343 - Application of Machine Learning Algorithms for the Estimation of the Concentration of Total Suspended Solids in the Colorado River Using Landsat 8 Operational Land Imager Data

The Colorado River housing several dams is a source of water to about 40 million people and 5.5 million acres of farmlands in the western United States of America and the Republic of Mexico. The quality of water in the river system is impacted by several factors including river modifications, anthropogenic activities, and drought events. Total suspended solids (TSS) is one of the water quality parameters (WQPs) that impair the quality of the river’s water. Increased concentration of TSS in the water body does not only impact its quality but also increases its turbidity thereby reducing light penetrability into the water for aquatic photosynthetic processes. Assessing and managing the concentrations of TSS in waterbodies is important to understand the changes they undergo to take pragmatic steps in protecting and restoring the quality of the water and to ensure the sustenance of the limited water resources. Monitoring of these parameters has been performed using traditional approaches involving laboratory analysis and field sampling. Conducting WQP assessment on a waterbody like the Colorado River is however labor and capital-intensive due to the length and the extent of the river. Remote sensing, however, provides an alternative to WQP monitoring due to its potential benefits in offering spatiotemporal coverages as well as their cost-effectiveness. This study utilized satellite images from the 30 m resolution Landsat 8 Operational Land Imager applied machine learning (ML) regression models including random forest, gradient boosting, adaptative boosting, and extreme gradient boosting for the estimation of the concentration of optically active TSS. Preliminary results showed immense potential for utilizing remote-sensed ML-based models for estimating the concentrations of TSS in the Colorado River System.