Current Status of the Algorithm Development and Improvement for GOCI-II Ocean Color Products

Abstract

GOCI-II was launched on the 19th of February in 2020 as the successor of Geostationary Ocean Color Imager-I (GOCI-I) onboard the Geo-Kompsat 2B satellite with a 10-year mission lifetime. GOCI-II captures the local area 10 times a day at hourly intervals with 250 m spatial resolution, along with the full-disk imaging over the Asia-Oceania region, selectively once a day. GOCI-II has passed the halfway point of its mission lifetime and produces 26 official L2 products with improved algorithms compared to the initial ones, which were developed using simulated data prior to launch. Here, the current status of the algorithm development and improvement for GOCI-II ocean color products is discussed, including atmospheric correction, chlorophyll-a, colored dissolved organic matter, total suspended matter, diffuse attenuation coefficient, and photosynthetically available radiation. We are also mentioning the GOCI-II data processing system, GOCI-II Ground System (G2GS), and the distribution system for the public data service.

1. Introduction

The ocean surface exhibits diverse colors reflecting the marine environment, and ocean color is the radiance emitted by the ocean, normalized by the irradiance reaching the ocean surface. Products from ocean color remote sensing are the results of interactions (i.e., absorption or scattering) of sunlight with water constituents like phytoplankton, mineral particles, and dissolved organic matter, etc. (Mobley, 1994; Werdell & McClain, 2019). Since the launch of the Coastal Zone Color Scanner (CZCS), Sea-viewing Wide Field-of-view Sensor (SeaWiFS), Moderate-resolution Imaging Spectroradiometer (MODIS), Medium-resolution Imaging Spectrometer (MERIS), Ocean and Land Color Instrument (OLCI), these satellite sensors have been widely used as ocean color remote sensing sensors for monitoring the marine ecosystem (Groom et al., 2019), however, all these satellite sensors were the polar orbit system, thus not able to properly delineate dynamic variation in optically complex coastal waters due to their poor temporal resolution. Ocean-color sensors on a geostationary orbit can overcome these limitations (Choi et al., 2012; International Ocean-Colour Coordinating Group, 2012).

Geostationary Ocean Color Imager-I (GOCI-I), the world’s first geostationary ocean color sensor, was successfully launched in 2010 and has observed the regional area around the Korean peninsula and provided ocean color products, such as chlorophyll-a concentration (Chl-a) and suspended sediment concentration (SSC), with high temporal resolution (Choi et al., 2014b; Wang et al., 2013). GOCI-I was effectively applied to studies of the movement of harmful algal blooms (HAB), sea surface currents, and sediment dynamics driven by tidal currents, using hourly observation data (Choi et al., 2014a; 2014b; Son et al., 2012; Yang et al., 2014). In addition, monitoring long-term scale marine ecosystems, like the habitat area of a certain fish, was also possible (Lee et al., 2023).

The second GOCI (GOCI-II), the successor of GOCI-I, was launched on the 19th of February in 2020 along with the Geostationary Environmental Monitoring Sensor (GEMS), a hyperspectral spectrometer designed for the observation of air pollutants and short-lived climate pollutants, onboard the Geo-Kompsat 2B satellite with a 10-year mission lifetime. The mission of GOCI-I officially ended in March 2021, and GOCI-II is currently in regular operation by the Korea Ocean Satellite Center (KOSC) of the Korea Institute of Ocean Science and Technology (KIOST). The near-real-time Level-1B and Level-2 products produced by KOSC are provided to the public by the National Ocean Satellite Center (NOSC) at the Korea Hydrographic and Oceanographic Agency (KHOA). GOCI-II has moderate spatial (250 m at nadir) and spectral resolution (12 spectral bands from 380 to 865 nm) with unprecedented temporal resolution (10 daytime images per day), and observes the Northeast Asian Sea, including the Korean Peninsula. Since the launch of GOCI-II, various studies for improving the accuracy of the product have been underway through full-scale calibration and validation (Cal/Val) activities.

Lee et al. (2023) demonstrated that GK-2A Advanced Meteorological Imager (AMI) total precipitable water (TPW) data could be applied to realistic water vapor absorption correction of GOCI-II color retrievals for the ocean. They revealed that AMI TPW can improve the accuracy and ability of GOCI-II ocean-color products to capture diurnal variability. Jang et al. (2024) examined the applicability of GOCI-II to the estimation of sea surface salinity (SSS) in the East China Sea (ECS). They addressed the challenge of maintaining continuous SSS estimation in the ECS post-GOCI mission and evaluated the accuracy and limitations of the GOCI-II-derived SSS, proposing future strategies to enhance its effectiveness. Kim et al. (2026) developed a spatially regularized variational optical flow (OF) framework optimized for GOCI-I and GOCI-II to retrieve spatially continuous surface current fields spanning interacting submesoscale-to-mesoscale regimes.

GOCI-II is currently experiencing sensor degradation, which should be calibrated to continuously generate reliable ocean color products along with continuous Cal/Val activities. Here, the current status of algorithm development and improvement for GOCI-II ocean color products is introduced, along with the public service scheme for the products. First, the GOCI-II mission, including user requirements, instrument design, etc., is introduced, followed by discussions of the algorithms for atmospheric correction, chlorophyll-a concentration, colored dissolved organic matter, total suspended matter, diffuse attenuation coefficient, and photosynthetically available radiation. Finally, we will mention the distribution system for the public data service.

2. GOCI Mission - User Requirements and Instrument Design

We significantly strengthened the requirements for GOCI-II to address both the advantages and limitations identified during the operation of GOCI-I. Key user requirements included improving spatial resolution, increasing observation frequency, and expanding the observation domain to full-disk coverage. In addition, enhanced observation performance—particularly in terms of signal-to-noise ratio (SNR), modulation transfer function (MTF), and ghost mitigation—was strongly emphasized. To meet these requirements, we proposed various structural improvements to the telescope system. We determined the final payload specifications through a comprehensive process that balanced user needs with technical feasibility, including constraints on space and system integration. This international collaboration among satellite/sensor developer companies and Korean research institutes, such as Korea Aerospace Research Institute (KARI) and KIOST, not only enabled the successful realization of GOCI-II but also provided a strong technological foundation for the domestically led development of GOCI-III.

The mission of GOCI-II is to ensure continuity of GOCI-I observations while expanding capabilities to monitor the Korean and global oceans, detect ocean disasters, support fisheries, and investigate climate change. Compared to its predecessor, GOCI-II significantly improves observational performance, achieving a spatial resolution of 250 m at nadir (from 500 m), increasing the number of spectral bands from 8 to 12, and enabling full-disk imaging over the Asia-Oceania region. Some of the spectral bands added in GOCI-II improve the accuracy of the atmospheric correction algorithm in highly turbid waters (620 and 709 nm), the Chl-a algorithm by enabling the OC4 algorithm (510 nm), and the red tide algorithm (709 nm). It acquires up to 10 images per day for local areas, allowing hourly monitoring of ocean color and capturing sub-diurnal variability. Additional ultraviolet and wide spectral bands further enhance atmospheric correction and navigation accuracy. To minimize operational interference with the co-mounted payload, GEMS, a coordinated observation strategy was implemented, in which observation time is shared within each hour while maintaining mission efficiency. The automated ground segment (G2GS) supports the processes of Level-2 data from acquisition to distribution.

The GOCI-II instrument design incorporates a compact three-mirror anastigmat (TMA) telescope with a 200 mm pupil diameter and a 7-Mega pixels 2D CMOS detector array. These design improvements result in enhanced SNR, improved MTF, and reduced straylight and image spurious response degradation (ISRD). We implemented additional internal structural refinements to minimize ghost effects and improve radiometric and geometric accuracy. As a result, GOCI-II provides high-quality ocean color products, including chlorophyll-a concentration, suspended sediments, dissolved organic matter, red tide detection, and marine environmental monitoring. Enhanced atmospheric correction algorithms, including spherical models for full-disk observations, and new band-based retrieval algorithms have been developed and validated. Overall, GOCI-II demonstrates more stable and precise performance than GOCI-I and serves as a critical technological foundation for future GOCI-III development. Table 1 summarizes the specifications of GOCI-I, GOCI-II, and GOCI-III.

