.6) is related to the relative increase in Si, Al, Fe and Ca (Figure 10.2), supported by the strong negative correlation (r - 0.98) between C content and the sum of contents of Si, Al, Fe and Ca quantified from XPS. Despite their high degree of weathering, preindustrial charcoal samples remain mainly aromatic, as shown from FTIR and NMR spectra (Figure 10.3 and Figure 10.7), with a significant fraction of carboxyl- C and carbonyl-C resulting from oxygenation through ageing. From forest to agricultural soil with the longest cultivation time, the FTIR signals from charcoal at ca. 1700 cm-1, corresponding to carboxylic acids, and at 1260 cm- 1, corresponding to phenolic acids and C–O bonds in COOH groups (Figure 10.3), decrease gradually. This corresponds to the deprotonation of (aromatic) carboxylic acid to carboxylate related to a progressive increase of soil pH from 3.5–4.0 in acidic forest soil to 7.5–8 after 200 yr of liming (Cheng et al., 2008a). Accordingly, the peak at 1375 cm-1 (Figure 10.3), attributed to COO- , increases with time of cultivation.

Inorganic composition of charcoal and organo-mineral association

As shown by atomic proportions quantified from XPS (Figure 10.2), the contents of Si, Al and Fe in charcoal increase sharply with ageing and, overall, cultivation, which indicates an increase in association with inorganic components of soil. Abundant coating of minerals on the surface of preindustrial charcoal particles, partly occluding the porosity, illustrates this effect (Figure 10.10a). The enrichment in Si, Al and Fe follows the increase in O content measured from XPS (Figure 10.2; r 0.96), which indicates that these elements are bonded mainly to O in soil minerals. We attribute the increase in association with soil minerals to the combined effect of ageing and liming. Whereas fresh charcoal is highly hydrophobic, ageing consists in the creation of O-rich functional groups at the surface by oxidation, mainly carboxylic and phenolic acids (Lehmann et al., 2005; Cheng et al., 2008a), enhances the polarity of charcoal particles (Criscuoli et al., 2014). Reactivity of organic molecules towards metal cations and mineral surfaces increases with polarity (Kleber et al., 2015). Surface functionalities of aged charcoal are dominated by carboxyl groups (Lehmann et al., 2005), which play a major role in SOM stabilization through organo-mineral association (Kramer et al., 2012; Kleber et al., 2015). Surfaces of clay minerals such as phyllosilicates and Fe or Al (hydr)oxides can adsorb aromatic carboxylic acids by way of both outer– or inner–sphere

238

Chapter 10. Properties of charcoal particles complexation depending on the adsorbate, mineral type and pH range (Guan et al., 2006). Therefore, the formation of polar, O-rich functional groups through ageing, mainly carboxyl groups, promotes the association of charcoal with soil minerals.

Figure 10.10. Scanning electron microscopy (SEM) photographs of a small charcoal fragment coated with minerals in the bulk soil of a charcoal kiln site (a) and of cracks visible at the surface of a pre-industrial charcoal (b). Over time of cultivation, frequent application of liming by farmers leads to gradual increase of soil pH and the consequent deprotonation of carboxylic acids (which have pKa between 4 and 5) to carboxylate anions, as inferred from FTIR spectra (Figure 10.3). Ionization of organic molecules increases their affinity for mineral surfaces (Kleber et al., 2015). Therefore, deprotonation of carboxylic acids through liming most likely favors mineral association to charcoal over time of cultivation. An increase in inorganic

239

Chapter 10. Properties of charcoal particles elements in charcoal may result from the coating of minerals at the surface of charcoal particles (Figure 10.10a) but also from precipitates in the porosity of charcoal (Figure 10.11). Porosity confers to charcoal a large specific surface area, which is an interface for the precipitation of minerals.

Figure 10.11. a) Scanning electron microscopic (SEM) photograph of inorganic precipitates in a charcoal from a 200 yr. cultivated soil (pH 7.85); b) energy dispersive X-ray spectroscopy (EDX) spectra of cell walls of charcoal, treated with a coating of PD-Au for charge stabilization; c) EDX spectra of inorganic precipitates in the porosity of charcoal previously coated with Pd-Au.

240

Chapter 10. Properties of charcoal particles

The energy dispersive X-ray spectroscopy (EDX) signature of foliated structures observed with scanning electron microscopy (SEM) in the charcoal (Figure 10.11) corresponds to the atomic composition of 2:1 phyllosilicates, possibly smectites, vermiculites or illite, as significant amounts of Fe, Mg and K were also identified. This corroborates the Si/Al ratio of ca. 2 calculated from XPS results for charcoal subject to cultivation since 200 yr., suggesting that inorganic components associated with charcoal are dominated by 2:1 phyllosilicates. In contrast, other pores of small diameter are not coated with precipitates (Figure 10.11). This might be related to pore occlusion (Pignatello et al., 2006) causing inaccessibility to solutes transported by water fluxes, indispensable for neoformation of secondary minerals. On the other hand, Nguyen et al. (2008) highlighted an increase in the physical breakdown of BC particles over time in cultivated soil, possibly caused by tillage and earthworm activity. An increase in the contact surface accessible to minerals by way of faster fragmentation of charcoal particles in agricultural soil is another factor that might contribute to explain the increase in the content of minerals associated with charcoal over time of cultivation. Large cracks at the surface of charcoal particles illustrate the advanced physical alteration (Figure 10.10b). Nevertheless, the effect of cultivation on the size of charcoal particles was not investigated here because repeated tillage dilutes the sites laterally. Sites initially 10 m in diameter in forest (Hardy et al., 2016) have a diameter up to 40 m after 200 yr of cultivation. This dilution at the sites over time interferes with the effect of cultivation on the number and size of charcoal particles in kiln soil. Whereas charcoal from the active kiln site contains ca. 1% of Ca, the content decreases through ageing in acidic forest soils (Figure 10.2), due to the leaching of “base” cations through natural re-acidification occurring over time after pyrolysis (Hardy et al., 2016). In contrast, Ca content in charcoal increases over time of cultivation in agricultural soils (Figure 10.2), with a close correlation with the carbonate content (Figure 10.4b). This, again, relates to the application of liming amendments, in the form of CaCO3 or

Ca(OH)2, deprotonating carboxylic acids (Figure 10.3) and saturating them with Ca2+. We observed from SEM–EDX a strong signal of Ca in the C–rich walls of charcoal (Figure 10.11), which supports the idea that Ca is strongly associated with the carbonaceous structure of charcoal. The strong affinity of Ca2+ for carboxylate groups is well known (Kalinichev and Kirkpatrick, 2007) and the consequent high affinity of Ca for aged charcoal was highlighted in earlier studies. (Hardy et al., 2017, 2016). Intriguingly, the content of

241

Chapter 10. Properties of charcoal particles carbonate in soil from which charcoal was extracted is systematically < 0.10% (data not shown), whereas charcoals from cultivated plots contain up to 5.73

% of carbonate (Figure 10.4a). This suggests CaCO3 precipitates preferentially in charcoal due to high specific surface area of charcoal and the high affinity of Ca for carboxyl groups at the surface of charcoal. This result provides further evidence that BC is an important interface for exchange between solid and liquid phases in soil, such as precipitation of minerals, including CaCO3. The preferential accumulation of carbonate in charcoal relative to total content in soil indicates that inorganic C, rarely taken into account in the C budget of BC, should not be overlooked, even at pH < 7.

Stability of charcoal

We investigated the thermal and chemical stability of charcoal by two oxidative methods, DSC and oxidation with dichromate, respectively. In agreement with the shift in atomic composition related to ageing, resistance to dichromate oxidation is less for pre-industrial charcoal samples than for freshly produced charcoal (Figure 10.9). Reduced resistance to dichromate oxidation of charcoal exposed to environmental conditions has been reported earlier and was attributed to the physical, chemical or biological weathering of charcoal (Ascough et al., 2011). Accordingly to the findings of Naisse et al. (2013), we found a positive correlation between the susceptibility of charcoal to be oxidized by dichromate and the H/C of the charcoal (Figure 10.9). The H/C ratio, which reflects the degree of aromatic condensation of char to some extent, is indicator of the stability and persistence of BC in the environment (Budai et al., 2013). Our results suggest that, through weathering, H/C ratio of charcoal re-increases, which might correspond to “decondensation” of graphene sheets in the structure of charcoal, possibly caused by physical breakdown coupled with (bio-)chemical oxidation starting at the edges of the graphene sheets and proceeding to their cores. Cultivation had an effect on both chemical and thermal resistance of charcoal to oxidation. The resistance to dichromate oxidation of pre-industrial charcoal aged in agricultural soils was lower than that of pre-industrial charcoal aged in forest (Figure 10.9). This shift of chemical stability was not apparent from FTIR or 13C-NMR-CPMAS spectra, which indicates that the link between the molecular composition of BC and resistance to oxidation is not straightforward. The smaller chemical stability of charcoal under cultivation might be explained by a combination of factors. Frequent tillage might unprotect charcoal pieces by the disruption of soil aggregates (Kuzyakov et

242

Chapter 10. Properties of charcoal particles al., 2009), and thereby accelerate the mechanical breakdown of charcoal particles (Nguyen et al., 2008). Liming might also accelerate the breakdown, as the fragmentation of charcoal particles is accelerated by basic treatments (Braadbaart et al., 2009). By a better soil aeration, tillage might promote the abiotic oxidation of charcoal, which is the first step in the weathering of BC (Cheng et al., 2006; Lehmann et al., 2009). Furthermore, the improved nutrient status related to the application of liming amendments and fertilizers favors biological activity in soil and therefore might accelerate the biological decomposition of the most labile fraction of charcoal, particularly by way of co-metabolism with natural SOM (Hamer et al., 2004; Kuzyakov et al., 2009). The decrease of chemical stability of charcoal particles aged in agricultural soil suggests that charcoal particles might decompose faster under cultivation. Nevertheless, Nguyen et al. (2008) proposed that physical protection of BC by adsorption to minerals and encapsulation in soil aggregates plays an important role in slowing down BC loss from cultivated soils. Association of charcoal with minerals was shown to increase with time of cultivation, which might improve the protection of charcoal against microbial decomposition and therefore balance the lower chemical resistance of charcoal aged in agricultural soil. This assumption is consistent with the high storage of BC in clay rich soils and in soils with high pH (Czimczik and Masiello, 2007; Reisser et al., 2016). Our data does not allow determining which of the two antagonist effects predominates (decrease of chemical resistance vs increase of association with minerals) for the persistence of charcoal in agricultural soil. DSC stresses the issue of BC definition and quantification in soil by highlighting the presence of at least three fractions of C with distinct thermal resistance in preindustrial charcoal (Figure 10.8a) which is not clearly reflected in the molecular composition, dominated by aromatic C (Figure 10.3 and Figure 10.7). Over time of cultivation, we reported a clear decrease in both temperature and area of the less thermally stable fraction of charcoal (peak 1) measured with DSC (Figure 10.8), with a strong negative correlation between peak temperature and Ca content (Figure 10.8c) and a strong positive correlation between peak area and O content (Figure 10.8a). This suggests that, with cultivation, the O-rich fraction of charcoal gets saturated with Ca2+ from liming, which catalyzes its combustion during DSC analysis. A comparable decrease in thermal stability related to the content of Fe3+ and Al3+ was reported for the humic fraction of Podzol soil (Schnitzer et al., 1964). Over time of cultivation, the O-rich, less thermally stable fraction of charcoal decreases (Figure 10.8b) and the most thermally stable fraction increases

243

Chapter 10. Properties of charcoal particles accordingly (Figure 10.8c). These results can be interpreted in two ways. The decrease of the less thermally stable fraction of charcoal under cropping might result from a faster decomposition of the oxidized, degraded fraction of charcoal resulting in a preferential enrichment of the O-poor, more thermally stable fraction. The greater survival over one century of the more condensed fraction of BC in Russian Chernozerms (Hammes et al., 2008), and the greater stability of BC remaining after 20 years of cultivation in soil cleared by forest fire in Kenya (Nguyen et al., 2008) are in agreement with this assumption. However, one could think the opposite: the smaller content of the O-rich, thermally less stable fraction of charcoal for the long cultivation times and the relative enrichment in the O-poor, thermally stable fraction might result from the slowdown of the degradation of charcoal related to a better stabilization of charcoal particles by association with minerals. The fact that some soils with large clay content and high pH store large amounts of BC (Czimczik and Masiello, 2007; Reisser et al., 2016) supports this latter point. Both assumptions are sound, albeit contradictory and we cannot determine whether cultivation slows down or accelerate the loss of charcoal from soil based on the results of this study.

10.5. Conclusion

For the first time, properties of preindustrial charcoal particles produced by pyrolysis in traditional mound kilns were studied along a chronosequence of land use change from forest to agricultural soil. We investigated the effect of ageing and cultivation on organic and inorganic properties of charcoal samples, as well as thermal and chemical resistance to oxidation. Compared with fresh charcoal, preindustrial charcoal has high O and H contents due to advanced weathering. Carbon bonds remain, however, dominated by aromatic-C, with a significant fraction of O-rich functional groups such as carboxyl, carbonyl and O-alkyl. Cultivation favors associating charcoal with soil minerals, which results from deprotonation of carboxylic acids to carboxylate ions under liming, enhancing their reactivity towards mineral surfaces. Soil minerals, including 2:1 phyllosilicates and CaCO3, either coat charcoal particles or precipitate inside the charcoal. Thermal stability of preindustrial charcoal explored with DSC revealed the presence of three fractions with distinct thermal resistance in charcoal. The O-rich, less thermally stable fraction decreased over time of cultivation, leading to the relative increase of the thermally most stable fraction of charcoal. This might result from (i) the preferential loss of the O-rich fraction due to improved soil conditions under cultivation or (ii) the slowdown of charcoal from oxidation 244

Chapter 10. Properties of charcoal particles by association with minerals. Both ageing and cultivation decrease the resistance of charcoal to dichromate oxidation, which was not reflected in the degree of aromaticity of charcoal. Resistance to chemical oxidation is negatively correlated with the H/C, which is generally considered to mirror the degree of aromatic condensation of char. Our results highlighted that land use significantly affects the properties of BC through the modification of soil conditions, which might influence the kinetics of BC loss from soil. A reliable estimation of the residence time of BC in cultivated soil is key for unveiling the potential contribution of biochar to terrestrial carbon storage.

Acknowledgements

We are grateful to F.-X. Henrard who took part in sampling, sample preparation and preliminary analyses as a MSc. student. We thank Pierre Eloy for his precious help for XPS analyzes and Laurence Ryelandt for SEM-EDX images. Funds were provided by the Fonds Spéciaux de Recherche (FSR)– Université catholique de Louvain.

245

Chapter 11. Carbon stocks by remote sensing

Chapter 11. Evaluation of carbon stocks at pre-industrial charcoal kiln sites by remote sensing along a chronosequence of a land use change from forest to cropland.

Summary

Previous results (Chapter 10) have indicated that a land use change from forest to cropland decreases the chemical stability of charcoal particles. Therefore, we hypothesized that a shift in land use from forest to cropland might accelerate the loss of charcoal because of an increase in pH, a greater microbial activity and the mechanical action of tillage. To address this issue, we aimed to estimate soil organic carbon (SOC) stocks for a number of pre- industrial charcoal kiln sites with contrasting time of cultivation, to follow the evolution of carbon stocks along a chronosequence of cultivation. In that goal, remote sensing provides the opportunity to acquire a large amount of data rapidly and at low cost. We acquired high resolution satellite imagery (Ikonos- 2 and Geoeye-1) to estimate SOC stocks based on the attenuation of reflectance of bare agricultural soil related to SOC enrichment at pre-industrial charcoal kiln sites. To calibrate the relationship between reflectance and SOC content, 178 soils were sampled in nine different fields and related to the pixel value of the panchromatic band of the satellite imagery. Regardless of the field, soil reflectance decreased linearly with SOC for concentrations < 40 g kg-1 but remained constant for higher values. Average reflectance of the field was much variable depending on soil surface conditions, strongly influenced by surface humidity and roughness. To normalize the relationship between reflectance and SOC content between fields with contrasting surface conditions, we calculated the excess of SOC content and the corresponding attenuation of reflectance by difference with the background of the soil unaffected by charcoal production. The accuracy of the model was tested by cross-validation, and average root mean square errors (RMSE) of 1.95 g kg-1 were obtained. The model was used to predict SOC stocks at kiln sites for a selection of fields with contrasting history of cultivation in the surroundings of Gembloux. SOC stocks ranged between 1.70 and 3.73 t per kiln site, with no clear effect of the history of cultivation on SOC stocks (P=0.105). Nevertheless, the limitation of our model to SOC concentrations < 40 g kg-1 did not allow to explore a large range of history of cultivation. Therefore, the

247

Chapter 11. Carbon stocks by remote sensing question of an effect of cultivation on kinetics of charcoal loss from soil remains open. It is likely that this limitation might be overcome by hyperspectral remote sensing, which better catches the (VIS-)NIR footprint of soil organic matter that consists of vibration over-tones of biochemical groups specific to soil organic matter.

11.1. Introduction

Under given soil conditions, black carbon has a residence time in soil of at least one or two order of magnitude longer than that of uncharred SOM. Nevertheless, the content of BC stored in soil is low regarding annual production rates from forest fires, which provides evidence for important losses of BC from soil (Masiello, 2004; Schmidt, 2004). Among mechanisms involved in BC loss, microbial decomposition has been proved to play a role (Baldock and Smernik, 2002; Hamer et al., 2004; Wengel et al., 2006), possibly by way of co-metabolism with labile organic molecules (Hamer et al., 2004; Kuzyakov et al., 2009). Large amounts of terrestrial BC can also be transported from soil in a dissolved form and build-up C stocks of riverine and oceanic waters and sediments (Hockaday et al., 2007; Jaffé et al., 2013). Cheng et al. (2008) have studied the evolution of properties of charcoal remnants at the sites of historic charcoal blast furnaces along a climatic gradient. Their results suggest that temperature and pH accelerate the decomposition of BC, which accords with results obtained from incubation studies of fresh biochars (Lehmann et al., 2015). Mechanical disturbance also promoted the decomposition of 14C-labelled biochar incubated in laboratory conditions (Kuzyakov et al., 2009), which suggest that tillage might accelerate the turnover of BC in soil. However, no clear evidence for such effect was recorded in field conditions (Skjemstad et al., 2004; Vasilyeva et al., 2011). A land use change from forest to cropland implies a shift in the quantity and quality of organic matter inputs as well as a number of farming practices such as liming, the use of (in)organic fertilizers and tillage that modify soil conditions and therefore might affect the loss of BC from soil. In Wallonia, large areas of land that were subject to charcoal production were deforested and converted into cropland since the late 18th century. Cultivation is known to decrease the stocks of uncharred SOC (Goidts and van Wesemael, 2007; Solomon et al., 2007; van Wesemael et al., 2010), particularly during the first 20 years after deforestation (Solomon et al., 2007a). This is mainly due to a decrease of the inputs of SOM, but also to an increase of mineralization rates due to a modification of soil conditions and agricultural practices. We assume

248

Chapter 11. Carbon stocks by remote sensing that a land use change from forest to cropland might decrease BC stocks like it was reported for uncharred SOC stocks. On the one hand, the shift in pH and the greater microbial activity related to the improved soil conditions might accelerate the decomposition of BC. On the other hand, tillage increases the breakdown of BC particles (Nguyen et al., 2008) and promotes soil aeration and destruction of aggregates, which might also accelerate BC decomposition. The presence of pre-industrial charcoal kiln sites in various fields with contrasting history of land-use (Chapter 4) provides a unique opportunity to test the long-term effect of cropping on the rates of BC loss from soil. To address this issue, an estimation of SOC stocks for an important number of sites is necessary. In that goal, proximal and remote sensing have the potential to acquire a large amount of data more rapidly and at lower cost than allowed by conventional sampling and chemical analysis. Field and airborne imaging spectroscopy are more and more used in soil monitoring, particularly to map topsoil organic carbon of bare agricultural soil (Bartholomeus et al., 2008; Croft et al., 2012; Hbirkou et al., 2012; Nocita et al., 2011; Stevens et al., 2010, 2008, 2006). For the mapping of limited areas of land, like the scale of an agricultural field, the use of drones is more and more frequent. Commonly, spectroscopic data are acquired with hyperspectral sensors in the visible (VIS) and the near infrared (NIR). The relationship between the signal and SOC content is calibrated by multivariate regressions such as partial least square regression (PLSR) along with principal components regression (PCR) to overcome the problem of high-dimensional, correlated predictors (Croft et al., 2012). Several factors other than SOC content affect soil reflectance in the VIS-NIR, including soil properties such as iron oxides, clay and carbonates or extraneous factors such as surface humidity and roughness (Ben-Dor et al., 2009; Mulder et al., 2011). Therefore, the accuracy of SOC prediction depends on the variability of properties and conditions of soil in the area of interest. Root mean square errors (RMSE) of prediction of 5.3–6.2 g kg-1 were obtained under a global calibration for a study area of 420 km² in Luxembourg, whereas RMSE were decreased by a factor up to 1.9 under local calibrations by soil texture and by geographic area (Stevens et al., 2010). Comparably to uncharred SOM, enrichment of charcoal attenuates soil reflectance in the VIS-NIR. Soil color is an indispensable feature to detect charcoal kiln sites in cropland because their relief has been completely erased by repeated tillage. Accordingly, aerial photographs and satellite imagery appeared to be ideal prospection tools to detect the sites on bare soil (Chapter 3). Spectral remote sensing provides a smaller density of information than

249

Chapter 11. Carbon stocks by remote sensing hyperspectral imaging spectroscopy. Nonetheless, the use of VIS-NIR spectral sensors is widespread in airborne and satellite imaging; hence, several image covers of Wallonia acquired recently are readily available, even though some of them are not calibrated radiometrically, such as orthoimages of the Public Service of Wallonia. The goal of this chapter was two-fold. First, we aimed to exploit the relationship between SOC content and the VIS-NIR reflectance of high spatial resolution satellite imagery to estimate BC stocks of pre-industrial charcoal kiln sites at the scale of a field. Second, we aimed to estimate BC stocks along a chronosequence of cultivation to assess whether the loss of charcoal-C at kiln site accelerates under cultivation.

11.2. Material and methods

Satellite imagery

We acquired five high resolution satellite archives covering areas that had been deforested for cultivation since the time of charcoal production (Figure 11.1) and that showed a number of bare fields containing charcoal kiln sites. Satellite imagery includes four Ikonos-2 and one Geoeye-1 images, covering about 750 km² in total (Table 11.1). Cloud cover was null for each of them. At 26° off-nadir, Ikonos-2 archives have a spatial resolution of 1 x 1 m for the panchromatic (PAN) band and of 4 x 4 m for the multispectral (MS) bands of the visible (VIS; blue, green and red) and the near infrared (NIR). Spectral wavelengths range of Ikonos-2 is 445–516 nm for the blue band, 506–595 nm for the green band, 632–698 nm for the red band, 757–853 nm for the NIR band and 526–929 nm for the PAN band. At 28° off-nadir, Geoeye-1 images have a spatial resolution of 0.5 x 0.5 m for the panchromatic (PAN) band and of 2 x 2 m for MS bands. Spectral wavelength range of Geoeye 1 is 450–510 nm for the blue band, 510–580 nm for the green band, 655–690 nm for the red band, 780–920 nm for the NIR band and 450–800 nm for the PAN band. Images were orthorectified with the program Envi 4.8. We used rationale polynomial coefficient equations (RPC) provided by the manufacturer, a digital elevation model (DEM) with a spatial resolution of 5 x 5 m provided by the Public Service of Wallonia and at least two ground control points (GCP) for each image. Chosen GCP were targets with a location immutable over time such as crossroads, identified on both the Ikonos-2/Geoeye-1 imagery and a georeferenced orthoimage of the Service Public of Wallonia, used as reference for differential correction.

