REVIEWS

Deforestation and reforestation impacts on soils in the tropics

Edzo Veldkamp 1 ✉ , Marcus Schmidt 1, Jennifer S. Powers2,3 and Marife D. Corre1 Abstract | Soils under natural, tropical forests provide essential ecosystem services that have been shaped by long-term​ soil–vegetation feedbacks. However, deforestation of tropical forest, with a net rate of 5.5 million hectares annually in 2010–2015, profoundly impacts soil properties and functions. Reforestation is also prominent in the tropics, again altering the state and functioning of the underlying soils. In this Review, we discuss the substantial changes in dynamic soil properties following deforestation and during reforestation. Changes associated with deforestation continue for decades after forest clearing eventually extend to deep subsoils and strongly affect soil functions, including nutrient storage and recycling, carbon storage and greenhouse gas emissions, erosion resistance and water storage, drainage and filtration. Reforestation reverses many of the effects of deforestation, mainly in the topsoil, but such restoration can take decades and the resulting soil properties still deviate from those under natural forests. Improved management of soil organic matter in converted land uses can moderate or reduce the ecologically deleterious effects of deforestation on soils. We emphasize the importance of soil knowledge not only in cross-disciplinary​ research on deforestation and reforestation but also in developing effective incentives and policies to reduce deforestation.

1,2 27 (Fig. 1) Slash-and-​ burn​ Soils in the tropics provide essential functions (such as natural forest losses . Substantial area that was 3–5 management nutrient storage and recycling , carbon (C) storage and formerly cleared for agriculture has also been abandoned, Cutting down and burning of greenhouse gas (GHG) emissions6,7, erosion resistance8 commonly because of soil degradation, and now pro- vegetation in an area, often as and water storage, drainage and filtration9–11), and soil vides areas for forest regeneration by secondary succession preparation for agricultural use. variability within and amongst landscapes promotes or replanting28. Indeed, replanted forest area has 12–14 Shifting cultivation biodiversity of tropical forests . The properties of increased rapidly, at approximately 1 million ha per Agricultural system in which soils under natural tropical forests reflect long-​term year between 1990 and 2015, a trend that is projected the area is fallowed in between soil-​vegetation feedbacks, influenced by the high pro- to continue27. Currently, the total area of primary forests periods of cultivation, allowing ductivity, regular litter input and permanent deep root in tropical countries (541 million ha) is only approxi- natural vegetation to return 15 and soils to recover. systems that are characteristic of these forests, all of mately half of regenerating forests (1,172 million ha), which stimulate activity of soil organisms16,17. Compared illustrating their global significance29,30. with agricultural land uses, tropical forests promote Deforestation and reforestation can lead to pro- 1Faculty of Forest Sciences efficient soil-​nutrient recycling18,19 and modulate soil found changes in dynamic soil properties that directly and Forest Ecology, Soil temperature and moisture20. and indirectly affect many soil functional processes31,32. Science of Tropical and Humans have shaped tropical forests and their soils For example, deforestation generally leads to lower Subtropical Ecosystems, 21,22 6,7,33–35 University of Goettingen, for millennia, especially since the onset of agriculture . soil organic carbon (SOC) stocks , higher soil Goettingen, Germany. For centuries, slash-​and-​burn management during bulk density36 and changes in soil pH (ref.37). Changes 2Department of Ecology, shifting cultivation was a main cause of forest clearing, in dynamic soil properties occur most rapidly in the Evolution, and Behavior, but was offset by natural regrowth during fallow periods23, organic-​matter-​rich topsoils with the highest biolog- University of Minnesota, thus, preventing significant net deforestation24. Tropical ical activity6,38. Deeper soil horizons (>50 cm) are also St. Paul, MN, USA. deforestation as a global process started during colonial affected by deforestation and reforestation, although 3 Department of Plant and times and intensified in the second half of the twentieth changes there occur more slowly and often become Microbial Biology, University 38 of Minnesota, St. Paul, century as a result of the demand-​driven expansion of substantial only after several decades . However, most 25 MN, USA. logging and agricultural land , typically supported by studies do not report long-term​ changes in dynamic soil 26 ✉e-mail:​ [email protected] government policies . At 5.5 million hectares (ha) per properties — of those reviewed here, only 48% included https://doi.org/10.1038/ year (based on the 2010–2015 period), tropical deforest- land uses ≥25 years old. Nevertheless, most studies s43017-020-0091-5 ation continues to be the largest contributor of global assumed that a steady state was reached within the period

590 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

42 Key points owing to the input of carbonate-​containing ashes that buffer soil pH to values between 6.5 and 7.5 (ref.41). • Deforestation leads to profound changes in dynamic soil properties that degrade In the absence of liming during land management, the most soil functions. net nutrient export by harvested products and leaching • The rate and degree of soil degradation following deforestation are a function of the losses43 cause a net release of H+ that can eventually lead inherent soil fertility and land-​use intensity. to a drop in soil pH over time41. Indeed, in >25-year-​ old​ • Changes in dynamic soil properties continue for decades following deforestation and croplands, soil pH decreases to values below those of the eventually extend to deep subsoils. original forest soils in the top 10 cm (Fig. 3a). Changes • Reforestation reverses some of the undesirable effects of deforestation on dynamic in pH following slash-​and-​burn management are less soil properties; however, the resulting soil conditions and their functions are pronounced in HAC soils compared with LAC soils substantially different from the previous soils under natural forests. (Supplementary Fig. 1). Conversion of forests to pastureland also rapidly studied or that further changes were insignificant7,35. increases soil pH owing to ash input44 (24% increase In combination with the undersampling of deeper soil in the top 10 cm and 9% increase in 10–50 cm, Fig. 3c), horizons, this assumption has led to the relatively wide- followed by a slow decrease in pH in subsequent years44 spread view that only topsoils are affected by deforest- (Fig. 3c). In contrast to croplands, >25-year-​ old​ pastures ation and reforestation, and that within approximately maintain soil pH values above those of the original for- one or two decades following deforestation, most ests (3–9% increase, Fig. 3c), likely because only small dynamic soil properties have reached steady-​state or amounts of base cations are exported through harvest equilibrium values7,35. and leaching11. As with crop and pastureland, tree cash Fallow periods In this Review, we examine the impacts of deforest- crop plantations (such as for rubber and oil palm) are Time during which arable land ation and reforestation on soils in the tropics. We often established using slash-and-​ burn​ management45,46; is not actively used in crop compile changes in dynamic soil properties following soil pH increases with age in these plantations, likely production. forest clearing and during forest regeneration, explicitly because of regular liming45 (Fig. 4a). Deforestation including results from studies that covered several dec- Most soils are negatively charged and retain posi- Removal of forest and ades and/or measurements from deeper subsoils. These tively charged nutrients (including K+, Ca2+, Mg2+ and conversion of land for other + quantitative results are framed in a broader context by NH4 ) in exchangeable form in the soil, referred to as uses. reviewing evidence of how changes in dynamic soil ECEC when measured at field pH (ref.40) (Box 1). LAC Forest regeneration properties affect important soil functions. Finally, we soils have low, pH-​dependent ECEC, with a value of −1 Re-​establishment of forest discuss how soil-management​ practices in land-use​ tran- 3.20 ± 0.15 cmolc kg in the natural forest soils included after disturbance. sitions affect soil functionality and identify important here. The ECEC is higher in HAC soils under natu- knowledge gaps. ral forests, with a value of 9.95 ± 1.05 cmol kg−1 in the Secondary succession c Ecological changes during the reviewed studies. Following slash-​and-​burn manage- regeneration of an ecosystem Impacts on dynamic soil properties ment, the inherently low ECEC of LAC soils quickly on disturbed or damaged land. Dynamic soil properties, which we use here to describe increases in both croplands42 (up to 35% increase com- characteristics that change over years to decades pared with forests, Supplementary Fig. 1a) and pastures44 Primary forests Native forests that lack owing to land-​use change and management, include (25–31% increase, Supplementary Fig. 1c). In contrast, substantial signs of human soil pH, effective cation-​exchange capacity (ECEC), the inherently high ECEC of HAC soils decreases with activity or disturbance, base saturation, bulk density (BD), SOC and soil C:N forest conversion to croplands (Supplementary Fig. 1b). sometimes referred to as ratio39. Anthropogenic activities during and following Eventually (after >25 years), ECEC values of both LAC old-growth​ forest. deforestation or reforestation, such as slash-​and-​burn and HAC soils fall below those of the original forested Dynamic soil properties management or the replanting of trees, alter these soils, related to long-​term SOC decreases following Soil properties that change properties both directly and indirectly (Fig. 2). In the deforestation. with disturbances and following sections, we present changes in dynamic Base saturation (the percentage of ECEC occupied by management. soil properties from a compilation of 120 studies of exchangeable bases) increases with increase in soil pH,

Topsoils forest-​related land-​use change in 35 tropical countries and a base saturation above 50% is considered favourable Uppermost layer of soil, (Supplementary material). because of the availability of bases as nutrients and the specifically, the top 10 cm for low amount of exchangeable Al, which can be toxic47. this Review. Soil pH, ECEC and base saturation. Although tropical Under natural forests, the base saturation of LAC soils is forests are located on diverse soil types, most remaining typically low (15.5% ± 0.8 in the reviewed studies), owing Subsoils Soil layers >10 cm; deeper natural forests are located on heavily weathered soils to the low soil pH, whereas the base saturation of HAC subsoils refers to >50 cm for with low-​activity clays (LAC soils, Fig. 1, Box 1), which soils is higher (66.6% ± 2.4 in the studies included here). this Review. have inherently low pH (from our dataset: 4.82 ± 0.02), Following slash-and-​ burn​ management, base saturation acid-​buffering capacity and fertility40. Compared with rapidly increases in LAC soils due to the base-containing​ Effective cation-exchange​ Fig. 3a,c capacity LAC soils, the inherent soil pH of high-​activity clay ashes ( , Supplementary Fig. 1a,c). In contrast, (ECEC). Negatively charged (HAC) soils under natural forests is typically higher such change is less pronounced in HAC soils, probably sites in the soil that adsorb (5.83 ± 0.07 in the reviewed studies), as they are buff- resulting from the higher initial base saturation and exchangeable cations, ered by the release of base cations (Ca2+, Mg2+, K+, Na+) the presence of weatherable minerals (Supplementary measured at field pH. during silicate weathering of primary or clay minerals41. Fig. 1b,d). After 25 years or more, the base saturation Liming In general, soil pH increases following slash-​and-​burn of pasture soils remains slightly elevated relative to the 48 Treatment of soil with lime, with management (Figs 3,4), as seen in all soil depths of original forested values (Fig. 3c, Supplementary Fig. 1c), the goal of reducing acidity. <10-​year-​old croplands (Fig. 3a, Supplementary Fig. 1), but is relatively lower in cropland soils (Fig. 3a,

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 591 Reviews

a

Equator

Forest cover loss Forest cover gain Forest cover in 2000 b

Equator

Soils with low-activity clays Soils with high-activity clays

Fig. 1 | Tropical tree cover change and soils based on dominant clay mineralogy. a | Tropical tree cover changes between 2000 and 2018 in areas that had greater than 60% tree cover in 2000 (ref.186). b | Tropical soils with high-​activity and low-activity​ clays, based on the SoilGrids250m dataset187 and categorized using the FAO World Reference Base for soil resources188. Pixels classified as Ferralsols, Acrisols, Lixisols, Nitisols and Sesquisols were categorized as soils with low-activity clays, while Cambisols, Luvisols, Vertisols, Planosols and Alisols (and others) were categorized as soils with high-activity​ clays. For panel b, forest areas were delineated based on FAO Global Forest Resources Assessment189 as tropical rainforests, tropical moist deciduous forests, tropical dry forests and tropical mountain forests. Tree cover data from: http://earthenginepartners.appspot.com/science-2013-​global-​forest.