Table 1. Major characteristics and specifications of GOCI-I, GOCI-II, and GOCI-III

3. Algorithms for Main Products and Services

3.1. Atmospheric Correction

Atmospheric correction is the first and most critical step in ocean color data processing: the water-leaving signal typically accounts for less than 10% of the top-of-atmosphere (TOA) signal in the visible bands, so a 1% uncertainty in the atmospheric path reflectance translates into more than 10% error in the retrieved remote-sensing reflectance (Rrs) (Eplee et al., 2001). The GOCI-II standard atmospheric correction algorithm (Ahn et al., 2021) is based on the two-near-infrared (NIR) band framework of Gordon and Wang (1994) and inherits the overall structure of the GOCI-I atmospheric correction (Ahn et al., 2012; 2015), with targeted modifications that exploit the two additional bands (620 nm and 709 nm) newly introduced on GOCI-II. After correction for gas absorption, Rayleigh scattering, sun glint (Cox & Munk, 1954), and whitecaps, aerosol reflectance in the visible bands is estimated from the two NIR bands (745 and 865 nm) under the black-pixel assumption using the spectral relationships of aerosol multiple-scattering reflectance (SRAMS) scheme of Ahn et al. (2016). By defining the NIR-to-visible extrapolation directly in the multiple-scattering domain, SRAMS avoids the nonlinearity residuals inherent to the single-scattering-based extrapolation used in earlier SeaWiFS-heritage algorithms. Over the turbid coastal waters of the Korean Peninsula and the Yellow Sea, the black-pixel assumption breaks down, and the NIR aerosol and water reflectance must instead be separated iteratively.

In the GOCI-I heritage scheme, this separation was anchored at 660 nm (Ahn et al., 2012), a band that is ill-suited to turbid water because it lies close to the secondary chlorophyll-a absorption peak near 670–680 nm and saturates relatively early as turbidity increases. The 620 and 709 nm bands newly added on GOCI-II overcome these limitations: 709 nm is essentially insensitive to chlorophyll-a absorption and saturates only at higher turbidity, while 620 nm is less affected by chlorophyll-a than 660 nm and by colored dissolved organic matter than 555 nm. Exploiting these bands, GOCI-II employs the spectral relationships of inherent optical properties (SRIOP) model of Ahn and Park (2020), which builds monotonic spectral relationships of the total absorption and backscattering coefficients among the 620, 709, 745, and 865 nm bands; SRIOP remains robust when several sediment types coexist and outperforms both the 660 nm-based GOCI-I scheme and the SeaWiFS-heritage NIR approach of Bailey et al. (2010) under high-turbidity conditions

The retrieved water-surface reflectance is subsequently converted to Rrs through bidirectional reflectance distribution function corrections (Morel et al., 2005; Wang, 2006). Residuals from sensor radiometric calibration and the radiative transfer model are compensated through system vicarious calibration (SVC), implemented following the sensor-independent approach of Franz et al. (2007) as adapted for GOCI-I by Ahn et al. (2015). The initial and current GOCI-II SVC gains were derived using Visible Infrared Imaging Radiometer Suite (VIIRS)-retrieved Rrs over the oligotrophic Philippine Sea as a reference (Ahn et al., 2021). Because of the short operational period at launch, SVC was applied only to the visible bands; NIR-band SVC remains pending and represents one of the principal sources of residual uncertainty in coastal retrievals.

Fig. 1 shows the validation of GOCI-II Rrs against in-situ measurements before and after applying the updated SVC gains. In the blue–green bands, the mean absolute percentage error (MAPE) decreases from 47–111% to 13–47%, with the largest improvements at 490 nm (54% →13%) and 660 nm (74% → 35%); the systematic positive bias of the uncalibrated product (+2–4 × 10^–³ sr^–¹ at the blue bands) is largely removed, leaving a small residual negative bias after calibration. The 412 nm band remains the most challenging, consistent with the known difficulty of short-wavelength atmospheric correction in the presence of absorbing aerosols Further improvement, particularly over turbid case-II waters, is expected once NIR-band SVC and additional in-situ matchups are incorporated through continuous Cal/Val activities.

Fig. 1. Validation of GOCI-II Rrs against in-situ measurements before (blue; uncalibrated) and after (red; calibrated) application of the updated system vicarious calibration gains. Scatter panels compare matchups at 412, 443, 490, 555, and 660 nm; the dashed line is the 1:1 reference, and the solid colored lines are ordinary-least-squares regressions. The number of matchups is N = 26 (uncalibrated) and N = 46 (calibrated); MAPE and mean bias are annotated in each panel. The bottom-right panel summarizes the mean bias ± one standard deviation as a function of wavelength, including the 680 nm band for which only calibrated data were available.

