Phosphorus Dynamics in an Artificially Drained Histosol
Geneviève Grenon
Department of Bioresource Engineering
McGill University, Montreal
Submitted December 2020
A Thesis Submitted to McGill University in partial fulfillment of the requirements for the degree
of Doctor of Philosophy
© Geneviève Grenon 2020
i Abstract
Over-use of phosphorus (P) and nitrogen (N) fertilization is common in agricultural
Histosols, or organic soils, and can be directly linked to eutrophication in lakes and rivers. In these soils, subsurface tile drainage is essential to remove excess water, thereby enhancing crop productivity. This is especially important given the fact that originally Histosols are derived from flat, low lying peatlands. Although studies have been conducted on P mitigation within organic soils, application of drainage water management (DWM) as a P mitigation strategy in these soils, has not been extensively studied. The Holland Marsh of Ontario is an organic soil agricultural area that contributes to the eutrophic conditions of Lake Simcoe. The overall objective of this research was to assess the P dynamics in subsurface tile drainage water under two DWM practices to improve P management on organic soils.
The first sub-objective of this thesis investigated the applicability of controlled water table management to mitigate high nutrient loads in drainage discharge. A field research was conducted over two years (2015 – 2016) to examine the impacts of a controlled drainage (CD) structure, placed on the collector tile line, on seasonal N and P concentrations and loads within the Holland
Marsh. The nutrient loads varied depending on CD discharge intensity which was influenced by annual spring thaw and high precipitation events that were more than 70% above the 30-year monthly average. The second sub-objective of the thesis was to assess the use of an artificial neural network (ANN) model to predict nutrient loads under various water management strategies. The
ANN model outperformed a nonlinear regression model. Furthermore, the results showed that seasonal winter-spring and summer models were more accurate when predicting the total P (TP) and total N (TN) loads, compared to the nitrate (NO3-N) and orthophosphate (PO4-P) loads. An investigation of different water table management scenarios, using ANN predictions, revealed that
ii raising the water table to 34 cm or less from the soil surface in the winter-spring allows for a potential reduction in N (14%) and P (34%) loads as a result of reduced drainage water outflow.
Furthermore, summer ANN management scenarios, indicated that if the water table was set to 77 cm from the soil surface, it has the potential to reduce both the water discharge and nutrient loads, as well as satisfy the water depth for optimal crop production.
In the Holland Marsh, many farmers use a pump drainage (PD) system, which pumps excess drainage water from a sump at the collector outlet and into a ditch, only on an as-needed basis. The third sub-objective of this study assessed the P movement through the soil in various P pools and their link to tile drainage water quality. This study quantified the different soil P pools found in organic soils under two water management practices (CD and PD) in 2016. Soil samples from three sampling periods through the growing season (pre-fertilizer, mid growing season, post- harvest) were analyzed for P by sequential fractionation as well as available Bray-1 and microbial biomass. In addition, the carrot tuber P content and the drainage discharge water quality were measured. The results identified calcium (Ca) bound P as the largest P pool, which acts as a sink within organic soils. The correlation analysis further identified the aluminum (Al)-iron (Fe) bound
P as a driving force for P movement in the soil, as it had the most significant relationship with both the soil P parameters and the drainage water quality. A linear and quadratic regression analysis of
TP found that the P concentration in drainage discharge was significantly related to the fertilizer application and to crop harvest at both sites. The 2016 P balance indicated that the P fertilizer inputs into the system were greater than the nutrient outflow through drainage loads and crop uptake, signifying that there is a constant P accumulation taking place. Overall, the P dynamics within the soil drives and changes the P concentrations within the drainage water identifying this interconnectivity of the nutrient cycle.
iii The final sub-objective of this study assessed the water quality loads and their relationship to discharge under the two DWM systems (CD and PD). The results indicated high N loads during periods of increased discharge volume, therefore higher under PD in 2015 (21 kg ha-1) and under
CD in 2016 (53 kg ha-1). The correlation and regression analysis found no relationship between the discharge and N loads, indicating that N dynamics are governed more by biochemical and hydrological soil characteristics than drainage volumes. The rate of P loads between 2015 and
2016 was found to decrease under CD. Additionally, the TP loads and discharge volumes resulted in a significant correlation and regression relationship under both DWM systems. This thesis concludes that both CD and PD are viable DWM strategies on Histosols for nutrient load reduction.
The thesis also concludes that the soil P dynamics influences the P concentrations from drainage discharge.
iv Résumé
La fertilisation au phosphore (P) et à l'azote (N) est surutilisée dans les Histosols agricoles, ou sols organiques, et peut être directement liée à l'eutrophisation des lacs et des rivières. Dans ces sols, le drainage souterrain est essentiel pour éliminer l'excès d'eau, ce qui améliore la productivité des cultures. Ceci est particulièrement important dans les Histosols étant donné qu'à l'origine, ils
étaient des tourbières plates et basses. Bien que des études aient été menées sur l'atténuation du P dans les sols organiques, l'application de la gestion des eaux de drainage (DWM) comme stratégie d'atténuation du P dans ces sols n'a pas été étudiée de manière approfondie. Le Holland Marsh de l'Ontario est une zone d'agriculture biologique qui contribue aux conditions d'eutrophisation du lac
Simcoe. L'objectif général de cette recherche était d'évaluer la dynamique du P dans les eaux de drainage souterrain en appliquant deux pratiques de DWM pour améliorer la gestion du P dans les sols organiques.
Le premier sous-objectif de cette thèse était d'étudier l'applicabilité de la gestion contrôlée de la nappe d’eau pour atténuer les charges élevées de nutriments dans les eaux de drainage. Une recherche sur le terrain a été menée pendant deux ans (2015 - 2016) pour examiner les impacts d'une structure de drainage contrôlé (CD), placée sur la ligne de tuiles collectrices, sur les concentrations et les charges saisonnières de N et P dans le Holland Marsh. Les charges en nutriments ont varié en fonction de l'intensité du drainage contrôlé, qui a été influencé par le dégel printanier annuel et les fortes précipitations qui ont dépassé de plus de 70% la moyenne en 30 ans.
Le deuxième sous-objectif de la thèse était d'évaluer l'utilisation d'un modèle de réseau neuronal artificiel (ANN) pour prévoir les charges en nutriments dans le cadre de diverses stratégies de gestion de l'eau. Le modèle ANN a eu de meilleures performances qu'un modèle de régression non linéaire. De plus, les résultats ont montré que les modèles saisonniers hiver-printemps et été sont
v plus précis pour les charges P totale (TP) et N total (TN), par rapport aux charges nitrate (NO3-N) et orthophosphate (PO4-P). L’évaluation de différents scénarios de gestion de la nappe d’eau, utilisant ANN, a démontré que l'élévation de 34 cm ou moins de la surface du sol en hiver- printemps permet une réduction potentielle des charges de N (14 %) et de P (34 %) en raison d’une réduction des eaux de drainage. En outre, les scénarios de gestion de l'ANN en été ont indiqué que si la nappe d’eau était fixée à 77 cm de la surface du sol, elle pourrait réduire à la fois la décharge d'eau et les charges en nutriments, aussi bien que de satisfaire la profondeur de l'eau pour une production agricole optimale.
Dans le Holland Marsh, de nombreux agriculteurs utilisent un système de drainage par pompage (PD), qui pompe l'excès d'eau de drainage d'un puit collecteur vers un fossé. Le troisième sous-objectif de cette étude consistait à évaluer le mouvement du P dans le sol dans différents bassins de P et leur lien avec la qualité de l'eau de drainage. Cette étude a quantifié les différents bassins de P du sol trouvés dans les sols organiques selon deux pratiques de gestion de l'eau (CD et PD) en 2016. Des échantillons de sol provenant de trois périodes d'échantillonnage au cours de la saison de croissance (pré-fertilisation, milieu de la saison de croissance, post-récolte) ont été analysés pour le P par fractionnement séquentiel ainsi que pour le Bray-1 et la biomasse microbienne. Le P des tubercules de carotte et la qualité des eaux de drainage ont aussi été mesurés.
Les résultats ont permis d'identifier le P lié au calcium (Ca) comme étant le plus grand réservoir de P, qui agit comme un puits dans les sols organiques. L'analyse de corrélation a également permis d'identifier le P lié à l'aluminium (Al)-fer (Fe) comme une force motrice pour le mouvement du P dans le sol, car il a la relation la plus significative avec les paramètres du P du sol et la qualité de l'eau de drainage. Une analyse de régression linéaire et quadratique de la TP a montré que la concentration de P dans les eaux de drainage était significativement liée à l'application d'engrais
vi et à la récolte des cultures sur les deux sites. Le bilan de P de 2016 a indiqué que les apports d'engrais P dans le système étaient supérieurs à la sortie d'éléments nutritifs par les charges de drainage et l'absorption des cultures, ce qui signifie qu'il y a une accumulation constante de P.
Dans l'ensemble, la dynamique du P dans le sol entraîne et modifie les concentrations de P dans l'eau de drainage, ce qui met en évidence cette interconnectivité du cycle des éléments nutritifs.
Le dernier sous-objectif de cette étude a évalué les charges de qualité de l'eau et leur relation avec les rejets dans les deux systèmes de gestion des eaux de drainage (CD et PD). Les résultats ont indiqué des charges d'azote élevées pendant les périodes d'augmentation du volume des rejets, donc plus élevées sous PD en 2015 (21 kg ha-1) et sous CD en 2016 (53 kg ha-1). Les analyses de corrélation et de régression n’ont permis d’établir aucune relation entre le débit et les charges d'azote, ce qui indique que la dynamique de l'azote est davantage régie par les caractéristiques biochimiques et hydrologiques du sol que par les volumes de drainage. Le taux de charges de P entre 2015 et 2016 a diminué sous la CD. En outre, les charges P et les volumes de décharge ont donné lieu à une corrélation et une relation de régression significatives dans les deux systèmes de
DWM. Cette thèse conclut que la CD et la PD sont toutes deux des stratégies de DWM viables sur les Histosols pour la réduction de la charge en nutriments. La thèse conclut également que la dynamique du P du sol influence les concentrations de P provenant des rejets de drainage.
vii Dedication
This piece of work is dedicated to perseverance, commitment, and the will to never give up through all of life’s trials. To all those who follow: the only lessons learned are those that we fight for.
viii Acknowledgements
Special thanks to my parents, Richard and Louise, and siblings, Mel and Nick, for always giving me the support, encouragement and the love I needed in the pursuit of my doctoral degree.
They continued to encourage and support me through all the hard times and without their love and support this doctorate would not have been possible. Thanks to all my friends for their support and understanding throughout these long years.
I am grateful to my supervisor Dr. Chandra Madramootoo for giving me this opportunity all those years ago. His encouragement, mentorship, dedication and support have allowed me to successfully journey through this Ph.D. program. Although multiple trials ensued, Chandra’s belief in me allowed all problems to become surmountable. Thank you for all the help.
I am grateful for the support from all members of the Water Innovation Lab, McGill
University, both past and present. Without their motivation and encouragement, the journey would have been so much harder. I say a very big thank you to Wendy Ouellette, Yari Aghil, Kaitlin
Lloyd, Cynthia Creze, Divya Gupta, Naresh Gaj, Samuel Ihuoma, Naeem Abbasi, Bhesram Singh,
Mfon Essien, Kosoluchukwu Ekwunife, Aidan De Sena, Naresh Arumugagounder Thangaraju,
Harsimar Sidana, and Anshika Jain. Many friendships were made that will last the test of time.
Special thanks to Kenton Ollivierre for his help and support during the field setup of the experiments and to Kerri Edwards for taking field samples every week for 2 years even through the winter months. Without their help, the project wouldn’t have been possible. I am grateful for all the help from Agriculture and Agri-Food Canada team for data, equipment and help in-field installation. Specifically, thanks to Patrick Handyside, Andrew Jamieson and Sonja Fransen.
Thanks also to technician Helene Lalande for all her support in the lab and the constant running of
ix the water quality and soil samples that were involved. For all the care you put into the results and the insights you gave, thank you.
Furthermore, I would like to thank all summer students that aided me in the field and lab throughout the years. Thank you to Paddy Enright, Laura Gilbert, Marjorie Macdonald, Rose
Seguin, Marie Norwood, and Rebecca Seltzer. A heartfelt thank you to the Muck Crops Research
Centre: Shawn Janse, Mary-Ruth McDonald, Kevin Vander Kooi and all their staff for providing machinery, support and storage facilities throughout the project. Finally, thank you to Dan Sopuch and Doug Weening for allowing their fields to be used throughout the research.
x Contributions of the Authors
Geneviève Grenon, the author of this thesis, prepared the first drafts of all the papers, undertook the revisions, and finalized the papers for submission to the journals. Grenon aided in the design of the experiments, collection of soil and water quality samples, as well as the drainage discharge data. She further conducted the lab analyses and worked with the summer students on the lab and field measurements, while she analyzed all the data reported in the manuscripts. Dr.
Chandra A. Madramootoo, James McGill Professor, Department of Bioresource Engineering was the primary supervisor of this thesis and provided valuable knowledge on all aspects of the research contributing to the conceptualization and validation of the research, and the review and editing of each paper, poster and oral presentation.
Bhesram Singh, a Ph.D. Candidate in Bioresource Engineering McGill University, contributed to the investigation through soil sampling, the maintenance of the research sites and the calibration and validation of all water discharge data. Also, B. Singh aided in the writing of the original draft for the review (Chapter 2) and assisted in the revision and editing of Chapters 3 and
5. Aidan De Sena, a Ph.D. Candidate in Bioresource Engineering McGill University, contributed to data acquisition through soil sampling and conducted the phosphorus fractionation analysis
(Chapter 4). Furthermore, A. De Sena contributed to the writing of the original draft of the review paper (Chapter 2) and the editing and review of Chapter 4. Dr. Christian von Sperber, Department of Geography, McGill University, provided valuable assistance through the review and editing of all papers (Chapters 2 - 5). Dr. Manish Kumar Goyal, Department of Civil Engineering, Indian
Institute of Technology Guwahati, provided assistance in coding in Matlab for the models used in
Chapter 3 and edited the manuscripts in Chapters 2 and 3. Dr. Abderrachid Hamrani provided assistance with coding in Matlab for the models used in Chapter 4. Dr. T.Q. Zhang, Agriculture
xi and Agri-Food Canada, Harrow Research and Development Centre provided assistance in editing the review manuscript (Chapter 2).
List of publications and scientific presentations related to the thesis:
A. Thesis components that have been or will be submitted for publication in peer-review journals
• Grenon, G., Singh, B., De Sena, A., Madramootoo, C.A., von Sperber, C., Goyal, M.K.,
Zhang, TQ., 2020. Phosphorus Fate, Transport and Management on Subsurface Drained
Agricultural Organic Soils: A Review. (Accepted: Environmental Research Letters,
https://doi.org/10.1088/1748-9326/abce81).
• Grenon, G., De Sena, A., Madramootoo, C.A., von Sperber, C., Hamrani, A., 2020.
Linking soil phosphorus pools to drainage water quality in intensively cropped organic
soils. (Accepted: Science of Total Environment).
• Grenon, G., Madramootoo, C.A., Singh, B., von Sperber, C., 2020. Nutrient Loads
Pollutant in Drainage Discharge from Two Methods of Water Management. (In
preparation).
• Grenon, G., Madramootoo, C.A., Singh, B., Goyal, M.K., von Sperber, C., 2020.
Seasonal effects of controlled drainage on water quality from agricultural organic soils
and the predictive use of artificial neural networks. (In preparation).
B. Thesis components that have been presented at scientific conferences
• Grenon, G., Madramootoo, C.A. Singh, B., von Sperber, C. 2020. Nutrient Loads in
Drainage Discharge from Two Methods of Water Management in Organic Soils.
Technical Talks/Webinar. Canadian Society for Bioengineering (CSBE-SCGAB). Oral
presentation. Online. September 11.
xii • Grenon, G., Madramootoo, C.A. 2018. Controlled Drainage to Improve Water Quality
in the Holland Marsh, Ontario. 61st annual Conference on Great Lakes Research
(IAGLR). Oral presentation. Scarborough, Ontario. June 18-22.
• Grenon, G., Madramootoo, C.A. Singh, B., Gaj. N. 2016. Water quality management
in the Holland Marsh, Ontario. Annual International Meeting of the American Society
of Agricultural and Biological Engineers (ASABE). Paper and poster presentation.
Orlando, Florida, July 17-20. doi: 10.13031/aim.20162456240
o Award: One of four outstanding NRES poster presentations
• Grenon, G., Madramootoo, C.A. 2016. Water quality management in the Holland
Marsh, Ontario. 69th National Conference and Annual Meeting of the Canadian Water
Resources Association (CWRA) Conference. Oral presentation. Montreal, Quebec, May
25-27.
• Grenon, G., Madramootoo, C.A. 2016. Agricultural management systems to improve
water quality in the Holland Marsh. 65th Annual Muck Vegetable Growers Conference.
Oral presentation. Bradford, Ontario, April 12 -13.
• Grenon, G., Madramootoo, C.A., Gaj, N. 2015. Agricultural management systems to
improve water quality into Lake Simcoe. Annual International Meeting of the American
Society of Agricultural and Biological Engineers (ASABE). Oral presentation. New
Orleans, Louisiana, July 28.
• Jamieson, A., Grenon, G., Madramootoo, C.A. 2015. Controlled drainage research to
quantify water quality benefits. 68th National Conference and Annual Meeting of the
Canadian Water Resources Association (CWRA). Paper and oral presentation
Winnipeg, Manitoba, June 2.
xiii • Grenon, G., Madramootoo, C.A. 2015. Agricultural management systems to improve
water quality into Lake Simcoe. 64th Annual Muck Vegetable Growers Conference. Oral
presentation. Bradford, Ontario, April 7-8.
• Madramootoo, C.A., Grenon G. 2015. Controlled drainage in the Holland Marsh,
Ontario. 57th Annual Convention of the Land Improvement Contractors of Ontario.
Invited presentation. London, Ontario, January 21.
xiv Table of Contents Abstract ...... ii Résumé ...... v Dedication ...... viii Acknowledgements ...... ix Contributions of the Authors ...... xi List of Figures ...... xviii List of Tables ...... xx List of Abbreviations ...... xxii 1 CHAPTER I ...... 1 General Introduction ...... 1 1.1 Background of the study ...... 1 1.2 Problem Statement ...... 2 1.3 Research Objectives ...... 3 1.4 Thesis Format ...... 4 2 CHAPTER II...... 5 Literature Review...... 5 2.1 Introduction ...... 5 2.2 Organic soils and its cultivation ...... 8 2.2.1 Soil structure ...... 10 2.2.2 Organic soil hydraulic characteristics ...... 11 2.2.3 Agricultural production effects ...... 13 2.3 Phosphorus dynamics in agricultural organic soils ...... 15 2.3.1 Abiotic soil P-pools ...... 19 2.3.2 Biotic soil P-pools ...... 23 2.3.3 Phosphorus forms in leached water ...... 25 2.3.4 Modeling P dynamics ...... 25 2.4 Phosphorus concentrations and loading under tile drainage ...... 26 2.4.1 Climatic effects ...... 28 2.5 Water management for organic soil P ...... 31 2.5.1 Controlled drainage (CD) ...... 32 2.5.2 Pump drainage (PD) ...... 33 2.5.3 Drainage through open channels ...... 34 2.6 Summary and recommendations ...... 34 Connecting text ...... 36 3 CHAPTER III ...... 38 Seasonal effects of controlled drainage on water quality from agricultural organic soils and the predictive use of artificial neural networks ...... 38 3.1 Abstract ...... 38 3.2 Introduction ...... 39
xv 3.3 Materials and Methods ...... 41 3.3.1 Study area ...... 41 3.3.2 Controlled drainage system ...... 43 3.3.3 Water quality analysis ...... 44 3.3.4 Seasonality statistics ...... 46 3.3.5 Mathematical Modeling of nutrient loads ...... 46 3.3.6 Water management scenario predictions using ANN ...... 50 3.4 Results and discussion ...... 51 3.4.1 Weather conditions ...... 51 3.4.2 Seasonality of tile drain discharge ...... 52 3.4.3 Seasonal effects of nutrient concentrations ...... 54 3.4.4 Seasonal effects of P and N load ...... 58 3.4.5 Performance comparison of ANN and nonlinear regression ...... 61 3.4.6 Load prediction performance using ANN ...... 64 3.4.7 Water table management predictions using ANN ...... 67 3.5 Conclusion ...... 71 Connecting Text ...... 72 4 CHAPTER IV ...... 73 Linking soil phosphorus pools to drainage water quality in intensively cropped organic soils ... 73 4.1 Abstract ...... 73 4.2 Introduction ...... 74 4.3 Materials and Methods ...... 76 4.3.1 Study area and soil sampling ...... 76 4.3.2 Soil phosphorus analyses ...... 78 4.3.3 Water quality analysis ...... 80 4.3.4 Soil-water P balance ...... 81 4.3.5 Statistical Analysis ...... 81 4.4 Results and Discussion ...... 83 4.4.1 Soil chemistry ...... 83 4.4.2 Relationship between the soil P pools and soil chemical properties ...... 87 4.4.3 Concentrations of DRP, DOP, and TP in drainage water ...... 90 4.4.4 Relationship between soil and water P parameters ...... 91 4.4.5 Regression models ...... 93 4.4.6 Phosphorus balance through the soil matrix ...... 97 4.5 Conclusions ...... 98 Connecting Text ...... 100 5 CHAPTER V ...... 101 Pollutant Nutrient Loads in Drainage Discharge from Two Methods of Water Management ... 101 5.1 Abstract ...... 101 5.2 . Introduction ...... 102
xvi 5.3 Methodology ...... 103 5.3.1 Study area ...... 103 5.3.2 Agronomic practices ...... 104 5.3.3 Water management ...... 105 5.3.4 Water quality analysis ...... 106 5.3.5 Statistical Analysis ...... 108 5.4 Results and Discussion ...... 109 5.4.1 Drainage discharge analysis ...... 109 5.4.2 Nitrogen load assessments ...... 112 5.4.3 Phosphorus load assessments ...... 114 5.5 Conclusion ...... 117 6 CHAPTER VI ...... 119 Summary and Conclusions ...... 119 6.1 General summary ...... 119 6.2 Contributions to knowledge ...... 122 6.3 Recommendations for future research...... 123 References ...... 124
xvii List of Figures
Figure 2.1: Abiotic and biotic phosphorus pools in the soil environment...... 18
Figure 3.1: Field research location from the Lake Simcoe Watershed and to the Holland Marsh
field location...... 42
Figure 3.2: The illustration of a subsurface tile drain outlet with the installation of a controlled
drainage (CD) structure...... 44
Figure 3.3: A typical ANN schematic with a three-layer feed-forward neural network...... 48
Figure 3.4: A 2015 and 2016 time series analysis of the total phosphorus (a) and total nitrogen (b)
concentrations ...... 56
Figure 3.5: Comparison of the predicted and measured TP loads for ANN (a/b) and nonlinear
regression (c/d) models...... 65
Figure 4.1: The change in the mineral concentrations Site 1 (CD) (a) and Site 2 (PD) (b) during
pre-fertilization (PF), growing season (GS) and post-harvest (PH) 2016 season...... 85
Figure 4.2: Pearson correlation (r) for TP between the soil parameters as well as the soil mineral
measured values at each site for 2016...... 89
Figure 4.3: Concentration of dissolved reactive P (DRP), dissolved organic P (DOP) and total P
(TP) in drainage water of Site 1 (a) and Site 2 (b) during the 2016 growing season...... 91
Figure 4.4: The representation of the P soil water balance found at sites 1 and 2...... 97
Figure 5.1: The illustration of the two drainage water management systems: (a) the installation of
a controlled drainage (CD) structure and (b) a pump drainage (PD) system ...... 107
Figure 5.2: A 2015 and 2016 time series analysis of the subsurface tile drain discharge from Site
1 (CD) (black) and Site 2 (PD) (grey)...... 111
xviii Figure 5.3: Pearson correlation (r) distribution of the discharge and the nutrient loads. Both Site 1
(CD) (bottom) and Site 2 (PD) (top) are represented in this distribution...... 113
Figure 5.4: The graphical representation of the normalized (by volume) P loads was released in
2015 and 2016 at both sites...... 116
xix List of Tables
Table 2.1: Mean inorganic P (Pi) and organic P (Po) concentrations in plant-available, Al/Fe-bound
non-labile and Ca-bound non-labile pools of arable organic soils...... 20
Table 2.2: The P concentration and load from different studies in Canada, the United States of
America and New Zealand ...... 29
Table 2.3: Mitigation measures for P used for organic soil agriculture ...... 32
Table 3.1: Monthly precipitation for 2015, 2016 and a 30-year average along with the annual total
and the totals for the spring and summer seasons...... 52
Table 3.2: Descriptive statistics for discharge, TP, PO4-P, NO3-N and TN concentration data for
2015 and 2016 separated into three seasons ...... 53
Table 3.3: Statistical representation of general linear model results identifying seasonal and yearly
nutrient concentration trends...... 58
Table 3.4: Load analysis for 2015 and 2016 separated into three seasons ...... 59
Table 3.5: Statistical representation of SAS results in identifying seasonal and yearly trends for TP
and TN loads as well as discharge...... 60
Table 3.6: Summary of descriptive statistics of water parameters and nutrient load...... 62
Table 3.7: Performance evaluation criteria parameters for spring and summer models of ANN and
nonlinear regression...... 63
Table 3.8: Performance of prediction evaluation for TP and TN ANN seasonal models in the
winter-spring and the summer seasons...... 66
Table 3.9: Predictive water quality loads through CD management of water table from multiple
scenarios in both the winter-spring and summer seasons using the ANN models...... 69
xx Table 4.1: Soil properties measured in October 2015 after the 2015 growing season and before the
2016 season...... 78
Table 4.2: P pool parameters found in the soil expressed as inorganic P (Pi) and organic P (Po) for
both sites, over the three sampling periods ...... 86
Table 4.3: Pearson correlation (r) between the TP and DRP water quality to the soil P chemical
analyses conducted at both sites and during the three sampling periods ...... 92
Table 4.4: Stepwise regression analysis for both sites annually as well as during the three sampling
periods...... 95
Table 4.5: Stepwise regression analysis at the three sampling periods (PF, GS, PH) ...... 96
Table 5.1: Soil properties measured in October 2015...... 104
Table 5.2: Annual discharge and nutrient loads for total P (TP), phosphate (PO4-P), total N (TN)
and nitrate (NO3-N) at Site 1 (CD) and Site 2 (PD)...... 110
Table 5.3: The linear regression analysis of discharge at site 1 (CD) and site 2 (PD) for nutrient
load assessment...... 114
xxi List of Abbreviations
AAFC, Agriculture Agri-Food Canada
Al, Aluminum
ANN, Artificial neural network
ANOVA, Analysis of Variance
Apr, April
BMP, Best Management Practice
C, Carbon
Ca, Calcium
CD, Controlled drainage
CHCl3, Chloroform cm, Centimeters
CO2, Carbon dioxide
Dec, December
DH2O, Deionized water
DOM, Dissolved organic matter
DOP, Dissolved Organic Phosphorus
DP, Dissolved Phosphorus
DRP, Dissolved Reactive Phosphorus
DWM, Drainage Water Management e, Uncertainty error
EAA, Everglades Agricultural Area
EDTA, Ethylenediaminetetraacetic acid
xxii EPIC, Environmental Policy Integrated Climate model
Fe, iron
FeO, Iron (II) oxide
GS, Growing Season
H2O2, Hydrogen peroxide
H2SO4, Sulfuric acid ha, Hectares
HCl, Hydrochloric acid
ICECREAM, ICE (Snow and ice models) CREAM (Chemical, runoff, erosion, agricultural management)
Jan, January
K, Potassium
K(h), Unsaturated hydraulic conductivity
Ksat, Saturated hydraulic conductivity
L, Liters
LSM, Least squared mean
M, Molarity mg, milligrams mL, milliliters mm, millimeters
Mn, Manganese
N, Nitrogen
NA, Not available
xxiii NaHCO3, Sodium bicarbonate
NaOH, Sodium hydroxide
NH4Cl, Ammonium Chloride
NMR, Nuclear magnetic resonance
NO3-N, Nitrate
NSE, Nash-Sutcliffe efficiency
NSERC, Natural Sciences and Engineering Research Council of Canada
O, Organic horizon
Of, fibric horizon
Oh, humic horizon
Om, mesic horizon
O2, Oxygen
Oct, October
OM, Organic Matter
OMAFRA, Ontario Ministry of Agricultural, Food, and Rural Affairs
P, Phosphorus
PD, pump drainage
PF, Pre-fertilizer
PH, Post-Harvest
Pi , Inorganic Phosphorus;
-3 PO4 , inorganic orthophosphate
PO4-P, orthophosphate
Po, Organic Phosphorus;
xxiv PNM, Pore Network Model
PP, Particulate Phosphorus r, Pearson correlation
R, Correlation coefficient
R2, Coefficient of determination
RMSE, Root mean square error
RZWQM2, Root zone water quality model 2 s, Seconds
SAS, Statistical Software Suite
Sept, September
SS, Suspended sediment
TDP, Total dissolved phosphorus
TN, Total nitrogen
TP, Total phosphorus
TRP, Total reactive phosphorus
TSP, Triple superphosphate
휇m, Micrometers
USA, United States of America
xxv 1 CHAPTER I
General Introduction
1.1 Background of the study
Eutrophication is the enrichment of superfluous nutrients, mostly nitrogen (N) and phosphorus (P), resulting in increased algal blooms within water bodies, and is a major environmental concern. Studies have found that P is of greater concern in fresh-water bodies, compared to N when looking at causes of eutrophication (Schindler and Fee 1974; Thomas et al.,
1995). The Florida Everglades, USA have found that even a small increase in P concentration in lakes and rivers can cause the immediate growth of algal blooms (McCormick and O'Dell 1996;
McCormick and Stevenson 1998; Noe et al., 2001).
Cultivation of organic soils within the watersheds of Lake Okeechobee (Daroub et al.,
2011), Lake Ontario (Longabucco and Rafferty, 1989) and Lake Simcoe (Winter et al., 2007) have been linked to the excessive eutrophication in their respective water bodies. Organic soils consist of 20% or more organic matter and account for 3% of land globally (Mukherjee and Lal, 2015;
Zheng et al., 2014; Silva, 2012). These soils are often low-lying with high water tables, necessitating the need for drainage infrastructure in the form of subsurface tiles, ditches, dykes, pumping stations and levees (Gambolati et al., 2006; Ilniki, 2003). Drainage management further reduces subsidence, which is a problem for organic soils (Ilniki, 2003). Within the Holland Marsh of Ontario, drainage water management (DWM) practices consist of a pump drainage (PD) system where the subsurface tile line drains into a collector well before the water is pumped into a ditch
(Miller 1979). These systems are turned on by the crop growers, when necessary (Bhadha et al.,
1 2017). Another DWM strategy is the installation of a controlled drainage (CD) structure to manage the water table within fields (Mejia et al., 2000; Skaggs et al., 2012; Williams et al., 2015).
These soils are important within the food production chain as they are ideal for intensive vegetable production. However, the inherent low N and P within these soils often lead to an overuse of inorganic fertilizer to increase crop productivity (Czuba and Hutchinson, 1980; Liator et al., 2004; Parent and Khiari, 2003; Guérin et al, 2007), thereby leading to eutrophication.
Extensive reviews have been conducted on P pollution from arable mineral soils for many decades
(Sims et al., 1998; Radcliffe et al., 2015; Kleinman et al., 2015; Christianson et al., 2016).
However, there is a lack of studies that focus on agricultural organic soils and mitigation practices that can be effective in reducing nutrient outflow in drainage water.
1.2 Problem Statement
Studies in organic soils have mainly focused on greenhouse gas emissions (Joosten, 2009;
Joosten et al., 2012; Fell et al., 2016) because of the large quantities of carbon (C) found in these soils. Phosphorus studies have been done on the assessment of the use of mineral soil P tests in organic soils, identifying the Bray-1 P test and the iron (II) oxide (FeO) filter paper extractable P as the soil tests that can be related to P concentration in water outflow from organic soils (Zheng et al., 2014, 2015). Results of mineral soil studies cannot be applied to P management in the organic soils because organic soils are unique in their biogeochemical and nutrient properties, as well as their hydrological characteristics. In Ontario, the Holland Marsh is an organic soil area that contributes $58 million to the provincial economy annually (Township of King, 2012). Intensive farming in the Holland Marsh has led to high fertilization rates to compensate for low inherent fertility, which has contributed to eutrophication in Lake Simcoe (Longabucco and Rafferty, 1989;
2 Sims et al.,2000; Miller, 1979). Most of the drainage water quality studies that have been
conducted within the Holland Marsh are outdated (Miller 1979; Cogger and Duxbury, 1984).
Presently, research into the effects of subsurface drainage P transport in organic soils is
limited (Miller 1979; Robinson, 1986) and there are little or no studies on nutrient loss mitigation
management practices on organic soils (Daroub et al., 2011). The DWM system is necessary for
organic soils to mitigate the inherent high-water table. However, the CD system has been
extensively studied on mineral soils (Elmi et al., 2002; Skaggs et al., 2010, 2012; Williams et al.,
2015; Yousseff et al., 2018), but not on Histosols. Furthermore, the study of current PD systems
on these soils also have limited studies (Bhadha et al., 2017). The effects of these DWM systems
on the drainage water quality (P concentrations and loads) is necessary to increase the
environmental sustainability of organic soils.
1.3 Research Objectives
The overall objective of this research was to evaluate the effect of drainage water
management on the P water quality from subsurface tile-drained agricultural organic soils. The
study assessed the P movement through the soil and into the drainage water, creating a complete
view of the P dynamics within the ecosystem. The research was conducted as a case study of the
Holland Marsh, located in Ontario, Canada.
The overall objective was achieved through the following specific sub-objectives: i. Assess the suitability of a controlled drainage structure as a water management strategy
through the evaluation of the seasonal variations in nutrient water quality and subsurface tile
drain discharge
3 ii. Evaluate the use of an artificial neural network model for nutrient load predictions on
cultivated organic soils. Furthermore, to assess the potential of the model to forecast various
water-table scenarios iii. Investigate the movement of P through the soil and into the drainage water by linking the
dynamics of different soil P-pools to changes in the chemistry of drainage water. iv. Evaluate the relationship between the nutrient loads and the drainage discharge under two
drainage water management practices: a controlled drainage structure, and a pump drainage
system.
1.4 Thesis Format
The format of this thesis is manuscript-based. The first chapter consists of the general
introduction presenting the background research, problem statement and objectives of the study.
The second chapter contains an extensive literature review on organic soils and the movement of
phosphorus (P) through subsurface tile-drained agricultural areas. The review article outlined the
challenges and uniqueness of organic soil agriculture, in comparison to mineral soils, as well as
highlight mitigation practices that have been used. It concluded with gaps in knowledge to be filled
through future research needs and perspectives that formed the core of this study. Chapters 3, 4,
and 5 present manuscripts on the results from this research. All figures and tables are integrated
into the papers and all references can be found at the end of the thesis.
4 2 CHAPTER II
Literature Review
2.1 Introduction
Organic soils, also known as muck or peat soils are composed of at least 20-35% organic matter (Kroetsch et al., 2011; Silva, 2012; Soil Survey Staff, 2014; Mukherjee and Lal, 2015) and account globally for approximately 4.23 million km2 (Xu et al., 2018), covering nearly 3% of the global land area (Joosten et al., 2012; Tubiello et al., 2016; Xu et al., 2018). Of the global peatlands converted for agricultural use, Tuebillo et al. (2016) estimates that 60% are in cool temperate/boreal climates, 34% in tropical and 5% in warm temperate areas. Arable organic soils are peatlands that were artificially drained and cleared of vegetation for agricultural purposes
(Keller and Medvedeff, 2016). In regions of North and South America, Europe and South Asia, organic soils are very favourable for the production of high value vegetable crops. There is a large economic return on crop production in these soils, as demonstrated in the Everglades, Florida,
USA, with an agricultural industry worth $1.5 billion annually (Aillery et al., 2001), the San
Joaquin Delta, California, USA, with $702 million in crop revenue (Delta Protection Commission,
2012), and the Holland Marsh, Ontario, Canada, with $58 million annual gross domestic product
(Township of King, 2012). There is an impetus to expand food production on organic soils to meet the nutritional diversity and food security of a growing world population. However, there are concerns as to how this can be achieved while maintaining the environmental safeguards of these very ecologically significant land areas.