250

Chapter 11. Carbon stocks by remote sensing

Table 11.1. Characteristics of satellite imagery Acquisition date and Cloud Area Number Name Imagery time cover (Km²) 1 Baisy-Thy Geoeye-1 17 april 2011, 10:42 0 83.8 2 Gembloux Ikonos-2 2 april 2009, 10:56 0 160.4 3 Andenne Ikonos-2 18 april 2010, 10:37 0 118.5 4 Fosses-la-Ville Ikonos-2 19 april 2010, 11:11 0 110.9 Ermeton-sur- 5 Ikonos-2 19 april 2010, 11:11 0 270.0 Biert

Figure 11.1. Location of the five satellite images on the map of Wallonia Spatial resolution of the PAN band is more adapted than MS bands to the short-distance gradient of carbon at the site of pre-industrial charcoal kilns. Accordingly, PAN pixels were preferred to MS data to calibrate the relationship between reflectance of bare soil and SOC content of the plow layer. Digital numbers (DN) of the PAN band were transformed into satellite -2 -1 -1 radiance (Lrad; mW cm sr µm ) according to equation (1) for Geoeye-1 and equation (2) for Ikonos 2:

퐿푟푎푑 = 0.17786 × 퐷푁 (1)

퐿푟푎푑 = 0.15412 × 퐷푁 (2) Coefficients of equations (1) and (2) result from the multiplication of the bandwidth (350 and 403 nm, respectively) with a calibration coefficient provided by the manufacturer. 251

Chapter 11. Carbon stocks by remote sensing

Satellite radiance data were then converted into planetary (or “top of atmosphere”) reflectance (ρtop) values with the program ArcMap 10.4 (ArcGIS) following equation (3):

π × Lrad × d² 휌푡표푝 = (3) 퐸푆푈푁×푐표푠휃푧 Where ESUN (137.58 mW cm-2 µm-1) is the mean solar irradiance for the PAN band, θz is the zenith angle (provided in the metadata of images) and d is the earth-sun distance, calculated based on Julian Day (jd; a continuous count of day used for astronomical calculations) following equation (4): 푑 = 1 − 0.01674 × cos (0.9856(푗푑 − 4)) (4)

Soil sampling and determination of SOC content

A specific sampling was realized to relate SOC content to the reflectance value of corresponding PAN pixels of satellite images. We chose nine fields whose soil was bare on the image. Selected fields contained at least four pre- industrial charcoal kiln sites and had relatively homogeneous surface conditions. Fields were sampled on the 19 December 2012 and the 2, 3, 4 and 8 April 2013, at 178 locations in total. In each field, we selected a minimum of four charcoal kiln sites and for each kiln site, we sampled soil at four locations along a carbon gradient from the center to the outside of the site (Figure 11.2). The last point was sampled at a sufficient distance from the site to be considered unaffected by charcoal production. Each soil sample was bulked from five cores sampled in the plow layer (0–25 cm) with a gouge auger, located at the center and at the four corners of a square of 1 x 1 m. We make the assumption that the fraction of charcoal at kiln sites in the subsoil below the plow layer is negligible, which is in agreement with the OC concentration measured in the subsoil (35–50 cm) of kiln site similar to that of adjacent reference soil (Chapter 8) The size of the areas of sampling accords with that of one pixel of the PAN band of the Ikonos 2 imagery. The center of each sampling area was located with a differential GPS Leica 1200, with the GPS carrier phase technique. Differential corrections of GPS coordinates were realized instantaneously by GSM connection to the red of reference Global Navigation Satellite System (GNSS) stations of Wallonia (WALCORS). Accuracy of the GPS position by this technique is of a few centimeters.

252

Chapter 11. Carbon stocks by remote sensing

Figure 11.2. Sixteen sampling points from four charcoal kiln sites in one of the plots that was used to calibrate the relationship between SOC content and pixel value of the panchromatic band of Ikonos 2 imagery. Soil samples were dried at 40 °C and sieved to 2 mm. The < 2 mm fraction was crushed to a powder with an oscillating ring crusher RS 200 (© Retsch). Total C content was determined by dry combustion with an elemental analyzer vario MAX (Elementar). Inorganic C content was analyzed by the modified- pressure calcimeter method (Sherrod et al., 2002) but content was null or below the detection limit of the apparatus. Therefore, total C was considered to correspond exclusively to organic C (and includes charcoal-C).

Bulk density

An estimation of bulk density is indispensable to transform SOC concentration into SOC stock in a given soil volume. For a selection of sampling areas (n = 32), an additional core was taken for the determination of bulk density. Undisturbed cores were sampled with a gouge auger of 4 cm of radius, and removed with caution from the gouge. Height of the soil cylinder was measured accurately to calculate the volume of the core. Soil volume was 520 (55) cm³ on average, which is more than 5 times the typical volume of 100 cm³ of Kopeckis traditionally used to determine bulk density of soil. We preferred to sample a larger volume of soil for a better representativeness of the bulk density of soil in the plow layer, which has a heterogeneous macroporosity because of frequent tillage. Mass of dry soil was determined

253

Chapter 11. Carbon stocks by remote sensing gravimetrically after drying at 105°C, and bulk density was calculated as the ratio between dry weight and soil volume. Bulk density of soil is known to depend on SOM content. To predict the bulk density of soil in absence of measurement, we calibrated a pedotransfer function according to Rawls (1983), which describes bulk density of soil as a function of SOM content :

100 휌푏 = 푆푂퐶 100−푆푂퐶/푓 (5) + 퐶,푆푂푀 푓퐶,푆푂푀 ×휌푏,푆푂푀 휌푏,푚푖푛 Where ρb is the bulk density of soil, SOC (%) is soil organic carbon content, fC,SOM is the fraction of C in SOM, ρb,SOM is the density of the organic fraction of soil and ρb,min is the density of the inorganic fraction of soil. Generally, fC,SOM is considered to be 0.58 for uncharred SOM. Based on elemental composition of charcoal particles extracted from 27 pre-industrial charcoal kiln sites (Chapter 4, Chapter 10), we calculated an average value of 0.61

(0.01) for charcoal. Therefore, we fixed the value of ρb,SOM at 0.595, which is a reasonable estimation of fC,SOM in pre-industrial kiln, soil where charcoal is mixed to uncharred SOM. Both ρb,min and ρb,SOM were determined empirically by fitting the model to experimental values in R 3.3.1.

Procedure of estimation of carbon stocks

We aimed to interpret the attenuation of reflectance caused by charcoal enrichment to map topsoil SOC and predict SOC stocks at pre-industrial charcoal kiln sites. Therefore, SOC content was related to DN, Lrad and ρtop of the corresponding pixel of the satellite imagery thanks to GPS coordinates of the sampling point. For each field of the calibration set, we observed that reflectance was strongly negatively correlated to SOC content. Nevertheless, the average level of reflectance for a same SOC content was much variable according to soil surface condition. In an attempt to obtain a unique relationship between SOC content and soil reflectance for the different fields, data were normalized by difference with the average level of reflectance and SOC of the field. Practically, we established the relationship between the excess of SOC (Δ SOC) of a sample and the corresponding attenuation of reflectance (Δ ρtop) of a pixels from the kiln site with respect to the average SOC level of the field in the area that was not affected by charcoal production.

The calibration of the relationship between Δ SOC content and Δ ρtop is detailed in the first part of the result section. The quality of the calibration fit

254

Chapter 11. Carbon stocks by remote sensing was tested by cross-validation after randomly splitting the data into four subgroups, with R 3.3.1. Once SOC content has been mapped based on soil reflectance, SOC stocks (kg) in the plow layer can be calculated in a given soil volume according to the equation:

푆푂퐶 푆푂퐶 푠푡표푐푘 = 푆 × 퐻 × 휌 × (6) 푏 1000 Where S is the surface (m²) of one pixel, H is the depth (m) of the plow layer, - -1 ρb is the bulk density (kg m ³) and SOC is the concentration of SOC (g kg ). At the scale of a field, some factors other than SOC can affect soil color, such as compaction by wheels of tractors or an increase in silt content, humidity or SOC in colluvial deposits. To limit the effect of such factors on the estimation of SOC stocks at pre-industrial charcoal kiln sites, the area affected by kiln sites was delimited by segmentation of images with eCognition Developer 64. Segmentation consists in splitting the image in spatial units that are spectrally homogeneous. For the delimitation of areas affected by pre-industrial charcoal kiln sites, we used a scale parameter of 40, a shape parameter of 0.1 and a compactness parameter of 0.5. To ensure that all pixels affected by charcoal enrichment are included in the estimation of SOC stocks at pre-industrial charcoal kiln sites, a circular buffer surrounding the areas identified as charcoal kiln site was drawn to increase the equivalent radius of the areas by 15 m. Carbon stocks were calculated with ArcMap 10.4 by summing the contribution of each pixel from areas delimited by these buffers.

Chronosequence of cultivation

To test the assumption that cultivation increases the loss of charcoal from pre- industrial charcoal kiln sites, we estimated carbon stocks based on remote sensing data for a selection of fields containing charcoal kiln sites along a chronosequence of land-use change from forest to cropland. In the surroundings of Gembloux, we identified three episodes of land use change from forest to cropland in areas affected by charcoal production thanks to the map of Ferraris (1770–1778), the map of Vandermaelen (1846–1854), the map of the “dépôt de la guerre” (1871–1875) and the first map of the “Institut de Cartographie Militaire” (1901) (see Chapter 4, Figure 4.4). For each time step of deforestation, we identified at least four bare fields with pre-industrial charcoal kiln sites on the Ikonos-2 imagery, and we estimated SOC stocks at kiln site in these fields based on soil reflectance. To interpret

255

Chapter 11. Carbon stocks by remote sensing differences in SOC stocks at charcoal kiln site as a consequence of contrasting history of cropping, we must assume that (i) each field has experienced a comparable history of charcoal production; (ii) deforested soils have experienced non-stop cultivation since deforestation and (iii) they have been subject to comparable agricultural practices since the time of deforestation. The validity of these assumptions will be discussed later in the text.

11.3. Results

The effect of SOC on soil reflectance

The black color of bare soil at kiln site (Figure 11.2) results from an attenuation of soil reflectance in the VIS-NIR caused by charcoal enrichment. Pixel values of a bare field with relatively homogeneous surface conditions and that has not been affected by charcoal production follows a normal distribution (Figure 11.3a), whereas that of a bare field that contains charcoal kiln sites spread towards low values as a result of the darkening of soil caused by charcoal enrichment (Figure 11.3b).

Figure 11.3. Frequency histograms of digital numbers from the panchromatic band of Ikonos-2 imagery. a) Field with bare soil containing no charcoal kiln site; b) Field with bare soil containing several charcoal kiln sites darkening the soil, which causes a queuing of the distribution towards low values. The dotted line indicates the mean value of the distribution.

Planetary reflectance (ρtop) was plotted against SOC content for the nine fields that were sampled to calibrate the relationship (Figure 11.4). Regardless of the field, ρtop seems to decrease linearly with SOC content. Nevertheless, the average reflectance varies from a field to another to a relatively large extent. The slope of the relationship also varies to some extent. Particularly, field 6 that spans over the largest range of SOC content with values up to 84.8 g kg-1 has a slope that is less steep than other fields (Figure 11.4).

256

Chapter 11. Carbon stocks by remote sensing

Figure 11.4. Planetary reflectance (ρtop) against SOC content at sampling locations from nine different fields.

In an attempt to generalize the relationship between ρtop and SOC content regardless of the field, we established the relationship between the excess of SOC (Δ SOC) of a sample with respect to charcoal-unaffected reference soil, and the corresponding attenuation of reflectance (Δ ρtop). This approach assumes that the slope of the relationship between ρtop and SOC is similar from one field to another, regardless of the average reflectance of the field. This step of normalization appeared to be quite successful, except for values of Δ SOC > 20 g kg-1 (Figure 11.5). Only fields 5 and 6 reached such values, because these fields were converted recently from grassland to cropland to produce silage maize in an agronomic region dedicated mainly to animal husbandry. Accordingly, they were subject to a shorter cultivation history and therefore were less diluted by tillage, which explains the large SOC content and the associated saturation of the signal.

257

Chapter 11. Carbon stocks by remote sensing

Figure 11.5. Attenuation of planetary reflectance (∆ ρtop) of pixels from charcoal kiln sites relative to the average value of pixels from the same field unaffected by charcoal production against the increase of SOC content (∆ SOC) at charcoal kiln site relative to the average SOC content of adjacent reference soil.

Prediction of SOC content with soil reflectance

After removing fields 5 and 6 from the data, we calibrated a linear model to predict ∆ SOC content from ∆ ρtop values (Figure 11.6a). We obtained similar determination coefficients by using ∆ Lrad and ∆ DN rather than ∆ ρtop (Figure 11.6b, c). The calibration fit was tested by cross-validation, randomly splitting the data into four subsets. An average root mean square error (RMSE) of 1.95 g kg-1 was obtained for predicted values (Table 11.2). Cross-validation was preferred to a strict validation (splitting the data into a calibration and a validation dataset) because data is relatively scarce (n=143) after the exclusion of field 5 and field 6. Table 11.2. Root mean square (RMSE) errors calculated by cross-validation. The dataset was split into four subsets. Cross-validation RMSE (g kg-1) 1 (n=35) 2.36 2 (n=36) 1.74 3 (n=36) 1.81 4 (n=36) 1.84 Mean±sd 1.95±0.28

258

Chapter 11. Carbon stocks by remote sensing

Figure 11.6. Relationships between the excess of SOC content (∆ SOC) at charcoal kiln site relative to adjacent reference soil and attenuation of the pixel value of high resolution satellite imagery, expressed as a) planetary reflectance (ρtop), b) digital numbers (DN) and c) radiance (Lrad). The model was applied to a selection of bare fields to map the increase of SOC content at charcoal kiln site relative to adjacent reference soil, unaffected by charcoal production. We mapped several fields from the surroundings of 259

Chapter 11. Carbon stocks by remote sensing

Gembloux, Andenne, Fosses-la-Ville and Baisy-Thy (Figure 11.7), but not from the surroundings of Ermeton-sur-Biert (Figure 11.1; Table 11.1. Characteristics of satellite imagery), to avoid the issue of saturation of the signal that was observed for the high concentrations of SOC related to short history of cropping. On the maps, we can observe that charcoal kiln sites have different sizes and shapes, with striations from tillage more or less visible. Maximum SOC content also varies from one field to another. The shape of the kiln sites of the fields from Fosses-la-Ville (Figure 11.7a) and Andenne (Figure 11.7c) is close to circular whereas that from Gembloux (Figure 11.7b) and Baisy-Thy (Figure 11.7b) are more elongated. Kiln sites of the field from Baisy-Thy contrast with kiln sites from other field because they have sharp striations, and because maximum SOC content is also clearly smaller (Figure 11.7b). Between the large sites of fields from Fosses-la-Ville and Andenne (Figure 11.7a, c), small spots are also visible.

Figure 11.7. Selection of maps of SOC concentration of the plow layer of four fields from a) Fosses-la-Ville, b) Gembloux, c) Andenne and d) Baisy-Thy, predicted with planetary reflectance of the panchromatic band of Ikonos-2 and Geoeye-1 satellite imagery.

260

Chapter 11. Carbon stocks by remote sensing

Bulk density

Coefficients of the pedotransfer function (equation 5) were calibrated thanks -3 to experimental data. The best fit estimated a ρb,min of 1528.7 kg m and a -3 ρb,SOM of 468.3 kg m for a fC,MO of 0.595. The quality of the prediction was relatively poor, with a coefficient of determination of 0.30 obtained by linear regression of predicted values against measured values. The pedotransfer function was compared to a simple linear regression model of bulk density against SOC content (Figure 11.8). By linear regression, we obtained a determination coefficient identical to that obtained for the pedotransfer function (R² = 0.30), and a very similar relationship for the range of value of experimental data (Figure 11.8). For the sake of convenience, we used the linear relationship to predict bulk density from SOC (g kg-1) content:

휌푏 = 1520.1 − 4.9096 푆푂퐶 (7)

Figure 11.8. Relationship between soil bulk density and SOC content in the plow layer of pre-industrial charcoal kiln sites. The solid line results from the linear regression between bulk density and SOC whereas the dotted line is the result of a pedotransfer function (Rawls, 1983) calibrated with experimental data.

Carbon storage in pre-industrial charcoal kiln sites along a chronosequence of cropping history

SOC concentration was then converted into SOC stocks according to equations 6 and 7 for a selection of fields in the surroundings of Gembloux with contrasting history of land use change from forest to cropland (Table 11.3). Total amount of SOC was divided by the number of kiln site of the field to obtain the average stock of SOC per kiln site. For the fields deforested the most recently, between 1871-75 and 1901, the average SOC stock per kiln site

261

Chapter 11. Carbon stocks by remote sensing was 2.93±0.67 (mean±s.d.) tons. For the fields deforested between 1846-54 and 1871-75, it was 2.17±0.55 tons and 2.32±0.18 tons for fields deforested between 1770-78 and 1846-54. According to these numbers, it is quite uncertain whether the period of deforestation has an effect on SOC storage at charcoal kiln site (P=0.105). Remarkably, two of the fields from the last period of deforestation have an average SOC stock of 3.73 tons per kiln site, which is considerably larger than SOC content measured in any other field. Table 11.3. Stocks of SOC content per kiln site estimated based on soil reflectance for a selection of fields deforested for cultivation. Bold numbers indicate mean±s.d. Period of deforestation Field Number of kiln Stock SOC per kiln sites site 1871-75 – 1901 1 7.5 3.73 2 2 2.76 3 4 2.04 4 7.5 2.71 5 6 2.60 6 5 3.73 2.93±0.67 1846-54 – 1871-75 1 6 1.70 2 4.5 2.78 3 8.5 2.49 4 4 1.70 2.17±0.55 1770-78 – 1846-54 1 7 2.58 2 6.5 2.18 3 5 2.24 4 4 2.27 2.32±0.18

11.4. Discussion

Estimations of SOC stocks by remote sensing

The quality of the prediction of SOC concentration based on VIS-NIR reflectance of satellite imagery depends on two main types of factors. The first is related to the natural variability of soil surface and the second is related to instrumental and methodological issues.

11.4.1.1. Variability of soil surface

Intrinsic soil properties such as texture, stone load, the content of iron oxides and mineralogy affect soil color (Ben-Dor et al., 2009; Mulder et al., 2011). Properties can change either abruptly or continuously depending on the distribution of the rock parent material, re-distribution by soil erosion and

262

Chapter 11. Carbon stocks by remote sensing transport or the history of land use. Variability of soil properties makes the calibration of models of prediction of SOC by remote sensing difficult on large areas and thereby accurate predictions often rely on local calibration datasets (Stevens et al., 2010). Fortunately, the soils that were deforested for cultivation since the time of charcoal production were almost exclusively Luvisols developed on quaternary loess. These were deforested selectively because of their high agricultural potential. Loess is probably the most homogeneous soil parent material at the scale of Wallonia, which is an advantage for an accurate prediction of SOC. Nevertheless, silt content can vary at the scale of the field according to the relief, because it accumulates preferentially in colluvial deposits. Some fields containing charcoal kiln relics also had variable contents of iron oxides (Figure 11.9a). As discussed in Chapter 3, reddish spots enriched with iron oxides might correspond to places where the iron ore was crushed before smelting in a low furnace, in times preceding the invention of blast furnace (personal communication of Christophe Colliou, archeologist). Iron oxides can interfere with the estimation of SOC content because they modify the reflectance of soil.

Figure 11.9. Short-distance variability of the color of bare soil. a) Field with pre- industrial charcoal kiln sites (black spots) and heterogeneous content of iron oxides (reddish areas are enriched with iron oxides); b) shift of topsoil color from very light to brown because of tillage (the red circle spots a farmer plowing the field). Moist spots darken the soil in the bottom left corner of the field. Extraneous factors also modify the reflectance of bare soil. First, the presence of crop residues and vegetation interferes with the spectral signature of soil. When soil is perfectly bare, surface humidity and roughness are main factors that influence the reflectance of soil (Figure 11.9b). Surface humidity decreases the overall reflectance of soil (Girard and Girard, 1989). It depends on precipitations, lateral and vertical redistribution of water, soil porosity and tillage (surface of aggregates dries rapidly after tilling). Drainage is also

263

Chapter 11. Carbon stocks by remote sensing slower in colluvial deposits. Surface roughness varies depending on tillage and precipitations that cause capping and crusting. The amount of relative shadow decreases with the size of soil aggregates, which makes the soil lighter (Girard and Girard, 1989). Like the size of aggregates, the direction of tillage is also very important because it influences the bidirectional reflectance distribution function (BRDF) of the soil, which describes how light is reflected by a surface. In our study, we limited the effect of surface humidity and roughness on the prediction of SOC by (i) selecting homogeneous soil units (field scale) and (ii) harmonizing SOC quantification between fields that had contrasting surface conditions by considering the surplus of SOC at kiln site as a function of the attenuation of reflectance relative to the average background of the field rather than absolute SOC content and reflectance values. Without this step of normalization, calibration of a single model for the prediction of SOC content based on PAN satellite reflectance would have been impossible. This highlights the strong influence of surface conditions on the relationship between SOC content and VIS-NIR spectral reflectance, which justifies the use of hyperspectral remote sensing for a reliable prediction of surface SOC content in absolute terms. The natural variability of SOC content at the scale of the field can also decrease the accuracy of the predictive model.

11.4.1.2. Instrumental and methodological issues

According to cross-validation results, RMSE of prediction of SOC concentrations is 1.95 g kg-1, which is within the range of values obtained with hyperspectral images and local calibration datasets for the prediction of uncharred SOC concentration (e.g. Stevens et al., 2010; Vaudour et al., 2016). Nevertheless, the main limitation of our model was related to the saturation of the signal at SOC content > 40 g kg-1, which prevented the prediction of SOC stocks in fields that had been recently converted to cropland. Prediction of SOC stocks by VIS-NIR hyperspectral remote sensing is not subject to such limitation (McCarty et al., 2002). This probably results from the poor spectral resolution of the satellite imagery, which makes the spectral signature of SOC little specific. Wavelengths of molecular vibration of organics and other soil constituents occur mainly in the mid infrared with weaker signals from vibration over-tones and combination bands occurring in the VIS-NIR (Croft et al., 2012; McCarty et al., 2002; Reeves, 2010; Viscarra Rossel et al., 2006). The use of mid-infrared spectroscopy in field conditions is not appropriate because of strong absorption of water that decreases the signal-to-noise ratio of other molecules (Reeves, 2010). Consequently, (VIS-)NIR hyperspectral 264

Chapter 11. Carbon stocks by remote sensing sensors are preferred for the monitoring of soil properties in field conditions because moisture has a smaller effect on (VIS-)NIR foot print of SOM. Pixels of the PAN band have the poorest possible spectral resolution, and therefore the signals of vibration over-tones and combination bands specific to SOC are diluted into a global attenuation of the spectra. As a result, a saturation of the PAN signal at large SOC concentration is not surprising, and it is very likely that the issue could be overcome by the use of hyperspectral data. Additionally to this spectral consideration, data had been resampled by cubic convolution by the provider. Nearest neighbor would probably have been more adapted to the short-distance gradient of SOC at pre-industrial charcoal kiln sites. We suspect that resampling might have smoothed slightly the minima of reflectance at pre-industrial kiln sites, which might contribute to explain why the attenuation of reflectance decreases at high carbon concentration. Another possible factor of explanation is related to the size of charcoal particles. Charcoal residues in the soil of pre-industrial kiln sites vary from microscopic particles to coarse fragments of up to > 1 cm large. The size of fire residues has been shown to decrease with time along a chronosequence of land use change from forest to cropland (Nguyen et al., 2008), possibly because of a faster fragmentation of particles under tillage. A higher level of fragmentation corresponds to a larger external surface area and therefore possibly to a larger attenuation of the reflectance for a given amount of charcoal. Sampling was a crucial point of the methodology which as implications on the accuracy and the robustness of the calibration model. First, we decided to sample the entire depth (0–25 cm) of the plow layer whereas remote sensing data provides information for soil surface only. This approach supposes that SOC concentration is relatively homogeneous on all the depth of the plow layer, which implies that surface reflectance is representative of carbon concentration of the plow layer. This assumption was essential for the calculation of SOC stocks in the plow layer based on soil reflectance. Moreover, the date of acquisition of remote sensing data and soil sampling were uncoupled. Therefore, sampling of all the depth of the plow layer aimed to increase the robustness of the model calibrated with satellite archives that do not correspond exactly with the date of sampling, assuming that the average concentration of SOC in the plow layer is less variable over time than that of surface soil. A main advantage of satellite with respect to airborne imagery is the revisit capability of the sensor. Consequently, it is interesting to have a robust SOC dataset that can be related to several images acquired at different

265

Chapter 11. Carbon stocks by remote sensing dates. Nevertheless, movement of topsoil by tillage erosion is relatively fast in cropland (Figure 11.10), and the more the date of sampling is close to the date of image acquisition, the more the prediction model is expected to be accurate.