Atmospheric correction is a critical preprocessing step for retrieving accurate Rrs in ocean color remote sensing. Continuous algorithmic refinement is essential to meet increasingly stringent international requirements. The initial GOCI-II gas absorption correction relied on external auxiliary data, such as meteorological forecast fields from the National Centers for Environmental Prediction (NCEP). However, due to limitations in spatiotemporal resolution and accuracy, these forecast model-based products introduced uncertainties into Rrs and subsequent ocean color products (Mélin et al., 2022; Tzortziou et al., 2014).

To overcome these constraints, we developed an advanced gas absorption correction model by integrating high-resolution, satellite-based gas concentration data from GK-2A (AMI) and GK-2B (GEMS) (Lee et al., 2023b). A major enhancement of this new model is the explicit inclusion of nitrogen dioxide (NO₂) correction, which was absent in the original GOCI-II scheme. By leveraging GEMS data, the model accounts for NO₂ spectral absorption, a factor particularly crucial in coastal regions and areas with high anthropogenic activity. Consequently, incorporating these near-real-time observations significantly reduces the uncertainties associated with coarse auxiliary inputs and atmospheric gas variability (Lee et al., 2023a; 2025). The performance of the fused GK-2A/B gas absorption correction model was comprehensively validated against ERA5 reanalysis data. Results demonstrated that GOCI-II products processed with the proposed model exhibit higher consistency with ERA5-based Rrs across all spectral wavelengths. Furthermore, the model achieved a substantial reduction in the daily variability of the Mean Absolute Percentage Difference (MAPD). Notably, it also proved its superior ability to capture localized atmospheric features, such as NO₂ hotspots, which are typically overlooked by lower-resolution models. In conclusion, this multi-sensor fusion approach successfully enhances the accuracy and temporal stability of GOCI-II atmospheric correction, establishing a robust framework for advanced ocean color monitoring.

Fig. 2. Time series of the daily mean absolute percentage difference (MAPD) for remote sensing reflectance (Rrs) at 412 nm (a) and 512 nm (b) over the year 2023. The blue line represents the Rrs processed with the fused GK-2A/B gas absorption correction model (GK-2 Rrs), while the red line indicates the baseline algorithm results (G2GS Rrs). The green dashed line denotes the 5% target accuracy threshold for Rrs.

3.2. Chlorophyll-a Concentration

The Chl-a retrieval algorithm is based on the selective absorption characteristics of phytoplankton, which strongly absorb blue light wavelengths while weakly absorbing green wavelengths. As Chl-a increases in the upper ocean, remote sensing reflectance in the blue bands decreases relative to that in the green band. Therefore, the blue-to-green band ratio is strongly correlated with Chl-a, which is a fundamental parameter for the OC4 algorithm.

Before launch, the original GOCI-II algorithm adopted the standard OC4 algorithm (O’Reilly et al., 1998; 2000), which has been operationally used for the early GOCI-II mission period. However, in optically complex coastal and turbid waters, such as those surrounding the Korean Peninsula, the algorithm needs to be updated for regional waters. After correcting GOCI-II in situ-satellite match-up observations during 2022–2023 (Fig. 3a), the Regional OC4 v1 algorithm was newly developed to reflect the regional bio-optical characteristics of waters surrounding the Korean Peninsula. The Regional OC4 v1 algorithm substantially improved Chl-a retrievals in waters surrounding the Korean Peninsula compared with the standard OC4 algorithm (Fig. 3b vs. Fig. 3c). Error statistics were reduced from MSE = 12.99 and MAPE = 105% to MSE = 0.56 and MAPE = 54%. The Regional OC4 v2 algorithm (Fig. 3d) was derived using 448 expanded GOCI-I match-up datasets, including reprocessed 510 nm remote sensing reflectance data. However, differences in data characteristics between GOCI-I and GOCI-II and several high-chlorophyll, turbid stations increased the error (MSE = 79.55, MAPE = 131%), indicating strong sensitivity to data quality and processing consistency.

Fig. 3. Comparison of each algorithm with in situ observations: (a) Locations of in situ-satellite match-up data observations used for algorithm development and validation, (b–e) Comparison between in situ Chl-a and satellite-derived Chl-a obtained using (b) the standard OC4 algorithm (initial operational algorithm), (c) the new Regional_OC4_v1 algorithm, (d) the new Regional_OC4_v2 algorithm, and (e) the new hybrid Regional_OC4_v1 + CI algorithm. The green line denotes the 1:1 relationship, and the red line indicates the linear regression fit. Statistical metrics, including mean squared error (MSE), mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), are shown in each panel.

