A high-resolution map of soil organic carbon in cropland of Southern China

CATENA, Journal Year: 2024, Volume and Issue: 237, P. 107813 - 107813

Published: Jan. 12, 2024

Language: Английский

---

Integrating additional spectroscopically inferred soil data improves the accuracy of digital soil mapping

Geoderma, Journal Year: 2023, Volume and Issue: 433, P. 116467 - 116467

Published: April 6, 2023

Digital soil mapping has been increasingly advocated as an efficient approach to deliver fine-resolution and up-to-date information in evaluating ecosystem services. Considering the great spatial heterogeneity of soils, it is widely recognized that more representative observations are needed for better capturing variation thus increase accuracy digital maps. In reality, budget field work laboratory analysis commonly limited due its high cost low efficiency. last two decades, being alternative wet chemistry, spectroscopy, such visible-near infrared (Vis-NIR), mid-infrared (MIR) spectroscopy developed measuring a rapid cost-effective manner enable collect (DSM). However, spectroscopically inferred (SI) data subject higher uncertainties than reference analysis. Many DSM practices integrated SI with into modelling while few studies addressed key question whether these non-errorless improve map DSM. this study, French Soil Monitoring Network (RMQS) Land Use Coverage Area frame Survey (LUCAS Soil) datasets were used evaluate potential from Vis-NIR MIR properties (i.e. organic carbon, clay, pH) at national scale. Cubist quantile regression forests spectral predictive modelling, respectively. For both RMQS LUCAS dataset, different scenarios regarding varying proportions tested models models. Repeated (50 times) external validation suggested adding additional can performance regardless (gain R2 proportion 3–19%) when (≤50%). Lower model led greater improvement Our results also showed lowered prediction intervals which may result underestimation uncertainty. The determination threshold on use needs be explored future studies.

Language: Английский

---

A framework for optimizing environmental covariates to support model interpretability in digital soil mapping

Geoderma, Journal Year: 2024, Volume and Issue: 445, P. 116873 - 116873

Published: April 4, 2024

A common practice in digital soil mapping (DSM) is to incorporate many environmental covariates into a machine-learning algorithm predict the spatial patterns of attributes. Variance inflation factor (VIF), principal component analysis (PCA), and recursive feature elimination (RFE) are three statistical methods that can be used reduce number covariates. This study aims 1) compare VIF PCA approaches; 2) identify an approach determine minimum DSM ensure model parsimony using RFE after VIF; 3) examine interpret impact on variability predicted properties. The area was province British Columbia (BC), Canada. legacy data for four properties make maps: organic carbon (SOC%), pH, clay%, coarse fragment (CF%). Seven models were made each property influence validation results by different produced various results. showed could reduced from 70 4 12 with only little or no difference concordance correlation coefficient (CCC) CCC pH 7 both 0.74, other properties, this negligible. obtained performance reducing not as effective when VIF. Moreover, related precipitation most important modeling SOC%, clay%. Topographic influential CF%. emphasizes potential benefits combining reduction achieve optimal outcomes generate parsimonious interpretable models.

Language: Английский

---

European topsoil bulk density and organic carbon stock database (0–20 cm) using machine-learning-based pedotransfer functions

Earth system science data, Journal Year: 2024, Volume and Issue: 16(5), P. 2367 - 2383

Published: May 16, 2024

Abstract. Soil bulk density (BD) serves as a fundamental indicator of soil health and quality, exerting significant influence on critical factors such plant growth, nutrient availability, water retention. Due to its limited availability in databases, the application pedotransfer functions (PTFs) has emerged potent tool for predicting BD using other easily measurable properties, while impact these PTFs' performance organic carbon (SOC) stock calculation been rarely explored. In this study, we proposed an innovative local modeling approach fine earth (BDfine) across Europe recently released BDfine data from LUCAS (Land Use Coverage Area Frame Survey Soil) 2018 (0–20 cm) relevant predictors. Our involved combination neighbor sample search, forward recursive feature selection (FRFS), random forest (RF) models (local-RFFRFS). The results showed that local-RFFRFS had good (R2 0.58, root mean square error (RMSE) 0.19 g cm−3, relative (RE) 16.27 %), surpassing earlier-published PTFs 0.40–0.45, RMSE 0.22 RE 19.11 %–21.18 %) global RF with without FRFS 0.56–0.57, 16.47 %–16.74 %). Interestingly, found best PTF = 0.84, 1.39 kg m−2, 17.57 performed close 0.85, 1.32 15.01 SOC predictions. However, still better (ΔR2 > 0.2) samples low stocks (< 3 m−2). Therefore, suggest is promising method prediction, would be more efficient when subsequently utilized calculating stock. Finally, produced two topsoil datasets (18 945 15 389 samples) at 0–20 cm local-RFFRFS, respectively. This dataset archived Zenodo platform https://doi.org/10.5281/zenodo.10211884 (S. Chen et al., 2023). outcomes study present meaningful advancement enhancing predictive accuracy BDfine, resultant enable precise hydrological biological modeling.

