What question did this study set out to answer?

To develop a machine learning framework for predicting geomorphometric and hydrological outputs in the Gangolli River Basin.

April 19, 2026Open Access

Machine learning-based predictive modelling and feature importance analysis of geomorphometric parameters in the Gangolli River Basin, Western Ghats, India

GVGopika VinodManipal Academy of Higher Education VSV. ShwethaManipal Academy of Higher Education MSMaddodi B SManipal Academy of Higher Education

Key Points

To develop a machine learning framework for predicting geomorphometric and hydrological outputs in the Gangolli River Basin.
Utilized machine learning techniques to analyze geomorphometric parameters.
Employed regression models, including Random Forest and Linear Regression, on a dataset from five sub-basins.
Assessed model performance using Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, and R2.
Random Forest achieved an RMSE of 0.156, indicating robust performance.
Linear Regression demonstrated a high explanatory power with an R2 of 0.934.
Findings highlight the efficacy of machine learning combined with geomorphometric analysis in watershed characterization.

Abstract

One of the world's hotspots for biodiversity, the Western Ghats are distinguished by their complicated topography, highly variable rainfall, and ecologically delicate river basins. Therefore, assessing flood risk and managing watersheds in this area require an understanding of geomorphometric influences on hydrological response. Effective watershed management depends on the morphological development and hydrological behaviour of river basins. Time, labour, and computational demands frequently limit traditional methods that rely on topographic maps, field surveys, and manual interpretation of Digital Elevation Models (DEMs). In the Gangolli River Basin, which is situated in the Udupi district of Karnataka, in the Western Ghats of India, this study offers a machine learning-based framework to forecast important geomorphometric and hydrological outputs, namely total stream length, number of streams, and drainage density. The number of streams, drainage density, and overall stream length were all predicted using different regression models. Regression-based models, such as Random Forest and Linear Regression, were used to process and analyze a large dataset that included aerial, linear, and relief-based variables from five sub-basins (secondary data). Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R2) were used to assess the model's performance. Random Forest yielded a lower RMSE (0.156), demonstrating its robustness in capturing complicated non-linear interactions, whereas Linear Regression achieved a significant explanatory power with an R2 of 0.934. Given the limited dataset size, the results should be interpreted as indicative and exploratory rather than definitive predictive outcomes. The findings show that in monsoon-dominated areas of the Western Ghats, combining machine learning with geomorphometric analysis offers a dependable, effective, and scalable method for initial watershed characterization and river basin planning.

KI fragen

Bookmark

View Full Paper