Evaluation of Satellite Sensors to Compute Chlorophyll-a Concentration in the Northeastern Arabian Sea: A Validation Approach

The primary productivity of an aquatic system like the Arabian Sea is majorly supported by the concentration of Chlorophyll-a (Chl-a/Ca) pigment. The present study is narrated to validate the Chl-a data set retrieved from the prominent ocean color sensors (OC3M-MODIS, OC-OCM2, and OC3V-VIIRS) with sea truth data, collected from 204 stations for three years (2015–2017). The in situ concentrations of Chl-a depict the geographic region under the mesotrophic and eutrophic spans with a mean of 1.36 mg m−3 (0.1 > Ca > 1.0 mg m−3). The ratio of CaOCM2/CaIn-situ was 0.97 ± 0.27 mg m−3 (n = 199), but the ratios were higher with CaVIIRS/CaIn-situ is 1.75 ± 0.79 mg m−3 (n = 170) and CaMODIS/CaIn-situ is 2.53 ± 1.42 mg m−3 (n = 158). The coefficient of determination proclaims a significant relationship for MODIS (R2 = 0.36; p < 0.001), followed by OCM2 (R2 = 0.32; p < 0.001) and VIIRS (R2 = 0.19; p < 0.001). The OCM2 showed the lowest RMSE at 0.13, which is relatively lower than the reference error limit by global ocean color missions at 0.35. In overall performance among the three algorithms evaluated for the region, the OCM2 will provide a better estimation of Chl-a with a prediction of 32% accuracy and 34.37% of bias. The log bias values for MODIS (0.35) and VIIRS (0.20) algorithms indicate the overestimation of Chl-a with in situ Chl-a, but the OCM2 algorithm is suitable in the region with a negligible bias (−0.03). The biogeochemical processes and ecosystem characteristics are dynamic from region to region, as yet in its urgent need to validate global sensors to fine-tune the regional algorithms periodically.


Introduction
The major driving force to restrain the earth's climate is the ocean, but the extent of interaction is limited in understanding. The major bottlenecks to understand the biogeochemical process due to limited onboard cruises and other available ocean-based programs. The researchers would get a synoptic view of physicochemical and biological changes with the higher resolutions of spatial and temporal scales from the data collected from satellite-borne ocean color sensors (Werdell et al., 2003). A precise assessment of phytoplankton biomass in the oceans is required to get an insight into the ocean carbon cycle. A large number of in situ data observations are required to validate the satellite products to build up algorithms (Lavigne et al., 2012). With the advancement of satellite technology and the development of precise sensors, viz., MODIS (Moderate Resolution Imaging Spectroradiometer; Esaias et al., 1998), OCTS (Ocean Color and Temperature Sensor, Tanii et al., 1991), MERIS (Medium Resolution Imaging Spectrometer, Rast & Bezy, 1995), and SeaWiFS (Sea-viewing Wide Field-ofview Sensor, McClain et al., 1998a, 1998b, the research investigations have been considered, especially in the coastal and marine research. The ability of remote sensing technology to cover wider spatiotemporal realms has made them the method of choice in marine ecological study. The color of the water of an aquatic system is a function of the concentration of components like organic carbon, other particulate matter, and pigments from phytoplankton (Morel & Prieur, 1977). In situ data as ground the truth can be used to validate and calibrate the satellite products and one such effort was attempted by the NASA Ocean Biology Processing Group (OBPG), in addition, maintained a repository of bio-optical data (Werdell & Bailey, 2005).
In the Earth's Carbon cycle and budget, the oceanic system plays a very prominent role to decode the same need to have a fair understanding of the magnum and primary production rate on a global scale and similarly in diverse and unique sub-systems of the oceans (Prasad & Haedrich, 1994). In aquatic systems, Chlorophyll-a concentration is most often used as a proxy for primary productivity and the difference over a wider spatial and temporal scale becomes important to understanding the carbon budget at a global scale (Platt et al., 1991;Sarmiento et al., 2004;Tomlinson et al., 2004). In the global carbon cycle, the phytoplankton plays a prominent role; it is high time to estimate the spatial and temporal distribution and variability of Chl-a for modern oceanography (Claustre et al., 2010). Although Chl-a is the most abundant biological oceanic measurement, however, the data generated is scarce in comparison with the number of other physical observations studied (e.g., temperature and salinity). Research investigations have been made on the empirical relationship between satellite-derived ocean color and in situ Chl-a in recent years (Clark, 1981;Shang et al., 2014;Tilstone et al., 2013).
The regional field data (ground truths) are limited from single to few field trips is often a limiting factor to validating satellite-derived Chl-a (Barbini et al., 2005). To validate the existing algorithms and newer ones to be developed, the high-quality in situ data over a larger temporal and spatial scale is a must to attain the objective. The Arabian Sea marine ecosystem is unique in more than one way; it is under marked continental influence and has a profound effect on the monsoon, which propels perennial productivity by ensuring vertical mixing of nutrients (Banse & English, 2000). The ground truth sampling is consorting with many limitations like time management, huge cost, sampling errors, disparities in sampling protocols, spatial coverage the paucity of wide-ranging observations, etc. The sensors developed for the ocean color products (SeaWiFS, MODIS Terra, MODIS Aqua, OCM2, and VIIRS) with the associated limitations will provide a synoptic view of the globe. There were very scanty in situ data sets accessible and few validation trials were attempted for available ocean color products before the present study region (Al-Naimi et al., 2017;Raman, 2013).
The key objective of the present research attempt is to generate in situ Chl-a data, which could characterize the regional productivity of the ecosystem and its dynamics. In addition, the study evaluates the performance of ocean color satellite sensors (MODIS, OCM2, and VIIRS) with in situ Chl-a data to suggest a suitable sensor database for large spatial scale application in the northeastern Arabian Sea.