Language: Английский

---

Fine-resolution mapping of cropland topsoil pH of Southern China and its environmental application

Geoderma, Journal Year: 2024, Volume and Issue: 442, P. 116798 - 116798

Published: Feb. 1, 2024

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

Language: Английский

---

Unveiling the explanatory power of environmental variables in soil organic carbon mapping: A global–local analysis framework

Geoderma, Journal Year: 2024, Volume and Issue: 449, P. 117011 - 117011

Published: Aug. 26, 2024

Soil organic carbon (SOC) is a critical component that affects soil quality and global cycling. Current SOC mapping approaches are based on the spatial stationarity relationship of formation processes. Nevertheless, pattern consequence different soil-forming factors processes operate at scales. In this work, we hypothesized covariation environmental variables might differ spatially, proposed (whole area) local analysis framework aimed to enhance our comprehension explanatory scale variation. This primarily incorporates Geographically Weighted correlation Multi-scale Regression (MGWR) model. With 216 farmland topsoil samples collected from Qilu Lake watershed in Yunnan Province, China (area 354 km2), explored both relationships between verify feasibility framework. Results showed power variation scale-dependent. Our revealed certain variables, which may explain variations SOC, often overlooked due their insignificant with (p > 0.05). For example, case study, porosity two landscape metrics characterize anthropogenic land use patterns can effectively SOC. They improved model performance MGWR, but not significant. The highlights necessity investigating scale.

Language: Английский

---

Multiple Environmental Variables as Covariates to Improve the Accuracy of Spatial Prediction Models for SOM on Karst Aera

Land Degradation and Development, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 12, 2025

ABSTRACT Aims accurately predicting the spatial distribution of soil organic matter (SOM) is essential for environmental management and carbon storage estimation. However, diversity sources variables poses a challenge in studying SOM. Methods order to address this issue, we propose leveraging multiple employing machine learning models, specifically Lightweight gradient boosting (LightGBM) random forest (RF), SOM distribution. 128 samples were collected from Caohai National Nature Reserve, their content was measured. Results study found that average 36.75 g/kg. Compared traditional linear regression models such as ordinary kriging (OK), least squares (OLS), geographically weighted (GWR), based on nonlinear regression, LightGBM RF, demonstrated higher cross‐validated coefficients determination ( R 2 ) 0.62 0.60, respectively, outperforming other models. Additionally, RF exhibited lower mean absolute error (MAE) root square (RMSE), indicating stability generalization capability. The among showed consistency, with observed southern near‐Caohai Lake regions northern farther Lake. Shapley additive explanations (SHAP) model highlighted agricultural land (AL), pH, Elevation (ELV) primary covariates influencing Conclusions provides valuable insights support estimation karst plateau region.

Language: Английский

---

National-scale mapping of soil organic carbon stock in France: New insights and lessons learned by direct and indirect approaches

Soil & Environmental Health, Journal Year: 2023, Volume and Issue: 1(4), P. 100049 - 100049

Published: Nov. 11, 2023

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

Language: Английский

---

The validity domain of sensor fusion in sensing soil quality indicators

Geoderma, Journal Year: 2023, Volume and Issue: 438, P. 116657 - 116657

Published: Sept. 4, 2023

Soil health has gained increasing attention under the rapid development of industrialization and requirement for green agriculture. Therefore, up-to-date soil information related to is urgently needed ensure food security biodiversity protection. Previous studies have shown potential proximal sensing in measuring information, while it remains challenging get cost-efficient robust estimates multiple indicators simultaneously via sensor fusion. In this study, we investigated visible near-infrared (vis-NIR), mid-infrared (MIR) spectroscopy as well three model averaging methods predicting properties, including organic matter (SOM), pH, cation exchange capacity (CEC). The are not only used fusion but also high-level fusion, which include Granger-Ramanathan (GR), Bayesian Model Averaging Spectral-Guided Ensemble Modelling (S-GEM). Here, S-GEM a recently proposed algorithm that can improve spectroscopic prediction by spectral ensemble modelling. Four widely models were evaluated, partial least square regression, Cubist, memory based learning convolutional neural network. For SOM, on algorithms was comparable Sensorsingle + Modelmultiple (MIR singly S-GEM) with R2 0.86. However, MIR performed best among all (LCCC 0.92, RMSE 3.66 g kg−1 RPIQ 3.68). 10-fold cross-validation results indicated 0.84, LCCC 0.90, 0.45 3.65. CEC, Sensormultiple GR 0.66, 0.80, 3.48 cmol 2.22. Our showed failed when performance sensors differed lot (△R2 > 0.2), use single therefore suggested case. When close < recommended. outcome study provide reference determining validity domain improving accuracy prediction.

Language: Английский

---

Improving model performance in mapping cropland soil organic matter using time-series remote sensing data

Journal of Integrative Agriculture, Journal Year: 2024, Volume and Issue: 23(8), P. 2820 - 2841

Published: Jan. 9, 2024

Faced with increasing global soil degradation, spatially explicit data on cropland organic matter (SOM) provides crucial for carbon pool accounting, quality assessment and the formulation of effective management policies. As a spatial information prediction technique, digital mapping (DSM) has been widely used to map at different scales. However, accuracy SOM maps is typically lower than other land cover types due inherent difficulty in precisely quantifying human disturbance. To overcome this limitation, study systematically assessed framework "information extraction-feature selection-model averaging" improving model performance using 462 samples collected Guangzhou, China 2021. The results showed that dynamic extraction, feature selection averaging could efficiently improve final predictions (R2: 0.48 0.53) without having obviously negative impacts uncertainty. Quantifying environment was an efficient way generate covariates are linearly nonlinearly related SOM, which improved R2 random forest from 0.44 extreme gradient boosting 0.37 0.43. FRFS recommended when there relatively few environmental (<200), whereas Boruta many (>500). granger-ramanathan approach average When structures initial models similar, number did not have significantly positive effects predictions. Given advantages these selected strategies over great potential high-accuracy any scales, so can provide more reliable references conservation policy-making.

Language: Английский

---