Figure 11.10. Pre-industrial charcoal kiln sites with marked striations that illustrate soil movement caused by tillage Sampling aimed to apprehend field variability in a way that suits the information of one pixel. In remote sensing, an area larger than the pixel size (up to three times) is generally sampled to integrate the effect of the point spread function (PSF) of the sensor (which corresponds to the answer of a sensor to a punctual source), or to attenuate the error related to resampling and orthorectification of the image. Nevertheless, SOC concentrations follow a sharp gradient at kiln sites, with large changes occurring over short distances. Therefore, we preferred to restrict the sampling area to the dimension of one pixel only (1 x 1 m) for more reliable SOC concentrations. Nevertheless, sampling of a smaller surface makes the model more sensitive to possible orthorectification problems. The characteristics of the sensor are of great importance. Data from a sensor depend on its spatial, spectral and radiometric resolutions, point spread function and viewing angle at the time of acquisition. In this work, a compromise between spatial and spectral resolution was needed. The red band was the most correlated with SOC content but the panchromatic band had a spatial resolution more adapted to apprehend the strong gradient of SOC at pre-industrial charcoal kiln site. We favored spatial resolution at the expense of spectral resolution. Atmospheric scattering and absorption phenomena interfering with the spectral measurements are considered as a main constraint to the widespread use of remote sensing for soil applications (Ben-Dor et al., 2009). 266

Chapter 11. Carbon stocks by remote sensing

Atmospheric effects can be corrected quite successfully by the empirical line method (Karpouzli and Malthus, 2003) or by physical models such as ATCOR (© Satellite Imaging Corporation). By using planetary reflectance, we neglected atmospheric effects. We justify this choice by the facts that (i) none of the images were affected by clouds covers, (ii) all images were acquired at the same time of the year (April) and (iii) data were normalized by difference with the average background of the field unaffected by charcoal kiln sites. Therefore, we assume that atmospheric corrections had little effect on the accuracy of the predictions.

Evolution of SOC stocks at charcoal kiln site over time of cultivation

Estimates of BC stocks at pre-industrial charcoal kiln sites along a chronosequence of cropping history did not allow drawing any clear conclusion on the effect of cultivation on the residence time of charcoal-C in soil. Average charcoal-C stocks were larger in fields that were subject to the shortest history of cultivation, but differences with sites that had been deforested earlier were small. In particular, two fields among those deforested the most recently contained more SOC per kiln site (3.73 tons) than any other field. If we remove these two fields from the dataset, differences in SOC stocks at kiln site between the three different episodes of deforestation diminish (P = 0.45). This result suggests that the sites might have not contained the same initial amount of charcoal-C. For instance, the number of times that charcoal production have occurred at one kiln site might be larger for some of the sites that remained forested for a longer period of time, given that most kiln sites were re-used periodically. Because of the saturation of the signal of the PAN band for high SOC concentrations, we were unable to apply our model to estimate SOC stocks at kiln site in fields that were converted to cropland more recently, like in the region of Ermeton-sur-Biert, where grasslands for animal breeding prevail on field crops. Consequently, our chronosequence lacks fields that were subject to a short period of cropping after deforestation. Nguyen et al., (2008) showed that major changes in BC stocks occurred in the first 20 years of cultivation of soils deforested by slash-and-burn. Their result suggest that we might have missed some effect of cropping on BC loss occurring on the short-term after deforestation, as we focused only on fields cropped for a long period of time (> 110 years). It is also worth to mention that uncertainty exists about the exact time of cultivation of the different fields. Historical maps provided a range of

267

Chapter 11. Carbon stocks by remote sensing possible dates of deforestation that is relatively large for each time step. For an extreme scenario assuming that fields with the longer history of cultivation were deforested in 1854 (last possible date according to the map of Vandermaelen) and that fields with the shorter history of cultivation were deforested in 1871 (first possible date according to the map of the “dépôt de la guerre”), the minimum possible time difference between fields with the shorter and the longer history of cropping is 17 years, which is a small difference with respect to the total time that has elapsed since deforestation. This consideration further supports the view that possible effects of cropping on the kinetics of BC loss might have been missed by our approach. In the dataset from forest soil, four sites were sampled on Haplic Luvisols (Chapter 7), including one from the surroundings of Gembloux, in one of the last areas that remained forested since the late 18th century. Diameter of this site was 10 m and the charcoal-rich topsoil had a depth of 0.38 m, with an average SOC concentration of 83.1 g kg-1. Bulk density was measured at five locations in the charcoal-rich layer of this site and was 1086 kg m-³ on average, which is remarkably close to the value of 1112 kg m-³ calculated from SOC content of the site by our linear model (equation 7). Making the assumption that depth, SOC concentration and bulk density are constant across the site, we calculated a stock of SOC of 2.69 t. Despite the absence of repetition, this value accords well with SOC stocks estimated by remote sensing in the sites from cropland, even those cultivated since > 150 years. From these results, we cannot conclude that cropping accelerates sensitively the loss of BC from pre-industrial charcoal kiln sites, but we could neither assert that cropping has no effect on the residence time of charcoal in soil. Soil disturbance has been shown to enhance the decomposition of biochar in a 3.2 years incubation experiment in laboratory conditions (Kuzyakov et al., 2009). This might be true for the most labile fraction of charcoal that reacts quite rapidly after introduction to soil (Sagrilo et al., 2014) but might not play a significant role for the more condensed, stable fraction of charcoal that decomposes very slowly in soil. Therefore, the question of an effect of cultivation on kinetics of charcoal loss remains open.

11.5. Conclusion

High resolution VIS-NIR satellite imagery was used successfully to predict SOC stocks at pre-industrial charcoal kiln site for bare fields, provided that SOC content was < 40 g kg-1. It is likely that this limitation might be overcome by VIS-NIR hyperspectral remote sensing, which provides a high density of 268

Chapter 11. Carbon stocks by remote sensing information including vibration over-tones and band combinations specific to organic functional groups. Removing this constraint of SOC concentration would make possible an estimation of SOC stocks at pre-industrial charcoal kiln sites for fields with a short history of cropping, to better assess whether cropping accelerates the loss of charcoal-C from the soil of pre-industrial kiln sites or not. The question remains open at this point.

Acknowledgements

Funds were provided by the General Directorate for Agriculture, Natural Resources and Environment (DGO3) - Public Service of Wallonia. We are particularly grateful to Vincent Brahy and Patrick Engels who made possible the acquisition of DigitalGlobe satellite imagery.

269

Chapter 12. Conclusions and perspectives

Chapter 12. General conclusions and perspectives

The aim of this work was to assess the long-term effect of pre-industrial charcoal kiln sites on properties of temperate soil, to understand long-term implications of a biochar soil amendment in temperate regions. In particular, we investigated how charcoal enrichment affected soil properties in contrasting soil conditions, and how soil conditions influenced the long-term evolution of chemical properties, stability and turnover of charcoal in soil. According to this general objective, four specific objectives were developed: (i) we assessed the magnitude of pre-industrial charcoal production in Wallonia in the late 18th century; (ii) we developed a procedure to identify and quantify charcoal in the soil of pre-industrial charcoal kiln sites; (iii) we assessed the effect of pre-industrial charcoal kiln sites on soil properties, under contrasting soil conditions; and (iv) we evaluated the influence of a land-use change from forest to cropland on the stability and dynamics of SOC in the soil of pre-industrial charcoal kiln sites. The main conclusions of this work are drawn with respect to these four specific objectives.

12.1. Magnitude of pre-industrial charcoal production in Wallonia

The forested area mapped by Ferraris (1770-1778) delimited the former forest area potentially available for charcoal production at the time of the peak of charcoal-based smelting. Accordingly, it was a key resource to orient the prospection for the detection of pre-industrial charcoal kiln sites. With the aim of estimating the number of sites at the scale of Wallonia, remote sensing successfully completed field prospection and allowed a rapid exploration of large areas of land. In areas that had been deforested since the time of charcoal production, the dark color of the soil related to the enrichment with charcoal residues made the detection of the sites possible on bare soil with VIS-NIR orthoimages or high resolution satellite imagery. Under forest, charcoal kiln sites were detected based on characteristic relief features appearing on a high resolution digital elevation model (DEM) derived from light detection and ranging (LiDAR) data. Despite a number of circular objects potentially interfering with the LiDAR detection of the sites (large tree trunks, shell holes, tree falls, burial mounds, backfilled materials, piles of branches after wood harvest, …) we obtained encouraging false positive (15.7 %) and false negative (29.3 %) detection rates. The presence of a dense herbaceous

271

Chapter 12. Conclusions and perspectives vegetation cover, particularly brambles, was identified as a cause of false negative detection in small forest bodies remaining in agricultural areas of the loess belt of Belgium, dominated by field crops. Indeed, a dense herbaceous cover decreases the signal-to-noise ratio of the ground by intercepting a main part of the LiDAR signal. The canopy of some coniferous plantations was also identified as a potential cause of false negative detection because of a strong interception of the LiDAR signal. For this reason, mixed and softwood forests were removed from the sampling area to estimate the number of sites at the scale of Wallonia. In total, we identified charcoal kiln sites in 93.9 % of the sampling plots, with a median site density of 1.2 sites per ha. Only few samples located far away from blast furnaces that were active in the late 18th century (located mainly in the “Entre-Sambre-et-Meuse” and the South of the province of Luxembourg) did not contain pre-industrial charcoal kiln sites. This result accord with historical records indicating that former production of charcoal was intimately related to the pre-industrial steel industry that used only charcoal as a combustible to smelt the iron ore until the early 19th century. The omnipresence of charcoal kiln relics in the forest mapped by Ferraris also supports the idea that a major part of wood resources were allocated to charcoal production in the late 18th century, which accords with the results of the historical approach based on the average production of pig iron of blast furnaces that were active in Wallonia in the late 18th century. In total, we estimated that the pre-industrial steel industry required 325.000 ha of forest to meet its demand for charcoal in 1790, which represents 75 % of the forested area on the map of Ferraris (1770–1778). This highlights the intense pressure exerted on the Walloon forest in the late 18th century through the production of charcoal. The scarcity of wood resources for charcoal production has been cited as one of the main reasons that caused the decline of the charcoal-based smelting in favor of the coal-based smelting. We estimated a total number of pre-industrial charcoal kiln sites of about 450.000 in Wallonia, with about 70 % of sites under forest, 20 % in agricultural areas (cropland or grassland) and about 10 % in artificialized areas. According to SOC stocks estimated for fields of the surrounding of Gembloux, we can reasonably assume that an average charcoal kiln site stores from two to three tons of SOC. This represents from 900.000 to 1.350.000 tons of SOC (or from 3.300.000 to 4.900.000 tons of CO2 equivalents) at the scale of Wallonia, including from 720.000 to 1.080.000 tons of charcoal-C. Given that the soils of Wallonia contain about 163.000.000 tons of SOC in the

272

Chapter 12. Conclusions and perspectives first meter (Van Wesemael & Brahy, 2007), SOC storage in charcoal kiln sites of Wallonia represents 0.55–0.83 % of SOC stocks. Considering that charcoal has a residence time in soil of one to two orders of magnitude longer than that of uncharred SOC, this contribution to soil carbon storage is appreciable.

12.2. Identification and quantification of charcoal in the soil of pre- industrial charcoal kiln sites

Pre-industrial charcoal kiln sites offered a unique opportunity to study the long-term effect of charcoal on the properties of soil in various conditions. Remote sensing data was crossed with the digital soil and geological maps of Wallonia for the detection of pre-industrial kiln sites under contrasting soil conditions, and historical maps were used to trace the land-use history of the sites. As a result, we identified a number of kiln sites under contrasting soil types and land use and a selection of sites was sampled and compared to adjacent reference soil unaffected by charcoal production. To accurately determine how charcoal influence soil properties and SOM dynamics on the long-term, it was crucial to quantify the contribution of charcoal-C to total SOC stocks in the soil of pre-industrial charcoal kiln sites, where charcoal is mixed to uncharred SOM. Quantification of BC in soil is a major challenge because it consists of a continuum of material with heterogeneous properties that overlap with those of uncharred SOM to some extent. Consequently, quantification of BC relies on operational definitions depending on specific objectives defined by researchers from very different fields in atmospheric, soil, sediment and paleoenvironmental sciences. Existing methods record systematic differences, up to more than two orders of magnitude, as they recover a different fraction of the BC continuum. In this work, we have investigated one thermal and one chemical method to discriminate between charcoal-C and uncharred SOC in the soil of pre-industrial charcoal kiln sites. First, we explored differential scanning calorimetry (DSC) as a means to discriminate between charcoal-C and uncharred SOC in pre-industrial charcoal kiln sites. Among the various existing methods to identify BC in soil, dynamic thermal analysis is particularly interesting because it provides a high density of information on the complete thermal continuum of soil organic matter. It also has the advantage to be rapid, inexpensive and reproducible and to require little equipment and sample preparation like any static thermal method. Despite the diversity of soils that were analyzed, total heat released during analysis was strongly linearly related to SOC content, which supports the view that heat fluxes recorded by DSC for the soils of this study can be

273

Chapter 12. Conclusions and perspectives interpreted directly in terms of SOM, without preliminary treatment to remove soil minerals such as gibbsite, kaolinite or halloysite that cause endotherms in the temperature range of combustion of SOM. Pre-industrial charcoals are more thermally stable than uncharred SOM, which reflects that the binding energy of C=C bonds from aromatic rings of charcoal is larger than that of C- C, C-H or C-O bonds from uncharred SOM. Aging was shown to decrease the thermal stability of charcoal, in particular by oxygenation of charcoal particles over time. We also highlighted that soil conditions influence the thermal signature of charcoal, particularly the concentration of Ca that governs the temperature of combustion of the O-rich fraction of charcoal. The variability of the thermal signature of SOM depends on the chemical composition of the samples, but also on experimental parameters such as heating rate, the weight of sample or material of the crucible (Fernández et al., 2010), which is a clear inconvenience of the technique for the characterization of SOM. Despite the fact that charcoal has a thermal footprint different from that of uncharred SOM, both signatures overlap, which stressed the issue of BC quantification in soil that is met with any approach of quantification related to the oxidative resistance of BC. This overlap makes impossible to define of a cut-off temperature value to separate reliably aged charcoal from uncharred SOM. Therefore, we proposed an approach to quantify SOC based on peak height of exotherms attributed to the combustion of charcoal relative to that attributed to the combustion of uncharred SOC. Charcoal-C content estimated by DSC overestimated by more than five times the amount of BPCA-C in soil, which highlights the extent to which BC can be underestimated by widely used BC quantification procedures. The calibration of our peak index has no general character for the quantification of BC because it depends on the shape of the thermogram of uncharred SOM and charcoal. Signature of BC is expected to be highly variable depending on the conditions of production of BC, its degree of degradation in soil and, to some extent, its inorganic composition, as illustrated by our data. However, our approach has the main advantage to consider all forms of C in charcoal, and not only an unknown fraction of aromatic-C from BC. The perspectives of this work to improve the method of quantification of BC by dynamic thermal analysis are (i) to better test the influence of uncharred SOM quality, of mineral soil and of BC quality on the estimation of BC content and (ii) to follow CO2 emissions over the thermal continuum by EGA rather than relying on heat fluxes, to get rid of the influence of soil minerals on the quantification of C pools.

274

Chapter 12. Conclusions and perspectives

Second, we aimed to test whether dichromate oxidation by the Walkley-Black procedure (which is a widely used method to quantify SOC in soil), discriminates between uncharred organic matter and charcoal residues in the topsoil of pre-industrial charcoal kiln sites. To test the influence of aging on the recovery of charcoal, the soil of a currently active charcoal kiln site was also analyzed. A significant fraction of SOC content of pre-industrial charcoal kiln soils was resistant to dichromate oxidation, which provided evidence that charcoal has a greater resistance to dichromate oxidation than uncharred organic matter. Nevertheless, 23.6 % of charcoal-C was recovered by the Walkley-Black procedure in the soil of the currently active kiln site against 65 % for the soil of pre-industrial kiln sites, which indicates that the resistance of charcoal to dichromate oxidation decreases with ageing. The recovery of charcoal increased to 90 % after boiling during digestion, providing evidence that heat catalyzes the oxidation of charcoal. The substantial oxidation of charcoal by dichromate and the variability of recovery according to the degree of alteration of charcoal and conditions of reaction support the idea that the quantification of BC based on its chemical resistance is challenging and can be subject to important biases if calibration is not adapted to the quality of BC of the sample of interest. Because the recovery of BC by the Walkley-Black method is incomplete, the presence of large amounts of BC in soils frequently affected by fires might be a significant cause of underestimation of SOC in regional and global databases.

12.3. The effect of pre-industrial charcoal kiln sites on soil properties, under contrasting soil conditions

The long-term effect of pre-industrial charcoal kiln sites on chemical soil properties of a variety of forest and cropland soils was investigated and related to the content of charcoal-C and uncharred SOC estimated by DSC. To trace the evolution of soil properties over time, soil characteristics of pre-industrial charcoal kiln sites were compared to that of a currently active kiln site. Our data has highlighted that most soil properties affected by the introduction of black carbon (or biochar) to soil evolve over time (Table 12.1). Long-term implications of charcoal enrichment were also shown to depend on land use and soil type to some extent. Main short-term benefits on plant growth from soil amendments with biochar result from liming effect and input of nutrients from wood ash, such as available P, K and Ca. We have shown that the long- term effect of charcoal on nutrient balance and acidity was very different from short-term effects. Under forest, our data revealed that about 200 years of natural re-acidification had decreased topsoil pH at kiln sites from neutral to 275

Chapter 12. Conclusions and perspectives

(very) acidic values, and that base cations from wood ash had been lixiviated to the subsoil. Concentrations of available nutrients were low in the charcoal- rich soil layer, particularly that of available P and exchangeable K+. This underlines that black carbon, because of its resistance to decomposition, does not contribute to renew the stocks of available nutrients by decomposition and therefore cannot substitute uncharred SOM in the biological cycling of nutrients (Jobbagy and Jackson, 2001). In contrast, we have highlighted that CEC of charcoal increases strongly over time in soil by surface oxidation of -1 charcoal particles to reach an average value per unit of C of 414 cmolc kg , which is about twice more elevated than that of uncharred SOM of adjacent -1 reference soils. However, CEC per unit of C up to > 800 cmolc kg have been reported in millennial terra preta soil (Glaser and Birk, 2012), which suggests that CEC might keep increasing over time through the continued slow oxidation of charcoal. This large CEC, mainly related to the creation of carboxyl groups, increases the potential of retention of nutrients of soil. Nevertheless, affinity of cations for carboxylate groups is much variable. The retention of Ca2+ and Mg2+ is promoted by aged charcoal but not that of K+. The small adsorption of K+ on aged charcoal coupled with the relative enrichment by other cations such as Ca2+ and Mg2+ suggests that K might be limiting for plant growth in the long-term in biochar-rich soil with little K supply. The availability of Cu was also decreased at kiln site but, in contrast to K, because it forms strong complexes with Cu, which reduces its availability. Our data also support the view that the benefit of hardwood biochar related to liming from ash is short-lived and becomes less important in soil that is regularly limed. When biochar is amended to an acidic soil in an experiment, we recommend to systematically including a control that was amended with an amount of CaCO3 equal to the quantity of CaCO3 equivalent introduced to soil with biochar, to evaluate whether benefits from biochar are related to liming only or not. It is important to consider the long-term implications of a biochar soil amendment on the balance of nutrients for an appropriate management as the use biochar as a soil amendment might become increasingly common in the future.

276

Table 12.1. Summary of the long-term evolution of charcoal’s effect on soil properties after > 150 years in soil.

Properties of Short-term Long-term charcoal Liming effect from wood ash; pH and base saturation In forest, natural re-acidification occurs, pH and base saturation decrease; in Acidity increase cropland, pH and BS remain high due to liming Organic C:N and C:P Large C:N and C:P ratios Large C:N and C:P ratios ratios Cation exchange CEC is sharply increased; CEC of charcoal is about twice larger than the CEC per Little effect on CEC capacity (CEC) unit of C uncharred SOC In forest, leaching of Ca over re-acidification of soil; content of Ca remains high due Exchangeable Ca2+ Input of Ca from wood ash to strong affinity for charcoal; in cropland, liming maintains high Ca content Exchangeable Mg2+ Input of Mg from wood ash Adsorption of Mg2+ regulated by “classic” properties of ion exchange Exchangeable K+ Large input of K from wood ash Poor affinity of charcoal for K+; preferential lixiviation relative to Ca2+ and Mg2+ Input of available P from wood ash; increase of P availability P availability diminishes over time; in forest, poor P availability in the soil enriched Plant available P in acidic soil related to the liming effect of wood ash with charcoal Little available N or N deficiency related to mineralization of No data for forest sites; in cropland, larger content of N-NO - at kiln site related to Plant available N 3 the N-poor labile fraction of charcoal mineralization of organic N that is larger at kiln site than in adjacent reference soil Input of Zn from wood ash; availability depends on initial soil Available Zn Available Zn remains large properties Input of Cu from wood ash; availability depends on soil Available Cu Decrease of Cu availability; strong complexes with aged charcoal properties; adsorption to phenolic groups of biochar Mineralogy No clear effect on mineralogy No clear effect on mineralogy

CO2 emissions Increases CO2 emissions CO2 emissions governed mainly by uncharred SOC content and pH

277

Chapter 12. Conclusions and perspectives

12.4. Evaluation of the stability and dynamics of charcoal in the soil of pre-industrial charcoal kiln sites, with respect to land use.

The conversion of a land from forest to cropland promotes the decomposition of labile SOM and decreases soil carbon stocks. We hypothesized that the dynamics of charcoal weathering and loss by decomposition might be also accelerated under cropping. This is an important question regarding the contribution of biochar to terrestrial carbon storage because applications of biochar are mainly to cropland soil. The question of the stability and dynamics of SOM at pre-industrial charcoal kiln site was addressed by three different ways (i) a laboratory incubation experiment of forest and cropland soils, (ii) a study of the chemical properties of charcoal particles extracted from the soil of pre-industrial charcoal kiln sites along a chronosequence of land-use change from forest to cropland and (iii) an evaluation of carbon stocks at kiln site by remote sensing.

Emissions of CO2 per unit of SOC were larger for the soils from cropland than for the soils from forest, which is attributed to the better soil fertility related to the application of organic and inorganic amendments and fertilizers.

Uncharred SOC content and pH explained a main part of the variability in CO2 emissions, regardless of the presence of charcoal. Clearly, an incubation experiment of a few months was too short to capture the turnover of charcoal-

C in soil, most likely because emissions of CO2 from charcoal are negligible with respect to that from uncharred SOM. The effect of charcoal enrichment on microbial community structure was small in forest and close to zero in cropland, where soil conditions are modified by the application of organic and inorganic fertilizers. This supports the view that on the long-term, when the labile fraction of charcoal has been degraded, the effect of charcoal on soil microbial properties is mainly indirect, related to a modification of trophic conditions such as acidity and, possibly, the availability of nutrients. The study of charcoal particles sampled along the chronosequence of land use change from forest to cropland has demonstrated that cropping had some effect on chemical properties of charcoal. Cultivation increased association of charcoal with soil minerals, which is favored by deprotonation of carboxylic acids under liming, thereby enhancing the reactivity of charcoal towards mineral surfaces. The large specific surface area of charcoal, related to its porosity, promotes the precipitation of 2:1 phyllosilicates and CaCO3. Both ageing and cultivation decreased the resistance of charcoal to dichromate oxidation, related to an increase of the H/C of charcoal. Differential scanning

278

Chapter 12. Conclusions and perspectives calorimetry revealed the presence of three fractions of distinct thermal resistance. Saturation of carboxylate groups with Ca2+ under liming decreased thermal resistance of the O-rich, less thermally stable fraction of charcoal. This fraction decreased over time of cultivation, leading to the relative increase of the thermally most stable fraction of charcoal. This might result from the preferential loss of the O-rich fraction or the slowdown of charcoal from oxidation by association with minerals. Our results highlighted that land use significantly affects the properties of BC through the modification of soil conditions, which might influence the kinetics of BC loss from soil. This decrease of chemical stability of charcoal particles suggests that cultivation might accelerate the loss of charcoal from soil by increased mineralization rates. Nevertheless, the evaluation of SOC stocks at kiln site by remote sensing for a selection of fields with contrasting history of cropping in the surroundings of Gembloux have not firmly indicated that a land use change from forest to cropland accelerated the loss of charcoal. Nonetheless, the chronosequence was limited to fields with a long history of cropping due to a limitation of our approach by remote sensing (the model to estimate SOC based on reflectance of satellite imagery was reliable only for small SOC concentrations). Therefore, the question of an effect of cultivation on kinetics of charcoal loss remains open.

