Estimating river discharge rate using remotely sensed thermal plumes -

Abstract

A methodology for river flow estimation from remotely sensed river thermal plume characteristics is introduced. Rule-based feature extraction tools were utilized to extract geometric characteristics of thermal plumes from 116 Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images of four rivers in Oregon State (Chetco, Coquille, Rogue and Siuslaw Rivers). The objective was to evaluate the ability of river thermal plume’s geometric characteristics (angle, length, deviation angle from the river channel centerline and mean tide level) to predict rivers discharge rate, measured independently through USGS stations. Multiple linear regression models using river plume’s geometric characteristics as independent variables accounted for most of the variability in the discharged flow (R²: 0.61, 0.54, 0.55, 0.57). Performance of these models was evaluated using statistical parameters and show satisfactory to good performance. Furthermore, high correlation is found between the river flow and the plume’s area and length variables. In fact, evaluation of the adopted plume’s area extraction method shows a Heidke Skill Score (HSS) values of 0.80, 0.75, 0.78 and 0.86, respectively for the four rivers. The proposed method is applied to estimate discharge from rivers with jetties at the mouth of the river.