Sentinel-3 Band Simulation to Estimate Chlorophyll-a Concentrations from Empirical Models on Amazon Case II Water Types
Sander de Carvalho, Lino Augusto1; Faria Barbosa, Cláudio Clemente2; Leão de Moraes Novo, Evlyn Márcia3
1National Institute for Space Research - INPE, BRAZIL; 2Nationa Institute for Space Research - INPE, BRAZIL; 3National Institute for Space Research, BRAZIL

Empirical Models to estimate chlorophyll-a concentration are among the most used in the remote sensing community. Historically, those models were applied to ocean monitoring, currently through ENVISAT - MERIS remote sensing images. However, in the last decade, an enormous effort has been carried out to make them suitable for Case II water types. In particular, tropical inland water bodies characterized by highly complex mixture of optically active components, represent a challenge, especially regarding the accuracy of chlorophyll-a estimates. Sentinel-3 Ocean and Land Color Instrument (OLCI) project presents as an improvement from MERIS instrument, an increase in the number of spectral bands, from 15 to 21 and standard products with 300 m spatial resolution. The simulation of the spectral response of OLCI could indicate the potential of this sensor for Amazon water environmental monitoring. This work presents the results of the simulation of Sentinel-3 - OLCI sensor spectral bands using 295 field spectra, collected at 4 different states in an Amazon River hydrological cycle. In order to compare the Sentinel performance with MERIS, the same data were used to simulate MERIS bands. The samples are from the Curuai floodplain lakes located in the Pará State in the Amazon Region. A computational search method was used to find the band set that best fits to the two and three bands empirical models for both simulated sensors. Results showed that Sentinel simulated bands displayed the best fit with data collected during "high water" which is characterized by high concentration of chlorophyll-a and low concentration of inorganic particle concentration. During this period the R 2 values were 0.90 for the two band model and 0.95 for the three band model. For the remaining hydrological states, the R2 values were smaller during the high concentration of inorganic particles which made it difficult to uncouple the chlorophyll-a out of the water volume reflectance. R2 value ranged from 0.80 during the receding water to 0.20 in the rising water, due to the high turbidity and low chlorophyll concentration. The results showed, as well, that the Sentinel-3 models for high water performed better then MERIS, which presents R2 of 0.87 for the two-band model and 0.94 for the three-band model.