ABSTRACT
Remote sensing of chlorophyll-a is challenging in water containing inorganic suspended sediments (i.e. non-volatile suspended solids, NVSS) and coloured dissolved organic matter (CDOM). The effects of NVSS and CDOM on empirical remote-sensing estimates of chlorophyll-a in inland waters have not been determined on a broad spatial and temporal scale. This study evaluated these effects using a long-term (1989–2012) data set that included chlorophyll-a, NVSS, and CDOM from 39 reservoirs across Missouri (USA). Model comparisons indicated that the machine-learning algorithm BRT (boosted regression trees, validation Nash–Sutcliffe coefficient = 0.350) was better than linear regression (validation Nash–Sutcliffe coefficient = 0.214) for chlorophyll-a estimate using Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) imagery. Only a small proportion of BRT model residuals could be explained by sediments or CDOM, and the observed trends in BRT residuals were different from the theoretical effects expected from NVSS and CDOM. Our results also indicated a small systematic bias by the BRT model, but it was not likely caused by NVSS or CDOM.
Acknowledgements
This work was supported by the Environmental Protection Agency (EPA), USA: [Grant Number R835203]. The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of US EPA, or any other agency of the US government. We thank Dr David W. Hyndman and Dr Stephen K. Hamilton for reviewing this manuscript and providing useful comments. We also appreciate the comments from the anonymous reviewers of this journal.
Disclosure statement
No potential conflict of interest was reported by the authors.