Abstract:Robust soil ecosystems are crucial for sustaining life on our planet and enhancing agricultural output. Soil erosion poses a significant threat to the health of these ecosystems. Gully erosion is capable of breaking down the soil ecosystem and degrades the quality of water in rivers and wetlands. This study aims to map and predict gully erosion susceptibility zones in the Birbhum district of West Bengal, India, using advanced machine learning models and nature-inspired optimization techniques. It addresses the significant issue of gully erosion, which degrades soil ecosystems and water quality. Eight machine learning models, namely Random Forest (RF), glmnet, C5.0, eXtreme Gradient Boosting (XGB), model averaged neural network (avNnet), AdaBoost, gradient boosting machine (GBM), and multivariate adaptive regression splines (MARS) algorithms were employed to map the susceptible zone for gully erosion with thirty-three conditioning parameters in the study area. Each of the main and sub factors were suitability ranked and weighted with the FuzzyAHP evaluation for integration with ML model. The parameters were selected through the novel nature inspired wrapper algorithms (Particle Swarm Optimization (PSO), Genetic Optimization (GO), Grey Wolf Optimization (GWO), Harris Hawk Optimization (HHO), and Dragon Fly Optimization (DFO)) and evaluated through multi-collinearity test as well as Boruta sensitivity analysis. Based on the validating parameters, such as accuracy, kappa, sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), and area under the receiver operating characteristic curve (AUROC) showed the RF, C5.0, XGB and AdaBoost models could be used for identifying susceptible gully erosion zones following 10-fold resampling technique. In terms of AUC, the RF, XGB and GBM models showed sound performance (> 0.92) along with good accuracy level (> 0.80). According to the Boruta sensitivity analysis, the top four parameter's namely elevation, distance to river, sand, and river density were the most influential factors. The severity of gully erosion was identified in the western and central zones of the study area having denser water bodies and comparatively higher run off depth. CMIP6 MIROC6 and EC-Earth3 climate data for ssp245 and ssp585 scenarios were also used for predicting future (2023–2060) gully erosion expansion trend in the study area for 10-fold resampling rate through the best performing ML models. The concerning situation of gully erosion in the study area required instant and promising steps to prevent soil losses through gully erosions promised to achieve the sustainable development goal (SDGs).

Land use/land cover classification using hyperspectral soil reflectance features in the Eastern Himalayas, India

Use of machine learning-based classification algorithms in the monitoring of Land Use and Land Cover practices in a hilly terrain

Spatial and temporal classification and prediction of LULC in Brahmani and Baitarni basin using integrated cellular automata models

Landuse classification of hyperspectral data by spectral angle mapper and support vector machine in humid tropical region of India

Land Use and Land Cover Classification with Hyperspectral Data: A comprehensive review of methods, challenges and future directions

Characterizing land use/land cover change dynamics by an enhanced random forest machine learning model: a Google Earth Engine implementation

Fusion of spectral and topographic features for land use mapping using a machine learning framework for a regional scale application

Comparative Analysis of Machine and Deep Learning Models for Soil Properties Prediction from Hyperspectral Visual Band

Use of support vector machine and cellular automata methods to evaluate impact of irrigation project on LULC

Spatio-temporal classification of land use and land cover and its changes in Kerala using remote sensing and machine learning approach

Land-use/cover change and future prediction by integrating the ML techniques of random forest and CA-Markov chain model of the Ganges alluvial tract of Eastern India

A machine learning approach to investigate the impact of land use land cover (LULC) changes on groundwater quality, health risks and ecological risks through GIS and response surface methodology (RSM)

Generation and classification of patch-based land use and land cover dataset in diverse Indian landscapes: a comparative study of machine learning and deep learning models

Land use land cover classification of remote sensing images based on the deep learning approaches: a statistical analysis and review

Insights from Big Spatial Data through Machine Learning Techniques for Prudent Management of Natural Resources

A comparative analysis of different pixel and object-based classification algorithms using multi-source high spatial resolution satellite data for LULC mapping

Accuracy of land use change detection using support vector machine and maximum likelihood techniques for open-cast coal mining areas

Imaging spectroscopic approach for land degradation studies: a case study from the arid land of India

Dimensionality reduction strategies for land use land cover classification based on airborne hyperspectral imagery: a survey

Detailed Land Use Classification in a Rare Earth Mining Area Using Hyperspectral Remote Sensing Data for Sustainable Agricultural Development

Integration of FuzzyAHP and machine learning algorithms for climate-driven gully erosion susceptibility mapping: predicting future trends in the eastern lateritic region, West Bengal, India