Indexed by:
Abstract:
Chlorophyll-a(Chl-a) concentration is one of the most important water quality parameters, and can be directly retrieved via remote sensing. It can be used to assess the water eutrophication in coastal waters. In this study, we proposed a random forest (RF) machine learning approach based on MODIS time-series images combining in-situ float data to retrieve Chl-a concentration in coastal waters of the Fujian Province, and compared with a traditional band ratio (BR) model. The RF model can well estimate the nearshore Chl-a concentration, and outperformed the BR model. The results were validated by using float measurement data. The R2 for RF model is as high as 0.87, while only 0.21 for BR model. The RMSE is reduced from 0.52 μg•L-1 for BR to 0.49 μg•L-1 for RF, and the MAPEs are 37.50% and 50.20% for RF and BR respectively, suggesting significant accuracy improvement by RF model. This study can provide a useful method for studying nearshore Chl-a concentration variability and monitoring water quality in the Fujian Province. ©2018, Science Press. All right reserved.
Keyword:
Reprint 's Address:
Email:
Source :
Acta Scientiae Circumstantiae
ISSN: 0253-2468
Year: 2018
Issue: 12
Volume: 38
Page: 4831-4839
Cited Count:
SCOPUS Cited Count: 6
ESI Highly Cited Papers on the List: 0 Unfold All
WanFang Cited Count:
Chinese Cited Count:
30 Days PV: 1
Affiliated Colleges: