Answers are grounded in this paper and cite their sources.
The Upper Indus Basin (UIB) is characterized by extreme topographic heterogeneity, making reliable geomorphometric characterization essential for hydrological and hazard-related modeling. However, the statistical and functional redundancy among commonly derived terrain parameters can introduce multicollinearity and unnecessary model complexity. In this study, we develop a quantitative framework to identify, evaluate, and correct redundancy among seven geomorphometric parameters, namely, elevation, slope, relief, roughness, aspect, hillshade, and a drainage index (DI) derived from 30 m Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) data. Pearson correlation, spatial autocorrelation, variance inflation factor (VIF), and principal component analysis (PCA) were used to quantify parameter dependence and develop a parsimonious terrain representation. The results demonstrate pronounced redundancy with a strong negative correlation between slope and the drainage index (r = −0.76), a near-perfect positive correlation between relief and roughness (r = 0.983), and a strong negative correlation between aspect and hillshade (r = −0.60). VIF analysis further confirmed substantial multicollinearity between relief and roughness (VIF ≈ 29.2–42.2) and between slope and the drainage index (VIF ≈ 4.9–9.0). In contrast, global Moran’s I values ranged from −0.005 to 0.006, with p-values of 0.3467–0.6638, indicating no statistically significant global spatial autocorrelation. PCA reduced the seven original parameters to five orthogonal components, achieving a 28.6% reduction in dimensionality while retaining approximately 95% of the variance; the maximum inter-component correlation was only 8.6 × 10 −8 . PC1 explained 29.2% of variance and represented terrain ruggedness and relief intensity, while PC2 explained 28.0% and represented drainage-related terrain organization, and PC3 explained 22.9% and represented slope orientation and illumination. Elevation was isolated predominantly in PC4 (loading = 0.989), confirming its distinct topographic contribution. Importantly, the PCA-based Random Forest model improved predictive performance relative to the seven-parameter model, increasing the coefficient of determination (R 2 ) from 0.94 to 0.99 and reducing root mean square error (RMSE) from 110.58 to 67.04, mean absolute error (MAE) from 58.57 to 25.51, and mean absolute percentage error (MAPE) from 1.549% to 0.680%. These results demonstrate that PCA provides an effective means of eliminating geomorphometric redundancy while preserving essential terrain information, establishing a robust and parsimonious topographic baseline for future hydrological, geomorphological, and hazard modeling in the UIB.
No takes yet. Share an insight, caveat, or question.
Khan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: