Water-Pixel Dilution Bias: A Systematic Measurement Error in Riverine Spatial Feature Engineering

This paper identifies and formalizes a previously unrecognized systematic measurement error in riverine spatial feature engineering, termed Water-Pixel Dilution Bias (WPDB). We derive a closed-form analytical equation that quantifies the bias caused by including open-water pixels in terrestrial land-cover normalization, introduce a flow-aware non-water normalization framework to eliminate the error, and validate the method across two independent river systems. The proposed framework improves the fidelity of spatial predictors and provides a general methodology applicable to riverine GIS, remote sensing, and environmental data science.