Spatial Prediction of Soil Continuous and Categorical Properties Using Deep Learning Approaches for Tamil Nadu, India DOI Creative Commons

Thamizh Vendan Tarun Kshatriya,

R. Kumaraperumal, S. Pazhanivelan

et al.

Agronomy, Journal Year: 2024, Volume and Issue: 14(11), P. 2707 - 2707

Published: Nov. 17, 2024

Large-scale mapping of soil resources can be crucial and indispensable for several the managerial applications policy implications. With machine learning models being most utilized modeling technique digital (DSM), implementation model-based deep methods spatial predictions is still under scrutiny. In this study, continuous (pH OC) categorical variables (order suborder) were predicted using learning–multi layer perceptron (DL-MLP) one-dimensional convolutional neural networks (1D-CNN) entire state Tamil Nadu, India. For training models, 27,098 profile observations (0–30 cm) extracted from generated database, considering series as distinctive stratum. A total 43 SCORPAN-based environmental covariates considered, which 37 retained after recursive feature elimination (RFE) process. The validation test results obtained each attributes both algorithms comparable with DL-MLP algorithm depicting attributes’ intricate organization details, compared to 1D-CNN model. Irrespective datasets, R2 RMSE values pH attribute ranged 0.15 0.30 0.97 1.15, respectively. Similarly, OC 0.20 0.39 0.38 0.42, Further, overall accuracy (OA) order suborder classification 39% 67% 35% 64%, explicit quantification covariate importance derived permutation implied that tried incorporate respect genesis study. Such approaches integrating soil–environmental relationships limited parameterization computing costs serve a baseline emphasizing opportunities in increasing transferability generalizability model while accounting associated dependencies.

Language: Английский

The problematic case of data leakage: A case for leave-profile-out cross-validation in 3-dimensional digital soil mapping DOI Creative Commons
Kingsley John, Daniel D. Saurette, Brandon Heung

et al.

Geoderma, Journal Year: 2025, Volume and Issue: 455, P. 117223 - 117223

Published: March 1, 2025

Language: Английский

Citations

2
Integrating multi-year crop inventories as a proxy for soil management within a digital soil mapping framework for predicting nitrogen indices DOI Creative Commons
Luke Laurence, Brandon Heung, Jin Zhang

et al.

Geoderma, Journal Year: 2024, Volume and Issue: 448, P. 116944 - 116944

Published: June 25, 2024

For the international digital soil mapping (DSM) community, adequate spatial estimates of nitrogen (N) mineralization have yet to be generated. This is due, in part, an inability capture critical N controls at regional and provincial scales. While influence climate, vegetation, relief are accessible predictors DSM, effect management known for its important on dynamics, but has hitherto been elusive mappers. purpose producing maps inform fertilizer management, intention this study was determine importance novel crop frequency layers, as a proxy through development scale DSMs total (TN), biological availability (BNA) estimate over growing season (GSN) calculated from TN BNA results. Crop covariates were developed that estimated particular type planted 10-year period, thus capturing cropping system tillage intensity. results 27% higher using layers support vector machine learner, with Lin's concordance correlation coefficient (concordance) 0.45. predictions increased by 24% stochastic gradient boosting learner final GSN showed least improvement (6%) resulted highest (0.47) learner. The stable pool, represented TN, climate importance; whereas, labile based measures, best predicted controlled organism covariates. successful inclusion into indicated number times forages potatoes period greatest importance. As intensity most pronounced potatoes, contribute biomass building organic matter levels, increasing years had positive pools.

Language: Английский

Citations

6
Satellite Soil Observation (Satsoil): Extraction of Bare Soil Reflectance for Soil Organic Carbon Mapping on Google Earth Engine DOI
Morteza Khazaei, Preston Sorenson, Ramata Magagi

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Improved estimation of soil organic carbon stock in subtropical cropland of Southern China based on digital soil mapping and multi-sources data DOI Creative Commons
Bifeng Hu, Qian Zhu, Modian Xie

et al.

Geo-spatial Information Science, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 24

Published: March 12, 2025

Language: Английский

Citations

0
Comparison of Machine Learning and Geostatistical Methods on Mapping Soil Organic Carbon Density in Regional Croplands and Visualizing Its Location‐Specific Dominators via Interpretable Model DOI Open Access
Bifeng Hu,

Yibo Geng,

Yi Lin

et al.

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

Published: March 17, 2025

ABSTRACT High‐precision soil organic carbon density (SOCD) map is significant for understanding ecosystem cycles and estimating storage. However, the current mapping methods are difficult to balance accuracy interpretability, which brings great challenges of SOCD. In present research, a total 6223 samples were collected, along with data pertaining 30 environmental covariates, from agricultural land located in Poyang Lake Plain Jiangxi Province, southern China. Furthermore, ordinary kriging (OK), geographically weighted regression (GWR), random forest (RF), empirical Bayesian (EBK), three hybrid models (RF‐OK, RF‐EBK, RF‐GWR), constructed. These used SOCD (soil density) study region high resolution m. After that, shapley additive explanations (SHAP) quantify global contribution spatially identify dominant factors that influence variation. The outcomes suggested compared single geostatistics model model, RF method emerged as most effective predictive showcasing superior performance (coefficient determination ( R 2 ) = 0.44, root mean squared error (RMSE) 0.61 kg m −2 , Lin's concordance coefficient (LCCC) 0.58). Using SHAP, we found properties contributed prediction (81.67%). At pixel level, nitrogen dominated 50.33% farmland, followed by parent material (8.11%), available silicon (8.00%), annual precipitation (5.71%), remaining variables accounted less than 5.50%. summary, our offered valuable enlightenment toward achieving between interpretability digital mapping, deepened spatial variation farmland

Language: Английский

Citations

0
Spatial-machine learning framework for rapid identification of soil cadmium risk in high geochemical background areas DOI
Cheng Li, Zhongfang Yang, Dong‐Xing Guan

et al.

Journal of Hazardous Materials, Journal Year: 2025, Volume and Issue: unknown, P. 138091 - 138091

Published: April 1, 2025

Language: Английский

Citations

0
Importance of Spectral Information, Seasonality, and Topography on Land Cover Classification of Tropical Land Cover Mapping DOI Creative Commons
Chansopheaktra Sovann, Stefan Olin, Ali Mansourian

et al.

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(9), P. 1551 - 1551

Published: April 27, 2025

Tropical forests provide essential ecosystem services, playing a critical role in climate regulation, biodiversity conservation, and regional hydrological cycles while also supporting livelihoods. However, they are increasingly threatened by deforestation land-use change. Accurate land cover (LC) mapping is vital to monitor these changes, but tropical challenging due complex spatial patterns, spectral similarities, frequent cloud cover. This study aims improve LC classification accuracy such heterogeneous forest region Southeast Asia, namely Kulen, Cambodia, which characterized natural forests, regrowth agricultural lands including cashew plantations croplands, using Sentinel-2 imagery, recursive feature elimination (RFE), Random Forest. We generated 65 variables of bands, indices, bi-seasonal differences, topographic data from Level-2A Shuttle Radar Topography Mission datasets. These were extracted 1000 random points per 12 classes reference polygons based on observed GPS points, Uncrewed Aerial Vehicle high-resolution satellite data. The models optimized through correlation-based filtering with hyperparameter tuning accuracy, validated via confusion matrices comparisons global national-scale products. Our results highlight the significant as elevation slope, along red-edge bands indices related tillage, leaf water content, greenness, chlorophyll, tasseled cap transformation for mapping. integration datasets improved particularly like semi-evergreen deciduous forests. Furthermore, reduced variable set 19, improving model efficiency without sacrificing accuracy. Combining selection methods classification, providing more reliable product that outperforms existing products proves valuable monitoring, management, use studies.

Language: Английский

Citations

0
Incorporating forest canopy openness and environmental covariates in predicting soil organic carbon in oak forest DOI

Su Lei,

Mehdi Heydari, Maryam Sadat Jaafarzadeh

et al.

Soil and Tillage Research, Journal Year: 2024, Volume and Issue: 244, P. 106220 - 106220

Published: July 1, 2024

Language: Английский

Citations

2
Spatial disaggregation of a legacy soil map to support digital soil and land evaluation assessments in Scotland DOI Creative Commons
Zisis Gagkas, Allan Lilly

Geoderma Regional, Journal Year: 2024, Volume and Issue: 38, P. e00833 - e00833

Published: July 1, 2024

In recent years, the importance of soils and soil functions has been recognised for supporting delivery ecosystem services realisation international initiatives, such as UN Sustainable Development Goals. At same time, Digital Soil Mapping (DSM) emerged a modelling technique that can satisfy increased end-user needs new datasets by producing fine resolution property maps to support complex digital land evaluation assessments. Spatial disaggregation is popular DSM used transform legacy more spatially-explicit datasets, which also be in conjunction with databases generate maps. this study, we performed spatial National Map Scotland (originally published at 1:250,000 scale) taxonomic level Series, specific objective facilitate production harmonised assessments through linking Scottish Database. We divided into Landscape Units similar landform characteristics trained probability random forest models within each Unit using area-proportion sampling both single- multiple- (complex) Series map units selected environmental covariates produce layers 50 m grid resolution. The performance disaggregated was evaluated prediction uncertainties individual types independent profile classifications. Evaluation results indicated algorithm successful promoting effective single polygons provided good accuracies most exception some least extensive typically found units. This attributed mainly algorithm's tendency favour dominant, classes, along its difficulty distinguish between spatially diverse areas. However, training instead nationally helped limit underestimation these minority overestimation dominant ones. addition, showed usefulness generated conditional probabilities exploring variability, especially areas river floodplains covered multiple alluvial non-alluvial soils. Overall, study demonstrates potential extract pedological knowledge embedded use it dynamic effectively readily-available easily-updated information from existing databases.

Language: Английский

Citations

2
Mapping the soil C:N ratio at the European scale by combining multi-year Sentinel radar and optical data via cloud computing DOI
Xinyue Wang,

Yajun Geng,

Tao Zhou

et al.

Soil and Tillage Research, Journal Year: 2024, Volume and Issue: 245, P. 106311 - 106311

Published: Sept. 24, 2024

Language: Английский

Citations

2