Supplementary Fig. 1a), likely owing to long-​term Bulk density, SOC and C:N ratios. Land-​use change base-​cation export. In tree cash crop plantations on in tropical regions impacts soil BD, an indicator of LAC soils45, liming (a common practice in these systems) compaction or porosity that is influenced by root bio- might have contributed to the higher base saturation of mass, biological activity, SOC content and texture51,52. the top 50 cm of soil in >25-year-​ old​ plantations relative Changes in BD following deforestation can be caused by to the original forests (Fig. 4a). the use of heavy machines53,54, modifications in macro- Reforestation of croplands or pastures often does not faunal activities55, foot traffic by grazing animals36 and result in clearly reversed trends for soil pH, ECEC and decreases in root biomass56 and SOC content57 (Fig. 2a). base saturation (Fig. 3b,d). For instance, the net transfer The mean BD of LAC soils under natural forests with a of base cations from the soil to trees in secondary forests clay content ≥60% (top 10 cm) is 0.84 ± 0.03 g cm−3 and can lead to a decrease in pH41 (Fig. 2b), but nutrients are 1.22 ± 0.03 g cm−3 for soils with ≥60% sand, according also returned to the soil through litterfall, which could to the data compiled here. Following deforestation, BD compensate for this effect11,49 by replenishing the base rapidly increases36,56 in the top 50 cm in both LAC and cations. The increased ECEC in the reforested cropland HAC soils (4–13% increase for croplands and 4–15% topsoil within 10 years is likely related to the increase increase for pastures, Fig. 3a,c, Supplementary Fig. 1). in SOC50 (Fig. 3b), as organic-​acid groups and phenolic There is a progressive increase in BD at deeper depths Secondary forests groups impact the pH-​dependent ECEC of LAC soils as croplands age58, probably related to the concurrent Forests established after the (Box 1). Generally, changes in pH and ECEC following decrease in SOC (Fig. 3a, Supplementary Fig. 1a,b). removal or disturbance of the original (primary) forests. reforestation take decades before possibly reaching val- Compared with the BD increases of soils in croplands ues similar to those of original forest soils, a notable con- and pastures that have been converted for at least Texture trast to the rapid changes in these properties following 25 years (15% and 22% increases, respectively, in the Composition of soil in terms of deforestation. Importantly, the data compiled here high- top 10 cm), the increases in tree cash crop plantations sand, silt and clay. light that there are fewer published studies on reforest- (9%) and agroforestry systems (12%) are lower (Fig. 4). Agroforestry ation compared with deforestation (Supplemental These differences are likely related to the permanent Growth of trees or shrubs and Table 1), which contributes to unclear trends in pH, root systems of tree-​based land uses in combination crop products concurrently. ECEC and base saturation under secondary forests. with high biological activity (stimulated by litter input)

592 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

and absence of grazing animals, promoting lower BD in temperature and moisture), affect soil biological activity the tree-covered​ systems. and decomposition rate. These changes, in turn, lead to Deforestation and reforestation lead to drastic the alteration of SOC, as it reflects a dynamic equilibrium changes in inputs of litter or organic residues, which, between input (litter or organic residues) and loss rates together with changes in microclimate (such as soil (decomposition), and depends on stabilization by clay

a Deforestation b Reforestation

Use of heavy Removal of Slash-and- Succession/ machines trees burn planting trees

1 3 6 Permanent Litter Permanent Litter Nutrient roots input roots input uptake

3 4 2 4 2 C:N C:N ratio ratio

Biological SOM pH Biological pH activity SOM activity 5 9 3 8 7

Bulk Base ECEC Bulk Base density saturation density ECEC saturation

Human activities Vegetation response Dynamic soil properties

c Dynamic soil properties Selected soil functions

SOM Nutrient storage and recycling

C:N ratio Carbon storage and greenhouse gas emissions ECEC Water storage and drainage pH and base Erosion resistance saturation Bulk density Water filtration

Fig. 2 | Linkages between land-use changes, soil properties and soil functions. a | Hypothesized linkages during and following deforestation between human activities (grey boxes with short-dashed​ outlines), vegetation response variables (red and blue boxes with long-​dashed outlines) and dynamic soil properties (red and blue boxes with solid outlines). Red boxes indicate a decrease in the property or function, blue indicates an increase, and red and blue together represent various or diverging responses. Deforestation often involves the use of heavy machines, which can increase soil bulk density53 (BD) (arrow 1). Removal of trees strongly reduces deep, permanent tree roots and litter input, which, in turn, lead to a reduction in the soil organic matter (SOM) content9 (arrow 2). Tree removal can also induce soil compaction, as litterfall is reduced and macropores collapse after remaining tree roots decompose. Slash-​and-burn​ decreases macrofaunal activity, which, together with tree removal, increases BD55,56 (arrow 3). As decomposed SOM has a lower C:N ratio than fresh litter, there is a reduction in the C:N ratio of the remaining SOM in the years following deforestation190 (arrow 4). A reduction in SOM further contributes to an increase in BD because the SOM has a lower density than the mineral soil fraction57 (arrow 5). Burning of the remaining slash leads to direct nutrient losses by volatilization and an input of nutrients through ashes on the topsoil, increasing the pH (arrow 6) and base saturation45,106 (arrow 7). Owing to the presence of low-activity​ clays and SOM, the effective cation-exchange​ capacity (ECEC) of many soils in the tropics depends on pH. An increase in pH raises the ECEC44 (arrow 8), whereas a decrease in SOM content has the opposite effect78 (arrow 9). b | Reforestation (secondary succession or replanted forest) often leads to opposite changes in dynamic soil properties because of the establishment of a permanent deep root system and increasing litter input. Deep and permanent roots will also lead to the uptake of nutrients, including base cations with congruent release of H+ by roots41, which can decrease soil pH. c | Linkages of dynamic soil properties to selected soil functions.

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 593 Reviews

59 60 61 Standing biomass content , clay mineralogy , soil-​aggregate formation on average, 17% of the original forests’ SOC contents 62 Total amount of biomass in an and soil biological activity . SOC also interacts with during the first 10 years following deforestation (Fig. 3a). area. many other soil properties, including BD, ECEC, bio­ However, topsoils of croplands continue to lose SOC: l­ogical activity and soil-​aggregate stability (Fig. 2a,b). In 29% on average after 10–25 years and 58% for croplands Nutrient-use​ efficiency 58 Fig. 3a The biomass produced per unit tropical forests, clay and base-​cation contents, rainfall >25 years (but <100 years) old ( , Supplementary of nutrients taken up by plants. and root biomass appear to be better predictors of SOC Fig. 1a,b). Several studies have reported substantial than standing biomass4,38. labile SOC stocks and microbial activity below 50-​cm SOC is a mixture of fractions that have a wide range depth in deeply weathered tropical forest soils9,38,65,66. of residence times (from less than a year to centuries Although the SOC contents in depths ≥50 cm of crop- or millennia62) and respond differently to land-​use lands ≤25 years old did not differ from the original for- change63. Changes in SOC stocks following deforestation ests (Fig. 3a), soils at ≥50-​cm depth and >25 years old and reforestation have been studied extensively because contain 35% less SOC content compared with the orig- of the prominent role that land-​use trajectories play in inal forests (Fig. 3a, Supplementary Fig. 1a,b). Thus, we the contemporary global C budget6,7,33,64. Earlier reviews did not find evidence of a new equilibrium in SOC con- have shown decreases between 18% and 25% in SOC tents within two decades of land-use​ change, neither in stocks of the top 0.3 m depth following deforestation for LAC nor in HAC soils, as was postulated previously7,35. annual crops6,7. Our Review similarly finds that, in both How long SOC losses will effectively continue follow- LAC and HAC soils, the top 10 cm of cropland soil loses, ing deforestation is unclear — a 2018 study suggests that the present-​day SOC pool turnover times in the Yucatán Peninsula continue to be reduced as a result Box 1 | Diversity of soils in the tropics of Mayan deforestation 1,000 years earlier63. Thus, soils Soils in tropical areas have Ferralsol Vertisol in large areas of the tropics likely continue to be a CO2 diverse properties, varying source even decades after clearing the original forests, fertility and different which might be underestimated by current C accounting reactions to land-use​ methods. For example, the Tier 1 approach in the IPCC changes6,191. Several FAO soil 20 groups188 (such as Ferralsols, Guidelines for National Greenhouse Gas Inventories only considers a default soil depth of 30 cm and that Acrisols, Lixisols, Nitisols and 40 changes in SOC stocks occur over a period of 20 years67. Sesquisols) or USDA soil 20 orders192 (such as Oxisols and For pastures, the direction and magnitude of changes Ultisols) largely occur in 60 in SOC stocks following deforestation have been linked tropical regions, and are old, to clay mineralogy and precipitation6, and the SOC highly weathered, deep (up to 80 stocks in some LAC soils even increase following 20 m), acidic and generally deforestation44. The dynamic nature of these changes low-fertility​ soils with little 100 40 becomes apparent in our dataset: in LAC soils, SOC con- primary mineral reserves for tents in <10-​year-old​ pastures increase in the top 10 cm replenishment of nutrients via weathering. These 120 (16%, Supplementary Fig. 1c) compared with the origi- low-activity​ clay (LAC) soils nal forests, likely as a result of the temporary increase in 68,69 cover about 44% of the 140 nutrient availability from ashes , the mineralization of (refs70,71) tropics (left image in 60 organic N and the fine-​root growth of grasses the figure, scale shown in the topsoil72. However, such increase is transitory, as in centimetres), mainly under 160 the topsoils of >25-​year-​old pastures lose 19% of SOC tropical rainforest or compared with the original forests (Fig. 3c). In HAC soils, savannah40,47, and are 180 topsoil SOC contents rapidly decrease and continue in dominated by a kaolinitic- >25-year-​ old​ pastures (Supplementary Fig. 1d), indicat- clay mineralogy. LAC soils 80 ing that SOC contents decrease on a longer timescale. have pH-dependent​ charge, mostly from hydroxide groups (-​OH) at the edges of clay minerals, as well as phenolic Reports of >25-year-​ old​ pastures with higher SOC stocks (-​OH) and organic-​acid groups (-​COOH) of soil organic matter. Depending on the pH than the original forests are mostly related to improved 65,73 of the soil, H+ ions are dissociated from these groups, resulting in negative charges, or management: the pastures were either fertilized , had 48 dissolved H+ ions are adsorbed, leading to positive charges, especially in the subsoil with significant C and N input from legumes or had low low soil organic matter content. In practice, these properties mean that the negative grazing intensities74. charge of LAC soils is low and, if the pH increases, more H+ will be dissociated and Deforestation for tree cash crop plantations reduces negative charge and effective cation-​exchange capacity (ECEC) increase. The anion SOC contents by 20–22% (Fig. 4a), which is detecta- exchange capacity of LAC soils increases with decreasing pH. The ECEC of LAC soils, ble down to 50-​cm depth in >25-​year-​old plantations. thus, depends on the pH and organic matter content of the soil. SOC stock losses from tree cash crop plantations may Other FAO soil groups (such as Luvisols, Cambisols, Vertisols and Alisols) or USDA soil be as high as 50% (refs38,75) and can be predicted by orders (Alfisols, Inceptisols and Vertisols) are younger, less weathered, often not deeper 38,58 than 1–2 m and generally fertile, covering approximately 32% of the tropics40 (right image the magnitude of SOC stocks in the original forests . in the box figure). The clay mineralogy of these high-activity​ clay soils is often dominated Diverse agroforestry systems often maintain SOC stocks by smectite, illite or vermiculite, and the soils have large, permanent negative charge, in comparison with the original forests, especially in which mainly originates from substitutions of ions in silicate-clay​ minerals. As this charge humid climates58,76,77, potentially owing to higher nutri- originates from inside the clay lattice, it cannot be neutralized by covalent bonding of ent and organic inputs in home gardens, N inputs from dissolved cations and the ECEC, thus, does not change with soil pH. Images courtesy legume shade trees43,78 and a high ecosystem nutrient-​ of N. A. Jelinski, University of Minnesota. use efficiency79. In drier regions, agroforestry systems

594 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

Cropland Pastures a b c d pH 0.3 0.3

0.2 0.2

0.1 0.1

0.0 0.0 Relative change –0.1 –0.1

Effective cation-exchange capacity

1.0 1.0

0.5 0.5

0.0 0.0 Relative change

Base saturation 6.0 6.0

4.0 4.0

2.0 2.0

Relative change 0.0 0.0

Bulk density 0.3 0.3

0.2 0.2

0.1 0.1

0.0 0.0 Relative change –0.1 –0.1

Soil organic carbon

0.5 0.5

0.0 0.0

–0.5 –0.5 Relative change

–1.0

C:N ratio

0.4 0.4

0.2 0.2

0.0 0.0

–0.2 –0.2 Relative change

Crop Forest Forest Pasture <10 years >25 years <10 years >25 years <10 years >25 years <10 years >25 years 10–25 years 10–25 years 10–25 years 10–25 years