279

Chapter 12. Conclusions and perspectives

12.5. Perspectives

Our data from pre-industrial charcoal kiln sites support the view that most effects of biochar on soil properties evolve over time. Therefore, in the perspective of amending soils with biochar, it is critical to verify that biochar contributes positively to soil ecosystem services on the long-term, since biochar is very persistent in soil. In this goal, Pre-industrial charcoal kiln sites in cropland of Wallonia provide the opportunity to address many unanswered questions, as they are natural models of soils amended with biochar >150 years ago, with several replicates in the same field. For example, they make possible to evaluate the influence of long-term charcoal enrichment on the emissions of greenhouse gases such as NO2 and on the biogeochemical cycling of elements in field conditions. Obviously, it is of prime importance to better assess how plant growth and crop yields respond to biochar on the long-term before allowing its large scale application to soil. There is a need to ensure that short-term benefits will persist over time, or at least that biochar will not decrease soil fertility on the long-term. We have mentioned the potential risk of unbalancing plant nutrition for some elements such as K, P and Cu under inadequate soil type or land management in presence of aged charcoal. We also need further insight into the interrelationship between biochar quality, application rates and soil and climatic conditions to optimize crop yields and environmental benefits related to soil carbon sequestration. For maximizing the return from biochar application to soil in terms of food security, soil amendment with biochar should be combined with (in)organic fertilizers to mimic the genesis of Amazonian terra preta, a model for sustainable agriculture in the humid tropics (Glaser et al., 2001) and possibly in temperate regions (Wiedner et al., 2015). Productivity and yields of different plant species could be mapped by precision agriculture for fields containing charcoal kiln sites and compared to SOC maps to document the response of the plant according to the amount of charcoal. It is very likely that some effects will be measured, as the footprint of charcoal kiln sites appears through the vegetation cover of some fields (Figure 12.1).

280

Chapter 12. Conclusions and perspectives

Figure 12.1. Visible impact of pre-industrial charcoal kiln sites (indicated by the black arrows) on the vegetation cover of a crop in Fosses-la-Ville, Belgium.

The concept of biochar application to soil relies on “a win-win-win scenario for simultaneously producing bioenergy, permanently sequestering carbon in soil and improving sustainably soil fertility” (Laird, 2008). If the objective of improving sustainably soil fertility is not met, it is quite clear that biochar strategy is not worth it.

281

References

Abiven, S., Schmidt, M.W.I., Lehmann, J., 2014. Biochar by design. Nature Geoscience 7, 326–327. doi:10.1038/ngeo2154 Aciego Pietri, J.C., Brookes, P.C., 2008. Relationships between soil pH and microbial properties in a UK arable soil. Soil Biology and Biochemistry 40, 1856–1861. doi:10.1016/j.soilbio.2008.03.020 Adler, E., 1977. Lignin chemistry-past, present and future. Wood Science and Technology 11, 169–218. doi:10.1007/BF00365615 Ameloot, N., Sleutel, S., Das, K.C., Kanagaratnam, J., de Neve, S., 2015. Biochar amendment to soils with contrasting organic matter level: Effects on N mineralization and biological soil properties. GCB Bioenergy 7, 135–144. doi:10.1111/gcbb.12119 Anne, P., 1945. Sur le dosage rapide du carbone organique des sols. Annales Agronomiques 15, 161–172. Antal, M.J., Grønli, M., 2003. The Art, Science, and Technology of Charcoal Production †. Industrial & Engineering Chemistry Research 42, 1619– 1640. doi:10.1021/ie0207919 Ascough, P.L., Bird, M.I., Francis, S.M., Thornton, B., Midwood, A.J., Scott, A.C., Apperley, D., 2011. Variability in oxidative degradation of charcoal: Influence of production conditions and environmental exposure. Geochimica et Cosmochimica Acta 75, 2361–2378. doi:10.1016/j.gca.2011.02.002 Baldock, J.A., Smernik, R.J., 2002. Chemical composition and bioavailability of thermally altered Pinus resinosa (Red pine) wood. Organic Geochemistry 33, 1093–1109. doi:10.1016/S0146-6380(02)00062-1 Bartholomeus, H.M., Schaepman, M.E., Kooistra, L., Stevens, a., Hoogmoed, W.B., Spaargaren, O.S.P., 2008. Spectral reflectance based indices for soil organic carbon quantification. Geoderma 145, 28–36. doi:10.1016/j.geoderma.2008.01.010 Bates, D., Machler, M., Bolker, B.M., Walker, S.C., 2014. Fitting Linear Mixed-Effects Models using lme4. Journal of Statistical Software 67, 1– 48. doi:10.18637/jss.v067.i01 Batjes, N.H., 2016. Harmonized soil property values for broad-scale modelling (WISE30sec) with estimates of global soil carbon stocks. Geoderma 269, 61–68. doi:10.1016/j.geoderma.2016.01.034 Batjes, N.H., 2014. Total carbon and nitrogen in the soils of the world. European Journal of Soil Science 65, 10–21. doi:10.1111/ejss.12114_2 Batjes, N.H., 1996. Total carbon and nitrogen in the soils of the world. 282

References

European Journal of Soil Science 47, 151–163. Baveye, P.C., Baveye, J., Gowdy, J., 2016. Soil “Ecosystem” Services and Natural Capital: Critical Appraisal of Research on Uncertain Ground. Frontiers in Environmental Science 4, 1–49. doi:10.3389/fenvs.2016.00041 Belhoste, J.-F., 1990. Une sylviculture pour les forges, XVIe–XIXe siècles., in: Woronoff, D. (Ed.), Forges et Forêts. Recherches Sur La Consommation Proto-Industrielle de Bois. Editions de l’Ecole des Hautes Etudes en Sciences Sociales, p. 219–261. Ben-Dor, E., Chabrillat, S., Demattê, J. a. M., Taylor, G.R., Hill, J., Whiting, M.L., Sommer, S., 2009. Using Imaging Spectroscopy to study soil properties. Remote Sensing of Environment 113, S38–S55. doi:10.1016/j.rse.2008.09.019 Bennett, R., Welham, K., Hill, R.A., Ford, A., 2012. A Comparison of Visualization Techniques for Models Created from Airborne Laser Scanned Data. Archeological Prospection 19, 41–48. doi:10.1002/arp.1414 Biederman, L.A., Harpole, W.S., 2013. Biochar and its effects on plant productivity and nutrient cycling: a meta-analysis. GCB Bioenergy 5, 202–214. doi:10.1111/gcbb.12037 Biester, H., Knorr, K.H., Schellekens, J., Basler, A., Hermanns, Y.M., 2014. Comparison of different methods to determine the degree of peat decomposition in peat bogs. Biogeosciences 11, 2691–2707. doi:10.5194/bg-11-2691-2014 Bird, M.I., Cali, J.A., 1998. A million-year record of fire in sub- Saharan Africa. Nature 394, 767–769. Bird, M.I., Gröcke, D.R., 1997. Determination of the abundance and carbon isotope composition of elemental carbon in sediments. Geochimica et Cosmochimica Acta 61, 3413–3423. doi:10.1016/S0016- 7037(97)00157-9 Bird, M.I., Wynn, J.G., Saiz, G., Wurster, C.M., McBeath, A., 2015. The Pyrogenic Carbon Cycle. Annual Review of Earth and Planetary Sciences 43, 273–298. doi:10.1146/annurev-earth-060614-105038 Blackwell, P., Krull, E., Butler, G., Herbert, A., Solaiman, Z., 2010. Effect of banded biochar on dryland wheat production and fertiliser use in southwestern Australia: An agronomic and economic perspective. Australian Journal of Soil Research 48, 531–545. doi:10.1071/SR10014 Bohn, U., Neuhäusl, R., Gollub, G., Hettwer, C., Neuhäuslová, Z., Raus, T., Schlüter, H., Weber, H., 2003. Map of the Natural Vegetation of Europe (Scale 1 : 2 500 000). Landwirtschaftsverlag, Münster.

283

References

Bollandsås, O.M., Risbøl, O., Ene, L.T., Nesbakken, A., Gobakken, T., Næsset, E., 2012. Using airborne small-footprint laser scanner data for detection of cultural remains in forests: An experimental study of the effects of pulse density and DTM smoothing. Journal of Archaeological Science 39, 2733–2743. doi:10.1016/j.jas.2012.04.026 Borchard, N., Ladd, B., Eschemann, S., Hegenberg, D., Möseler, B.M., Amelung, W., 2014. Black carbon and soil properties at historical charcoal production sites in Germany. Geoderma 232–234, 236–242. doi:10.1016/j.geoderma.2014.05.007 Braadbaart, F., Poole, I., van Brussel, A.A., 2009. Preservation potential of charcoal in alkaline environments: an experimental approach and implications for the archaeological record. Journal of Archaeological Science 36, 1672–1679. doi:10.1016/j.jas.2009.03.006 Brodowski, S., Amelung, W., Haumaier, L., Abetz, C., Zech, W., 2005a. Morphological and chemical properties of black carbon in physical soil fractions as revealed by scanning electron microscopy and energy- dispersive X-ray spectroscopy. Geoderma 128, 116–129. doi:10.1016/j.geoderma.2004.12.019 Brodowski, S., Rodionov, A., Haumaier, L., Glaser, B., Amelung, W., 2005b. Revised black carbon assessment using benzene polycarboxylic acids. Organic Geochemistry 36, 1299–1310. doi:10.1016/j.orggeochem.2005.03.011 Brown, R., 2009. Biochar Production Technology, in: Lehmann, J., Joseph, S. (Eds.), Biochar for Environmental Management: Science and Technology. Earthscan, London, pp. 127–146. Bruun, S., Clauson-Kaas, S., Bobuľská, L., Thomsen, I.K., 2014. Carbon dioxide emissions from biochar in soil: role of clay, microorganisms and carbonates. European Journal of Soil Science 65, 52–59. doi:10.1111/ejss.12073 Bryson, C.E., 1987. Surface potential control in XPS. Surface Science 189– 190, 50–58. doi:10.1016/S0039-6028(87)80414-4 Budai, A., Wang, L., Gronli, M., Strand, L.T., Antal, M.J., Abiven, S., Dieguez-Alonso, A., Anca-Couce, A., Rasse, D.P., 2014. Surface Properties and Chemical Composition of Corncob and Miscanthus Biochars: Effects of Production Temperature and Method. Journal of Agricultural and Food Chemistry 62, 3791–3799. doi:10.1021/jf501139f Budai, A., Zimmerman, A.R., Cowie, A.L., Webber, J.B.W., Singh, B.P., Glaser, B., Masiello, C.A., Andersson, D., Shields, F., Lehmann, J., Camps Arbestain, M., Williams, M., Sohi, S., Joseph, S., 2013. Biochar Carbon Stability Test Method : An assessment of methods to determine biochar carbon stability [WWW Document]. International Biochar

284

References

Initiative. URL www.biochar- international.org/sites/default/files/IBI_Report_Biochar_Stability_Test _Method_Final.pdf Calvelo Pereira, R., Camps Arbestain, M., Kaal, J., Vazquez Sueiro, M., Sevilla, M., Hindmarsh, J., 2014. Detailed carbon chemistry in charcoals from pre-European Māori gardens of New Zealand as a tool for understanding biochar stability in soils. European Journal of Soil Science 65, 83–95. doi:10.1111/ejss.12096 Cayuela, M.L., van Zwieten, L., Singh, B.P., Jeffery, S., Roig, A., Sánchez- Monedero, M.A., 2014. Biochar’s role in mitigating soil nitrous oxide emissions: A review and meta-analysis. Agriculture, Ecosystems & Environment 191, 5–16. doi:10.1016/j.agee.2013.10.009 Certini, G., 2005. Effects of fire on properties of forest soils: a review. Oecologia 143, 1–10. doi:10.1007/s00442-004-1788-8 Chan, K.Y., Van Zwieten, L., Meszaros, I., Downie, A., Joseph, S., 2007. Agronomic values of green waste biochar as a soil amendment. Australian Journal of Soil Research 45, 629–634. doi:10.1071/SR07109 Cheng, C.-H., Lehmann, J., 2009. Ageing of black carbon along a temperature gradient. Chemosphere 75, 1021–7. doi:10.1016/j.chemosphere.2009.01.045 Cheng, C.-H., Lehmann, J., Engelhard, M.H., 2008a. Natural oxidation of black carbon in soils: Changes in molecular form and surface charge along a climosequence. Geochimica et Cosmochimica Acta 72, 1598– 1610. doi:10.1016/j.gca.2008.01.010 Cheng, C.-H., Lehmann, J., Thies, J.E., Burton, S.D., 2008c. Stability of black carbon in soils across a climatic gradient. Journal of Geophysical Research 113, G02027. doi:10.1029/2007JG000642 Cheng, C.-H., Lehmann, J., Thies, J.E., Burton, S.D., Engelhard, M.H., 2006. Oxidation of black carbon by biotic and abiotic processes. Organic Geochemistry 37, 1477–1488. doi:10.1016/j.orggeochem.2006.06.022 Cheng, C.-H., Lin, T.-P., Lehmann, J., Fang, L.-J., Yang, Y.-W., Menyailo, O. V., Chang, K.-H., Lai, J.-S., 2014. Sorption properties for black carbon (wood char) after long term exposure in soils. Organic Geochemistry 70, 53–61. doi:10.1016/j.orggeochem.2014.02.013 Chestworth, W., 2004. REDOX , SOILS , AND CARBON SEQUESTRATION. Edafologia 11, 37–43. Chia, C.H., Downie, A., Munroe, P., 2015. Characteristics of biochar: physical and structural properties, in: Lehmann, J., Joseph, S. (Eds.), Biochar For Environmental Management: Science, Technology and Implementation. Earthscan, Routledge, London, pp. 89–109.

285

References

Chidumayo, E.N., 1994. Effects of wood carbonization on soil and initial development of seedlings in miombo woodland, Zambia. Forest Ecology & Management 70, 353–357. Cohen-Ofri, I., Weiner, L., Boaretto, E., Mintz, G., Weiner, S., 2006. Modern and fossil charcoal: aspects of structure and diagenesis. Journal of Archaeological Science 33, 428–439. doi:10.1016/j.jas.2005.08.008 Cottenie, A., Kiekens, L., Verloo, M., 1975. Principles of soil and substrate analysis with regard to mobility of nutrient elements. Pédologie 25, 134– 142. Crane-Droesch, A., Abiven, S., Jeffery, S., Torn, M.S., 2013. Heterogeneous global crop yield response to biochar: a meta-regression analysis. Environmental Research Letters 8, 44049. doi:10.1088/1748- 9326/8/4/044049 Cressler, W.L., 2001. Evidence of Earliest Known Wildfires. Palaios 16, 171– 174. Criscuoli, I., Alberti, G., Baronti, S., Favilli, F., Martinez, C., Calzolari, C., Pusceddu, E., Rumpel, C., Viola, R., Miglietta, F., 2014. Carbon sequestration and fertility after centennial time scale incorporation of charcoal into soil. PloS One 9, 1–11. doi:10.1371/journal.pone.0091114 Croft, H., Kuhn, N.J., Anderson, K., 2012. On the use of remote sensing techniques for monitoring spatio-temporal soil organic carbon dynamics in agricultural systems. Catena 94, 64–74. doi:10.1016/j.catena.2012.01.001 Cross, A., Sohi, S.P., 2011. The priming potential of biochar products in relation to labile carbon contents and soil organic matter status. Soil Biology and Biochemistry 43, 2127–2134. doi:10.1016/j.soilbio.2011.06.016 Czimczik, C.I., Masiello, C.A., 2007. Controls on black carbon storage in soils. Global Biogeochemical Cycles 21, GB3005. doi:10.1029/2006GB002798 Czimczik, C.I., Preston, C.M., Schmidt, M.W.I., Schulze, E.-D., 2003. How surface fire in Siberian Scots pine forests affects soil organic carbon in the forest floor: Stocks, molecular structure, and conversion to black carbon (charcoal). Global Biogeochemical Cycles 17, GB001956. doi:10.1029/2002GB001956 Davidson, E.A., Janssens, I.A., Marks, D., Murdock, M., Ahl, R.S., Woods, S.W., Zuuring, H.R., Balisky, A.C., Burton, P.J., Bardgett, R.D., Wardle, D.A., Bond-Lamberty, B., Wang, C., Gower, S.T., Briggs, M.A., Lautz, L.K., McKenzie, J.M., Gordon, R.P., Hare, D.K., Brown, E., Pregitzer, S., Reed, D., Burton, J., Burnett, B.N., Meyer, G.A., McFadden, L.D., Buttle, J.M., Inc, C.S., Chauvin, G.M., Flerchinger, 286

References

G.N., Link, T.E., Marks, D., Winstral, A.H., Seyfried, M.S., Chorover, J., Troch, P.A., Rasmussen, C., Brooks, P.D., Pelletier, J.D., Breshears, D.D., Ciocca, F., Lunati, I., Giesen, N. van de, Parlange, M.B., Clark, P.E., Seyfried, M.S., Harris, B., Dakin, J.P., Pratt, D.J., Bibby, G.W., Ross, J.N., Davidson, E.A.J., Janssens, I.A., Dornbush, M.E., Raich, J.W., Ebel, B.A., Ebel, B.A., Hinkley, E.S., Martin, D.A., Flerchinger, G.N., Cooley, K.R., Hanson, C.L., Seyfried, M.S., Flerchinger, G.N., Seyfried, M.S., Groffman, P., Driscoll, C., Fahey, T., Hardy, J., Fitzhugh, R., Tierney, G., Gutiérrez-Jurado, H.A., Vivoni, E.R., Cikoski, C., Harrison, J.B.J., Bras, R.L., Istanbulluoglu, E., Hanson, C.L., Pierson, F.B., Johnson, G.L., Hardy, J., Groffman, P., Fitzhugh, R., Henry, K., Welman, A., Demers, J., Scientific, H., Hausner, M.B., Suárez, F., Glander, K.E., Giesen, N. van de, Selker, J.S., Tyler, S.W., Hillel, D., Jenny, H., Jonas, T., Rioxen, C., Sturm, M., Stoeckli, V., Kang, S., Kim, S., Ho, S., Lee, D., Kirchner, J.W., Kurashima, T., Horiguchi, T., Tateda, M., Leguizamón, E.S., Fernandez-Quintanilla, C., Barroso, J., Gonzalez-Andujar, J.L., Link, T.E., Flerchinger, G.N., Unsworth, M.H., Marks, D., Lutz, J.A., Martin, K.A., Lundquist, J.D., Mellander, P.-E., Stähli, M., Gustafsson, D., Bishop, K., Minasny, B., Whelan, B.M., Triantafilis, J., McBratney, A.B., Moffett, K., Tyler, S.W., Torgersen, T., Menon, M., Selker, J.S., Gorelick, S.M., Persaud, N., Chang, A.C., Philip, J.R., Prasad, R., Tarboton, D.G., Liston, G.E., Luce, C.H., Seyfried, M.S., Rao, P.S.C., Wagenet, R.J., Rasmussen, C., Brantley, S., Richter, D.D.B., Blum, A., Dixon, J., White, A.F., Rose, L., Krause, S., Cassidy, N.J., Schaetzl, R.J., Knapp, B.D., Isard, S.A., Selker, J.S., Selker, J.S., Thevenaz, L., Huwald, H., Mallet, A., Luxemberg, W., Giesen, N. van de, Selker, J.S., Tyler, S., Giesen, N. van de, Selker, J.S., Giesen, N. van de, Westhoff, M., Luxemburg, W., Parlange, M.B., Seyfried, M.S., Seyfried, M.S., Flerchinger, G.N., Murdock, M.D., Hanson, C.L., Seyfried, M.S., Grant, L.E., Marks, D., Winstral, A., McNamara, J., Seyfried, M.S., Murdock, M.D., Staff, S.S., Stockmann, U., Adams, M.A., Crawford, J.W., Field, D.J., Henakaarchchi, N., Jenkins, M., Tian, Y.Q., Davies-Colley, R.J., Gong, P., Thorrold, B.W., Tyler, S.W., Burak, S.A., McNamara, J.P., Lamontagne, A., Selker, J.S., Dozier, J., Tyler, S.W., Selker, J.S., Hausner, M.B., Hatch, C.E., Torgersen, T., Thodal, C.E., Schladow, S.G., Westhoff, M.C., Savenije, H.H.G., Luxemburg, W.M.J., Stelling, G.S., Giesen, N.C. van de, Selker, J.S., Wilding, L.P., Lin, H., Wilson, D.J., Western, A.W., Grayson, R.B., Wundram, D., Paper, R., Loffler, J., 2006. Temperature sensitivity of soil carbon decomposition and feedbacks to climate change. Nature 440, 165–73. doi:10.1038/nature04514 Day, P.R., 1965. Particle fractionation and particle size analysis., in: Black, C.A. (Ed.), Methods of Soil Analysis. Part 1. Physical and Mineralogical

287

References

Properties, Including Statistics of Measurement and Sampling. American Society of Agronomy, Soil Science Society of America, Madison, WI, USA, pp. 545–567. De la Rosa, J.M., Knicker, H., López-Capel, E., Manning, D.A.C., González- Perez, J.A., González-Vila, F.J., 2008. Direct Detection of Black Carbon in Soils by Py-GC/MS, Carbon-13 NMR Spectroscopy and Thermogravimetric Techniques. Soil Science Society of America Journal 72, 258. doi:10.2136/sssaj2007.0031 Deforce, K., Boeren, I., Adriaenssens, S., Bastiaens, J., De Keersmaeker, L., Haneca, K., Tys, D., Vandekerkhove, K., 2013. Selective woodland exploitation for charcoal production. A detailed analysis of charcoal kiln remains (ca. 1300–1900 AD) from Zoersel (northern Belgium). Journal of Archaeological Science 40, 681–689. doi:10.1016/j.jas.2012.07.009 Dell’Abate, M.T., Benedetti, A., Trinchera, A., Dazzi, C., 2002. Humic substances along the profile of two Typic Haploxerert. Geoderma 107, 281–296. doi:10.1016/S0016-7061(01)00153-7 Delvaux, A., 1998. Les charbonniers d’antan. Forêt Wallonne 38, 2–5. Demeyer, A., Nkana, J.C.V., Verloo, M.G., 2001. Characteristics of wood ash and infuence on soil properties and nutrient uptake : an overview. Bioresource Technology 77, 287–295. Devereux, B., Amable, G.S., Crow, P., 2008. Visualisation of LiDAR terrain models for archaeological feature detection. Antiquity 82, 470–479. Dorban, M., 1988. Trois siècles de consommation forestière dans le Duché de Luxembourg 1500–1830 : un bilan sous le régime français (Département des forêts)., in: Woronoff, D. (Ed.), Révolution et Espaces Forestiers: Colloque Des 3 et 4 Juin 1987. Editions L’Harmattan, Paris, France. Downie, A.E., Van Zwieten, L., Smernik, R.J., Morris, S., Munroe, P.R., 2011. Terra Preta Australis: Reassessing the carbon storage capacity of temperate soils. Agriculture, Ecosystems & Environment 140, 137–147. doi:10.1016/j.agee.2010.11.020 Dussart, F., Wilmet, J., 1970. Les “aires de faude” dans les défrichements du XIXe siècle en Thiérache (Entre-Sambre-et-Meuse). Leur repérage par photo-interprétation. Bulletin de La Société Géographique de Liège 6, 169–178. Elmquist, M., Gustafsson, O., Andersson, P., 2004. Quantification of sedimentary black carbon using the chemothermal oxidation method : an evaluation of ex situ pretreatments and standard additions approaches. Limnology and Oceanography: Methods 2, 417–427. Emrich, W., 1985. Handbook of Charcoal Making. D. Reidel Publishing Company, Dordrecht, The Netherlands.

