Using process-oriented model output to enhance machine learning-based soil organic carbon prediction in space and time

Monitoring and modelling soil organic carbon (SOC) in space and time can help us to better understand soil carbon dynamics and is of key importance to support climate change research and policy. Although machine learning (ML) has attracted a lot of attention in the digital soil mapping (DSM) community for its powerful ability to learn from data and predict soil properties, such as SOC, it is better at capturing soil spatial variation than soil temporal dynamics. By contrast, process-oriented (PO) models benefit from mechanistic knowledge to express physiochemical and biological processes that govern SOC temporal changes. Therefore, integrating PO and ML models seems a promising means to represent physically plausible SOC dynamics while retaining the spatial prediction accuracy of ML models. In this study, a hybrid modelling framework was developed and tested for predicting topsoil SOC stock in space and time for a regional cropland area located in eastern China. In essence, the hybrid model uses predictions of the PO model in unsampled years as additional training data of the ML model, with a weighting parameter assigned to balance the importance of SOC values from the PO model and real measurements. The results indicated that temporal trends of SOC stock modelled by PO and ML models were largely different, while they were notably similar between the PO and hybrid models. Cross-validation showed that the hybrid model had the best performance (RMSE = 0.29 kg m-2), with a 19 % improvement compared with the ML model. We conclude that the proposed hybrid framework not only enhances space-time soil carbon mapping in terms of prediction accuracy and physical plausibility, it also provides insights for soil management and policy decisions in the face of future climate change and intensified human activities.

Carbon Farming practices assessment: Modelling spatial changes of Soil Organic Carbon in Flanders, Belgium

Carbon sequestration in soils is a strategy to mitigate climate change and promote sustainable soil management. Since the European Union (EU) stimulates the reduction of greenhouse gases (GHG) from the atmosphere, the necessity to explore innovative approaches to sequester carbon in agricultural landscapes is becoming urgent. Carbon Farming (CF) has emerged as a promising program to mitigate climate change in agriculture but there is still a lack of agreement on which tools can be used to calculate Soil Organic Carbon (SOC) dynamics in this context. Using the RothC model a spatial analysis of SOC in the agricultural parcels of Flanders, Belgium was performed. Two among the various CF practices were simulated: a use of cover crops (CC) and the most common crop rotations adopted in the area, enriched with the use of cover crops. The performances of the model were evaluated and compared to other studies in areas with similar climate and environments. The selected CF practices can mitigate the carbon emissions from agricultural soils up to 60 % of the current projections. The most sensitive variables in the RothC model that affect the final total SOC, and thus determining the model outcome, are the Business As Usual (BAU) carbon inputs and the initial carbon content. For these variables the Pearson Correlation Coefficient with the change in SOC reached values of -0.78 and -0.50 respectively. To achieve net carbon sequestration in the agricultural soils of Flanders, Belgium, more effective solutions need to be evaluated. Furthermore, a larger amount and accessibility of data are required to reach better modelling performances.

Modification of the RothC model to evaluate the inconsistent effect of conservation tillage on SOC stock and a suggestion of a national-scale assessment framework

Simulation of conservation tillage effect on soil organic carbon (SOC) stock on the national scale is essential for Tier 3 level greenhouse gas inventory in the agricultural sector. However, the conservation tillage effects varied depending on different soil conditions, potentially leading to inaccurate national assessments. This study aimed to propose a framework for estimating the national scale impact of conservation tillage on SOC. As even in the most commonly used SOC dynamic model, the Rothamsted Carbon Model (RothC), does not reflect the conservation tillage effect in an explicit way, we modified it by developing the tillage rate modifiers (TRMs). First, we investigated the conditions for the inconsistent conservation tillage effects using the decision tree analysis based on 210 field experiment data from the mid-latitude region. The results highlighted that soil sand content and the existing SOC stock were the main factors driving the inconsistencies. After we categorized into four distinctive conditions, the TRMs for each condition were parameterized using a genetic algorithm. The average TRMs were 0.88 in the soils with sand content >37.6 % and 1.58 in the soils with sand content ≤37.6 %, indicating that conservation tillage is more effective in coarse-textured soil, and there is a risk of decreasing SOC stock in the latter condition. Using the modified RothC model, a three-step national-scale simulation framework was suggested: compiling country-specific data, establishing baseline and conservation tillage scenarios, and modeling conservation tillage effects with uncertainty analysis. Our approach also defined the maximum conservation tillage area, factoring in local cropping systems and soil conditions. Our refined RothC model with TRMs provides a nuanced understanding of conservation tillage effects, emphasizing the role of soil characteristics. The proposed national-scale simulation framework offers a reliable tool for evaluating conservation tillage impact on SOC, ensuring more accurate greenhouse gas inventories in agriculture.