The Regional OC4 v1+CI algorithm (Fig. 3e) combined the v1 algorithm with the CI approach to reduce overestimation in low-chlorophyll waters. It showed similar or slightly improved performance compared with v1 (MAE = 0.47, MAPE = 50.86%), particularly improving stability in low-chlorophyll regions. Thus, for the new GOCI-II algorithm, the Regional OC4 v1+CI hybrid algorithm will be applied to provide stable Chl-a retrieval (Fig. 4). Regional OC4 v1 was used in moderate-to-high chlorophyll waters, while the CI algorithm was applied in low-chlorophyll waters to reduce the OC4 overestimation bias (Hu et al., 2012). A transition zone was introduced to ensure a smooth change between the two algorithms and avoid discontinuities in the output. This hybrid approach provides more stable Chl-a estimates across diverse coastal and open-ocean conditions in East Asian waters and is expected to improve both the operational and reprocessed GOCI-II products.

Fig. 4. Updated GOCI-II Chl-a algorithm (Regional_OC4_v1 + CI) combining the regionally optimized Regional_OC4_v1 algorithm and the CI algorithm to provide stable Chl-a retrievals across both low- and high-concentration waters. The left panel shows the relationship between the band ratio and log(Chl-a) for the Regional_OC4_v1 algorithm (red line) compared with the Standard_OC4_algorithm (black line), while the right panel illustrates the characteristics of the CI-based algorithm in low-chlorophyll waters.

3.3. Absorption Coefficient of Colored Dissolved Organic Matter

Colored Dissolved Organic Matter (CDOM) is a fraction of dissolved organic matter that absorbs light within the water column and undergoes photodegradation by solar radiation at the surface, playing an important role in the marine carbon cycle (Park et al., 2021). The absorption spectrum of CDOM is characterized by an exponential decrease as wavelength increases, resulting in strong absorption characteristics within the blue wavelength range. The absorption coefficient of dissolved organic matter (aDOM) is an essential input parameter for estimating dissolved organic carbon concentrations, calculating euphotic depth, and modeling marine primary productivity.

The GOCI-II aDOM algorithm was developed based on the model by Siswanto et al. (2011) to account for the complex optical properties of coastal waters in the Northeast Asian Sea (Korea Ocean Satellite Center, 2021a). While semi-analytical algorithms are primarily optimized for Case-I (open-ocean) waters, the algorithm by Siswanto et al. (2011) was specifically designed for Case-II waters, such as the Yellow Sea and East China Sea, characterized by high concentrations of organic matter and suspended sediments. The empirical model used the Rrs ratio between the blue (443 nm) and green (555 nm) bands. The initial GOCI-II algorithm employed modified coefficients of the original Siswanto model. GOCI-II produces the aDOM at 440 nm, hereafter referred to as G2aDOM. To overcome the limitations of the initial G2aDOM, the algorithm has been refined using an extensive in-situ dataset (Rrs and aDOM) accumulated since the GOCI-II launch. Multi-year data (2010–2024) from diverse regions were collected via research vessels and ocean research stations to incorporate spatiotemporal validations. Out of a total of 644 in-situ samples, 626 were used for coefficient recalibration, while 18 samples were reserved for independent validation.

The validation results demonstrated that the refined algorithm significantly reduced overall error metrics compared to the original version (Fig. 5). The RMSE decreased from 0.0310 to 0.0277 m^–1, and the MAPE improved by approximately 5% (from 63.89% to 58.92%). Furthermore, rRMSE decreased from 51.33% to 45.73%, confirming the enhanced stability of the new algorithm. Notably, accuracy improved in the South China and East China Seas, suggesting that the new G2aDOM algorithm more effectively reflects the region’s complex coastal optics. While the overestimation tendency in the East Sea was slightly mitigated compared to the initial model, a residual bias remains. The residual overestimation in the East Sea may be attributed to its relatively clear, Case-I-like optical properties compared to the turbid coastal waters of the South Sea and the East China Sea. To further enhance accuracy, future research will explore modified functional forms by incorporating additional spectral bands or ratios as input parameters.