<10 cm 10 to <50 cm 50 to 600 cm

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 595 Reviews

◀ Fig. 3 | Changes in soil properties during land-use change for croplands and Such land-​use conversion directly causes the species pastures. Relative change ([converted − reference]/reference); bars are bootstrapped richness of detritivores and omnivores to decrease, 95% confidence intervals from 10,000 randomizations) in dynamic soil properties and reduces the biomass of omnivores and reliant following deforestation for cropland (panel a), reforestation of cropland (panel b), predators12. However, the same land-use​ conversion can deforestation for pastures (panel c) and reforestation of pastures (panel d). Old-​growth forest is the reference land use for deforestation (panels a and c), whereas cropland indirectly increase biomass and species richness through (panel b) or pasture (panel d) is the reference land use for secondary forest. Results are the bottom-​up cascading effects of microclimate, SOM presented as aggregated years for young (<10 years), intermediate (10–25 years) and old and soil nutrients, which then enhances omnivore bio- (>25 years) croplands, pastures and secondary forests. The number of observations are mass and the species richness of detritivores and their reported in Supplemental Table 1. Changes in soil properties progress to deeper depths predators12. The net impacts of land-​use change for tree decades after deforestation; reforestation restores some soil properties, mainly in the cash crop plantations are decreases in functional diver- topsoil, but their values do not reach those of the original forests. Symbols adapted sity of the decomposer community, decomposition83 and courtesy of the Integration and Application Network, University of Maryland Center for soil-​N-​cycling rates; the latter is associated with reduc- Environmental Science (http://ian.umces.edu/symbols/). tions in microbial biomass and SOM84. The differences between direct and indirect changes highlight that over- have the tendency to lose SOC58. From our dataset, all changes in ecosystem functions cannot be linked to agroforestry systems indeed show an average decrease a single soil function. of 24% in SOC contents after >25 years, which is only Within the microbial community, conversion of for- detectable in the top 10 cm (Fig. 4b). ests to tree cash crop plantations can cause decreases The C:N ratio of soil organic matter (SOM) typically in symbiotrophic fungi (ectomycorrhizae, arbuscular decreases with increasing turnover time80. The data here and ericoid mycorrhizae, which are important for P and show that soil C:N ratios in ≤25-year-​ old​ croplands were organic N acquisition) and increases in saprotrophic similar to those of the original forests, but decrease after and pathotrophic fungi, as seen in Indonesian planta- >25 years (Fig. 3a), suggesting that the lower SOC con- tions; this restructuring of the fungal community sug- tents of older croplands might have slow turnover times. gests enhanced pathogens in plantations85. The bacterial In contrast, an increase in soil C:N ratio with pasture community also differs between plantation soils and age is often observed in unfertilized pastures (for exam- forest soils in Indonesia, with the relative abundance ple, the >10-cm​ depth in >25-year-​ old​ pastures, Fig. 3c), of Alphaproteobacteria decreasing and Acidobacteria which is interpreted as a progressive decline in N availa- increasing with increased land-​use intensity, and the bility with pasture age71, potentially in combination with diversities of the bacterial and archaeal communi- the invasion of woody shrubs with a high litter C:N ratio ties increasing in plantations relative to forest86. These in old, degraded pastures65. changes were related to increases in soil pH, base sat- Reforestation of cropland and pastures through sec- uration, C:N ratio and extractable P due to liming and ondary succession81 or tree plantations82 reverses the chemical fertilizations in tree cash crop plantations on decreasing trend in SOC contents, leading to a 14% LAC soils86,87. increase of SOC in the topsoils within 10–25 years for Understanding the changes in populations and secondary forests converted from former croplands communities of soil-​dwelling invertebrates during (Fig. 3b) and an 18% increase of SOC from former pas- secondary succession and determining the role of micro- tures (Fig. 3d). Reforestation of former croplands also bial symbionts (like mycorrhizal fungi88,89 and N-​fixing increases the C:N ratios of SOM in the topsoil (Fig. 3b). bacteria90) in facilitating successional dynamics of Although secondary forests restore some of the SOC trees are two major themes addressed by soil-​biology contents lost following deforestation and generally research. The regeneration of secondary forest often Species richness reverse trends in BD and C:N ratios, these appear to be leads to rapid increases in litterfall production and Number of species in a mainly limited to the topsoil (Fig. 3b,d). A pattern starts decomposition, with fluxes approaching those in older community. to emerge showing that the SOC, BD and C:N ratios in forests within several decades91,92. However, populations subsoils are restored much more slowly than in topsoils, and communities of litter and soil invertebrates do not Detritivores Organism that feeds on dead and do not reach values comparable to the original for- necessarily change in concert with litterfall quantities 92 biomass. est within a few decades. Obviously, the low number during succession , and, instead, can be more respon- of observations for secondary forests >25 years old in sive to changes in litter chemistry93. The abundances of Symbiotrophic which subsoils were sampled (Supplemental Table 1) many macro-​invertebrate groups increase with forest Receiving nutrients by contribute to the uncertainty of this statement. regeneration92,94, although some groups of organisms exchanging resources with host cells. do not — Earthworms, for example, decrease in suc- Soil biology cessional forests compared with pastures93. Similarly, Mycorrhizae The changes in dynamic soil properties with land-​use the diversity and species composition of invertebrates Symbiotic relationship change are mediated by soil biological communities in secondary forests relative to older forests show vary- between plant and fungus in a rooting system. that, in turn, can alter soil functions, including decom- ing patterns, with communities becoming increasingly position and/or nutrient cycling83,84, greenhouse gas similar to old forests in some studies95 but not in others92. Saprotrophic emissions17,70 and water storage20,32 (Fig. 2c), and biocon- Inconsistent methodology94 hampers the determination Receiving nutrients by breaking trol through predation12,16. For example, conversion of of what drives the divergent responses of soil biological down dead host cells. tropical forests to tree cash crop plantations decreases communities to forest regeneration, which, ultimately, species richness Pathotrophic macro-invertebrate​ and biomass, driven might depend on a number of factors beyond dynamic Receiving nutrients by harming by decreases in litterfall and overall energy fluxes, soil properties, such as proximity to sources of colonizers host cells. and shifts from predator to omnivore dominance16. and plant-​species composition.

596 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

Tree cash crop plantations Agroforestry Fig. 4 | Changes in soil properties during land-use change for tree cash crop plantations and agroforestry. a b Relative change ([converted − reference]/reference); pH bars are bootstrapped 95% confidence intervals from 10,000 0.2 randomizations) in dynamic soil properties following deforestation for tree cash crop plantations (panel a) and 0.1 agroforestry (panel b). Old-growth​ forest is the reference land use. Results are presented as aggregated years for young ( 10 years), intermediate (10–25 years) and old 0.0 < (>25 years) land uses. The number of observations are Relative change reported in Supplemental Table 1. Diverse agroforestry –0.1 systems appear to be more successful than tree cash crop plantations in maintaining soil organic matter and effective Effective cation-exchange capacity cation-​exchange capacity. Symbols adapted courtesy of 1.5 the Integration and Application Network, University of Maryland Center for Environmental Science (http://ian. 1.0 umces.edu/symbols/).

0.5 Deforestation and soil function 0.0

Relative change Soil functions emerge from a combination of inher- –0.5 ent soil properties (soil texture, soil mineralogy) and dynamic soil properties and processes2 (Fig. 2c), and are Base saturation critical to ecosystem function. Here, we review how 3.0 the drastic changes in dynamic soil properties follow- ing deforestation affect nutrient storage and recycling, 2.0 C storage and GHG emissions, water storage, drainage 1.0 and filtration, and erosion resistance (Fig. 5). 0.0 Nutrient storage and recycling. Nutrient storage and Relative change –1.0 recycling are arguably the dominant soil functions influencing primary production in the humid tropics96, and are strongly linked to multiple dynamic soil proper- Bulk density ties that drastically change with deforestation (Fig. 2a,c). 0.3 In natural, tropical forests97–99, N enters the system 0.2 mainly through N fixation, N cycling and accumu- lation rates are large, and the systems are often not 0.1 N-limited​ 98,100 (Fig. 5). Rock-derived​ nutrients, such as P

0.0 and base cations, are derived from weathering (mainly

Relative change in HAC soils) or dust (mainly in primary-​mineral-​ –0.1 depleted LAC soils)101,102 (Fig. 5). Whereas HAC soils typ- ically have higher soil P availability and less efficient P Soil organic carbon recycling5, the majority of LAC soils have a high capacity 103 0.4 for P sorption (Fig. 5a), explaining the low availabil- ity and efficient recycling of P in many natural, tropical 0.2 forests104 (Fig. 5b). 0.0 During slash-​and-​burn management, as much as 95–98% of N in the slash is volatilized, whereas 27–47% –0.2 of P and 10–48% of base cations can be exported as ash Relative change 11,105 –0.4 particles . However, there is also a significant increase of mineral N, P and base cations in the soil afterwards42, C:N ratio owing to the input from ashes, oxidation of organic material to mineral nutrients from high temperatures 0.2 during burning and high mineralization rates follow- ing clearing37,106,107 (Fig. 5). The pH-​related, transitory 0.0 increase in ECEC following deforestation improves the LAC soils’ capacity to retain base cations11, although –0.2 this increase is not enough to prevent substantial leach- Relative change ing losses following pulse release of nutrients from the burning of slash106. In unfertilized systems, SOM con- tinues to decrease following deforestation, leading to Slash Forest Forest (Fig. 5) Downed vegetation produced <10 years >25 years <10 years >25 years decreasing N and P availability and correspond- 10–25 years 10–25 years 108,109 during slash-and-​ burn​ ing crop yields . In these systems, harvest exports <10 cm 10 to <50 cm 50 to 600 cm management. of N and P typically surpass their inputs, unless there is

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 597 Reviews

a Soils with high-activity clays 1 Forest 2 3 Cropland 4 Secondary forest

b Soils with low-activity clays 1 Forest 2 3 Cropland 4 Secondary forest

Rock-derived nutrients Base-cation-leaching losses Net soil C and GHG emissions Parent material Rooting depth Atmosphere-derived nutrients Nitrogen- leaching losses Net soil C and GHG sink Nutrient and water cycling

Fig. 5 | Deforestation and reforestation impacts on soil functions. losses mirror nutrient cycling and soils increasingly become a net C and a | Land-​use-​change trajectory on high-​activity clay (HAC) soils. Stage 1: GHG sink. b | Land-​use-​change trajectory on low-​activity clay (LAC) soils. tropical natural forests are relatively rich in nitrogen (N) and rock-​derived Stage 1: tropical natural forests on LAC soils are relatively rich in N but low nutrients, and nutrient leaching is substantial. Nutrients and water are in rock-​derived nutrients, with substantial N leaching and low base-cation​ efficiently cycled with large carbon (C) and nutrient stocks in vegetation leaching. Water and rock-derived​ nutrients are efficiently cycled, and C and and soil. Soils are a net nitrous-​oxide source and methane sink. Stage 2: nutrient stocks are largely stored in vegetation. Soils are a net nitrous-​oxide slash-​and-​burn management leads to a net transfer of nutrients from source and a methane sink. Stage 2: after slash-​and-​burn management, vegetation to the soil surface. N leaching is large and base-​cation leaching N leaching is relatively small because of substantial anion-​exchange is moderated by the large effective cation-exchange​ capacity in HAC soils. capacity in the subsoil, whereas base-​cation leaching is relatively large Net soil C and greenhouse gas (GHG) emissions are high. Stage 3: the because of small effective cation-​exchange capacity in LAC soils. Stage 3: relatively long cropping phase with high initial productivity declines with relatively short cropping phases as compared with HAC soils have high time. Nutrient-​leaching losses, net soil C and GHG emissions decrease initial productivity, followed by rapid decline with time (unless fertilized). with years of unfertilized cropping, owing to nutrient uptake and harvest Nutrient-​leaching losses, net soil C and GHG emissions quickly decrease. export by crops and slowly declining soil nutrient stocks. Stage 4: secondary Stage 4: secondary succession occurs slowly, with N limitation during the succession occurs quickly and experiences N limitation during the early early decades. Nutrient and water cycling between vegetation and soil, net years, later mitigated by increased N fixation. Nutrient and water cycling C and GHG sinks in soils increase over a longer timescale than in HAC soils. between vegetation and soil increase with time; consequently, leaching Arrow width represents relative flux size.

598 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

significant nutrient redistribution within a farm caused topsoil, which enhances the likelihood of denitrification- 110 by management, erosion and water movement or related increases in soil N2O emission if N fertilizer is unless N is added by leguminous shade trees43, crops applied126. or plants mixed in pastures48. The gradual decline of ECEC in LAC soils following deforestation (Fig. 3a,c, Water storage, drainage and filtration. Water storage Supplementary Fig. 1a,c) may limit the efficient reten- in soils determines the water supply, which impacts tion of nutrients32. In contrast, the base-cation​ stocks of the length of the growing season, and, as water drains HAC soils are normally sufficient for decades following through soil, it interacts with biogeochemically active deforestation, and weathering of primary minerals can soil components (organic matter, clay minerals), plant replenish lost base cations43 (Supplementary Fig. 1b,d). roots and the soil microbial community. Together, these Diverse agroforestry systems are more successful than components can alter the composition of solutes of the other land uses in retaining soil N, P and other nutrients, drained water — referred to as the soil-filtering​ function although reports vary from no difference in soil-nutrient​ — and can moderate nitrate, phosphate and Al leach- levels compared with the original forests58,111 (observed ing, which are pollution hazards to ground and surface mainly in humid climates) to substantial decreases58,108 waters127. Water storage and drainage both depend on (observed mainly in drier climates). the total porosity and pore-​size distribution of the soil, which are linked to soil depth, texture, structure and Carbon storage and GHG emissions. An estimated BD (ref.40) (Fig. 2c), whereas filtration is related mainly 1,037 Pg C is stored as SOC in the top 3 m of soil in to charge characteristics (such as ECEC; Fig. 2c) and soil tropical biomes (tropical evergreen forests, tropical redox potential32. deciduous forests and tropical savannahs), which is Natural forests on LAC soils have substantial nitrate 44% of the global estimate112. These substantial SOC leaching (9–10 kg N per ha per year, refs11,106,128), stocks are vulnerable to losses with land-​use conver- owing to their large N-​cycling rates, although the sion. Historically, 50 Pg C of SOC losses can be related anion-exchange capacity (AEC) of LAC subsoils129 to deforestation in the tropics, which is 37% of the global (Box 1) can temporarily impede nitrate leaching to estimate of 133 Pg C of the net SOC losses since the onset groundwater130,131 (Fig. 5b). Natural forests on HAC soils 113 of agriculture . This C is released largely as CO2 during have lower N leaching (1–5 kg N per ha per year) and decomposition of labile SOM and combustion dur- mostly as dissolved organic N (ref.132) (Fig. 5a). P leach- 114 11,32 ing slash-​and-​burn ; thus, CO2 emissions are a major ing is typically low in LAC soils independent of land side effect of deforestation115. use due to the P sorption capacity of sesquioxides103.

Other potent GHGs, methane (CH4) and nitrous Base-​cation leaching is lower from natural forests on 11,106 oxide (N2O), are both produced and consumed in trop- LAC soils than on HAC soils because of reservoirs of ical forests by soil microbial processes31,116. Well-drained​ primary minerals in the latter that releases bases through tropical-​forest soils contribute about 25% (6.4 Tg per weathering43 (Fig. 5a,b). 117 year) to the global CH4 sink . Annual CH4 uptake is Deforestation for crops or pastures reduces the lower in soils with high clay contents and during the wet leaf-​area index and the rooting depth, causing reduced season118,119, due to diffusion limitation of atmospheric evapotranspiration133,134 (Fig. 5a,b), increased drainage135 120 CH4 to methanotrophic bacteria in soils . Thus, the and, consequently, increased localized discharge to increase in soil BD following deforestation (Figs 3a,c,4) streams and rivers (for example, a 25% increase of dis- 53 136 reduces soil gas diffusivity , which reduces soil CH4 charge within a local watershed in the Amazon basin) . Denitrification uptake121 or can even turn soils into net CH sources122. In contrast, as a consequence of reduced evapotranspi- Microbial process where 4 − Soil N availability also affects CH4 fluxes, either through ration and moisture recycling, large-​scale deforestation nitrate (NO3 ) is reduced to

NO, N2O and, ultimately, N2. inhibition by mineral N following N fertilizer applica- can decrease the amount of atmospheric moisture and 133,134 tion (caused by the non-​specific behaviour of the CH4 reduce regional precipitation . Slash-​and-​burn Anion-exchange​ capacity monooxygenase enzyme)123 or because methanotrophic management leads to strongly elevated nitrate leach- (AEC). Positively charged sites 121 in the soil that adsorb bacteria are N-limited​ . ing from LAC soils (in one study, 207–344 kg nitrate-​N 106 exchangeable anions. Natural tropical forests are a significant global per ha from the top 25 cm in Brazil) , because of the

N2O source (approximately 27% of terrestrial biome accelerated N-​mineralization and nitrification follow- Sesquioxides sources124), owing to fast N-​cycling rates and high soil ing forest clearing. However, this nitrate leaching can Oxides with three oxygen water contents31. Clearing and burning of tropical for- partly be moderated by their AEC. For example, the atoms for every two atoms of another element, mostly as ests for use as cropland can lead to transitory increases subsoil accumulated 345–1,875 kg nitrate-​N per ha in aluminium oxide (Al2O3) or iron of mineral N, available C and N2O emissions that last fertilized Ferralsols (LAC soils), where the estimated oxide (Fe O ) in soils. 2 3 for several days, followed by elevated N2O emissions nitrate-holding​ capacity ranged from 17 to 32 Mg N per that can last for weeks until crops start competing ha (ref.131). In contrast, HAC soils have low AEC, and Leaf-area​ index 43 One-​sided green leaf area per with N2O-​producing microorganisms for available N substantial nitrate-​leaching losses can occur . In LAC 107 unit ground area, used as a (ref. ). The progressive declines in SOC and availa- soils, the base cations in ashes lead to considerable measure of greenness and ble N during the first years following deforestation70,84 leaching losses during the first year106, which are only vegetation. eventually lead to lower soil N2O fluxes, especially in partly offset by the transitory increase in ECEC fol- unfertilized agricultural systems, relative to the origi- lowing slash-​and-​burn management (Supplementary Nitrification 125 Microbial process where nal forest . However, the increase in soil BD following Fig. 1a,c). The ECEC of HAC soils is inherently higher organic N or ammonia is forest conversion to cropland and pastures (Fig. 3a,c, and not dependent on the pH (Supplementary Fig. 1b,c). oxidized to nitrate. Supplementary Fig. 1) can lead to high water content in the Owing to decreasing SOC with age of land use (Fig. 1a),

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 599 Reviews

which influences the ECEC (Fig. 2a,c), the ability to retain 26% of the mature-​forest biomass within 19 years11. In base cations in both LAC and HAC soils decreases with this section, we review how changes in dynamic soil years. Ultimately, this leads to low base-​cation leach- properties following reforestation affect soil functions ing in LAC soils, as weathering of primary minerals (Figs 2c,5). is low or absent11,32, whereas in HAC soils, base-​cation leaching remains higher, probably due to inputs from Nutrient storage and recycling. During secondary suc- weathering43. cession, soil-nutrient​ storage and recycling increase with time150. For example, soil mineral N levels and foliar N Erosion resistance. Water erosion is often reported follow- contents are very low in <10-​year-​old secondary forest ing deforestation137,138, which can lead to removal of top- following agricultural abandonment on LAC soils, and soil, reducing crop-​production potential and lowering increase over decadal timescales18. Although mobili- surface-water​ quality139; thus, the ability of soils to resist zation of soil organic N has also been reported as an erosion is a crucial soil function. Soil type, properties important N source151, biological N fixation appears to (including SOM content and BD) and geomorpholog- be the main source of N during secondary succession150 ical stability influence the impact of deforestation on and strongly affects the rate at which soil N and SOC soil erosion (Fig. 2c). In many soils (such as HAC soils contents are restored (Fig. 3b,d). The faster rate of N or LAC soils with horizons that impede vertical water accrual in less weathered HAC soils152 compared with flow), reduced vegetation cover and a low soil-aggregate​ heavily weathered LAC soils18,122 suggests that the avail- stability following slash-and-​ ​burn management in both ability of rock-derived​ nutrients important for N fixation croplands140,141 and pastures61,142 increase the suscepti- (such as P and Mo)153 limits the rates at which N cycling bility of surface soils to the impact of raindrops. If the is restored in LAC soils (Fig. 5). Secondary forests on LAC reduced infiltration rates are not sufficient to drain soils possibly meet part of their N demand by accessing high rainfall intensity, overland flow can occur8 and nitrate adsorbed in the subsoil154 (Fig. 5b), a potential cause severe water erosion, especially on sloping or geo- N-​acquisition strategy that has mostly been ignored in morphologically unstable areas137,138,143. However, some studies on secondary succession. deforested LAC Ferralsol or Oxisol soils are relatively Total P stocks in soils under secondary succession resistant to water erosion130. For example, in LAC soils appear to be relatively stable155,156. In LAC soils, total P is in the Amazon, deforestation for soya bean increased mostly in unavailable forms156 and mycorrhizal fungi11,150 BD by 25% (ref.144) and corresponding water-infiltration​ probably stimulate internal P recycling154, a slow pro- rates decreased by almost 70% (ref.130). Nevertheless, cess that could constrain biomass accumulation in forest water erosion was negligible because infiltration rates regrowth157. This process could also cause older second- remained sufficient to drain high rainfall intensity and ary forests on LAC soils to exhibit efficient P cycling, minimize surface run-​off130, alongside geomorpholog- for example, indicated by increasing litterfall N:P ratios ical stability145 and stable microaggregates containing as secondary forests progress18 (Fig. 5b). The strong P sesquioxides146. sorption by Fe and Al (hydr)oxides in highly weath- Serious erosion, ultimately, can remove the organic- ered LAC soils also contributes to the slow recovery of matter-rich​ topsoil. In western Africa, erosion following P cycling11,103. In HAC soils, chemical weathering can deforestation resulted in soil textural changes (6% less contribute substantially to the increase in P availability sand and 40% more gravel), topsoil loss (0.04–0.06 mm during succession43. These observations could partially per year) and a reduction in topsoil water-​storage explain why secondary succession is faster on HAC soils capacity of 15 mm in croplands compared with the compared with LAC soils. original forests135. Erosion-​induced translocation and In contrast with the slow increases in vegetation P burial of SOC and nutrients can be substantial147 — stocks, base-​cation accumulation and restoration via erosion-induced​ continental nutrient fluxes in the trop- throughfall and litterfall in secondary forests occur ics are estimated to be ~15 Tg N per year and ~10 Tg P remarkably quickly, and exceed the input through per year, which are on the same order of magnitude as atmospheric deposition considerably11. In Brazil, a fertilizer applications148. Although SOC burial in deposi- 19-​year-​old secondary forest already displayed cation tional areas can contribute to long-​term SOC storage148, cycling in throughfall and litterfall comparable to that present-​day erosion rates greatly surpass the rate at of a natural forest11, probably the result of the exchangea- which soils are newly formed, making erosion one of ble base cations stored in soils following deforestation158. the most destructive forms of soil degradation149. For example, a degraded pasture soil contains 124 kg K per ha, 654 kg Ca per ha and 146 kg Mg per ha in the Reforestation and soil function topsoil, which are sufficient to support a growing sec- Regeneration of secondary forest and the recovery of ondary forest11. Comparable exchangeable base-​cation dynamic soil properties and soil functions depend on stocks have been reported from other studies68. the soil type, climate and former land-​use intensity, management or duration30,150. In general, the relative Carbon storage and GHG emissions. The important recovery of standing biomass on HAC soils is faster role of secondary forests in C storage and climate reg- than on LAC soils30 — secondary forests can reach ulation is often attributed to above-​ground biomass Water erosion a biomass comparable with natural forests within accumulation. However, secondary forests also restore 81 Removal of soil by water (as 30 years on HAC soils (Fig. 5). In contrast, a secondary SOC (Fig. 3b,d), the rates of which appear to depend opposed to wind, for example). forest on a LAC Ferralsol in Brazil accrues only about on the land-​use type before secondary succession, the