Sampling Design
To generate robust and accurate algorithms, in addition, to validating satellite-derived Chl-a concentration at a regional scale will be limited because of the insufficiency of in situ datasets. The study area comes under the northeastern Arabian Sea and is located on the northwest coast of India's Exclusive Economic Zone (EEZ). The in situ samplings were carried out at two locations, i.e., off Veraval and Okha coasts, Gujarat (Arabian Sea; Fig. 1). Onboard scientific sampling cruises were conducted at monthly temporal intervals for a period of three years (2015 to 2017). The in situ samples were collected randomly from the 204 sampling stations during the study period, but the sample size was less during the monsoon season (June-September) due to the erratic sea conditions. To understand the temporal variations in Chl-a, the data were categorized into four main seasons, i.e., winter monsoon (December-March), spring inter-monsoon (April-May), monsoon (June to September), and fall intermonsoon (October-November) (Raman, 2013).

Estimation of In Situ Chlorophyll-a Concentration
The most practiced conventional approaches to measure Chl-a are fluorometry (Holm-Hansen et al., 1965), spectrophotometry (Lorenzen, 1967), and chromatography (Mantoura & Llewellyn, 1983) with different levels of accuracy and precision. Among three approaches, the most accurate method was high-performance liquid chromatography (HPLC, Hooker, et al., 2009), which provides the concentrations of a large spectrum of phytoplankton accessory pigments in addition to Chl-a. The fluorescence method is the only one that has not been included in the global reanalysis of Chl-a, for example, open ocean climatologies are among the three main approach methodologies that exist for measuring Chla (i.e., water sampling, ocean color, and induced fluorescence), (Gregg & Conkright, 2001). The spectrophotometric approach was adopted to evaluate the performance of the Chl-a value of different satellite sensors because we aimed principally to analyze Chl-a, not for any pigments. The surface water samples (0.5 L for each) were collected at each sampling station. After collection, water samples were protected from sunlight and stored in chilled water to arrest the degradation of Chlorophyll-a pigments on-board. The collected samples were properly labeled on board and kept in the icebox with gel ice in the dark and cold region of the boat. The sample holding time was about four hours before reaching the laboratory for filtration and further analysis. Samples were filtered through 47 mm GF/ F filter paper using a vacuum filter for immediate filtration to avoid time lag. Followed, place the filter paper in a 90% acetone solution; later it was kept in the dark for 24 h in a refrigerator. The extracted material was separated using a centrifuge (R 24 Research Centrifuge, REMI, India) and the absorbance concentration was calculated by using a spectrophotometer at different wavelengths of 750 nm, 665 nm, 645 nm, and 630 nm (Vase et al., 2018). Decant the supernatant into a 10 mm path length cuvette and measure the extinction at different wavelengths against a cell containing 90% acetone. The amount of Chlorophyll-a in the sample was computed using the formula (Strickland & Parsons, 1972): where E is the absorbance at different wavelengths above and corrected by the 750 nm, and C is the amount of Chlorophyll-a estimated.

Satellite/In Situ Matchups
To evaluate the satellite Chlorophyll-a of selected sensors with in situ data, we have collected all images of  (Seegers et al., 2018). The variable range and target error budget was estimated for the OCM2 algorithm at 0.05 to 30 mg m -3 and \ 30%, respectively (Navalgund, 2008). Out of a total in situ data points, the obtained matchups for the satellite products are, i.e., MODIS (n = 158), VIIRS (n = 170), and OCM2 (n = 199) due to the sun glint, temporal mismatch of satellite coverage, cloudy/haze, and monsoonal conditions during the sampling period. The 8-day composite L3 Chl-a data product with a high resolution of * 4 km 2 /pixel was retrieved to evaluate the variability in the sensors dataset in the region. A stringent comparison between satellite and in situ data should limit the time difference between the two measurements to within ± 2-3 h. The in situ Chl-a datasets were a matchup in time (8-day temporal composite matchup) and spatial (latitude and longitude) with the satellite-derived dataset (Al-Naimi et al., 2017). The 8-day composite products were preferred for the region instead of daily satellite data, while the temporal threshold may not decrease the accuracy (Johnson et al., 2013).

