Comparison of reflectance algorithms in multisensor images for time series mapping of Chlorophyll-A Concentrationin Tiab Rural Wetland

The oceans cover about 70% of the earth's surface and contain the most water on Earth, as well as important marine ecosystems. In generally, global waters are classified into two types of water (the first case and the second case). In waters of the first type, such as the waters of the open ocean, phytoplankton dominate the inherent optical properties of water. However Case-2 waters, like coastal waters, are complex waters that are affected by a variety of active light compounds such as phytoplankton, colored dissolved organic matter and Total suspended matter. Coastal wetlands are considered as the Case-2 water. These types of areas are dynamic environments that are threatened by the entry of pollutants and because the wetlands have a calm environment and away from open sea waves, they are exposed to the accumulation of natural and human pollution. As a result, the identification and monitoring of coastal and marine pollution is essential to minimize their destructive effects on human health and the environment and economic damage to coastal communities.  Phytoplankton are floating or scattered single-celled algae that travel primarily through water waves. Chlorophyll-a considered as an indicator of the abundance of phytoplankton and biomass in oceanic, coastal and lake waters. Field and laboratory methods are difficult and time consuming and weak for spatial and temporal observations. In contrast to the weakness of field methods, remote sensing methods can provide the spatial perspective needed to gather information on ocean and coastal water surface on a regional and global scale. The purpose of this study was to compare and evaluate atmospheric correction methods (high atmospheric radiation and high atmospheric reflectance) on the algorithm for estimating the concentration of chlorophyll A based on blue and green bands (OC2) in Landsat-8 and Sentinel-2 data, evaluating the results using Field data and finally the time series mapping of chlorophyll-a concentration.

In this study, Landsat 8, Sentinel 2 satellite time series data and field data collected from the study area were used. First, the satellite images used in ENVI 5.3.1 software were converted to Surface Reflectance and Top of Atmosphere Reflectance. Then, MATLAB 2018a software was used for image processing and coding. to estimate the chlorophyll-A concentration, the bio-optical algorithm OC2 was used, which in fact uses a nonlinear relationship to link between field data and satellite data. In order to evaluate the results two statistical parameters R2 and RMSE were used.

Based on the analysis of field data, the concentration of chlorophyll-A in all sampled stations was less than 1 mg/m3. Water in the Surface Reflectance and Top of Atmosphere Reflectance Sentinel 2 and Landsat 8 data had a relatively similar spectral signature at wavelengths, due to the similarity in the spectral signature of water on the satellites used, covering the same spectral range in the Landsat 8 and Sentinel 2 satellites systems. The OC2 algorithm had amounts R2 (0.91 and 0.64) and RMSE (0.13 and 0.33) in Landsat 8 and Sentinel 2 Surface Reflectance data, respectively, while Landsat 8 and Sentinel 2 Top of Atmosphere Reflectance data had amounts R2 (0.12 and 0.53) and RMSE (0.45 and 0.51), respectively. The time series of chlorophyll-A concentration estimated using surface reflectance data (Landsat 8) corresponds to the natural conditions of the region, However, the time series of chlorophyll-a concentrations using the surface reflectance data (Sentinel 2) during the seasons estimated the chlorophyll concentration to be uniformly and downward. The reason for this poor performance in the Sentinel 2 is the lack of sufficient field data for calibration.

In this study, we tried to evaluate and compare the reflectance algorithms (Landsat 8 and Sentinel 2) in the OC2 algorithm. Preliminary results indicate that the type of satellite data used (surface reflectance and Top Atmosphere reflectance) is of great importance for entering the OC2 bio-optical algorithm because the satellite image to enter the OC2 algorithm must be surface reflectance data and atmospheric correction that In fact, these algorithms are sensitive to high-atmosphere reflectance data. In general, the results showed that 10 field data is enough to calibrate with Landsat 8 data, but for Sentinel 2 data, more than 10 numbers field data must be calibrated to obtain a good result.