600 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

duration and intensity of the previous land use81,159, and changing in ways that can promote secondary-​forest the nutrient availability for the regrowing forest18,160,161. regrowth. Owing to their inherent difference in soil fer- Indeed, within a few decades, SOC stocks in the top- tility, relative recovery in biomass of secondary forests on soils of secondary forests can be restored to levels close HAC soils is faster than in LAC soils30 (Fig. 5). Regardless, to those of the original forests159,162. Restoration can important differences remain between soils under take considerably longer for subsoils (Fig. 3b,d), which mature secondary forests and soils under old-​growth is in line with recent radiocarbon measurements of forest, and full restoration to the original forests’ soil deep soils163. properties requires more than just a few decades, 150 (Fig. 3) Changes in soil CH4 fluxes during secondary suc- especially in the subsoils . cession depend mainly on soil water content and BD 122,164 restoration . In secondary forests, CH4 uptake is Long-term​ impacts typically higher than in the previous land uses, owing There is little information available on the long-​term to decreases in BD and soil water content164. In mature effects of deforestation on soil properties and soil func-

secondary forests, CH4 uptake is restored to pre-clearing​ tions in the tropics. Most of the studies that we discussed levels122. Young secondary forests are typically N-limited​ used space-​for-​time substitution, a method that is typi-

and exhibit low N2O fluxes that are only a small frac- cally used for land uses that are several decades old at the tion compared with those of natural forests18. As sec- maximum, as there is only limited landowner memory ondary forests progress, soil mineral N and foliar N when collecting information such as time of clearing, 18 168 concentrations increase , and soil N2O emissions management and previous land uses . Furthermore, the slowly restore to pre-​clearing levels as N-​cycling pro- original reference land use (natural forest) is often no cesses recuperate, which can take 40 years or more of longer found in the region, especially when deforestation LAC soils18,122. happened several decades ago. Studies that examine centuries-​old changes nor- Water storage, drainage and filtration. During secondary mally derive information indirectly, such as from succession or reforestation, the diminished water cycling archaeological studies22 or from sediment cores of between soil and vegetation following deforestation134 is lakes169 or oceans170. The Yucatán Peninsula (home to normally reversed, exhibiting increased evapotranspira- large, ancient Mayan populations) is probably one of tion and decreased run-off​ to streams165 (Fig. 5). If soil BD the best-​studied regions that faced substantial deforest- following reforestation eventually decreases (although ation followed by extended periods of forest regrowth not observed within 25 years of secondary succession (for the past millennium, approximately). Studies show in the studies reviewed here; Fig. 3b,d) and water-​ that the long-​lasting effects of soil erosion associated infiltration rates increase, total run-​off will be reduced with ancient Mayan systems can still be measured and peak run-​off rates will be moderated166. Overall, on formerly eroded hillslopes, as shown by buried water will recycle more efficiently through evapotran- soils, soil base-​cation contents, 13C isotopic evidence

spiration in densely vegetated secondary forest than on of C4 crops such as maize and evidence of young soils agricultural land133,134. formed since abandonment171. Furthermore, radiocar- During secondary succession, nutrients (N, P and bon ages of plant waxes in sediment cores from this base cations) stored in biomass and their recycling region suggest short turnover times of SOC associated through litterfall increase11,150. Moreover, nutrients with Mayan deforestation. Even 1,000 years of second- closely linked to accrual of SOM or redistributed by ary succession did not recover SOC turnover times to fine roots from deep-​soil to topsoil depths19,131 can also pre-​deforestation values63, and the authors suggested increase in availability with succession18, and, hence, lead that prolonged periods of reforestation are needed to to an increase in dissolved N leaching from young to old accumulate slow-​cycling SOC pools, especially in the secondary forests on LAC soils, as seen in the Amazon11. subsoils. How general these long-​term effects are is However, the base-cation-​ rich​ ash inputs in pastures on presently unclear, as the Yucatán Peninsula is dominated LAC soils (Fig. 3c, Supplementary Fig. 1c) result in high by HAC soils (Fig. 1b) and long-​term SOC stabilization base-​cation leaching; under secondary forests, uptake mechanisms in other regions dominated by LAC soils by and litterfall from trees reequilibrate (or recycle) are probably quite different. the base cations to the acidic condition of LAC soils, and, In the past decades, it has become apparent that ultimately, decrease base-​cation leaching as secondary extensive areas that are nowadays considered old-growth​ forests regrow11 (Fig. 5b). forests were used or modified in the past by human 22 Space-for-​ time​ substitution activity . These impacts extend to their soils: evidence Studying ecological processes Erosion resistance. Reforestation increases erosion of prehistoric intensive drainage systems in the Bolivian at different aged sites, resistance, owing to the establishment of dense vegeta- Amazon172; the pre-​Columbian formation of fertile soil assumed to represent different 138,165 stages of developments; used tion cover , and the soil-​aggregate stability in sec- patches rich in organic matter in parts of the Amazon 61 173,174 especially in studies of ondary forests improves compared with pastures and basin (terra preta) ; and indications of intensi- long-term​ processes. croplands140. Notably, tree plantations normally have fied weathering related to deforestation in Iron Age similar effects on soil erosion as secondary forests, unless Central Africa170. Thus, many soils under old-​growth C crops 4 removal of understory vegetation in tree plantations forests are probably not as ‘pristine’ as was previously Plants that use the C4 167 carbon-fixation​ pathway, as increases soil crusting and the risk of erosion . believed. The soils under such forests are also witness opposed to the C3 In summary, dynamic soil properties and soil func- to their remarkable resilience, once human interference carbon-fixation​ pathway. tions are strongly affected during secondary succession, is discontinued.

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 601 Reviews

181 Kaolinitic Perspectives comprehensive simulation models . For example, in Layered silicate clays formed New studies that include older land-​use systems and land surface models, SOC and nutrient cycling of soils through the weathering of deeper soil depths now demonstrate unequivocally that in the tropics182 is often still based on the Century model. aluminium silicates with the dynamic soil properties continue to change for several However, Century was developed and calibrated for soils formula Al Si O (OH) . 2 2 5 4 decades following deforestation. These changes are not of the Great Plains of North America59, where SOC and Gibbsite limited to the top 0.3 m or possibly even the top 1 m nutrient levels and dynamics are governed by distinctly An aluminium-hydroxide​ (refs38,58) but probably reach as deep as the soil itself, different climate, vegetation and parent material183 than mineral, with the formula easily several metres for LAC soils in the tropics175. in the tropics. Al(OH) . 3 Furthermore, these changes are not limited to SOC Nevertheless, substantial advances have been made contents but also include many other dynamic soil in understanding the long-​term impacts of tropical properties (Figs 3,4) and soil functions (Fig. 2c). land-use​ change on soils. Interdisciplinary deforestation Important knowledge gaps remain. Topography and studies that include soil as an important provider of eco- landforms, which influence water and element redis- system functions17,20 and as a moderator of ecological tribution within a landscape, are rarely considered in shifts following deforestation12 have started to emerge. studies on deforestation and reforestation176,177. Very In one notable study, agronomic experiments are related little is known about the mechanisms of SOC stabiliza- to nutrient depletion following deforestation109. Studies tion and their interactions. Furthermore, our knowledge of the ancient Mayan region, where ancient erosion of soil microbially mediated and chemical processes in patterns are moderating modern tree distribution171 subsoils (especially those involving charge character- and where radiocarbon ages of plant waxes in sedi- istics) is limited. Better knowledge of these processes ment cores are connected to SOC stability, give exciting is needed to understand and manage nutrient storage new insights. and cycling, soil C sequestration and GHG produc- Finally, we put forward some recommendations based tion and consumption120. In addition, clay mineralogy, on our emerging understanding of the long-term​ effects normally considered an inherent soil property, could of deforestation and reforestation on dynamic soil prop- be more dynamic following land-​use changes than erties and functions. First, investigations in deforested commonly assumed. Tropical forests act as a biologi- or reforested soils should include subsoils (and the full cal pump178, resulting in kaolinitic clay minerals instead soil profile when possible) and land-use​ systems that are of gibbsite in the topsoils as the typical end product of several decades old, and have a stronger focus on refor- tropical weathering179,180, with important implica- ested soils, which are studied less than deforested soils tions for soil charge characteristics and associated soil (Supplemental Table 1). Soils should also be better rep- functions (Fig. 2c). resented in land surface models — for example, there are While soil scientists have gained novel insight now better estimates of SOC turnover rates38,184, includ- into the effects of deforestation and reforestation on ing from subsoils based on radiocarbon measurements163, soils in the past decades, this knowledge has often than the default values given by models such as the not yet reached cross-​disciplinary efforts to address Century model. Second, almost all available data in our deforestation and reforestation in the tropics. The Review originated from studies using space-​for-​time limited cross-​disciplinary collaboration is probably substitution, and there are no long-​term longitudinal best illustrated by the poor representation of soils in experiments on deforestation and reforestation that include soil processes and functions. Such studies would Box 2 | Societal impacts of soil management greatly enhance our understanding on dynamic changes in vegetation–soil interactions. Third, there is a need In the tropics, forests and savannah are cleared for agriculture (despite increases in productivity of tropical agriculture in the past decades193), owing to policies addressing for multi-​taxa analyses of soil biological communities land-ownership​ redistribution194, transmigration195 and other social issues. Many of (macro­fauna, microfauna and microbial composition) these permanent agricultural systems lead to decreases in dynamic soil properties and with their linkages to soil properties and functions, based functions, which can, ultimately, lead to marginal crop yields109 and severe land on comparable experimental designs, which are particu- degradation149. Soil organic matter (SOM) could be targeted to counteract such larly missing for reforestation. Such investigation is ongo- declines, as it is the only dynamic soil property that is central to all soil functions and ing in forest-to-​ plantation​ conversions in Indonesia12, and processes. The highest SOM losses occur in croplands, followed by tree cash crop now also includes plantation-​management practices185, plantations, pastures and agroforestry. Improved SOM management in croplands can although relating soil biological changes to soil func- 196,197 196 include minimal tillage , inclusion of a fallow and/or legume rotation , and return tions is still not fully explored. Finally, basic soil physical, of mulch or crop residues without burning198. Maintenance of understory vegetation chemical and biological data should be placed in online helps to maintain SOM levels in monoculture tree plantations, as does the return of organic waste199. In pastures, improved SOM management includes the use of data repositories alongside their publication. Although deep-rooting​ grass species73, often in combination with legumes48, and appropriate increasingly required by journals, such data deposition animal stocking rates adjusted to pasture productivity74. The resilience of diverse is not yet the standard. Given the critical role that soils agroforestry systems against SOM losses58,77 is likely the result of a combination of play following deforestation and secondary succession, organic inputs near homesteads108, diverse plant life forms that increase nutrient-use​ and the potential consequences for society (Box 2), it is efficiencies79, permanent rooting systems and the inclusion of N-​fixing trees43. Thus, the imperative that soil scientific knowledge is better com- agricultural techniques for improved management exist. However, farmers are more municated to colleagues in other disciplines, such as likely to plant trees or invest in improved SOM and nutrient management if they have a ecologists, anthropologists, climatologists, archaeologists secure tenure, access to markets to sell their produce, access to knowledge and the and economists. financial means to make their management profitable200. Policies that support these prerequisites are critical steps towards better adjusted land-use​ systems in the tropics. Published online 15 September 2020

602 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

1. Koch, A. et al. Soil security: solving the global soil Assessment 2015. For. Ecol. Manag. 352, 9–20 53. Chauvel, A., Grimaldi, M. & Tessier, D. Changes in soil crisis. Glob. Policy 4, 434–441 (2013). (2015). pore-space​ distribution following deforestation and 2. Vogel, H.-J. et al. A systemic approach for modeling 28. Busch, J. et al. Potential for low-​cost carbon dioxide revegetation: an example from the Central Amazon soil functions. SOIL 4, 83–92 (2018). removal through tropical reforestation. Nat. Clim. Basin, Brazil. For. Ecol. Manag. 38, 259–271 (1991). 3. Corre, M. D., Veldkamp, E., Arnold, J. & Wright, S. J. Change 9, 463–466 (2019). 54. Kayombo, B. & Lal, R. Effects of soil compaction by Impact of elevated N input on soil N cycling and losses 29. Morales-​Hidalgo, D., Oswalt, S. N. & Somanathan, E. rolling on soil structure and development of maize in in old-​growth lowland and montane forests in Panama. Status and trends in global primary forest, protected no-​till and disc ploughing systems on a tropical alfisol. Ecology 91, 1715–1729 (2010). areas, and areas designated for conservation of Soil. Tillage Res. 7, 117–134 (1986). 4. Cusack, D. F., Markesteijn, L., Condit, R., Lewis, O. T. biodiversity from the Global Forest Resources 55. Lal, R. Effects of macrofauna on soil properties in & Turner, B. L. Soil carbon stocks across tropical Assessment 2015. For. Ecol. Manag. 352, 68–77 tropical ecosystems. Agric. Ecosyst. Environ. 24, forests of Panama regulated by base cation effects on (2015). 101–116 (1988). fine roots. Biogeochemistry 137, 253–266 (2018). 30. Poorter, L. et al. Biomass resilience of Neotropical 56. Ghuman, B. S., Lal, R. & Shearer, W. Land clearing and 5. Vitousek, P. M. & Sanford, R. L. Nutrient cycling secondary forests. Nature 530, 211–214 (2016). use in the humid Nigerian tropics: I. Soil physical in moist tropical forest. Annu. Rev. Ecol. Syst. 17, 31. Davidson, E. A., Keller, M., Erickson, H. E., Verchot, L. V. properties. Soil Sci. Soc. Am. J. 55, 178–183 (1991). 137–167 (1986). & Veldkamp, E. Testing a conceptual model of soil 57. Minasny, B. & Hartemink, A. E. Predicting soil 6. Powers, J. S., Corre, M. D., Twine, T. E. & Veldkamp, E. emissions of nitrous and nitric oxides. BioScience 50, properties in the tropics. Earth Sci. Rev. 106, 52–62 Geographic bias of field observations of soil carbon 667–680 (2000). (2011). stocks with tropical land-​use changes precludes 32. Kurniawan, S. et al. Conversion of tropical forests to 58. Hombegowda, H. C., van Straaten, O., Köhler, M. spatial extrapolation. Proc. Natl Acad. Sci. USA 108, smallholder rubber and oil palm plantations impacts & Hölscher, D. On the rebound: soil organic carbon 6318–6322 (2011). nutrient leaching losses and nutrient retention stocks can bounce back to near forest levels when 7. Don, A., Schumacher, J. & Freibauer, A. Impact of efficiency in highly weathered soils. Biogeosciences agroforests replace agriculture in southern India. SOIL tropical land-use​ change on soil organic carbon stocks 15, 5131–5154 (2018). 2, 13–23 (2016). - a meta-​analysis. Glob. Change Biol. 17, 1658–1670 33. Detwiler, R. P. Land use change and the global carbon 59. Parton, W. J., Stewart, J. W. B. & Cole, C. V. Dynamics (2011). cycle: the role of tropical soils. Biogeochemistry 2, of C, N, P and S in grassland soils: a model. 8. Chaves, J. et al. Land management impacts on runoff 67–93 (1986). Biogeochemistry 5, 109–131 (1988). sources in small Amazon watersheds. Hydrol. Process. 34. Guo, L. B. & Gifford, R. M. Soil carbon stocks and land 60. López-​Ulloa, M., Veldkamp, E. & de Koning, G. H. J. 22, 1766–1775 (2008). use change: a meta analysis. Glob. Change Biol. 8, Soil carbon stabilization in converted tropical pastures 9. Nepstad, D. C. et al. The role of deep roots in the 345–360 (2002). and forests depends on soil type. Soil Sci. Soc. Am. J. hydrological and carbon cycles of Amazonian forests 35. Davidson, E. A. & Ackerman, I. L. Changes in soil 69, 1110–1117 (2005). and pastures. Nature 372, 666–669 (1994). carbon inventories following cultivation of previously 61. Paul, S., Flessa, H., Veldkamp, E. & López-​Ulloa, M. 10. Elsenbeer, H. Hydrologic flowpaths in tropical untilled soils. Biogeochemistry 20, 161–193 (1993). Stabilization of recent soil carbon in the humid tropics rainforest soilscapes-​a review. Hydrol. Process. 15, 36. Veldkamp, E. Organic carbon turnover in three tropical following land use changes: evidence from aggregate 1751–1759 (2001). soils under pasture after deforestation. Soil Sci. Soc. fractionation and stable isotope analyses. 11. Markewitz, D., Davidson, E., Moutinho, P. & Am. J. 58, 175–180 (1994). Biogeochemistry 87, 247–263 (2008). Nepstad, D. Nutrient loss and redistribution after 37. Nye, P. H. & Greenland, D. J. Changes in the soil after 62. Amundson, R. The carbon budget in soils. Annu. Rev. forest clearing on a highly weathered soil in Amazonia. clearing tropical forest. Plant Soil 21, 101–112 (1964). Earth Planet. Sci. 29, 535–562 (2001). Ecol. Appl. 14, 177–199 (2004). 38. van Straaten, O. et al. Conversion of lowland tropical 63. Douglas, P. M. J. et al. A long-​term decrease in the 12. Barnes, A. D. et al. Direct and cascading impacts of forests to tree cash crop plantations loses up to persistence of soil carbon caused by ancient Maya tropical land-use​ change on multi-​trophic biodiversity. one-half of stored soil organic carbon. Proc. Natl Acad. land use. Nat. Geosci. 11, 645–649 (2018). Nat. Ecol. Evol. 1, 1511–1519 (2017). Sci. USA 112, 9956–9960 (2015). 64. Marín-​Spiotta, E. & Sharma, S. Carbon storage in 13. Clark, D. B., Palmer, M. W. & Clark, D. A. Edaphic 39. Tugel, A. J. et al. Soil change, soil survey, and natural successional and plantation forest soils: a tropical factors and the landscape-​scale distributions of resources decision making. Soil Sci. Soc. Am. J. 69, analysis. Glob. Ecol. Biogeogr. 22, 105–117 (2013). tropical rain forest trees. Ecology 80, 2662–2675 738–747 (2005). 65. Trumbore, S. E., Davidson, E. A., Barbosa de (1999). 40. Sanchez, P. A. Properties and Management of Soils in Camargo, P., Nepstad, D. C. & Martinelli, L. A. 14. Jones, M. M. et al. Explaining variation in tropical the Tropics 2nd edn (Cambridge Univ. Press, 2019). Belowground cycling of carbon in forests and pastures plant community composition: influence of 41. van Breemen, N., Mulder, J. & Driscoll, C. T. of eastern Amazonia. Glob. Biogeochem. Cycles 9, environmental and spatial data quality. Oecologia Acidification and alkalinization of soils. Plant Soil 75, 515–528 (1995). 155, 593–604 (2008). 283–308 (1983). 66. Veldkamp, E., Becker, A., Schwendenmann, L., 15. Canadell, J. et al. Maximum rooting depth of 42. Andriesse, J. P. & Schelhaas, R. M. A monitoring study Clark, D. A. & Schulte-​Bisping, H. Substantial labile vegetation types at the global scale. Oecologia 108, on nutrient cycles in soils used for shifting cultivation carbon stocks and microbial activity in deeply weathered 583–595 (1996). under various climatic conditions in tropical Asia. III. soils below a tropical wet forest. Glob. Change Biol. 9, 16. Barnes, A. D. et al. Consequences of tropical land use The effects of land clearing through burning on 1171–1184 (2003). for multitrophic biodiversity and ecosystem functioning. fertility level. Agric. Ecosyst. Environ. 19, 311–332 67. Intergovernmental Panel on Climate Change (IPCC) Nat. Commun. 5, 5351 (2014). (1987). 2019 Refinement to the 2006 IPCC Guidelines for 17. Grass, I. et al. Trade-​offs between multifunctionality 43. Dechert, G., Veldkamp, E. & Brumme, R. Are partial National Greenhouse Gas Inventories (IPCC, 2019). and profit in tropical smallholder landscapes. nutrient balances suitable to evaluate nutrient 68. Reiners, W. A., Bouwman, A. F., Parsons, W. F. J. Nat. Commun. 11, 1186 (2020). sustainability of land use systems? Results from a case & Keller, M. Tropical rain forest conversion to pasture: 18. Davidson, E. A. et al. Recuperation of nitrogen cycling study in Central Sulawesi, Indonesia. Nutr. Cycling changes in vegetation and soil properties. Ecol. Appl. in Amazonian forests following agricultural Agroecosyst. 72, 201–212 (2005). 4, 363–377 (1994). abandonment. Nature 447, 995–998 (2007). 44. Neill, C. et al. Soil carbon and nitrogen stocks 69. Hamer, U., Potthast, K., Burneo, J. I. & Makeschin, F. 19. Jobbágy, E. G. & Jackson, R. B. The distribution of soil following forest clearing for pasture in the Nutrient stocks and phosphorus fractions in mountain nutrients with depth: global patterns and the imprint southwestern Brazilian Amazon. Ecol. Appl. 7, soils of Southern Ecuador after conversion of forest to of plants. Biogeochemistry 53, 51–77 (2001). 1216–1225 (1997). pasture. Biogeochemistry 112, 495–510 (2013). 20. Clough, Y. et al. Land-​use choices follow profitability 45. Allen, K., Corre, M. D., Kurniawan, S., Utami, S. R. 70. Veldkamp, E., Davidson, E., Erickson, H., Keller, M. at the expense of ecological functions in Indonesian & Veldkamp, E. Spatial variability surpasses land-​use & Weitz, A. Soil nitrogen cycling and nitrogen oxide smallholder landscapes. Nat. Commun. 7, 13137 change effects on soil biochemical properties of emissions along a pasture chronosequence in the (2016). converted lowland landscapes in Sumatra, Indonesia. humid tropics of Costa Rica. Soil Biol. Biochem. 31, 21. Lewis, S. L., Edwards, D. P. & Galbraith, D. Increasing Geoderma 284, 42–50 (2016). 387–394 (1999). human dominance of tropical forests. Science 349, 46. Carlson, K. M. et al. Effect of oil palm sustainability 71. Wick, B., Veldkamp, E., de Mello, W. Z., Keller, M. 827–832 (2015). certification on deforestation and fire in Indonesia. & Crill, P. Nitrous oxide fluxes and nitrogen cycling 22. Roberts, P., Hunt, C., Arroyo-​Kalin, M., Evans, D. Proc. Natl Acad. Sci. USA 115, 121–126 (2018). along a pasture chronosequence in Central Amazonia, & Boivin, N. The deep human prehistory of global 47. Sanchez P. A. & Logan T. J. Myths and Science About Brazil. Biogeosciences 2, 175–187 (2005). tropical forests and its relevance for modern the Chemistry and Fertility of Soils in the Tropics (Soil 72. van Dam, D., van Breemen, N. & Veldkamp, E. Soil conservation. Nat. Plants 3, 17093 (2017). Science Society of America and American Society of organic carbon dynamics: variability with depth in 23. Ribeiro Filho, A. A., Adams, C., Manfredini, S., Agronomy, 1992). forested and deforested soils under pasture in Costa Aguilar, R. & Neves, W. A. Dynamics of soil chemical 48. Stahl, C. et al. Continuous soil carbon storage of old Rica. Biogeochemistry 39, 343–375 (1997). properties in shifting cultivation systems in the tropics: permanent pastures in Amazonia. Glob. Change Biol. 73. Fisher, M. J. et al. Carbon storage by introduced a meta-analysis.​ Soil Use Manag. 31, 474–482 23, 3382–3392 (2017). deep-rooted grasses in the South American savannas. (2015). 49. Bautista-​Cruz, A. & del Castillo, R. F. Soil changes Nature 371, 236–238 (1994). 24. Jarosz, L. Defining and explaining tropical deforestation: during secondary succession in a tropical montane 74. Navarrete, D., Sitch, S., Aragão, L. E. O. C. & shifting cultivation and population growth in colonial cloud forest area. Soil Sci. Soc. Am. J. 69, 906–914 Pedroni, L. Conversion from forests to pastures in the Madagascar (1896–1940). Econ. Geogr. 69, 366–379 (2005). Colombian Amazon leads to contrasting soil carbon (1993). 50. Marin-​Spiotta, E., Silver, W. L., Swanston, C. W. & dynamics depending on land management practices. 25. Lambin, E. F. et al. The causes of land-​use and Ostertag, R. Soil organic matter dynamics during Glob. Change Biol. 22, 3503–3517 (2016). land-cover change: moving beyond the myths. 80 years of reforestation of tropical pastures. 75. Chiti, T., Grieco, E., Perugini, L., Rey, A. & Valentini, R. Glob. Environ. Change 11, 261–269 (2001). Glob. Change Biol. 15, 1584–1597 (2009). Effect of the replacement of tropical forests with tree 26. Rudel, T. K., Defries, R., Asner, G. P. & Laurance, W. F. 51. Silver, W. L. et al. Effects of soil texture on plantations on soil organic carbon levels in the Jomoro Changing drivers of deforestation and new opportunities belowground carbon and nutrient storage in a district, Ghana. Plant Soil 375, 47–59 (2014). for conservation. Conserv. Biol. 23, 1396–1405 lowland Amazonian forest ecosystem. Ecosystems 3, 76. Kirsten, M., Kimaro, D. N., Feger, K.-H. & Kalbitz, K. (2009). 193–209 (2000). Impact of land use on soil organic carbon stocks in the 27. Keenan, R. J. et al. Dynamics of global forest area: 52. Oades, J. & Waters, A. Aggregate hierarchy in soils. humid tropics of NE Tanzania. J. Plant. Nutr. Soil Sci. results from the FAO Global Forest Resources Soil Res. 29, 815–828 (1991). 182, 625–636 (2019).

