POSTER SESSION 3
15:30–17:30, Wednesday, September 16
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
ABSTRACT 1107 | POSTER 187
CURVATURE-RED NORMALIZED DIFFERENCE INDEX FOR PARTICULATE ORGANIC CARBON RETRIEVAL FROM GOCI-I AND GOCI-II IN KOREAN COASTAL WATERS
Satellite retrieval of particulate organic carbon (POC) in coastal waters is challenged by suspended sediment and colored dissolved organic matter, which decouple blue-band reflectance from POC and cause global algorithms to fail. This study presents the Curvature-Red Normalized Difference Index (CRNDI), a three-predictor algorithm for POC retrieval from GOCI-I and GOCI-II data over Korean coastal waters. Band sensitivity analysis across 213 in-situ observations revealed a sign reversal in the Rrs(660)/Rrs(555), POC correlation between Case-2 (r = +0.22) and Case-1 (r = −0.28) waters, confirming that red-wavelength spectral curvature carries particle-load information unavailable to blue-only formulations, motivating a predictor combining spectral curvature at 660 nm, a three-band normalized difference index, and a near-infrared modulation term. Independent validation on 91 withheld matchups yielded R² = 0.605 and MAPE = 24.4% with in-situ radiometry and R² = 0.452 and MAPE = 28.6% with GOCI-I reflectance, outperforming all baselines, including Stramski et al. (2008) and Son et al. (2009). The satellite R² improvement from 0.109 to 0.452 over the locally tuned baseline is the most operationally significant result. Turbidity-stratified validation confirmed reliable performance below 2.0 mg/L (R² = 0.794, MAPE = 22.2%) and failure above 5.7 mg/L, attributed to irreducible optical decoupling rather than algorithmic limitation. Spring performance was best (R² = 0.616), coinciding with phytoplankton bloom dominance, while winter degraded severely. Bootstrap resampling yielded per-pixel uncertainties of 21.3% and 24.9% under in-situ and satellite inputs. CRNDI uses five bands compatible with both GOCI sensors, enabling operational implementation and providing a physically grounded retrieval framework for turbid coastal environments globally.
Md Rakesur Rahman*, Korea Institute of Ocean Science and Technology (KIOST), [email protected], https://orcid.org/0000-0003-0404-5137
Deuk-Jae Hwang, Korea Institute of Ocean Science and Technology, [email protected]
Eunna Jang, Korea Institute of Ocean Science and Technology, [email protected]
Young Baek Son, Korea Institute of Ocean Science and Technology, Jeju Research Center, [email protected]
Taehee Lee, Korea Institute of Ocean Science and Technology, Jeju Research Center, [email protected]
Jong-Kuk Choi, Korea Institute of Ocean Science and Technology, [email protected]
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
