Multi-model ensemble improved the prediction of trends in soil organic carbon stocks in German croplands

Riggers, Catharina; Poeplau, Christopher; Don, Axel; Bamminger, Chris; Höper, Heinrich; Dechow, René

doi:10.1016/j.geoderma.2019.03.014

Cited by 48 publications

(40 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Several studies have indicated that the RothC model is most sensitive to C inputs (Gottschalk et al, 2012;Stamati et al, 2013;Riggers et al, 2019). In our study, analyses were performed to test the sensitivity effect on SOC changes of the different modifications (other than C inputs) implemented in the model, using RothC_4.…”

Section: Sensitivity Analysismentioning

confidence: 99%

Simulating soil organic C dynamics in managed grasslands under humid temperate climatic conditions

Jebari

Álvaro‐Fuentes

Pardo

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. Temperate grasslands are of paramount importance in terms of soil organic carbon (SOC) dynamics. Globally, research on SOC dynamics has largely focused on forests, croplands and natural grasslands, while intensively managed grasslands has received much less attention. In this regard, we aimed to improve the prediction of SOC dynamics in managed grasslands under humid temperate regions. In order to do so, we modified and recalibrated the SOC model RothC, originally developed to model the turnover of SOC in arable topsoils, which requires limited amount of readily available input data. The modifications proposed for the RothC are: (1) water content up to saturation conditions in the soil water function of RothC to fit the humid temperate climatic conditions, (2) entry pools that account for particularity of exogenous organic matter (EOM) applied (e.g., ruminant excreta), (3) annual variation in the carbon inputs derived from plant residues considering both above- and below-ground plant residue and rhizodeposits components as well as their quality, and (4) the livestock treading effect (i.e., poaching damage) as a common problem in humid areas with higher annual precipitation. In the paper, we describe the basis of these modifications, carry out a simple sensitivity analysis and validate predictions against data from existing field experiments from four sites in Europe. Model performance showed that modified RothC reasonably captures well the different modifications. However, the model seems to be more sensitive to soil moisture and plant residues modifications than to the other modifications. The applied changes in RothC model could be appropriate to simulate both farm and regional SOC dynamics from managed grassland-based systems under humid temperate conditions.

show abstract

Section: Sensitivity Analysismentioning

confidence: 99%

Simulating soil organic C dynamics in managed grasslands under humid temperate climatic conditions

Jebari

Álvaro‐Fuentes

Pardo

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Thus, we decided to stick to the conventional assumption well proven by Bolinder et al (2007) and provided regionally sound mean values of NPP below (equal root-C org input) as a starting point for future research. A SOC modeling study on German arable long-term monitoring sites using five different C org input estimation methods (Riggers et al 2019) supported this procedure: C org input estimated by the here presented regional approach led to lower model errors than the original one of Bolinder et al (2007). This is most likely because the latter summarized studies mainly from North America.…”

Section: Discussionmentioning

confidence: 56%

“…However, C org allocation coefficients are needed to convert yield data into root-and shoot-derived C org input. Keel et al (2017) and Riggers et al (2019) demonstrated that the choice of allocation coefficients used for C org input estimation strongly influences the SOC trends modeled. Region-specific up-to-date allocation coefficients and harvest indices are required to minimize this source of error.…”

Section: Introductionmentioning

confidence: 99%

Exports and inputs of organic carbon on agricultural soils in Germany

Jacobs

Poeplau²,

Weiser³

et al. 2020

Nutr Cycl Agroecosyst

Self Cite

View full text Add to dashboard Cite

The quantity and quality of organic carbon (Corg) input drive soil Corg stocks and thus fertility and climate mitigation potential of soils. To estimate fluxes of Corg as net primary production (NPP), exports, and inputs on German arable and grassland soils, we used field management data surveyed within the Agricultural Soil Inventory (n = 27.404 cases of sites multiplied by years). Further, we refined the concept of yield-based Corg allocation coefficients and delivered a new regionalized method applicable for agricultural soils in Central Europe. Mean total NPP calculated for arable and grassland soils was 6.9 ± 2.3 and 5.9 ± 2.9 Mg Corg ha−1 yr−1, respectively, of which approximately half was exported. On average, total Corg input calculated did not differ between arable (3.7 ± 1.8 Mg ha−1 yr−1) and grassland soils (3.7 ± 1.3 Mg ha−1 yr−1) but Corg sources were different: Grasslands received 1.4 times more Corg from root material than arable soils and we suggest that this difference in quality rather than quantity drives differences in soil Corg stocks between land use systems. On arable soils, side products were exported in 43% of the site * years. Cover crops were cultivated in 11% of site * years and contributed on average 3% of the mean annual total NPP. Across arable crops, total NPP drove Corg input (R2 = 0.47) stronger than organic fertilization (R2 = 0.11). Thus, maximizing plant growth enhances Corg input to soil. Our results are reliable estimates of management related Corg fluxes on agricultural soils in Germany.

show abstract

“…The stacking approach can take advantage of the characteristics of different machine learning algorithms, reduce the variance of the single machine learning model, and provide better and more stable predictions. Previous studies have shown the potential of stacking models in improving the accuracy of soil property maps of soil organic carbon [33][34][35][36][37], soil total nitrogen [38], soil class [39], soil pH [40], and soil texture [41].…”

Section: Introductionmentioning

confidence: 99%

Improving Soil Thickness Estimations Based on Multiple Environmental Variables with Stacking Ensemble Methods

Luo

Jin

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

Spatially continuous soil thickness data at large scales are usually not readily available and are often difficult and expensive to acquire. Various machine learning algorithms have become very popular in digital soil mapping to predict and map the spatial distribution of soil properties. Identifying the controlling environmental variables of soil thickness and selecting suitable machine learning algorithms are vitally important in modeling. In this study, 11 quantitative and four qualitative environmental variables were selected to explore the main variables that affect soil thickness. Four commonly used machine learning algorithms (multiple linear regression (MLR), support vector regression (SVR), random forest (RF), and extreme gradient boosting (XGBoost) were evaluated as individual models to separately predict and obtain a soil thickness distribution map in Henan Province, China. In addition, the two stacking ensemble models using least absolute shrinkage and selection operator (LASSO) and generalized boosted regression model (GBM) were tested and applied to build the most reliable and accurate estimation model. The results showed that variable selection was a very important part of soil thickness modeling. Topographic wetness index (TWI), slope, elevation, land use and enhanced vegetation index (EVI) were the most influential environmental variables in soil thickness modeling. Comparative results showed that the XGBoost model outperformed the MLR, RF and SVR models. Importantly, the two stacking models achieved higher performance than the single model, especially when using GBM. In terms of accuracy, the proposed stacking method explained 64.0% of the variation for soil thickness. The results of our study provide useful alternative approaches for mapping soil thickness, with potential for use with other soil properties.

show abstract

Multi-model ensemble improved the prediction of trends in soil organic carbon stocks in German croplands

Cited by 48 publications

References 51 publications

Simulating soil organic C dynamics in managed grasslands under humid temperate climatic conditions

Simulating soil organic C dynamics in managed grasslands under humid temperate climatic conditions

Exports and inputs of organic carbon on agricultural soils in Germany

Improving Soil Thickness Estimations Based on Multiple Environmental Variables with Stacking Ensemble Methods

Contact Info

Product

Resources

About