Prediction and Mapping of Topsoil Organic Carbon Content in the Provence Coal Field, France: A Machine Learning and Deep Learning Approach DOI
Mounir Oukhattar, Mounir Oukhattar, Sébastien Gadal

и другие.

Опубликована: Янв. 1, 2023

Soil organic carbon content (SOC) plays a crucial role in cycle management and soil fertility. In this study, spatial modelling approach of the dynamics SOC distribution between 2003 2022, as well its relationship with land use/land cover (LULC) change Provence coal field (PCF) France, was carried out by performing regression using random forest (RF), support vector machine (SVM), gradient boosting (GBM), deep neural network (DNN) fed 21 predictors data from 162 sites. Predictors were extracted multispectral images. The results show that forests contain significantly more than other types LULC (average 69.3g/kg), while arable has lowest average (8.9g/kg). Although shows certain proportionality cover, soils artificial areas have relatively high (33.6g/kg). correlations suggest LULC, topography, parameters environmental indices are main factors influencing PCFs. RF model proved to be best for predicting SOC. maximum overall accuracy (OA) maps generated reached (0.84) coefficient determination (R2) 0.81 compared an OA=0.76 R2=0.83 GBM model, OA=0.75 R2=0.6 DNN OA=0.71 R2=0.3 SVM. However, visual point view, showed better match reality PCF indicating learning effectively captures features reducing significant variations. study provide important guidance management, which could prove beneficial mitigating climate through sustainable use practices.

Язык: Английский

A critical systematic review on spectral-based soil nutrient prediction using machine learning DOI
Shagun Jain, Divyashikha Sethia, K. C. Tiwari

и другие.

Environmental Monitoring and Assessment, Год журнала: 2024, Номер 196(8)

Опубликована: Июль 4, 2024

Язык: Английский

Процитировано

7
Geospatial prediction of total soil carbon in European agricultural land based on deep learning DOI
Dorijan Radočaj, Mateo Gašparović,

Petra Radočaj

и другие.

The Science of The Total Environment, Год журнала: 2023, Номер 912, С. 169647 - 169647

Опубликована: Дек. 26, 2023

Язык: Английский

Процитировано

16
Digital mapping of soil organic carbon in a plain area based on time-series features DOI Creative Commons
Kun Yan, Decai Wang,

Yongkang Feng

и другие.

Ecological Indicators, Год журнала: 2025, Номер 171, С. 113215 - 113215

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0
A partitioned conditioned Latin hypercube sampling method considering spatial heterogeneity in digital soil mapping DOI Creative Commons
Biao Huang, Guangjun Yang, Jing Lei

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 14, 2025

The design of sampling methods is crucial in digital soil mapping for organic carbon (SOC), as it directly affects prediction precision and reliability. While based on environmental variables are widely used, the spatial heterogeneity properties poses challenges by introducing variability influential driving factors across subregions, potentially reducing accuracy. To address this, a partitioned conditioned Latin hypercube (PcLHS) method explicitly considering proposed. PcLHS first employs regionalization with dynamically constrained agglomerative clustering partitioning (REDCAP) to partition study area into relatively homogeneous subregions. Key then identified using Boruta Variance Inflation Factor method, followed (cLHS) select training points within each subregion. Finally, selected combined form complete dataset. A case SOC northeastern France demonstrated that consistently outperformed traditional methods, achieving lower root mean square error (RMSE, 0.40-0.43), higher coefficient determination (R2, 0.36-0.44), improved concordance correlation (CCC, 0.58-0.63). Compared other reduced RMSE 4-11%, increased R2 18-46%, CCC 14-29%. These results highlight necessity establish an effective heterogeneous landscapes.

Язык: Английский

Процитировано

0
Soil Organic Carbon (SOC) Prediction using Super Learner Algorithm Based on the Remote Sensing Variables DOI Creative Commons
Y. Jo, Palash Panja, Hanseup Kim

и другие.

Environmental Challenges, Год журнала: 2025, Номер unknown, С. 101160 - 101160

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Soil Organic Carbon Sequestration Potential, Storage, and Influencing Mechanisms in China DOI
Jinhua Cao, Zipeng Zhang, Jianli Ding

и другие.

Land Degradation and Development, Год журнала: 2025, Номер unknown

Опубликована: Май 8, 2025

ABSTRACT The soil organic carbon sequestration potential (SOC sp ) has important implications for the global cycle and responses to climate change. However, there is a dearth of spatial information specifically China within this field, our knowledge regarding factors influencing SOC remains somewhat limited. To solve problem, study utilized legacy data collected in 1980s (1979–1984s), combined with climatic landscape zoning, adopted digital mapping techniques produce prediction models density five designated depths. results showed that accuracy top (0–30 cm) model was higher than subsoil (30–100 model. highest northwestern, northern, eastern lowest southeastern Tibetan Plateau northeastern China. Scale‐ location‐specific effects environmental on SOCs were observed, two‐factor being stronger those their one‐factor counterparts. Spatial differentiation characteristics drivers between topsoil layers show significant zonal differences. In layer, vegetation are dominant arid zone, while semi‐arid zone mainly regulated by land use; use together dominate zones. study, we provide support pathway change mitigation processes, emphasizing importance in‐depth studies mechanisms dynamics through its driving mechanisms.