Organic soils, although high in organic matter and therefore rich in carbon (C), oxygen and hydrogen, have a low intrinsic soil phosphorus (P) content, which requires frequent P fertilization
5 due to the high P demand for intensive vegetable production (Czuba and Hutchinson, 1980; Liator et al., 2004; Parent and Khiari, 2003; Guérin et al, 2007). P fertilizer is regularly lost through export pathways to the surrounding environment, including runoff and agricultural tile drainage
(Gentry et al., 2007; Chikhaoui et al., 2008). The tile drainage discharge, as source of nutrient pollution, is a major factor in the growing eutrophication of lakes and rivers (Rockwell et al., 2005;
Schindler et al., 2008; King et al., 2015). Eutrophication is the enrichment of water bodies with superfluous nutrients, mostly nitrogen (N) and P, resulting in increased algal blooms, and subsequently leading to the removal of dissolved oxygen upon their decomposition (Spivakov et al., 1999). Cultivation of organic soils within the watersheds of Lake Okeechobee (Daroub et al.,
2011), Lake Ontario (Longabucco and Rafferty, 1989) and Lake Simcoe (Winter et al., 2007) have been linked to the excessive eutrophication in their respective water bodies.
Studies into the effects of subsurface drainage P transport in organic soils are limited
(Miller 1979; Robinson, 1986). Many studies concentrate on soil P tests in organic soils and their link to P within the drainage water, rather than the actual P exported in drainage effluent (Zheng et al., 2014, 2015). Additionally, studies rarely analyze, both, soil P and drainage dynamics in organic soils, as well as their combined effects on P pollution. Some studies have demonstrated that soil P was positively correlated to dissolved P concentrations in drainage waters (Miller,
1979).
In general, agricultural P export studies have primarily concentrated on surface runoff as the major pathway of P transport, compared to subsurface tile drainage, however the flat topography and high water holding capacity of organic soils decrease the runoff potential (Skaggs et al., 1994; Sims et al., 1998; Algoazany et al., 2007; Eastman et al., 2010; Christianson et al.,
2016). Subsurface drainage, also known as tile or artificial drainage, is used in the humid regions
6 of northwestern Europe, the USA and in Canada to lower the water table in the crop root zone
(Madramootoo, 1990; Blann et al., 2009; Gramlich et al., 2018). Recently, there has been an increased interest in better understanding the effects of subsurface drainage in relation to P pollution of water bodies. Sims et al. (1998) concluded that subsurface drainage can be a significant P export pathway in tile drained fields, especially when soils are P-saturated and thereby reducing the P sorption capacity. King et al. (2015) concluded that the hydrology of subsurface drained lands needs further study to better understand P losses under improved drainage and water table control.
Extensive reviews have been conducted on P pollution from arable mineral soils for many decades (Sims et al., 1998; Radcliffe et al., 2015; Kleinman et al., 2015; Christianson et al., 2016).
However, to our knowledge no review has been conducted on drainage impacts of P in organic soils specifically; rather they have only been studied in mineral soils (Thomas et al., 1995,
Haygarth and Jarvis, 1999; King et al., 2015; Gramlich et al., 2018). Increased understanding of nutrient movement in organic soils due to intensive agriculture can contribute to enhanced environmental practices required for sustainable crop production. The objectives of this review are to: i) critically review P losses from subsurface drained organic soils, with a focus on P in the soil- water continuum; ii) determine the effectiveness of drainage water management strategies to reduce P leaching; and iii) identify gaps in knowledge regarding subsurface drainage and P dynamics in peat soils. The review will assess these objectives through the collation of published literature found to be pertaining to P on tile drained organic soils.
7 2.2 Organic soils and its cultivation
Organic soils are highly productive cropping systems throughout the world, in particular for vegetable production. In Europe, a total of about 125,000 km2 of peatland are used for agriculture which represents about 14% of the total peatland area. The largest agricultural areas are located in Russia (70,400 km2), Germany (12,000 km2), Belarus (9631 km2), Poland (7620 km2), and the Ukraine (5000 km2). In some European countries the proportion of cultivation of peatlands to total cultivated land is very high, for example in Hungary (98%), in The Netherlands
(85%), in Germany (80%), in Denmark and Poland (70%) (Lucas et al.,1982; Lappalainen, 1996;
Parent and Ilnicki, 2002; Roßkopf et al.,2015). In the United States, the Florida Everglades is the largest organic soil agricultural area at over 2300 km2 (Kolka et al., 2016), while in Canada
170,000 km2 are used for agriculture (Oleszczuk et al., 2008). Canadian organic soil areas include
Ste-Clothilde, Napierville and Sherrington in Southern Quebec (Browne, 1950) and the Holland
Marsh area within the Lake Simcoe Watershed Ontario (Winter et al., 2007). Subtropical and tropical areas can be found in Asia where in China 2,610 km2 of peatland are used for agriculture
(Laine et al., 2009), and in Indonesia there is a total of 42,000 km2 of peatland used for agriculture,
(Oleszczuk et al., 2008) with Kalimantan, having the largest area of 2,200 km2 (Jauhiainen et al.,
2014; Konecny et al., 2016; Dommain et al., 2018). In South America, it is estimated that 3,140 km2 have been drained, mainly for agriculture (Joosten et al., 2012). Organic soil areas can also be found in South Africa, specifically in northern areas of the KwaZulu-Natal Province (Gabriel et al., 2018). To date, only a small percentage have been converted for agricultural use. In
Indonesia, 20% of organic cropland has been converted for agriculture, while in Malaysia 32%,
Chine 25%, U.S.A. 10%, Europe 14% and Canada 15% have been converted (Oleszcuk et al.,
2008). The potential for expansion of these organic areas could increase in the future, given the
8 northward shift of the agricultural climate zone due to climate warming in countries such as
Canada, Russia, Sweden and Finland (Rosenzweig and Parry, 1994; King et al.,2018).
Apart from their value for agriculture, peatlands are large carbon sinks storing about 16-
24% of earth’s soil carbon (Bridgham et al., 2006; Tubeillo et al., 2016). Prolonged lowering of the water table for cultivation will cause enhanced peat mineralization and, peat subsidence leading to a continuous loss of stored carbon to the atmosphere (Kasimir-Klemedtsson et al., 1997; Zauft et al., 2010; Fell et al.,2016). Peat soils develop through the accumulation of organic matter in areas where an excess of water slows down the decomposition of plant materials. These soils are important habitats for flora and fauna and regulate hydrological cycles. Peatlands describe a diverse range of ecosystems like fens, swamps, or bogs and vary in hydrological sources and dominant vegetation (Kolka et al., 2016). Though often considered as sinks for N and P (Liator et al., 2004), these various types of peatlands can vary in their P status as well, either being oligotrophic (McCray et al., 2012) or eutrophic (Kolka et al., 2016). However, overall, P is more limited in peatland soils compared to N (Hill et al., 2014) and due to the peat accumulation, P often exists primarily in organic forms (Castillo and Wright, 2008). In areas of organic soil farming in
Quebec and Ontario, N and P leaching from these soils are a pollution source for eutrophic algae blooms within surrounding surface waters (Winter et al., 2007; Guérin et al., 2011).
The movement of water is a key driver of the biogeochemical processes in drained lands and a good understanding of the effects of improved drainage on field and landscape hydrology is essential to the understanding of its effects on drainage water quality (Skaggs et al., 1994). Over the past decade, the use of subsurface tile drainage has expanded rapidly in agriculture, yet its potential effect on nutrient loss from agricultural peatlands is not well understood (Kennedy et al.,
2018). Even though the nexus between water flow and solute transport properties in peat soils have
9 been identified, little attention has been given towards investigating this link and few documented studies exist, in particular with regards of P (Gharedaghloo et al., 2018).
2.2.1 Soil structure
The structure of peat and level of decomposition of the organic matter influences its hydraulic and nutrient transport properties. The soil profile can be defined by the organic (O) horizon, which has three distinct horizons within its description. The fibric horizon (Of) consists of readily identifiable fibric material of botanical origin, while the mesic horizon (Om) is material within a stage of partial decomposition with altered both physically and biochemically. The final horizon, most often found in agricultural use is the humic horizon (Oh) which is composed of material in an advanced stage of decomposition with low saturated water holding capacity
(National Research Council of Canada, 1998).
When looking at peat from the microscopic level, there are four types of pores: (i) relatively large pores which are highly irregular and interconnected; (ii) smaller open pores; (iii) dead-end pores, and (iv) closed or partially closed pores. Depending on the nature of these pores, peat is classified as a dual-porosity medium, which includes a “mobile region” through which water and nutrients move relatively easily, and an “immobile region,” where fluid flow velocity, is negligible
(Rezanezhad et al., 2016). The distribution of these pore regions in the peat soils are critical to water movement. Rapid equilibration of solutes and water occur between the mobile and immobile regions resulting in pore spaces that are not affected by advective fluxes but have similar chemical concentrations as other regions. The immobile pores largely regulate microbial biogeochemical cycling in organic soils (McCarter et al., 2020). Water storage, flow and transport of nutrients in peat, including the division of pore water between mobile and immobile regions, are influenced
10 by the depth variations of porosity and pore size distribution, as well as the stratigraphy of the peat soil layer (Rezanezhad et al., 2016).
Another consideration in estimating the movement of water in the unsaturated zone within organic soils is associated with changing pathways as a consequence of drying and shrinkage.
Gharedaghloo et al. (2018) in a recent study on Sphagnum-dominated peat found that the common macro-scale representation of groundwater flow and transport processes, typically utilized in mineral soil, do not adequately simulate variability in flow pathways and dynamics in organic soils. Wang et al (2020) found that the structure of the peat, and the orientation of the undecomposed plant material can direct water flow in soils. A pore network model (PNM) was found to better represent the flow and transport processes within the individual pore structure of the peat soils, changing flow paths and the anisotropy and heterogeneity in Ksat (Gharedaghloo et al., 2018). A significant challenge, however, would be to extrapolate a PNM from a point scale to a field scale, which could improve flux and P loss estimation from the soil profile.
2.2.2 Organic soil hydraulic characteristics
The unique physical properties of peat, including low bulk density and multi-porosity, combined with its ability to swell and shrink upon wetting and drying, limits the application of concepts and methods used to describe the porous media properties of mineral soils (Dettmann et al., 2014; Caron et al., 2015; Rezanezhad et al., 2016). The total porosity of peat soil is complex with pore size ranging from relatively large, inter-particle pores that can actively transmit water and solutes, to relatively small, closed, and dead-end pores formed by the remains of plant cells
(Rezanezhad et al., 2016), which destabilizes the soil matrix.
The geo-mechanical instability of organic soils is well-documented (Huat et al., 2009;
Radcliffe et al., 2015). Water retention capacity for organic soils has been extensively reported
11 (Schwärzel et al., 2002; Hallema et al., 2015; Bechtold et al., 2018; Kennedy et al., 2018; Wallor et al., 2018a, 2018b). However, these results may not be applicable to field scale cultivated organic soils (Hallema et al., 2015), as agricultural practices accelerate the soil-forming processes, leading to changes in the soil physical, chemical and hydraulic properties (Kechavarzi et al., 2010;
Kroetsch et al., 2011).
Managed peat soils exhibit different physical and hydraulic properties, in comparison to other types of soil. The saturated hydraulic conductivity (Ksat) in peat soils has been observed to increase with pore water pressure (Lafond et al., 2014). The unsaturated hydraulic conductivity,
K(h), in peat soils is also dependent on pore water pressure (Hemond and Goldman, 1985; Waine et al., 1985). Further, previous studies have confirmed significant anisotropy in organic soil profiles (Kechavarzi et al., 2010).
Water repellency or hydrophobicity, which is common in peat soils, significantly influences the hydrology within the soil profile; and is particularly related to the degree of decomposition of the material, as well as the initial wetting (Caron et al., 2015). Water repellency is partly responsible for the strong hysteretic properties of these organic layers (Naasz et al., 2008).
With hysteresis, the relationship between hydraulic conductivity, water content and matric potential depends on the direction of wetting (Caron et al., 2015). Preferential flow in peat soils is also a significant hydraulic factor influencing the fate and transport of P. Preferential flow is common in non-structured soils, owing to the development of unstable wetting fronts. Such factors as increasing hydraulic conductivity with depth and strong water repellency in the soil profile, have been found to result in fingers or preferential flow paths in peat soils (Dekker and Ritsema,
2000). As such, the direct and indirect effects of water repellency can lead to an increase in the unstable flow phenomenon in these soils (Lafond et al., 2014). Due to the difficulties in predicting
12 water movement in organic soils, evaluating water fluxes for drainage and irrigation management is complex, requiring direct measurements of flux rates and estimation of flow mechanics (Lafond et al., 2014). The uncertainty associated with the unstable flows may result in difficulties in estimating or predicting P losses from organic soils.
2.2.3 Agricultural production effects
During the past century, major drainage projects occurred using dykes and canals to transform peatland areas into intensive agricultural hubs. Organic soils under arable use are often intensive in nature, growing vegetables like carrots (Daucus carota), celery (Apium graveolens) and onions (Allium cepa), Asian crops such as choy sum (Brassica chinensis) and yow choy
(Brassica rapa), and sugarcane (Saccharum officinarum) in tropical areas (USGS, 1997; Castillo and Wright, 2008; McDonald et al., 2013, 2014). Organic soils are ideal for agricultural production due to their proximity to a water source and ease of root growth, but the success of arable systems on organic soils is dependent on climate, environmental regulations, geomorphology, hydrological regimes, and soil physico-chemical characteristics (Ilnicki, 2003). As organic soils vary considerably in their nutrient status and other physico-chemical properties, farmers employ various strategies to transform the organic soils into suitable substrate for horticultural production. One strategy employed by farmers in western Europe is to incorporate a sand cover into the peat soils, to increase the load bearing capacity of these soils for agricultural production (Ilnicki, 2003).
Furthermore, farmers can amend the organic soils with substantial amounts of lime and fertilizer to adjust pH and nutrient availability to appropriate levels (Castillo and Wright, 2008; Ewing et al., 2012).
The use of P fertilizer on these soils is required to support intensive horticulture production
(Hoffman et al., 1962). However, P fertilization efficiency depends on the amount of precipitation
13 or irrigation, which may allow P to immediately enter surface runoff (Carpenter et al., 1998;
-1 Mbonimpa et al., 2014). Miller (1979) found that up to 37 kg-PO4 ha was lost annually from cultivated organic soil through subsurface tile drainage in Ontario, attributed to over-fertilization.
Additional studies in Ontario comparing carrots and onions yields and change in soil P through the growing season found that there was not a significant increase between crop yield on fields with or without fertilizer. Furthermore, the fertilizer treatments (with or without application) did not show a significant change in the soil P content over time, signifying that the soils were not storing excess P from fertilizer application but that it was being lost through leaching (McDonald et al., 2013, 2014). A study conducted in Quebec showed that the recommended rates of fertilizer
P were an environmental risk, leading to 50% of soluble phosphate accumulating within the top
40 cm (Asselin, 1997). These findings show that over-fertilization can lead to the point where P no longer assimilates within the soil, but rather is fully discharged into the drainage water. Thus, present P application rates are often excessive, and can lead to increased water pollution (Asselin,
1997; Parent and Khiari, 2003). The P pools within the rhizosphere have not been studied to our knowledge, on organic soils, however in soils with inherent low P under legume production, it was found that the addition of fertilizer P resulted in an increase in labile P pools that differed with the ability of the crop to take up P (Hassan et al., 2012). High P build-up in the soil, known as legacy
P (Reddy et al., 2011; Sharpley et al., 2015; Withers et al., 2017) has global implications. In
Florida, approximately 61% of stored P in the Southern Everglades Ecosystem is from the agricultural area (Reddy et al., 2011).
One of the major issues with intensive cultivation of organic soils is increased microbial and chemical oxidation of organic matter and concomitant losses of carbon dioxide to the atmosphere. This loss of organic matter is accelerated by drainage and leads over time to
14 subsidence (Castillo and Wright, 2008). Millette et al. (1982) found that in organic soils, subsidence occurs at a rate between 1 and 7 cm per year in Quebec, while Mirza and Irwin (1964) found a rate of 3.3 cm loss from the Holland Marsh, Ontario. This loss of soil reduces the long- term sustainability of arable farming on organic soils. However, subsidence can be minimized in drained organic soils through drainage infrastructure of subsurface tiles, ditches, pumping stations and levees (Gambolati et al., 2006; Ilnicki, 2003). Given that these areas are often low-lying with high water tables, there is an increased risk of flooding (Ilnicki, 2003). Therefore, water management is practiced on these soils through the installation of tile drains, ditches, and dykes.
Cultivation of organic soils accelerates soil-forming processes such as decomposition and humification, which leads to changes in the physical and chemical soil properties (Kechavarzi et al., 2010; Kroetsch et al., 2011; Liu et al., 2019), and its hydraulic properties. Decomposition of peat soils is accelerated by tillage, fertilization, and drainage practices (Hallema et al., 2015).
Kechavarzi et al. (2010) reported that decomposition and humification result in degradation of the soil structure and shrinkage, which can lead to a decrease in water storage, water transmission and water retention. Further difficulties arise in predicting the drainage properties of cultivated organic soils, as different degrees of decomposition can be found in the same soil profile, leading to stratification where a coarse textured layer (mesic or fibric) is overlaid by a finer layer (Lafond et al., 2014).
2.3 Phosphorus dynamics in agricultural organic soils
Phosphorus (P) is a major nutrient required for crop production but is only available in limited quantities in nature. P is a rock-derived nutrient which enters the soil solution through the dissolution of primary and secondary minerals. The P cycle includes only a negligible gaseous
15 phase and consequently nature depends on the continuing input of P through the weathering of soils (Filippelli, 2008, 2017). Soils in tropical ecosystems are often highly weathered which leads to the fixation and occlusion of P impeding biological uptake. These systems sustain their primary production by efficient recycling mechanisms and by the utilization of P from atmospheric deposition of dust (Chadwick et al.,1999; Bristow et al.,2010).
P exists in a variety of forms, either as organic P (Po) or inorganic P (Pi) (Figure 2.1). Plants
3- and microorganisms usually take up P from the soil solution as inorganic orthophosphate (PO4 ).
Orthophosphate enters the soil solution via dissolution and desorption from primary and secondary minerals or via the hydrolysis of Po by extracellular enzymes, so called phosphatases. These processes operate on different timescales from seconds to millennia and therefore the supply of available P is often limited in natural ecosystems.
Anthropogenic activities have had a severe impact on the environmental P cycle such as increased use of P fertilizer, land-use change and sewage waste contamination. This has led to the global environmental problems of eutrophication, algal blooms, hypoxia and coastal dead zones
(Filippelli, 2017). Intensive cropping relies on the application of P fertilizers so that P is available for immediate uptake by crops (Mbonimpa et al., 2014). Without the application of this essential nutrient, available P in agricultural soils can be as low as 3-25 mg P kg-1 (Negassa and Leinweber,
2009). This is due to the reactivity of the negatively charged oxygen moieties in a phosphate molecule, which are prone to adsorbing to soil particle surfaces via anion exchange (moderately labile P), unless these sites are P-saturated (Whalen and Sampedro, 2010; Doolette and Smernik,
2011). This moderately labile P is in equilibrium with the available P pool, replenishing the depleted pool when kinetically favorable (Whalen and Sampedro, 2010; Doolette and Smernik,
2011; Stutter et al., 2015).
16 The microbiome’s function is crucial in driving agricultural P dynamics. Just as P is available to plants in the soil solution, the same is true for microorganisms, competing directly with plants and crops for limited P (Jones and Oburger, 2011; Wasaki and Maruyama, 2011; Dodd and Sharpley, 2015). Microbes can immobilize both, Pi and some Po compounds (Oberson et al.,
2001; Jansa et al., 2011; Jones and Oburger, 2011). While immobilized, P undergoes modifications and transformations within the membranes of the microbes (Oberson et al., 2011). Sudden fluctuations in the environment, such as decreases in organic C availability, increased P concentrations, drying and rewetting and lysis of microorganisms, can lead to the release of microbial P into the environment (Turner and Haygarth 2001). Upon release, much of this microbial P is Po (Bünemann et al., 2011).
Depending on the mineral composition of the agricultural soil, Pi can also form recalcitrant ionic complexes with aluminum (Al) and iron (Fe), known as sesquisoxides (Al/Fe-bound non- labile P) as well as precipitate with calcium (Ca) (Ca-bound non-labile P). The initial formation of these Al, Fe, and Ca phosphate complexes are amorphous, but can become crystalline structures such as variscite, strengite, and hydroxyapatite, respectively (Negassa and Leinweber, 2009;
Doolette and Smernik, 2011; George et al., 2011). However, these reactions are pH dependent.
When soil pH is more acidic, Pi will react with Al and Fe, while when more alkaline, Ca phosphates are formed (Jones and Oburger, 2011; Shen et al., 2011).
17
Figure 2.1: Abiotic and biotic phosphorus pools in the soil environment.
18 2.3.1 Abiotic soil P-pools
Extensive research has been conducted to characterize P in different abiotic and biotic soil pools. Abiotic pools are often assessed with the Hedley fractionation method (Hedley et al.,1982).
This procedure follows a sequential extraction with chemical solutions of varying strength to separate operationally defined P pools. Distribution of P among these pools differs based on soil order, soil age and land-use and has been cataloged in Cross and Schlesinger (1995) for natural soils, and Negassa and Leinweber (2009) for natural and agricultural soils. While these reviews thoroughly assessed the available literature on P pools in abiotic soils, both studies overlooked organic soils, most likely due to the sparse available research published in the literature.
In the past decade, more studies have been conducted on the abiotic P pools in arable organic soils (Table 2.1). Like arable mineral soils, arable organic soils have larger P pools compared to their natural counterparts due to the accumulation of fertilizer P throughout their cultivation history. Despite their differences in soil physical-chemical properties, the TP stocks of arable organic soils (340–2000 kg ha-1) are comparable to arable mineral soils (390–2100 kg ha-
1 -1 ). However, Ca-bound non-labile Pi in arable organic soils (86–975 kg ha ) can be nearly twice the amount found in arable mineral soils (62–458 kg ha-1; Negassa and Leinweber, 2009).
19 Table 2.1: Mean inorganic P (Pi) and organic P (Po) concentrations in plant-available, Al/Fe-bound non-labile and Ca-bound non-labile pools of arable organic soils from Canada, Germany, Israel, and the United States. Crop type and P fertilizer inputs are provided.
Moderately Al/Fe-bound P Available P† Ca-bound Nonlabile P¶ Total P Study Crop Labile P‡ Nonlabile P§ Fertilizer Pi Po Pi Po Pi Po Pi Po kg ha–1 kg ha-1 Ontario, Canada Carrot 50 25 2.9 50 5.4 190 92 480 50 890 De Sena Carrot 25 37 4.5 100 7.6 460 130 980 62 1800 (2017) Carrot 80 49 3.2 61 3.7 150 73 780 54 1200 Audette et al. Arable 50†† 25§§ NA# 14¶¶ NA# 67 NA# 86††† NA# 450‡‡‡ (2018) Québec, Canada Parent et al. Vegetables NA# 70‡‡ NA# 44 59 71 96 NA# NA# 340 (1992) Parent et al. Arable NA# 240‡‡ NA# 150 48 71 180 NA# NA# 680 (2014) Florida, USA Castillo and Sugarcane 20–50†† 0.80 NA# NA# NA# 88 170 280††† NA# 540 Wright (2008) Saxony-Anhalt, Germany Schlichting et Arable 15.2 400‡‡ NA# 130 180 120 500 680## NA# 2000 al. (2002)
Hula Valley, Israel Liator et al. Arable 20–50 0.12 NA# 58 6.2 180 59 420 NA# 720 (2004) † Sequentially extracted with deionized water.
‡ Sequentially extracted with 0.5M NaHCO3 (pH 8.5).
§ Sequentially extracted with 0.1M NaOH solution.
20 ¶ Sequential extracted with 1M HCl solution.
# NA = not available.
†† Conventional fertilizer application rates for the region.
‡‡ Sequentially extracted with resin
§§ Sequentially extracted with 1M NH4Cl.
¶¶ Sequentially extracted with 0.11 M borate dithionite.
## Sequentially extracted with 1M H2SO4 solution.
††† Sequentially extracted with 0.5M HCl solution.
‡‡‡ Po was only listed as a total, and not its distribution in individual pools.
21 The plant-available Pi pool in organic soils expresses great variability ranging from as low as 0.12–0.80 kg P ha–1 (Liator et al., 2004; Castillo and Wright, 2008) to three orders of magnitude greater with 240–400 kg P ha-1 (Schlichting et al., 2002; Parent et al., 2014). This may stem from variables such as cultivation history (e.g. fertilization practices, intensive and long-term vs. extensive and short-term horticultural production); time of year when samples were collected (after fertilization or post-harvest); or inherent mineral content of the parent peat material (ombrogenous peatlands vs minerogenous peatlands) (Kolka et al., 2016). Studies in Finland found that the risk of leaching is highest in organic soils with little inorganic mineral compounds (Kaila, 1959;
Saarela et al., 2004). Certain organic soil studies may also have greater dissolved organic matter
(DOM) concentrations which could compete with P for mineral sorption sites, resulting in greater available P. By modeling P sorption to goethite, Weng et al. (2012) determined that P adsorption diminished by 37-97% with DOM present. In addition, DOM may increase the negative charge on soil surfaces, which repels P (Guppy et al., 2005). However, the role of DOM in P interactions is not clear, as DOM may provide low energy binding sites that keep P weakly sorbed (Johnston et al., 2009), and should be studied further, especially in organic soils.
Fractionation studies demonstrate the capacity of arable organic soils to retain P in mineral fractions, subsequently reducing their risk of P export. Castillo and Wright (2008) further exhibited the P retention of arable organic soils, determining that the Ca-bound non-labile pool grew by 8,
21 and 40% after a 21-day incubation of cultivated organic soils with 10, 50 and 150 kg-P ha–1, respectively. The availability of minerals in these soils can result from the subsidence of organic soils, which releases metals bound in humic complexes and oxidizes minerals to more reactive forms like Fe (III) (Liator et al., 2004; Zak et al., 2008). Tillage also influences the P retention capacity of arable organic soils by exposing mineral surfaces for P to bind, further exacerbating
22 soil subsidence, which brings the surface layer closer to the bedrock (Graham et al., 2005; Castillo and Wright, 2008). This mineral substrate heavily influences the mineralogy of the soil and is therefore a controlling factor in the fate of soil P. Most of the arable organic soil studies found in
Table 2.1 mention the presence of a Ca-rich bedrock which could explain the importance of the
Ca-bound non-labile Pi pool in these soils (Liator et al., 2004; Castillo and Wright, 2008; De Sena,
2017; Audette et al., 2018). However, if arable organic soils were cultivated on ombrogenous peatlands, mineral inputs from the bedrock are minimal due to their deep organic layer isolating the topsoil. Most organic soils are naturally acidic (pH ≤ 4). Therefore, most arable organic soils receive amendments of lime and gypsum to raise the pH which can be a significant source of Ca
(Liator et al., 2004; Negassa and Leinweber, 2009). The application of certain mineral fertilizers
(e.g. calcium monophosphate) can introduce Ca to these fields (Vu et al., 2010). While the studies presented in Table 2.1 either did not use calcium contained fertilizers or do not mention their fertilizer source, previous applications during cultivation history may have contained such fertilizers. While these arable organic soils demonstrate the ability to retain P in mineral fractions, those with greater available Pi may constitute a eutrophication risk. P present in the soil solution can be susceptible to leaching during a precipitation or irrigation event (Castillo and Wright, 2008).
Consequently, the geochemistry of abiotic P pools in organic soils is crucial for understanding the risk for organic soils in exporting P to water bodies.
2.3.2 Biotic soil P-pools
Microorganisms are an important pool of phosphorus in soils. Microbial reactions are the processes that predominately regulate the availability of P in soils (Cross and Schlesinger, 1995), contributing to the accessibility of plant-available P to crops and soil microorganisms. Studies
3- have found that the PO4 fluxes in organic soils are controlled by the soil microbial community
23 present, as they can immobilize P in their biomass or lower the redox potential due to their oxygen consumption, therefore releasing P bound to Fe (Noe et al., 2001). Microbial P has been found to relate directly to the microbial biomass (Annaheim et al., 2015). The microbial biomass consists to a large part of decomposed plant residues found in the soil (Richardson and Simpson, 2011).
The uptake, cycling and release of P by microorganisms in soils strongly influence its availability for plants in ecosystems. Microbial turnover of phosphorus, which can be defined as the sum of all microbial mediated transformations and related fluxes of phosphorus, is regulated by two factors: temporal fluctuations of microbial P and microbial activity (Wardle, 1998, Oberson and
Joner, 2005). Studies, which employed radioisotope techniques, indicate that microbial uptake of
P does not necessarily correspond to a net change in the microbial P-pool, suggesting equilibrium between P-uptake and P-release (Oberson et al., 2001; Oehl et al., 2001; Kouno et al., 2002). Thus, there can be a large P-flux through the microbial biomass without any net changes in microbial P.
Chen et al. (2003) calculated the turnover time of microbial phosphorus in forest and grassland soils using measurements of the temporal net changes in microbial phosphorus, whereas, Oehl et al. (2001) estimated the turnover time in a soil where no net change in microbial P occurred, by applying a radioisotope technique. The turnover times of the two studies showed a large variation, ranging from 70 days (Oehl et al., 2001) to 1.25 years (Chen et al., 2003). Hagerty et al. (2014) found that microbial C had a faster turnover rate in organic soils compared to mineral, however, to our knowledge, there is no information on microbial turnover times of P in organic soils.
The microbial biomass P was found to increase in the Everglades wherever there was high
P enrichment (Qualls and Richardson, 2000). A study by Ivanoff et al. (1998) found that the microbial biomass P in cultivated organic soil was 21% of TP. Noe et al. (2001) suggested that the microbial biomass P relationship to TP is more important when there is a lack of nutrients within
24 the soils. Other studies have found that P can limit microbial activities and growth in peat soils
-3 (Amador and Jones, 1993). Furthermore, these authors found that PO4 can stimulate the respiration rates in peat soils with intermediate TP content (385 mg P kg-1). Further research has found that adding nutrients to the soil changes the microbial biomass more readily in P-poor soils, but that even in P-poor soils, mineralization of organic P is driven more by microbial C than P
(Heuck et al., 2015).
2.3.3 Phosphorus forms in leached water
Phosphorus can be found in both organic and inorganic forms within water. The form of P most often studied is total P (TP), which includes all forms of P found in water without differentiation between the states of P. The P present in runoff regularly transitions between dissolved and particulate states and can be found in multiple forms, including dissolved orthophosphate (Spivakov et al., 1999). A major portion of P enters rivers and lakes in the form of particulate P, which can then dissolve through weathering and mineralization, either becoming total dissolved P (TDP) or binding with minerals. The particulate state of P includes both the reactive and organic suspended P particles (Spivakov et al., 1999). Dissolved reactive P (DRP), or
PO4-P, is readily available to aquatic biota, thereby causing degradation of flora and fauna in water bodies (Sharpley, 1993; Maguire and Sims, 2002; McDowell et al., 2001; Hoepting, 2009; Zheng et al., 2014).
2.3.4 Modeling P dynamics
Process-based models such as ICECREAM (Rekolainen and Posch, 1993; Larsson et al.,
2007; Qi et al., 2018), EPIC (Jones et al., 1984; Qi and Qi, 2016) and RZWQM2 (Qi and Qi.,
2016) can be used in P studies on mineral soils. However, transport models for P are not able to provide accurate P runoff estimations under diverse management scenarios, and have not been
25 designed for organic soils, limiting the use of the models (Sharpley et al., 2017). Mineral and organic soils have different soil properties, P dynamics, and therefore, researchers cannot draw inferences from mineral soils to study P dynamics in organic soils.
Process-based models require multiple specific soil physical-chemical parametric data points to create a robust simulation. Both deterministic and statistical models can be used to analyze water quality data and fill gaps in data. Statistical models include knowledge-based systems, genetic algorithms, artificial neural networks (ANNs), fuzzy inference system (Chau,
2006), evolutionary algorithms (Nicklow et al., 2010) and support vector machine models
(Raghavendra and Deka, 2014). Tiyasha et al. (2020) found through a review of 200 studies, for river water quality, from 2000 until 2020 that artificial intelligence models are the ‘perfect tool’ for water quality monitoring and management, even though there are still gaps in knowledge to be addressed. The ANN models can be trained to predict water quality using various hydrologic input parameters (Maier and Dandy, 1996; Sengorur et al., 2015; Chang et al., 2016) and have been used extensively in basin rainfall-runoff models (Hsu et al., 1995; Trafalis et al., 2002; Dahamsheh and
Aksoy, 2009) and in river water quality or flow models (Maier and Dandy, 1996; Bowden et al.,
2005; Tiyasha et al., 2020). However, there is a lack of studies using ANN for agricultural water quality monitoring.
2.4 Phosphorus concentrations and loading under tile drainage
Studies have found that P is of greater concern in fresh-water bodies, compared to N when looking at causes of eutrophication (Schindler and Fee 1974; Thomas et al., 1995). Factors affecting P concentrations and drainage include fertilizer application, crop management practices, irrigation practices, climate, soil properties and site conditions (Skaggs et al., 1994). Other studies
26 found that weather and soil properties have a larger impact on P loss due to drainage than beneficial management practices (BMPs) (Bergström et al., 2015; Kleinman et al., 2015). Studies, documented in Table 2.2, show the evidence of excess P loss from agricultural areas. Arable organic soils in New York were found to contribute approximately 55 to 86% of the P load that enters Lake Ontario from Oak Orchard Creek (Longabucco and Rafferty, 1989). Further studies of the Lake Simcoe watershed in Ontario found that organic soils contribute 1-5% of the TP load into the lake while only covering 1% of land within the watershed (Winter et al., 2007). The mandatory BMP program, Everglades Forever Act (Daroub et al., 2009), implemented since 1995, for the Everglades Agricultural Area have reduced the P load outflow by 50% from the agricultural area to the Lake Okeechobee (Daroub et al, 2011). Despite this successful reduction of P loads, studies within the Everglades have found that even small increase in P concentration in lakes and rivers can cause the immediate growth of algal blooms (McCormick and O’Dell 1996; McCormick and Stevenson 1998; Noe et al., 2001). However, despite the implementation of BMPs in the
Everglades, the P load entering Lake Okeechobee has been excessive to the point that the water body can no longer assimilate P (Havens and James, 2005).
Saarela et al. (2004) found that nearly 1 kg-P ha-1 was leached annually from arable fields in Finland with a weakly humified organic soil, while other studies have found up to 2.3% of total soil P can be leached (Martin et al., 1997). Excessive P leaching and P loads have been consistently linked to organic soils despite their large proportion of non-labile P (Table 2.1); this suggests over- fertilization and that soil-water fluxes may play a larger role than soil characteristics. Martin et al.
(1997) found that in drained organic soils there was a decrease in P concentrations as the water table rose closer to the soil surface. Furthermore, Thomas et al. (1995) found that P loss can be
27 reduced in organic soils when drainage water is retained within channels and field drainage infrastructure.
2.4.1 Climatic effects
Large areas of organic soils have been known to create a microclimate, which results in higher temperature amplitudes and air humidity (Ilnicki, 2003), and may affect P concentrations in subsurface drainage effluent and TP loads entering water bodies. A study by Lang et al. (2010) shows a positive correlation between precipitation, and the drainage volume at 10 different farms on organic soils. Further studies have found a positive relationship between rainfall and rise in water levels and drainage outflows, and between drainage and increased P in runoff (Nicholls and
MacCrimmon, 1974; Williams et al., 2015). This corroborates with Longabucco and Rafferty
(1989) whose study concludes that approximately half of the P loads from peat soils around Lake
Ontario are from runoff during the initial spring thaw and rainfall events during the months of
March, April and May. The discharge of water is a major source of P and therefore, nutrient release into the environment is linked to precipitation and freeze-thaw events in spring. A model using changes in temperature and precipitation within the Everglades predicts that a decrease in precipitation and an increase in temperature results in greater mineralization of Po forms (Orem et al., 2015), and thus would increase P release into water bodies.