Fig. 5. Validation of G2aDOM algorithms against in-situ measurements across different sea areas. (a) Validation results of the initial G2aDOM algorithm based on the model by Siswanto et al. (2011). (b) Validation results of the newly refined G2aDOM algorithm with optimized coefficients using multi-year in-situ measurements.

3.4. Total Suspended Matter

Total suspended matter (TSM) is a key water-quality variable for monitoring turbidity, sediment resuspension, erosion and deposition, river plume dispersion, and changes in the underwater light field. This parameter is especially important in the GOCI-II observation region because the Yellow Sea, Korean coastal waters, and the East China Sea include shallow shelves, tidal flats, river-influenced waters, and areas affected by coastal engineering activities. The initial GOCI-II TSM algorithm is a single-band switching algorithm that does not use a band-ratio approach, but instead applies different empirical relationships using Rrs (620) or Rrs (709) according to the magnitude of Rrs (620), thereby accounting for low-, moderate-, and high-turbidity conditions (Korea Ocean Satellite Center, 2021b). This structure is simple and suitable for operational processing, but it has limitations in very turbid waters because single-band reflectance can become less sensitive or saturated under extremely high suspended-particle concentrations.

To overcome this limitation, the TSM retrieval approach was improved by introducing a new algorithm form that combines a spectral band ratio with the absorption coefficient of suspended particles derived from the quasi-analytical algorithm (QAA) (Min, 2016). This band-ratio and semi-analytical QAA-based approach was intended to represent better the inherent optical properties (IOPs) of suspended particles and to extend the applicability of GOCI-II TSM retrieval from relatively clear waters to highly turbid coastal waters.

Field observations were conducted within the GOCI-II observation area from 2022 to 2025 to evaluate and improve the GOCI-II TSM algorithm. Among 506 field stations, 114 TSM matchup data points were obtained. Recalibration of the existing switching algorithm reduced RMSE by 0.09 g m^–³ and MAPE by 24.94% in the field-based evaluation. By comparison, the aTSM-based multiple nonlinear model, derived using the absorption coefficient of suspended particles estimated from the QAA model, reduced RMSE by 11.10 g m^–³ and MAPE by 43.80%, while showing a more stable distribution across different turbidity ranges. As a result, this model showed improved performance, with a correlation coefficient of 0.9218, absolute percent difference (APD) MAPE of 49.25%, relative percent difference (RPD) of 35.45%, and RMSE of 2.7081 g m^–³. Further in-situ observations and continuous coefficient updates are required to improve the robustness of TSM retrieval across the full turbidity gradient within the GOCI-II observation area.

3.5. Diffuse Attenuation Coefficient

The diffuse attenuation coefficient for downwelling irradiance, Kd, describes the decrease of downwelling light with depth and is a fundamental optical parameter for evaluating water transparency, underwater visibility, euphotic depth, and the light environment for primary production. Because Kd responds to the combined effects of phytoplankton, suspended particles, CDOM, and pure water, it is particularly useful for diagnosing optically complex coastal waters. The current GOCI-II Kd algorithm is based on the semi-analytical model of Lee et al. (2005), which estimates Kd from inherent optical properties, especially absorption and backscattering coefficients, together with solar zenith angle. In the GOCI-II processing chain, these input IOPs are retrieved from Rrs using a QAA-based procedure; therefore, Kd accuracy strongly depends on atmospheric correction, Rrs quality, and the reliability of QAA-derived IOPs. The QAA model used as input for the initial GOCI-II Kd algorithm was originally developed by Lee et al. (2005) for global waters and was therefore not fully optimized for the optically complex waters around Korea. As a result, the retrieval rate of the initial GOCI-II Kd product was very low, with valid Kd pixels accounting for less than 30% of the available observations in most regions. This limitation indicated that improvement of the QAA model was necessary as a prerequisite for enhancing the GOCI-II Kd algorithm.

Therefore, the QAA v6 procedure was modified using field observations collected within the GOCI-II observation area from 2022 to 2025. The improvement focused on empirical coefficient updates in steps 3, 8a, and 8b, modification of the u (λ) relationship for highly turbid waters in step 2, and refinement of the spectral slope parameter Y in step 5. In parallel, the Kd algorithm was also updated using the semi-analytical model of Lee et al. (2013), which more flexibly accounts for ultraviolet and visible wavelengths than the original model. Based on the field dataset, the improved-QAA-only case produced Kd values closer to the observations in the range of 0.03–0.40 m–¹ and reduced the mean bias, RMSE, and MAPE across spectral bands by 0.006, 0.010, and 5.294%, respectively. The combined case, in which both the improved QAA and updated Kd algorithms were applied, also improved RMSE and MAPE by 0.002 and 4.868%, respectively, compared with the original algorithm. As a result, this model showed improved performance, with a correlation coefficient of 0.95, APD (MAPE) of 44.44%, RPD of 40.61%, and RMSE of 0.1290 m^–1. Continued field observations and further updates to the QAA coefficient remain necessary for robust Kd retrieval in diverse coastal waters.

