Search

Soil pH is one of the critical indicators of soil quality. A fine resolution map of soil pH is urgently required to address practical issues of agricultural production, environmental protection, and ecosystem functioning, which often fall short of meeting the demands for local applications. To fill this gap, we used data from an extensive survey of 13,424 surface soil samples (0–0.2 m) across the cropland of Jiangxi Province in Southern China. Using digital soil mapping techniques with 46 covariates, we produced a 30 m resolution soil map of topsoil pH in cropland of Southern China. We integrate different variable selection algorithms and machine learning methods. Our results indicate the Random Forest with covariates selected by recursive feature selection had the best performance for mapping soil pH with r of 0.583 and RMSE of 0.41. The prediction interval coverage probability for our prediction was 0.92, indicating a low estimated prediction uncertainty. Climate was identified as the most critical variable for predicting soil pH with a contribution of 37.42 %, followed by soil properties (29.09 %), soil management (21.86 %), parent material (6.22 %), and biota (5.39 %) factors. The mean topsoil pH in the cropland of Jiangxi Province was 5.21, indicating a great pressure of soil acidification in this region. The high soil pH values were mainly distributed in the Northern, Western, and Eastern parts of the survey region while low soil pH values were majorly located in the central part. Compared with past surveys in 1980 s, there was no significant soil pH change in the surveyed region. Our soil pH map can provide important implications and guidance decisions on soil heavy metal pollution remediation, precision agriculture, and prevention of soil acidification.

Soil organic carbon (SOC) plays a crucial role in soil health and global carbon cycling, therefore accurate estimates of its spatial distribution are important for managing soil health and mitigating global climate change. Digital soil mapping shows its potential to provide accurate and high-resolution spatial distribution of SOC across scales. To convert SOC content to SOC density (SOCD), two inference trajectories exist for predicting SOCD in digital soil mapping: the direct approach (calculate-then-model) and indirect approach (model-then-calculate). However, there is a lack of comprehensive exploration regarding the differences in their performance in SOCD estimates, particularly in regions characterized by diverse pedoclimatic conditions. To bridge this knowledge gap, we evaluated the two approaches based on model performance of SOCD in France. Using 916 topsoils (0−20 ​cm) from the LUCAS Soil 2018 and 24 environmental covariates, random forest model and forward recursive feature selection were used to build the spatial predictive models of SOCD using direct and indirect approaches. The results show that, using random forest model and full covariates, both approaches show moderate performance (R2 ​= ​0.28−0.32). By utilizing forward recursive feature selection model, the number of predictors was reduced from 24 to 9, enhancing model performance for direct approach (R2 of 0.35), with no improvement for indirect approach (R2 of 0.28). The mean SOCD of the French topsoil was 5.29 and 6.14 ​kg ​m−2 by direct and indirect approaches, resulting in SOC stock of 2.8 and 3.3 ​Pg, respectively. We found that the direct approach clearly underestimated the high SOCD (>9 ​kg ​m−2), while the indirect approach performed much better for high SOCD. Our findings serve as a valuable reference for SOCD mapping, thereby providing a scientific basis for maintaining soil health.

Soil visible and near-infrared (Vis-NIR, 350–2500 nm) spectroscopy has been proven as an alternative to conventional laboratory analysis due to its advantages being rapid, cost-effective, non-destructive and environmentally friendly. Different variable selection methods have been used to deal with the high redundancy, heavy computation, and model complexity of using full spectra in spectral modelling. However, most previous studies used a linear algorithm in the variable selection, and the application of a non-linear algorithm remains poorly explored. To address the current knowledge gap, based on a regional soil Vis-NIR spectral library (1430 soil samples), we evaluated seven variable selection algorithms together with three predictive algorithms in predicting seven soil properties. Our results showed that Cubist overperformed partial least squares regression (PLSR) and random forests (RF) in most soil properties (R2 > 0.75 for soil organic matter, total nitrogen and pH) when using the full spectra. Most of variable selection can greatly reduce the number of spectral bands and therefore simplified predictive models without losing accuracy. The results also showed that there was no silver bullet for the optimal variable selection algorithm among different predictive algorithms: (1) competitive adaptive reweighted sampling (CARS) always performed best for the PLSR algorithm, followed by forward recursive feature selection (FRFS); (2) recursive feature elimination (RFE) and genetic algorithm (GA) generally had better accuracy than others for the Cubist algorithm; and (3) FRFS had the best model performance for the RF algorithm. In addition, the performance was generally better when the algorithm used in the variable selection matched the predictive algorithm. The outcome of this study provides a valuable reference for predicting soil information using spectroscopic techniques together with variable selection algorithms. [ABSTRACT FROM AUTHOR]