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 603 Reviews

77. Kassa, H., Dondeyne, S., Poesen, J., Frankl, A. & 100. Kaspari, M. & Powers, J. S. Biogeochemistry and 124. Tian, H. et al. Global methane and nitrous oxide Nyssen, J. Impact of deforestation on soil fertility, soil geographical ecology: Embracing all twenty-​five emissions from terrestrial ecosystems due to multiple carbon and nitrogen stocks: the case of the Gacheb elements required to build organisms. Am. Nat. 188, environmental changes. Ecosyst. Health Sustain. 1, catchment in the White Nile Basin, Ethiopia. Agric. S62–S73 (2016). 1–20 (2015). Ecosyst. Environ. 247, 273–282 (2017). 101. Kennedy, M. J., Chadwick, O. A., Vitousek, P. M., 125. Veldkamp, E., Purbopuspito, J., Corre, M. D., 78. Dechert, G., Veldkamp, E. & Anas, I. Is soil Derry, L. A. & Hendricks, D. M. Changing sources of Brumme, R. & Murdiyarso, D. Land use change effects degradation unrelated to deforestation? Examining base cations during ecosystem development, Hawaiian on trace gas fluxes in the forest margins of Central soil parameters of land use systems in upland Central Islands. Geology 26, 1015–1018 (1998). Sulawesi, Indonesia. J. Geophys. Res. Biogeosci. 113, Sulawesi, Indonesia. Plant Soil 265, 197–209 (2004). 102. Bristow, C. S., Hudson-​Edwards, K. A. & Chappell, A. G02003 (2008). 79. Hiremath, A. J. & Ewel, J. J. Ecosystem nutrient use Fertilizing the Amazon and equatorial Atlantic with 126. Weitz, A. M., Linder, E., Frolking, S., Crill, P. M.

efficiency, productivity, and nutrient accrual in model West African dust. Geophys. Res. Lett. 37, L14807 & Keller, M. N2O emissions from humid tropical tropical communities. Ecosystems 4, 669–682 (2001). (2010). agricultural soils: effects of soil moisture, texture 80. Pineiro, G., Oesterheld, M., Batista, W. B. & 103. Bortoluzzi, E. C., Pérez, C. A. S., Ardisson, J. D., and nitrogen availability. Soil. Biol. Biochem. 33, Paruelo, J. M. Opposite changes of whole-​soil vs. Tiecher, T. & Caner, L. Occurrence of iron and 1077–1093 (2001). pools C:N ratios: a case of Simpson’s paradox with aluminum sesquioxides and their implications for the 127. Keesstra, S. et al. Soil as a filter for groundwater implications on nitrogen cycling. Glob. Change Biol. P sorption in subtropical soils. Appl. Clay Sci. 104, quality. Curr. Opin. Environ. Sustain. 4, 507–516 12, 804–809 (2006). 196–204 (2015). (2012). 81. de Koning, G. H. J., Veldkamp, E. & López-​Ulloa, M. 104. Hedin, L. O., Vitousek, P. M. & Matson, P. A. Nutrient 128. Schwendenmann, L. & Veldkamp, E. The role of Quantification of carbon sequestration in soils losses over four million years of tropical forest dissolved organic carbon, dissolved organic nitrogen, following pasture to forest conversion in northwestern development. Ecology 84, 2231–2255 (2003). and dissolved inorganic nitrogen in a tropical wet Ecuador. Glob. Biogeochem. Cycles 17, 1098 (2003). 105. Mackensen, J., Hölscher, D., Klinge, R. & Fölster, H. forest ecosystem. Ecosystems 8, 339–351 (2005). 82. Silver, W. L., Ostertag, R. & Lugo, A. E. The potential Nutrient transfer to the atmosphere by burning of 129. Lehmann, J. et al. Subsoil retention of organic and for carbon sequestration through reforestation of debris in eastern Amazonia. For. Ecol. Manag. 86, inorganic nitrogen in a Brazilian savanna Oxisol. abandoned tropical agricultural and pasture lands. 121–128 (1996). Soil Use Manag. 20, 163–172 (2004). Restor. Ecol. 8, 394–407 (2000). 106. Klinge, R., Araujo Martins, A. R., Mackensen, J. & 130. Neill, C. et al. Watershed responses to Amazon soya 83. Krashevska, V. et al. Micro-​decomposer communities Fölster, H. Element loss on rain forest conversion in bean cropland expansion and intensification. Philos. and decomposition processes in tropical lowlands as East Amazonia: comparison of balances of stores and Trans. R. Soc. B Biol. Sci. 368, 20120425 (2013). affected by land use and litter type. Oecologia 187, fluxes. Biogeochemistry 69, 63–82 (2004). 131. Rasiah, V. & Armour, J. D. Nitrate accumulation under 255–266 (2018). 107. Weitz, A. M., Veldkamp, E., Keller, M., Neff, J. cropping in the Ferrosols of Far North Queensland wet 84. Allen, K., Corre, M. D., Tjoa, A. & Veldkamp, E. Soil & Crill, P. M. Nitrous oxide, nitric oxide, and methane tropics. Aust. J. Soil Res. 39, 329–341 (2001). nitrogen-​cycling responses to conversion of lowland fluxes from soils following clearing and burning of 132. Goller, R., Wilcke, W., Fleischbein, K., Valarezo, C. forests to oil palm and rubber plantations in Sumatra, tropical secondary forest. J. Geophys. Res. Atmos. & Zech, W. Dissolved nitrogen, phosphorus, and sulfur Indonesia. PLoS ONE 10, e0133325 (2015). 103, 28047–28058 (1998). forms in the ecosystem fluxes of a montane forest in 85. Brinkmann, N. et al. Intensive tropical land use 108. Moebius-Clune,​ B. N. et al. Long-​term soil quality Ecuador. Biogeochemistry 77, 57–89 (2006). massively shifts soil fungal communities. Sci. Rep. 9, degradation along a cultivation chronosequence in 133. Aragão, L. E. O. C. The rainforest’s water pump. 3403 (2019). western Kenya. Agric. Ecosyst. Environ. 141, 86–99 Nature 489, 217–218 (2012). 86. Berkelmann, D. et al. How rainforest conversion to (2011). 134. Spracklen, D. V., Arnold, S. R. & Taylor, C. M. agricultural systems in Sumatra (Indonesia) affects 109. Ngoze, S. et al. Nutrient constraints to tropical Observations of increased tropical rainfall preceded active soil bacterial communities. Front. Microbiol. 9, agroecosystem productivity in long-​term degrading by air passage over forests. Nature 489, 282–285 2381 (2018). soils. Glob. Change Biol. 14, 2810–2822 (2008). (2012). 87. Schneider, D. et al. Impact of lowland rainforest 110. Haileslassie, A., Priess, J. A., Veldkamp, E. & 135. Giertz, S., Junge, B. & Diekkrüger, B. Assessing the transformation on diversity and composition of soil Lesschen, J. P. Nutrient flows and balances at the field effects of land use change on soil physical properties prokaryotic communities in Sumatra (Indonesia). and farm scale: Exploring effects of land-​use strategies and hydrological processes in the sub-​humid tropical Front. Microbiol. 6, 1339 (2015). and access to resources. Agric. Syst. 94, 459–470 environment of West Africa. Phys. Chem. Earth Parts 88. Janos, D. P. Mycorrhizae influence tropical succession. (2007). A/B/C 30, 485–496 (2005). Biotropica 12, 56–64 (1980). 111. Kassa, H., Dondeyne, S., Poesen, J., Frankl, A. & 136. Davidson, E. A. et al. The Amazon basin in transition. 89. Bachelot, B. et al. Associations among arbuscular Nyssen, J. Agro-​ecological implications of forest and Nature 481, 321–328 (2012). mycorrhizal fungi and seedlings are predicted to agroforestry systems conversion to cereal-​based 137. Kassa, H., Frankl, A., Dondeyne, S., Poesen, J. change with tree successional status. Ecology 99, farming systems in the White Nile Basin, Ethiopia. & Nyssen, J. Sediment yield at southwest Ethiopia’s 607–620 (2018). Agroecol. Sustain. Food Syst. 42, 149–168 (2018). forest frontier. Land Degrad. Dev. 30, 695–705 90. Gei, M. et al. Legume abundance along successional 112. Jobbágy, E. G. & Jackson, R. B. The vertical distribution (2019). and rainfall gradients in Neotropical forests. Nat. Ecol. of soil organic carbon and its relation to climate and 138. Molina, A., Vanacker, V., Balthazar, V., Mora, D. & Evol. 2, 1104–1111 (2018). vegetation. Ecol. Appl. 10, 423–436 (2000). Govers, G. Complex land cover change, water and 91. Ostertag, R., Marín-​Spiotta, E., Silver, W. L. & 113. Sanderman, J., Hengl, T. & Fiske, G. J. Soil carbon sediment yield in a degraded Andean environment. Schulten, J. Litterfall and decomposition in relation debt of 12,000 years of human land use. Proc. Natl J. Hydrol. 472–473, 25–35 (2012). to soil carbon pools along a secondary forest Acad. Sci. USA 114, 9575–9580 (2017). 139. Labrière, N., Locatelli, B., Laumonier, Y., Freycon, V. chronosequence in Puerto Rico. Ecosystems 11, 114. Drake, T. W. et al. Mobilization of aged and biolabile & Bernoux, M. Soil erosion in the humid tropics: 701–714 (2008). soil carbon by tropical deforestation. Nat. Geosci. 12, A systematic quantitative review. Agric. Ecosyst. 92. Cole, R. J., Selmants, P., Khan, S. & Chazdon, R. 541–546 (2019). Environ. 203, 127–139 (2015). Litter dynamics recover faster than arthropod 115. Meijide, A. et al. Measured greenhouse gas budgets 140. Islam, K. R. & Weil, R. R. Land use effects on soil biodiversity during tropical forest succession. challenge emission savings from palm-​oil biodiesel. quality in a tropical forest ecosystem of Bangladesh. Biotropica 52, 22–33 (2020). Nat. Commun. 11, 1089 (2020). Agric. Ecosyst. Environ. 79, 9–16 (2000). 93. Zou, X. & Gonzalez, G. Changes in earthworm density 116. Conrad, R. Microbial ecology of methanogens and 141. Le Bissonnais, Y. et al. Soil aggregate stability in and community structure during secondary succession methanotrophs. Adv. Agron. 96, 1–63 (2007). Mediterranean and tropical agro-​ecosystems: effect in abandoned tropical pastures. Soil Biol. Biochem. 117. Dutaur, L. & Verchot, L. V. A global inventory of the of plant roots and soil characteristics. Plant Soil 424,

29, 627–629 (1997). soil CH4 sink. Glob. Biogeochem. Cycles 21, GB4013 303–317 (2018). 94. Stone, M. J., Shoo, L., Stork, N. E., Sheldon, F. & (2007). 142. Garcıa-́Oliva,​ F., Sanford, R. L. & Kelly, E. Effects of Catterall, C. P. Recovery of decomposition rates 118. Veldkamp, E., Koehler, B. & Corre, M. D. Indications slash-​and-burn management on soil aggregate organic and decomposer invertebrates during rain forest of nitrogen-limited​ methane uptake in tropical forest C and N in a tropical deciduous forest. Geoderma 88, restoration on disused pasture. Biotropica 52, soils. Biogeosciences 10, 5367–5379 (2013). 1–12 (1999). 230–241 (2020). 119. Matson, A. L., Corre, M. D., Langs, K. & Veldkamp, E. 143. Sidle, R. C. et al. Erosion processes in steep terrain — 95. Meloni, F. & Varanda, E. M. Litter and soil arthropod Soil trace gas fluxes along orthogonal precipitation Truths, myths, and uncertainties related to forest colonization in reforested semi-​deciduous seasonal and soil fertility gradients in tropical lowland forests management in Southeast Asia. For. Ecol. Manag. Atlantic forests: Arthropod colonization in Atlantic of Panama. Biogeosciences 14, 3509–3524 (2017). 224, 199–225 (2006). forest soils. Restor. Ecol. 23, 690–697 (2015). 120. Koehler, B. et al. An in-​depth look into a tropical 144. Nagy, R. C. et al. Soil carbon dynamics in soybean 96. Cleveland, C. C. et al. Relationships among net lowland forest soil: nitrogen-​addition effects on the cropland and forests in Mato Grosso, Brazil. J. Geophys.

primary productivity, nutrients and climate in tropical contents of N2O, CO2 and CH4 and N2O isotopic Res. Biogeosci. 123, 18–31 (2018). rain forest: a pan-​tropical analysis: Nutrients, climate signatures down to 2-m depth. Biogeochemistry 111, 145. Driessen, P. M. Lecture Notes on the Major Soils of and tropical NPP. Ecol. Lett. 14, 939–947 (2011). 695–713 (2012). the World (Food and Agriculture Organization of the 97. Matson, A. L., Corre, M. D., Burneo, J. I. & Veldkamp, E. 121. Hassler, E. et al. Soil fertility controls soil–atmosphere United Nations, 2001). Free-​living nitrogen fixation responds to elevated carbon dioxide and methane fluxes in a tropical 146. Tisdall, J. M. & Oades, J. M. Organic matter and water-​ nutrient inputs in tropical montane forest floor and landscape converted from lowland forest to rubber and stable aggregates in soils. J. Soil. Sci. 33, 141–163 canopy soils of southern Ecuador. Biogeochemistry oil palm plantations. Biogeosciences 12, 5831–5852 (1982). 122, 281–294 (2015). (2015). 147. Haileslassie, A., Priess, J., Veldkamp, E., Teketay, D. 98. Hedin, L. O., Brookshire, E. N. J., Menge, D. N. L. & 122. Keller, M. & Reiners, W. A. Soil-​atmosphere exchange & Lesschen, J. P. Assessment of soil nutrient depletion Barron, A. R. The nitrogen paradox in tropical forest of nitrous oxide, nitric oxide, and methane under and its spatial variability on smallholders’ mixed farming ecosystems. Annu. Rev. Ecol. Evol. Syst. 40, 613–635 secondary succession of pasture to forest in the systems in Ethiopia using partial versus full nutrient (2009). Atlantic lowlands of Costa Rica. Glob. Biogeochem. balances. Agric. Ecosyst. Environ. 108, 1–16 (2005). 99. Cusack, D. F., Silver, W. & McDowell, W. H. Biological Cycles 8, 399–409 (1994). 148. Quinton, J. N., Govers, G., Van Oost, K. & nitrogen fixation in two tropical forests: ecosystem-​ 123. Veldkamp, E., Weitz, A. M. & Keller, M. Management Bardgett, R. D. The impact of agricultural soil erosion level patterns and effects of nitrogen fertilization. effects on methane fluxes in humid tropical pasture on biogeochemical cycling. Nat. Geosci. 3, 311–314 Ecosystems 12, 1299–1315 (2009). soils. Soil Biol. Biochem. 33, 1493–1499 (2001). (2010).