Fig. 6. Validation of GOCI-II-derived TSM against in-situ TSM measurements.

Fig. 7. Validation of GOCI-II-derived Kd (490) against in-situ Kd (490) measurements.

3.6. Photosynthetically Available Radiation

Photosynthetically Available Radiation (PAR) refers to the solar energy flux utilized by chlorophyll-a for photosynthesis, typically defined as the number of photons within the visible spectrum ranging from 400 to 700 nm. Standard ocean color sensors, such as the operational MODIS and VIIRS, as well as the decommissioned Sea WiFS and MERIS, have traditionally provided PAR—a critical input for estimating primary production—as a standard product. In contrast, GOCI-I, the predecessor to GOCI-II, did not offer PAR as an official product, which posed significant challenges for its widespread application. To address this, a dedicated PAR retrieval algorithm for GOCI-II was developed. Grounded in plane-parallel theory, this algorithm maintains a fundamental structure consistent with the MODIS PAR model but has been specifically refined to accommodate the unique geostationary characteristics of the GOCI series (Frouin & McPherson, 2012; Hwang et al., 2022). As detailed by Hwang (2025), the model derives daily PAR by integrating instantaneous PAR (iPAR) values retrieved at each hourly observation slot, thereby effectively leveraging GOCI-II’s high temporal resolution. This process is exemplified in Fig. 8, which illustrates the diurnal progression of iPAR from UTC 00 to 08 and the resulting daily integrated PAR map.

Fig. 8. GOCI-II iPAR and daily PAR images captured on May 10 (unit: Ein/m²/day).

Compared to polar-orbiting sensors, which inherently suffer from accuracy degradation by assuming that a single daily snapshot represents the entire day’s atmospheric condition (Frouin et al., 2003), GOCI-II is uniquely positioned to produce highly accurate daily PAR values. Traditional algorithms often fail to capture non-linear, rapid changes in cloud optical thickness due to their limited temporal sampling. In contrast, GOCI-II overcomes this limitation by physically tracking these high-frequency cloud dynamics hourly, effectively minimizing temporal aliasing errors. This dynamic accounting of atmospheric transmittances allows for more precise primary production estimates in complex regions like the waters surrounding the Korean Peninsula, where meteorological variability is highly pronounced. Validation studies have demonstrated that GOCI-II PAR exhibits superior agreement with in-situ observations compared to both polar-orbiting MODIS and geostationary Advanced Himawari Imager (AHI). Furthermore, recent research has expanded to encompass iPAR in addition to daily integrated values (Hwang et al., 2025). Rigorous Cal/Val processes are currently underway to facilitate the distribution of iPAR as an official standard data product for the GOCI-II mission.

3.7. GOCI-II Ground System

The observation signals from GOCI-II are processed into ocean products immediately upon reception by the ground system via an antenna. This conversion process involves radiometric calibration, which transforms digital counts into physical quantities; geometric correction, which aligns satellite imagery with the Earth’s coordinate system; and atmospheric correction, which isolates pure water-leaving radiance. These procedures are executed by the Data Processing System (DPS), Image Navigation and Registration (INR), and the GOCI-II Ground System (G2GS) established at KIOST (Han et al., 2019; Park et al., 2020).

In particular, the G2GS is the core infrastructure for Level 2 product generation. Algorithms enhanced through Cal/Val activities are implemented into the System Operational Environment (SOE) following trial operations and verification within the Algorithm Development Environment (ADE). Although the update intervals vary by product, the reliability of the products is continuously maintained by performing algorithm updates approximately every 1 to 2 years. Furthermore, to facilitate more systematic quality assessment in consideration of the 10-year design life of GOCI-II, the Radiometric and Geometric Performance Assessment Module (RGPAM) was developed and is currently operated by KIOST (Lee et al., 2023). This system facilitates real-time monitoring of sensor performance and geometric correction accuracy, enabling the prompt identification of transient fluctuations or long-term anomalies in satellite data, which ultimately assists in detecting sensor degradation. Recently, the integration of a quality assessment function for Rrs —a core product of Level 2 data—has significantly enhanced the overall quality management capabilities for GOCI-II data.

