HIRBERT

HIRBERT: Hyperspectral Imaging of River Bathymetry using an Ensemble of Regression Trees

This standalone software application provides a means of inferring water depth from passive optical remotely sensed data via a machine learning-based workflow: Hyperspectral Imaging of River Bathymetry using an Ensemble of Regression Trees, or HIRBERT for short. This approach could provide accurate depth estimates in clear-flowing, relatively shallow streams but is less likely to yield reliable bathymetric information in more turbid, deeper rivers. The primary inputs to the algorithm are paired observations of depth and spectral reflectance, which are used to train an ensemble of regression trees and perform an accuracy assessment via cross-validation. The resulting model can then be applied to the image from which the spectra were obtained to produce a continuous map of water depth. The software also provides tools to facilitate interpretation of the depth retrieval model.