28 Table 2.2: The P concentration and load from different studies in Canada, the United States of America and New Zealand
Location and Description TP / PO4-P concentration TP / PO4-P load Study PO -P concentration varies TP cultivated spring thaw load Ontario: 2 separate areas in the Holland 4 from 0.001 to 0.62 mg L-1 on was between 0.0036 to 1.51 kg ha- Nicholls and Marsh (one cultivated, one not), For spring uncultivated and from 0.003 1 while the uncultivated was from MacCrimmon, 1974 thaw load in 1971 (March to end April). and 0.59 mg L-1 for cultivated 0.0026 to 0.17 kg ha-1 New York: Elba Oak swamp, Grinnel PO -P concentration varies PO -P load varies annually Duxbury and farm. 3 locations were used from 1975 to 4 4 between 0.5 to 7.6 mg L-1 between 0.6 to 30.7 kg ha-1 Peverly, 1978 1977 Ontario: Organic soils south of Chatham in TP concentration varies PO -P load varies annually Kent County measured from 1971 to 1975 4 Miller, 1979 between 1.14 to 18.23 mg L-1 between 1.6 to 26.8 kg ha-1 at three field sites New York: organic soils in Western New Annual TP loads varies from 0.6 Cogger and York State using two sites to determine to 36 kg ha-1 Duxbury, 1984 factors effecting P loss New York: Elba Oak Creek with one year TP concentration was found Annual TP load for three sites (1984-1985) of sampling five stream to vary between 0.20 and 1.0 varied between 0.63 and 1.76 kg Longabucco and -1 -1 locations that account for the major mg L , with the mean being ha , while the PO4-P loads varied Rafferty, 1989 drainage of the organic soil farmland 0.72 mg L-1 between 0.34 and 1.37 kg ha-1 Florida: 4 field sites were used in the TP concentration ranged from Annual TP load for sugarcane was Everglades Agricultural Area looking at P 0.08 to 1.5 mg L-1 for 0.72 kg ha-1 and for radish was Izuno et al., 1991 release under different crop type from sugarcane and 0.06 to 0.62 0.88 kg ha-1 1988-1989 mg L-1 for radishes Florida: Two fields in the Everglades Nutrient Removal Project were used in a TP concentration averaged column study that looked at the P with four between 0.14 and 0.35 mg L-1 Martin et al., 1997 different water table levels for 30 days between the field sites each trial Ontario: Holland Marsh sub-watershed, TP concentration was Annual TP load varies between values from the discharge water at the between 0.25 – 0.83 mg L-1 1.5 kg ha-1 in 1990/91 to 0.39 kg Winter et al., 2007 Pump Station (1990-2004) ha-1 in 1994/95
29 Florida: 10 farms in Everglades Annual TP load varies between Agricultural Area from 1992-2002, Daroub et al., 2011 0.083 kg ha-1 to 0.56 kg ha-1 average monthly load New Zealand: Marginal lands on the Annual TP load on organic soils McDowell and south-island use of artificial drainage on was found to be 57.8 kg ha-1 Monaghan, 2015 dairy farms over an 18-month period
30 2.5 Water management for organic soil P
Reduction in P is critical to eutrophication management. Several studies have looked at various agronomic and water management practices aimed at reducing P losses from cultivated organic soils. This section focuses primarily on water management as a mitigation strategy and it is imperative that further research be conducted to assess potential BMPs for P reduction from intensively farmed organic soils (Carpenter, 2008; Schindler et al., 2008). Several studies
(Minasny and McBratney, 2006a, 2006b; Miles et al., 2013; McDowell and Monaghan, 2015) have identified significant gaps in our understanding of how P losses via surface runoff and tile drainage are influenced by hydroclimatic drivers and agronomic practices, and the need for improved
BMPs. In recent decades, mitigation measures have been implemented in agricultural watersheds, such as the Everglades (Daroub et al., 2011). Some organic soil regions have adopted these BMPs
(Table 2.3). Studies in the Florida Everglades have used DWM to maintain and control water levels in fields and control soil subsidence (Clayton and Jones, 1941; Stephens, 1955; Skaggs et al.,
2012). Since 1995, the Florida Everglades Agricultural Authority has a mandatory BMP program where DWM is an option through improved infrastructure (Daroub et al., 2011). However, although BMPs exist, the enforcement of their use and their effectiveness in the long-term are not always apparent.
31 Table 2.3: Mitigation measures for P used for organic soil agriculture
Practice Author Location Key Finding Rice as a cover crop has the added advantage Selected cover Jones et al., of reducing soil subsidence, destroying soil crops for P Everglades 1994 pests and pathogens and uptake of residual reduction. fertilizers. Slow drained sugarcane plots exhibited significantly higher TP concentrations than the Controlled Izuno et al., fast-drained plots. However, TP loads were Drainage in open Everglades 1995 significantly higher (0.97 kg ha-1) for fast channel farmlands drained plots than for the slow drained plots (0.67 kg ha-1). Thomas et al., presented several BMPs for mitigating P loss, including water table Review of Several North Carolina, management, control drainage, retaining Thomas et al., Mitigation South Carolina, drainage water from vegetable and sugarcane 1995 measures Georgia, Florida fields (on sugarcane or fallow areas). The review provided an amalgamation of both organic and mineral soils. Controls runoff by reducing the surface flows. This increases deposition and interaction between incoming nutrients and soil matrices, Buffer zones and and plant and microbial nutrient processes. Newbold et Southeastern constructed Implementation of buffer zones and al., 2010 Pennsylvania wetlands constructed wetlands led to particulate phosphorus concentration being lowered by 22%, but this removal was balanced by a 26% increase in soluble reactive phosphorus. Water P load reduction can be obtained by lowering management: Daroub et al., drainage volume and improving internal Everglades Detention of water 2011 drainage. Installation of culverts with riser in farmlands. boards and land levelling reduced outflows. Grassed peatland buffer zone reduced total reactive phosphorus (TRP) and suspended O'Driscoll et Glennamong, Buffer zones sediment (SS) loads by 18% and 33%, al., 2014 Ireland respectively, released from an upstream clear- felled blanket peat site.
2.5.1 Controlled drainage (CD)
Drainage water management (DWM) through the use of a controlled drainage (CD)
structure permits the regulation of the water table in fields. The use of CD in mineral soils has been
32 studied extensively (Elmi et al., 2002; Jamieson et al., 2003; Skaggs et al., 2010; Skaggs et al.,
2012; Williams et al., 2015; Carstensen et al., 2016; Schott et al., 2017; Youssef et al., 2018).
Multiple studies have shown that using CD can reduce drainage outflow and nutrient losses from farmlands (Williams et al., 2015; Lalonde et al., 1996; Schott et al., 2017; Wesstrom et al., 2001).
Consequently, farmers can drain their fields by removing gates such to improve field-machine trafficability, or to raise the gates during dry periods to maintain adequate soil-water in the root zone for crop growth (Mejia et al., 2000; Stämpfli and Madramootoo, 2006; Sanchez Valero et al.,
2007; Skaggs et al., 2012; Williams et al., 2015). Williams et al. (2015) reported that although nutrient loads are correlated to reduced discharge, the P concentrations are not affected by water management. Multiple studies on mineral soils showed precipitation and spring-thaw as drivers for discharge events (Jamieson et al., 2003; King et al., 2015; Gramlich et al., 2018).
2.5.2 Pump drainage (PD)
The high-water tables associated with organic soils, has led to the common use of pump drainage (PD) to discharge excess water from agricultural fields. Miller (1979) found that the
Holland Marsh growers used tile drainage that flowed into a collector well or a ditch which was then pumped into a municipal ditch connected to the river. Therefore, PD is the current DWM practice in multiple organic soil agricultural areas. In the Everglades Agricultural Area (EAA),
Bhadha et al. (2017) noted that drainage farms through PD was a common practice from June to
October. Furthermore, the timing of PD varied depending on storm events and needs. This signifies that the PD is only active when necessitated by the growers. Other studies using PD, setup the pump and leave it on continuously, allowing for constant activation when drainage water reaches a certain level in the sump (Poole et al., 2018).
33 2.5.3 Drainage through open channels
Open channels or ditches are used extensively in organic soil farmlands to provide drainage. A review conducted by Needelman et al. (2007) on improved management of agricultural ditches indicated several instances where ditches provided a sink for nutrients (N and P). Ditches play an important role in moderating downstream P losses. The most comprehensive information in this regard, on organic soils, is the work conducted in the Everglades, Florida, which utilized
DWM in open channels (Izuno et al., 1995). Contrary to what was expected, slowing drainage rates resulted in an increase in TP and total dissolved P loads rather than the expected reduction in
TP from particulate matter removal in the drainage channel. The high nutrient loads occurred following heavy rainfall, which caused the release of excess P from the system as the water flushed out the nutrients (Izuno et al., 1995). Capone et al. (1995), also stated that retaining water within the soil profile through DWM, instead of allowing complete drainage, would greatly aid in reducing nutrient loading in the canals and Lake Okeechobee.
2.6 Summary and recommendations
In comparison to mineral soil, few studies have been conducted on water and P transport from organic soils. Organic soils differ in characteristics from mineral soils. The soil porosity is different for organic soils with ‘mobile’ and ‘immobile’ regions in the soil matrix and four types of soil pores, leading to increased preferential flow, hydrophobicity and varying expansion and contraction parameters compared to mineral soils. Furthermore, organic soils can have higher Ca bound P and TP concentration in drainage water. The objective of this paper was to review subsurface P loss and to assess the effectiveness of DWM strategies to control these losses on organic soils. Based on the literature reviewed and our field measurements, the following
34 conclusions can be drawn: i) water movement, which is a key driver for P transport, is affected by several physical soil parameters which changed during wetting and drying, and bio-chemical degradation and humification of the soil organic matter; ii) Ca-bound Pi is the dominant P pool within organic soils, demonstrating their potential for P retention through the accumulation of fertilizer P; iv) organic soils have been linked to the excessive leaching of P, leading to large TP loads into the surrounding water bodies; and v) DWM is a potential mitigation measure for reducing P in surface waters and the environment. These conclusions led to the identification of the following gaps in knowledge for future study: i) longer-term studies are required on the effectiveness of DWM strategies as methods to reduce TP loads from cultivated organic soils; ii) effects of DWM strategies on movement of P within the soil water continuum and iii) assessment of potential mathematical models for nutrient flow within organic soils.
35 Connecting text
The literature review in Chapter II showed that phosphorus (P) is a concern for eutrophic algal blooms in freshwater bodies. Organic soils are often over-fertilized to counteract the inherent low nitrogen (N) and P found in these soils. There are multiple misconceptions when assessing the
P dynamics within these soils as there is a lack of studies focused on drainage water management practices, specifically using a controlled drainage (CD) structure. A CD structure can increase the water available to plants and decrease nutrient loads from agricultural fields. Understanding the function of CD as a mitigation management practice on organic soils can be beneficial for the management of P. Therefore, it is imperative to investigate the use of CD and its effects on nutrient concentration and loads to assess the potential of the use of this mitigation practice on organic soils. In this study we conducted a two-year (2015-2016) field study within the Holland Marsh, an important agricultural organic soil area in Ontario. Chapter III of this thesis investigated seasonal relationship between discharge, and N and P concentration and loads under a CD system.
Furthermore, an artificial neural network (ANN) was used to assess the predictability of seasonal loads under various water table scenarios.
This study is titled “Seasonal effects of controlled drainage on water quality from agricultural organic soils and the predictive use of artificial neural networks”. The author of the thesis was responsible for conceptualization, methodology, validation, formal analysis, investigation, data curation, and writing the original draft followed by all review and editing. Dr. Madramootoo provided supervision, aided in conceptualization, funding acquisition and the review and editing.
B. Singh aided in the investigation and with reviewing the paper, while Dr. Goyal provided software code and valuable advice through the review. Dr. von Sperber provided valuable advice through the review and editing of the manuscript. In order to ensure consistency with the thesis
36 format, the original draft has been modified, and the cited references are listed in the reference section. The funding for this project was provided by Dr. Chandra A. Madramootoo from the
Natural Sciences and Engineering Research Council of Canada (NSERC) Strategic Projects Grant
(447528 – 13).
37 3 . CHAPTER III
Seasonal effects of controlled drainage on water quality from agricultural organic soils and
the predictive use of artificial neural networks
3.1 Abstract
Organic soils are mainly used for the intensive production of high-value vegetable crops, which demand high fertilizer applications. This production system, however, contributes to eutrophication in surrounding water bodies, which necessitates control of non-point source pollution. The Holland Marsh is an organic soil vegetable farming area in Ontario, Canada, whose excess nutrients drain into Lake Simcoe, contributing to the eutrophication of the lake. In this study, controlled drainage (CD) was assessed to reduce nitrogen (N) and phosphorus (P) loads from tile drains. The research conducted from 2015 to 2016, examined the impacts of the CD on seasonal nutrient concentrations and loads. In 2015, there was no significant difference amongst the winter-spring and summer total P (TP) or total nitrogen (TN) loads. Whereas, in 2016, there was a significant difference between the winter-spring and summer loads for both nutrients. In both years, there were no nutrient loads in the fall due to CD. A nonlinear regression and an artificial neural network (ANN) model were used to develop seasonal relationships of nutrient loads based on inputs of fertilizer application, drain discharge and precipitation. The ANN performed better in all models compared to the nonlinear regression. The evaluation of the ANN model found that running individual models for both winter-spring and summer seasons accurately predicted the TP and TN loads. An investigation of different water table management scenarios, using ANN predictions, identified raising the CD water table to 34 cm or less from the soil surface in the winter-spring allowing for a reduction in nutrient loads through reduced water outflow. The
38 management scenarios for the summer season identified that the water table of 77 cm from the soil surface, could both reduce the water discharge and would satisfy the water depth for optimal crop production.
3.2 Introduction
The inflow of nutrients, specifically nitrogen (N) and phosphorus (P), in a freshwater ecosystem from its surrounding intensively cultivated agricultural lands, can cause eutrophication
(Elser et al., 2007; Schindler et al., 2012; Van Esbroeck et al., 2017; Krasa et al., 2019). Organic soils, also known as Histosols or muck soils, are intensively used for crop production. These soils require high fertilizer applications due to the inherent low nutrient status of these soils (Czuba and
Hutchinson, 1980; Porter and Sanchez, 1992; Kroetsch et al., 2011; Zheng et al., 2014). Excess fertilizer leaches into the environment, becoming a major contributor to excess N and P entering water bodies. Studies have found that a reduction of P in freshwater systems, more than N, is critical to eutrophication control (Carpenter, 2008; Schindler et al., 2008). Instances of excessive
P loads from cultivated organic soils have been observed in New York (Longabucco and Rafferty,
1989), Florida Everglades (McCormick and Stevenson 1998; Noe et al., 2001) and Ontario (Winter et al., 2007).
In Ontario, Canada, the Holland Marsh is a significant arable organic soil region that was partially converted from a wetland for agricultural use in 1925. The Holland Marsh contributes between 1 and 5% of the total P load entering Lake Simcoe and is therefore considered a source of anthropogenic pollution (Winter et al., 2007; Zheng et al., 2015). Due to the high water holding capacity and higher water tables encountered on organic soils, subsurface tile drains are installed to support crop production (Grozav and Rogobete, 2012; Hallema et al., 2015). Though there have
39 been studies on soil nutrients in the Holland Marsh (McDonald et al., 2013, 2014; Zheng et al.,
2014, 2015), research on drainage water quality is limited.
A controlled drainage (CD) structure has the potential to mitigate the release of nutrients to the environment by controlling the water table height and therefore the amount of water exiting the system. CD has been mostly implemented on mineral soils and resulted in reduced discharge, nutrient loads and in some cases nutrient concentrations in the drainage outflow (Elmi et al., 2000;
Mejia and Madramootoo, 1998; Jamieson et al., 2003; Williams et al., 2015a). However, the use of a CD structure has limited studies on agricultural organic soils, therefore their water quality benefits are unknown for these soils.
Field evaluation studies of water quality management practices involve the labour and cost- intensive process of acquiring water quality data. Both deterministic and statistical models can be used to analyze water quality data and fill gaps in data. Such modelling approaches can reduce the costs of water quality data collection. Artificial neural networks (ANNs) can be trained to predict water quality using various hydrologic input parameters (Maier and Dandy, 1996; Sengorur et al.,
2015; Chang et al., 2016). Unlike ANN, process-based models, such as ICECREAM (Rekolainen and Posch, 1993; Larsson et al., 2007; Qi et al., 2018), and EPIC (Jones et al., 1984; Qi and Qi,
2016) models, requires multiple specific ancillary parametric data points to create a robust simulation. Furthermore, these models have not been calibrated for organic soil properties. ANN is a method of data analysis that uses machine learning and analytical model building to understand the relationship between multiple inputs with the output variable, without detailed knowledge of the internal process-based functions (Lek et al., 1999; El-Din and Smith, 2002; Ha and Stenstrom,
2003). ANN has been used extensively to model basin rainfall-runoff processes (Hsu et al., 1995;
Trafalis et al., 2002; Dahamsheh and Aksoy, 2009), to predict water quality in river systems (Maier
40 and Dandy, 1996; Bowden et al., 2005), and has been dependently used to estimate dissolved P and ammonium-nitrogen in runoff from croplands (Kim and Gilley, 2008). Furthermore, the ANN model can reduce the need for intensive and long-term field sampling of water quality allowing for increased water management assessment.
Studies on Histosols agriculture tend to focus on the assessment of soil tests and their application on organic soils (Zheng at al., 2014, 2015) or entire watersheds (Longabucco and
Rafferty, 1989; Daroub et al., 2011; Winter et al., 2007). There are limited studies on the drainage water quality from agricultural organic soils or the application and benefits of CD from these soils.
The objective of the study was to assess the effects of CD on the seasonal variations of nutrient water quality and subsurface tile drain discharge, on cultivated organic soils. Furthermore, the study aims to evaluate the use of two mathematical models; ANN and a nonlinear regression model, to explore the relationship of fertilizer application, tile discharge and precipitation on seasonal and total annual nutrient loads. Finally, the study will use the models in a predictive approach to evaluate different CD water table scenario and their potential effects on water management.
3.3 Materials and Methods
3.3.1 Study area
The Holland Marsh is located within the Lake Simcoe Watershed, Ontario, and is surrounded by canals to divert excess water around the area to reduce flooding (Figure 3.1). This area consists of a temperate climate with more than five months (November to April) of the year under snow cover. The temperature of the area over 30 years, ranges from -12 to 11 °C without the wind chill during the cold months, while during the summer season (May to September) the
41 temperature ranges from 9 to 24 °C (Government of Canada, 2018). The study was conducted across two years (2015-2016) on a 4.2 ha field (44.064900, -79.587878) with subsurface tile drains.
The soil of the field is classified as a muck soil, which accounts for approximately 6% of the soil in Simcoe County and averages 1.5 meters in depth (Hoffman et al., 1962). The general soil properties for the study site were analyzed by AgriDirect using a composite soil sample taken post- harvest 2015. The soil properties consisted of an organic matter content of 68%, a bulk density of
0.31 g cm-3, a pH of 6.6, and a soil Mehlich P of 655 mg kg-1. Carrots (Daucus carota Bergen) were cultivated for the duration of the study. The mineral fertilizer application by the grower in
May 2015 consisted of 20 kg-P ha-1 and 210 kg Potassium (K) ha-1, while in May 2016 it consisted of 10 kg-N ha-1, 20 kg-P ha-1 and 205 kg-K ha-1. Seeding occurred in May, soon after the fertilization, and the crop harvest started in September both years.
Figure 3.1: Field research location from the Lake Simcoe Watershed and to the Holland Marsh field location.
42 3.3.2 Controlled drainage system
A water CD structure (Figure 3.2) was installed at the outlet of the collector tile line. This structure allowed for manual control of water table height in the field with the placement of removable gates that stack on top of each other to the desired height. The water level strategy adopted at the experimental site was to raise the water table during the winter-spring to reduce tile flow and to lower the water table during the summer growing season for optimal carrot production water level. Similar strategies were implemented in other studies as drainage reduction during the spring-thaw period is critical for nutrient load reduction (Adeuya et al., 2012; Gunn et al., 2015;
Skaggs et al., 2012). Therefore, during the colder months of November to April, the removable gates were placed between 30 and 40 cm below the soil surface to reduce drain flow during the non-growing season (Gunn et al., 2015). In May, the gates were removed to allow water to drain from the field for fertilizer application and seeding. Following seeding, the gates were placed at the height of 75 to 80 cm from the soil surface for the summer season, as the optimal water table depth for carrot crops is 75 – 90 cm from the soil surface (McDonald and Chaput, 2010). However, in 2015, the height of the gates varied from 70 cm to free drainage between mid-June and early
July to reduce the pooling of water caused by excess rain. The CD gates were raised post-harvest in early November 2015. In 2016, they were lowered mid-April to dry the field and then raised in
July at the request of the crop grower because of the lack of precipitation.
Discharge measurements were taken using a compound weir, which was placed as the uppermost gate. The compound weir consisted of 11° V-notch to estimate flows less than 1 L s-1 and a rectangular weir to calculate flows above 1 L s-1. The discharge was continuously recorded at 15-minute intervals using a pressure transducer within the control structure.
43
Figure 3.2: The illustration of a subsurface tile drain outlet with the installation of a controlled drainage
(CD) structure which allows for the manual adjustment of the water table height in a field with removable gates that stack on top of each other.
3.3.3 Water quality analysis
Water samples were collected during periods of drain discharge, using an ISCO 6712 portable auto-sampler at 4-hour intervals (Teledyne ISCO, Lincoln, Nebraska). Also, discrete grab samples were taken throughout the year. Samples were not taken during periods of the winter freeze, and during periods of low water elevation as a result of dry weather conditions. The water samples were stored in a 4°C refrigerator at the Guelph University Muck Crops Research Station
(MCRS) before being transported in coolers with ice to the McGill University labs for analysis.
3.3.3.1 Nutrient Concentrations
Water samples were taken throughout the year to allow for a seasonal analysis of nutrient concentrations, even during periods without discharge. Water analyses consisted of nitrate (NO3-
44 N), total N (TN), phosphate (PO4-P), and total P (TP), using the Lachat XYZ Sampler (Hach
Company, Loveland, Colorado). To start, 30 to 50 ml of each water sample were filtered using a
0.45 µm filter, and the remainder was left unfiltered. NO3-N and PO4-P were measured by analyzing the filtered water samples with colometry (Murphy and Riley, 1962; Lachat Instruments,
2003) . The unfiltered water samples were analyzed for TN and TP using the potassium persulfate digestion method (Dayton et al., 2017; Ebina et al., 1983). The analysis was conducted by combining 4 ml of the water sample with 4 ml of persulfate solution in a test tube and autoclaving the samples for one hour. For each analysis, both blank distilled water samples and standards of N and P were included for the calibration of the sample results. Results were acceptable when there was 5% error or less through the calibration of the Lachat using standards of N and P.
3.3.3.2 Nutrient Load
Calculation of N and P loads was done using NO3-N, TN, PO4-P, TP concentrations and continuous discharge. The load was calculated using linear interpolation, which studies have found to be one of the most common methods of load calculation with the least amount of uncertainty
(Tiemeyer et al., 2010; Williams et al., 2015b). The linear interpolation was used to estimate the hourly nutrient concentrations over short time series intervals. The load was calculated using the following equation (Williams et al., 2015b):
푛 푖푛푡 Load = 퐾 (∑푗=1 푄푗퐶푗 ) (3.1)
where K is the conversion factor to account for the change in units, 푄푗 is the hourly
푖푛푡 discharge and 퐶푗 is the hourly nutrient concentrations derived from the interpolation. Although the linear interpolation is found to be the standard method for annual load calculation, uncertainty and error can be found through interpolating values (Williams et al., 2015b).
45 3.3.4 Seasonality statistics
The year was divided into seasons as follows: i) January to April signifying the winter- spring season, ii) May to September representing the summer growing season, and iii) October to
December representing the fall season. The winter-spring was grouped together to represent the spring-thaw period, which can start anywhere from early February to late March as weather in
Ontario varies. Furthermore, with weather constantly changing, it would be a misrepresentation to separate these two ‘seasons’, combining them to present unified results allowed for better result assessment. Overall seasonal variability was assessed through an analysis of variance (ANOVA) for NO3-N, TN, PO4-P, and TP concentrations and loads, as well as the discharge. To identify trends within each year (winter-spring, summer, fall) and between years (2015 and 2016), a least square mean test (LSM) with Tukey post hoc adjustment for multiple comparisons was conducted.
The statistical significance of p < 0.05 was used throughout the statistical analysis.
3.3.5 Mathematical Modeling of nutrient loads
The mathematical models chosen for this study were a nonlinear regression and an ANN model. Two years of data were used in the model simulations to establish a relationship between fertilizer application and the hydrological parameters of discharge and precipitation as input parameters to the following output loads: NO3-N, TN, PO4-P, and TP. All available nutrient and hydrological parameters measured were assessed using an ad-hoc model-free approach (Maier et al., 2010) based on domain knowledge to select input parameters. The nutrient load from subsurface tile drainage only occurs during periods of discharge. Fertilizer application of N and P are yearly inputs, based on data provided by the crop grower. Although more parameters such as soil moisture or temperature, would have been optimal to use, a lack of continuous data prevented their use. The dataset was broken down into seasonal models for winter-spring (January to April)
46 and summer (May to September). In total the models were run with 241 data points during the winter-spring and 305 data points during the summer. Both the nonlinear regression and the ANN models used 70% of the data to train (calibrate) and 30% to test (validate) the models. As previously mentioned, there was no tile drain flow in the fall of both years.
3.3.5.1 Nonlinear regression
The nonlinear regression model was used to explore the integrated effects between precipitation and discharge, and N and P loads. The regression analysis is done using a quadratic function with several observation combinations between the independent variables of precipitation and discharge and the response variable of nutrient loads. The quadratic equation was defined as follows:
2 2 2 Nonlinear regression = 푎푋1 + 푏푋2 + 푐푋3 + 푑푋1푋2 + 푒푋1푋3 + 푓푋2푋3 + 휀 (3.2)
Where X1, X2 and X3 are the input parameters (fertilizer, precipitation and discharge) and a, b, c, d, e, f are constants generated in the model, with 휀 being the error term calculated within the model.
3.3.5.2 Artificial Neural Networks (ANNs)
An ANN model employs a structure closely related to neural pathways with powerful computing power to analyze nonlinear relationships (Singh et al., 2009). The ANN model is useful in situations where the characterization of parameters is difficult to explain using simple physical equations, which can be the case for P models (Antonopoulos et al., 2016; Radcliffe et al., 2015).
A feed-forward model allows the input signals to propagate through the network in a forward motion, from the input layer to the hidden inner layer, and finally to the output layer (Antonopoulos et al., 2016; Ajmera and Goyal, 2012). A three-layer feed-forward network is depicted in Figure
3.3. The input layer consists of the introduction of the input parameters to the model, as well it
47 computes the weighted sum of the input parameters. The middle layer, also known as the hidden layer, processes the data and the output layer that produces the results from the ANN model (Singh et al., 2009; Ajmera and Goyal, 2012). Each layer comprises the interconnectivity between nodes, determined-connection-weights, and the activation function of the model that allows for the brain- like function of the model (Hornik et al., 1989; Ajmera and Goyal, 2012; Kumar et al., 2013). The nodes are nonlinear functions that allow the input signals to be modified according to the weight and the interaction between other parameters within the model. The nodes have an internal activation function given by the input parameters, which sends signals to other nodes, creating a complex system of links and information processing to create the model output (Fausett, 1994;
Singh et al., 2009; Antonopoulos et al., 2016).
Figure 3.3: A typical ANN schematic with a three-layer feed-forward neural network.
Seventy percent of the dataset was randomly assigned to train (calibrate) the ANN model, while the remaining thirty percent of the data was used to test (validate) the model. The ANN model was run multiple times to account for model accuracy. The limited number of input parameters (three parameters) can affect the results, meaning there can be increased inaccuracy to the model caused by limited inputs. Among the training algorithms, backpropagation is the most
48 popular for training the ANN model (Zhang and Govindaraju, 2000). In this study and others
(Khalil et al., 2011; Ajmera and Goyal, 2012), the Levenberg–Marquardt algorithm was one of the main algorithms used for the overall efficiency of the ANN model training. The Levenberg-
Marquardt algorithm incorporates both a gradient descent method and the Gauss-Newton method which increases the optimization of the algorithm in solving nonlinear least-square problems and was found to perform better than other ANN training models (Ajmera and Goyal, 2012). Maier et al. (2010) found that the Levenberg-Marquardt algorithm has been used frequently when looking at water quality within river systems and can improve the computational efficiency of ANN calibration as a second-order calibration method. Other artificial intelligence models as well as differing training methods used by ANN models could affect the performance of the models (Maier et al., 2010; Tiyasha et al., 2020). However, the lack of comparable studies in agricultural areas allows for the unbiased assessment of our model.
3.3.5.3 Performance evaluation of models
The performance of the models was evaluated using the following statistical measures:
Nash-Sutcliffe efficiency (NSE), correlation coefficient (R), and root mean squared error (RMSE), defined as follows:
푛 2 ∑푖=1(푥−푦) Nash-Sutcliffe Efficiency (NSE) = 1 − 푛 ̅ 2 (3.3) ∑푖=1(푥−푋)
∑ 푋푌 Correlation Coefficient (R) = (3.4) √∑ 푋2 ∑ 푌2
∑푛 (푥−푦)2 Root Mean Squared Error (RMSE) = √ 푖=1 (3.5) 푛
Where n is the number of observations; x is the observed data; y is the predicted data; X = x - 푥̅, where 푥̅ is the mean of the observed data; and Y = y - 푦̅, where 푦̅ is the mean of the predicted data.
49 3.3.5.4 Predicting nutrient loads
Model training was done with the 2015 data, and the model performance was tested by predicting the 2016 nutrient loads, using the ANN model. The performance of the predictive results was assessed using the R function and by identifying the percentage of uncertainty. The uncertainty error was measured using the following equation by Williams et al. (2015b):
퐿표푎푑 −퐿표푎푑 푒(%) = 푝푟푒푑 × 100 (3.6) 퐿표푎푑
where e is the uncertainty, Loadpred as the predicted load found by the model and Load is the nutrient load calculated by the interpolation of the hourly dataset.
3.3.6 Water management scenario predictions using ANN
Following the investigation in the performance of the ANN summer and winter-spring models, the study assessed multiple water management scenarios. The water management scenarios consisted of varying the discharge released from the field through the CD structure by increasing or decreasing the water table height. Changing the discharge by altering the set water table, allows for various management scenarios to be considered. Therefore, modelling of multiple scenarios during the winter-spring and summer seasons allows for an increased understanding of the best organic soil water management strategy.
The model was calibrated using the actual data for summer 2015 and winter-spring 2016, while the prediction scenarios were done using the actual precipitation and fertilizer information along with the calculated discharge values at different set water table heights. In winter-spring
2016, the water table was lowered on April 18th to allow the field to dry for farming activities. This lowering of the water table was incorporated into the model to maintain accuracy. Additionally, in summer 2015, the gate depth varied from 65 to 120 cm, with nine days of free drainage to
50 accommodate agronomic needs. The model scenarios maintained a constant water table height while also accounting for the days of free drainage.
The model was run multiple times with an NSE value of > 0.80 during the testing and training of the model. The load predictions were made only for TN and TP loads, representing the largest portion of the nutrients in the water quality. Furthermore, the ANN models for these TN and TP loads had less uncertainty in comparison to the calculated loads. Each model was run three times to account for model variance, with the NSE set at 85% or greater when running the models.
3.4 Results and discussion
The results are presented in two parts: the first consists of understanding the impacts of CD on the seasonality of tile drainage discharge, as well as nutrient concentrations and loads during the two study years. Secondly, the relationship between the fertilizer application and hydrological parameters of precipitation and discharge, with the N and P loads was assessed with the nonlinear regression and ANN models. The resulting models with better overall performance were used to predict the 2016 seasonal loading of N and P, using 2015 data to train the model. Following the
2016 predictions, management scenarios with varying seasonal water discharge were considered.
3.4.1 Weather conditions
The monthly precipitation for 2015, 2016 and a 30-year average is presented in Table 3.1.
The annual precipitation was 675 and 630 mm for 2015 and 2016, respectively, compared to the
30-year average of 774 mm, suggesting both years were drier than average. Furthermore, the precipitation during the winter-spring season was 29% less than the 30-year average in 2015, with the summer season (368 mm) being slightly lower than the long-term average of 371 mm. In 2016, the winter-spring season was nearly equal to the average. However, the summer season (234 mm)
51 was 36% below the long-term average. Intense precipitation occurred in March 2016, where there was 80 mm of precipitation, compared to the 30-year average of 46 mm, an increase of 72%.
3.4.2 Seasonality of tile drain discharge
Table 3.2 presents the seasonal discharge separated by season (winter-spring, summer and fall). The periods with higher discharge (summer 2015; winter-spring 2016) correspond to high monthly precipitation. The largest daily discharge of 37.59 mm day-1 occurred on June 28th, 2015 with 1.6*106 L of drainage water released from the CD structure. The precipitation for June was
160 mm, leading to the lowering of the CD gates to lower the water table, thus reducing the waterlogging of the crop root zone.
Table 3.1: Monthly precipitation from the Pearson Toronto Airport weather station for 2015, 2016 and a
30-year average along with the annual total and the totals for the spring and summer seasons (Government of Canada, 2018).
30-Year monthly 2015 (mm) 2016 (mm) average (mm) January 54.3 31.4 38.4 February 47.3 31.2 45.6 March 46.5 14.3 80 April 72.6 78.8 59.8 May 74.9 62.8 34.2 June 75.3 160.2 26.4 July 79.8 24.4 39.8 August 67.5 61.6 66.8 September 70.4 62 66.4 October 63.3 67.6 40.6 November 67.1 35.4 55.2 December 55.7 45.6 77.4 Total 775 675 631 Jan-April 221 156 224 May - Sept 368 371 234
52 Table 3.2: Descriptive statistics for discharge, TP, PO4-P, NO3-N and TN concentration data for 2015 and
2016 separated into three seasons: spring (January to April), summer (May to September) and fall (October to December).
Type Variables Year Season Mean Median St. Dev. Min Max Jan-Apr 0.90 N/A 1.53 0 6.24 2015 May-Sep 1.79 N/A 5.02 0 37.6 Discharge Oct-Dec 0 N/A 0 0 0 Water (mm day-1) Jan-Apr 3.13 N/A 5.38 0 23.2 2016 May-Sep 1.22 N/A 3.15 0 14.1 Oct-Dec 0 N/A 0 0 0 Jan-Apr 0.11 0.084 0.12 0.02 1.11 2015 May-Sep 0.52 0.16 0.63 0.07 2.25 Total P Oct-Dec 0.53 0.29 0.65 0.17 2.29 (mg L-1) Jan-Apr 0.086 0.077 0.09 0.03 0.83 2016 May-Sep 0.58 0.23 0.61 0.1 1.9 Oct-Dec 0.46 0.56 0.28 0.09 0.77 Jan-Apr 0.056 0.022 0.11 0.01 1.05 2015 May-Sep 0.11 0.068 0.11 0.02 0.5 PO4-P Oct-Dec 0.16 0.15 0.07 0.02 0.29 (mg L-1) Jan-Apr 0.059 0.058 0.01 0.03 0.09 2016 May-Sep 0.11 0.088 0.06 0.02 0.28 Oct-Dec 0.29 0.36 0.16 0.07 0.49 Concentrations Jan-Apr 6.34 3.75 4.97 0.92 19.34 2015 May-Sep 6.43 4.37 4.09 2.70 16.95 Total N Oct-Dec 3.81 4.08 0.99 1.72 5.09 (mg L-1) Jan-Apr 15.58 16.69 4.99 3.19 23.98 2016 May-Sep 5.90 5.83 1.60 2.94 8.79 Oct-Dec 3.12 3.86 1.38 1.40 4.70 Jan-Apr 4.21 1.69 4.39 0.03 17.05 2015 May-Sep 4.23 2.81 3.80 0.46 13.96 NO3-N Oct-Dec 1.24 0.93 0.88 0.38 3.03 (mg L-1) Jan-Apr 13.16 14.39 4.77 0.85 22.22 2016 May-Sep 3.58 4.02 1.72 0.35 6.62 Oct-Dec 1.63 1.86 0.82 0.55 2.56
53 Studies have shown that some of the most significant discharge events occur during spring thaw (Jamieson et al., 2003; King et al., 2014; Lam et al., 2016). This period of high discharge was corroborated by the 2016 results, where 67% of the tile discharge occurred in the winter-spring due to the freeze-thaw, as well as the increased precipitation in March 2016 (avg: 46 mm; 2016:
80 mm). There was no discharge during the fall of either year. It can thus be seen that under CD, discharge is limited to between February and July for both 2015 and 2016. By raising the height of the gates between 30 and 40 cm from the soil surface for the colder periods of the year, CD reduced drainage discharge, thus minimizing eutrophication impacts to the Holland River.
3.4.3 Seasonal effects of nutrient concentrations
The annual and seasonal variations in nutrient concentrations are shown in Table 3.2 for all seasons in 2015 and 2016. Although there was no water discharge from the CD structure during the fall season, the discrete manual grab samples were allowed from fall nutrient concentration assessment. Figure 3.4 shows a time series representation of 2015-2016 TP (3.4a) and the TN(3.4b) concentrations along with daily precipitation and discharge (mm day-1). The TP and TN concentrations were shown to represent the total nutrient released from the subsurface tile drains.
Figure 3.4b shows that the TN concentrations increase during the spring thaw and decrease in the summer season for both 2015 and 2016. According to previous studies (Gilliam et al., 1979; Mejia and Madramootoo, 1998; Elmi et al., 2000; Skaggs et al., 2012), the increased water saturation in the soil during the winter months followed by the winter-spring discharge, increases the biogeochemical N processes. Therefore, the increase in N concentrations (Table 3.2) during the winter-spring period identifies water movement as a leading factor for N movement through the soil and into the drainage water. Although other studies have found N fertilizer application to be a factor to N loss in the leachate (Petrovic, 1990; Easton and Petrovic, 2004); the lack of N fertilizer
54 application in the study, eliminated this factor from consideration. The higher concentrations of
NO3-N were observed during the winter-spring thaw events of both years. However, 87% of the
-1 NO3-N concentrations were less than the recommended limit of 13 mg L (Chambers et al., 2012;
CCME, 2012). Therefore, although there is an increased release of N concentrations during the winter-spring, the concentrations are usually within the recommended limit and not a major environmental concern.