288

References

Evrard, R., 1956. Forges anciennes. Editions Solédi, Liège, Belgium. Fabbri, D., Torri, C., Spokas, K.A., 2012. Analytical pyrolysis of synthetic chars derived from biomass with potential agronomic application (biochar). Relationships with impacts on microbial carbon dioxide production. Journal of Analytical and Applied Pyrolysis 93, 77–84. doi:10.1016/j.jaap.2011.09.012 Fairhead, J., Leach, M., 2009. Amazonian dark earths in Africa?, in: Woods, W.I., Teixeira, W.G., Lehmann, J., Steiner, C., WinklerPrins, A.M.G.A., Rebellato, L. (Eds.), Amazonian Dark Earths: Wim Sombroek’s Vision. Springer-Verlag, Heidelberg, Germany, pp. 265–278. Falcão, N.P.S., Clement, C.R., Tsai, S.M., Comerford, N.B., 2009. Pedology , Fertility , and Biology of Central Amazonian Dark Earths, in: Woods, W.I., Teixeira, W.G., Lehmann, J., Steiner, C., WinklerPrins, A.M.G.A., Rebellato, L. (Eds.), Amazonian Dark Earths: Wim Sombroek’s Vision. Springer-Verlag, Heidelberg, Germany, pp. 213–228. Feltz, C., Incourt, A.-F., 1995. Itinéraire de la sidérurgie du XVIe au XXe siècle en Sud-Ardenne et Gaume., Hommes et Paysages. Hommes et Paysages, Société Royale Belge de Géographie, Coédition Fondation Universitaire Luxembourgeoise. Fernández, J.M., Plante, A.F., Leifeld, J., Rasmussen, C., 2010. Methodological considerations for using thermal analysis in the characterization of soil organic matter. Journal of Thermal Analysis and Calorimetry 104, 389–398. doi:10.1007/s10973-010-1145-6 Foard, G., 2001. Medieval woodland, agriculture and industry in Rockingham Forest, Northamptonshire. Medieval Archaeology 45, 41–95. doi:10.1179/med.2001.45.1.41 Foereid, B., Lehmann, J., Major, J., 2011. Modeling black carbon degradation and movement in soil. Plant and Soil 345, 223–236. doi:10.1007/s11104-011-0773-3 Forbes, M.S., Raison, R.J., Skjemstad, J.O., 2006. Formation, transformation and transport of black carbon (charcoal) in terrestrial and aquatic ecosystems. Science of the Total Environment 370, 190–206. doi:10.1016/j.scitotenv.2006.06.007 Francoeur, L.B., Molard, F.E., Lenormand, L.S., Payen, A.M., Robiquet, P.J., 1824. Dictionnaire technologique, ou Nouveau dictionnaire universel des arts et métiers, et de l’économie industrielle et commerciale, Volume 5. Thomine et Fortic, Paris, France. Frey, S.D., Elliott, E.T., Paustian, K., 1999. Bacterial and fungal abundance and biomass in conventional and no-tillage agroecosystems along two climatic gradients. Soil Biology and Biochemistry 31, 573–585. doi:10.1016/S0038-0717(98)00161-8 289

References

Gaudin, S., 1996. Quelques éléments d’histoire forestière et généralités sur la forêt en France et dans le monde. [WWW Document]. BTSA Gestion Forestière. URL http://le.mago.pagesperso- orange.fr/Acrobat/Histoire.pdf (accessed 9.6.16). Gebhardt, A., 2007. Impact of charcoal production activities on soil profiles: the micromorphological point of view. ArchéoSciences 31, 127–136. Gillard, A., 1971. L’industrie du fer dans les localités du comté de Namur et de l’Entre-Sambre-et-Meuse de 1345 à 1600. Editions Crédit communal, Bruxelles. Girard, M.-C., Girard, C.-M., 1989. Télédétection appliquée - Zones tempérées et intertropicales. Masson, Paris. Glaser, B., 2007. Prehistorically modified soils of central Amazonia: a model for sustainable agriculture in the twenty-first century. Philosophical Transactions of the Royal Society B: Biological Sciences 362, 187–196. doi:10.1098/rstb.2006.1978 Glaser, B., Balashov, E., Haumaier, L., Guggenberger, G., Zech, W., 2000. Black carbon in density fractions of anthropogenic soils of the Brazilian Amazon region. Organic Geochemistry 31, 669–678. doi:10.1016/S0146-6380(00)00044-9 Glaser, B., Birk, J.J., 2012. State of the scientific knowledge on properties and genesis of Anthropogenic Dark Earths in Central Amazonia (terra preta de Índio). Geochimica et Cosmochimica Acta 82, 39–51. doi:10.1016/j.gca.2010.11.029 Glaser, B., Haumaier, L., Guggenberger, G., Zech, W., 2001. The “Terra Preta” phenomenon: a model for sustainable agriculture in the humid tropics. Naturwissenschaften 88, 37–41. doi:10.1007/s001140000193 Glaser, B., Haumaier, L., Guggenberger, G., Zech, W., 1998. Black carbon in soils : the use of benzenecarboxylic acids as specific markers. Organic Geochemistry 29, 811–819. Glaser, B., Lehmann, J., Zech, W., 2002. Ameliorating physical and chemical properties of highly weathered soils in the tropics with charcoal - a review. Biology and Fertility of Soils 35, 219–230. doi:10.1007/s00374- 002-0466-4 Glaser, B., Woods, W.I. (Eds.), 2004. Amazonian dark earths: explorations in space and time. Springer, Berlin. Goblet d’Alviella, F., 1927. Histoire des bois et forêts de Belgique, des origines à la fin du régime autrichien. Tome 1. Paris, Ed. Lechevalier ; Bruxelles, Ed. Maurice Lamertin. Goidts, E., van Wesemael, B., 2007. Regional assessment of soil organic carbon changes under agriculture in Southern Belgium (1955–2005).

290

References

Geoderma 141, 341–354. doi:10.1016/j.geoderma.2007.06.013 Goldberg, E.D., 1985. Black carbon in the environment. John Wiley & Sons, New York. Gómez-Luna, B.E., Rivera-Mosqueda, M.C., Dendooven, L., Vázquez- Marrufo, G., Olalde-Portugal, V., 2009. Charcoal production at kiln sites affects C and N dynamics and associated soil microorganisms in Quercus spp. temperate forests of central Mexico. Applied Soil Ecology 41, 50–58. doi:10.1016/j.apsoil.2008.08.007 Gomez, J.D., Denef, K., Stewart, C.E., Zheng, J., Cotrufo, M.F., 2014. Biochar addition rate influences soil microbial abundance and activity in temperate soils. European Journal of Soil Science 65, 28–39. doi:10.1111/ejss.12097 Groenewoudt, B., 2005. Charcoal Burning and Landscape Dynamics in the Early Medieval Netherlands : Arts and Crafts in Medieval Rural Environment. Ruralia VI, 327–337. Grossman, J.M., O’Neill, B.E., Tsai, S.M., Liang, B., Neves, E., Lehmann, J., Thies, J.E., 2010. Amazonian anthrosols support similar microbial communities that differ distinctly from those extant in adjacent, unmodified soils of the same mineralogy. Microbial Ecology 60, 192– 205. doi:10.1007/s00248-010-9689-3 Guan, X.H., Chen, G.H., Shang, C., 2006. Combining kinetic investigation with surface spectroscopic examination to study the role of aromatic carboxyl groups in NOM adsorption by aluminum hydroxide. Journal of Colloid and Interface Science 301, 419–427. doi:10.1016/j.jcis.2006.05.031 Guo, L.B., Gifford, R.M., 2002. Soil carbon stocks and land use change: a meta analysis. Global Change Biology 8, 345–360. doi:DOI 10.1046/j.1354-1013.2002.00486.x Gustafsson, Ö., Bucheli, T.D., Kukulska, Z., Andersson, M., Largeau, C., Rouzaud, J.N., Reddy, C.M., Eglinton, T.I., 2001. Evaluation of a protocol for the quantification of black carbon in sediments. Global Biogeochemical Cycles 15, 881–890. Hamer, U., Marschner, B., Brodowski, S., Amelung, W., 2004. Interactive priming of black carbon and glucose mineralisation. Organic Geochemistry 35, 823–830. doi:10.1016/j.orggeochem.2004.03.003 Hammes, K., Schmidt, M.W.I., 2009. Changes of Biochar in Soil, in: Lehmann, J., Joseph, S. (Eds.), Biochar for Environmental Management: Science and Technology. Earthscan, London, pp. 169–181. Hammes, K., Schmidt, M.W.I., Smernik, R.J., Currie, L.A., Ball, W.P., Nguyen, T.H., Louchouarn, P., Houel, S., Gustafsson, Ö., Elmquist, M.,

291

References

Cornelissen, G., Skjemstad, J.O., Masiello, C. a., Song, J., Peng, P., Mitra, S., Dunn, J.C., Hatcher, P.G., Hockaday, W.C., Smith, D.M., Hartkopf-Fröder, C., Böhmer, A., Lüer, B., Huebert, B.J., Amelung, W., Brodowski, S., Huang, L., Zhang, W., Gschwend, P.M., Flores- Cervantes, D.X., Largeau, C., Rouzaud, J.-N., Rumpel, C., Guggenberger, G., Kaiser, K., Rodionov, A., Gonzalez-Vila, F.J., Gonzalez-Perez, J. a., de la Rosa, J.M., Manning, D. a. C., López-Capél, E., Ding, L., 2007. Comparison of quantification methods to measure fire-derived (black/elemental) carbon in soils and sediments using reference materials from soil, water, sediment and the atmosphere. Global Biogeochemical Cycles 21, GB3016. doi:10.1029/2006GB002914 Hammes, K., Torn, M.S., Lapenas, a. G., Schmidt, M.W.I., 2008. Centennial black carbon turnover observed in a Russian steppe soil. Biogeosciences 5, 1339–1350. doi:10.5194/bg-5-1339-2008 Hansotte, G., 1980. La métallurgie et le commerce international du fer dans les Pays-Bas autrichiens et la Principauté de Liège pendant la seconde moitié du XVIIIe siècle. Académie royale de Belgique, Bruxelles. Hardy, B., Cornelis, J.-T., Houben, D., Lambert, R., Dufey, J.E., 2016. The effect of pre-industrial charcoal kilns on chemical properties of forest soil of Wallonia, Belgium. European Journal of Soil Science 67, 206– 216. doi:10.1111/ejss.12324 Hardy, B., Cornelis, J.-T., Houben, D., Leifeld, J., Lambert, R., Dufey, J., 2017. Evaluation of the long-term effect of biochar on properties of temperate agricultural soil at pre-industrial charcoal kiln sites in Wallonia, Belgium. European Journal of Soil Science 68, 80–89. doi:10.1111/ejss.12395 Hardy, B., Dufey, J., 2012a. Estimation des besoins en charbon de bois et en superficie forestière pour la sidérurgie wallonne préindustrielle (1750- 1830) Première partie : les besoins en charbon de bois. Revue forestière française 477–487. Hardy, B., Dufey, J., 2012b. Estimation des besoins en charbon de bois et en superficie forestière pour la sidérurgie wallonne préindustrielle (1750- 1830) Deuxième partie : les besoins en superficie forestière. Revue forestière française 799–806. Hardy, B., Dufey, J.E., 2015. Les aires de faulde en forêt wallonne: Repérage, morphologie et distribution spatiale. Forêt.nature 135, 21–32. Harvey, O.R., Kuo, L.-J., Zimmerman, A.R., Louchouarn, P., Amonette, J.E., Herbert, B.E., 2012. An index-based approach to assessing recalcitrance and soil carbon sequestration potential of engineered black carbons (biochars). Environmental Science & Technology 46, 1415–21.

292

References

doi:10.1021/es2040398 Haumaier, L., Zech, W., 1995. Black carbon—possible source of highly aromatic components of soil humic acids. Organic Geochemistry 23, 191–196. doi:10.1016/0146-6380(95)00003-W Hbirkou, C., Pätzold, S., Mahlein, A.-K., Welp, G., 2012. Airborne hyperspectral imaging of spatial soil organic carbon heterogeneity at the field-scale. Geoderma 175–176, 21–28. doi:10.1016/j.geoderma.2012.01.017 Hedges, J.I., Eglinton, G., Hatcher, P.G., Kirchman, D.L., Arnosti, C., Derenne, S., Evershed, R.P., Kögel-Knabner, I., de Leeuw, J.W., Littke, R., Michaelis, W., Rullkötter, J., 2000. The molecularly uncharacterized component of nonliving organic matter in natural environments. Organic Geochemistry 31, 945–958. doi:http://dx.doi.org/10.1016/S0146- 6380(00)00096-6 Heidenreich, I., Hess, W.M., Ban, L.L., 1968. A test object and criteria for high resolution electron microscopy. Journal of Applied Christallography 1, 1–19. Heitkötter, J., Marschner, B., 2015. Interactive effects of biochar ageing in soils related to feedstock, pyrolysis temperature, and historic charcoal production. Geoderma 245–246, 56–64. doi:10.1016/j.geoderma.2015.01.012 Helling, C.S., Chesters, G., Corey, R.B., 1964. Contribution of organic matter and clay to soil cation-exchange capacity as affected by the pH of the saturating solution. Soil Science Society of America Journal 28, 517– 520. doi:10.2136/sssaj1964.03615995002800040020x Hernandez-Soriano, M.C., Kerré, B., Goos, P., Hardy, B., Dufey, J., Smolders, E., 2015. Long-term effect of biochar on the stabilization of recent carbon: soils with historical inputs of charcoal. GCB Bioenergy 1–11. doi:10.1111/gcbb.12250 Hesse, R., 2010. LiDAR-derived Local Relief Models - a new tool for archaeological prospection. Archeological Prospection 17, 67–72. doi:10.1002/arp.374 Heuschling, X., Van der Maelen, P., 1838. Essai sur la statistique générale de la Belgique. A l’établissement géographique, Faubourg de Flandre, Bruxelles. Hiederer, R., Köchy, M., 2012. Global soil organic carbon estimates and the harmonized world soil database, EUR 25225EN, Luxembourg. doi:10.2788/13267 Hilscher, A., Knicker, H., 2011. Degradation of grass-derived pyrogenic organic material, transport of the residues within a soil column and

293

References

distribution in soil organic matter fractions during a 28 months microcosm experiment. Organic Geochemistry 42, 42–54. doi:10.1016/j.orggeochem.2010.10.005 Hockaday, W.C., Grannas, A.M., Kim, S., Hatcher, P.G., 2007. The transformation and mobility of charcoal in a fire-impacted watershed. Geochimica et Cosmochimica Acta 71, 3432–3445. doi:10.1016/j.gca.2007.02.023 Houben, D., Meunier, C., Pereira, B., Sonnet, P., 2011. Predicting the degree of phosphorus saturation using the ammonium acetate-EDTA soil test. Soil Use and Management 283–293. doi:10.1111/j.1475- 2743.2011.00353.x Houbrechts, G., Petit, F., 2004. Evolution des techniques sidérurgiques préindustrielles et aperçu des critères de localisation de la métallurgie en “Terre de Durbuy.” Terre de Durbuy 89, 3–29. Houghton, R.A., 2007. Balancing the Global Carbon Budget. Annu. Rev. Earth Planet. Sci 35, 313–47. doi:10.1146/annurev.earth.35.031306.140057 Hoyois, G., 1953. L’Ardenne et l’Ardennais. L’Evolution Economique et Sociale d’une région. Tome 2. Bruxelles-Paris: Éditions universitaires; Gembloux: J.Duculot, imprimeur-éditeur. Hoyois, G., 1949. L’Ardenne et l’Ardennais. L’Evolution Economique et Sociale d’une Région. Tome 1. Bruxelles-Paris: Editions universitaires; Gembloux, Duculot, J., Imprimeur-éditeur. Institut pour la Forêt, 2012. La forêt qui produit, 2012 [WWW Document]. URL http://www.institut-foret.com/-La-foret- qui-produit IPCC, 2014. Climate Change 2014: Synthesis Report., Core Writing Team, R.K. Pachauri and L.A. Meyer (Eds.), Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Geneva, Switzerland,. IUSS Working Group WRB, 2014. World reference base for soil resources 2014. International soil classification system for naming soils and creating legends for soil maps. World Soil Resources Reports N° 106. Food and Agriculture Organization of the United Nations, Rome, Italy. Jacquemin, F., Kervyn, T., Branquart, E., Delahaye, L., Dufrêne, M., Claessens, H., 2014. Les Forêts Anciennes en Wallonie. 1ère Partie: Concepts Généraux. Forêt Wallonne 131, 34–49. Jaffé, R., Ding, Y., Niggemann, J., Vähätalo, A. V, Stubbins, A., Spencer, R.G.M., Campbell, J., Dittmar, T., 2013. Global charcoal mobilization from soils via dissolution and riverine transport to the oceans. Science 340, 345–7. doi:10.1126/science.1231476

294

References

Jeffery, S., Verheijen, F.G.A., van der Velde, M., Bastos, A.C., 2011. A quantitative review of the effects of biochar application to soils on crop productivity using meta-analysis. Agriculture, Ecosystems & Environment 144, 175–187. doi:10.1016/j.agee.2011.08.015 Jin, H., 2010. Characterization of microbial life colonizing bicohar and biochar-amended soils. PhD Dissertation. Cornell University, Ithaca, NY. doi:10.1017/CBO9781107415324.004 Jobbagy, E.G., Jackson, R.B., 2001. The distribution of soils nutrients with depth: Gobal patterns and the imprint of plants. Biogeochemistry 53, 51– 77. doi:10.1023/A:1010760720215 Jobbagy, E.G., Jackson, R.B., 2000. The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological Applications 10, 423–436. Jones, D.L., Murphy, D.V., Khalid, M., Ahmad, W., Edwards-Jones, G., DeLuca, T.H., 2011. Short-term biochar-induced increase in soil CO2 release is both biotically and abiotically mediated. Soil Biology and Biochemistry 43, 1723–1731. doi:10.1016/j.soilbio.2011.04.018 Joseph, S.D., Camps-Arbestain, M., Lin, Y., Munroe, P., Chia, C.H., Hook, J., van Zwieten, L., Kimber, S., Cowie, A., Singh, B.P., Lehmann, J., Foidl, N., Smernik, R.J., Amonette, J.E., 2010. An investigation into the reactions of biochar in soil. Australian Journal of Soil Research 48, 501– 515. doi:10.1071/SR10009 Kalinichev, A.G., Kirkpatrick, R.J., 2007. Molecular dynamics simulation of cationic complexation with natural organic matter. European Journal of Soil Science 58, 909–917. doi:10.1111/j.1365-2389.2007.00929.x Kardol, P., Wardle, D. a., 2010. How understanding aboveground- belowground linkages can assist restoration ecology. Trends in Ecology and Evolution 25, 670–679. doi:10.1016/j.tree.2010.09.001 Karpouzli, E., Malthus, T., 2003. The empirical line method for the atmospheric correction of IKONOS imagery. International Journal of Remote Sensing 24, 1143–1150. doi:10.1080/0143116021000026779 Kasozi, G.N., Nkedi-kizza, P., Gao, B.I.N., 2010. Catechol and Humic Acid Sorption onto a Range of Laboratory- Produced Black Carbons (Biochars). Environmental Science & Technology 44, 6189–6195. Keiluweit, M., Nico, P.S., Johnson, M.G., Kleber, M., 2010. Dynamic molecular structure of plant biomass-derived black carbon (biochar). Environmental Science & Technology 44, 1247–53. doi:10.1021/es9031419 Kerré, B., Bravo, C.T., Leifeld, J., Cornelissen, G., Smolders, E., 2016. Historical soil amendment with charcoal increases sequestration of non-

295

References

charcoal carbon : a comparison among methods of black carbon quantification. European Journal of Soil Science 67, 324–331. doi:10.1111/ejss.12338 Kerré, B., Willaert, B., Smolders, E., 2017. Lower residue decomposition in historically charcoal-enriched soils is related to increased adsorption of organic matter. Soil Biology and Biochemistry 104, 1–7. doi:10.1016/j.soilbio.2016.10.007 Kervyn, T., Jacquemin, F., Branquart, E., Delahaye, L., Dufrêne, M., Claessens, H., 2014. Les forêts anciennes en Wallonie. 2ème partie : Cartographie. Forêt Wallonne 133, 38–52. Kim, J.S., Sparovek, G., Longo, R.M., De Melo, W.J., Crowley, D., 2007. Bacterial diversity of terra preta and pristine forest soil from the Western Amazon. Soil Biology and Biochemistry 39, 684–690. doi:10.1016/j.soilbio.2006.08.010 Kleber, M., Eusterhues, K., Keiluweit, M., Mikutta, C., Mikutta, R., Nico, P.S., 2015. Mineral-Organic Associations: Formation, Properties, and Relevance in Soil Environments. Advances in Agronomy 130, 1–140. doi:10.1016/bs.agron.2014.10.005 Knicker, H., 2011a. Pyrogenic organic matter in soil: Its origin and occurrence, its chemistry and survival in soil environments. Quaternary International 243, 251–263. doi:10.1016/j.quaint.2011.02.037 Knicker, H., 2011b. Solid state CPMAS 13C and 15N NMR spectroscopy in organic geochemistry and how spin dynamics can either aggravate or improve spectra interpretation. Organic Geochemistry 42, 867–890. doi:10.1016/j.orggeochem.2011.06.019 Knicker, H., 2010. “Black nitrogen” – an important fraction in determining the recalcitrance of charcoal. Organic Geochemistry 41, 947–950. doi:10.1016/j.orggeochem.2010.04.007 Knicker, H., Müller, P., Hilscher, A., 2007. How useful is chemical oxidation with dichromate for the determination of “Black Carbon” in fire-affected soils? Geoderma 142, 178–196. doi:10.1016/j.geoderma.2007.08.010 Knicker, H., Wiesmeier, M., Dick, D.P., 2008. A simplified method for the quantification of pyrogenic organic matter in grassland soils via chemical oxidation. Geoderma 147, 69–74. doi:10.1016/j.geoderma.2008.07.008 Kokalj, Z., Zaksek, K., Ostir, K., 2011. Application in sky-view factor for the visualisation of historic landscape features in lidar-derived relief models. Antiquity 85, 263–273. Kramer, M.G., Sanderman, J., Chadwick, O. a., Chorover, J., Vitousek, P.M., 2012. Long-term carbon storage through retention of dissolved aromatic

296

References

acids by reactive particles in soil. Global Change Biology 18, 2594– 2605. doi:10.1111/j.1365-2486.2012.02681.x Kucerík, J., Kovár, J., Pekar, M., 2004. Thermoanalytical investigation of lignite humic acids fractions. Journal of Thermal Analysis and Calorimetry 76, 55–65. doi:10.1023/B:JTAN.0000027803.24266.48 Kuhlbusch, T.A.J., 1998. Black carbon and the carbon cycle. Science 280, 1903–4. Kuhlbusch, T.A.J., Crutzen, P.J., 1996. Black carbon, the global carbon cycle, and atmospheric carbon dioxide. Biomass Burning and Global Change 1, 160–69. Kurth, V.J., MacKenzie, M.D., DeLuca, T.H., 2006. Estimating charcoal content in forest mineral soils. Geoderma 137, 135–139. doi:10.1016/j.geoderma.2006.08.003 Kuznetsova, A., Brockhoff, P.B., Christensen, R.H.B., 2015. lmerTest: Tests in Linear Mixed Effects Models. R package version 2.0-29. [WWW Document]. URL https://cran.r- project.org/web/packages/lmerTest/index.html Kuzyakov, Y., Subbotina, I., Chen, H., Bogomolova, I., Xu, X., 2009. Black carbon decomposition and incorporation into soil microbial biomass estimated by 14C labeling. Soil Biology and Biochemistry 41, 210–219. doi:10.1016/j.soilbio.2008.10.016 Laird, D.A., 2008. The charcoal vision: a win-win-win scenario for simultaneously producing bioenergy, permanently sequestering carbon, while improving soil and water quality. Agronomy Journal 100, 178– 181. doi:10.2134/agronj2007.0161 Laird, D.A., Chappell, M.A., Martens, D.A., Wershaw, R.L., Thompson, M., 2008. Distinguishing black carbon from biogenic humic substances in soil clay fractions. Geoderma 143, 115–122. doi:10.1016/j.geoderma.2007.10.025 Lakanen, E., Erviö, R., 1971. A comparison of eight extractans for the determination of plant available micronutrients in soils. Acta Agralia Fennica 123, 223–232. Lambrechts, T., Gustot, Q., Couder, E., Houben, D., Iserentant, A., Lutts, S., 2011. Comparison of EDTA-enhanced phytoextraction and phytostabilisation strategies with Lolium perenne on a heavy metal contaminated soil. Chemosphere 85, 1290–8. doi:10.1016/j.chemosphere.2011.07.034 Lehmann, J., 2007a. Bio-energy in the black. Frontiers in Ecology and the Environment 5, 381–387. Lehmann, J., 2007b. A handful of carbon. Nature 447, 10–11.