3.8. GOCI-II Data Service

The GOCI-II data service has been systematically developed to facilitate efficient access to and utilization of satellite-derived ocean color data by a wide range of users, including researchers, governmental agencies, and industry stakeholders. KOSC operates an integrated data service framework that provides near real-time access to L1B and L2 products generated by the G2GS. These data are transmitted in real time to the NOSC of KHOA and are subsequently distributed to both governmental institutions and the public.

Standard data products, including Rrs, Chl-a, TSM, aDOM, are routinely processed and disseminated through the official NOSC data portal. The short latency between satellite observation and data distribution enables timely monitoring of dynamic ocean phenomena, such as algal blooms, sediment transport, and variations in coastal water quality. To enhance accessibility, the data service platform supports multiple data formats (e.g., NetCDF and image-based formats), as well as both manual download and automated access via application programming interfaces (Open API). In addition, large-volume data access is supported through web directory services and OPeNDAP. Visualization tools are also provided, allowing users to efficiently assess spatial and temporal variability without requiring advanced data processing capabilities. These features significantly reduce barriers to data access and promote broader utilization of GOCI-II products. In addition, NOSC produces and provides L3 and L4 data, such as Sargassum horneri, Sea Surface Temperature, and Sea Surface Current, to enhance convenience and usability for the public and related organizations.

Fig. 9. System overview for GOCI-II data processing and performance monitoring.

Furthermore, value-added services have been developed to support specific scientific and operational needs. These include composite products (e.g., daily and monthly averages), anomaly detection outputs, and regionally optimized products for the Northeast Asian seas. The high temporal resolution of GOCI-II, with hourly observations, enables the generation of diurnal products that provide detailed insights into short-term ocean dynamics, which are difficult to capture using conventional polar-orbiting sensors.

Continuous efforts are being made to improve service quality through user feedback and system enhancements. Updates to processing algorithms, driven by ongoing Cal/Val activities, are promptly implemented in the operational system. In addition, quality flags and uncertainty estimates are provided alongside the data to support robust scientific analysis and informed decision-making. To further promote the use of GOCI-II data, user support services are provided, including technical documentation, algorithm theoretical basis documents (ATBDs), and user training programs. These initiatives contribute to strengthening the user community and expanding the application of geostationary ocean color data in various domains, such as marine environmental monitoring, fisheries management, and climate research.

4. Conclusions

GOCI-II, the successor of the world’s first geostationary ocean color satellite sensor GOCI-I, was successfully launched in 2020 and has been in regular operation since October 2020. It has been improved from GOCI-I in terms of SNR, spatial resolution, spectral resolution, and temporal resolution, reflecting its user requirements. GOCI-II captures the local area 10 times a day at hourly intervals, along with full-disk imaging over the Asia-Oceania region, selectively once a day. Now that it has passed the halfway point of the mission lifetime, it has produced 26 official L2 products of high quality, including the main ocean color products. These have been actively applied to the studies on improving atmospheric correction algorithms, monitoring sea surface salinity, deriving sea surface current fields, etc.

Although GOCI-II is currently experiencing sensor degradation, an algorithm to compensate for it has been developed and improved to ensure the continuous generation of reliable ocean color products. The observation signals from GOCI-II are processed into L2 products immediately upon reception using the G2GS system, which are transmitted in real time to NOSC of KHOA and are subsequently distributed to governmental institutions and the public through the website. We expect the high-quality GOCI-II products to be produced until the end of the mission and to be widely used for marine environmental monitoring, fisheries management, climate research, etc.

List of Acronyms and Abbreviations

##

Author Contributions

Conceptualization, Formal analysis, Funding acquisition: All authors; Supervision: Choi JK; Writing–original draft, Writing–review & editing: All authors.

Conflicts of Interest

No potential conflict of interest relevant to this article was reported.

Funding

This research was supported by the Korea Institute of Marine Science and Technology Promotion (KIMST), funded by the Ministry of Oceans and Fisheries, Republic of Korea (Grant No. RS-2022-KS221660).

Data Availability Statement

None.