The average TP concentration was lower during the winter-spring thaw, compared to other periods of the year. These lower TP concentrations were caused by higher volumes of water discharge diluting the TP concentrations during the spring. The results further showed that only
9% of the TP concentrations are below or equal to the Provincial Water Quality Objectives of 0.03 mg L-1 (Chambers et al., 2012; MOE, 1994). Therefore, P concentrations within the drainage discharge are a cause for concern. In 2015, the highest concentration of TP was 2.29 mg L-1, observed in early October. While, in 2016, the maximum TP was of 1.90 mg L-1, observed in the summer, mid-July. The 20 kg-P ha-1 fertilizer application in May of 2015 and 2016 was a factor in the increase of TP and PO4-P concentrations through the summer and into the fall seasons, as seen in Figure 3.4a. Easton and Petrovic (2004) found that P concentrations increased immediately following fertilizer application. TP concentrations increased gradually through the summer and early fall seasons before declining towards the end of both years. This increase in concentrations coincides with fertilizer application in both years.
55 40 2.5 a.
) 35
y
)
a
d
L /
2 /
g
m m
m 30
(
(
e n
g o
i
r
t
a
a h
25 r t
c 1.5
n
s
i
e
c
D
n d
20 o
n
C
a
s
)
u
r m
1 o m
15 h
(
p
s
n
o
o
h
i
t
P
a
t
10 l
i
a
p
t i
0.5 o
c
e T r
P 5
0 0
5 5 5 5 6 6 6 6
5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6
1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
------
------
r l t r l t
r r
n g v c n g v c
n b y p n b y p
c c
a p u a p u
e e
a e a u u e o a e a u u e o
J J
J J J J
O O
F A S F A S
D D
M A N M A N
M M Precipitation (mm) Discharge (mm/day) Total P concentration (mg/L)
40 25 b.
) 35
y
a
d
) /
20 L
/
m g
m 30
(
m
(
e
g
n
r
o
a
i t
h 25
a c
15 r
s t
i
n
e
D
c d
20 n
n
o
a
C
)
n m
10 e g
m 15
(
o
r
t
n i
o
i N
t
l
a t
10 a
i
t
p o
i 5
T
c
e r
P 5
0 0
5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
------
l t l t
r r
r r
c c
n b y n g p v n b y n g p v
c c
a u a u
p e p e
a e a e a e a e
u u o u u o
J J
J J
J O J O
F S F S
A D A D
M M
A N A N
M M Precipitation (mm) Discharge (mm/day) Total N concentration (mg/L)
Figure 3.4: A 2015 and 2016 time series analysis of the total phosphorus (a) and total nitrogen (b) concentrations, represented by black triangles, along with the daily precipitation (black bars) and subsurface tile drain discharge (grey bar) from the field. In (a) an increasing TP concentration trend can be seen starting midsummer (June/July) each year, with a decrease occurring in the late fall season (December).
Furthermore, in (b) an increase in TN concentration can be seen between February and March, corresponding with the winter-spring thaw.
56 The ANOVA analysis (Table 3.3) showed a significant difference (p < 0.05) between years, as well as between seasons, for all nutrient concentrations. The TN and NO3-N concentrations had a significant difference between the winter-spring values of 2015 and 2016. This difference is evident from the increase in concentrations observed from 2015 to 2016, as shown in Table 3.2.
This increase in concentrations between the winter-spring periods can be linked to the intense and continual tile drain discharge from February until June 2016. Furthermore, there was an increase in variability in 2016 between the seasons, compared to 2015, where no significant variance was found between seasons. Studies report that water management effects are more correlated to the reduction in discharge. However, additional factors such as fertilizer application, crop management practices, climate, soil, and site conditions, affect nutrient concentrations (Skaggs et al., 1994; Williams et al., 2015a). Similarly, our study found that although the N concentrations have a relationship with tile discharge, other agronomic factors including past fertilizer application and N present in the soil, also had an impact on the seasonal concentration variations.
TP concentrations were found to be significantly different between all seasons and between both years (Table 3.3), with no visible seasonal trends. PO4-P was found to have no significant differences when comparing 2015 and 2016 in the winter-spring, summer and fall. However, a similar trend was found within each year such that the winter-spring PO4-P concentrations were diluted by the continuous discharge; concentrations increased with fertilizer application in the early summer, and finally decreased in the fall, after harvest. The similarity within each year identifies the PO4-P concentrations as a proper valuation of seasonality trends.
57 Table 3.3: Statistical representation of general linear model results identifying seasonal and yearly nutrient
concentration trends. Furthermore, the table identifies the periods in the year (spring, summer and fall)
where similarity and differences between the values exist.
Interactions TP PO4-P TN NO3-N F value 2531.89 4.13 133.92 152.59 Year Pr > F < 0.0001* 0.0424* < 0.0001* < 0.0001* F value 75.42 14.26 28.38 28.97 Seasonal Pr > F < 0.0001* < 0.0001* < 0.0001* < 0.0001* Least Squares Mean Pr > t Winter-Spring - Summer 2015 < 0.0001* 0.0362* 0.0007* 0.0828 Winter-Spring - Fall 2015 < 0.0001* 0.3427 0.9985 0.7851 Summer - Fall 2015 < 0.0001* 0.0127* 0.2786 0.1392 Winter-Spring 2015 – Winter-Spring 2016 < 0.0001* 0.3962 < 0.0001* < 0.0001* Summer 2015 - Summer 2016 < 0.0001* 0.1805 0.9435 0.9386 Fall 2015 - Fall 2016 < 0.0001* 0.0652 0.9996 0.7422 Winter-Spring - Summer 2016 < 0.0001* 0.3995 < 0.0001* < 0.0001* Winter-Spring - Fall 2016 < 0.0001* < 0.0001* < 0.0001* < 0.0001* Summer - Fall 2016 0.0003* 0.0362* 0.7369 0.0828 * Values where there is a significant difference of < 0.05.
3.4.4 Seasonal effects of P and N load
Nutrient concentrations were used with continuous discharge values to calculate the daily
values of N and P loads, as seen in Table 3.4. The CD gates were raised in the fall (October –
December) season, following harvest, preventing tile drainage discharge from occurring (Table
3.2), which led to no nutrient loads during the fall seasons. In 2015, between 60 and 80% of the
-1 -1 -1 -1 nutrient loads (TN 30 kg ha ; NO3-N 25 kg ha ; TP 0.27 kg ha ; PO4-P 0.21 kg ha ) occurred
during the summer. These loads were mainly caused, in the case of P, by fertilizer inputs, and for
both N and P by the rainfall in June 2015 (160 mm), which was 112% higher than the 30-year
average (75 mm). The high precipitation resulted in a cumulative discharge of 8.4*106 L in June
58 2015. Tiemeyer and Kahle (2014) found peatlands in Germany to be a source of diffuse N pollution with the loads occurring during the wettest part of the year.
Table 3.4: Load analysis for 2015 and 2016 separated into three seasons: spring (January to April) and the summer (May to September) seasons, in addition to annually.
Time period Total P PO4-P Total N NO3-N Year (Seasonal) (kg ha-1) (kg ha-1) (kg ha-1) (kg ha-1) Jan - April 0.18 0.09 8.76 6.17 May - Sept 0.27 0.21 29.99 25.09 2015 Oct - Dec 0.0 0.0 0.0 0.0 Annual 0.45 0.30 38.75 31.26 Jan - April 0.33 0.24 45.89 38.80 May - Sept 0.17 0.12 14.35 11.61 2016 Oct - Dec 0.0 0.0 0.0 0.0 Annual 0.50 0.36 60.24 50.40
PO4-P is a form of P that is known as dissolved reactive P, which is immediately bioavailable to aquatic plants and, therefore, of great concern to environmental mitigation practices
-1 (Tan and Zhang, 2011). The PO4-P load in 2015 accounts for 67% (0.30 kg ha ) of the annual TP load discharged into the waterways. Consequently, more than half of the P released from the agricultural field can be immediately used by the surrounding aquatic environment to impact the ecosystem negatively.
In 2016, the majority of these loads were discharged during the winter-spring season (TN
-1 -1 -1 -1 46 kg ha ; NO3-N 39 kg ha ; TP 0.33 kg ha ; PO4-P 0.24 kg ha ), correlating to the precipitation of 80 mm in March. Higher NO3-N loads were also found during the wetter seasons by other studies (Lee et al., 2020). The annual TP load accounts for approximately 3% of the 20 kg-P ha-1 applied fertilizer for both 2015 and 2016. Therefore, only a small portion of the applied P is released through tile drain discharge (Sanchez Valero et al., 2007), large amounts of P was sequestered within the soil or taken up by crops, as found in other studies (Sims et al., 1998; Easton
59 and Petrovic, 2004). According to the Ontario Ministry of Agriculture, Food and Rural Affairs
(OMAFRA) soil fertility guide (Munroe et al., 2018), about 242.5 mg-P kg-1 is taken up by the
crop. The remaining P becomes unavailable to the plant, binding with aluminum, iron and other
minerals (Sims et al., 1998).
Statistical analysis (Table 3.5) of the TN and TP loads shows significant differences across
seasons, but no significant differences between the two years. Significance across both season and
year was only found in NO3-N loads. The significant difference between the winter-spring and
summer loads of 2016 was linked to the discharge pattern, with the CD gates being lowered in
April, thereby prolonging the winter-spring discharge. The raising of the gates in July limited the
nutrient loads.
Table 3.5: Statistical representation of SAS results in identifying seasonal and yearly trends for TP and TN
loads as well as discharge. The seasons are represented by spring and summer only as there was no nutrient
load during the fall season.
Interactions between Seasons TP load PO4-P load TN load NO3-N load Discharge F value 0.23 0.14 3.62 4.76 3.64 Year Pr > F 0.6297 0.707 0.0576 0.0295* 0.0571 F value 7.56 12.41 15.65 21.22 10.16 Seasonal Pr > F 0.0006* < 0.0001* < 0.0001* < 0.0001* < 0.0001* Least Squares Mean (LSM) Pr > t Winter-Spring - Summer 2015 0.1004 0.9048 0.1537 0.0384* 0.2535 Winter-Spring 2015 – 0.0521 0.0441* < 0.0001* < 0.0001* < 0.0001* Winter-Spring 2016 Summer 2015 - Summer 2016 0.0202* 0.3216 0.1999 0.1109 0.5321 Winter-Spring - Summer 2016 0.0097* 0.0012* < 0.0001* < 0.0001* 0.0003* * Values where there is a significant difference of < 0.05.
60 3.4.5 Performance comparison of ANN and nonlinear regression
The descriptive statistics of the input and output values for discharge, precipitation, fertilizer and nutrient loads are summarized in Table 3.6. The ANN and nonlinear regression model performance (Table 3.7) found that the summer models had a higher correlation (R) for training from the nonlinear regression models for both N and P (TN: 0.96, NO3-N: 0.95, TP: 0.91, PO4-P:
0.88), compared to the winter-spring models (TN: 0.80, NO3-N: 0.79, TP: 0.84, PO4-P: 0.81).
Furthermore, the ANN model for TP had a lower Nash-Sutcliffe efficiency (NSE) in the winter- spring training (0.92) and testing (0.81) compared to the summer model (NSE: training 1.00; testing: 0.98). Similar results were found for the other nutrients, with higher accuracy for the summer models compared to the winter-spring models. As fertilizer application only occurs in the summer season, the results are aligned with past studies, which have found that the increase in input parameters increases the accuracy of the predicted output (Palani et al., 2008; Singh et al.,
2009; Ajmera and Goyal, 2012; Keshavarzi et al., 2015). The limited number of available annual input parameters could have further restricted the performance of the models. The assessment of other parameters would be needed for further study as Maier et al. (2010) found through a review of past ANN water quality studies in river systems that input selection can have a significant impact on model performance. As all model inputs within this study affected either the discharge or nutrient concentration which are the factors that make up the nutrient load, these parameters were found to be significant to the model. However, further parameters to consider include soil moisture, soil temperature, and irrigation inputs.
61 Table 3.6: Summary of descriptive statistics of water parameters and nutrient load.
P N Total precipitation Discharge Daily TP load Daily PO4-P Daily TN load Daily NO3-N fertilizer fertilizer (mm) (mm day-1) (g day-1) load (g day-1) (g day-1) load (g day-1) (kg ha-1) (kg ha-1) Winter-Spring Count 241 241 241 241 241 241 241 241 Mean 0 0 1.57 2.02 8.92 5.78 952.42 783.6 Standard 0 0 3.71 4.11 17.47 11.66 1794.96 1483.8 Deviation Minimum 0 0 0 0 0 0 0 0 Maximum 0 0 24 23.20 87.69 59.22 8642.92 6997.4 Summer Count 305 305 305 305 305 305 305 305 Mean 0.13 0.03 1.98 1.51 5.98 4.45 610.51 503.67 Standard 1.62 0.52 5.38 3.84 16.30 11.47 2049.51 1643.7 Deviation Minimum 0 0 0 0 0 0 0 0 Maximum 20 9 35.80 37.59 182.56 120.4 26285.96 20392
62 Table 3.7: Performance evaluation criteria parameters for spring and summer models of ANN and nonlinear regression.
ANN Nonlinear regression
Model NSE R RMSE (g day-1) NSE R RMSE (g day-1) Training 0.92 0.96 4.81 0.65 0.84 10.19 TP Testing 0.81 0.93 6.86 0.59 0.83 6.55 Training 0.93 0.97 3.08 0.60 0.81 7.49 PO4-P Testing 0.96 0.98 2.20 0.79 0.91 2.91 Winter-Spring Training 0.96 0.98 336.74 0.58 0.80 1184.72 TN Testing 0.92 0.96 491.64 0.62 0.83 602.75 Training 0.93 0.97 386.27 0.56 0.79 986.12 NO3-N Testing 0.87 0.93 551.73 0.56 0.80 552.16 Training 1.00 1.00 0.28 0.81 0.91 7.27 TP Testing 0.97 0.98 2.68 0.44 0.78 7.23 Training 1.00 1.00 0.78 0.75 0.88 6.10 PO4-P Testing 0.98 0.99 1.25 0.52 0.81 4.32 Summer Training 0.98 0.99 157.77 0.93 0.96 337.46 TN Testing 0.94 0.99 912.39 0.49 0.71 1479.06 Training 1.00 1.00 79.59 0.90 0.95 577.34 NO3-N Testing 0.86 0.93 505.32 0.47 0.70 576.54
63 The ANN models outperformed the nonlinear regression models with a higher NSE and R, and lower root mean squared error (RMSE) values. The RMSE for TN ranged from 158 to 492 g day-1 for the ANN model, while for nonlinear regression, they ranged from 337 to 1479 g day-1.
Improved model performance can also be observed with the ANN model via the TP NSE and R and the TN NSE and R values. Figure 3.5 shows the fit of the ANN and nonlinear regression models between the training (Figure 3.5a and 3.5c) and testing (Figure 3.5b and 3.5d) measured and predicted TP winter-spring loads. Figure 5 illustrates a better R and NSE in the training phase with 0.96 and 0.92, respectively (Figure 3.5a) for the ANN, compared to the nonlinear regression model (0.84 and 0.65; Figure 3.5c). Similar results were found for the testing (Figure 3.5b and
3.5d) with ANN having a better performance and smaller error values than the nonlinear model
(Hornik et al., 1989; Liu et al., 2009). Furthermore, when looking at water quality predictive models in recent years, there has been an increased use of ANN models with high implementation and predictive rates (Tiyasha et al., 2020).
3.4.6 Load prediction performance using ANN
The results in Table 3.8 show the performance of the winter-spring (January to April) and summer (May to September) ANN models utilizing 2015 data for training, with the 2016 data used as an unknown allowing for predictive comparisons between the model results and the calculated nutrient loads. It can be seen that both the winter-spring and summer models perform well with high R values in training and testing. However, there is a higher accuracy found in the summer training NSE values, which signifies a better fit for the model. The higher accuracy of the summer model can be equated to the input of the fertilizer application data. Furthermore, the summer 2015 dataset has more variability of measured data points than the 2016 summer, allowing for a better calibration of the model.
64
Figure 3.5: Comparison of the predicted and measured TP loads for ANN (a/b) and nonlinear regression (c/d) models. The best fit linear trend is shown for the ANN winter-spring training (a) and testing (b) models and the nonlinear regression winter-spring training (c) and testing (d) models.
A higher correlation can be seen by the fit of the linear trend line in the ANN models, compared to the nonlinear models.
65 Table 3.8: Performance of prediction evaluation for TP and TN ANN seasonal models in the winter-spring (January to April) and the summer (May to September) seasons using 2015 for training and predicting 2016.
ANN Load comparison
RMSE Predicted Actual load Model NSE R R e% (g/day) load (kg ha-1) (kg ha-1)
Training 0.88 0.95 3.52 Total P 0.39 0.33 0.56 16.31 Testing 0.89 0.94 5.07 Training 0.90 0.95 2.58 PO4-P 0.17 0.24 0.73 -31.44 Testing 0.93 0.99 1.28 Winter-Spring Training 0.99 0.99 77.10 Total N 40.82 45.89 0.90 -11.05 Testing 0.86 0.95 199.94 Training 0.99 0.99 50.74 NO3-N 24.40 38.80 0.60 -37.11 Testing 0.92 0.97 151.92 Training 1.00 1.00 0.31 Total P 0.16 0.17 1.00 -1.16 Testing 1.00 1.00 0.47 Training 1.00 1.00 0.22 PO4-P 0.16 0.12 0.91 33.55 Testing 0.99 1.00 1.29 Summer Training 1.00 1.00 52.02 Total N 15.66 14.35 0.97 9.18 Testing 0.99 1.00 426.45 Training 1.00 1.00 15.23 NO3-N 13.57 11.61 0.98 16.92 Testing 0.97 0.99 326.0
66 The ANN winter-spring TN model predicted a load of 40.82 kg ha-1, which had an uncertainty of approximately 11%, and the TP model predicted 0.39 kg ha-1 with 16% uncertainty.
There were higher uncertainties found in the resulting loads from the NO3-N (37.11% uncertainty) and the PO4-P (31.44% uncertainty) models. The higher uncertainty percentages were found in the winter-spring models signifying that the training dataset was insufficient for predicting a high accuracy for 2016 winter-spring nutrient loads. The 2015 winter-spring dataset had less seasonal variation and was not as robust as the 2016 discharge data. The ANN models need large data sets for training to improve the accuracy of the model (Tiyasha et al., 2020). Therefore, if more data were available to train and test the model, perhaps the predictive loads would have higher accuracy.
This was seen in the summer models where greater predictive accuracy was found for both the TP
(0.16 kg ha-1) and TN (15.66 kg ha-1) loads. However, the uncertainty was still high for both the
PO4-P (33.55%) and NO3-N (16.92%) loads. Nevertheless, the performance of the TP and TN summer models was very high, with R values equal to 1 and comparable NSE results of 0.99 to
1.00. Although ANN has been used to predict P loads (Sengorur et al., 2015; Chang et al., 2016), as well as N concentrations and loads (Chen et al., 2009; He et al., 2011; Al-Mahallawi et al.,
2012) in groundwater, streams and rivers. The use of ANN in predicting nutrient loading from agriculture is limited.
3.4.7 Water table management predictions using ANN
The ANN model was able to predict known 2016 nutrient loads with more accuracy for
TN and TP loads and therefore, can be used under various water table management scenarios. The winter-spring season 2016 had precipitation (224 mm) closer to the 30-year average (221 mm), allowing for scenario results to reflect an average rainfall season. The water table scenarios varied between 20 to 50 cm to represent the benefits in nutrient loads if raising or lowering the height of
67 the water table during the winter-spring season 2016. The ANN model predicted (Table 3.9) a large difference between the actual loads at 34 cm (TN 46 kg ha-1; TP 0.33 kg ha-1) and the loads at 25 cm (TN 31 kg ha-1; TP 0.23 kg ha-1), however, no changes occurred between 34 and 30 cm water table. These changes show that the higher the water table is set in the winter-spring season, the fewer nutrients are released into the surrounding water bodies. Furthermore, the predictions at
40 cm (TN 53 kg ha-1; TP 0.51 kg ha-1) allow increased water discharge levels and significant nutrient loads, which would negatively affect the environment. Multiple studies have found that the main nutrient discharge occurs during spring thaw (Nicholls and MacCrimmon, 1974;
Longabucco and Rafferty, 1989; Williams et al., 2015). Therefore, the management of water levels before the growing season is crucial. Maintaining a higher water level within the field can reduce the amount of nutrient loads draining from agricultural fields. Overall, the ANN predictions found that a water table set to 34 cm or above from the soil surface, has the potential to reduce the TN and TP loads through the reduction of drainage discharge.
68 Table 3.9: Predictive water quality loads through CD management of water table from multiple scenarios in both the Winter-Spring and Summer seasons using the ANN models.
Winter – Spring Summer Load (kg/ha) Standard deviation Load (kg/ha) Standard deviation 20 CM 0.174 0.028 65 CM 0.105 0.011 25 CM 0.234 0.032 70 CM 0.100 0.001 30 CM 0.327 0.017 77 CM 0.106 0.002 TP Actual (34 cm) 0.335 NA Actual (Varying depths) 0.270 NA 40 CM 0.510 0.016 90 CM 1.141 0.301 45 CM 0.574 0.028 110 CM 0.418 0.037 50 CM 0.673 0.108 120 CM 0.772 0.032 20 CM 22.45 2.60 65 CM 12.30 0.54 25 CM 30.96 9.83 70 CM 12.56 1.62 30 CM 45.55 10.74 77 CM 12.33 0.33 TN Actual (34 cm) 45.89 NA Actual (Varying depths) 29.99 NA 40 CM 53.20 2.22 90 CM 103.24 17.38 45 CM 55.63 1.62 110 CM 48.46 5.30 50 CM 66.43 8.35 120 CM 115.39 15.77
69 The management of the water table during the summer period is more subjective to the needs of the farmer. The 2015 summer season had similar rainfall (371 mm) to the 30-year average
(368 mm). However, the height of the gates within the CD structure varied significantly during
2015 to reduce water ponding in the field. Therefore, the depths of the gates were varied from 65 cm to 120 cm with nine days of free drainage, to accommodate the agricultural needs of 2015. The carrot crops require the water table to be between 75 and 90 cm (McDonald and Chaput, 2010).
The scenarios found that keeping the water table depth at 77 cm from the soil surface allowed for a reduction in both the TN (12.33 kg ha-1) and TP (0.11 kg ha-1) loads compared to varying the water table. Furthermore, a lower water table (90 – 120 cm from the soil surface) increased the nutrient output significantly (TN 48 – 115 kg ha-1; TP 0.42 – 1.1 kg ha-1). However, the variation in the nutrient load results of 90 cm from the soil surface or below, identifies an increased margin of error. A concern was that data scarcity within the training and testing of the model could have impacted the model performance (Chen et al., 2017). Although ANN has been used in river systems approximately 210 times between 1999 and 2007, its use for water quality analysis is limited to a few studies (Maier et al., 2010). As the prediction uses the initial model at 70-80 cm from the soil surface, the decrease in the water table to 90 cm or below, reduces the accuracy of the model.
Overall, the results showed 77 cm from the soil surface as a viable water table depth that can maintain a good water table for carrot crops. As well, this gate height identified a potential decrease in nutrient loads. Increased datasets and a more stable water table for comparison in the future can increase the accessibility of using the ANN as a predictive model in agriculture.
Moreover, there is a need for a robust ANN development approach (Maier et al., 2010).
70
3.5 Conclusion
This two-year (2015-16) study assessed the impact of CD on the seasonal variations of N and P concentrations and loads in drainage water from an organic soil field in the Holland Marsh,
Ontario. The results showed that the N concentrations increased during the spring thaw due to increased drain discharge, while the P concentrations increased during the summer seasons due to fertilizer application in May of both years. The PO4-P concentrations had significant seasonal variation between the winter-spring and summer but similar annual trends, signifying that it can be used for seasonal assessment. Seasonally, the TN and TP load varied depending on CD discharge events, spring thaw, and high precipitation events that are more than 70% above the 30- year monthly average. The results from this study suggest that CD can greatly limit the nutrient loads by limiting the amount of drainage discharge from organic soil agricultural areas.
The ANN model was able to accurately predict nutrient loads in drainage outflow, using fertilizer, tile discharge and precipitation as input parameters. The ANN model for all nutrient loads had higher accuracy in predicting the summer load, as there was greater variability in input values for the training of the model, resulting in a more robust model to be used for predicting nutrient loads. Predicting nutrient loads under various CD scenarios by fluctuating the set water table height found that during the winter-spring, maintaining the gate height at 34 cm or less from the soil surface can potentially reduce the nutrient outflows. Furthermore, changing the water table height in the summer period found that keeping the CD gates at 77 cm can potentially reduce the nutrient load outflow while maintaining the water level depth needed to satisfy the soil-water-air environment for the crops. The study shows the potential for researchers to use the ANN model to predict N and P loads in the absence of intensive and long-term sampling in future water quality projects.
71
Connecting Text
The previous two chapters assessed the water quality and drainage water management systems on agricultural organic soils. However, the literature review in Chapter II identified a lack of studies that combine both the phosphorus (P) dynamics in both the soil and water. The differences identified between mineral and organic soils show that mineral soil studies cannot be used to understand organic soils. As the soil P dynamics are a driving factor for P leaching into the drainage water, it is imperative to study the dynamics of the soil-water system. Chapter IV of this thesis incorporated the soil P pools (including a P fractionation, Bray-1 available P, microbial biomass P, and root P content) into an assessment of P water quality from an organic soil agricultural field. The P was measured at two sites within the Holland Marsh for one year (2016).
This study titled “Linking soil phosphorus pools to drainage water quality in intensively cropped organic soils”. G. Grenon was responsible for the conceptualization, methodology, analysis and validation of results, investigation and resources, and writing the original draft along with all revisions and editing. Dr. Madramootoo provided supervision, conception, validation of results and valuable advice throughout the review and editing process. A. De Sena provided aid through investigation and data analysis, Dr. Hamrani provided support through Matlab coding. Dr. von Sperber provided valuable advice through the review and editing of the manuscript. To ensure consistency with the thesis format, the original draft has been modified, and the cited references are listed in the reference section. The funding for this project was provided by Dr. Chandra A.
Madramootoo from the Natural Sciences and Engineering Research Council of Canada (NSERC)
Strategic Projects Grant (447528 – 13).
72
4 . CHAPTER IV
Linking soil phosphorus pools to drainage water quality in intensively cropped organic
soils
4.1 Abstract
High fertilization rates are often recommended for the cultivation of organic soils, which over time, have led to phosphorus (P) pollution into receiving water bodies via subsurface tile drainage. Limited studies have documented the P pools within organic soils or their link to tile drainage water quality. This study quantified the different soil P pools found in organic soils under two water management practices: controlled drainage (CD) structure and pump drainage (PD) system. Soil samples were analyzed by sequential fractionation as well as available Bray-1 P test, root P content and microbial biomass. Drainage water samples were also analyzed for total P (TP), dissolved organic P, and dissolved reactive P (DRP). Correlation analyses assessed the relationships between the soil parameters, and with the drainage water quality parameters. The results identified calcium (Ca) bound P as the largest P pool. Its correlation with other P pools suggests that it acts as a P sink in organic soils. The correlation analysis further identified the aluminum (Al)-iron (Fe) bound P as a driving force for P movement in the soil, as it had the most significant relationship with the other parameters. A linear and quadratic regression analysis of TP found that the fertilizer and root P content were significantly related to the drainage water quality at both sites. A P balance at each site indicated that more fertilizer was being applied each year than was being released, causing an accumulation of P in the soil.
73
4.2 Introduction
Phosphorus (P) in drainage water released from cultivated organic soils severely impacts surrounding waterways as it promotes algal blooms and eutrophication. Specifically, drainage effluent from the Holland Marsh agricultural area contributes approximately 1 to 5% of the P loads to Lake Simcoe in southern Ontario, Canada (Winter et al., 2007). Other locations where this also occurs include the Great Lakes (Longabucco and Rafferty, 1989; Rockwell et al., 2005), the Baltic
Sea (Conley et al., 2002; de Jonge et al., 2002), and Lake Okeechobee (Daroub et al., 2011). The primary source of P in drainage water is from P fertilizer to the fields. Over-fertilization of P can occur because this nutrient is often a limiting factor for crop growth due to high P fixation in soils, which leads to low plant availability (Shen et al., 2011). Studies have shown that the management of P from agricultural areas is integral to the health of freshwater ecosystems (Schindler and Fee
1974; Thomas et al., 1995).
Peatlands are often cleared and drained for agriculture (Keller and Medvedeff, 2016).
Natural peat soils consist mainly of organic matter with only low quantities of mineral nutrients such as P. This deficiency in P requires high P fertilization rates for the intensive cultivation of these soils (Czuba and Hutchinson, 1980; Parent and Khiari, 2003; Guérin et al, 2007).
Furthermore, organic soils have a high water-holding capacity and a naturally high-water table necessitating the installation of subsurface tile drains to support crop production (Grozav and
Rogobete, 2012; Hallema et al., 2015). However, the outlets of these subsurface drainage systems are a conduit for P leaching into surrounding water bodies (Rockwell et al., 2005; Schindler et al.,
2008; King et al., 2015). The use of water management systems, including controlled drainage structures, pumps, levees, ditches, and canals, help with the removal of excess soil-water in organic soil areas (Thomas et al., 1995; Gambolati et al., 2006; Ilnicki, 2003). Strategies for the mitigation
74 of P pollution from drainage discharge cannot be implemented without an understanding of the P dynamics in the soil-water continuum.
Within the soil, P can be found in two major forms: inorganic P (Pi), and organic P (Po).
These two forms of P vary in function and fate within the soil matrix (Hansen et al., 2004). Plants
3- and microorganisms take up P from the soil solution as orthophosphate (PO4 ). Orthophosphate enters the soil solution either through dissolution and desorption from primary and secondary minerals or through the hydrolysis of Po by extracellular enzymes. However, orthophosphate in the soil solution quickly reacts with the surfaces of soil particles and can bind to iron and aluminum at pH < 7 or precipitate with Ca at pH > 7. These processes form inorganic phosphate compounds that are not directly bioavailable (Frossard et al.,2000; Hinsinger, 2001). For this reason, intensive cropping systems rely on the application of P fertilizers to maintain the supply of bioavailable P for crop uptake. However, excess P from fertilizers can accumulate in soils and eventually be lost from agricultural fields via erosion, runoff and leaching, leading to the water quality problems hitherto mentioned (McCann & Easter, 1999; Mbonimpa et al., 2014).
The majority of studies in the past have dealt with P losses from agricultural fields via erosion and runoff. Much less attention has been given to the impact of leaching, especially for arable organic soils. Directly linking changes in water quality to leaching losses of P is a difficult task, because the transfer of P through the soil-water continuum is slow, which delays and blurs the signal. Assessing the effects of fertilization and drainage practices on P transfer to freshwater bodies, therefore, requires both a detailed assessment of the biogeochemistry of P in soils and the continuous monitoring of the water quality in adjacent streams or lakes. Few studies have assessed the correlation between the biogeochemistry of P in soils and drainage water quality. Zheng et al.
(2015) found that the concentrations of dissolved reactive P (DRP or orthophosphate in solution)
75 in the leachate strongly correlated with the concentrations of P in the Bray-1 P test, more so than other standard P tests (Mehlich III and Olsen). A more recent study in arable organic soils showed that the amount of P that leaches strongly depends on the presence of aluminum (Al) and iron (Fe) oxides in the soil profile indicating that other geochemical soil P-pools might control leaching losses of P (Riddle et al.,2018).
To date, no study has combined a comprehensive analysis of the different biogeochemical soil P-pools with continuous monitoring of drainage water quality in subsurface drained organic soils. The objective of this study was to incorporate the soil P pools into an assessment of P water quality from an organic soil agricultural field. This was attained by linking the dynamics of different soil P-pools to changes in the chemistry of drainage water and by assessing the effect of different drainage practices on the transfer of P from arable organic soils to freshwater bodies. To achieve this objective, P in different biogeochemical soil pools and drainage water outflow were measured at two field sites in the intensively cropped organic soils of the Holland Marsh of
Ontario.
4.3 Materials and Methods
4.3.1 Study area and soil sampling
This study was conducted in 2016 using two commercial carrot (Daucus carota Bergen) fields on organic soils in the Holland Marsh, Ontario. Although two years of water quality data
(2015-2016: Chapter 3) was collected, the increased labor and laboratory costs of soil sampling allowed for only one year of analyses. The temperature during the growing season (May to
October) in 2016 averaged at 20.4 C with the 30-year climate average being 18.4 C. The precipitation for the 2016 growing season (274 mm) was 36.4% less than the 30-year average (431
76 mm). The precipitation from May to July ranged from 26 to 40 mm, which was significantly less than the 30-year average (75 – 80 mm). Overall, 2016 was a warmer and dryer year compared to the average (Government of Canada, 2018).
The two selected fields differed in their water management practices. Site 1 was a 4.2 ha field (44°03'54.0"N 79°35'16.8"W) equipped with controlled drainage (CD) structure installed on the collector tile line, near the edge of the field to control the water table level within the field.
This structure was installed 2 m deep with stackable gates allowing for the manual control of the water table height and, in turn, the volume of drain discharge from the field. Site 2 was a 5.6 ha field (44°02'49.2"N 79°35'20.4"W) with a pump drainage (PD) system consisting of subsurface tile lines that drain their water into a collection well before being pumped into a ditch. The pump is activated during periods of intense rainfall and spring-thaw depending on growers’ management strategies.
The initial soil properties (post-harvest 2015), including physical and chemical properties, as well as fertilizer application rates, can be found in Table 4.1. Using a random composite sampling scheme at both sites, 30 soil core samples were taken at 0 – 20 cm depth and then mixed to create one composite sample. Soil sampling took place three times throughout the growing season: pre-fertilization (PF), mid growing season (GS), and post-harvest (PH). The first sampling event (PF) took place in early May, preceding both the fertilization (Table 4.1) and seeding of the fields. The second sampling event (GS) occurred in early July to represent the fields after fertilization when the crops were already germinated and growing. The final sampling event (PH), occurred mid-October, after the harvest at both sites. The soil samples were taken back to the
McGill laboratory for analyses.
77
Harvest yield (Table 4.1) was measured at each site by calculating the amount of crop within a 1 m2 area. Six yield samples were taken, counted and weighed. The marketable yield was obtained following the removal of those carrots that were diseased. The average carrot yield for the Holland Marsh area in 2016 was approximately 50,830 kg ha-1 (Mailvaganam, 2018).
Table 4.1: Soil properties measured in October 2015 after the 2015 growing season and before the 2016 season.
Site 1 (CD) Site 2 (PD) Physical Properties Bulk Density (g cm-3) 0.31 0.20 Organic matter (%) 68 78 pH 6.6 6.2 Mehlich III Analysis P (kg ha-1) 406 324 Ca (kg ha-1) 21,775 16,285 Mg (kg ha-1) 1,439 1,649 Al (mg kg-1) < 40 < 40 Fe (mg kg-1) 475 353 Fertilization N (kg ha-1) 10 35 P (kg ha-1) 20 35 K (kg ha-1) 205 240 Marketable Yield Carrot yield (kg ha-1) 56,142 51,500
4.3.2 Soil phosphorus analyses
4.3.2.1 Phosphorus fractionation
The P sequential fractionation procedure was followed as given by Zhu et al. (2013), a modification of Hedley et al. (1982), using successive chemical extracts to operationally define the abiotic stabilization of P in the soil environment. Phosphorus fractionation was conducted
78 stepwise with deionized water (DH2O), 0.5M sodium bicarbonate (NaHCO3), 0.1M sodium hydroxide (NaOH), and 1M hydrochloric acid (HCl). Sequential soil extractions target operationally defined P pools. The DH2O extract targets available P, the 0.5M NaHCO3 targets moderately available P, the 0.1M NaOH extract targets less available P bound to Fe and Al, and the 1M HCl extract targets P bound to Ca (Tiessen & Moir, 1993).
Briefly, soil samples were air-dried for 72 hours, then ground and sieved with a 2mm mesh before being oven-dried for 48 hours at 60°C. A 2.0 g oven-dried soil sample was then extracted with 40 ml of the respective solvent for P fractionation, shaken for 30 min and centrifuged for 15 minutes before vacuum filtered to separate the filtrate and the residue for analysis. The process was replicated three times and repeated for the other solvents. For each extract, an aliquot was removed for digestion according to Ebina et al. (1983). TP and Pi were measured with the molybdate blue-ascorbic acid method (Watanabe and Olsen, 1965) on digested and undigested extracts, respectively. Po was calculated as the difference between total P and Pi. We refrained from determining the residual P pools by hot digestion with concentrated acids because the focus of this study was on those P pools, which are likely to become mobilized in organic soils under varying hydrological conditions. The resulting HCl extract was also analyzed for aluminum (Al), iron (Fe), magnesium (Mg), and calcium (Ca) content with flame spectrometry (Varian 220FS).