297

References

Lehmann, J., Abiven, S., Kleber, M., Pan, G., Singh, B.P., Sohi, S.P., Zimmerman, A.R., 2015. Persistence of biochar in soil, in: Lehmann, J., Joseph, S. (Eds.), Biochar For Environmental Management: Science, Technology and Implementation. Earthscan, Routledge, London, pp. 235–282. Lehmann, J., Czimczik, C., Laird, D., Sohi, S., 2009. Stability of Biochar in Soil, in: Lehmann, J., Joseph, S. (Eds.), Biochar for Environmental Management: Science and Technology. Earthscan, London, pp. 183– 205. Lehmann, J., Gaunt, J., Rondon, M., 2006. Bio-char Sequestration in Terrestrial Ecosystems – A Review. Mitigation and Adaptation Strategies for Global Change 11, 395–419. doi:10.1007/s11027-005- 9006-5 Lehmann, J., Joseph, S., 2015. Biochar for environmental management: an introduction, in: Lehmann, J., Joseph, S. (Eds.), Biochar For Environmental Management: Science, Technology and Implementation. Earthscan, Routledge, London, pp. 1–13. Lehmann, J., Joseph, S., 2009. Biochar for Environmental Management : An Introduction, in: Lehmann, J., Joseph, S. (Eds.), Biochar for Environmental Management: Science and Technology. Earthscan, London, pp. 1–12. Lehmann, J., Liang, B., Solomon, D., Lerotic, M., Luizão, F., Kinyangi, J., Schäfer, T., Wirick, S., Jacobsen, C., 2005. Near-edge X-ray absorption fine structure (NEXAFS) spectroscopy for mapping nano-scale distribution of organic carbon forms in soil: Application to black carbon particles. Global Biogeochemical Cycles 19, GB1013. doi:10.1029/2004GB002435 Lehmann, J., Pereira, J., Steiner, C., Nehls, T., Zech, W., Glaser, B., 2003. Nutrient availability and leaching in an archaeological Anthrosol and a Ferralsol of the Central Amazon basin : fertilizer, manure and charcoal amendments. Plant and Soil 249, 343–357. Lehmann, J., Rillig, M.C., Thies, J.E., Masiello, C. a., Hockaday, W.C., Crowley, D., 2011. Biochar effects on soil biota – A review. Soil Biology and Biochemistry 43, 1812–1836. doi:10.1016/j.soilbio.2011.04.022 Lehmann, J., Solomon, D., 2010. Organic Carbon Chemistry in Soils Observed by Synchrotron- Based Spectroscopy. Synchrotron-Based Techniques in Soils and SedimentsDevelopments in Soil Science 34, 289–312. doi:10.1016/S0166-2481(10)34010-4 Leifeld, J., 2007. Thermal stability of black carbon characterised by oxidative differential scanning calorimetry. Organic Geochemistry 38, 112–127. doi:10.1016/j.orggeochem.2006.08.004

298

References

Lepoivre, A., Septembre, G., 1941. Le charbon de bois : traité pratique de carbonisation. Chiron, Paris, France. Lepoivre, M., 1940. Le charbon de bois. La Nature 3064. Lettens, S., De Vos, B., Quataert, P., van Wesemael, B., Muys, B., van Orshoven, J., 2007. Variable carbon recovery of Walkley-Black analysis and implications for national soil organic carbon accounting. European Journal of Soil Science 58, 1244–1253. doi:10.1111/j.1365- 2389.2007.00916.x Lettens, S., Van Orshoven, J., Van Wesemael, B., Muys, B., Perrin, D., 2005. Soil organic carbon changes in landscape units of Belgium between 1960 and 2000 with reference to 1990. Global Change Biology 11, 2128– 2140. doi:10.1111/j.1365-2486.2005.001074.x Li, J., Li, B., Zhang, X.C., 2002. Comparative studies of thermal degradation between Larch lignin and Manchurian Ash lignin. Polymer Degradation and Stability 78, 279– 285. Liang, B., Lehmann, J., Sohi, S.P., Thies, J.E., O’Neill, B., Trujillo, L., Gaunt, J., Solomon, D., Grossman, J., Neves, E.G., Luizão, F.J., 2010. Black carbon affects the cycling of non-black carbon in soil. Organic Geochemistry 41, 206–213. doi:10.1016/j.orggeochem.2009.09.007 Liang, B., Lehmann, J., Solomon, D., Kinyangi, J., Grossman, J., O’Neill, B., Skjemstad, J.O., Thies, J., Luizão, F.J., Petersen, J., Neves, E.G., 2006. Black Carbon Increases Cation Exchange Capacity in Soils. Soil Science Society of America Journal 70, 1719–1730. doi:10.2136/sssaj2005.0383 Liang, B., Lehmann, J., Solomon, D., Sohi, S., Thies, J.E., Skjemstad, J.O., Luizão, F.J., Engelhard, M.H., Neves, E.G., Wirick, S., 2008. Stability of biomass-derived black carbon in soils. Geochimica et Cosmochimica Acta 72, 6069–6078. doi:10.1016/j.gca.2008.09.028 Liao, Y., Wang, S., Ma, X., 2004. Study of reaction mechanisms in cellulose pyrolysis. American Chemical Society, Division of Fuel Chemistry 49, 407–411. Lim, B., Cachier, H., 1996. Determination of black carbon by chemical oxidation and thermal treatment in recent marine and lake sediments and Cretaceous-Tertiary clays. Chemical Geology 131, 143–154. Liu, X., Zhang, A., Ji, C., Joseph, S., Bian, R., Li, L., Pan, G., Paz-Ferreiro, J., 2013. Biochar’s effect on crop productivity and the dependence on experimental conditions—a meta-analysis of literature data. Plant and Soil 373, 583–594. doi:10.1007/s11104-013-1806-x Ludemann, T., 2012. Airborne laser scanning of historical wood charcoal production sites - A new tool of kiln site anthracology at the landscape level. Saguntum: Papeles Del Laboratorio de Arqueología de Valencia

299

References

13, 247–252. Ludemann, T., 2011. Scanning the historical and scientific significance of charcoal production – local scale , high resolution kiln site anthracology at the landscape level. Sagvtvm Extra 11, 23–24. Ludemann, T., 2010. Past fuel wood exploitation and natural forest vegetation in the Black Forest, the Vosges and neighbouring regions in western Central Europe. Palaeogeography, Palaeoclimatology, Palaeoecology 291, 154–165. doi:10.1016/j.palaeo.2009.09.013 Maestrini, B., Nannipieri, P., Abiven, S., 2014. A meta-analysis on pyrogenic organic matter induced priming effect. GCB Bioenergy 7, 577–590. doi:10.1111/gcbb.12194 Major, J., Lehmann, J., Rondon, M., Goodale, C., 2010. Fate of soil-applied black carbon: Downward migration, leaching and soil respiration. Global Change Biology 16, 1366–1379. doi:10.1111/j.1365- 2486.2009.02044.x Manyà, J.J., 2012. Pyrolysis for biochar purposes: a review to establish current knowledge gaps and research needs. Environmental Science & Technology 46, 7939–54. doi:10.1021/es301029g Mao, J.-D., Johnson, R.L., Lehmann, J., Olk, D.C., Neves, E.G., Thompson, M.L., Schmidt-Rohr, K., 2012. Abundant and stable char residues in soils: implications for soil fertility and carbon sequestration. Environmental Science & Technology 46, 9571–6. doi:10.1021/es301107c Masiello, C.A., 2004. New directions in black carbon organic geochemistry. Marine Chemistry 92, 201–213. doi:10.1016/j.marchem.2004.06.043 Masiello, C.A., Druffel, E.R.M., Currie, L.A., 2002. Radiocarbon measurements of black carbon in aerosols and ocean sediments. Geochimica et Cosmochimica Acta 66, 1025–1036. Masiello, C.A., Gallagher, M.E., Randerson, J.T., Deco, R.M., Chadwick, O.A., 2008. Evaluating two experimental approaches for measuring ecosystem carbon oxidation state and oxidative ratio. Journal of Geophysical Research-Biogeosciences 113, G03010–G03010. doi:10.1029/2007JG000534 Masiello, C., Druffel, E., 1998. Black carbon in deep-Sea sediments. Science 280, 1911–1913. doi:10.1126/science.280.5371.1911 McBeath, A. V., Smernik, R.J., 2009. Variation in the degree of aromatic condensation of chars. Organic Geochemistry 40, 1161–1168. doi:10.1016/j.orggeochem.2009.09.006 McBeath, A. V., Smernik, R.J., Schneider, M.P.W., Schmidt, M.W.I., Plant, E.L., 2011. Determination of the aromaticity and the degree of aromatic

300

References

condensation of a thermosequence of wood charcoal using NMR. Organic Geochemistry 42, 1194–1202. doi:10.1016/j.orggeochem.2011.08.008 McBeath, A. V., Wurster, C.M., Bird, M.I., 2015. Influence of feedstock properties and pyrolysis conditions on biochar carbon stability as determined by hydrogen pyrolysis. Biomass and Bioenergy 73, 155– 173. doi:10.1016/j.biombioe.2014.12.022 McCarty, G.W., Reeves, J.B., Reeves, V.B., Follett, R.F., Kimble, J.M., 2002. Mid-Infrared and near-infrared diffuse reflectance spectroscopy for soil carbon measurement. Soil Sci. Soc. Am. J. 66, 640–646. doi:10.2136/sssaj2002.6400 Meersmans, J., Van Wesemael, B., De Ridder, F., Dotti, M.F., De Baets, S., Van Molle, M., 2009. Changes in organic carbon distribution with depth in agricultural soils in northern Belgium, 1960-2006. Global Change Biology 15, 2739–2750. doi:10.1111/j.1365-2486.2009.01855.x Metson, A.J., 1956. Methods of Chemical Analysis for Soil Survey Samples, New Zealand Soil Bureau, Bulletin 12. Mikan, C.J., Abrams, M.D., 1995. Altered forest composition and soil properties of historic charcoal hearths in southeastern Pennsylvania. Canadian Journal of Forest Research 25, 687–696. Minasny, B., Sulaeman, Y., Mcbratney, A.B., 2011. Is soil carbon disappearing? The dynamics of soil organic carbon in Java. Global Change Biology 17, 1917–1924. doi:10.1111/j.1365-2486.2010.02324.x Mulder, V.L., de Bruin, S., Schaepman, M.E., Mayr, T.R., 2011. The use of remote sensing in soil and terrain mapping — A review. Geoderma 162, 1–19. doi:10.1016/j.geoderma.2010.12.018 Naisse, C., Alexis, M., Plante, A., Wiedner, K., Glaser, B., Pozzi, A., Carcaillet, C., Criscuoli, I., Rumpel, C., 2013. Can biochar and hydrochar stability be assessed with chemical methods? Organic Geochemistry 60, 40–44. doi:10.1016/j.orggeochem.2013.04.011 Nguyen, B.T., Lehmann, J., 2009. Black carbon decomposition under varying water regimes. Organic Geochemistry 40, 846–853. doi:10.1016/j.orggeochem.2009.05.004 Nguyen, B.T., Lehmann, J., Hockaday, W.C., Joseph, S., Masiello, C.A., 2010. Temperature sensitivity of black carbon decomposition and oxidation. Environmental Science & Technology 44, 3324–3331. doi:10.1021/es903016y Nguyen, B.T., Lehmann, J., Kinyangi, J., Smernik, R., Riha, S.J., Engelhard, M.H., 2008. Long-term black carbon dynamics in cultivated soil. Biogeochemistry 89, 295–308. doi:10.1007/s10533-008-9220-9

301

References

Nguyen, T.H., Brown, R. a, Ball, W.P., 2004. An evaluation of thermal resistance as a measure of black carbon content in diesel soot, wood char, and sediment. Organic Geochemistry 35, 217–234. doi:10.1016/j.orggeochem.2003.09.005 Nigussie, A., Kissi, E., 2011. Effect of Charcoal Production on Soil Properties in Southwestern Ethiopia. Middle-East Journal of Scientific Research 9, 807–813. Nocita, M., Kooistra, L., Bachmann, M., Müller, A., Powell, M., Weel, S., 2011. Predictions of soil surface and topsoil organic carbon content through the use of laboratory and field spectroscopy in the Albany Thicket Biome of Eastern Cape Province of South Africa. Geoderma 167–168, 295–302. doi:10.1016/j.geoderma.2011.09.018 Noirot, 1843. De l’influence de l’entrée des fers étrangers sur les produits des forêts. Annales Forestières 626–634. Novak, J.M., Cantrell, K.B., Watts, D.W., Busscher, W.J., Johnson, M.G., 2014. Designing relevant biochars as soil amendments using lignocellulosic-based and manure-based feedstocks. Journal of Soils and Sediments 14, 330–343. doi:10.1007/s11368-013-0680-8 Nye, P.H., Greenland, D.J., 1964. Changes in the soil after clearing tropical forest. Plant and Soil 21, 101–112. Ogundele, A.T., Eludoyin, O.S., Oladapo, O.S., 2011. Assessment of impacts of charcoal production on soil properties in the derived savanna , Oyo state , Nigeria. Journal of Soil Science and Environmental Management 2, 142–146. Oguntunde, P.G., Fosu, M., Ajayi, A.E., van de Giesen, N., 2004. Effects of charcoal production on maize yield, chemical properties and texture of soil. Biology and Fertility of Soils 39, 295–299. doi:10.1007/s00374- 003-0707-1 Ohno, T., 1992. Neutralization of Soil Acidity and Release of Phosphorus and Potassium by Wood Ash. Journal of Environmental Quality 21, 433– 438. doi:10.2134/jeq1992.00472425002100030022x Ohno, T., Erich, S.M., 1990. Effect of wood ash application on soil pH and soil test nutrient levels. Agriculture, Ecosystems & Environment 32, 223–239. doi:10.1016/0167-8809(90)90162-7 Olarieta, J.R., Padrò, R., Masip, G., Rodríguez-Ochoa, R., Tello, E., 2011. “Formiguers”, a historical system of soil fertilization (and biochar production?). Agriculture, Ecosystems & Environment 140, 27–33. doi:10.1016/j.agee.2010.11.008 Panckoucke, C.-J., 1783. Encyclopédie méthodique. Planches. Publication : A Paris, chez Panckoucke, libraire, hôtel de Thou, rue des Poitevins ; A

302

References

Liège, chez Plomteux, imprimeur des États. Peel, M.C., Finlayson, B.L., McMahon, T. a., 2007. Updated world map of the K ̈oppen-Geiger climate classification. Hydrology and Earth System Sciences Discussions, European Geosciences Union 4, 439–473. doi:10.1127/0941-2948/2006/0130 Peltre, C., Fernández, J.M., Craine, J.M., Plante, A.F., 2013. Relationships between Biological and Thermal Indices of Soil Organic Matter Stability Differ with Soil Organic Carbon Level. Soil Science Society of America Journal 77, 2020. doi:10.2136/sssaj2013.02.0081 Pignatello, J.J., Kwon, S., Lu, Y., 2006. Effect of natural organic substances on the surface and adsorptive properties of environmental black carbon (char): attenuation of surface activity by humic and fulvic acids. Environmental Science & Technology 40, 7757–7763. Plante, A.F., Fernández, J.M., Haddix, M.L., Steinweg, J.M., Conant, R.T., 2011. Biological, chemical and thermal indices of soil organic matter stability in four grassland soils. Soil Biology and Biochemistry 43, 1051–1058. doi:10.1016/j.soilbio.2011.01.024 Plante, A.F., Fernández, J.M., Leifeld, J., 2009. Application of thermal analysis techniques in soil science. Geoderma 153, 1–10. doi:10.1016/j.geoderma.2009.08.016 Preston, C.M., Schmidt, M.W.I., 2006. Black (pyrogenic) carbon: a synthesis of current knowledge and uncertainties with special consideration of boreal regions. Biogeosciences 3, 397–420. doi:10.5194/bg-3-397-2006 Purmalis, O., Porsnovs, D., Klavins, M., 2011. Differential Thermal Analysis of Peat and Peat Humic Acids. Scientific Journal of Riga Technical University - Material Science and Applied Chemistry 24, 89–94. Qualls, R.G., Haines, B.L., Swank, W.T., Tyler, S.W., 2000. Soluble Organic and Inorganic Nutrient Fluxes in Clearcut and Mature Deciduous Forests. Soil Science Society of America Journal 64, 1068–1077. Raab, A., Takla, M., Raab, T., Nicolay, A., Schneider, A., Rösler, H., Heußner, K.-U., Bönisch, E., 2014. Pre-industrial charcoal production in Lower Lusatia (Brandenburg, Germany): Detection and evaluation of a large charcoal-burning field by combining archaeological studies, GIS- based analyses of shaded-relief maps and dendrochronological age determination. Quaternary International 367, 1–12. doi:10.1016/j.quaint.2014.09.041 R Core Team, 2012. R: A language and environment for statistical computing. Vienna, Austria. Rawls, W.J., 1983. Estimating soil bulk density from particle size analysis and organic matter content. Soil Science 135, 123–125.

303

References

Reeves, D.W., 1997. The role of soil organic matter in maintaining soil quality in continuous cropping systems. Soil and Tillage Research 43, 131–167. doi:10.1016/s0167-1987(97)00038-x Reeves, J.B., 2010. Near- versus mid-infrared diffuse reflectance spectroscopy for soil analysis emphasizing carbon and laboratory versus on-site analysis: Where are we and what needs to be done? Geoderma 158, 3–14. Regelink, I.C., Weng, L., Lair, G.J., Comans, R.N.J., 2015. Adsorption of phosphate and organic matter on metal (hydr)oxides in arable and forest soil: a mechanistic modelling study. European Journal of Soil Science 66, 867–875. doi:10.1111/ejss.12285 Reimer, P., 2013. IntCal13 and Marine13 Radiocarbon Age Calibration Curves 0–50,000 Years cal BP. Radiocarbon 55, 1869–1887. doi:10.2458/azu_js_rc.55.16947 Reisser, M., Purves, R.S., Schmidt, M.W.I., Abiven, S., 2016. Pyrogenic Carbon in Soils: A Literature-Based Inventory and a Global Estimation of Its Content in Soil Organic Carbon and Stocks. Frontiers in Earth Science 4, 1–14. doi:10.3389/feart.2016.00080 Risbøl, O., Bollandsås, O.M., Nesbakken, A., Ørka, H.O., Næsset, E., Gobakken, T., 2013. Interpreting cultural remains in airborne laser scanning generated digital terrain models: effects of size and shape on detection success rates. Journal of Archaeological. Archaeological Science 40, 4688–4700. Rodella, A.A., Saboya, L.V., 1999. Calibration for conductimetric determination of carbon dioxide. Soil Biology and Biochemistry 31, 2059–2060. doi:10.1016/S0038-0717(99)00046-2 Rösler, H., Bönisch, E., Schopper, F., Raab, T., Raab, A., 2012. Pre-industrial charcoal production in southern Brandenburg and its impact on the environment., in: Kluiving, S., Guttmann-Bond, E. (Eds.), Landschape Archeology between Art and Science. Amsterdam University Press, Amsterdam, pp. 167–178. Rousk, J., Baath, E., Brookes, P.C., Lauber, C.L., Lozupone, C., Caporaso, J.G., Knight, R., Fierer, N., 2010. Soil bacterial and fungal communities across a pH gradient in an arable soil. The ISME Journal 4, 1340–1351. doi:10.1038/ismej.2010.58 Rumpel, C., Alexis, M., Chabbi, A., Chaplot, V., Rasse, D.P., Valentin, C., Mariotti, A., 2006. Black carbon contribution to soil organic matter composition in tropical sloping land under slash and burn agriculture. Geoderma 130, 35–46. doi:10.1016/j.geoderma.2005.01.007 Sagrilo, E., Jeffery, S., Hoffland, E., Kuyper, T.W., 2014. Emission of CO2 from biochar-amended soils and implications for soil organic carbon. 304

References

GCB Bioenergy 7, 1294–1304. doi:10.1111/gcbb.12234 Sanyal, S.K., De Datta, S.K., 1991. Chemistry of phosphorus transformations in soil, in: Stewart, B.A. (Ed.), Advances in Soil Science, Volume 16. Springer-Verlag, New York, pp. 1–120. Satoh, T., 1984. Organo-mineral complex status in soils. I. thermal analytical characteristics of humus in the soils. Soil Science and Plant Nutrition 30, 1–12. doi:10.1080/00380768.1984.10434663 Schenkel, Y., 1999. Modelisation des Flux Massiques et Energetiques dans la Carbonisation du Bois en Four Cornue. PhD Dissertation. Universite des Sciences Agronomiques de Gembloux, Gembloux, Belgium. Schenkel, Y., Bertaux, P., Vanwijnsberghe, S., Carre, J., 1998. An evaluation of the mound kiln carbonization technique. Biomass and Bioenergy 14, 505–516. Schmidt, M.W.I., 2004. Biogeochemistry: carbon budget in the black. Nature 427, 305–307. doi:10.1038/427305a Schmidt, M.W.I., Noack, A.G., 2000a. Black carbon in soils and sediments: Analysis, distribution, implications, and current challenges. Global Biogeochemical Cycles 14, 777–793. doi:10.1029/1999GB001208 Schmidt, M.W.I., Skjemstad, J.O., Czimczik, C.I., Glaser, B., Prentice, K.M., Gelinas, Y., Kuhlbusch, T.A.J., 2001. Comparative analysis of black carbon in soils. Global Biogeochemical Cycles 15, 163–167. Schmidt, M.W.I., Torn, M.S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I.A., Kleber, M., Kögel-Knabner, I., Lehmann, J., Manning, D.A.C., Nannipieri, P., Rasse, D.P., Weiner, S., Trumbore, S.E., 2011. Persistence of soil organic matter as an ecosystem property. Nature 478, 49–56. doi:10.1038/nature10386 Schneider, A., Takla, M., Nicolay, A., Raab, A., Raab, T., 2015. A Template- matching Approach Combining Morphometric Variables for Automated Mapping of Charcoal Kiln Sites. Archeological Prospection 22, 45–62. doi:10.1002/arp Schnitzer, M., Turner, R.C., Hoffman, I., 1964. A thermo-gravimetric study of organic matter of representative Canadian podzol soils. Canadian Journal of Soil Science 44, 7–13. Schweingruber, F.H., 1990. Microscopic wood anatomy: Structural variability of stems and twigs in recent and subfossil woods from Central Europe. Eidgenössische Forschungsanstalt WSL, Birmensdorf, Switzerland. Scott, A.C., Damblon, F., 2010. Charcoal: Taphonomy and significance in geology, botany and archaeology. Palaeogeography, Palaeoclimatology, Palaeoecology 291, 1–10. doi:10.1016/j.palaeo.2010.03.044

305

References

Scott, A.C., Glasspool, I.J., 2006. The diversification of Paleozoic fire systems and fluctuations in atmospheric oxygen concentration. Proceedings of the National Academy of Sciences of the United States of America 103, 10861–10865. doi:10.1073/pnas.0604090103 Sergides, C.A., Jassim, J.A., Chughtai, A.R., Smith, D.M., 1987. The structure of hexane soot. Part III: Ozonation studies. Applied Spectroscopy 41, 482–492. Sherrod, L.A., Dunn, G., Peterson, G.A., Kolberg, R.L., 2002. Inorganic Carbon Analysis by Modified Pressure-Calcimeter Method. Soil Science Society of America Journal 66, 299–305. Shirley, D.A., 1972. High-Resolution X-Ray Photoemission Spectrum of the Valence Bands of Gold. Physical Review B 5, 4709–4714. doi:10.1103/PhysRevB.5.4709 Singh, N., Abiven, S., Torn, M.S., Schmidt, M.W.I., 2012. Fire-derived organic carbon in soil turns over on a centennial scale. Biogeosciences 9, 2847–2857. doi:10.5194/bg-9-2847-2012 Skjemstad, J.O., Spouncer, L.R., Cowie, B., Swift, R.S., 2004. Calibration of the Rothamsted organic carbon turnover model (RothC ver. 26.3), using measurable soil organic carbon pools. Australian Journal of Soil Research 42, 79–88. doi:10.1071/SR03013 Skjemstad, J.O., Taylor, J.A., 2008. Does the walkley-black method determine soil charcoal? Communications in Soil Science and Plant Analysis 30, 2299–2310. doi:10.1080/00103629909370373 Sleutel, S., Bouckaert, L., Buchan, D., Van Loo, D., Cornelis, W.M., Sanga, H.G., 2012. Manipulation of the soil pore and microbial community structure in soil mesocosm incubation studies. Soil Biology and Biochemistry 45, 40–48. doi:10.1016/j.soilbio.2011.09.016 Smernik, R.J., 2009. Biochar and Sorption of Organic Compounds, in: Lehmann, J., Joseph, S. (Eds.), Biochar for Environmental Management: Science and Technology. Earthscan, London, pp. 289–300. Smith, N.J.H., 1980. Anthrosols and human carrying capacity in Amazonia. Annals of the Association of American Geographers 70, 553–566. Solomon, D., Lehmann, J., Kinyangi, J., Amelung, W., Lobe, I., Pell, A., Riha, S., Ngoze, S., Verchot, L., Mbugua, D., Skjemstad, J., Schäfer, T., 2007a. Long-term impacts of anthropogenic perturbations on dynamics and speciation of organic carbon in tropical forest and subtropical grassland ecosystems. Global Change Biology 13, 511–530. doi:10.1111/j.1365-2486.2006.01304.x Solomon, D., Lehmann, J., Thies, J., Schäfer, T., Liang, B., Kinyangi, J., Neves, E., Petersen, J., Luizão, F., Skjemstad, J., 2007b. Molecular

