Machine learning-based prediction of land use and land cover, and estimation of surface runoff using the GIS-based NRCS-CN model

Uttarakhand is a geodynamically active mountainous region located in the Central Himalaya, and is prone to multi-hazards such as floods, flash floods, earthquakes and landslides. The region is experiencing rapid changes in land use and land cover (LULC), which are increasing the frequency and intensity of natural hazards. The aim of the present study is to assess LULC changes from 2017 to 2023 using different machine learning (ML) algorithms and to predict future LULC for 2040. To analyse the influence of LULC variations, surface runoff is also estimated for the corresponding years. LULC maps for 2017, 2020, and 2023 are prepared in Google Earth Engine (GEE) using Sentinel-2 MultiSpectral Instrument (MSI) Level-1C data, and the Classification and Regression Trees (CART), Random Forest (RF), and Support Vector Machine (SVM) algorithms are applied. The future predicted LULC map for 2040 was carried out using the Random Forest algorithm in GEE. Corresponding to the LULC of each year, sub-basin-wise surface runoff was estimated using the NRCS-CN (National Resources Conservation Services-Curve Number) model utilizing significant parameters. The findings indicate that the RF model has the highest accuracy for LULC classification in all three years. Vegetation and barren land dominate the region, and settlement is expected to increase to about 6.76% of Uttarakhand by 2040. Predicted runoff for 2040 indicates dominance of low to moderate runoff in the study region, while very high to extreme runoff (>500 mm) is observed in the sub-basins of Girthiganga, Rishiganga, Dharamganga, Goriganga, Kuthi Yankti, and upper Kali in the Chamoli and Pithoragarh districts. The results highlight the significant influence of LULC dynamics on surface runoff and underscore their importance for flood risk management and sustainable land-use planning. • ML models (CART, RF, SVM, SmileRF) predicted LULC dynamics (2017–2040) in Uttarakhand • NRCS-CN model quantified basin-scale runoff using rainfall, soil, slope, and LULC data • Projected settlement growth (∼6.8%) and vegetation loss alter hydrological response • Chamoli and Pithoragarh basins indicate >500 mm runoff and elevated flood risk • Framework supports sustainable land-use planning and flood resilience strategies