4.3.2.2 Soil-Available P
Soil-available P was analyzed according to the Bray-1 P test (Bray and Kurtz, 1945; Sims,
2000). The soil samples were sieved (< 2 mm) and air-dried for several days. A soil sample of 2 g was placed in a 250 mL flask with 100mL of acid ammonium fluoride extraction solution and was shaken for five minutes. Once shaken, the solution was filtered, and the extracts were analyzed by
79 colorimetry using the molybdate blue-ascorbic acid method (Murply and Riley, 1962). 20% of samples were analyzed twice to calibrate the results along with standards and blanks.
4.3.2.3 Microbial biomass P indices
Microbial biomass P (Pi fraction only) was estimated by the fumigation-extraction method
(Brookes et al., 1985). The soil was sieved (< 2 mm) and incubated at 25 °C for multiple days. The samples for fumigation were placed in large vacuum desiccators with chloroform (CHCl3) and sealed for 24 hours before the vacuum was released. The fumigated and non-fumigated samples were extracted using the Olsen P test (0.5 M NaHCO3, pH 8.5) and shaken for 30 minutes, filtered, and then analyzed for Pi using the molybdate blue-ascorbic acid method (Watanabe and Olsen,
1965).
4.3.2.4 Root P content
Harvested carrot roots were dried at 105 °C for 48 hours and ground. Samples (0.16 g) were digested with H2SO4, H2O2, and 4.4 mL of a lithium sulphate selenium solution. (Parkinson and Allen, 1975; Schneider et al., 2016). The solution was left to react overnight before being heated to 340 °C, then diluted to 100 mL using DH2O and analyzed using molybdate blue colorimetric analysis (Murply and Riley, 1962). Only the root was analyzed, as studies have found that the majority of the P content remains in the root with the leaves accounting for only 24% of P content (Cole, 1984).
4.3.3 Water quality analysis
Water samples were collected during periods of drain discharge, using an ISCO 6712 portable auto-sampler at 4-hour intervals. Additional discrete grab samples were taken throughout the year, except during periods of low water tables to obtain continuous P concentration data. The
80 water samples were stored in a 4 °C refrigerator at the Guelph University Muck Crops Research
Station (MCRS) before being transported in coolers with ice to McGill University for analysis.
Water samples were analyzed for dissolved reactive phosphorus (DRP), total dissolved P
(TDP), and total P (TP) using the Lachat XYZ Sampler (Hach Company, Loveland, Colorado).
DRP was determined by first filtering ((≤ 0.45 µm) 30-50 mL of each water sample and then doing a colorimetric analysis with the molybdate blue-ascorbic acid method (Murply and Riley, 1962).
TDP and TP were determined by digesting aliquots of the filtered and unfiltered water samples, respectively, with the potassium persulfate digestion method (Dayton et al., 2017; Ebina et al.,
1983). The dissolved organic P (DOP) was measured by subtracting the DRP from the TDP, leaving only the organic P.
4.3.4 Soil-water P balance
A soil water P balance was calculated using the initial soil P (Table 4.1) as the baseline of
P availability and the fertilizer application rate as the input. The outputs consisted of the root P uptake and the TP water quality load calculated through non-linear interpolation between the water quality and drain discharge at both sites. The values were converted to kg ha-1 for consistency of analysis.
4.3.5 Statistical Analysis
All data analyses were conducted using Matlab (MathWorks, 2019) and JMP (SAS
Institute, 2019). The soil P fractions were evaluated with a non-parametric Mann-Whitney test to assess the significant differences between the three periods of analysis (PF, GS, PH). The significant differences were determined at p-value < 0.05. Multiple statistical methods were used to assess the relationship between the soil P fractions and the P drainage water concentration including, correlation, and stepwise regression.
81
The measured P pools (i.e., soil-available P, microbial biomass P, root P, and all abiotic P pools) were assessed using the Pearson correlation analysis (r) to determine the relationship between the soil parameters at each site. The r analysis was also used to determine which P pools had a higher statistical significance to the water quality from the subsurface tile drainage at each site and during the different sampling periods (PF, GS, PH). The Pearson correlation is calculated using the following equation:
푛(∑ 푥푦)−(∑ 푥)(∑ 푦) 푟 = (4.1) √[푛(∑ 푥2)−(∑ 푥)2][푛(∑ 푦2) −(∑ 푦)2]
where x and y are parameter values, and n is the total number of integers. All forms of P were evaluated for relationships and then compared.
There is a lack of studies that use both soil P and drainage P water concentrations within one model. Multiple studies into soil P fractionation use either a linear or stepwise regression models for statistical analysis (Schlichting et al., 2002; Liator et al., 2004; Castillo and Wright,
2008). Therefore, precedent suggests that a stepwise regression model will assess all facets of the input parameters (soil) and their relationship to the output parameter (water). A stepwise regression model allowed for the finding of parameter(s) with a statistically significant effect on drainage water quality. The analysis allowed for both a linear and quadratic regression using forward and backward regression to find the input parameters with the best fit to the model (Bendel et al.,
1977). The input parameters consisted of all measured P parameters, while the output parameter was the P water concentrations found on the day of soil analysis. All scenarios were run for DRP and TP to assess the different responses of the soil P pools to the water quality. All models used a significance of p < 0.05. Additionally, the coefficient of determination (R2), and root mean squared error (RMSE) were obtained to assess the effect of each statistical model.
82
4.4 Results and Discussion
4.4.1 Soil chemistry
The soil mineral concentrations (Figure 4.1) are different between the two sites, and across the different time-periods (PF, GS, PH). The concentration of Ca is dominant, with Al being the least present. Furthermore, all mineral concentrations were highest during GS under CD, although only the Ca concentrations were significantly higher than PF mineral concentrations. There was a greater reduction in the mineral concentrations between PF and PH at CD. Under PD, Al and Mg were highest during GS, with Ca and Fe being higher during PF.
The concentration of Pi and Po in the different pools from sequential fractionation is shown in Table 4.2 along with microbial biomass P, root P, and available Bray-1 P. Although the use of the Hedley fractionation does not identify specific P pools bound to the minerals, as seen in recent literature (Gu et al., 2020), the over or underestimation of the P pools allows for comparison to past research. In all extracts, the amount of Pi was larger than the amount of Po. Large amounts of
Pi found in arable soils indicate a historical accumulation of P (legacy P) through long term application of inorganic fertilizers (Cogger and Duxbury, 1984; Withers et al., 2014). DH2O extractable P represented the smallest pool and HCl extractable P the largest pool in the soils similar to the findings of McCray et al. (2012), who quantified different soil P pools in organic soils in the Everglades Agricultural Area of Florida (EAA). The significantly high Ca content
(Figure 4.1) can be linked to the high HCl fraction of P found in the soil. Studies in the past have suggested that P bound to Ca can function as a sink of P within organic soils (Oberson et al., 2001;
He et al., 2008). Indeed, our results show that the HCl Pi fraction increased at both sites from PF
(CD: 750 and PD: 715 mg kg-1) to PH (CD: 1523 and PD: 1633 mg kg-1). In contrast, the NaOH
-1 fraction Pi (617 mg kg ) was highest during the PF period under CD and then continuously
83 decreased over the growing season. For each sampling period, there was substantially more NaOH-
P found under CD, which is likely caused by the higher Fe content in the soil (Table 4.1).
Furthermore, the NaOH-P pool contained a larger amount of Pi than Po, which is in contrast to many studies conducted in cropping systems on mineral soils (e.g. Schlichting et al., 2002; von
Sperber et al., 2017; Bauke et al., 2018). The largest pool of Po for both sites was the NaOH, followed by the HCl. Liator et al. (2004) found that the NaOH fraction in the soil was highly correlated to the Al, Fe, and manganese (Mn) mineral concentrations in the soil. This fraction is, therefore, correlated to the high Fe concentration found in our soils (Figure 4.1 and Table 4.1).
The pH in the soil can further affect the P reactions as with more acidic soils Pi will react with Al and Fe. However, in more alkaline or neutral soils, the Pi will form Ca phosphates (Jones and
Oburger, 2011; Shen et al., 2011). The soils in this study were more neutral (pH 6.4-6.6) and will, therefore, more readily bind with Ca. Other studies also found a high amount of Ca bound P in organic soils (Negassa and Leinweber, 2009). The NaHCO3 fraction was found to be constant at both sites in contrast to other studies which found this pool to be more transitory (Tiessen et al.,
1984; Zheng et al., 2002; Zheng et al., 2004). This might indicate the reason why we did not find any increase in the GS months.
84
24000 a. 22000 Site 1 (CD) 20000
) 18000
g
k /
g 16000
m
(
s n
o 14000
i
t
a
r t
n 12000
e
c
n
o c
10000
l
a
r
e n
i 8000 M 6000
4000
2000
0 Calcium Aluminum Iron Magnesium Pre-Fertilizer Growing Season Post-Harvest
24000 b. 22000
20000 Site 2 (PD)
18000
)
g k
/ 16000
g
m
(
s
n 14000
o
i
t
a r
t 12000
n
e
c n
o 10000
c
l
a
r e
n 8000
i M 6000
4000
2000
0 Calcium Aluminum Iron Magnesium Pre-Fertilizer Growing Season Post-Harvest
Figure 4.1: The change in the mineral concentrations Site 1 (CD) (a) and Site 2 (PD) (b) during pre- fertilization (PF), growing season (GS) and post-harvest (PH) 2016 season. The standard deviation is represented by the error bars.
85
Table 4.2: P pool parameters found in the soil, along with the standard deviation (St Dev) for each
parameter. The parameters are expressed as inorganic P (Pi) and organic P (Po) for both sites, over the three
sampling periods (pre-fertilization: PF; growing season: GS; post-harvest: PH).
Site 1 (CD) Site 2 (PD) PF GS PH PF GS PH
Pi 46.94 44.65 49.13 81.60 80.77 86.12
DH2O St Dev 0.57 1.51 0.36 0.47 7.87 0.88 -1 (mg kg ) Po 5.76 3.00 8.66 5.94 2.35 7.55 St Dev 1.28 1.66 0.50 0.85 1.35 0.64
Pi 145.83 117.60 132.02 100.40 100.93 101.13
NaHCO3 St Dev 6.17 3.37 2.60 9.36 7.86 4.22 -1 (mg kg ) Po 10.91 8.10 10.46 8.31 0.00 10.11 St Dev 5.32 3.38 7.77 3.01 0.00 3.49
Pi 671.03 583.63 518.95 295.33 237.92 233.58 NaOH St Dev 36.62 26.24 37.42 28.70 5.29 6.80 -1 (mg kg ) Po 145.96 205.73 135.47 95.44 174.88 96.68 St Dev 37.59 15.54 22.38 24.98 24.17 5.75
Pi 750.12 1487.57 1522.95 715.49 1572.12 1632.90 HCl St Dev 12.12 323.84 275.08 68.71 5.98 55.87 -1 (mg kg ) Po 36.06 153.05 49.47 19.77 199.45 52.29 St Dev 19.39 129.36 41.77 15.06 54.13 66.25
Bray 1 Pi 241.70 222.52 215.23 187.19 314.69 261.94 (mg kg-1) St Dev 23.05 18.58 29.75 37.36 49.58 60.88
Microbial P Pi 157.32 244.82 268.97 138.18 66.53 95.51 (mg kg-1) St Dev 119.25 95.09 113.05 14.13 12.56 33.98 Root P* Total P 0.00 NA 243.33 0.00 NA 284.84 (mg kg-1) St Dev 0.00 NA 35.12 0.00 NA 37.54 * Root P was analyzed only in terms of total P and therefore expressed in those terms here.
**NA represents the values not analyzed due to the inability to harvest mid-growing season
86
Increased microbial P throughout the sampling periods was found under CD. Phosphorus fertilizer addition in the soil can decrease the microbial P (Clarholm, 1993; Grierson et al., 1998;
Oberson and Joner, 2005). Therefore, the lower microbial P under PD can be linked to the high application of P fertilizer (Table 4.1). Furthermore, the increase in microbial P content (GS and
PH) following fertilizer application can be linked to the increase in available P present in the soil
(Oberson and Joner, 2005; Heuck et al., 2015). The available Bray-1 P found higher values at PD during the GS and the PH (262 – 315 mg kg-1). Zheng et al. (2015) found on organic soils that
Bray-1 P ranged from 15 to 214 mg kg-1, which was lower than the range found in this study (187 to 315 mg kg-1). However, Zheng et al. (2015) took all Bray-1 samples in the spring or late fall, which according to our values (PD-PF: 187 mg kg; CD-PH: 215 mg kg-1) were some of the lowest concentrations.
4.4.2 Relationship between the soil P pools and soil chemical properties
The correlation between the soil chemical properties and TP (Figure 4.2) shows that Al,
Ca, and Mg all had a positive correlation with fertilizer P, emphasizing that the fertilizer application affects multiple soil processes. Magnesium is an important co-factor for most phosphatases within the soil (Jones and Oburger, 2011; von Sperber et al., 2014), while Al can readily bind with the Pi, fixing it in the soil (Spivakov et al., 1999; Negassa and Leinweber, 2009).
The negative relationship between the NaHCO3 and the NaOH fractions with the HCl fraction indicates that the Ca bound P is a sink within the soil system (Jarosch et al., 2015; Gu et al., 2020).
The analysis showed that the NaOH bound P was correlated with most soil chemical properties at both sites, signifying the dynamic nature of this compound. Studies have found that the NaOH fraction, correlated to the Al and Fe oxides, rise in solubility with increased pH values from acidic to neutral (Hinsinger, 2001; Shen et al., 2011). Riddle et al. (2018) further found that the Al and
87
Fe oxides were correlated to the TP leaching in a study on organic soils. Both sites had a strong negative correlation between the root P content and the soil mineral concentrations and the NaOH pool. The correlation between the NaOH P fraction to the Al and Fe minerals within the soil has been found in multiple studies (Porter and Sanchez, 1992; Villapando and Graetz, 2001;
Janardhanan and Daroub, 2010). Here, we observe a continuous decline of the NaOH-Pi over the growing season at both sites (Table 4.2), which indicates that this pool is dynamic and might drive leaching losses from these systems. A similar observation has been reported in a study in Sweden
(Riddle et al., 2018), which suggested that Fe and Al bound P has a strong influence on P leaching in arable organic soils.
The microbial biomass P predominantly had a negative correlation with other P parameters at the sites. This is in keeping with the results of Amador and Jones (1993), where the increase in
P fertilizer and other P pools will decrease the microbial biomass P present in the soil. Noe et al.
(2001) suggested that the microbial biomass P relationship to soil TP is more important when there is a lack of nutrients in the soil. Although there are high negative correlations between multiple parameters, the root uptake in both fields was negatively correlated with all the minerals. The large amounts of P fertilizer applied, both present and past, have allowed for the increase in fertility levels on these organic soils. However, this has caused large fractions to become unavailable due to over-application (Schroder et al., 2011), as found in the identified negative correlation between nutrient concentration and P (Figure 4.2).
88
Fertilizer Fertilizer Fertilizer
DH2O DH2O DH2O
NaHCO3 NaHCO3 NaHCO3
NaOH NaOH NaOH
HCl HCl HCl
Bray-1 P Bray-1 P Bray-1 P S )
i
t D
e
C
2
(
Microbial Microbial
(
1 Microbial Biomass P
P
Biomass P Biomass P e
D
t i
) S Tuber P Tuber P Tuber P
Al Al Al
Ca Ca Ca
Fe Fe Fe
Mg Mg Mg
Legend -1 -0.8 -0.6 -0.4 -0.2 – 0.2 0.4 0.6 0.8 1
Figure 4.2: Pearson correlation (r) for TP between the soil parameters as well as the soil mineral measured values at each site for 2016. Only the strongly correlated (negative and positive) values are shown. 89
4.4.3 Concentrations of DRP, DOP, and TP in drainage water
The variations in the DRP, DOP, and TP concentrations at each site for April to October were identified in Figure 4.3. Under CD (Figure 4.3a), no samples were taken from early
September to late October as there was not enough water present in the CD structure due to the low levels of precipitation found from May to July. The reduced water levels were not a factor for the PD system as subsurface drainage continues to collect within the well regardless of whether the farmer pumps water from the collector well or not. An increase in TP concentration was found beginning in June at both sites. This increase reflects the increase of available P in the soil after the application of fertilizer. Although the increase is less apparent in the DRP concentrations, there is an increase in the concentrations at both sites. Higher DRP concentrations were found under
PD, which was caused by the pooling of the nutrients in the well for extended periods, allowing for the settling of nutrients disturbed during pumping. DRP is the form of P that is bioavailable to plants (Tan and Zhang, 2011), which is, therefore, a concern as it allows the ready use of P by algae, thereby creating toxic water environments (Longabucco and Rafferty, 1989; Daroub et al.,
2011). The results further showed that all TP concentrations were above the Provincial Water
Quality Objectives of 0.03 mg L-1 (Chambers et al., 2012; MOE, 1994). The highest concentration of TP under CD was 1.89 mg L-1 in July, while for PD it was 1.21 mg L-1 in early August. The
DRP concentrations across both sites ranged from 0.023 to 0.1 mg L-1, compared to the findings of Nicholls and MacCrimmon (1974), who measured concentrations ranging from 0.03 to 0.59 mg
L-1.
90
2 Fertilizer a. 1.8 application 1.6 Start of
) harvest
L 1.4
/ g
m 1.2
(
s
u 1
r
o h
p 0.8
s o
h 0.6 P 0.4 0.2 0
2
0
1
1
2
0
0
1
2
2
0
1
2
2
0
1
1
2
3
0
1
2
2
0
1
1
2
0
7
4
1
8
5
1
8
5
2
9
6
3
0
7
3
0
7
4
1
7
4
1
8
5
2
9
6
2
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
4
5
5
5
5
6
6
6
6
6
7
7
7
7
8
8
8
8
8
9
9
9
9
0
0
0
0
1
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
Dissolved Reactive P (mg/L) Dissolved Organic P (mg/L) Total P (mg/L)
1.6 Fertilizer 1.4 b. Start of 1.2
) harvest
L /
g 1
m
(
s
u 0.8
r
o
h p
s 0.6
o h P 0.4
0.2
0
2
0
1
1
2
0
0
1
2
2
0
1
2
2
0
1
1
2
3
0
1
2
2
0
1
1
2
7
4
1
8
5
1
8
5
2
9
6
3
0
7
3
0
7
4
1
7
4
1
8
5
2
9
6
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
4
5
5
5
5
6
6
6
6
6
7
7
7
7
8
8
8
8
8
9
9
9
9
0
0
0
0
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
Dissolved Reactive P (mg/L) Dissolved Organic P (mg/L) Total P (mg/L)
Figure 4.3: Concentration of dissolved reactive P (DRP), dissolved organic P (DOP) and total P (TP) in drainage water of Site 1 (a) and Site 2 (b) during the 2016 growing season.
4.4.4 Relationship between soil and water P parameters
The relationship of the soil P chemical properties to the outflow of P from the subsurface tile drainage differ at each site because of the agronomic practices and water management practices used by each crop grower. The correlation between the factors (Table 4.3) identifies the highest
91 correlation under CD for TP was found to be the HCl pool (r = 0.76), while for DRP, it was the
DH2O pool (r = 0.81). The NaOH was the highest correlations found under PD for both TP (r =
0.76) and DRP (r = 0.89). The results indicate that soil dynamics are site-specific. Strong negative correlations identify the specific P pool as sinks, while the positive correlations indicate more readily sources of P. The Bray-1, microbial biomass, and root P were only significantly correlated for PD. The variability in the correlated results can be linked to the difference in water management systems. Studies have found that the analysis of soil P can be both temporally and spatially site- specific (Baum et al., 2003). The soil properties between each study location differed chemically through soil and drainage management practices, which can affect the rate and reaction between the P bioavailability and soil nutrient content (Sharpley, 1995). The increased amount of Ca initially found under PD (Table 4.1) can be a factor in the correlation between HCl and DRP (Table
4.3). The historical management of the fields, which are under different growers, can also lead to varying correlations between the soil P and the drainage water quality.
Table 4.3: Pearson correlation (r) between the TP and DRP water quality from the tile drainage to the soil
P chemical analyses conducted at both sites and during the three sampling periods (Pre-fertilization, mid- growing season, and post-harvest). The significant negative correlations are shaded grey with strong positive correlations in bold.
Site 1 (CD) Site 2 (PD) r TP DRP TP DRP Fertilizer NA NA NA NA
DH2O -0.63 0.81 -0.76 -0.21
NAHCO3 -0.95 0.09 -0.42 -0.05 NaOH 0.04 -0.65 0.76 0.89 HCl 0.74 0.34 0.24 -0.99 Bray-1 P -0.26 -0.29 0.40 -0.72
92
Microbial 0.19 0.28 -0.45 0.78 Biomass P Root P content -0.15 NA -0.76 NA *NA = Not assigned
4.4.5 Regression models
A regression analysis of the entire April to October period (Table 4.4) identified the parameters that were closely linked to the output of TP and DRP at each site. The quadratic regression model for CD highlighted the importance of root P content, the NaHCO3 fraction, and the fertilizer application on TP concentrations, while only the DH2O fraction had a relationship to the DRP concentrations. Under PD, both fertilizer and root P influenced TP outflow, as well as
DH2O. The fertilizer application is an addition to the soil that increased the P concentration in the drainage discharge, while the root P content analysis occurred following harvest during which P concentrations decreased. The DRP outflow, identified correlation only with the HCl fraction.
Studies in the past on organic soils also reported that HCl-P was the largest fraction (Schlichting et al., 2002; Liator et al., 2004), while the DH2O and the NaHCO3 are both related to the availability of P in the soil. The fertilizer application and root P content, representing the P input and P output from the agricultural system, were the constant variables in the TP regression models at both sites. Therefore, this relationship signifies that the concentrations of TP in drainage water from the fields are partially dependent on both the crop type and the amount of P fertilizer applied.
4.4.5.1 Regression models for different sampling periods
The regression analysis (Table 4.5) for the 3 different sampling periods (PF, GS, PH) found no difference between the linear and the quadratic regressions. With the PF regression analysis,
2 the DH2O pool was the most significant parameter for both TP (p < 0.05; R = 0.96) and DRP (p
2 < 0.05; R = 1.0). The DH2O-P remained relatively constant at each site and represents only a 93 slight fraction, as found in other studies (McCray et al., 2012). In the assessment of GS, the NaOH had a relationship with the TP model, while the fertilizer application had a relationship with the
DRP concentration. Negassa and Leinweber (2009) stated that fertilizer application leads to greater available P, which shows a clear link from fertilizer to soil available P and additionally, the water
DRP concentration. The fertilizer correlation with TP during the GS corroborates results from previous studies (McDonald et al., 2013; 2014) which found that much of the applied P fertilizer leaches from the agricultural fields, as the soils are already saturated with P.
Furthermore, the relationship between NaOH and TP in the drainage water within the GS identifies this fraction of P as a predictor of P loss from the fields. This is corroborated by past studies (Cogger and Duxbury, 1984; Sharpley, 1995; Negassa and Leinweber, 2009). When
2 assessing the PH period, the regression models for TP correlated to DH2O (p < 0.08; R = 0.95), however, the p-value was not significant, and therefore the regression model was not considered to be strong. For the DRP PH models, both NaOH and HCl (p < 0.01; R2 = 1) were significant. He et al. (2008) found an eight-fold increase in the HCl fraction, which represents the P retained by the soil through historical fertilizer application (Vu et al., 2010).
The NaOH fraction was lower compared to the other sampling periods during post-harvest, which can be caused by the replenishment of the available pools. This was confirmed during laboratory experiments (Oberson et al., 2001; He et al., 2008). Soil sampling during the different periods (PF, GS, PH) indicated that the DH2O, representing bioavailable P, as well as the P bound to Al, Fe, and Ca, were all predictors of P in tile drainage flows from organic soils. Furthermore, there was a significant relationship between the water quality P and the P fertilizer application, which can be linked to over-application, as seen in past studies (McDonald et al., 2013; 2014).
94
Table 4.4: Stepwise regression analysis for both sites annually as well as during the three sampling periods. The stepwise regression was completed
at the 1st degree (linear regression) and the 2nd degree (quadratic regression).
Predictors Stepwise R2 RMSE p-value Microbial regression Fertilizer DH O NAHCO NaOH HCl Bray-1 P Root P 2 3 Biomass Results for TP (mg/L) at Site 1 - Annual Linear 0.996 0.0656 3.89E-05 ● ● ● st 1 -degree regression equation TP = 1 + Root P + NaHCO3*Fertilizer ● Quadratic 0.997 0.0634 3.39E-05 ● ● ●● nd 2 2 -degree regression equation TP = 1 + NAHCO3 + Root P + Fertilizer + NaHCO3 Results for TP (mg/L) at Site 2 - Annual Linear 0.993 0.0262 9.79E-06 ● ● ● st 1 -degree regression equation TP = 1 + Root P + Fertilizer + DH2O ● Quadratic 1 0.00293 1.43E-06 ● ● ●● nd- 2 2 degree regression equation TP = 1 + Root P + Fertilizer*DH2O + DH2O Results for DRP (mg/L) at Site 1 - Annual Linear 0.658 0.114 8.01E-03 ● st 1 -degree regression equation DRP = 1 + DH2O ● Quadratic 0.853 0.807 0.00315 ●● nd 2 2 -degree regression equation DRP = 1 + DH2O + DH2O Results for DRP (mg/L) at Site 2 – Annual Linear/ 0.985 0.0117 1.15E-07 ● Quadratic 1st and 2nd-degree regression equation DRP = 1 + HCl ● 1st-degree regression terms ●● 2nd-degree regression terms 95
Table 4.5: Stepwise regression analysis at the three sampling periods (PF, GS, PH) in 2016. The stepwise regression was completed at the 1st degree
(linear regression) and the 2nd degree (quadratic regression).
R2 RMSE p-value Linear and quadratic stepwise regression results Results for TP (mg L-1) Pre-Fertilization (May)
0.957 0.0279 7.09E-04 TP = 1 + DH2O Results for TP (mg L-1) Growing Season (July) 0.977 0.111 2.02E-04 TP = 1 + NaOH Results for TP (mg L-1) Post-Harvest (October)
0.949 0.0842 0.001 TP = 1 + DH2O Results for DRP (mg L-1) Pre-Fertilization (May)
0.999 0.00423 2.25E-07 DRP = 1 + DH2O
Results for DRP (mg L-1) Growing Season (July) 0.998 0.00273 9.75E-07 DRP = 1 + Fertilizer Results for DRP (mg L-1) Post-Harvest (October) 0.998 0.00978 0.000123 DRP = 1 + NaOH + HCl
96
4.4.6 Phosphorus balance through the soil matrix
The soil water P balance, found in Figure 4.4, identifies the initial soil P and the P fertilizer application in 2016 (Table 4.1). Furthermore, the output from the system was the P root uptake by crop, calculated by using 2016 yield measurements along with the root P content (Table 4.2) to acquire the value of root P uptake in hectares, and the TP load, calculated from the drainage discharge and TP concentrations through a non-linear interpolation. The drainage discharge for
2016 under CD was 23,760 m3, while for PD it was 2,797 m3, which resulted in TP loads of 0.58 kg ha-1 for CD and 0.18 kg ha-1 for PD. The P balance indicates that the fertilizer input was greater than the output at both sites. Moreover, increased fertilizer application had no significant effect on the crop root P uptake.
Site 1 Input: P fertilizer: 20 kg ha-1 Output: P crop root uptake: 13.66 kg ha-1 Soil TP water drainage load outflow: 0.58 kg ha-1 Initial soil P (Mehlich III): 406 kg ha-1
Site 2 Input: P fertilizer: 35 kg ha-1 Output: P crop root uptake: 14.67 kg ha-1 Soil TP water drainage load outflow: 0.18 kg ha-1 Initial soil P (Mehlich III): 324 kg ha-1
Figure 4.4: The representation of the P soil water balance found at sites 1 and 2. The balance includes the application of P fertilizer, initial soil P (Mehlich III), root P content and TP load found in tile drainage water.
Hamilton et al. (1975) found that the carrot root uptake from organic soil in Quebec was
25.8 kg ha. More recently, Parent and Khiari (2003) found that in Quebec total P content in soils
97 ranged from 375-1960 mg kg-1 with a P uptake by carrots being approximately 34 kg ha-1, while in Finnish soils they range from 190 – 2350 mg kg-1. The results from this study have a P content in soil (Table 4.1) within the Quebec range, however, the P uptake was lower. Overall, under CD with current fertilizer application, the annual P added to the soil was 5.76 kg ha-1, while under PD the annual P added to the field in 2016 was 20.15 kg ha-1. The legacy P, or initial soil P, is the historical accumulation of P in the soil (Kleinman et al., 2011; Sharpley et al., 2013). Therefore, both sites will continue to accumulate legacy P with current management practices, increasing the release of P into the drainage water and the environment. Zhang et al. (2020) found that halting P fertilizer application can decrease the legacy P in soils.
4.5 Conclusions
This study assessed the impact of P pools within the soil on the P outflow from drainage discharge on two organic soil sites in the Holland Marsh of Ontario, Canada. The drainage discharge occurred mainly between the PF and GS sampling periods with a higher discharge rate under CD. The drainage P water concentrations at both fields increased following fertilizer application. Within the soil, the largest P parameter was the Ca-bound P, which acts as a sink of P within these fields. The HCl and DH2O P fractions under CD and the NaOH P fraction under PD were found to correlate with the P drainage water concentration. The different drainage water management practices at each site affected the soil chemical dynamics, through changes in water table fluctuations and soil-water content. The CD raised the water table within the field creating a wetter dynamic, allowing for increased fluctuations within the soil Al-Fe bound P. The bioavailability of P was affected by the amount of microbial biomass P found in the soil, which is dependent on the fertilizer application and drainage practices. The regression analysis found a
98 statistical significance (p value of < 0.05) in the relationship between the fertilizer application, root
P, and available P with the drainage P water concentrations for both sites. The changes in drainage management affected the P pool dynamics within the soil and also the drainage P water concentrations. The 2016 P balance revealed that at both sites, the amount of P in the soils increased under CD and PD by approx. 5 kg ha-1 yr-1 and 20 kg ha-1 yr-1, respectively, leading to the accumulation of legacy P. Overall, the HCl and NaOH soil P fractions, the amount of P fertilizer applied, and the P uptake by crop were all important factors controlling the loss of P via drainage water. Future research needs to include the analysis of the dynamics of the residual P fractions and assess soil P dynamics through long-term studies to identify trends and peaks within the soil that translate into the water quality to create a sustainable agricultural system.
99 Connecting Text
The literature review in Chapter II found that most organic soil agricultural areas are drained wetlands and highlighted the use of tile drains, pumps, dykes, ditches and other water management practices used to manage the water table. Although a controlled drainage (CD) structure is not conventionally used on these soils, water management does occur. Many farmers within the Holland Marsh use subsurface tiles that drain into a collector well and the water is then pumped into a ditch. This system is only activated when needed and is a form of drainage water management. Chapter V, therefore, examined the effects of a pump drainage (PD) system in comparison to a CD system, on water quality outflow from subsurface tile drains.
This study titled “Nutrient loads in drainage water from two methods of drainage water management”. The author of this thesis was responsible for the conceptualization, methodology, analysis and validation of results, investigation and resources needed, and writing the original draft along with all revisions and editing. Dr. Madramootoo provided supervision, conception, validation of results and valuable advice through review and editing of paper. B. Singh provided aid through investigation and data calibration. Dr. von Sperber provided valuable advice through the review and editing of the manuscript. This paper is being prepared for journal submission. In order to ensure consistency with the thesis format, the original draft has been modified, and the cited references are listed in the reference section. The funding for this project was provided by
Dr. Chandra A. Madramootoo from the Natural Sciences and Engineering Research Council of
Canada (NSERC) Strategic Projects Grant (447528 – 13).
100 5 . CHAPTER V
Pollutant Nutrient Loads in Drainage Discharge from Two Methods of Water Management
5.1 Abstract
The release of available nitrogen (N) and phosphorus (P) from agricultural tile drainage contributes to the eutrophication in water bodies. To mitigate the agricultural impact of nutrient release, drainage water management has been proposed as a beneficial management practice, to limit N and P in tile drainage effluent. This study assessed the nutrient water quality under two drainage water management systems for two years (2015-2016): controlled drainage (CD) comprising a series of stackable gates to manually control the water table level; and, pumped drainage (PD) which uses a submersible pump within a collector well that discharges effluent when activated. The N loads were found under CD to be higher in 2016 (53 kg ha-1) than 2015 (35 kg ha-1) due to periods of increased discharge volume. Furthermore, the rate of P loads decreased from 2015 to 2016. The N loads for the PD system were nearly 10 times higher in 2015 compared to 2016, caused by the variations in drainage volumes. Limited significance was found in both the correlation and regression analysis between the discharge and N loads, suggesting that soil biogeochemical and hydrological properties impact N release into drainage outflow. Furthermore, the discharge volume was found to have statistical significance with the total P loads under both water management systems. Overall, water management in the form of CD or PD can reduce the nutrient load environmental impacts through reduced discharge volumes.
101 5.2. Introduction
Agricultural effluent in the form of nitrogen (N) and phosphorus (P) have been found to cause eutrophic algal blooms in water bodies (Elser et al., 2007; Schindler et al., 2012; Van
Esbroeck et al., 2017). Agricultural organic soils have low inherent soil N and P, which can lead to over-fertilization and increase the release of nutrients from fields (Czuba and Hutchinson, 1980;
Parent and Khiari, 2003; Guérin et al., 2007; Gramlich et al., 2018). Examples can be found in
New York where arable organic soils contribute up to 39% of the total P (TP) load entering Lake
Ontario (Longabucco and Rafferty, 1989), and in Ontario where 1 to 5% of P loads into Lake
Simcoe are from organic soil agriculture (Winters et al., 2007). Miller (1979) found that nitrate
-1 (NO3-N) loads in the drainage water can range from 37 to 245 kg ha from agricultural organic soils due to the N fertilizer application.
As organic soils are usually found in areas of drained wetlands, these soils have inherent high-water tables and require some form of drainage water management practice to make them suitable for agriculture, such as dykes, levees, pumps, subsurface tile drains, and ditches (Thomas et al., 1995; Gambolati et al., 2006; Ilnicki, 2003). Furthermore, drainage infrastructure has also been implemented on organic soils in the form of tile drainage to reduce the effects of subsidence that occurs with intensive agricultural practices (Gambolati et al., 2006; Ilnicki 2003).
Drainage water management (DWM) has been proposed as a beneficial management practice to limit the release of N and P. Controlled drainage (CD) is a form of DWM that utilizes a control structure placed on the collector tile line, close to the outlet, to control the water table level within the field (Elmi et al., 2002; Skaggs et al., 2010, 2012; Carstensen et al., 2016).
Research into CD has found that NO3-N loads can be reduced by 18 to 79% in Midwest US on mineral soils (Skaggs et al., 2012). Further research has found N and P concentrations are not
102 affected by CD, but the inherent reduction in discharge, leads to reduced nutrient loads (Skaggs et al., 2012; Williams et al., 2015). Another form of DWM is pump drainage (PD), where a submersible pump is used to evacuate drainage water from a collector well into a ditch. PD can be set to activate when water reaches a certain level in the sump (Poole et al., 2018), or can be activated by the crop grower, should he anticipate the water table rising into the root zone or field surface (Bhadha et al., 2017). In agricultural organic soils, such as the Holland Marsh, Ontario,
PD is naturally used by farmers to drain excess water from these areas (Miller, 1979). These PD systems are only used during periods of excess water. As opposed to PD, which uses one set water table level, CD allows for greater flexibility of drainage outflow by the modulation of the set water table during field operations (seeding, tillage), and winter/summer months, preventing excess drainage (Cooper et al., 1991; Drury et al., 1999; Williams et al., 2015). There have been limited studies on these two DWM systems and their benefits under cultivated organic soil.
To better understand the implementation of the two different DWM systems on organic soils, a study was conducted on two farms within the Holland Marsh, Ontario, Canada, over two years (2015-2016). The objective was to assess the relationship between nutrient loads and the drainage discharge from two different DWM practices (CD and PD).
5.3 Methodology
5.3.1 Study area
The study took place during the 2015 and 2016 growing seasons on two agricultural fields in the Holland Marsh, Ontario, Canada (44.0415°N, 79.6001°W). Holland Marsh is a drained peatland where approximately 60% of the area is used for intensive, high-value crop production.
The area provides approximately 14% of provincial vegetable production and contributes $58
103 million to the provincial economy (Township of King, 2012). The temperature during the growing season (May to September) of 2015 and 2016 ranged from 14.6 to 24.3 C, which was slightly above the 30-year average of 13.5 to 21.8 C. During 2015 and 2016, precipitation was lower than the 30-year average (774 mm) with 675 mm and 630 mm, respectively (Government of Canada,
2018).