306

References

signature and sources of biochemical recalcitrance of organic C in Amazonian Dark Earths. Geochimica et Cosmochimica Acta 71, 2285– 2298. doi:10.1016/j.gca.2007.02.014 Sombroek, S.W.G., Nachtergaele, F.O., Hebel, A., 1993. Amounts, dynamics and sequestering of carbon in tropical and subtropical soils. Ambio 22, 417–426. Song, J., Peng, P., Huang, W., 2002. Black Carbon and Kerogen in Soils and Sediments. 1. Quantification and Characterization. Environmental Science & Technology 36, 3960–3967. doi:10.1021/es025502m Steierer, B., 2011. Highlights on wood charcoal: 2004–2009. FAOSTAT- ForesSTAT. Steiner, C., Garcia, M., Zech, W., 2009. Effects of charcoal as slow release nutrient carrier on N-P-K dynamics and soil microbial population: Pot experiments with ferralsol substrate, in: Amazonian Dark Earths: Wim Sombroek’s Vision. pp. 325–338. doi:10.1007/978-1-4020-9031-8_17 Stevens, A., Udelhoven, T., Denis, A., Tychon, B., Lioy, R., Hoffmann, L., van Wesemael, B., 2010. Measuring soil organic carbon in croplands at regional scale using airborne imaging spectroscopy. Geoderma 158, 32– 45. doi:10.1016/j.geoderma.2009.11.032 Stevens, A., van Wesemael, B., Bartholomeus, H., Rosillon, D., Tychon, B., Ben-Dor, E., 2008. Laboratory, field and airborne spectroscopy for monitoring organic carbon content in agricultural soils. Geoderma 144, 395–404. doi:10.1016/j.geoderma.2007.12.009 Stevens, A., van Wesemael, B., Vandenschrick, G., Touré, S., Tychon, B., 2006. Detection of Carbon Stock Change in Agricultural Soils Using Spectroscopic Techniques. Soil Science Society of America Journal 70, 844. doi:10.2136/sssaj2005.0025 Stolz, C., Böhnke, S., Grunert, J., 2012. Reconstructing 2500 years of land use history on the Kemel Heath (Kemeler Heide), southern Rhenish Massif, Germany. E&G Quaternary Science Journal 61, 169–183. doi:10.3285/eg.61.2.05 Strickland, M.S., Rousk, J., 2010. Considering fungal: Bacterial dominance in soils - Methods, controls, and ecosystem implications. Soil Biology and Biochemistry 42, 1385–1395. doi:10.1016/j.soilbio.2010.05.007 Tallier, P.A., 2004. Entre déboisement, boisement et reboisement, deux siècles d’histoire des forêts belges (1750–1950). Forêt Wallonne 68, 8– 19. Tan, K.H., Hajek, B.F., Barshad, I., 1986. Thermal analysis techniques., in: Klute, A. (Ed.), Methods of Soil Analysis, Part I. Physical and Mineralogical Methods. Agronomy Monograph, Vol. 9. American

307

References

Society of Agronomy, Madison, WI Todor, pp. 151–183. Tarnocai, C., Canadell, J.G., Schuur, E. a G., Kuhry, P., Mazhitova, G., Zimov, S., 2009. Soil organic carbon pools in the northern circumpolar permafrost region. Global Biogeochemical Cycles 23, 1–11. doi:10.1029/2008GB003327 Tiessen, H., Moir, J.O., 1993. Total and organic carbon, in: Carter, E.D. (Ed.), Soil Sampling and Methods of Analysis. Lewis Publishers, Ann Arbor, MI, pp. 187–211. Todd-Brown, K.E.O., Randerson, J.T., Hopkins, F., Arora, V., Hajima, T., Jones, C., Shevliakova, E., Tjiputra, J., Volodin, E., Wu, T., Zhang, Q., Allison, S.D., 2014. Changes in soil organic carbon storage predicted by Earth system models during the 21st century. Biogeosciences 11, 2341– 2356. doi:10.5194/bg-11-2341-2014 Tryon, E.H., 1948. Effect of charcoal on certain physical, chemical, and biological properties of forest soils. Ecological Monographs 18, 81–115. doi:10.2307/1948629 Uehara, G., Gillman, P.G., 1981. The Mineralogy, Chemistry, and Physics of Tropical Soils with Variable Charge Clays. Westview Press, Ann Arbor, MI. Van Ranst, E., Verloo, M., Demeyer, A., Pauwels, J.M., 1999. Manual for the Soil Chemistry and Fertility Laboratory. NUGI 835, University of Ghent, Ghent. doi:D/1999/8536/1 van Wesemael, B., Paustian, K., Meersmans, J., Goidts, E., Barancikova, G., Easter, M., 2010. Agricultural management explains historic changes in regional soil carbon stocks. Proceedings of the National Academy of Sciences of the United States of America 107, 14926–30. doi:10.1073/pnas.1002592107 Vandenberghe, C., De Toffoli, M., Bachelart, F., Imbrecht, O., Lambert, R., Marcoen, J.M., 2013. Contrôle de l’azote potentiellement lessivable dans le sol en début de période de lixiviation. Établissement des valeurs de référence. Biotechnologie, Agronomie, Société et Environnement 17, 231–236. Vasilyeva, N. a., Abiven, S., Milanovskiy, E.Y., Hilf, M., Rizhkov, O. V., Schmidt, M.W.I., 2011. Pyrogenic carbon quantity and quality unchanged after 55 years of organic matter depletion in a Chernozem. Soil Biology and Biochemistry 43, 1985–1988. doi:10.1016/j.soilbio.2011.05.015 Vaudour, E., Gilliot, J.M., Bel, L., Lefevre, J., Chehdi, K., 2016. Regional prediction of soil organic carbon content over temperate croplands using visible near-infrared airborne hyperspectral imagery and synchronous field spectra. International Journal of Applied Earth Observation and 308

References

Geoinformation 49, 24–38. Vazquez-Marrufo, G., Serrato-Flores, R., Frias-Hernandez, J.T., Jimenez- Magdaleno, L.A., Olalde-Portugal, V., 2003. Microsite soil changes associated with traditional charcoal production in Quercus temperate forest in central Mexico. Phyton 72, 85–99. Viscarra Rossel, R. a., Walvoort, D.J.J., McBratney, a. B., Janik, L.J., Skjemstad, J.O., 2006. Visible, near infrared, mid infrared or combined diffuse reflectance spectroscopy for simultaneous assessment of various soil properties. Geoderma 131, 59–75. doi:10.1016/j.geoderma.2005.03.007 Viscarra Rossel, R. a., Webster, R., Bui, E.N., Baldock, J. a., 2014. Baseline map of organic carbon in Australian soil to support national carbon accounting and monitoring under climate change. Global Change Biology 20, 2953–2970. doi:10.1111/gcb.12569 Wagner, J., 1921. La sidérurgie luxembourgeoise avant la découverte du gisement des minettes. Histoire technique du bon vieux temps. Imprimerie Paul Schroell, Diekirch, Grand-Duché du Luxembourg. Walkley, A., 1947. A critical examination of a rapid method for determining organic carbon in soils – effect of variations in digestion conditions and of inorganic soil constituents. Soil Science 63, 251–264. Walkley, A., Black, I.A., 1934. An Examination of the Degtjareff method for determining soil organic matter, and a proposed modification of the chromic acid titration method. Soil Science 37, 29–38. Wardle, D. a, Nilsson, M.-C., Zackrisson, O., 2008. Fire-Derived Charcoal Causes Loss of Forest Humus. Science 320, 629-. doi:10.1126/science.1154960 Waschkies, C., Hüttl, R.F.J., 1999. Microbial degradation of geogenic organic C and N in mine spoils. Plant and Soil 213, 221—230. doi:10.1023/A:1004539502221 Wehr, A., Lohr, U., 1999. Airborne laser scanning—an introduction and overview. ISPRS Journal of Photogrammetry and Remote Sensing 54, 68–82. doi:10.1016/S0924-2716(99)00011-8 Wengel, M., Kothe, E., Schmidt, C.M., Heide, K., Gleixner, G., 2006. Degradation of organic matter from black shales and charcoal by the wood-rotting fungus Schizophyllum commune and release of DOC and heavy metals in the aqueous phase. Science of the Total Environment 367, 383–393. Wiedemeier, D.B., Abiven, S., Hockaday, W.C., Keiluweit, M., Kleber, M., Masiello, C.A., McBeath, A. V., Nico, P.S., Pyle, L.A., Schneider, M.P.W., Smernik, R.J., Wiesenberg, G.L.B., Schmidt, M.W.I., 2015.

309

References

Aromaticity and degree of aromatic condensation of char. Organic Geochemistry 78, 135–143. doi:10.1016/j.orggeochem.2014.10.002 Wiedner, K., Glaser, B., 2015. Traditional use of biochar, in: Lehmann, J., Joseph, S. (Eds.), Biochar For Environmental Management: Science, Technology and Implementation. Earthscan, Routledge, London, pp. 15–38. Wiedner, K., Schneeweiß, J., Dippold, M.A., Glaser, B., 2015. Anthropogenic Dark Earth in Northern Germany — The Nordic Analogue to terra preta de Índio in Amazonia. CATENA 132, 114–125. doi:10.1016/j.catena.2014.10.024 Wolbach, W.S., Anders, E., 1989. Elemental carbon in sediments: Determination and spectroscopic analysis in the presence of kerogen. Geochimica et Cosmochimica Acta 1637–1647. Woolf, D., Amonette, J.E., Street-Perrott, F.A., Lehmann, J., Joseph, S., 2010. Sustainable biochar to mitigate global climate change. Nature Communications 1, 1–9. doi:10.1038/ncomms1053 Woronoff, D., 1994. Histoire de l’industrie en France: du XVIe siècle à nos jours. Éditions du Seuil, Paris, France. Woronoff, D., 1990. Forges et forêts : recherches sur la consommation proto- industrielle de bois. Editions de l’Ecole des Hautes Etudes en Sciences Sociales, Paris. Wurster, C.M., Saiz, G., Schneider, M.P.W., Schmidt, M.W.I., Bird, M.I., 2013. Quantifying pyrogenic carbon from thermosequences of wood and grass using hydrogen pyrolysis. Organic Geochemistry 62, 28–32. doi:10.1016/j.orggeochem.2013.06.009 Yang, H., Yan, R., Chen, H., Lee, D.H., Zheng, C., 2007. Characteristics of hemicellulose, cellulose and lignin pyrolysis. Fuel 86, 1781–1788. doi:10.1016/j.fuel.2006.12.013 Yao, Y., Gao, B., Zhang, M., Inyang, M., Zimmerman, A.R., 2012. Effect of biochar amendment on sorption and leaching of nitrate, ammonium, and phosphate in a sandy soil. Chemosphere 89, 1467–71. doi:10.1016/j.chemosphere.2012.06.002 Young, M.J., Johnson, J.E., Abrams, M.D., 1996. Vegetative and edaphic characteristics on relic charcoal hearths in the Appalachian mountains. Vegetatio 125, 43–50. Zimmerman, A.R., Gao, B., Ahn, M.-Y., 2011. Positive and negative carbon mineralization priming effects among a variety of biochar-amended soils. Soil Biology and Biochemistry 43, 1169–1179. doi:10.1016/j.soilbio.2011.02.005 Zimov, S. a., Davydov, S.P., Zimova, G.M., Davydova, a. I., Schuur, E. a G.,

310

References

Dutta, K., Chapin, I.S., 2006. Permafrost carbon: Stock and decomposability of a globally significant carbon pool. Geophysical Research Letters 33, 1–5. doi:10.1029/2006GL027484

311

Appendices

Appendix 1. Soil properties of the 19 charcoal kiln sites of Wallonia (K) and adjacent reference soils (R). The names of soil profiles are based on their soil unit according to the WRB 2014 classification (Arenosols (AR), Cambisols (CM), Luvisols (LV), and Podzols (PZ)). We differentiated between acidic Cambisols (CMA) and calcareous Cambisols (CMC). The charcoal kiln CMA6 was sampled twice because it had a particular shape and most of the charcoal-enriched soil had accumulated at the edge of the platform.

CEC Ca/ Mg/ Na/ P /mg/k K/ cmol pH7 Depth/ av cmol cmol c cmol Site K/R Horizon pHH2O pHKCl OC/% N/% C:N -1 c c −1 c BS/% cm g kg cmolc kg−1 kg−1 kg−1 kg−1/ AR1 K topsoil 0−40 5.0 4.7 5.29 0.21 25.4 6.3 13.99 0.58 0.03 0.03 23.3 62.9 AR1 K topsoil 40−50 5.3 5.0 3.90 0.11 35.5 5.8 14.56 0.62 0.01 0.02 19.3 79.0 AR1 K subsoil 50−60 5.5 5.1 0.85 0.03 24.5 3.7 4.97 0.25 0.01 0.03 5.1 100 AR1 K subsoil 60−80 5.5 5.1 0.40 0.03 12.3 1.8 2.71 0.22 0.01 0.03 3.2 94.1 AR1 K subsoil 80−120 5.6 5.0 0.25 0.03 8.3 1.5 1.81 0.19 0.02 0.01 2.8 72.5 AR1 R topsoil 0−10 4.0 3.6 2.22 0.17 12.9 20.9 0.78 0.18 0.09 0.03 4.8 22.5 AR1 R subsoil 10−20 4.2 3.7 1.18 0.10 12.2 10.7 0.25 0.07 0.03 0.02 4.2 9.0 AR1 R subsoil 20−40 4.4 4.1 0.58 0.05 11.7 <1.0 0.16 0.04 0.02 0.01 3.3 6.9 AR1 R subsoil 40−80 4.2 3.9 0.28 0.03 8.5 3.3 0.21 0.05 0.03 0.02 3.3 9.2 AR1 R subsoil 80−120 4.5 4.1 0.16 0.02 6.9 3.2 1.45 0.22 0.05 0.02 3.4 50.3 AR2 K topsoil 0−20 4.5 3.4 3.20 0.16 20.4 26.5 0.72 0.10 0.08 0.03 14.9 6.2 AR2 K topsoil 20−54 5.4 4.2 1.48 0.06 26.2 9.6 3.06 0.12 0.03 0.04 7.1 45.4 AR2 K subsoil 54−75 5.6 4.4 0.27 0.02 14.9 6.7 0.60 0.06 0.01 0.03 1.2 58.0 AR2 K subsoil 75−95 5.6 4.6 0.17 0.02 10.5 3.4 0.65 0.07 0.02 0.03 0.8 92.3 AR2 R topsoil 0−8 4.1 3.2 3.07 0.19 16.5 43.5 0.79 0.19 0.14 0.04 6.1 18.9 AR2 R topsoil 8−18 4.5 3.9 1.58 0.09 17.5 17.4 0.20 0.06 0.04 0.02 4.5 7.2 AR2 R subsoil 18−35 4.8 4.4 0.60 0.05 12.3 6.9 0.11 0.04 0.02 0.02 1.4 12.8 AR2 R subsoil 35−60 4.8 4.4 0.29 0.03 10.2 3.1 0.12 0.03 0.02 0.02 0.8 22.7 CMA1 K topsoil 0−30 4.2 3.3 4.57 0.19 23.5 18.3 0.47 0.09 0.08 0.09 17.3 4.2 CMA1 K topsoil 30−35 4.3 3.4 2.97 0.11 26.6 9.3 0.29 0.06 0.05 0.06 12.1 3.8 CMA1 K topsoil 35−45 4.1 3.4 4.30 0.14 30.8 5.7 0.67 0.09 0.05 0.09 18.5 4.8 CMA1 K subsoil 45−60 4.6 3.8 0.21 0.01 32.2 4.6 0.31 0.07 0.03 0.03 1.9 23.6 CMA1 K subsoil 60−77 4.3 3.8 0.34 0.04 8.4 4.6 0.90 0.20 0.05 0.16 4.2 31.3 CMA1 K subsoil 77−100 5.2 4.4 0.23 0.02 10.8 3.4 0.19 0.05 0.03 0.05 1.7 18.7 CMA1 R topsoil 0−30 3.6 3.2 2.93 0.18 16.5 50.9 0.26 0.06 0.07 0.00 8.2 4.8 CMA1 R subsoil 30−60 4.0 3.8 0.73 0.05 15.0 11.3 0.12 0.03 0.03 0.02 3.0 6.7 CMA1 R subsoil 60−80 3.9 3.3 0.17 0.02 9.4 7.4 0.12 0.03 0.03 0.02 1.7 11.7 CMA2 K topsoil 0−31 3.8 3.0 4.48 0.17 26.9 15.4 0.73 0.07 0.07 0.01 17.8 4.9 CMA2 K topsoil 31−40 4.0 3.2 1.98 0.05 39.0 6.3 0.74 0.05 0.01 0.00 10.2 7.9 CMA2 K topsoil 40−50 3.9 3.2 2.40 0.07 35.6 6.0 1.39 0.07 0.02 0.01 14.0 10.6 CMA2 K subsoil 50−85 4.4 4.0 0.20 0.01 22.1 4.8 0.21 0.03 0.01 0.02 1.5 18.8 CMA2 K subsoil 85−96 4.2 3.7 0.32 0.03 12.9 7.8 0.44 0.05 0.02 0.02 2.7 19.8 CMA2 K subsoil 96−107 4.3 3.8 0.41 0.03 12.0 10.1 0.69 0.06 0.03 0.02 3.7 21.5 CMA2 K subsoil 107−120 4.7 4.2 0.41 0.04 10.9 6.6 1.24 0.12 0.03 0.02 3.4 41.1 CMA2 R topsoil 0−10 4.0 3.2 2.70 0.19 14.0 25.1 0.46 0.16 0.09 0.02 4.4 16.9 CMA2 R subsoil 10−14 3.8 3.3 1.05 0.07 14.6 20.0 0.24 0.07 0.03 0.02 2.1 17.9 CMA2 R subsoil 14−25 3.8 3.3 1.15 0.08 14.7 25.2 0.25 0.06 0.03 0.02 4.9 7.5 CMA2 R subsoil 25−44 4.3 4.0 0.66 0.05 13.5 10.7 0.15 0.04 0.04 0.02 4.5 5.5 CMA2 R subsoil 44−75 4.4 4.1 0.34 0.03 11.8 3.8 0.18 0.05 0.07 0.02 4.9 6.7

313

Appendices

Depth/ OC/ N/ Pav/ Ca/ Mg/ K/ Na/ CEC pH7/ BS/ Site K/R Horizon pHH2O pHKCl C:N −1 −1 −1 −1 −1 −1 cm % % mg kg cmolc kg cmolc kg cmolc kg cmolc kg cmolc kg % K topsoil 0-38 4.3 3.9 7.65 0.33 23.2 22.3 8.19 0.34 0.11 0.06 30.5 28.5 K topsoil 38−60 4.9 4.5 6.89 0.21 33.2 23.2 18.22 0.45 0.03 0.07 34.9 53.8 K topsoil 60−75 5.1 4.8 2.09 0.08 26.2 19.9 8.55 0.32 0.04 0.04 10.8 83.1

CMA3 K subsoil 75−108 5.2 4.8 1.04 0.08 13.7 11.8 6.46 0.39 0.05 0.04 8.5 81.4 K subsoil 108−130 5.2 4.7 0.52 0.05 9.9 4.9 4.84 0.32 0.05 0.03 6.0 87.8 K subsoil 130−160 5.3 4.7 0.27 0.04 7.7 4.1 3.48 0.27 0.05 0.05 4.1 94.1 R topsoil 0−10 4.4 4.1 3.10 0.24 12.7 27.3 4.83 0.44 0.16 0.05 9.1 60.0 R subsoil 10−25 4.5 3.8 1.31 0.10 12.9 10.0 1.87 0.13 0.05 0.03 6.0 34.4 R subsoil 25−50 4.6 4.0 0.53 0.06 9.5 5.6 1.63 0.12 0.03 0.03 4.5 40.0

CMA3 R subsoil 50−80 4.7 4.2 0.35 0.04 9.0 4.4 1.48 0.11 0.03 0.03 3.5 47.0 R subsoil 80−110 4.8 4.3 0.25 0.03 8.4 4.4 1.18 0.06 0.02 0.02 2.6 49.0 K topsoil 0−17 3.7 3.3 11.96 0.39 31.0 12.5 0.37 0.12 0.14 0.05 44.2 1.5 K topsoil 17−30 3.9 3.5 5.34 0.12 44.2 2.9 0.26 0.03 0.05 0.02 24.1 1.5

CMA4 K subsoil 30−40 3.9 3.6 1.74 0.06 30.7 4.0 0.20 0.02 0.03 0.05 9.8 3.0 K subsoil 40−68 3.9 3.6 0.38 0.04 9.7 2.3 0.19 0.03 0.05 0.05 6.7 4.8 R topsoil 0−12 3.7 3.3 5.48 0.32 17.0 26.8 0.57 0.18 0.16 0.05 15.2 6.4 R subsoil 12−42 3.9 3.5 1.34 0.07 18.1 6.8 0.17 0.07 0.08 0.04 8.8 4.1 CMA4 R subsoil 42−86 3.9 3.6 0.34 0.06 6.0 3.1 0.16 0.12 0.11 0.04 9.9 4.4 K topsoil 0−40 4.6 4.1 10.09 0.50 20.3 6.5 6.03 1.17 0.26 0.10 43.3 17.5 K topsoil 40−51 5.0 4.6 2.74 0.16 16.7 5.1 7.92 1.82 0.15 0.11 18.6 53.8

CMA5 K subsoil 51−72 5.1 4.6 1.02 0.12 8.5 5.2 5.41 1.89 0.16 0.09 13.8 55.0 K subsoil 72−94 5.1 4.3 0.59 0.09 6.3 3.8 5.02 2.44 0.30 0.07 12.8 61.4 R topsoil 0−9 4.0 3.5 6.31 0.48 13.2 19.4 0.68 0.29 0.30 0.07 21.8 6.2 R subsoil 9−20 4.1 3.7 3.14 0.24 13.0 5.8 0.36 0.14 0.17 0.05 15.0 4.8

CMA5 R subsoil 20−58 4.1 3.7 0.89 0.12 7.2 4.4 0.60 0.26 0.11 0.04 13.0 7.7 R subsoil 58−92 4.4 3.6 0.26 0.09 3.0 4.2 1.50 1.45 0.14 0.06 15.9 19.9 K topsoil 0−10 3.9 3.3 18.22 0.84 21.7 30.6 0.49 0.23 0.17 0.10 60.6 1.6 K subsoil 10−30 4.4 3.9 3.22 0.19 17.2 5.2 1.10 0.30 0.10 0.06 16.6 9.4

CMA6 CMA6 Center K subsoil 30−55 5.5 4.1 1.67 0.13 12.8 4.7 1.34 0.51 0.12 0.05 11.0 18.5 R topsoil 0−15 3.8 3.3 9.29 0.51 18.3 65.1 0.82 0.29 0.18 0.07 37.6 3.6 R subsoil 15−30 4.0 3.7 4.30 0.26 16.8 15.1 0.24 0.11 0.11 0.05 22.2 2.3

CMA6 CMA6 Center R subsoil 30−60 4.1 3.8 2.09 0.15 14.2 7.1 0.23 0.09 0.11 0.04 13.0 3.5 K topsoil 0−40 4.2 3.6 20.35 0.89 22.8 6.5 2.28 1.02 0.12 0.27 79.1 4.7 K topsoil 40−70 4.4 3.7 14.48 0.51 28.2 4.4 2.92 1.87 0.12 0.22 63.6 8.1 K topsoil 70−85 4.1 3.6 7.31 0.27 27.3 5.0 0.90 0.74 0.09 0.13 15.0 12.5 K subsoil 85−105 4.1 3.9 2.14 0.13 17.0 4.9 0.30 0.58 0.08 0.11 15.2 7.1

CMA6 BorderCMA6 K subsoil 105−125 4.2 3.9 1.02 0.09 11.4 4.7 0.22 0.49 0.10 0.10 10.5 8.7 K subsoil 125−150 4.6 4.2 0.52 0.06 8.0 3.4 0.47 0.77 0.11 0.10 8.7 16.7 R topsoil 0−25 4.0 3.7 4.44 0.28 15.6 8.4 0.39 0.12 0.14 0.05 20.8 3.4 R subsoil 25−45 4.2 3.9 1.99 0.15 13.1 5.7 0.58 0.21 0.09 0.05 11.2 8.4 R subsoil 45−65 4.7 4.4 0.93 0.10 9.3 4.8 1.35 0.51 0.09 0.10 7.4 27.8