Feasibility of the 4 per 1000 aspirational target for soil carbon: A case study for France

Increasing soil organic carbon (SOC) stocks is a promising way to mitigate the increase in atmospheric CO₂ concentration. Based on a simple ratio between CO₂ anthropogenic emissions and SOC stocks worldwide, it has been suggested that a 0.4% (4 per 1000) yearly increase in SOC stocks could compensate for current anthropogenic CO₂ emissions. Here, we used a reverse RothC modelling approach to estimate the amount of C inputs to soils required to sustain current SOC stocks and to increase them by 4‰ per year over a period of 30 years. We assessed the feasibility of this aspirational target first by comparing the required C input with net primary productivity (NPP) flowing to the soil, and second by considering the SOC saturation concept. Calculations were performed for mainland France, at a 1 km grid cell resolution. Results showed that a 30%-40% increase in C inputs to soil would be needed to obtain a 4‰ increase per year over a 30-year period. 88.4% of cropland areas were considered unsaturated in terms of mineral-associated SOC, but characterized by a below target C balance, that is, less NPP available than required to reach the 4‰ aspirational target. Conversely, 90.4% of unimproved grasslands were characterized by an above target C balance, that is, enough NPP to reach the 4‰ objective, but 59.1% were also saturated. The situation of improved grasslands and forests was more evenly distributed among the four categories (saturated vs. unsaturated and above vs below target C balance). Future data from soil monitoring networks should enable to validate these results. Overall, our results suggest that, for mainland France, priorities should be (1) to increase NPP returns in cropland soils that are unsaturated and have a below target carbon balance and (2) to preserve SOC stocks in other land uses.

Optimizing multifunctional agroecosystems in irrigated dryland agriculture to restore soil carbon - Experiments and modelling

Irrigated dryland agroecosystems could become more sustainable if crop and soil management enhanced soil organic carbon (SOC). We hypothesized that combining high inputs from cover crops with no-tillage will increase long-term SOC stocks. Caatinga shrublands had been cleared in 1972 for arable crops and palm plantations before implementing field experiments on Mango and Melon systems (established in 2009 and 2012, respectively). Each of the two experiments were managed with no-till (NT) or conventional till (CT), and three types of cover cropping, either a plant mixture of 75% (PM1) or 25% (PM2) legumes, or spontaneous vegetation (SV). The RothC model was used with a daily timestep to simulate the soil moisture dynamics and C turnover for this dry climate. Carbon inputs were between 2.62 and 5.82 Mg C ha^-1 year^-1 and increased the depleted SOC stocks by 0.08 to 0.56 Mg C ha^-1 year^-1. Scenarios of continuous biomass inputs of ca. 5 Mg C ha^-1 year^-1 for 60 years are likely to increase SOC stocks in the mango NT beyond the original Caatinga SOC by between 19.2 and 20.5 Mg C ha^-1. Under CT similar inputs would increase SOC stocks only marginally above depletion (2.75 to 2.47 Mg C ha^-1). Under melon, annual carbon inputs are slightly greater (up to 5.5 Mg C ha^-1 year^-1) and SOC stocks would increase on average by another 8% to 22.3 to 20.6 Mg C ha^-1 under NT and by 8 Mg C ha^-1 under CT. These long-term simulations show that combining NT with high quality cover crops (PM1, PM2) would exceed SOC stocks of the initial Caatinga within 20 and 25 years under irrigated melon and mango cultivation, respectively. These results present a solution to reverse prior loss of SOC by replacing CT dryland agriculture with irrigated NT plus high input cover crops agroecosystems.

Projected trends of soil organic carbon stocks in Meghalaya state of Northeast Himalayas, India. Implications for a policy perspective

Agricultural and forestry activities can affect soil organic carbon (SOC) levels and CO₂ emissions from terrestrial ecosystems due to land use changes. In Northeast Himalayas, studies on the effects of forest conversion to temporary agricultural lands (jhum) on the loss of SOC and soil quality degradation have received the attention of policy makers and scientific research. Presently, local communities are now oriented towards the settled plantations systems with modern cash crops such as tea and rubber, that could act as potential SOC sinks. However, no information on SOC dynamics and simulation studies after land-use change from temporary agricultural lands (jhum) to settled cultivations and under climate change (CC) conditions are available for the Meghalaya state. Applying the RothC model, we focused on four different scenarios including the conversion from jhum to settled cultivation (rubber plantations and tea gardens), as well as continuous jhum cultivation and jhum to jhum with a period of secondary succession. Simulations under CC conditions indicated that SOC stocks significantly increased by 1.20 t C ha^-1 yr^-1 in tea gardens compared to rubber and jhum scenarios. Conversely, SOC stocks slightly decreased by 0.07 t C ha^-1 yr^-1 in rubber plantations, while the regrowth of a natural vegetation cover as secondary succession following the abandonment of the jhum fields, showed a lower SOC decrease (0.18 t C ha^-1 yr^-1) compared to the continuous jhum cultivation (0.24 t C ha^-1 yr^-1). Thus, for CC mitigation in a policy perspective, tea gardens could represent the best land use to store increasing amounts of SOC in the long-term perspective and optimize farmers' incomes, while in rubber plantations SOC storage is limited in time. Jhum cultivation can benefit in terms of productivity and profitability by extending the duration of the secondary succession period.