5.3.2 Agronomic practices
A composite soil sample was collected post-harvest 2015 and analyzed through Agri-Direct
(Quebec, Canada) for chemical soil properties at both sites (Table 5.1). The cultivated crop for both fields was carrots (Daucus carota Bergen) and planting was conducted after fertilizer application in May. Fertilizer application rates in the spring for both sites are documented in Table
5.1. The crop harvest started in October for both sites.
Table 5.1: Soil properties measured in October 2015
Site 1 (CD) Site 2 (PD) Physical Properties Bulk Density (g cm-3) 0.31 0.20 Organic Matter (%) 68 78 pH 6.6 6.2 Mehlich III Analysis P (mg kg-1) 655 810 Ca (mg kg-1) 35,056 40,625 Mg (mg kg-1) 2,320 4,111 Al (mg kg-1) < 40 < 40 Fe (mg kg-1) 475 353 Fertilization 2015 N (kg ha-1) 0 30 P (kg ha-1) 20 30 K (kg ha-1) 210 205 Fertilization 2016
104 N (kg ha-1) 10 35 P (kg ha-1) 20 35 K (kg ha-1) 205 242
5.3.3 Water management
At the CD site, a 4.2 ha tile-drained field was equipped with a water control structure
(Figure 5.1a). The structure was installed at the outlet of the collector tile line allowing for the manual control of the water table through the use of stackable gates. These gates prevented the outflow of water from the field, unless the water table rose above the set gate level. A compound weir was installed at the top-most gate, which allowed discharge measurements. The weir consisted of 11° V-notch to estimate flows less than 1 L s-1 and a rectangular weir to calculate flows above 1 L s-1. The discharge was continuously recorded at 15-minute intervals using a pressure transducer within the water control structure. The water table level was adjusted according to the growing season (Adeuya et al., 2012; Gunn et al., 2015; Skaggs et al., 2012) with the gate levels being placed between 30 and 40 cm from the soil surface from November to April to reduce drain outflow (Gunn et al., 2015), and lowered between May and October to 70 to 80 cm for optimal crop production (McDonald and Chaput, 2010). However, intense precipitation in June
2015 resulted in ponding, leading to the removal of the gates and subsequent free drainage for approximately 6 weeks. July 2016, by request from the grower, the gates were set with the water table 70-80 cm from the soil surface to counteract the lack of precipitation by restricting drainage and conserving water within the field.
At the PD site, a 5.6 ha field with a tile drainage system (Figure 5.1b) discharged into a collector well. Within the collector well, a float-controlled water pump activated to discharge access water from the well into a ditch that drained into the West Holland River. The pump was activated at the grower’s discretion during periods of spring-thaw and rainfall events to mitigate
105 water ponding within the field. An H-flume with a sonic sensor was installed at the pump outlet to measure pump discharge at one-minute intervals. A log was kept with a record of the start and end of pumping events. These two DWM systems are being studied simultaneously, however, the research was not a comparison between the systems as the field management, fertilizer application, subsurface tile drain spacing, soil physical and chemical properties are all different. Therefore, these systems were studied to assess water quality impacts on each separate system.
5.3.4 Water quality analysis
Water samples were taken by an ISCO 6712 portable auto-sampler (Teledyne ISCO,
Lincoln, Nebraska) at both sites. Under CD, during periods of water discharge, water samples were taken at 4-hour intervals. At the PD site, during the periods where the submersible pump was active, discharge events lasted as short as 30 seconds to more than 2 minutes as the sump within the pump regained its level. The water from the field would re-fill the collector well and reactivate the pump, continuously raising the water level in the H-flume between 2 and 10 cm. To prevent over-sampling, the portable auto-sampler was set to take water samples once the water level within the H-flume reached 5 cm, 10 times to prevent over sampling. These two sampling regimes allowed for a similar number of water samples to be taken from the sites over 7 days. Additionally, discrete grab samples were taken at both sites weekly, except during periods where the temperature fell below -10 °C or where the low water table elevation made it impractical. All samples were stored in a 4 °C refrigerator at the University of Guelph Muck Crops Research Station (MCRS), before being transported in coolers with ice to the McGill University labs for analysis.
106 a. b. PUMP DRAINAGE
WATER LEVEL IN SOIL WATER PUMPED INTO DITCH PUMP SUBSURFACE TILE DRAIN
SATURATED SOIL
Figure 5.1: The illustration of the two drainage water management systems: (a) the installation of a controlled drainage (CD) structure on the subsurface tile line which allows for the manual adjustment of the water table height in a field with removable gates that stack on top of each other and (b) a pump drainage (PD) system where the water from the subsurface tile lines drain into a collector well and the water is then pump out of the system and into a ditch.
107 Water samples were assessed for nitrate (NO3-N), total N (TN), dissolved inorganic phosphate (PO4-P), and total P (TP) concentrations, using the Lachat XYZ Sampler (Hach
Company, Loveland, Colorado). Approximately 40 mL of water from each sample was filtered using a 0.45 µm filter. A colometry analysis was used to determine NO3-N and PO4-P concentrations from filtered water samples (Murply and Riley, 1962; Lachat Instruments, 2003), while TP and TN measurements were made by first digesting unfiltered samples at a 1:1 ratio with potassium persulphate in an autoclave (temp; 1 h) followed by analysis with the above colorimetric methods (Dayton et al., 2017; Ebina et al., 1983). The analysis was conducted by placing a 1:1 ratio of sample to persulfate solution in a test tube and autoclaving the samples for one hour.
Distilled water controls and standards of N and P were included for quality control and the calibration of the sample results, respectively. Nutrient loads were calculated using the measured
PO4-P, TP, NO3-N and TN concentrations and CD and PD drain discharges. Nonlinear interpolation using a power equation was used for load calculations, due to its greater accuracy
(Phillip et al., 1999).
5.3.5 Statistical Analysis
The Pearson correlation (r) was run between the nutrient loads and the discharge to assess the strength of the relationship between the parameters. In addition, a linear regression analysis was used to assess the relationship between the nutrient loads and the drain discharge. The statistical significance of the regression was investigated at a significance level of p < 0.05, as well as through its coefficient of determination (R2) and its correlation coefficient (R).
108 5.4 Results and Discussion
5.4.1 Drainage discharge analysis
The discharge time series for both years under the two DWM systems are shown in (Figure
5.2). The 2015 spring-thaw in March-April showed discharge trends at both sites. The hydrograph identifies a large peak under PD (Apr 11: 29.4 mm day-1), while under CD there was continuous smaller daily discharge (Apr 22: 6.2 mm day-1). The precipitation for June-July accounted for a total of 185 mm of precipitation, a 19% increase on the 30-year average (155 mm). The rainfall in
2015 occurred from June 7th to July 8th, causing peak discharge events at both sites. The discharge under CD occurred immediately following the start of the increased rainfall, from June 13th to July
10th, while PD showed the discharge occurring between June 22nd and August 1st. The majority of the discharge in 2015 occurred due to increased rainfall in June-July. This direct correlation was also found in by Helmers et al. (2012) with large precipitation periods causing a dramatic increase in discharge. In 2016, higher flow rates were experiences under CD, typically of the order of 2 to
23 mm day-1 between March 6th to May 30th. The PD system with reduced pumping had consistently low discharge with a peak flow rate of approximately 2 mm day-1 between March and the end of July. Studies have shown that some of the most significant discharge events occur during the spring thaw, which was seen in 2016 (Jamieson et al., 2003; King et al., 2014; Lam et al.,
2016).
The annual volume of discharge from both sites (Table 5.2) found that under CD there were greater discharge volumes in 2016 (23,760 m3) compared to 2015 (16,025 m3). The PD system, however, found higher discharge volumes in 2015 (13,699 m3), compared to 2016 (2,797 m3). March 2016 found increased monthly precipitation (80 mm) compared to the 30-year average
(46.5 mm). The high volumes in 2016 were caused by the lowering of the gates within the CD
109 structure in early March to facilitate drainage of the field due to increased rainfall and to allow the lowering of the water table within the field for fertilization and seeding. Multiple studies have shown that using CD can reduce the drainage outflow by 18 to 79% compared to free drainage
(Williams et al., 2015; Lalonde et al., 1996; Wesström et al., 2001). Although free drainage was not measured, CD limited the periods of discharge to spring-thaw and intense rainfall events (June-
July 2015, March 2016), thereby reducing the total drainage discharge. The PD protocols were governed on the management decisions of growers, which limited the periods that the pump was used. The pump, when on, activates the pump only when it reaches the set level, affected the rate and flow patterns of the discharge water (Ayars et al., 2006). Variability in the management of PD can be an influencing factor on total drainage discharge. PD is used during high precipitation events to reduce the chance of flooding (Daroub et al., 2005; Bhadha et al., 2017).
Table 5.2: Annual discharge and nutrient loads for total P (TP), phosphate (PO4-P), total N (TN) and nitrate
(NO3-N) at Site 1 (CD) and Site 2 (PD).
Total Volume TP Load PO4-P Load TN load NO3-N Load Year (m3) (kg ha-1) (kg ha-1) (kg ha-1) (kg ha-1)
2015 16,025 0.60 0.28 34.73 24.63 CD 2016 23,760 0.58 0.34 52.86 43.76 2015 13,699 0.90 0.69 20.63 5.88 PD 2016 2,797 0.18 0.06 2.09 0.67
110 40
35
30
)
y
a
d /
m 25
m
(
e g
r 20
a
h
c
s
i D
15
y
l
i a
D 10
5
0 Jan Feb Apr May Jul Sep Oct Dec Feb Mar May Jul Aug Oct Dec 2015 2016 Discharge Site 1 (CD) Discharge Site 2 (PD)
Figure 5.2: A 2015 and 2016 time series analysis of the subsurface tile drain discharge from Site 1 (CD) (black) and Site 2 (PD) (grey).
111 5.4.2 Nitrogen load assessments
The TN loads (Table 5.2) found during the two years of study, showed a 1.5-time increase in 2016 loads under CD compared to 2015. The continuous release of water in 2015 under CD through the freeze-thaw period allowed for an increase in the biogeochemical N processes (Mejia and Madramootoo, 1998; Elmi et al., 2000; Skaggs et al., 2012). Another factor in increased amounts of N under CD can be associated with the increased N mineralization occurring in the soil because of subsidence and organic matter oxidation (Santos et al., 2020). Terry (1980) found that approximately 686 kg N ha-1 is mineralized annually per centimetre of soil subsidence that occurs within organic soil agricultural fields in the Everglades Agricultural Area (EAA). These properties vary with location and can affect the N loads found in the drainage water. Figure 5.3 illustrates the significance of TN and NO3-N under CD. The NO3-N loads under CD accounted for approximately 71 and 83% of the TN loads each year respectively. Nitrogen transport from agricultural fields is mainly in the dissolved form, specifically NO3-N (Zak et al., 2008). This could be a factor in high NO3-N loads under CD in 2016 (Table 5.2) highlighted by the increased volume of drainage water. Duxbury and Peverly (1978) found the NO3-N load to range between 39 and 87
-1 kg ha . Past studies have found that the use of CD can decrease the NO3-N concentration in drainage water and can reduce the load through drainage water reduction (Reddy, 1982; Skaggs et al., 1994; Martin et al., 1997; Dinnes et al., 2002).
Under PD the TN and NO3-N loads were higher in 2015. The NO3-N loads accounted for approximately 29 and 32% of the TN loads in 2015 and 2016, respectively. The higher discharge in 2015 increased the nutrient loads during that period. The correlation analysis (Figure 5.3) found a significant relationship to discharge under PD. The time between pump activation may result in the accumulation of suspended particles that have the time to settle within the sediments (Bhadha
112 et al., 2017), in the tile drains and collectors well. Jia et al. (2012) found in a study on 16
agricultural field wells under subsurface drainage and sub-irrigation, that well locations, as well
as soil physical and chemical properties affected the water quality. Fertilizer management is
another factor that affects NO3-N loads (Skaggs et al., 1994) as approximately 65% of applied N
can be lost to leaching or volatilization (Cameron et al., 2013).
Site 2 (PD) Dis TP TN PO4 NO3 Correlation Dis (R ) Legend TP 0 TN 0.4 PO4 0.6 NO3 0.8 Site 1 (CD) 1
Figure 5.3: Pearson correlation (r) distribution of the discharge and the nutrient loads. Both Site 1 (CD)
(bottom) and Site 2 (PD) (top) are represented in this distribution.
The regression analysis (Table 5.3) found that there was a significant relationship between
no3 N loads under CD and discharge. However, the correlation (Figure 5.3) did not identify the same 0.85805 significance between the N loads and discharge. Under the PD system, the regression analysis <.0001
0.81476 found no significant relationship between discharge and the NO3-N loads (p-value 0.59). <.0001 Therefore, the N loads are influenced by more than the DWM practices, including fertilizer 0.98155 <.0001 application, soil biogeochemical properties and the hydrological movement of water through 0.84454 <.0001 individual fields (Kennedy et al., 2018). 1
113
no3 Table 5.3: The linear regression analysis of discharge at site 1 (CD) and site 2 (PD) for nutrient load assessment.
Equation R2 Pr>F Correlation (R) SITE 1 (CD) TP y= -0.87 + 5.56x 0.86 < 0.0001 0.93
PO4-P y= -0.35 + 2.86x 0.89 < 0.0001 0.91 TN y= -55.09 + 400.1x 0.84 < 0.0001 0.94
NO3-N y= 41.56 + 294.4x 0.74 < 0.0001 0.86 SITE 2 (PD) TP y= -1.93 + 16.48x 0.88 < 0.0001 0.94
PO4-P y= -6.63 + 14.61x 0.30 < 0.0001 0.99 TN y= -98.39 + 380.8x 0.98 < 0.0001 0.55
NO3-N y= 30.5 + 0.2x 0.001 0.59 0.04
5.4.3 Phosphorus load assessments
Carpenter (2008) found that P management, more than N, was the critical nutrient to manage in freshwater systems to reduce eutrophication. The TP loads (Table 5.2) under CD were less than 5% different across the two years. The PO4-P loads found were approximately 47 and
59% of the TP loads in 2015 and 2016 respectively. Although intense storm events represent a small portion of the annual rainfall, they can account for the majority of the P load (Bhadha et al.,
2017). The rainfall of June-July 2015 and March 2016 increased the discharge volume in both years during these periods. The PO4-P loads (Table 5.2) under CD were represented by 46 and
58% of the TP load in 2015 and 2016 respectively. PO4-P, also known as dissolved reactive P, in the form of P that is immediately bioavailable to aquatic plants and therefore of greater concern to the environment (Hoepting, 2009; Tan and Zhang, 2011; Zheng et al., 2014; 2015). Factors affecting the amount of PO4-P found in the water outflow are legacy P, fertilizer input, soil P dynamics (Gramlich et al., 2018). The correlation distribution (Figure 5.3) found a strong positive correlation between TP loads and discharge under CD. Williams et al. (2015) reported that the
114 nutrient loads are not affected by water management but are more correlated to the reduction in discharge, which correlates with our study. To assess the normalized rate of the P loads (Figure
5.4), a graphical representation of the loads over a set discharge was conducted. The rate of P release allowed for the identification of trends within the water management systems. The rating assessment found that between the two years, there was a decrease of 54% in the rate of TP release under CD. The 2016 TP loads occurred in the spring-thaw period, the high volume of water allowed for the dilution of the TP concentration and therefore effecting the loads. Other studies
(Nash et al., 2015) have proposed that under CD the drainage water increased P uptake, decreasing available P in water.
The P loads under PD varied between the two years of study (Table 5.2). The 2015 TP load
(0.90 kg ha-1) was 5 times greater than the 2016 load (0.18 kg ha-1). The reduction in TP loads is due to the reduction of the discharge volume, identifying the volume as a significant factor in P load reduction. Studies have found that hydrological conditions are a major factor in nutrient concentrations and loads (Kennedy et al., 2018). This collaborated with the correlation results
(Figure 5.3) that found a strong positive correlation between TP loads and both discharge and PO4-
P loads under PD. The water velocity of flow affects the particle distribution of P (Garcia, 2000;
Bhadha et al., 2017), which the PD system allows to be constant when turned on perhaps reducing the P loads through the low flow rates. The rate of P loads (Figure 5.4) allows for the comparison of the annual loads as the normalization of the volume removes the variation in drainage flows.
The discharge in 2016 was 5 times less than 2015, however the TP rate of release was similar in both years. The constant flow rates under a PD system may generate a TP load that is comparable year to year. Other factors that may affect the P concentrations and therefore loads are the soil physical and chemical properties (Jia et al., 2012). The PO4-P loads (Table 5.2) under PD, represent
115 77 and 35% of the TP annual loads, respectively. The complex factors affecting the P loads highlight the significance of the P export pathway through subsurface drainage outflow, as other studies have also found (Sims et al., 1998; King et al., 2015).
0.07 )
e 0.06
m
u
l o
v 0.05
/
a
h /
g 0.04
k
(
s
d 0.03
a
o
l
P
0.02
f
o
e t
a 0.01 R 0 TP PO4-P TP PO4-P Site 1 (CD) Site 2 (PD) 2015 2016
Figure 5.4: The graphical representation of the normalized (by volume) P loads was released in 2015 and
2016 at both sites. The rate is depicted over an equal volume, to identify trends within the overall water quality.
The regression analysis (Table 5.3) found a significant relationship between the TP and the discharge under both DWM practices. These results concur with the correlation analysis (Figure
2 4.3). However, the regression found no significance (R = 0.3) in the PO4-P load relationship with a discharge under PD, which was also found in the correlation analysis. The poor correlation may be associated with several factors, as indicated by previous studies which found the interaction between P concentration in water, suspended particles, hydrological movement and decomposition of organic matter are all factors that affect the P dynamics within the water (Sheng et al., 1998;
Koshi-Vähälä and Hartikainen, 2001; Bhadha et al., 2017). The relationship between discharge
116 and P can further be affected by the fertilizer application, leading to the saturation of P within the soil and a reduction in its integration, increase P concentration in drainage discharge (Miller, 1979;
Longabucco and Rafferty, 1989; Daroub et al., 2011). Studies within the Holland Marsh on fertilizer treatments found that the application of P fertilizer did not show a significant change to the soil P content, but that most of it was lost through leaching (McDonald et al., 2013, 2014).
Therefore, the initial P content in the soil (Table 5.1) can affect the assimilation or leaching of P fertilizer. Research from organic soils in the Everglades Agricultural Area showed similar results
(Thomas et al., 1995; Daroub et al., 2011).
5.5 Conclusion
Drainage water management has been extensively studied in mineral soil agriculture.
However, limited studies have assessed the implementation of a CD structure or the current practice of a PD system and its effects on the pollutant nutrient loads from organic soil agriculture.
The management of these two DWM systems varied significantly and had different impacts on water quality. The N loads under CD were affected by increased periods of discharge, specifically during the spring-thaw. Furthermore, the P loads were higher in 2015 and the rate of P decreased by 35% in 2016. The N loads from the PD system was 10 times higher in 2015 than 2016 because of the variations in drainage volumes between the two years. The low correlation between the discharge and N loads indicate an effect of biogeochemistry and hydrological soil movement on soil N dynamics. The P loads under PD were 5 times higher in 2015, compared to 2016, which was caused by the reduction in discharge volume by nearly the same percentage. Overall, the management of P loads is critical for the reduction eutrophication within freshwater aquatic systems. The correlation between TP loads and discharge was found in both DWM systems
117 showing the importance of discharge to P management. Both DWM systems have the potential to reduce discharge volumes and therefore reduce nutrient loads through higher water tables for CD and reduced periods of pump activation for the PD system. However, increased comparative studies on the variance between the two DWM systems are needed to identify which system can have long-term beneficial effects on P management.
118 6 . CHAPTER VI
Summary and Conclusions
6.1 General summary
The application of drainage water management has been studied extensively on mineral soil. However, the same cannot be said for organic soils, and the findings of mineral soils cannot be applied to mitigate the nutrients in tile drainage discharge in organic soils. This is because organic soils have unique characteristics that affect their biogeochemical and hydrological soil properties. Furthermore, organic soils are inherently low in nitrogen (N) and phosphorus (P), which leads to over-fertilization and increased loss of nutrients in drainage outflow. Organic soils have a high-water table and there is a constant need for water management in multiple forms including, tile drainage, ditches, dykes, and pumps. Intensive agricultural cultivation in highly organic soils of the Holland Marsh has been linked to increased eutrophic algal blooms in adjacent lakes and rivers. This thesis assessed the application of a controlled drainage (CD) structure and the current pumped drainage (PD) practices to mitigate nutrient loads and improve P management in the soil.
A two-year intensive field study was conducted using two field sites under a carrot crop rotation in the Holland Marsh, a prominent organic soil area known for its intensive vegetable production in Ontario. The study evaluated the discharge at both sites, along with the total N (TN), nitrate (NO3-N), total P (TP), and dissolved reactive P (DRP) concentrations and loads. The CD water quality was assessed from seasonal patterns and for the use of an artificial neural network
(ANN) model to predict nutrient loads in water management scenarios done through changing water table depth. The results indicate that the N concentrations increased during the spring-thaw,
119 while the P concentrations were positively affected by fertilizer application. Furthermore, the use of a CD structure reduced the drainage discharge, allowing no release during the fall period
(October – December). The use of the machine learning ANN model found that it was possible to use hydrological (precipitation and discharge) parameters and fertilizer application to model seasonal nutrient load predictions. The ANN model predictions were validated using the 2015 data.
The validated model was used to predict various drainage discharge scenarios by modifying the water table height. The winter-spring (January – April) model found that a set water table of 34 cm or above the soil surface, greatly reduced TP and TN loads, while during the summer period
(May – September) a water table depth of 77 cm was able to reduce the nutrient loads, while also allowing for suitable water table depth for carrot crop growth.
The results from the water quality assessment of the two sites identified a gap in understanding the P dynamics. To address this gap, it is essential to take in account the movement of P through various P pools within the soil that can affect the amount of P concentration within the drainage water quality. Therefore, a soil P fractionation was done along with the Bray-1 P test, the microbial biomass and the root P uptake analysis to evaluate the relationship between the inorganic P (Pi) and organic (Po) fractions within the soil and the TP and DRP water concentrations.
The study results show that the calcium (Ca) bound P was identified as a sink for P within these soils. Furthermore, the NaOH P fraction was found to be dynamic and a significant pool that affects the movement of P through the soil. The fertilizer application and root P uptake content had significant relationships to drainage TP concentration at both sites. The valuation of a P balance found that there was less annual accumulation of P at site 1, compared to site 2. Overall, the study concluded that the amount of fertilizer application and the drainage water management practices are equally important to the valuation of P movement on organic soils.
120 The two different water management systems, CD and PD, affect the water quality discharged from agricultural fields. The study assessed the effects of these systems on the water quality loads and their relationship to the drainage discharge. The N loads under CD were higher in 2016 due to the increase in discharge and effluent during the spring-thaw. While under PD, the
N loads were higher in 2015. The correlation analysis found no relationship between the N loads and discharge under PD signifying that soil physical and chemical properties affect the movement of N. The P loads were greater under both water management systems in 2015, compared to 2016.
The CD annual rate of P release found a decrease of 35% between the years. The P loads under
PD were 5 times higher in 2015, compared to 2016, which was caused by the reduction in discharge volume by nearly the same percentage. The correlation between TP loads and discharge was found in both DWM systems, signifying the importance of discharge to P management.
The findings of this thesis confirmed the necessity of drainage water management practices on agricultural organic soils to reduce the nutrient load outflow from the tile drain discharge.
Although the current PD practices that are used by farmers is effective in reducing the N loads from tile drainage discharge. Additionally, this study documented the potential use of ANN to predict the most accurate water management protocol while using a CD structure. Finally, the thesis increased our understanding of the impacts that various P parameters have on the drainage water quality outflow, allowing for an increased knowledge base for decision making.
121 6.2 Contributions to knowledge
As a result of the research conducted, the following are the contributions to knowledge:
1. This thesis demonstrated for the first time that controlled drainage (CD) in the form of a
gated water table control structure can be used to manage drainage water and reduce
nutrient loads in Histosols. Growers and policymakers can use this management strategy
to mitigate nutrient loss from subsurface drainage, reducing environmental degradation.
2. The artificial neural network (ANN) model was used for the first time to predict
phosphorus loads and forecast future drainage scenarios in organic soils. The biophysical
phosphorus models are not calibrated for organic soils. This major gap in knowledge can
effectively be addressed through ANN as demonstrated by this thesis.
3. This thesis proposed a novel and holistic approach which integrates the soil phosphorus
(P) dynamics and the drainage P concentrations to increase understanding into the soil-
water continuum. Past studies have not assessed the soil phosphorus dynamics in relation
to drainage water quality. This approach provides policymakers with a more reliable
method for strategic decision making to reduce P concentrations in Histosols.
122 6.3 Recommendations for future research
1. Although we identified ANN as a possible prediction model data from multiple years of study
are needed to increase the accuracy of the seasonal models (winter-spring and summer), further
assessment in the potential predictive accuracy is needed through a comparison between field
and model data. The assessment of increased input variables would also increase the
applicability of ANN models for use in agricultural systems.
2. Future study on PD and CD should be a long-term study with multiple PD locations to assess
the various management practices of farmers and their decision-making policies for activating
the drainage system. Therefore, this will clarify the effects of the two systems on nutrient loads,
crop yield and soil hydrological movement.
3. The focus of this study was phosphorus water quality and drainage water management;
however, a two-year study is extremely limited. Conducting a study with various fertilizer
rates, water management strategies and different crop types while analyzing the soil-water
phosphorus dynamics will fill the gap of knowledge in P movement from soil to drainage water.
4. Future work should also be on the effects of new technology on the assessment of phosphorus
geochemistry in organic soils through the use of isotope tracing techniques (32P/33P and oxygen
isotope ratios in phosphates), X-ray absorption near-edge spectroscopy, and 31P nuclear
magnetic resonance, in conjunction with sequential fractionation and water quality analysis
5. Future work should include the creation of phosphorus transport indices to predict nutrient loss
from organic soils specifically. Histosols are an important part of future food security and
therefore these soils must be environmentally sustainable.
123 References
Adeuya, R., Utt, N., Frankenberger, J., Bowling, L., Kladivko, E., Brouder, S., & Carter, B. (2012).
Impacts of drainage water management on subsurface drain flow, nitrate concentration, and
nitrate loads in Indiana. Journal of Soil and Water Conservation, 67(6), 474–484.
https://doi.org/10.2489/jswc.67.6.474
Aillery, M., Shoemaker, R., & Caswell, M. (2001). Agriculture and Ecosystem Restoration in
South Florida: Assessing Trade‐Offs from Water‐Retention Development in the Everglades
Agricultural Area. American Journal of Agricultural Economics, 83(1), 183–195.
https://doi.org/10.1111/0002-9092.00146
Ajmera, T. K., & Goyal, M. K. (2012). Development of stage–discharge rating curve using model
tree and neural networks: An application to Peachtree Creek in Atlanta. Expert Systems with
Applications, 39(5), 5702–5710. https://doi.org/10.1016/j.eswa.2011.11.101
Al-Mahallawi, K., Mania, J., Hani, A., & Shahrour, I. (2012). Using of neural networks for the
prediction of nitrate groundwater contamination in rural and agricultural areas.
Environmental Earth Sciences, 65(3), 917–928. https://doi.org/10.1007/s12665-011-1134-5
Algoazany, A. S., Kalita, P. K., Czapar, G. F., & Mitchell, J. K. (2007). Phosphorus Transport
through Subsurface Drainage and Surface Runoff from a Flat Watershed in East Central
Illinois, USA. Journal of Environmental Quality, 36(3), 681–693.
https://doi.org/10.2134/jeq2006.0161
Amador, J., & Jones, R. D. (1993). Nutrient limitations on microbial respiration in peat soils with
different total phosphorus content. Soil Biology and Biochemistry, 25(6), 793–801.
https://doi.org/10.1016/0038-0717(93)90125-U
124 Annaheim, K. E., Doolette, A. L., Smernik, R. J., Mayer, J., Oberson, A., Frossard, E., &
Bünemann, E. K. (2015). Long-term addition of organic fertilizers has little effect on soil
organic phosphorus as characterized by 31P NMR spectroscopy and enzyme additions.
Geoderma, 257–258, 67–77. https://doi.org/10.1016/j.geoderma.2015.01.014
Antonopoulos, V. Z., Gianniou, S. K., & Antonopoulos, A. V. (2016). Artificial neural networks
and empirical equations to estimate daily evaporation: application to Lake Vegoritis, Greece.
Hydrological Sciences Journal, 61(14), 2590–2599.
https://doi.org/10.1080/02626667.2016.1142667
Asselin, M. (1997). Computer model for rational use of fertilizers and the correction of nutrient
imbalances in vegetable crops on organic soils (In French). Canada-Quebec Agreement.
Sustainable Environmental Agriculture Report, 13(67130811), 046.
Audette, Y., O’Halloran, I. P., Nowell, P. M., Dyer, R., Kelly, R., & Voroney, R. P. (2018).
Speciation of Phosphorus from Agricultural Muck Soils to Stream and Lake Sediments.
Journal of Environmental Quality, 47(4), 884–892. https://doi.org/10.2134/jeq2018.02.0068
Ayars, J. E., Christen, E. W., & Hornbuckle, J. W. (2006). Controlled drainage for improved water
management in arid regions irrigated agriculture. Agricultural Water Management, 86(1-2),
128-139. https://doi.org/10.1016/j.agwat.2006.07.004
Bauke, S. L., von Sperber, C., Tamburini, F., Gocke, M. I., Honermeier, B., Schweitzer, K., …
Amelung, W. (2018). Subsoil phosphorus is affected by fertilization regime in long-term
agricultural experimental trials. European Journal of Soil Science, 69(1), 103–112.
https://doi.org/10.1111/ejss.12516
Baum, C., Leinweber, P., & Schlichting, A. (2003). Effects of chemical conditions in re-wetted
peats on temporal variation in microbial biomass and acid phosphatase activity within the
125 growing season. Applied Soil Ecology, 22(2), 167–174. https://doi.org/10.1016/S0929-
1393(02)00129-4
Beauchemin, S., Simard, R. R., & Cluis, D. (1998). Forms and Concentration of Phosphorus in
Drainage Water of Twenty-Seven Tile-Drained Soils. Journal of Environmental Quality,
27(3), 721-728. https://doi.org/10.2134/jeq1998.00472425002700030033x
Bechtold, M., Dettmann, U., Wöhl, L., Durner, W., Piayda, A., & Tiemeyer, B. (2018). Comparing
Methods for Measuring Water Retention of Peat Near Permanent Wilting Point. Soil Science
Society of America Journal, 82(3), 601–605. https://doi.org/10.2136/sssaj2017.10.0372
Bendel, R. B., & Afifi, A. A. (1977). Comparison of stopping rules in forward "stepwise"
regression. Journal of the American Statistical association, 72(357), 46-53.
https://doi.org/10.1080/01621459.1977.10479905
Bergström, L., Kirchmann, H., Djodjic, F., Kyllmar, K., Ulén, B., Liu, J., … Villa, A. (2015).
Turnover and Losses of Phosphorus in Swedish Agricultural Soils: Long-Term Changes,
Leaching Trends, and Mitigation Measures. Journal of Environmental Quality, 44(2), 512–
523. https://doi.org/10.2134/jeq2014.04.0165
Bhadha, J. H., Lang, T. A., & Daroub, S. H. (2017). Influence of suspended particulates on
phosphorus loading exported from farm drainage during a storm event in the Everglades
Agricultural Area. Journal of Soils and Sediments, 17(1), 240-252.
https://doi.org/10.1007/s11368-016-1548-5
Blann, K. L., Anderson, J. L., Sands, G. R., & Vondracek, B. (2009). Effects of Agricultural
Drainage on Aquatic Ecosystems: A Review. Critical Reviews in Environmental Science and
Technology, 39(11), 909–1001. https://doi.org/10.1080/10643380801977966
126 Bowden, G. J., Dandy, G. C., & Maier, H. R. (2005). Input determination for neural network
models in water resources applications. Part 1—background and methodology. Journal of
Hydrology, 301(1–4), 75–92. https://doi.org/10.1016/j.jhydrol.2004.06.021
Bray, R. H., & Kurtz, L. T. (1945). Determination of Total, Organic and Available Forms of
Phosphorus in Soils. Soil Science, 59(1), 39–46. https://doi.org/10.1097/00010694-
194501000-00006
Bridgham, S. D., Megonigal, J. P., Keller, J. K., Bliss, N. B., & Trettin, C. (2006). The carbon
balance of North American wetlands. Wetlands. https://doi.org/10.1672/0277-
5212(2006)26[889:TCBONA]2.0.CO;2
Bristow, C. S., Hudson-Edwards, K. A., & Chappell, A. (2010). Fertilizing the Amazon and
equatorial Atlantic with West African dust. Geophysical Research Letters, 37(14), n/a-n/a.
https://doi.org/10.1029/2010GL043486
Brookes, P. C., Landman, A., Pruden, G., & Jenkinson, D. S. (1985). Chloroform fumigation and
the release of soil nitrogen: A rapid direct extraction method to measure microbial biomass
nitrogen in soil. Soil Biology and Biochemistry, 17(6), 837–842.
https://doi.org/10.1016/0038-0717(85)90144-0
Browne, F. S. (1950). Organic soil management for vegetables. Publications. Department of
Agriculture, Canada, 853.
Bünemann, E. K., Prusisz, B., & Ehlers, K. (2011). Characterization of Phosphorus Forms in Soil
Microorganisms. https://doi.org/10.1007/978-3-642-15271-9_2
Cameron, K. C., Di, H. J., & Moir, J. L. (2013). Nitrogen losses from the soil/plant system: a
review. Annals of Applied Biology, 162(2), 145–173. https://doi.org/10.1111/aab.12014
127 Capone, L. T., Izuno, F. T., Bottcher, A. B., Sanchez, C. A., Coale, F. J., & Jones, D. B. (1995).
Nitrogen Concentrations in Agricultural Drainage Water in South Florida. Transactions of
the ASAE, 38(4), 1089–1098. https://doi.org/10.13031/2013.27926
Caron, J., Price, J. S., & Rochefort, L. (2015). Physical Properties of Organic Soil: Adapting
Mineral Soil Concepts to Horticultural Growing Media and Histosol Characterization.
Vadose Zone Journal, 14(6), vzj2014.10.0146. https://doi.org/10.2136/vzj2014.10.0146
Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., & Smith, V. H.
(1998). Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological
Applications. https://doi.org/10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2
Carpenter, S. R. (2008). Phosphorus control is critical to mitigating eutrophication. Proceedings
of the National Academy of Sciences, 105(32), 11039–11040.
https://doi.org/10.1073/pnas.0806112105
Carstensen, M., Poulsen, J., Ovesen, N., Børgesen, C., Hvid, S., & Kronvang, B. (2016). Can
controlled drainage control agricultural nutrient emissions? Evidence from a BACI
experiment combined with a dual isotope approach. Hydrology and Earth System Sciences
Discussions. https://doi.org/10.5194/hess-2016-303
Castillo, M. S., & Wright, A. L. (2008). Soil phosphorus pools for Histosols under sugarcane and
pasture in the Everglades, USA. Geoderma, 145(1–2), 130–135.
https://doi.org/10.1016/j.geoderma.2008.03.006
CCME. (2012). Canadian Environmental Quality Guidelines (CEQG) Summary Table: Nitrate.
Canadian Council of the Ministers of the Environment. Retrieved from: http://st-
ts.ccme.ca/en/index.html?chems=140&chapters=all.
128 Chadwick, O. A., Derry, L. A., Vitousek, P. M., Huebert, B. J., & Hedin, L. O. (1999). Changing
sources of nutrients during four million years of ecosystem development. Nature, 397(6719),
491–497. https://doi.org/10.1038/17276
Chambers, P. A., Culp, J. M., Roberts, E. S., & Bowerman, M. (2012). Development of
Environmental Thresholds for Streams in Agricultural Watersheds. Journal of Environmental
Quality, 41(1), 1–6. https://doi.org/10.2134/jeq2011.0338
Chang, F.-J., Chen, P.-A., Chang, L.-C., & Tsai, Y.-H. (2016). Estimating spatio-temporal
dynamics of stream total phosphate concentration by soft computing techniques. Science of
The Total Environment, 562, 228–236. https://doi.org/10.1016/j.scitotenv.2016.03.219
Chau, K. (2006). A review on integration of artificial intelligence into water quality modelling.
Mar. Pollut. Bull. 52, 726–733. https://doi.org/10.1016/j.marpolbul. 2006.04.003.
Chen, C. ., Condron, L. ., Davis, M. ., & Sherlock, R. . (2003). Seasonal changes in soil phosphorus
and associated microbial properties under adjacent grassland and forest in New Zealand.