Soil visible and near-infrared (Vis-NIR, 350–2500 nm) spectroscopy has been proven as an alternative to conventional laboratory analysis due to its advantages being rapid, cost-effective, non-destructive and environmentally friendly. Different variable selection methods have been used to deal with the high redundancy, heavy computation, and model complexity of using full spectra in spectral modelling. However, most previous studies used a linear algorithm in the variable selection, and the application of a non-linear algorithm remains poorly explored. To address the current knowledge gap, based on a regional soil Vis-NIR spectral library (1430 soil samples), we evaluated seven variable selection algorithms together with three predictive algorithms in predicting seven soil properties. Our results showed that Cubist overperformed partial least squares regression (PLSR) and random forests (RF) in most soil properties (R2 > 0.75 for soil organic matter, total nitrogen and pH) when using the full spectra. Most of variable selection can greatly reduce the number of spectral bands and therefore simplified predictive models without losing accuracy. The results also showed that there was no silver bullet for the optimal variable selection algorithm among different predictive algorithms: (1) competitive adaptive reweighted sampling (CARS) always performed best for the PLSR algorithm, followed by forward recursive feature selection (FRFS); (2) recursive feature elimination (RFE) and genetic algorithm (GA) generally had better accuracy than others for the Cubist algorithm; and (3) FRFS had the best model performance for the RF algorithm. In addition, the performance was generally better when the algorithm used in the variable selection matched the predictive algorithm. The outcome of this study provides a valuable reference for predicting soil information using spectroscopic techniques together with variable selection algorithms.

[Display omitted] • A finest map (30 m) of soil pH in cropland of Southern China was produced. • The performance of different variable selection methods was compared. • The performance of different spatial predictive algorithms on mapping soil pH was evaluated. • We analyzed the application of our produced pH map on soil heavy metal pollution. • The best applicable scenarios of RFE and FRFS were validated and analyzed. Soil pH is one of the critical indicators of soil quality. A fine resolution map of soil pH is urgently required to address practical issues of agricultural production, environmental protection, and ecosystem functioning, which often fall short of meeting the demands for local applications. To fill this gap, we used data from an extensive survey of 13,424 surface soil samples (0–0.2 m) across the cropland of Jiangxi Province in Southern China. Using digital soil mapping techniques with 46 covariates, we produced a 30 m resolution soil map of topsoil pH in cropland of Southern China. We integrate different variable selection algorithms and machine learning methods. Our results indicate the Random Forest with covariates selected by recursive feature selection had the best performance for mapping soil pH with r of 0.583 and RMSE of 0.41. The prediction interval coverage probability for our prediction was 0.92, indicating a low estimated prediction uncertainty. Climate was identified as the most critical variable for predicting soil pH with a contribution of 37.42 %, followed by soil properties (29.09 %), soil management (21.86 %), parent material (6.22 %), and biota (5.39 %) factors. The mean topsoil pH in the cropland of Jiangxi Province was 5.21, indicating a great pressure of soil acidification in this region. The high soil pH values were mainly distributed in the Northern, Western, and Eastern parts of the survey region while low soil pH values were majorly located in the central part. Compared with past surveys in 1980 s, there was no significant soil pH change in the surveyed region. Our soil pH map can provide important implications and guidance decisions on soil heavy metal pollution remediation, precision agriculture, and prevention of soil acidification. [ABSTRACT FROM AUTHOR]

• We evaluated the potential of pedotransfer function in MAOC prediction. • Forward recursive feature selection performed well in model parsimony and performance. • Cubist had better model performance than Random Forest and Gradient Boosted Machine. • Model ensemble improved model performance and robustness. Soil organic carbon (SOC) sequestration is a promising natural climate solution for capturing atmospheric CO 2 , and it provides crucial co-benefits in improving soil functions and services at the same time. Given that SOC is not a single and uniform entity, a deep understanding of SOC response to environmental changes requires additional information on SOC fractions with distinct characteristics such as particulate organic carbon (POC) and mineral associated organic carbon (MAOC). Despite their great importance, POC and MAOC information is still scarce in the soil databases, particularly on a broad scale. Pedotransfer function (PTF) is a good strategy to estimate missing soil properties, while its application in SOC fractions has been poorly explored. Based on 352 representative mineral topsoil samples (0–20 cm) across Europe, we evaluated the potential of MAOC prediction using machine learning based PTF (random forest (RF), Cubist, and gradient boosted machine (GBM)) together with predictor selection methods (recursive feature elimination (RFE) and forward recursive feature selection (FRFS)). The repeated validation (100 times) showed that MAOC could be well predicted by machine learning based PTFs (R2 of 0.877–0.9, RMSE of 2.994–3.269 g kg−1). RFE can effectively reduce the number of predictors from 21 to 12 with comparable performance to the models using all predictors. The proposed FRFS algorithm had the best model parsimony with only 6 predictors (SOC, silt + clay, nitrogen, nitrogen deposition, soil erosion and sand) and performed similar to or even better than RFE. In combination with FRFS, Cubist performed best among the three machine learning models (R2 of 0.9, RMSE of 2.994 g kg−1). Our results also showed that five model ensemble methods had similar model performance and can improve model accuracy and robustness compared to a single machine learning model. This study provides a valuable reference for coupling PTF and legacy soil databases to increase the spatial coverage and the performance of machine learning based SOC fraction predictions. [ABSTRACT FROM AUTHOR]