CMA6 BorderCMA6 R subsoil 65−100 4.9 4.4 0.59 0.08 7.2 3.8 1.77 0.86 0.12 0.06 16.2 17.4

314

Appendices

Depth/ OC/ N/ Pav/ Ca/ Mg/ K/ Na/ CEC pH7/ BS/ Site K/R Horizon pHH2O pHKCl C:N −1 −1 −1 −1 −1 −1 cm % % mg kg cmolc kg cmolc kg cmolc kg cmolc kg cmolc kg % CMA7 K topsoil 0−21 4.0 3.4 18.76 0.69 27.1 12.2 0.44 0.13 0.19 0.12 49.9 1.7 CMA7 K topsoil 21−28 4.0 3.3 9.37 0.30 30.9 6.8 0.50 0.08 0.08 0.08 32.2 2.3 CMA7 K subsoil 28−45 4.2 3.7 1.81 0.11 16.2 5.0 0.32 0.09 0.07 0.05 8.9 6.0 CMA7 K subsoil 45−66 4.2 3.9 1.16 0.12 9.3 5.0 0.36 0.10 0.08 0.05 7.5 7.9 CMA7 R topsoil 0−10 3.6 3.2 7.50 0.94 8.0 48.8 0.38 0.18 0.26 0.06 21.8 4.1 CMA7 R subsoil 10−22 3.8 3.5 3.49 0.24 14.8 10.3 0.22 0.09 0.13 0.05 13.8 3.5 CMA7 R subsoil 22−37 4.4 4.0 1.78 0.16 11.1 6.8 0.15 0.05 0.08 0.04 8.8 3.6 CMA7 R subsoil 37−57 4.4 4.0 0.99 0.12 8.5 5.2 0.14 0.04 0.08 0.04 7.3 4.0 CMA8 K topsoil 0−35 4.9 3.9 10.65 0.46 23.0 9.6 4.10 1.15 0.21 0.06 35.6 15.5 CMA8 K topsoil 35−47 5.1 4.1 3.60 0.15 24.5 4.8 4.47 1.62 0.13 0.09 15.4 41.0 CMA8 K subsoil 47−67 5.0 3.9 1.09 0.07 15.0 4.3 1.86 1.48 0.14 0.07 10.4 34.5 CMA8 K subsoil 67−86 4.8 3.8 0.40 0.06 7.0 2.6 1.36 1.61 0.18 0.05 8.7 37.0 CMA8 R topsoil 0−11 4.4 3.5 7.66 0.51 15.2 47.5 1.04 0.41 0.31 0.05 23.8 7.6 CMA8 R topsoil 11−24 4.4 3.7 3.04 0.18 16.7 11.4 0.27 0.15 0.15 0.04 12.4 4.9 CMA8 R subsoil 24−45 4.6 3.8 0.78 0.08 10.1 6.6 0.20 0.13 0.12 0.03 8.5 5.6 CMA8 R subsoil 45−80 4.5 3.5 0.32 0.06 5.4 3.9 0.24 0.24 0.15 0.04 9.6 6.9 CMC1 K topsoil 0−20 4.6 4.2 9.01 0.50 18.2 8.6 7.68 1.76 0.14 0.11 28.0 34.7 CMC1 K subsoil 20−34 5.0 4.4 2.00 0.11 18.3 6.9 6.39 1.68 0.09 0.09 13.8 59.9 CMC1 K subsoil 34−64 5.3 4.8 0.68 0.06 10.9 4.1 6.23 1.71 0.08 0.06 11.7 69.3 CMC1 K subsoil 64−90 5.4 5.0 0.18 0.02 7.7 1.7 4.13 1.11 0.07 0.04 6.4 84.2 CMC1 K subsoil 90−110 5.6 5.3 0.17 0.02 6.8 3.2 5.91 1.65 0.07 0.05 9.3 82.7 CMC1 R topsoil 0−15 5.2 4.9 2.77 0.26 10.7 7.2 8.77 3.42 0.25 0.06 14.7 85.5 CMC1 R subsoil 15−38 5.1 4.7 1.21 0.12 10.5 2.2 6.69 3.29 0.13 0.06 14.5 70.2 CMC1 R subsoil 38−60 5.3 4.9 0.51 0.06 8.3 4.4 7.92 3.92 0.13 0.06 12.6 96.1 CMC1 R subsoil 60−80 5.5 5.1 0.34 0.05 6.7 3.9 9.15 5.16 0.18 0.07 15.9 92.0 CMC1 R subsoil 80−110 5.6 5.2 0.29 0.06 5.3 4.0 10.45 6.76 0.19 0.08 19.3 90.9 CMC2 K topsoil 0−21 5.3 5.0 8.89 0.52 17.1 10.1 32.9 1.13 0.24 0.16 35.5 90.7 CMC2 K subsoil 21−42 6.0 5.6 1.68 0.12 14.0 6.6 23.5 0.43 0.22 0.11 17.4 100.0 CMC2 K subsoil 42−90 6.4 6.1 0.51 0.07 7.7 5.3 28.9 0.41 0.25 0.10 13.1 100.0 CMC2 R topsoil 0−15 6.4 5.8 3.92 0.37 10.6 11.4 24.8 1.35 0.41 0.10 22.2 100.0 CMC2 R subsoil 15−30 6.7 6.0 1.81 0.17 10.7 10.4 25.5 0.85 0.40 0.07 17.9 100.0 CMC2 R subsoil 30−56 7.0 6.4 1.00 0.11 9.0 10.9 35.5 0.74 0.53 0.06 14.5 100.0 CMC3 K topsoil 0−19 5.5 4.5 8.00 0.37 21.4 11.4 9.07 3.07 0.35 0.05 27.9 45.1 CMC3 K subsoil 19−30 5.6 4.8 2.38 0.10 24.1 4.7 6.94 3.73 0.28 0.01 13.3 83.1 CMC3 K subsoil 30−49 6.0 5.4 0.48 0.06 8.6 4.2 9.18 7.31 0.37 0.02 14.4 100.0 CMC3 K subsoil 49−68 6.6 5.9 0.39 0.04 8.9 15.6 9.83 7.67 0.32 0.05 12.8 100.0 CMC3 K subsoil 68−76 7.0 6.3 0.13 0.04 3.0 15.2 11.37 7.37 0.29 0.06 11.5 100.0 CMC3 R topsoil 0−17 5.2 4.1 2.11 0.18 11.6 9.5 2.60 2.13 0.19 0.04 11.0 45.3 CMC3 R subsoil 17−28 5.8 4.5 0.94 0.10 9.6 5.5 5.71 4.27 0.22 0.04 10.8 94.9 CMC3 R subsoil 28−47 6.3 5.5 0.43 0.07 6.1 6.2 9.16 7.08 0.29 0.06 10.6 100.0 CMC3 R subsoil 47−62 7.0 6.3 0.10 0.05 2.1 28.9 10.4 6.68 0.25 0.10 13.5 100.0 CMC3 R subsoil 62−73 7.6 6.8 0.19 0.04 5.0 13.0 12.5 7.00 0.25 0.11 9.8 100.0

315

Appendices

Depth/ OC/ N/ Pav/ Ca/ Mg/ K/ Na/ CEC pH7/ BS/ Site K/R Horizon pHH2O pHKCl C:N −1 −1 −1 −1 −1 −1 cm % % mg kg cmolc kg cmolc kg cmolc kg cmolc kg cmolc kg % LV1 K topsoil 0−38 3.8 3.4 9.96 0.41 24.4 10.3 1.13 0.17 0.19 0.08 32.0 4.9 LV1 K topsoil 38−46 4.2 3.7 3.72 0.11 32.8 4.4 2.21 0.20 0.06 0.08 15.6 16.4 LV1 K subsoil 46−70 4.2 3.8 0.37 0.03 13.9 2.5 0.84 0.14 0.07 0.05 4.3 25.9 LV1 K subsoil 70−100 4.0 3.6 0.31 0.04 7.7 3.4 0.99 0.61 0.25 0.07 14.6 13.2 LV1 R topsoil 0−6 3.4 3.0 9.78 0.63 15.5 161.8 2.00 0.50 0.27 0.07 24.5 11.6 LV1 R topsoil 6−14 3.6 3.2 3.83 0.22 17.2 60.5 0.42 0.13 0.11 0.06 12.3 5.8 LV1 R subsoil 14−30 3.9 3.5 0.73 0.04 18.7 8.3 0.32 0.06 0.07 0.04 5.8 8.5 LV1 R subsoil 30−72 3.9 3.6 0.31 0.04 8.5 5.5 0.56 0.34 0.23 0.05 11.5 10.3 LV2 K topsoil 0−46 4.4 4.0 9.59 0.36 26.6 12.9 9.04 0.78 0.14 0.11 36.0 28.1 LV2 K topsoil 46−59 4.7 4.3 6.39 0.18 35.7 15.0 9.80 0.99 0.09 0.12 20.4 53.9 LV2 K subsoil 59−73 4.8 4.3 1.58 0.05 29.8 6.1 4.75 0.98 0.10 0.09 10.3 57.4 LV2 K subsoil 73−90 4.7 4.0 0.23 0.03 9.1 2.7 3.73 1.20 0.16 0.08 6.9 75.1 LV2 R topsoil 0−10 3.7 3.3 6.01 0.36 16.7 61.4 1.32 0.36 0.27 0.05 17.5 11.5 LV2 R topsoil 10−20 3.8 3.5 2.99 0.17 18.1 17.6 0.40 0.13 0.15 0.03 10.8 6.7 LV2 R subsoil 20−40 4.0 3.6 0.65 0.04 14.7 6.6 0.31 0.11 0.14 0.03 7.1 8.4 LV2 R subsoil 40−60 4.1 3.7 0.28 0.03 8.9 5.5 1.11 0.65 0.19 0.04 9.5 21.1 LV2 R subsoil 65−90 4.2 3.6 0.20 0.03 6.0 5.0 2.26 1.86 0.23 0.07 11.9 37.2 LV3 K topsoil 0−35 3.9 3.5 6.23 0.30 20.5 6.7 0.42 0.14 0.21 0.04 20.5 4.0 LV3 K topsoil 35−45 4.1 3.7 3.35 0.14 24.3 4.8 0.38 0.12 0.15 0.06 15.1 4.7 LV3 K subsoil 45−55 4.4 3.8 1.47 0.07 21.8 3.9 1.54 0.33 0.07 0.06 9.5 20.9 LV3 K subsoil 55−68 4.5 3.8 0.49 0.04 12.3 3.7 2.01 0.57 0.07 0.06 6.8 40.3 LV3 K subsoil 68−90 4.4 3.8 0.24 0.04 6.9 2.6 1.87 0.88 0.12 0.07 7.8 37.5 LV3 K subsoil 90−110 4.2 3.7 0.14 0.03 5.3 2.0 1.36 1.21 0.22 0.06 10.5 27.2 LV3 R topsoil 0−7 3.8 3.3 5.74 0.39 14.7 54.4 2.16 0.60 0.34 0.05 15.5 20.4 LV3 R subsoil 7−42 3.9 3.6 0.63 0.06 11.2 7.1 0.20 0.12 0.13 0.04 7.8 6.3 LV3 R subsoil 42−90 3.9 3.6 0.18 0.03 6.2 3.6 0.23 0.16 0.23 0.04 11.0 6.0 LV4 K topsoil 0−25 3.7 3.1 8.49 0.32 26.9 53.1 1.15 0.22 0.36 0.03 31.6 5.6 LV4 K topsoil 25−38 3.9 3.4 7.96 0.19 42.3 8.0 0.67 0.10 0.21 0.01 22.9 4.3 LV4 K subsoil 38−46 4.2 3.6 1.02 0.07 15.2 7.0 0.70 0.19 0.16 0.00 6.2 17.1 LV4 K subsoil 46−63 4.4 3.8 0.44 0.05 8.5 8.5 2.03 0.65 0.16 0.01 7.3 39.4 LV4 K subsoil 63−83 4.5 3.8 0.29 0.05 6.0 10.3 3.35 0.94 0.20 0.02 9.7 46.7 LV4 K subsoil 83−103 4.8 3.9 0.28 0.05 5.4 9.8 5.62 1.75 0.27 0.03 11.9 64.6 LV4 R topsoil 0−7 3.8 3.1 8.92 0.54 16.6 193.8 1.06 0.34 0.45 0.05 23.5 8.1 LV4 R subsoil 7−12 4.1 3.5 1.84 0.12 15.5 39.8 0.36 0.15 0.24 0.03 8.8 8.9 LV4 R subsoil 12−31 4.4 3.8 0.84 0.08 9.9 10.1 0.21 0.10 0.25 0.03 6.7 8.8 LV4 R subsoil 31−51 4.2 3.7 0.43 0.06 7.2 9.9 0.25 0.13 0.30 0.04 8.1 8.8 LV4 R subsoil 51−71 4.2 3.7 0.24 0.05 4.9 9.8 0.27 0.17 0.31 0.04 10.8 7.3 LV4 R subsoil 71−91 4.4 3.7 0.24 0.07 3.2 12.1 0.99 0.55 0.34 0.06 12.0 16.2

316

Appendices

Depth/ OC/ N/ Pav/ Ca/ Mg/ K/ Na/ CEC pH7/ BS/ Site K/R Horizon pHH2O pHKCl C:N −1 −1 −1 −1 −1 −1 cm % % mg kg cmolc kg cmolc kg cmolc kg cmolc kg cmolc kg % PZ1 K topsoil 0−19 3.9 3.0 3.98 0.11 37.4 4.8 0.75 0.09 0.06 0.03 14.3 6.5 PZ1 K topsoil 19−29 4.0 3.3 2.33 0.06 38.0 11.1 0.50 0.05 0.03 0.04 7.4 8.5 PZ1 K topsoil 29−40 4.2 3.6 3.92 0.07 55.6 4.1 6.05 0.13 0.03 0.02 17.4 35.7 PZ1 K subsoil 40−65 4.7 4.3 0.10 0.00 73.6 1.1 0.34 0.03 0.00 0.01 0.3 100.0 PZ1 K subsoil 65−75 4.8 4.5 0.17 0.02 9.9 9.1 0.63 0.04 0.00 0.01 1.0 70.9 PZ1 K subsoil 75−96 4.9 4.6 0.13 0.02 8.6 9.0 0.56 0.04 0.00 0.01 0.8 74.5 PZ1 K subsoil 96−120 5.0 4.7 0.08 0.01 7.5 7.9 0.57 0.04 0.01 0.01 0.8 83.0 PZ1 R topsoil 0−9 4.3 3.3 1.72 0.12 14.8 13.5 0.28 0.09 0.08 0.02 2.4 19.3 PZ1 R subsoil 9−19 4.1 3.3 0.57 0.04 13.6 9.1 0.16 0.04 0.02 0.02 1.1 21.0 PZ1 R subsoil 19−32 4.0 3.6 0.68 0.06 11.7 21.0 0.19 0.04 0.01 0.01 2.3 11.6 PZ1 R subsoil 32−48 4.4 4.0 0.44 0.04 11.2 9.3 0.15 0.03 0.01 0.01 1.4 14.6 PZ1 R subsoil 48−70 4.5 4.1 0.15 0.02 9.2 7.1 0.14 0.02 0.01 0.01 0.9 18.7 PZ2 K topsoil 0−30 4.0 2.9 3.81 0.13 29.2 10.6 1.34 0.13 0.04 0.04 17.8 8.8 PZ2 K topsoil 30−50 4.5 3.7 4.84 0.09 56.9 5.5 10.48 0.16 0.02 0.07 24.4 43.9 PZ2 K subsoil 50−65 4.9 4.3 0.36 0.00 101.5 5.3 1.02 0.05 0.00 0.02 1.3 80.9 PZ2 K subsoil 65−78 5.0 4.6 0.06 0.01 8.6 3.2 0.34 0.03 0.00 0.02 0.2 100.0 PZ2 K subsoil 78−88 4.7 4.2 0.39 0.03 12.4 44.8 0.26 0.04 0.01 0.03 2.3 14.9 PZ2 K subsoil 88−98 4.8 4.4 0.27 0.02 13.3 26.4 0.20 0.03 0.01 0.02 1.6 17.1 PZ2 K subsoil 98−108 4.7 4.3 0.31 0.02 12.7 17.4 0.11 0.02 0.03 0.03 1.7 11.0 PZ2 K subsoil 108−140 4.8 4.5 0.13 0.01 10.6 9.1 0.07 0.02 0.01 0.02 0.7 17.8 PZ2 R topsoil 0−11 4.6 3.4 1.78 0.12 14.9 19.0 0.33 0.12 0.08 0.03 3.2 17.6 PZ2 R subsoil 11−40 4.7 4.0 0.10 0.01 9.4 4.9 0.10 0.02 0.00 0.02 0.3 42.6 PZ2 R subsoil 40−46 4.0 3.5 0.95 0.06 14.8 84.0 0.16 0.04 0.02 0.02 4.4 5.5 PZ2 R subsoil 46−57 4.2 3.8 0.58 0.04 14.8 23.1 0.11 0.03 0.01 0.02 2.3 7.5 PZ2 R subsoil 57−65 4.6 4.3 0.38 0.03 12.6 11.6 0.10 0.02 0.01 0.02 1.2 12.0 PZ2 R subsoil 65−90 4.4 4.0 0.10 0.01 8.3 7.5 0.09 0.02 0.03 0.01 1.1 14.2

317

Appendices

Appendix 2. The X-Ray spectra of the clay fraction of topsoil samples from a subset of paired charcoal kiln sites and reference soils: (a) site 7, (b) site 8 and (c) site 9. The arrows indicate the interplanar spacing corresponding to each peak, in Angström (Å) On the < 2 mm soil, sand (50–2000 µm) was separated from silt and clay by wet sieving in an ultrasonic bath. Silt (2–50 µm) was separated from clay (0– 2 µm) by gravimetric sedimentation after dispersion with sodium hexametaphosphate (Na-HMP). The clay fraction was treated with NaOCl at 40°C and dithionite citrate bicarbonate (DCB). Subsamples were saturated with K+ or Mg2+ using chloride salts. The K-saturated samples were subjected to four heat treatments (20, 105, 300 and 550°C) and one Mg-saturated sample was treated with ethylene glycol (Robert & Tessier, 1974). Clay minerals were identified by X-ray diffraction (XRD) with CuKα radiation in a Bruker Advance diffractometer. The XRD spectra have a comparable signature in the topsoil of the kiln to that of the reference soil, which suggests that charcoal production had a limited effect on clay mineralogy. Kaolinite is present in each sample (peak at 7.0 Å that disappears after heating at 550°C), which supports the idea that temperature was below 550°C in the soil subjected to charcoal production. Illite is present in each sample (peak at 10 Å for each treatment). Vermiculite is detected in each sample as well (peak at 14 Å for the K 20°C, Mg and Mg glycol spectra), but in a smaller amount than kaolinite and illite. Heating at 300 and 550°C shrinks vermiculites interstratified with illites, which decreases the peak at 14 Å and increases the peak at 10 Å. Trace amounts of chlorite cause a residual peak at 14 Å after heating at 550°C. Treatment with ethylene glycol swells smectites, which results in a loose peak around 17 Å. In the absence of ethylene glycol, smectites contribute to the signal at 14 Å with vermiculites. Smectite content is small for all sites except for one (Figure S4, c). This charcoal kiln site was in the middle of a thalweg where colluvium accumulated. The greater susceptibility of swelling clay to water erosion may explain the enrichment in smectites. Moreover, the reference soil was sampled slightly higher on the bank slope, which might explain the smaller enrichment of the reference soil.

318

Appendices

319

Appendices

320

Appendices

321

Appendices

Appendix 3. Raw data for charcoal particles (Chapter 10)

Land Time of Organic Organic ID C N H O use cultivation C O % % % Yrs % mass % mass % mass mass mass mass Active 1 0 71.28 0.87 1.77 11.34 71.03 10.34 kiln 2 Forest 0 54.28 0.52 2.61 26.67 54.03 25.67 3 Forest 0 50.82 0.70 2.73 27.36 50.56 26.33 4 Forest 0 52.30 0.51 2.74 27.81 52.20 27.41 5 Cropland 0 48.07 0.55 2.88 27.31 47.55 25.25 6 Cropland 0 49.30 0.63 2.89 28.77 48.69 26.34 7 Cropland 0 48.43 0.50 2.90 27.17 47.97 25.32 8 Cropland 2 46.54 0.53 2.99 26.91 45.90 24.35 9 Cropland 2 45.43 0.51 2.92 27.57 44.95 25.64 10 Cropland 2 47.96 0.44 2.84 28.74 47.51 26.94 11 Cropland 23 47.62 0.67 2.80 28.22 47.38 27.25 12 Cropland 23 48.38 0.59 2.60 27.18 47.95 25.48 13 Cropland 23 50.02 0.62 2.79 27.27 49.60 25.60 14 Cropland 23 47.20 0.56 2.75 25.72 46.55 23.11 15 Cropland 23 45.10 0.49 2.85 27.24 44.47 24.73 16 Cropland 23 46.36 0.50 2.83 27.31 45.69 24.63 17 Cropland 130 45.69 0.60 2.62 25.38 45.00 22.61 18 Cropland 130 45.76 0.70 2.72 24.91 45.00 21.85 19 Cropland 130 45.14 0.82 2.58 26.19 44.29 22.81 20 Cropland 155 46.05 0.66 2.55 25.37 45.16 21.82 21 Cropland 155 44.59 0.64 2.47 25.81 43.62 21.93 22 Cropland 155 46.14 0.62 2.47 24.79 45.17 20.91 23 Cropland 200 43.28 0.64 2.53 25.19 42.14 20.60 24 Cropland 200 45.57 0.61 2.37 24.21 44.53 20.04 25 Cropland 200 45.13 0.53 2.44 25.96 44.14 22.01 26 Cropland 200 43.05 0.65 2.46 25.24 41.94 20.80 27 Cropland 200 41.25 0.56 2.51 25.63 40.30 21.86 28 Cropland 200 42.12 0.49 2.46 26.06 41.44 23.33

322

Appendices

Sample Land Time of 2- CO3 Si Al Fe Ca CW&B ID use cultivation % yrs % mass % mass % mass % mass % mass mass Active 1 0 1.24 1.459 0.863 0.156 0.602 4.67 kiln 2 Forest 0 1.25 2.399 1.819 0.311 0.061 8.58 3 Forest 0 1.29 3.149 2.018 0.418 0.353 10.58 4 Forest 0 0.50 2.265 1.243 0.328 0.036 4.92 5 Cropland 0 2.58 3.233 2.38 0.669 0.735 14.96 6 Cropland 0 3.04 2.743 2.068 0.368 0.739 18.94 7 Cropland 0 2.30 2.969 1.935 0.587 0.636 8 Cropland 2 3.20 3.279 2.424 0.798 0.652 26.91 9 Cropland 2 2.41 2.955 2.381 0.86 0.587 16.45 10 Cropland 2 2.25 3.144 2.585 0.552 0.609 11 Cropland 23 1.21 2.62 1.806 0.235 0.7215 13.08 12 Cropland 23 2.13 3.708 2.852 0.8 0.383 13.04 13 Cropland 23 2.09 3.211 2.203 0.822 0.458 14 Cropland 23 3.26 3.949 2.686 0.589 0.834 12.37 15 Cropland 23 3.14 4.072 2.806 0.552 0.763 15.67 16 Cropland 23 3.34 3.01 2.297 0.657 0.697 17 Cropland 130 3.46 5.217 3.481 1.046 0.789 10.75 18 Cropland 130 3.82 4.542 2.695 1.249 0.966 14.78 19 Cropland 130 4.23 5.692 2.608 1.352 1.079 20 Cropland 155 4.43 5.859 2.823 0.947 1.04 13.88 21 Cropland 155 4.84 5.301 3.086 0.74 1.252 15.77 22 Cropland 155 4.85 5.507 2.721 0.874 1.214 23 Cropland 200 5.73 6.613 3.718 0.752 1.118 14.78 24 Cropland 200 5.20 6.473 3.469 0.954 1.067 14.78 25 Cropland 200 4.94 5.461 2.794 1.081 1.072 26 Cropland 200 5.56 5.598 3.206 1.004 0.867 27 Cropland 200 4.71 5.351 3.339 1.245 0.958 11.61 28 Cropland 200 3.41 5.39 3.119 0.955 1.048 18.22

323