Ocean color as a proxy to predict sea surface salinity of the Banda Sea
Sam Wouthuyzen (a,f*), Edi Kusmanto (a), Muhammad Fadli (b,f), Gentio Harsono (c,d), Gerry Salamena (b,e,f), Johanes Lekalette (b,f) and Augy Syahailatua (b,f)
a) Research Centre for Oceanography – Indonesian Institute of Sciences. Jl. Pasir Putih 1, Ancol Timur, Jakarta 14430, Indonesia.
b) Centre for Deep Sea Research – Indonesian Institute of Sciences. Jl. Y. Syaranamual, Poka, Ambon 97233 – Indonesia
c) Centre for Hydrography and Oceanography – Indonesian Navy, Jakarta, Indonesia.
d) Departement of Remote Sensing Technology, Indonesian Defence University, Bogor, Indonesia
e) Graduate Research School, College of Science and Engineering, James Cook University, Australia
f) Centre of Excellence for Tuna Conservation – Ambon, Jl. Y. Syaranamual, Poka, Ambon 97233 – Indonesia
Abstract
Salinity is an important conservative tracer in the ocean considered as a proxy to explain physical and chemical processes in the system (e.g. upwelling process and nutrient flux) controlling biological activities (e.g. primary production). Combining with temperature and chlorophyll-a (Chl-a), these three oceanographic parameters are important to reveal the water quality of a marine system supporting fishery. In a significantly large spatial scale system of ocean processes such as upwelling systems, the availability of spatial and temporal sea surface salinity (SSS), sea surface temperature (SST) and Chl-a data is essential to be used for this water quality purpose and is mostly sourced from remote sensing-based measurements. However, the satellite-derived SSS dataset (~4 to 9 years long) is not as temporally adequate as SST and Chl-a datasets (~3 decades long) thus, preventing a comprehensively spatio-temporal analysis of this water quality aspect. Since SSS can be approximated using satellite-derived ocean color products having the similar temporal length of datasets to the available SST and Chl-a datasets, predicted SSS can be produced from these ocean color products to fill the gap of the existing SSS dataset. Here, we estimated SSS from ocean color products of Aqua-MODIS satellite with a spatial resolution of 4 km by developing an empirical model. Ocean products used in this study were remote sensing reflectances (Rrs) at a range of blue (412, 433, 469 and 488 nm), green (531, 547 and 555 nm) and red wavelength (645, 667 and 678 nm). We also used absorption coefficients due to detritus material non-algae, Gelbstof and CDOM (ADG) at 443 nm and the absorption coefficient due to phytoplankton (APH) at 443 nm. We chose Banda Sea as our area of interest due to its large-scale upwelling system (~300 km x 300 km) providing an important ocean process related to fishery and the availability of in-situ salinity measurements in this location (i.e. CTD casts from series of R/V Baruna Jaya III, VII and VIII cruises and Argo floats), which a part of these datasets will be used to validate our empirical SSS model. As results, we found that ADG-using empirical model (polynomial regression order 5) produces the highest correlation to SSS with R2 of 0.940. The average of all Blue and Green Bands ratio of B/G (R2: 0.903), Blue Chromaticity [B/(B+G+R)] (R2: 0.917) and Green chromaticity [(G/(B+G+R)] (R2: 0.836) also show the similar result, while the rest ocean color products in our empirical model only produce weak correlation to SSS (R2<0.7). Due to its accuracy, ADG was selected as empirical model to estimate SSS. The root mean square error (RMSE) of this model was significantly small (0.130 psu). The predicted SSS using ADG is well validated to Argo floats (RMSE < 0.3 psu) and Aquarius satellite measuring SSS (RMSE: 0.15 psu). The accuracy of ADG-derived salinity at 443 nm indicates that this approach can be used to investigate upwelling process in
Keywords: Sea Surface Salinity, Ocean Color, empirical model, Banda Sea
Topic: Marine Geodesy and Satellite Oceanography