Язык: Английский

Процитировано

0
Spatioemporal dynamics and driving forces of soil organic carbon changes in an arid coal mining area of China investigated based on remote sensing techniques DOI Creative Commons

Xuting Yang,

Xiao Bai, Wanqiang Yao

и другие.

Ecological Indicators, Год журнала: 2023, Номер 158, С. 111453 - 111453

Опубликована: Дек. 22, 2023

Soil organic carbon (SOC) undergoes rapid changes due to human production activities, which have an impact on the land cycle and ultimately global change. As one of main coal mining significantly impacts soil cycle. However, lack remote sensing modeling in areas, spatio-temporal driving mechanisms SOC areas remain unclear. Therefore, this study investigated determined data from 300 sampling points (depth 0–20 cm) located arid area China. Remote images were then used established a density (SOCD) prediction model within Random Forest (RF) achieve digital mapping stocks (SOCS). The spatiotemporal SOCS analyzed using mapping, influencing mechanism was revealed path analysis. results showed that constructed SOCD predictive meets demand for (R2 ≥ 0.74, p < 0.01, RMSE ≤ 1.72 kg/m2). Under combined influence reclamation, total amount surface exhibited fluctuating upward trend 1990 (6.34 Tg) 2021 (7.73 Tg), with annual growth rate 0.038 Tg/a. spatial distribution generally increased southeast northwest. Precipitation, Normalized Difference Vegetation Index (NDVI), use positively correlated distribution, while temperature, elevation, erosion, intensity negatively SOCS. degree factors as follows: NDVI > erosion precipitation elevation temperature. negative mainly indirect, through disturbance vegetation, erosion. uneven ground subsidence stretching caused by contribute intensified vegetation degradation affected area, leading reduction did not decrease under high mining, related increase area. In study, based evaluate temporal can serve valuable references scientific improvement ecological environment rational planning construction, well low-carbon reclamation compensation assessments.

Язык: Английский

Процитировано

8
Prediction of Spatial Distribution of Soil Organic Carbon in Helan Farmland Based on Different Prediction Models DOI Creative Commons
Yuhan Zhang, Youqi Wang,

Yi‐Ru Bai

и другие.

Land, Год журнала: 2023, Номер 12(11), С. 1984 - 1984

Опубликована: Окт. 27, 2023

Soil organic carbon (SOC) is widely recognized as an essential indicator of the quality arable soils and health ecosystems. In addition, accurate understanding spatial distribution soil content for precision digital agriculture important. this study, in topsoil was determined using four common machine learning methods, namely back-propagation neural network model (BPNN), random forest algorithm (RF), geographically weighted regression (GWR), ordinary Kriging interpolation method (OK), with Helan County study area. The prediction accuracies different models were compared conjunction multiple sources auxiliary variables. BPNN (MRE = 0.066, RMSE 0.257) > RF 0.186, 3.320) GWR 0.193, 3.595) OK 0.198, 4.248). Moreover, trends SOC predicted similar: high western area low eastern region. better handled nonlinear relationship between multisource variables presented finer information differentiation. These results provide important theoretical basis data support to explore trend content.

Язык: Английский

Процитировано

3
Soil Carbon Stock Modelling in the Forest-Tundra Ecotone Using Drone-Based Lidar DOI
Claire Céline Devos, Erik Næsset, Mikael Ohlson

и другие.

Опубликована: Янв. 1, 2024

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

Процитировано

0
Prediction and Mapping of Topsoil Organic Carbon Content in the Provence Coal Field, France: A Machine Learning and Deep Learning Approach DOI
Mounir Oukhattar, Mounir Oukhattar, Sébastien Gadal

и другие.

Опубликована: Янв. 1, 2023

Soil organic carbon content (SOC) plays a crucial role in cycle management and soil fertility. In this study, spatial modelling approach of the dynamics SOC distribution between 2003 2022, as well its relationship with land use/land cover (LULC) change Provence coal field (PCF) France, was carried out by performing regression using random forest (RF), support vector machine (SVM), gradient boosting (GBM), deep neural network (DNN) fed 21 predictors data from 162 sites. Predictors were extracted multispectral images. The results show that forests contain significantly more than other types LULC (average 69.3g/kg), while arable has lowest average (8.9g/kg). Although shows certain proportionality cover, soils artificial areas have relatively high (33.6g/kg). correlations suggest LULC, topography, parameters environmental indices are main factors influencing PCFs. RF model proved to be best for predicting SOC. maximum overall accuracy (OA) maps generated reached (0.84) coefficient determination (R2) 0.81 compared an OA=0.76 R2=0.83 GBM model, OA=0.75 R2=0.6 DNN OA=0.71 R2=0.3 SVM. However, visual point view, showed better match reality PCF indicating learning effectively captures features reducing significant variations. study provide important guidance management, which could prove beneficial mitigating climate through sustainable use practices.

Язык: Английский

Процитировано

0