604 | November 2020 | volume 1 www.nature.com/natrevearthenviron Reviews

149. Amundson, R. et al. Soil and human security in the 170. Bayon, G. et al. Intensifying weathering and land use National Park, India. Plant Soil 198, 137–145 21st century. Science 348, 1261071 (2015). in Iron Age Central Africa. Science 335, 1219–1222 (1998). 150. Powers, J. S. & Marín-​Spiotta, E. Ecosystem processes (2012). 191. Paul, S., Veldkamp, E. & Flessa, H. Differential and biogeochemical cycles in secondary tropical forest 171. Beach, T., Dunning, N., Luzzadder-​Beach, S., response of mineral-​associated organic matter in succession. Annu. Rev. Ecol. Evol. Syst. 48, 497–519 Cook, D. E. & Lohse, J. Impacts of the ancient Maya tropical soils formed in volcanic ashes and marine (2017). on soils and soil erosion in the central Maya Lowlands. Tertiary sediment to treatment with HCl, NaOCl, and

151. Russell, A. E. & Raich, J. W. Rapidly growing tropical Catena 65, 166–178 (2006). Na4P2O7. Soil Biol. Biochem. 40, 1846–1855 (2008). trees mobilize remarkable amounts of nitrogen, 172. Lombardo, U. & Prümers, H. Pre-​Columbian human 192. Soil Survey Staff. Keys to Soil Taxonomy (US in ways that differ surprisingly among species. occupation patterns in the eastern plains of the Llanos Department of Agriculture, Natural Resources Proc. Natl. Acad. Sci. USA 109, 10398–10402 (2012). de Moxos, Bolivian Amazonia. J. Archaeol. Sci. 37, Conservation Service, 2014). 152. Saynes, V., Hidalgo, C., Etchevers, J. D. & Campo, J. E. 1875–1885 (2010). 193. Foley, J. A. et al. Solutions for a cultivated planet. Soil C and N dynamics in primary and secondary 173. Arroyo-​Kalin, M. The Amazonian formative: crop Nature 478, 337–342 (2011). seasonally dry tropical forests in Mexico. Appl. Soil. Ecol. domestication and anthropogenic soils. Diversity 2, 194. Alston, L. J., Libecap, G. D. & Mueller, B. Land reform 29, 282–289 (2005). 473–504 (2010). policies, the sources of violent conflict, and implications 153. Barron, A. R. et al. Molybdenum limitation of 174. Glaser, B. & Birk, J. J. State of the scientific knowledge for deforestation in the Brazilian Amazon. J. Environ. asymbiotic nitrogen fixation in tropical forest soils. on properties and genesis of Anthropogenic Dark Econ. Manag. 39, 162–188 (2000). Nat. Geosci. 2, 42–45 (2009). Earths in Central Amazonia (terra preta de Índio). 195. Gatto, M., Wollni, M. & Qaim, M. Oil palm boom and 154. Szott, L. T., Palm, C. A. & Buresh, R. J. Ecosystem Geochim. Cosmochim. Acta 82, 39–51 (2012). land-​use dynamics in Indonesia: The role of policies fertility and fallow function in the humid and subhumid 175. Richter, D. D. & Markewitz, D. How deep is soil? and socioeconomic factors. Land Use Policy 46, tropics. Agrofor. Syst. 47, 163–196 (1999). BioScience 45, 600–609 (1995). 292–303 (2015). 155. Batterman, S. A. et al. Key role of symbiotic dinitrogen 176. Borneman, J. & Triplett, E. W. Molecular microbial 196. Jantalia, C. P. et al. Tillage effect on C stocks of a fixation in tropical forest secondary succession. Nature diversity in soils from eastern Amazonia: evidence for clayey Oxisol under a soybean-​based crop rotation 502, 224–227 (2013). unusual microorganisms and microbial population in the Brazilian Cerrado region. Soil Tillage Res. 95, 156. Lawrence, D. & Schlesinger, W. H. Changes in soil shifts associated with deforestation. Appl. Environ. 97–109 (2007). phosphorus during 200 years of shifting cultivation in Microbiol. 63, 2647–2653 (1997). 197. Six, J. et al. Soil organic matter, biota and aggregation Indonesia. Ecology 82, 2769–2780 (2001). 177. Powers, J. S. & Veldkamp, E. Regional variation in temperate and tropical soils - Effects of no-​tillage. 157. Markewitz, D., Figueiredo, R., de, O. & Davidson, E. A. in soil carbon and δ13C in forests and pastures of Agronomie 22, 755–775 (2002).

CO2-driven cation leaching after tropical forest northeastern Costa Rica. Biogeochemistry 72, 198. Comte, I. et al. Physicochemical properties of soils clearing. J. Geochem. Explor. 88, 214–219 (2006). 315–336 (2005). in the Brazilian Amazon following fire-​free land 158. Markewitz, D. et al. Control of cation concentrations 178. Lucas, Y. The role of plants in controlling rates and preparation and slash-​and-burn practices. Agric. in stream waters by surface soil processes in an products of weathering: importance of biological Ecosyst. Environ. 156, 108–115 (2012). Amazonian watershed. Nature 410, 802–805 pumping. Annu. Rev. Earth Planet. Sci. 29, 135–163 199. Abu Bakar, R., Darus, S. Z., Kulaseharan, S. & (2001). (2001). Jamaluddin, N. Effects of ten year application of 159. Orihuela-Belmonte,​ D. E. et al. Carbon stocks and 179. Kleber, M., Schwendenmann, L., Veldkamp, E., empty fruit bunches in an oil palm plantation on soil accumulation rates in tropical secondary forests at the Rößner, J. & Jahn, R. Halloysite versus gibbsite: chemical properties. Nutr. Cycling Agroecosyst. 89, scale of community, landscape and forest type. Agric. Silicon cycling as a pedogenetic process in two lowland 341–349 (2011). Ecosyst. Environ. 171, 72–84 (2013). neotropical rain forest soils of La Selva, Costa Rica. 200. Clay, D., Reardon, T. & Kangasniemi, J. Sustainable 160. Davidson, E. A. et al. Nitrogen and phosphorus Geoderma 138, 1–11 (2007). intensification in the highland tropics: Rwandan limitation of biomass growth in a tropical secondary 180. Lucas, Y., Luizao, F. J., Chauvel, A., Rouiller, J. & farmers’ investments in land conservation and soil forest. Ecol. Appl. 14, 150–163 (2004). Nahon, D. The relation between biological activity fertility. Econ. Dev. Cult. Change 46, 351–377 161. Lu, D., Moran, E. & Mausel, P. Linking Amazonian of the rain forest and mineral composition of soils. (1998). secondary succession forest growth to soil properties. Science 260, 521–523 (1993). Land Degrad. Dev. 13, 331–343 (2002). 181. Bouma, J. et al. Hydropedological insights when Acknowledgements 162. Mekuria, W., Veldkamp, E., Corre, M. D. & Haile, M. considering catchment classification. Hydrol. Earth We thank the following researchers for providing their original Restoration of ecosystem carbon stocks following Syst. Sci. 15, 1909–1919 (2011). data from their publications: Marleen de Blécourt; Tommaso exclosure establishment in communal grazing lands in 182. Krinner, G. et al. A dynamic global vegetation model Chiti, Ute Hamer, Hennok Kassa, Maximilian Kirsten, Wolde Tigray, Ethiopia. Soil Sci. Soc. Am. J. 75, 246–256 for studies of the coupled atmosphere-​biosphere Mekuria, Diego Navarrete, Jan Nyssen, Iván Prieto, Amin (2011). system. Glob. Biogeochem. Cycles 19, GB1015 Soltangheisi, Clément Stahl and Oliver van Straaten. We 163. Shi, Z. et al. The age distribution of global soil carbon (2005). thank Oliver van Straaten for making the maps. We thank inferred from radiocarbon measurements. Nat. Geosci. 183. Jenny, H. Factors of Soil Formation. A System of Boniface Massawe for assistance with the soil profile images. 13, 555–559 (2020). Quantitative Pedology (McGraw-​Hill, 1941). E.V. and M.D.C. acknowledge financial support by the 164. Palm, C. A. et al. Nitrous oxide and methane fluxes in 184. de Blécourt, M., Brumme, R., Xu, J., Corre, M. D. Deutsche Forschungsgemeinschaft (DFG, German Research six different land use systems in the Peruvian Amazon. & Veldkamp, E. Soil carbon stocks decrease following Foundation — Project ID 192626868 — SFB 990) as part of Glob. Biogeochem. Cycles 16, 1073 (2002). conversion of secondary forests to rubber (Hevea project A05. 165. Brown, A. E., Zhang, L., McMahon, T. A., Western, A. W. brasiliensis) plantations. PLoS ONE 8, e69357 & Vertessy, R. A. A review of paired catchment studies (2013). Author contributions for determining changes in water yield resulting from 185. Darras, K. F. A. et al. Reducing fertilizer and avoiding The authors contributed equally to all aspects of the article. alterations in vegetation. J. Hydrol. 310, 28–61 (2005). herbicides in oil palm plantations — Ecological and 166. Ogden, F. L., Crouch, T. D., Stallard, R. F. & Hall, J. S. economic valuations. Front. For. Glob. Change 2, 65 Competing interests Effect of land cover and use on dry season river runoff, (2019). The authors declare no competing interests. runoff efficiency, and peak storm runoff in the seasonal 186. Hansen, M. C. et al. High-​resolution global maps tropics of Central Panama. Water Resour. Res. 49, of 21st-century​ forest cover change. Science 342, Peer review information 8443–8462 (2013). 850–853 (2013). Nature Reviews Earth & Environment thanks Mark Bonner, 167. Lacombe, G. et al. Contradictory hydrological impacts 187. Hengl, T. et al. SoilGrids250m: Global gridded soil Gervasio Piñeiro and the other, anonymous, reviewer(s) for of afforestation in the humid tropics evidenced by information based on machine learning. PLoS ONE their contribution to the peer review of this work. long-term​ field monitoring and simulation modelling. 12, e0169748 (2017). Hydrol. Earth Syst. Sci. 20, 2691–2704 (2016). 188. IUSS Working Group WRB. World Reference Base for Publisher’s note 168. de Blécourt, M., Gröngröft, A., Baumann, S. & Soil Resources 2014. International Soil Classification Springer Nature remains neutral with regard to jurisdictional Eschenbach, A. Losses in soil organic carbon stocks System for Naming Soils and Creating Legends for Soil claims in published maps and institutional affiliations. and soil fertility due to deforestation for low-​input Maps (FAO, 2015). agriculture in semi-​arid southern Africa. J. Arid. 189. FAO. Global Forest Resources Assessment 2015: Supplementary information Environ. 165, 88–96 (2019). How are the World’s Forests Changing? (FAO, 2016). Supplementary information is available for this paper at 169. Garcin, Y. et al. Early anthropogenic impact on 190. Saikh, H., Varadachari, C. & Ghosh, K. Changes in https://doi.org/10.1038/s43017-020-0091-5. Western Central African rainforests 2,600 y ago. carbon, nitrogen and phosphorus levels due to Proc. Natl Acad. Sci. USA 115, 3261–3266 (2018). deforestation and cultivation: a case study in Simlipal © Springer Nature Limited 2020

NATure RevIewS | EarTh & EnvironmenT volume 1 | November 2020 | 605