Analysis of Various Remote Sensing Algorithms for the Estimation of Chlorophyll-a Concentration in the Coastal Waters of Semarang, Indonesia

Phytoplankton are a key component of aquatic systems worldwide. Also known as microalgae, they are responsible for forming the base of food chains and cycling carbon. Therefore, it becomes important to monitor the amount of phytoplankton in a water body. The main way this is done is by measuring the amount of chlorophyll-a (chl-a) in the water. Chl-a can be measured through field missions. Although this method is one of the most accurate, it is rather time consuming and cannot cover a large spatial extent. Algorithms derived from remote sensing measurements can aid in fixing the temporal and spatial issues. In this study, three algorithms were applied on Sentinel 2 images to see how well they estimated chl-a for the coastal waters of Semarang, Indonesia. Semarang is one of the top five largest cities in Indonesia. Two important features of the city are the east and west flood canals that were constructed to deal with the heavy rains and tidal flooding. Beside their function as flood control, both canals may also contribute to the enrichment of the coastal water due to the run-off from the upland. In the present study, the three algorithms were tested against 140 points of in-situ measurements taken off the coast of the east and west flood canals. The satellite images used for the algorithms are from the same days and times as the collection of field data. Two of the algorithms picked used a blue-green band ratio. The first is the OC2 which uses two bands and the second is the OC3 which uses three bands. These two algorithms were not expected to perform well in areas that had higher amounts of colored dissolved organic matter (CDOM) and suspended sediments (SPM). The third algorithm chosen was the normalized difference chlorophyll index (NDCI) which uses bands in the red and NIR to minimize the confounding effects of the other optically active constituents. The results show that the OC2 and OC3 have some reliability for the area. These two algorithms did much better than the NDCI which had no correlation with the in-situ data. However, the estimation of chl-a for the OC2 and OC3 does not perform as well as the waters become more turbid. Therefore, even though the blue-green algorithms performed better than the NDCI, a new algorithm should be developed that can accurately estimate the chl-a concentration for clear and turbid waters in coastal areas of Semarang.