Forest Ecology and Management, 177(1–3), 539–557. https://doi.org/10.1016/S0378-
1127(02)00450-4
Chen, D., Lu, J., & Shen, Y. (2009). Artificial neural network modelling of concentrations of
nitrogen, phosphorus and dissolved oxygen in a non-point source polluted river in Zhejiang
Province, southeast China. Hydrological Processes, n/a-n/a.
https://doi.org/10.1002/hyp.7482
Chen, L., Sun, C., Wang, G., Xie, H., & Shen, Z. (2017). Modeling Multi-Event Non-Point Source
Pollution in a Data-Scarce Catchment Using ANN and Entropy Analysis. Entropy, 19(6),
265. https://doi.org/10.3390/e19060265
129 Chikhaoui, M., Madramootoo, C. A., Eastman, M., & Michaud, A. (2008). Estimating Preferential
Flow to Agricultural Tile Drains. 2008 Providence, Rhode Island, June 29 - July 2, 2008.
https://doi.org/10.13031/2013.24812
Christianson, L. E., Harmel, R. D., Smith, D., Williams, M. R., & King, K. (2016). Assessment
and Synthesis of 50 Years of Published Drainage Phosphorus Losses. Journal of
Environmental Quality, 45(5), 1467–1477. https://doi.org/10.2134/jeq2015.12.0593
Clarholm, M. (1993). Microbial biomass P, labile P, and acid phosphatase activity in the humus
layer of a spruce forest, after repeated additions of fertilizers. Biology and Fertility of Soils,
16(4), 287–292. https://doi.org/10.1007/BF00369306
Clayton, B.C., and L.A. Jones. (1941). Controlled drainage in the northern Everglades of Florida.
Agricultural Engineering, 22.
Cogger, C., & Duxbury, J. M. (1984). Factors Affecting Phosphorus Losses from Cultivated
Organic Soils. Journal of Environmental Quality, 13(1), 111–114.
https://doi.org/10.2134/jeq1984.00472425001300010020x
Cole, A. J. (1984). Nitrogen, Phosphorus, Potassium, Calcium and Copper as Plant Nutrients for
Carrots Grown on Wood Fen Peat. Irish Journal of Agricultural Research, 23(2/3), 191–199.
Retrieved from https://www.jstor.org/stable/25556091
Conley, D. J., Humborg, C., Rahm, L., Savchuk, O. P., & Wulff, F. (2002). Hypoxia in the Baltic
Sea and Basin-Scale Changes in Phosphorus Biogeochemistry. Environmental Science &
Technology, 36(24), 5315–5320. https://doi.org/10.1021/es025763w
Cooper, R. L., Fausey, N. R., & Streeter, J. G. (1991). Yield Potential of Soybean Grown under a
Subirrigation/Drainage Water Management System. Agronomy Journal, 83(5), 884.
https://doi.org/10.2134/agronj1991.00021962008300050021x
130 Cross, A. F., & Schlesinger, W. H. (1995). A literature review and evaluation of the. Hedley
fractionation: Applications to the biogeochemical cycle of soil phosphorus in natural
ecosystems. Geoderma, 64(3–4), 197–214. https://doi.org/10.1016/0016-7061(94)00023-4
Czuba, M., & Hutchinson, T. C. (1980). Copper and Lead Levels in Crops and Soils of the Holland
Marsh Area—Ontario. Journal of Environmental Quality, 9(4), 566–575.
https://doi.org/10.2134/jeq1980.00472425000900040006x
Dahamsheh, A., & Aksoy, H. (2009). Artificial neural network models for forecasting intermittent
monthly precipitation in arid regions. Meteorological Applications, 16(3), 325–337.
https://doi.org/10.1002/met.127
Daroub SH, Lang TA, Diaz OA, Chen M, Stuck JD (2005) Everglades Agricultural Area BMPs
for reducing particulate phosphorus trans- port. Final Report submitted to Florida Department
of Environmental Protection, Tallahassee, FL
Daroub, S. H., Lang, T. A., Diaz, O. A., & Grunwald, S. (2009). Long-term Water Quality Trends
after Implementing Best Management Practices in South Florida. Journal of Environmental
Quality, 38(4), 1683–1693. https://doi.org/10.2134/jeq2008.0462
Daroub, S. H., Van Horn, S., Lang, T. A., & Diaz, O. A. (2011). Best Management Practices and
Long-Term Water Quality Trends in the Everglades Agricultural Area. Critical Reviews in
Environmental Science and Technology, 41(sup1), 608–632.
https://doi.org/10.1080/10643389.2010.530905
Dayton, E. A., Whitacre, S., & Holloman, C. (2017). Comparison of three persulfate digestion
methods for total phosphorus analysis and estimation of suspended sediments. Applied
Geochemistry, 78, 357–362. https://doi.org/10.1016/j.apgeochem.2017.01.011
131 de Jonge, V. N., Elliott, M., & Orive, E. (2002). Causes, historical development, effects and future
challenges of a common environmental problem: eutrophication. In Nutrients and
Eutrophication in Estuaries and Coastal Waters (pp. 1–19). https://doi.org/10.1007/978-94-
017-2464-7_1
De Sena, A. (2017). Characterizing the organic phosphorus species in Histosols of the Holland
Marsh, Canada. M.Sc. Thesis. McGill University, Montreal.
Dekker, L. ., & Ritsema, C. J. (2000). Wetting patterns and moisture variability in water repellent
Dutch soils. Journal of Hydrology, 231–232, 148–164. https://doi.org/10.1016/S0022-
1694(00)00191-8
Delta Protection Commission. (2012). Economic sustainability plan for the Sacramento-San
Joaquin River Delta, Sacramento, CA. Delta Protection Commission.
https://www.waterboards.ca.gov/waterrights/water_issues/programs/bay_delta/california_
waterfix/exhibits/docs/CDWA%20et%20al/SDWA_137.pdf (accessed 29 Jan. 2019).
Dettmann, U., Bechtold, M., Frahm, E., & Tiemeyer, B. (2014). On the applicability of unimodal
and bimodal van Genuchten–Mualem based models to peat and other organic soils under
evaporation conditions. Journal of Hydrology, 515, 103–115.
https://doi.org/10.1016/j.jhydrol.2014.04.047
Dinnes, D. L., Karlen, D. L., Jaynes, D. B., Kaspar, T. C., Hatfield, J. L., Colvin, T. S., &
Cambardella, C. A. (2002). Nitrogen management strategies to reduce nitrate leaching in
tile-drained midwestern soils. Agronomy Journal, 94(1), 153.
https://doi.org/10.2134/agronj2002.0153
132 Dodd, R. J., & Sharpley, A. N. (2015). Recognizing the role of soil organic phosphorus in soil
fertility and water quality. Resources, Conservation and Recycling, 105, 282–293.
https://doi.org/10.1016/j.resconrec.2015.10.001
Dommain, R., Frolking, S., Jeltsch-Thömmes, A., Joos, F., Couwenberg, J., & Glaser, P. H. (2018).
A radiative forcing analysis of tropical peatlands before and after their conversion to
agricultural plantations. Global Change Biology, 24(11), 5518–5533.
https://doi.org/10.1111/gcb.14400
Doolette, A. L., & Smernik, R. J. (2011). Soil Organic Phosphorus Speciation Using Spectroscopic
Techniques. https://doi.org/10.1007/978-3-642-15271-9_1
Drury, C. F., Tan, C. S., Welacky, T. W., Oloya, T. O., Hamill, A. S., & Weaver, S. E. (1999).
Red clover and tillage influence on soil temperature, water content, and corn emergence.
Agronomy Journal, 91(1), 101-108.
https://doi.org/10.2134/agronj1999.00021962009100010016x
Duxbury, J. M., & Peverly, J. H. (1978). Nitrogen and Phosphorus Losses from Organic Soils.
Journal of Environmental Quality, 7(4), 566–570.
https://doi.org/10.2134/jeq1978.00472425000700040020x
Eastman, M., Gollamudi, A., Stämpfli, N., Madramootoo, C. A., & Sarangi, A. (2010).
Comparative evaluation of phosphorus losses from subsurface and naturally drained
agricultural fields in the Pike River watershed of Quebec, Canada. Agricultural Water
Management, 97(5), 596–604. https://doi.org/10.1016/j.agwat.2009.11.010
Easton, Z. M., & Petrovic, A. M. (2004). Fertilizer Source Effect on Ground and Surface Water
Quality in Drainage from Turfgrass. Journal of Environmental Quality, 33(2), 645–655.
https://doi.org/10.2134/jeq2004.6450
133 Ebina, J., Tsutsui, T., & Shirai, T. (1983). Simultaneous determination of total nitrogen and total
phosphorus in water using peroxodisulfate oxidation. Water Research, 17(12), 1721–1726.
https://doi.org/10.1016/0043-1354(83)90192-6
El-Din, A. G., & Smith, D. W. (2002). A neural network model to predict the wastewater inflow
incorporating rainfall events. Water Research, 36(5), 1115–1126.
https://doi.org/10.1016/S0043-1354(01)00287-1
Elmi, A. A., Madramootoo, C., & Hamel, C. (2000). Influence of water table and nitrogen
− management on residual soil NO3 and denitrification rate under corn production in sandy
loam soil in Quebec. Agriculture, Ecosystems & Environment, 79(2–3), 187–197.
https://doi.org/10.1016/S0167-8809(99)00157-7
Elmi, A. A., Madramootoo, C., Egeh, M., Dodds, G., & Hamel, C. (2002). Water Table
Management as a Natural Bioremediation Technique of Nitrate Pollution. Water Quality
Research Journal, 37(3), 563–576. https://doi.org/10.2166/wqrj.2002.037
Elser, J. J., Bracken, M. E. S., Cleland, E. E., Gruner, D. S., Harpole, W. S., Hillebrand, H., …
Smith, J. E. (2007). Global analysis of nitrogen and phosphorus limitation of primary
producers in freshwater, marine and terrestrial ecosystems. Ecology Letters, 10(12), 1135–
1142. https://doi.org/10.1111/j.1461-0248.2007.01113.x
Ewing, J. M., Vepraskas, M. J., Broome, S. W., & White, J. G. (2012). Changes in wetland soil
morphological and chemical properties after 15, 20, and 30years of agricultural production.
Geoderma, 179–180, 73–80. https://doi.org/10.1016/j.geoderma.2012.02.018
Fausett, L. (1994). Fundamentals Of Neural Network Architectures, Algorithms, and Applications.
Inc.,New Jersey.
134 Fell, H., Roßkopf, N., Bauriegel, A., & Zeitz, J. (2016). Estimating vulnerability of agriculturally
used peatlands in north-east Germany to carbon loss based on multi-temporal subsidence data
analysis. CATENA, 137, 61–69. https://doi.org/10.1016/j.catena.2015.08.010
Filippelli, G. M. (2008). The Global Phosphorus Cycle: Past, Present, and Future. Elements, 4(2),
89–95. https://doi.org/10.2113/GSELEMENTS.4.2.89
Filippelli, G.M. (2017). The global phosphorus cycle. In: R. Lal, B.A. Stewart, editors, Soil
phosphorus. CRC Press, Boca Raton. p. 1–21.
Frossard, E., Condron, L. M., Oberson, A., Sinaj, S., & Fardeau, J. C. (2000). Processes Governing
Phosphorus Availability in Temperate Soils. Journal of Environmental Quality, 29(1), 15–
23. https://doi.org/10.2134/jeq2000.00472425002900010003x
Gabriel, M., Toader, C., Faul, F., Roßkopf, N., Grundling, P., Huyssteen, C., … Zeitz, J. (2018).
Physical and hydrological properties of peat as proxies for degradation of south african
peatlands: Implications for conservation and restoration. Mires and Peat.
https://doi.org/10.19189/MaP.2018.OMB.336
Gambolati, G., Putti, M., Teatini, P., & Gasparetto Stori, G. (2006). Subsidence due to peat
oxidation and impact on drainage infrastructures in a farmland catchment south of the Venice
Lagoon. Environmental Geology, 49(6), 814–820. https://doi.org/10.1007/s00254-006-0176-
6
Garcia R.M. (2000) Water tables and drainage uniformity in the Everglades Agricultural Area.
Master’s Thesis. Florida Atlantic University, Boca Raton, FL
Gentry, L. E., David, M. B., Royer, T. V., Mitchell, C. A., & Starks, K. M. (2007). Phosphorus
Transport Pathways to Streams in Tile-Drained Agricultural Watersheds. Journal of
Environmental Quality, 36(2), 408–415. https://doi.org/10.2134/jeq2006.0098
135 George, T. S., Fransson, A.-M., Hammond, J. P., & White, P. J. (2011). Phosphorus Nutrition:
Rhizosphere Processes, Plant Response and Adaptations. https://doi.org/10.1007/978-3-642-
15271-9_10
Gharedaghloo, B., Price, J. S., Rezanezhad, F., & Quinton, W. L. (2018). Evaluating the hydraulic
and transport properties of peat soil using pore network modeling and X-ray micro computed
tomography. Journal of Hydrology, 561, 494–508.
https://doi.org/10.1016/j.jhydrol.2018.04.007
Gilliam, J. W., Skaggs, R. W., & Weed, S. B. (1979). Drainage Control to Diminish Nitrate Loss
from Agricultural Fields. Journal of Environmental Quality, 8(1), 137–142.
https://doi.org/10.2134/jeq1979.00472425000800010030x
Government of Canada. (2018). Daily Data Report for Toronto International Airport, Ontario.
Historical Data. [July, 2018].
http://climate.weather.gc.ca/historical_data/search_historic_data_e.html.
Graham, S. A., Craft, C. B., McCormick, P. V., & Aldous, A. (2005). Forms and accumulation of
soil P in natural and recently restored peatlands - Upper Klamath Lake, Oregon, USA.
Wetlands. https://doi.org/10.1672/0277-5212(2005)025[0594:FAAOSP]2.0.CO;2
Gramlich, A., Stoll, S., Stamm, C., Walter, T., & Prasuhn, V. (2018). Effects of artificial land
drainage on hydrology, nutrient and pesticide fluxes from agricultural fields – A review.
Agriculture, Ecosystems & Environment, 266, 84–99.
https://doi.org/10.1016/j.agee.2018.04.005
Grierson, P. F., Comerford, N. B., & Jokela, E. J. (1998). Phosphorus mineralization kinetics and
response of microbial phosphorus to drying and rewetting in a Florida Spodosol. Soil Biology
and Biochemistry, 30(10–11), 1323–1331. https://doi.org/10.1016/S0038-0717(98)00002-9
136 Grozav, A., & Rogobete, G. (2012). Drainage of histosols. Research Journal of Agricultural
Science, 44(3), 49–54.
Gu, C., Dam, T., Hart, S. C., Turner, B. L., Chadwick, O. A., Berhe, A. A., … Zhu, M. (2020).
Quantifying Uncertainties in Sequential Chemical Extraction of Soil Phosphorus Using
XANES Spectroscopy. Environmental Science & Technology, 54(4), 2257–2267.
https://doi.org/10.1021/acs.est.9b05278
Guérin, J., Parent, L.-É., & Abdelhafid, R. (2007). Agri-environmental Thresholds using Mehlich
III Soil Phosphorus Saturation Index for Vegetables in Histosols. Journal of Environmental
Quality, 36(4), 975–982. https://doi.org/10.2134/jeq2006.0424
Guérin, J. É., Parent, L. É., & Si, B. C. (2011). Spatial and seasonal variability of phosphorus risk
indexes in cultivated organic soils. Canadian Journal of Soil Science, 91(2), 291-302.
https://doi.org/10.4141/cjss10089
Gunn, K. M., Fausey, N. R., Shang, Y., Shedekar, V. S., Ghane, E., Wahl, M. D., & Brown, L. C.
(2015). Subsurface drainage volume reduction with drainage water management: Case
studies in Ohio, USA. Agricultural Water Management, 149, 131–142.
https://doi.org/10.1016/j.agwat.2014.10.014
Guppy, C. N., Menzies, N. W., Moody, P. W., & Blamey, F. P. C. (2005). Competitive sorption
reactions between phosphorus and organic matter in soil: a review. Soil Research, 43(2), 189.
https://doi.org/10.1071/SR04049
Ha, H., & Stenstrom, M. K. (2003). Identification of land use with water quality data in stormwater
using a neural network. Water Research, 37(17), 4222–4230. https://doi.org/10.1016/S0043-
1354(03)00344-0
137 Hagerty, S. B., van Groenigen, K. J., Allison, S. D., Hungate, B. A., Schwartz, E., Koch, G. W.,
… Dijkstra, P. (2014). Accelerated microbial turnover but constant growth efficiency with
warming in soil. Nature Climate Change, 4(10), 903–906.
https://doi.org/10.1038/nclimate2361
Hallema, D. W., Périard, Y., Lafond, J. A., Gumiere, S. J., & Caron, J. (2015). Characterization of
Water Retention Curves for a Series of Cultivated Histosols. Vadose Zone Journal, 14(6),
vzj2014.10.0148. https://doi.org/10.2136/vzj2014.10.0148
Hamilton, H. A., & Bernier, R. (1975). N–P–K Fertilizer Effects on Yield, Composition and
Residues of Lettuce, Celery, Carrot and Onion Grown on an Organic Soil in Quebec.
Canadian Journal of Plant Science, 55(2), 453–461. https://doi.org/10.4141/cjps75-071
Hansen, J. C., Cade-Menun, B. J., & Strawn, D. G. (2004). Phosphorus Speciation in Manure-
Amended Alkaline Soils. Journal of Environment Quality, 33(4), 1521.
https://doi.org/10.2134/jeq2004.1521
Hassan, H. M., Marschner, P., McNeill, A., & Tang, C. (2012). Growth, P uptake in grain legumes
and changes in rhizosphere soil P pools. Biology and Fertility of Soils.
https://doi.org/10.1007/s00374-011-0612-y
Havens, K. E., & James, R. T. (2005). The phosphorus mass balance of lake okeechobee, Florida:
Implications for eutrophication management. Lake and Reservoir Management.
https://doi.org/10.1080/07438140509354423
Haygarth, P. M., & Jarvis, S. C. (1999). Transfer of Phosphorus from Agricultural Soil. In
Advances in Agronomy (pp. 195–249). https://doi.org/10.1016/S0065-2113(08)60428-9
138 He, B., Oki, T., Sun, F., Komori, D., Kanae, S., Wang, Y., … Yamazaki, D. (2011). Estimating
monthly total nitrogen concentration in streams by using artificial neural network. Journal of
Environmental Management, 92(1), 172–177. https://doi.org/10.1016/j.jenvman.2010.09.014
He, Z., Honeycutt, C. W., Cade-Menun, B. J., Senwo, Z. N., & Tazisong, I. A. (2008). Phosphorus
in Poultry Litter and Soil: Enzymatic and Nuclear Magnetic Resonance Characterization. Soil
Science Society of America Journal, 72(5), 1425–1433.
https://doi.org/10.2136/sssaj2007.0407
Hedley, M. J., Stewart, J. W. B., & Chauhan, B. S. (1982). Changes in Inorganic and Organic Soil
Phosphorus Fractions Induced by Cultivation Practices and by Laboratory Incubations. Soil
Science Society of America Journal, 46(5), 970–976.
https://doi.org/10.2136/sssaj1982.03615995004600050017x
Helmers, M., Christianson, R., Brenneman, G., Lockett, D., & Pederson, C. (2012). Water table,
drainage, and yield response to drainage water management in southeast Iowa. Journal of Soil
and Water Conservation, 67(6), 495-501. https://doi.org/10.2489/jswc.67.6.495
Hemond, H. F., & Goldman, J. C. (1985). On Non-Darcian Water Flow in Peat. The Journal of
Ecology, 73(2), 579. https://doi.org/10.2307/2260495
Heuck, C., Weig, A., & Spohn, M. (2015). Soil microbial biomass C: N: P stoichiometry and
microbial use of organic phosphorus. Soil Biology and Biochemistry, 85, 119–129.
https://doi.org/10.1016/j.soilbio.2015.02.029
Hill, B. H., Elonen, C. M., Jicha, T. M., Kolka, R. K., Lehto, L. L. P., Sebestyen, S. D., & Seifert-
Monson, L. R. (2014). Ecoenzymatic stoichiometry and microbial processing of organic
matter in northern bogs and fens reveals a common P-limitation between peatland types.
Biogeochemistry, 120(1–3), 203–224. https://doi.org/10.1007/s10533-014-9991-0
139 Hinsinger, P. (2001). Bioavailability of soil inorganic P in the rhizosphere as affected by root-
induced chemical changes: a review. Plant and Soil, 237, 173–195.
https://doi.org/10.1023/A:1013351617532
Hoepting, C. (2009). Elba muck soil nutrient survey summary. Part II of III: Phosphorus,
potassium and nitrogen. Cornell Cooperative Extension Vegetable Program, Cornell
University, Ithaca, NY. https://rvpadmin.cce.cornell.edu/uploads/doc_193.pdf.
Hoffman, D.W., Wicklund, R.E., & Richards, N.R. (1962). Soil survey of Simcoe County, Ontario.
Report No. 29 of the Ontario Soil Survey. Research Branch, Canada Department of
Agriculture and the Ontario Agricultural College. (accessed 30 Jan. 2019).
Hornik, K., Stinchcombe, M., & White, H. (1989). Multilayer feedforward networks are universal
approximators. Neural Networks, 2(5), 359–366. https://doi.org/10.1016/0893-
6080(89)90020-8
Hsu, K., Gupta, H. V., & Sorooshian, S. (1995). Artificial Neural Network Modeling of the
Rainfall-Runoff Process. Water Resources Research, 31(10), 2517–2530.
https://doi.org/10.1029/95WR01955
Huat, B. B. K., Asadi, A., & Kazemian, S. (2009). Experimental Investigation on Geomechanical
Properties of Tropical Organic Soils and Peat. American Journal of Engineering and Applied
Sciences, 2(1), 184–188. https://doi.org/10.3844/ajeas.2009.184.188
Ilnicki, P. (2003). Agricultural Production Systems for Organic Soil Conservation. In Organic
Soils and Peat Materials for Sustainable Agriculture (pp. 187–200).
https://doi.org/10.1201/9781420040098-10
140 Ivanoff, D. B., Reddy, K. R., & Robinson, S. (1998). Chemical Fractionation of Organic
Phosphorus in Selected Histosols 1. Soil Science, 163(1), 36–45.
https://doi.org/10.1097/00010694-199801000-00006
Izuno, F. T., Sanchez, C. A., Coale, F. J., Bottcher, A. B., & Jones, D. B. (1991). Phosphorus
Concentrations in Drainage Water in the Everglades Agricultural Area. Journal of
Environmental Quality, 20(3), 608–619.
https://doi.org/10.2134/jeq1991.00472425002000030018x
Izuno, F. T., Bottcher, A. B., Coale, F. J., Sanchez, C. A., & Jones, D. B. (1995). Agricultural
BMPs for Phosphorus Reduction in South Florida. Transactions of the ASAE, 38(3), 735–
744. https://doi.org/10.13031/2013.27887
Jamieson, A., Madramootoo, C. A., & Enright, P. (2003). Phosphorus losses in surface and
subsurface runoff from a snowmelt event on an agricultural field in Quebec. Canadian
Biosystems Engineering / Le Genie Des Biosystems Au Canada.
Janardhanan, L., & Daroub, S. H. (2010). Phosphorus Sorption in Organic Soils in South Florida.
Soil Science Society of America Journal, 74(5), 1597–1606.
https://doi.org/10.2136/sssaj2009.0137
Jansa, J., Finlay, R., Wallander, H., Smith, F. A., & Smith, S. E. (2011). Role of Mycorrhizal
Symbioses in Phosphorus Cycling. https://doi.org/10.1007/978-3-642-15271-9_6
Jarosch, K. A., Doolette, A. L., Smernik, R. J., Tamburini, F., Frossard, E., & Bünemann, E. K.
(2015). Characterisation of soil organic phosphorus in NaOH-EDTA extracts: A comparison
of 31P NMR spectroscopy and enzyme addition assays. Soil Biology and Biochemistry, 91,
298–309. https://doi.org/10.1016/j.soilbio.2015.09.010
141 Jauhiainen, J., Kerojoki, O., Silvennoinen, H., Limin, S., & Vasander, H. (2014). Heterotrophic
respiration in drained tropical peat is greatly affected by temperature—a passive ecosystem
cooling experiment. Environmental Research Letters, 9(10), 105013.
https://doi.org/10.1088/1748-9326/9/10/105013
Jia, X., DeSutter, T. M., Lin, Z., Schuh, W. M., & Steele, D. D. (2012). Subsurface Drainage and
Subirrigation Effects on Water Quality in Southeast North Dakota. Transactions of the
ASABE, 55(5), 1757–1769. https://doi.org/10.13031/2013.42368
Johnston, A. E., Poulton, P. R., & Coleman, K. (2009). Chapter 1 Soil Organic Matter. In Advances
in Agronomy (pp. 1–57). https://doi.org/10.1016/S0065-2113(08)00801-8
Jones, C. A., Cole, C. V., Sharpley, A. N., & Williams, J. R. (1984). A Simplified Soil and Plant
Phosphorus Model: I. Documentation. Soil Science Society of America Journal, 48(4), 800–
805. https://doi.org/10.2136/sssaj1984.03615995004800040020x
Jones, D. & Oburger, E. (2011). Solubilization of phosphorus by soil microorganisms. In E.
Bünemann, A. Oberson, & E. Frossard, (Eds.) Phosphorus in action (pp. 245-274). Berlin,
Heidelberg: Springer. https://doi.org/10.1007/978-3-642-15271-9_7
Jones, D.B., G.H. Snyder, J. Alvarez. (1994). Chapter 12 Rice and other aquatic crops. In A. B.
Bottcher and F. T. Izuno (Eds.), Everglades Agricultural Area (EAA): Water, soil, crop, and
environmental management. Gainesville, FL: University Press of Florida.
Joosten, H. (2009). The Global Peatland CO2 Picture: Peatland status and drainage related
emissions in all the countries of the world.
Joosten, H., Tapio-Biström, M.-L., & Tol, S. (2012). Peatlands - guidance for climate change
mitigation through conservation, rehabilitation and sustainable use. In Mitigation of Climate
Change in Agriculture (MICCA) Programme series 5.
142 Kaila, A. (1959). Retention of phosphate by peat samples. The Journal of the Scientific
Agricultural Society of Finland, 31(1), 215–224. doi:10.23986/afsci.71488
Kasimir-Klemedtsson, Å., Klemedtsson, L., Berglund, K., Martikainen, P., Silvola, J., & Oenema,
O. (1997). Greenhouse gas emissions from farmed organic soils: a review. Soil Use and
Management, 13(s4), 245–250. https://doi.org/10.1111/j.1475-2743.1997.tb00595.x
Kechavarzi, C., Dawson, Q., & Leeds-Harrison, P. B. (2010). Physical properties of low-lying
agricultural peat soils in England. Geoderma, 154(3–4), 196–202.
https://doi.org/10.1016/j.geoderma.2009.08.018
Keller, J. K., & Medvedeff, C. A. (2016). Soil organic matter. In Wetland Soils: Genesis,
Hydrology, Landscapes, and Classification: Second Edition (pp. 165–188).
Kennedy, C. D., Alverson, N., Jeranyama, P., & DeMoranville, C. (2018). Seasonal dynamics of
water and nutrient fluxes in an agricultural peatland. Hydrological Processes, 32(6), 698–
712. https://doi.org/10.1002/hyp.11436
Keshavarzi, A., Sarmadian, F., Omran, E.-S. E., & Iqbal, M. (2015). A neural network model for
estimating soil phosphorus using terrain analysis. The Egyptian Journal of Remote Sensing
and Space Science, 18(2), 127–135. https://doi.org/10.1016/j.ejrs.2015.06.004
Khalil, B., Ouarda, T. B. M. J., & St-Hilaire, A. (2011). Estimation of water quality characteristics
at ungauged sites using artificial neural networks and canonical correlation analysis. Journal
of Hydrology, 405(3–4), 277–287. https://doi.org/10.1016/j.jhydrol.2011.05.024
Kim, M., & Gilley, J. E. (2008). Artificial Neural Network estimation of soil erosion and nutrient
concentrations in runoff from land application areas. Computers and electronics in
agriculture, 64(2), 268-275. https://doi.org/10.1016/j.compag.2008.05.021.
143 King, K. W., Fausey, N. R., & Williams, M. R. (2014). Effect of subsurface drainage on
streamflow in an agricultural headwater watershed. Journal of Hydrology, 519, 438–445.
https://doi.org/10.1016/j.jhydrol.2014.07.035
King, K. W., Williams, M. R., Macrae, M. L., Fausey, N. R., Frankenberger, J., Smith, D. R., …
Brown, L. C. (2015). Phosphorus Transport in Agricultural Subsurface Drainage: A Review.
Journal of Environmental Quality, 44(2), 467–485. https://doi.org/10.2134/jeq2014.04.0163
King, M., Altdorff, D., Li, P., Galagedara, L., Holden, J., & Unc, A. (2018). Northward shift of
the agricultural climate zone under 21st-century global climate change. Scientific Reports,
8(1), 7904. https://doi.org/10.1038/s41598-018-26321-8
Kleinman, P., Sharpley, A., Buda, A., McDowell, R., & Allen, A. (2011). Soil controls of
phosphorus in runoff: Management barriers and opportunities. Canadian Journal of Soil
Science, 91(3), 329–338. https://doi.org/10.4141/cjss09106
Kleinman, P. J. A., Smith, D. R., Bolster, C. H., & Easton, Z. M. (2015). Phosphorus Fate,
Management, and Modeling in Artificially Drained Systems. Journal of Environmental
Quality, 44(2), 460–466. https://doi.org/10.2134/jeq2015.02.0090
Kolka, R., S.D. Bridgham, C.-L. Ping. (2016). Soils of peatlands: Histosols and Gelisols. In: M.J.
Vepraskas and C.B. Craft, editors, Wetland soils: Genesis, hydrology, landscapes, and
classification. CRC Press, Boca Raton. p. 277–310.
Konecny, K., Ballhorn, U., Navratil, P., Jubanski, J., Page, S. E., Tansey, K., … Siegert, F. (2016).
Variable carbon losses from recurrent fires in drained tropical peatlands. Global Change
Biology, 22(4), 1469–1480. https://doi.org/10.1111/gcb.13186
144 Koski-Vähälä, J., & Hartikainen, H. (2001). Assessment of the Risk of Phosphorus Loading Due
to Resuspended Sediment. Journal of Environmental Quality, 30(3), 960–966.
https://doi.org/10.2134/jeq2001.303960x
Kouno, K., Wu, J., & Brookes, P. . (2002). Turnover of biomass C and P in soil following
incorporation of glucose or ryegrass. Soil Biology and Biochemistry, 34(5), 617–622.
https://doi.org/10.1016/S0038-0717(01)00218-8
Krasa, J., Dostal, T., Jachymova, B., Bauer, M., & Devaty, J. (2019). Soil erosion as a source of
sediment and phosphorus in rivers and reservoirs – Watershed analyses using
WaTEM/SEDEM. Environmental Research, 171, 470–483.
https://doi.org/10.1016/j.envres.2019.01.044
Kroetsch, D. J., Geng, X., Chang, S. X., & Saurette, D. D. (2011). Organic Soils of canada: Part
1. Wetland Organic soils. Canadian Journal of Soil Science, 91(5), 807–822.
https://doi.org/10.4141/cjss10043
Kumar, A. R. S., Goyal, M. K., Ojha, C. S. P., Singh, R. D., Swamee, P. K., & Nema, R. K. (2013).
Application of ANN, Fuzzy Logic and Decision Tree Algorithms for the Development of
Reservoir Operating Rules. Water Resources Management, 27(3), 911–925.
https://doi.org/10.1007/s11269-012-0225-8
Lachat Instruments. (2003). Determination of nitrate/nitrite in surface and wastewaters by flow
injection analysis. QuickChem Method, 10-107.
https://www.uvm.edu/bwrl/lab_docs/protocols/Nitrate_water_lachat.pdf
Lafond, J. A., Bergeron Piette, É., Caron, J., & Théroux Rancourt, G. (2014). Evaluating fluxes in
Histosols for water management in lettuce: A comparison of mass balance, evapotranspiration
145 and lysimeter methods. Agricultural Water Management, 135, 73–83.
https://doi.org/10.1016/j.agwat.2013.12.016
Laine, J., Minkkinen, K., & Trettin, C. (2013). Direct Human Impacts on the Peatland Carbon
Sink. In Carbon Cycling in Northern Peatlands (pp. 71–78).
https://doi.org/10.1029/2008GM000808
Lalonde, V., Madramootoo, C. A., Trenholm, L., & Broughton, R. S. (1996). Effects of controlled
drainage on nitrate concentrations in subsurface drain discharge. Agricultural Water
Management. https://doi.org/10.1016/0378-3774(95)01193-5
Lam, W. V., Macrae, M. L., English, M. C., O’Halloran, I. P., Plach, J. M., & Wang, Y. (2016).
Seasonal and event-based drivers of runoff and phosphorus export through agricultural tile
drains under sandy loam soil in a cool temperate region. Hydrological Processes, 30(15),
2644–2656. https://doi.org/10.1002/hyp.10871
Lang, T. A., Oladeji, O., Josan, M., & Daroub, S. (2010). Environmental and management factors
that influence drainage water P loads from Everglades Agricultural Area farms of South
Florida. Agriculture, Ecosystems & Environment, 138(3–4), 170–180.
https://doi.org/10.1016/j.agee.2010.04.015
Lappalainen, E., Ed. (1996). Global Peat Resources. Int. Peat Soc. Geological Survey of Jyskä,
Finland, 359 pp.
Larsson, M. H., Persson, K., Ulén, B., Lindsjö, A., & Jarvis, N. J. (2007). A dual porosity model
to quantify phosphorus losses from macroporous soils. Ecological Modelling, 205(1–2), 123–
134. https://doi.org/10.1016/j.ecolmodel.2007.02.014
Lee, C.-M., Hamm, S.-Y., Cheong, J.-Y., Kim, K., Yoon, H., Kim, M., & Kim, J. (2020).
Contribution of nitrate-nitrogen concentration in groundwater to stream water in an
146 agricultural head watershed. Environmental Research, 184, 109313.
https://doi.org/10.1016/j.envres.2020.109313
Lek, S. (1999). Predicting stream nitrogen concentration from watershed features using neural
networks. Water Research, 33(16), 3469–3478. https://doi.org/10.1016/S0043-
1354(99)00061-5
Liator, M. I., Reichmann, O., Auerswald, K., Haim, A., & Shenker, M. (2004). The Geochemistry
of Phosphorus in Peat Soils of a Semiarid Altered Wetland. Soil Science Society of America
Journal, 68(6), 2078–2085. https://doi.org/10.2136/sssaj2004.2078
Liu, D., Yuan, Y., & Liao, S. (2009). Artificial neural network vs. nonlinear regression for gold
content estimation in pyrometallurgy. Expert Systems with Applications, 36(7), 10397–
10400. https://doi.org/10.1016/j.eswa.2009.01.038
Liu, H., Zak, D., Rezanezhad, F., & Lennartz, B. (2019). Soil degradation determines release of
nitrous oxide and dissolved organic carbon from peatlands. Environmental Research Letters,
14(9), 094009. https://doi.org/10.1088/1748-9326/ab3947
Longabucco, P., & Rafferty, M. R. (1989). Delivery of Nonpoint‐Source Phosphorus from
Cultivated Mucklands to Lake Ontario. Journal of Environmental Quality, 18(2), 157–163.
https://doi.org/10.2134/jeq1989.00472425001800020005x
Lucas, R.E. (1982). Organic soils (Histosols): formation, distribution, physical and chemical
properties and management for crop production. Michigan State University, Agricultural
Experiment Station, East Lansing.
Madramootoo, C. A. (1990). Assessing Drainage Benefits on a Heavy Clay Soil in Quebec.
Transactions of the ASAE, 33(4), 1217–1223. https://doi.org/10.13031/2013.31460
147 Maguire, R. O., & Sims, J. T. (2002). Soil Testing to Predict Phosphorus Leaching. Journal of
Environment Quality. https://doi.org/10.2134/jeq2002.1601
Maier, H. R., & Dandy, G. C. (1996). The Use of Artificial Neural Networks for the Prediction of
Water Quality Parameters. Water Resources Research, 32(4), 1013–1022.
https://doi.org/10.1029/96WR03529
Maier, H. R., Jain, A., Dandy, G. C., & Sudheer, K. P. (2010). Methods used for the development
of neural networks for the prediction of water resource variables in river systems: Current
status and future directions. Environmental Modelling & Software, 25(8), 891–909.
https://doi.org/10.1016/j.envsoft.2010.02.003
Mailvaganam, S. (2018) Horticulture Crops, Carrots: Area, Production, Farm Value, Price and
Yield, Ontario 2016. Statistics Ontario Ministry of Agriculture, Food and Rural Affairs
(OMAFRA). http://www.omafra.gov.on.ca/english/stats/hort/index.html
Martin, H. W., Ivanoff, D. B., Graetz, D. A., & Reddy, K. R. (1997). Water Table Effects on
Histosol Drainage Water Carbon, Nitrogen, and Phosphorus. Journal of Environmental
Quality. https://doi.org/10.2134/jeq1997.00472425002600040018x
Mbonimpa, E. G., Yuan, Y., Nash, M. S., & Mehaffey, M. H. (2014). Sediment and total
phosphorous contributors in Rock River watershed. Journal of Environmental Management.
https://doi.org/10.1016/j.jenvman.2013.11.030
McCann, L., & Easter, K. W. (1999). Transaction Costs of Policies to Reduce Agricultural
Phosphorous Pollution in the Minnesota River. Land Economics, 75(3), 402.
https://doi.org/10.2307/3147186
McCarter, C. P. R., Rezanezhad, F., Quinton, W. L., Gharedaghloo, B., Lennartz, B., Price, J., …
Van Cappellen, P. (2020). Pore-scale controls on hydrological and geochemical processes in
148 peat: Implications on interacting processes. Earth-Science Reviews.
https://doi.org/10.1016/j.earscirev.2020.103227
Mccormick, P. V., & O’Dell, M. B. (1996). Quantifying periphyton responses to phosphorus in
the Florida Everglades: A synoptic-experimental approach. Journal of the North American
Benthological Society. https://doi.org/10.2307/1467798
McCormick, P. V., & Stevenson, R. J. (1998). Periphyton as a Tool for Ecological Assessment
and Management in the Florida Everglades. Journal of Phycology, 34(5), 726–733.
https://doi.org/10.1046/j.1529-8817.1998.340726.x
McCray, J. M., Wright, A. L., Luo, Y., & Ji, S. (2012). Soil Phosphorus Forms Related to
Extractable Phosphorus in the Everglades Agricultural Area. Soil Science, 177(1), 31–38.
https://doi.org/10.1097/SS.0b013e31823782da
McDonald M.R., & Chaput, J. (2010). Management of Organic Soils. OMAFRA. [November 10,
2019]. http://www.omafra.gov.on.ca/english/crops/facts/93-053.htm.
McDonald, M.R., Vander Kooi, K., Kessel, C., & Nemeth. D. (2013). Evaluation of phosphorus
requirements on organic (muck) soil in carrots, 2013. In: M.R. McDonald et al., editors, Muck
vegetable cultivar trial & research report 2013. University of Guelph Muck Crops Research
Station. Report No. 63. p. 52–54.
McDonald, M.R., Vander Kooi, K., Kessel, C., & Nemeth. D. (2014). Evaluation of phosphorus
requirements on organic (muck) soil on onions, 2014. In: M.R. McDonald et al., editors,
Muck vegetable cultivar trial & research report 2014. University of Guelph Muck Crops
Research Station. Report No. 64. p. 100–101.
McDowell, R. W., Sharpley, A. N., Condron, L. M., Haygarth, P. M., & Brookes, P. C. (2001).
Processes controlling soil phosphorus release to runoff and implications for agricultural
149 management. Nutrient Cycling in Agroecosystems.
https://doi.org/10.1023/A:1014419206761
McDowell, R. W., & Monaghan, R. M. (2015). Extreme Phosphorus Losses in Drainage from
Grazed Dairy Pastures on Marginal Land. Journal of Environmental Quality.
https://doi.org/10.2134/jeq2014.04.0160
Mejia, M. N., & Madramootoo, C. A. (1998). Improved Water Quality through Water Table
Management in Eastern Canada. Journal of Irrigation and Drainage Engineering, 124(2),
116–122. https://doi.org/10.1061/(ASCE)0733-9437(1998)124:2(116)
Mejia, M. N., Madramootoo, C. A., & Broughton, R. S. (2000). Influence of water table
management on corn and soybean yields. Agricultural Water Management, 46(1), 73–89.
https://doi.org/10.1016/S0378-3774(99)00109-2
Miles, J. J., Eimers, M. C., North, R. L., & Dillon, P. J. (2013). Spatial distribution and temporal
variability in the forms of phosphorus in the Beaver River subwatershed of Lake Simcoe,
Ontario, Canada. Inland Waters. https://doi.org/10.5268/IW-3.2.531
Miller, M. H. (1979). Contribution of Nitrogen and Phosphorus to Subsurface Drainage Water
from Intensively Cropped Mineral and Organic Soils in Ontario. Journal of Environmental
Quality, 8(1), 42–48. https://doi.org/10.2134/jeq1979.00472425000800010011x
Millette, J.A., B. Vigier, and R.S. Broughton. 1982. An evaluation of the drainage and subsidence
of some organic soils in Quebec. Can. Agric. Eng. 24:5–10.
Minasny, B., & McBratney, A. B. (2006). Chapter 12 Latin Hypercube Sampling as a Tool for
Digital Soil Mapping. Developments in Soil Science. https://doi.org/10.1016/S0166-
2481(06)31012-4
150 Minasny, B., & McBratney, A. B. (2006). A conditioned Latin hypercube method for sampling in
the presence of ancillary information. Computers and Geosciences.
https://doi.org/10.1016/j.cageo.2005.12.009
Mirza, C., & Irwin, R. W. (1964). Determination of Subsidence of an Organic Soil in Southern
Ontario. Canadian Journal of Soil Science. https://doi.org/10.4141/cjss64-035
MOE. (1994). Water management: Policies, guidelines, and provincial water quality objectives of
the Ministry of environment and energy. Government of Ontario, 1-61. Ontario Ministry of
Environment and Energy. http://agrienvarchive.ca/download/water_qual_object94.pdf
Mukherjee, A., & Lal, R. (2015). Tillage effects on quality of organic and mineral soils under on-
farm conditions in Ohio. Environmental Earth Sciences, 74(2), 1815–1822.
https://doi.org/10.1007/s12665-015-4189-x
Munroe, J. (ed). (2018). Soil Fertility Handbook, Publication 611, 3rd edition. Ontario Ministry of
Agricultural Food and Rural Affairs OMAFRA.
http://www.omafra.gov.on.ca/english/crops/pub611/pub611.pdf
Murphy, J., & Riley, J. P. (1962). A modified single solution method for the determination of
phosphate in natural waters. Analytica Chimica Acta, 27, 31–36.
https://doi.org/10.1016/S0003-2670(00)88444-5
Naasz, R., Michel, J. C., & Charpentier, S. (2008). Water repellency of organic growing media
related to hysteretic water retention properties. European Journal of Soil Science.
https://doi.org/10.1111/j.1365-2389.2007.00966.x
Nash, P. R., Nelson, K. A., Motavalli, P. P., Nathan, M., & Dudenhoeffer, C. (2015). Reducing
Phosphorus Loss in Tile Water with Managed Drainage in a Claypan Soil. Journal of
Environmental Quality, 44(2), 585–593. https://doi.org/10.2134/jeq2014.04.0146
151 National Research Council of Canada. (1998). The Canadian system of soil classification, third
edition. In Agriculture and Agri-Food Canada Publication 1646.
Needelman, B. A., Kleinman, P. J. A., Strock, J. S., & Allen, A. L. (2007). Improved management
of agricultural drainage ditches for water quality protection: An overview. Journal of Soil and
Water Conservation.
Negassa, W., & Leinweber, P. (2009). How does the Hedley sequential phosphorus fractionation
reflect impacts of land use and management on soil phosphorus: A review. Journal of Plant
Nutrition and Soil Science, 172(3), 305–325. https://doi.org/10.1002/jpln.200800223
Newbold, D. J., Herbert, S., Sweeney, B. W., Kiry, P., & Alberts, S. J. (2010). Water quality
functions of a 15-year-old riparian forest buffer system. Journal of the American Water
Resources Association. https://doi.org/10.1111/j.1752-1688.2010.00421.x
Nicholls, K. H., & MacCrimmon, H. R. (1974). Nutrients in Subsurface and Runoff Waters of the
Holland Marsh, Ontario. Journal of Environmental Quality, 3(1), 31–35.
https://doi.org/10.2134/jeq1974.00472425000300010010x
Nicklow, J., Asce, F., Reed, P., Asce, M., Savic, D., Dessalegne, T., Asce, M., Harrell, L., Asce,
M., Chan-hilton, A., Asce, M., Karamouz, M., Asce, F., Minsker, B., Asce, M., Ostfeld, A.,
Asce, M., Singh, A., Asce, M., Zechman, E. & Asce, M. (2010). State of the Art for Genetic
Algorithms and Beyond in Water Resources Planning and Management 412–432.
Noe, G. B., Childers, D. L., & Jones, R. D. (2001). Phosphorus Biogeochemistry and the Impact
of Phosphorus Enrichment: Why Is the Everglades so Unique? Ecosystems, 4(7), 603–624.
https://doi.org/10.1007/s10021-001-0032-1
152 O’Driscoll, C., O’Connor, M., Asam, Z. ul Z., De Eyto, E., Rodgers, M., & Xiao, L. (2014).
Creation and functioning of a buffer zone in a blanket peat forested catchment. Ecological
Engineering. https://doi.org/10.1016/j.ecoleng.2013.10.029
Oberson, A., Friesen, D. K., Rao, I. M., Bühler, S., & Frossard, E. (2001). Phosphorus
transformations in an Oxisol under contrasting land-use systems: The role of the soil
microbial biomass. Plant and Soil, 237(2), 197–210.
https://doi.org/10.1023/A:1013301716913
Oberson, A., & Joner, E. J. (2005). Microbial turnover of phosphorus in soil. In Organic
phosphorus in the environment (pp. 133–164). https://doi.org/10.1079/9780851998220.0133
Oberson, A., Pypers, P., Bünemann, E. K., & Frossard, E. (2011). Management Impacts on
Biological Phosphorus Cycling in Cropped Soils. https://doi.org/10.1007/978-3-642-15271-
9_17
Oehl, F., Oberson, A., Probst, M., Fliessbach, A., Roth, H. R., & Frossard, E. (2001). Kinetics of
microbial phosphorus uptake in cultivated soils. Biology and Fertility of Soils.
https://doi.org/10.1007/s003740100362
Oleszczuk, R., Regina, K., Szajdak, L., Höper, H., & Maryganova. V. (2008). Impacts of
agricultural utilization of peat soils on the greenhouse gas balance. Peatlands and climate
change, pp.70-97.
Orem, W., Newman, S., Osborne, T. Z., & Reddy, K. R. (2014). Projecting Changes in Everglades
Soil Biogeochemistry for Carbon and Other Key Elements, to Possible 2060 Climate and
Hydrologic Scenarios. Environmental Management. https://doi.org/10.1007/s00267-014-
0381-0
153 Palani, S., Liong, S.-Y., & Tkalich, P. (2008). An ANN application for water quality forecasting.
Marine Pollution Bulletin, 56(9), 1586–1597.
https://doi.org/10.1016/j.marpolbul.2008.05.021
Parent, L. E., Parent, S. É., & Ziadi, N. (2014). Biogeochemistry of soil inorganic and organic
phosphorus: A compositional analysis with balances. Journal of Geochemical Exploration.
https://doi.org/10.1016/j.gexplo.2014.01.030
Parent, L. E., & Ilnicki, P. (2003). Organic soils and peat materials for sustainable agriculture. In
Organic Soils and Peat Materials for Sustainable Agriculture.
https://doi.org/10.1201/9781420040098
Parent, L. E., & Khiari, L. (2003). Nitrogen and Phosphorus Balance Indicators in Organic Soils.
In Organic Soils and Peat Materials for Sustainable Agriculture (pp. 105–136).
https://doi.org/10.1201/9781420040098-6
Parent, L.E., Sasseville, L., Ndayegamiye, A., & Karam. A. (1992). The P status of cultivated
organic soils in Québec. Proceedings of the 9th International Peat Congress, Uppsala,
Sweden. p. 400–410. 22–26 Jun.
Parkinson, J. A., & Allen, S. E. (1975). A wet oxidation procedure suitable for the determination
of nitrogen and mineral nutrients in biological material. Communications in Soil Science and
Plant Analysis, 6(1), 1–11. https://doi.org/10.1080/00103627509366539
Petrovic, A. M. (1990). The Fate of Nitrogenous Fertilizers Applied to Turfgrass. Journal of
Environmental Quality. https://doi.org/10.2134/jeq1990.00472425001900010001x
Phillips, J. M., Webb, B. W., Walling, D. E., & Leeks, G. J. L. (1999). Estimating the suspended
sediment loads of rivers in the LOIS study area using infrequent samples. Hydrological
154 Processes, 13(7), 1035–1050. https://doi.org/10.1002/(SICI)1099-
1085(199905)13:7<1035::AID-HYP788>3.0.CO;2-K
Plach, J. M., Macrae, M. L., Ali, G. A., Brunke, R. R., English, M. C., Ferguson, G., … Van
Esbroeck, C. J. (2018). Supply and Transport Limitations on Phosphorus Losses from
Agricultural Fields in the Lower Great Lakes Region, Canada. Journal of Environmental
Quality, 47(1), 96–105. https://doi.org/10.2134/jeq2017.06.0234
Poole, C. A., Skaggs, R. W., Youssef, M. A., Chescheir, G. M., & Crozier, C. R. (2018). Effect of
Drainage Water Management on Nitrate Nitrogen Loss to Tile Drains in North Carolina.
Transactions of the ASABE, 61(1), 233–244. https://doi.org/10.13031/trans.12296
Porter, P. S., & Sanchez, C. A. (1992). The Effects of Soil Properties on Phosphorus Sorption by
Everglades Histosols. Soil Science, 154(5), 387–398. https://doi.org/10.1097/00010694-
199211000-00007
Qi, H., & Qi, Z. (2016). Simulating phosphorus loss to subsurface tile drainage flow: a review.
Environmental Reviews, 25(2), 150–162. https://doi.org/10.1139/er-2016-0024
Qi, H., Qi, Z., Zhang, T. Q., Tan, C. S., & Sadhukhan, D. (2018). Modeling Phosphorus Losses
through Surface Runoff and Subsurface Drainage Using ICECREAM. Journal of
Environmental Quality, 47(2), 203–211. https://doi.org/10.2134/jeq2017.02.0063
Qualls, R. G., & Richardson, C. J. (2000). Phosphorus Enrichment Affects Litter Decomposition,
Immobilization, and Soil Microbial Phosphorus in Wetland Mesocosms. Soil Science Society
of America Journal. https://doi.org/10.2136/sssaj2000.642799x
Radcliffe, D. E., Reid, D. K., Blombäck, K., Bolster, C. H., Collick, A. S., Easton, Z. M., … Smith,
D. R. (2015). Applicability of Models to Predict Phosphorus Losses in Drained Fields: A
155 Review. Journal of Environmental Quality, 44(2), 614–628.
https://doi.org/10.2134/jeq2014.05.0220
Raghavendra, S., Deka, P.C., 2014. Support vector machine applications in the field of hydrology:
a review. Appl. Soft Comput. J. 19, 372–386. https://doi.org/10.1016/j. asoc.2014.02.002.
Reddy, K. R. (1982). Mineralization of Nitrogen in Organic Soils. Soil Science Society of America
Journal, 46(3), 561–566. https://doi.org/10.2136/sssaj1982.03615995004600030024x
Reddy, K. R., Newman, S., Osborne, T. Z., White, J. R., & Fitz, H. C. (2011). Phosphorous cycling
in the greater everglades ecosystem: Legacy phosphorous implications for management and
restoration. Critical Reviews in Environmental Science and Technology.
https://doi.org/10.1080/10643389.2010.530932
Rekolainen, S., & Posch, M. (1993). Adapting the CREAMS Model for Finnish Conditions.
Hydrology Research, 24(5), 309–322. https://doi.org/10.2166/nh.1993.10
Rezanezhad, F., Price, J. S., Quinton, W. L., Lennartz, B., Milojevic, T., & Van Cappellen, P.
(2016). Structure of peat soils and implications for water storage, flow and solute transport:
A review update for geochemists. Chemical Geology.
https://doi.org/10.1016/j.chemgeo.2016.03.010
Richardson, A. E., & Simpson, R. J. (2011). Soil Microorganisms Mediating Phosphorus
Availability Update on Microbial Phosphorus. Plant Physiology, 156(3), 989–996.
https://doi.org/10.1104/pp.111.175448
Riddle, M., Bergström, L., Schmieder, F., Kirchmann, H., Condron, L., & Aronsson, H. (2018).
Phosphorus Leaching from an Organic and a Mineral Arable Soil in a Rainfall Simulation
Study. Journal of Environmental Quality, 47(3), 487–495.
https://doi.org/10.2134/jeq2018.01.0037
156 Robinson, M. (1986). Changes in catchment runoff following drainage and afforestation. Journal
of Hydrology, 86(1–2), 71–84. https://doi.org/10.1016/0022-1694(86)90007-7
Rockwell, D. C., Warren, G. J., Bertram, P. E., Salisbury, D. K., & Burns, N. M. (2005). The U.S.
EPA Lake Erie Indicators Monitoring Program 1983–2002: Trends in Phosphorus, Silica, and
Chlorophyll a in the Central Basin. Journal of Great Lakes Research, 31(SUPPL. 2), 23–34.
https://doi.org/10.1016/S0380-1330(05)70302-6
Rosenzweig, C., & Parry, M. L. (1994). Potential impact of climate change on world food supply.
Nature. https://doi.org/10.1038/367133a0
Roßkopf, N., Fell, H., & Zeitz, J. (2015). Organic soils in Germany, their distribution and carbon
stocks. CATENA, 133, 157–170. https://doi.org/10.1016/j.catena.2015.05.004
Saarela, I., Järvi, A., Hakkola, H., & Rinne, K. (2004). Phosphorus status of diverse soils in Finland
as influenced by long-term P fertilisation 2. Changes of soil test values in relation to P balance
with references to incorporation depth of residual and freshly applied P. Agricultural and
Food Science. https://doi.org/10.2137/1239099042643099
Sanchez Valero, C., Madramootoo, C. A., & Stämpfli, N. (2007). Water table management impacts
on phosphorus loads in tile drainage. Agricultural Water Management, 89(1–2), 71–80.
https://doi.org/10.1016/j.agwat.2006.12.007
Santos, A. de C., McCray, J. M., Daroub, S. H., Rowland, D. L., Ji, S., & Sandhu, H. (2020).
Nitrogen Assessment of Shallow Florida Histosols. Communications in Soil Science and
Plant Analysis, 1–14. https://doi.org/10.1080/00103624.2020.1798990
Schindler, D. W., & Fee, E. J. (1974). Experimental Lakes Area: Whole-Lake Experiments in
Eutrophication. Journal of the Fisheries Research Board of Canada, 31(5), 937–953.
https://doi.org/10.1139/f74-110
157 Schindler, D. W., Hecky, R. E., Findlay, D. L., Stainton, M. P., Parker, B. R., Paterson, M. J., …
Kasian, S. E. M. (2008). Eutrophication of lakes cannot be controlled by reducing nitrogen
input: Results of a 37-year whole-ecosystem experiment. Proceedings of the National
Academy of Sciences, 105(32), 11254–11258. https://doi.org/10.1073/pnas.0805108105
Schindler, D. W., Hecky, R. E., & McCullough, G. K. (2012). The rapid eutrophication of Lake
Winnipeg: Greening under global change. Journal of Great Lakes Research, 38, 6–13.
https://doi.org/10.1016/j.jglr.2012.04.003
Schlichting, A., Leinweber, P., Meissner, R., & Altermann, M. (2002). Sequentially extracted
phosphorus fractions in peat-derived soils. Journal of Plant Nutrition and Soil Science,
165(3), 290–298. https://doi.org/10.1002/1522-2624(200206)165:3<290::AID-
JPLN290>3.0.CO;2-A
Schneider, K. D., Cade-Menun, B. J., Lynch, D. H., & Voroney, R. P. (2016). Soil Phosphorus
Forms from Organic and Conventional Forage Fields. Soil Science Society of America
Journal, 80(2), 328–340. https://doi.org/10.2136/sssaj2015.09.0340
Schott, L., Lagzdins, A., Daigh, A. L. M., Craft, K., Pederson, C., Brenneman, G., & Helmers, M.
J. (2017). Drainage water management effects over five years on water tables, drainage, and
yields in southeast Iowa. Journal of Soil and Water Conservation.
https://doi.org/10.2489/jswc.72.3.251
Schröder, J. J., Smit, A. L., Cordell, D., & Rosemarin, A. (2011). Improved phosphorus use
efficiency in agriculture: A key requirement for its sustainable use. Chemosphere, 84(6), 822–
831. https://doi.org/10.1016/j.chemosphere.2011.01.065
158 Schwärzel, K., Renger, M., Sauerbrey, R., & Wessolek, G. (2002). Soil physical characteristics of
peat soils. Journal of Plant Nutrition and Soil Science. https://doi.org/10.1002/1522-
2624(200208)165:4<479::AID-JPLN479>3.0.CO;2-8
Sengorur, B., Koklu, R., & Ates, A. (2015). Water Quality Assessment Using Artificial
Intelligence Techniques: SOM and ANN—A Case Study of Melen River Turkey. Water
Quality, Exposure and Health, 7(4), 469–490. https://doi.org/10.1007/s12403-015-0163-9
Sharpley, A. N. (1995). Soil phosphorus dynamics: agronomic and environmental impacts.
Ecological Engineering, 5(2–3), 261–279. https://doi.org/10.1016/0925-8574(95)00027-5
Sharpley, A. N. (1993). An Innovative Approach to Estimate Bioavailable Phosphorus in
Agricultural Runoff Using Iron Oxide‐Impregnated Paper. Journal of Environmental Quality.
https://doi.org/10.2134/jeq1993.00472425002200030026x
Sharpley, A. N., Bergström, L., Aronsson, H., Bechmann, M., Bolster, C. H., Börling, K., …
Withers, P. J. A. (2015). Future agriculture with minimized phosphorus losses to waters:
Research needs and direction. Ambio. https://doi.org/10.1007/s13280-014-0612-x
Sharpley, A., Jarvie, H. P., Buda, A., May, L., Spears, B., & Kleinman, P. (2013). Phosphorus
Legacy: Overcoming the Effects of Past Management Practices to Mitigate Future Water
Quality Impairment. Journal of Environmental Quality, 42(5), 1308–1326.
https://doi.org/10.2134/jeq2013.03.0098
Sharpley, A., Kleinman, P., Baffaut, C., Beegle, D., Bolster, C., Collick, A., … Weld, J. (2017).
Evaluation of Phosphorus Site Assessment Tools: Lessons from the USA. Journal of
Environmental Quality. https://doi.org/10.2134/jeq2016.11.0427
159 Shen, J., Yuan, L., Zhang, J., Li, H., Bai, Z., Chen, X., … Zhang, F. (2011). Phosphorus Dynamics:
From Soil to Plant. Plant Physiology, 156(3), 997–1005.
https://doi.org/10.1104/pp.111.175232
Sheng, Y. P., Chen, X., & Schofield, S. (1998). Hydrodynamic vs. non-hydrodynamic influences
on phosphorus dynamics during episodic events, 613-622
https://doi.org/10.1029/CE054p0613
Silva, G. (2012). Keeping muck soils sustainable. MSU Extension. Michigan State University.
https://www.canr.msu.edu/news/keeping_muck_soils_sustainable (accessed 29 Jan. 2019).
Sims, J. T., Simard, R. R., & Joern, B. C. (1998). Phosphorus Loss in Agricultural Drainage:
Historical Perspective and Current Research. Journal of Environment Quality, 27(2), 277.
https://doi.org/10.2134/jeq1998.00472425002700020006x
Sims, T. J. (2000). A phosphorus sorption index. In Methods of Phosphorus Analysis for Soils,
Sediments, Residuals, and Waters (pp. 22–23).
Singh, K. P., Basant, A., Malik, A., & Jain, G. (2009). Artificial neural network modeling of the
river water quality—A case study. Ecological Modelling, 220(6), 888–895.
https://doi.org/10.1016/j.ecolmodel.2009.01.004
Skaggs, R. W., Brevé, M. A., & Gilliam, J. W. (1994). Hydrologic and water quality impacts of
agricultural drainage∗. Critical Reviews in Environmental Science and Technology, 24(1), 1–
32. https://doi.org/10.1080/10643389409388459
Skaggs, R. W., Fausey, N. R., & Evans, R. O. (2012). Drainage water management. Journal of
Soil and Water Conservation, 67(6), 167A-172A. https://doi.org/10.2489/jswc.67.6.167A
Skaggs, R. W., Youssef, M. A., Gilliam, J. W., & Evans, R. O. (2010). Effect of controlled
drainage on water and nitrogen balances in drained lands. Transactions of the ASABE.
160 Soil Survey Staff. (2014). Keys to soil taxonomy. 12th ed. United States Department of
Agriculture-Natural Resources Conservation Service (USDA-NRCS).
Spivakov, B. Y. A., Maryutina, T. A., & Muntau, H. (1999). Phosphorus Speciation in Water and
Sediments. Pure and Applied Chemistry, 71(11), 2161–2176.
https://doi.org/10.1351/pac199971112161
Stämpfli, N., & Madramootoo, C. A. (2006). Dissolved Phosphorus Losses in Tile Drainage under
Subirrigation. Water Quality Research Journal, 41(1), 63–71.
https://doi.org/10.2166/wqrj.2006.007
Stephens, J.C. (1955). Drainage of peat and muck lands. In: Water. The 1955 yearbook of
agriculture. US Government Printing Office, Washington, DC. p. 539–557.
Stutter, M. I., Shand, C. A., George, T. S., Blackwell, M. S. A., Dixon, L., Bol, R., … Haygarth,
P. M. (2015). Land use and soil factors affecting accumulation of phosphorus species in
temperate soils. Geoderma. https://doi.org/10.1016/j.geoderma.2015.03.020
Tan, C. S., & Zhang, T. Q. (2011). Surface runoff and sub-surface drainage phosphorus losses
under regular free drainage and controlled drainage with sub-irrigation systems in southern
Ontario. Canadian Journal of Soil Science, 91(3), 349–359.
https://doi.org/10.4141/cjss09086
Terry, R. E. (1980). Nitrogen Mineralization in Florida Histosols. Soil Science Society of America
Journal, 44(4), 747–750. https://doi.org/10.2136/sssaj1980.03615995004400040018x
Thomas, D. L., Perry, C. D., Evans, R. O., Izuno, F. T., Stone, K. C., & Wendell Gilliam, J. (1995).
Agricultural Drainage Effects on Water Quality in Southeastern U.S. Journal of Irrigation
and Drainage Engineering, 121(4), 277–282. https://doi.org/10.1061/(ASCE)0733-
9437(1995)121:4(277)
161 Tiemeyer, B., & Kahle, P. (2014). Nitrogen and dissolved organic carbon (DOC) losses from an
artificially drained grassland on organic soils. Biogeosciences, 11(15), 4123–4137.
https://doi.org/10.5194/bg-11-4123-2014
Tiemeyer, B., Kahle, P., & Lennartz, B. (2010). Designing Monitoring Programs for Artificially
Drained Catchments. Vadose Zone Journal, 9(1), 14. https://doi.org/10.2136/vzj2008.0181
Tiessen, H., & Moir, J. (2007). Characterization of Available P by Sequential Extraction. In Soil
Sampling and Methods of Analysis, Second Edition (pp. 5–229).
https://doi.org/10.1201/9781420005271.ch25
Tiessen, H., Stewart, J. W. B., & Cole, C. V. (1984). Pathways of Phosphorus Transformations in
Soils of Differing Pedogenesis. Soil Science Society of America Journal, 48(4), 853–858.
https://doi.org/10.2136/sssaj1984.03615995004800040031x
Tiyasha, Tung, T. M., & Yaseen, Z. M. (2020). A survey on river water quality modelling using
artificial intelligence models: 2000–2020. Journal of Hydrology, 585, 124670.
https://doi.org/10.1016/j.jhydrol.2020.124670
Township of King. (2012). King Township’s integrated community sustainability plan. Township
of King.
Trafalis, T. ., Richman, M. ., White, A., & Santosa, B. (2002). Data mining techniques for
improved WSR-88D rainfall estimation. Computers & Industrial Engineering, 43(4), 775–
786. https://doi.org/10.1016/S0360-8352(02)00139-0
Tubiello, F. N., Biancalani, R., Salvatore, M., Rossi, S., & Conchedda, G. (2016). A worldwide
assessment of greenhouse gas emissions from drained organic soils. Sustainability
(Switzerland). https://doi.org/10.3390/su8040371
162 Turner, B. L., & Haygarth, P. M. (2001). Biogeochemistry: Phosphorus solubilization in rewetted
soils. Nature.
USGS. (1997). Land and people: Finding a balance - Everglades. United States Geological Survey.
Van Esbroeck, C. J., Macrae, M. L., Brunke, R. R., & McKague, K. (2017). Surface and subsurface
phosphorus export from agricultural fields during peak flow events over the nongrowing
season in regions with cool, temperate climates. Journal of Soil and Water Conservation,
72(1), 65–76. https://doi.org/10.2489/jswc.72.1.65
Villapando, R. R., & Graetz, D. A. (2001). Phosphorus Sorption and Desorption Properties of the
Spodic Horizon from Selected Florida Spodosols. Soil Science Society of America Journal,
65(2), 331–339. https://doi.org/10.2136/sssaj2001.652331x von Sperber, C., Stallforth, R., Du Preez, C., & Amelung, W. (2017). Changes in soil phosphorus
pools during prolonged arable cropping in semiarid grasslands. European Journal of Soil
Science, 68(4), 462–471. https://doi.org/10.1111/ejss.12433 von Sperber, C., Kries, H., Tamburini, F., Bernasconi, S. M., & Frossard, E. (2014). The effect of
phosphomonoesterases on the oxygen isotope composition of phosphate. Geochimica et
Cosmochimica Acta, 125, 519–527. https://doi.org/10.1016/j.gca.2013.10.010
Vu, D. T., Tang, C., & Armstrong, R. D. (2010). Transformations and availability of phosphorus
in three contrasting soil types from native and farming systems: A study using fractionation
and isotopic labeling techniques. Journal of Soils and Sediments, 10(1), 18–29.
https://doi.org/10.1007/s11368-009-0068-y
Waine, J., Brown, J. M. B., & Ingram, H. A. P. (1985). Non-Darcian transmission of water in
certain humified peats. Journal of Hydrology, 82(3–4), 327–339.
https://doi.org/10.1016/0022-1694(85)90025-3
163 Wallor, E., Herrmann, A., & Zeitz, J. (2018). Hydraulic properties of drained and cultivated fen
soils part II — Model-based evaluation of generated van Genuchten parameters using
experimental field data. Geoderma, 319, 208–218.
https://doi.org/10.1016/j.geoderma.2017.12.012
Wallor, E., Rosskopf, N., & Zeitz, J. (2018). Hydraulic properties of drained and cultivated fen
soils part I - Horizon-based evaluation of van Genuchten parameters considering the state of
moorsh-forming process. Geoderma. https://doi.org/10.1016/j.geoderma.2017.10.026
Wang, M., Liu, H., Zak, D., & Lennartz, B. (2020). Effect of anisotropy on solute transport in
degraded fen peat soils. Hydrological Processes, 34(9), 2128–2138.
https://doi.org/10.1002/hyp.13717
Wardle, D. A. (1998). Controls of temporal variability of the soil microbial biomass: A global-
scale synthesis. Soil Biology and Biochemistry. https://doi.org/10.1016/S0038-
0717(97)00201-0
Wasaki, J., & Maruyama, H. (2011). Molecular Approaches to the Study of Biological Phosphorus
Cycling. https://doi.org/10.1007/978-3-642-15271-9_4
Watanabe, F. S., & Olsen, S. R. (1965). Test of an Ascorbic Acid Method for Determining
Phosphorus in Water and NaHCO 3 Extracts from Soil. Soil Science Society of America
Journal, 29(6), 677–678. https://doi.org/10.2136/sssaj1965.03615995002900060025x
Weng, L., Van Riemsdijk, W. H., & Hiemstra, T. (2012). Factors Controlling Phosphate
Interaction with Iron Oxides. Journal of Environmental Quality.
https://doi.org/10.2134/jeq2011.0250
164 Wesström, I., Messing, I., Linnér, H., & Lindström, J. (2001). Controlled drainage - Effects on
drain outflow and water quality. Agricultural Water Management.
https://doi.org/10.1016/S0378-3774(00)00104-9
Whalen, J., & Sampedro, L. (2010). Soil ecology and management. 1st ed. CABI, Cambridge,
MA.
Williams, M. R., King, K. W., & Fausey, N. R. (2015). Drainage water management effects on tile
discharge and water quality. Agricultural Water Management, 148, 43–51.
https://doi.org/10.1016/j.agwat.2014.09.017
Williams, M. R., King, K. W., Macrae, M. L., Ford, W., Van Esbroeck, C., Brunke, R. I., … Schiff,
S. L. (2015). Uncertainty in nutrient loads from tile-drained landscapes: Effect of sampling
frequency, calculation algorithm, and compositing strategy. Journal of Hydrology, 530, 306–
316. https://doi.org/10.1016/j.jhydrol.2015.09.060
Winter, J. G., Eimers, M. C., Dillon, P. J., Scott, L. D., Scheider, W. A., & Willox, C. C. (2007).
Phosphorus inputs to Lake Simcoe from 1990 to 2003: Declines in tributary loads and
observations on lake water quality. Journal of Great Lakes Research, 33(2), 381–396.
https://doi.org/10.3394/0380-1330(2007)33[381:PITLSF]2.0.CO;2
Withers, P. J. A., Sylvester-Bradley, R., Jones, D. L., Healey, J. R., & Talboys, P. J. (2014). Feed
the Crop Not the Soil: Rethinking Phosphorus Management in the Food Chain.
Environmental Science & Technology, 48(12), 6523–6530.
https://doi.org/10.1021/es501670j
Withers, P. J. A., Hodgkinson, R. A., Rollett, A., Dyer, C., Dils, R., Collins, A. L., … Sylvester-
Bradley, R. (2017). Reducing soil phosphorus fertility brings potential long-term
165 environmental gains: A UK analysis. Environmental Research Letters.
https://doi.org/10.1088/1748-9326/aa69fc
Xu, J., Morris, P. J., Liu, J., & Holden, J. (2018). PEATMAP: Refining estimates of global peatland
distribution based on a meta-analysis. Catena. https://doi.org/10.1016/j.catena.2017.09.010
Youssef, M. A., Abdelbaki, A. M., Negm, L. M., Skaggs, R. W., Thorp, K. R., & Jaynes, D. B.
(2018). DRAINMOD-simulated performance of controlled drainage across the U.S. Midwest.
Agricultural Water Management. https://doi.org/10.1016/j.agwat.2017.11.012
Zak, D., Gelbrecht, J., Wagner, C., & Steinberg, C. E. W. (2008). Evaluation of phosphorus
mobilization potential in rewetted fens by an improved sequential chemical extraction
procedure. European Journal of Soil Science. https://doi.org/10.1111/j.1365-
2389.2008.01081.x
Zauft, M., Fell, H., Glasser, F., Rosskopf, N., & Zeitz, J. (2010). Carbon storage in the peatlands
of Mecklenburg-Western Pomerania , north-east Germany. Mires and Peat.
Zhang, B., & Govindaraju, R. S. (2000). Prediction of watershed runoff using Bayesian concepts
and modular neural networks. Water Resources Research, 36(3), 753–762.
https://doi.org/10.1029/1999WR900264
Zhang, T., Wang, Y., Tan, C. S., & Welacky, T. (2020). An 11-Year Agronomic, Economic, and
Phosphorus Loss Potential Evaluation of Legacy Phosphorus Utilization in a Clay Loam Soil
of the Lake Erie Basin. Frontiers in Earth Science, 8(May), 1–8.
https://doi.org/10.3389/feart.2020.00115
Zheng, Z. M., Zhang, T. Q., Kessel, C., Tan, C. S., O’Halloran, I. P., Wang, Y. T., … Van Eerd,
L. L. (2015). Approximating Phosphorus Leaching from Agricultural Organic Soils by Soil
166 Testing. Journal of Environmental Quality, 44(6), 1871–1882.
https://doi.org/10.2134/jeq2015.05.0211
Zheng, Z. M., Zhang, T. Q., Wen, G., Kessel, C., Tan, C. S., O’Halloran, I. P., … Speranzini, D.
(2014). Soil Testing to Predict Dissolved Reactive Phosphorus Loss in Surface Runoff from
Organic Soils. Soil Science Society of America Journal, 78(5), 1786–1796.
https://doi.org/10.2136/sssaj2014.02.0065
Zheng, Z., MacLeod, J. A., Sanderson, J. B., & Lafond, J. (2004). Soil Phosphorus Dynamics After
Ten Annual Applications of Mineral Fertilizers and Liquid Dairy Manure: Fractionation and
Path Analyses. Soil Science, 169(6), 449–456.
https://doi.org/10.1097/01.ss.0000131225.05485.25
Zheng, Z., Simard, R. R., Lafond, J., & Parent, L. E. (2002). Pathways of Soil Phosphorus
Transformations after 8 Years of Cultivation under Contrasting Cropping Practices. Soil
Science Society of America Journal, 66(3), 999–1007.
https://doi.org/10.2136/sssaj2002.9990
Zhu, Y., Wu, F., He, Z., Guo, J., Qu, X., Xie, F., … Guo, F. (2013). Characterization of Organic
Phosphorus in Lake Sediments by Sequential Fractionation and Enzymatic Hydrolysis.
Environmental Science & Technology, 47(14), 7679–7687.
https://doi.org/10.1021/es305277g
167