Using measured stocks of biomass and litter carbon to constrain modelled estimates of sequestration of soil organic carbon under contrasting mixed-species environmental plantings

Reforestation of agricultural land with mixed-species environmental plantings of native trees and shrubs contributes to abatement of greenhouse gas emissions through sequestration of carbon, and to landscape remediation and biodiversity enhancement. Although accumulation of carbon in biomass is relatively well understood, less is known about associated changes in soil organic carbon (SOC) following different types of reforestation. Direct measurement of SOC may not be cost effective where rates of SOC sequestration are relatively small and/or highly spatially-variable, thereby requiring intensive sampling. Hence, our objective was to develop a verified modelling approach for determining changes in SOC to facilitate the inclusion of SOC in the carbon accounts of reforestation projects. We measured carbon stocks of biomass, litter and SOC (0-30cm) in 125 environmental plantings (often paired to adjacent agricultural sites), representing sites of varying productivity across the Australian continent. After constraining a carbon accounting model to observed measures of growth, allocation of biomass, and rates of litterfall and litter decomposition, the model was calibrated to maximise the efficiency of prediction of SOC and its fractions. Uncertainties in both measured and modelled results meant that efficiencies of prediction of SOC across the 125 contrasting plantings were only moderate, at 39-68%. Data-informed modelling nonetheless improved confidence in outputs from scenario analyses, confirming that: (i) reforestation on agricultural land highly depleted in SOC (i.e. previously under cropping) had the highest capacity to sequester SOC, particularly where rainfall was relatively high (>600mmyear^-1), and; (ii) decreased planting width and increased stand density and the proportion of eucalypts enhanced rates of SOC sequestration. These results improve confidence in predictions of SOC following environmental reforestation under varying conditions. The calibrated model will be a useful tool for informing land managers and policy makers seeking to understand the dynamics of SOC following such reforestation.

Simulating effects of changing climate and CO(2) emissions on soil carbon pools at the Hubbard Brook experimental forest

Carbon (C) sequestration in forest biomass and soils may help decrease regional C footprints and mitigate future climate change. The efficacy of these practices must be verified by monitoring and by approved calculation methods (i.e., models) to be credible in C markets. Two widely used soil organic matter models - CENTURY and RothC - were used to project changes in SOC pools after clear-cutting disturbance, as well as under a range of future climate and atmospheric carbon dioxide (CO(2) ) scenarios. Data from the temperate, predominantly deciduous Hubbard Brook Experimental Forest (HBEF) in New Hampshire, USA, were used to parameterize and validate the models. Clear-cutting simulations demonstrated that both models can effectively simulate soil C dynamics in the northern hardwood forest when adequately parameterized. The minimum postharvest SOC predicted by RothC occurred in postharvest year 14 and was within 1.5% of the observed minimum, which occurred in year 8. CENTURY predicted the postharvest minimum SOC to occur in year 45, at a value 6.9% greater than the observed minimum; the slow response of both models to disturbance suggests that they may overestimate the time required to reach new steady-state conditions. Four climate change scenarios were used to simulate future changes in SOC pools. Climate-change simulations predicted increases in SOC by as much as 7% at the end of this century, partially offsetting future CO(2) emissions. This sequestration was the product of enhanced forest productivity, and associated litter input to the soil, due to increased temperature, precipitation and CO(2) . The simulations also suggested that considerable losses of SOC (8-30%) could occur if forest vegetation at HBEF does not respond to changes in climate and CO(2) levels. Therefore, the source/sink behavior of temperate forest soils likely depends on the degree to which forest growth is stimulated by new climate and CO(2) conditions.

Soil salinity decreases global soil organic carbon stocks

Saline soils cover 3.1% (397 million hectare) of the total land area of the world. The stock of soil organic carbon (SOC) reflects the balance between carbon (C) inputs from plants, and losses through decomposition, leaching and erosion. Soil salinity decreases plant productivity and hence C inputs to the soil, but also microbial activity and therefore SOC decomposition rates. Using a modified Rothamsted Carbon model (RothC) with a newly introduced salinity decomposition rate modifier and a plant input modifier we estimate that, historically, world soils that are currently saline have lost an average of 3.47 tSOC ha(-1) since they became saline. With the extent of saline soils predicted to increase in the future, our modelling suggests that world soils may lose 6.8 Pg SOC due to salinity by the year 2100. Our findings suggest that current models overestimate future global SOC stocks and underestimate net CO2 emissions from the soil-plant system by not taking salinity effects into account. From the perspective of enhancing soil C stocks, however, given the lower SOC decomposition rate in saline soils, salt tolerant plants could be used to sequester C in salt-affected areas.