Hydro-climatic trends and land use change impacts on river discharge: insights from statistical and machine learning approaches
摘要
In monsoon-dominated basins, assessment of combined effects of hydro-meteorological variability and land use/land cover (LULC) transformations on river discharge remains a critical yet less addressed challenge. Though climatic or land-use influences have often been examined separately, there is a gap in an integrated framework linking long-term trend analysis with predictive modeling. This study has addressed such gap by evaluating the concurrent impacts of hydro-meteorological trends and LULC changes on a river discharge with test case of Upper Subarnarekha River Basin, Jharkhand (India), considering a 30-year period (1991–2020) on decadal basis. Daily river discharge, rainfall, and temperature data from four active hydrological stations were analysed at seasonal and annual scales using non-parametric Mann–Kendall (MK) trend test and Sen’s slope estimator. Multi-temporal Landsat imagery helped to assess LULC dynamics, revealing a 355% expansion in built-up areas and 52.38% increase in agricultural land between 1991 and 2020. Rolling correlation analysis illustrated temporally varying relationships between hydro-climatic variables and discharge, consistently identifying rainfall as the dominant driver. Hydro Elastic Boosted Forest (HEBF) ensemble approach integrating Random Forest (RF) and ElasticNet techniques was employed for assessing feature importance. Hydro-meteorological variables were used as predictors and LULC data were incorporated exclusively in post-model analysis to explain observed discharge variability. The model showed strong performance, Root Mean Square Error (RMSE): 4.08–5.67 m3s−1; Mean Absolute Error (MAE): 3.49–5.58 m3s−1; R2: 0.87–0.95), reliably capturing nonlinear climate-discharge relationships. These findings reveal combined climatic and land use impacts on river discharge, informing adaptation, forecasting, and sustainable monsoon water management.