Exploring Random Forest Machine Learning and Remote Sensing Data for Streamflow Prediction: An Alternative Approach to a Process-Based Hydrologic Modeling in a Snowmelt-Driven Watershed

Physically based hydrologic models require significant effort and extensive information for development, calibration, validation. The study explored the use of random forest regression (RFR), a supervised machine learning (ML) model, as an alternative to physically Soil Water Assessment Tool (SWAT) predicting streamflow in Rio Grande Headwaters near Del Norte, snowmelt-dominated mountainous watershed Upper Basin. Remotely sensed data were used analysis (RFML) RStudio processing synthesizing. RFML model outperformed SWAT accuracy demonstrated its capability this region. We implemented customized approach RFR assess model’s performance three training periods, across 1991–2010, 1996–2010, 2001–2010; results indicated that improved with longer implying trained on more extended period is better able capture parameters’ variability reproduce accurately. variable importance (i.e., IncNodePurity) measure revealed snow depth minimum temperature consistently top two predictors all periods. paper also evaluated how well performs reproducing conventional approach. needed time set up calibrate, delivering acceptable annual mean simulation, satisfactory index agreement (d), coefficient determination (R2), percent bias (PBIAS) values, but monthly simulation warrants further exploration adjustments. recommends exploring snowmelt runoff processes, dust-driven sublimation effects, detailed topographic input parameters update routine flow estimation. provide critical enhancing prediction, which valuable research water resource management, including snowmelt-driven semi-arid regions.