Statistical Analysis
The in situ and satellite Chl-a concentrations were compared to understand the disparities/similarities in variability, correlation, accuracy, and error. The performance indicators like a satellite to in situ ratios, semi-interquartile range (SIQR), median satellite to in situ ratio, and median absolute percentage difference (MPD) were estimated to understand the overestimation/underestimation of satellite Chl-a. The indication of the spread of the data and measurement of uncertainty is explained by SIQR in addition to how accurately the satellite Chl-a agrees with in situ measurements.
where Q 1 is the 25th percentile, and Q 3 is the 75th percentile (Q 2 would be the median value). The median ratio indicates an overall bias. The MPD (median absolute percentage difference) is calculated as the median of the individual absolute percentage differences (Bailey & Werdell, 2006): where X is the satellite value, and Y is the in situ value.
The root mean square (RMS) and the mean difference (bias) in percentage are used to assess the differences between the in situ and satellite-derived data (Marrari et al., 2006): where S is the satellite data, I is in situ data, and n is the number of data pairs. For a normally distributed x, RMS should equal the standard deviation. Further, because the natural distribution of Chl-a is lognormal (Campbell, 1995), error estimates were also made on the logarithmically transformed (base 10) data: Generally, error estimates cannot be expressed as percentages because they are transformed logarithmically. These error estimates have been used in recent publications to validate SeaWiFS global and regional estimates of Chla (Zhang et al., 2006) and to describe the performance of the ocean color algorithms (O'Reilly et al., 2000).

Results
The significant temporal variations (monthly) were shown in Fig. 2, illustrating the maximum during the winter monsoon (1.52 ± 0.48 mg m -3 ) and minimum during the monsoon season (1.25 ± 0.50 mg m -3 ) (p \ 0.05). The representation of log_Chl-a data in the histogram plot depicts the study region was categorized under the mesotrophic and eutrophic categories (Chl-a concentration of [ 0.1 mg m -3 ) (Fig. 3). The median of Chl-a value was highest for the MODIS at 2.92 mg m -3 (ranged from 0.28 to 18.75 mg m -3 ), followed by VIIRS as 2.43 mg m -3 (ranged from 0.36 to 11.82 mg m -3 ), OCM2 as 1.35 mg m -3 (ranged from 0.57 to 2.38 mg m -3 ) and in situ as 1.36 mg m -3 (ranged from 0.50-2.96 mg m -3 ) (Table 1; Fig. 4).
The mean ratio of C a OCM2 /C a In-situ is closer to 1 with a narrow deviation (i.e., 0.97 ± 0.27 mg m -3 ), in contrast to the higher ratio for C a VIIRS /C a In-situ (1.75 ± 0.79 mg m -3 ) and C a MODIS /C a In-situ (2.53 ± 1.42 mg m -3 ), which depicts the overestimation of Chl-a for the later satellite dataset (Fig. 5). The correlation analysis between the in situ and satellite Chl-a data set revealed a better correlation with MODIS (0.602, p \ 0.001), followed by VIIRS (0.589, p \ 0.001), and OCM2 (0.562, p \ 0.001) (Fig. 6). The OCM2 satellite data product is having less   The plot indicates the ratio of Chl-a between in situ and satellite data products significantly lower for OCM2 than in other datasets (MODIS and VIIRS). The slope, coefficient of determination (R 2 ), and root mean square of the regression (RMSE) are listed in Table 2 for the scatter plots of satellite Chl-a data with in situ data (Fig. 7). The high R 2 and slopes close to unity suggest that the validation comparisons are in agreement with the measured dynamic range. The slope values are significantly deviating from the unity in the case of three regression equations, i.e., MODIS (3.73), VIIRS (1.25), and OCM2 (0.40). The highest slope value was noticed for the MODIS algorithm due to the widespread data (0.28 to 18.75 mg m -3 ). The R 2 values for the three regression equations are not so strong but show significant variability (p \ 0.001), out of which the highest value was observed for MODIS (R 2 = 0.36), followed by OCM2 (R 2 = 0.32) and VIIRS (R 2 = 0.19) ( Table 2 and Fig. 7). Out of the three sensors studied, the MODIS can explain the higher 36.0% variability of in situ Chl-a, followed by OCM2 (32.0%) and VIIRS (19.0%).
The scattered plots symbolize the wide distribution of regression points around the 1:1 line. The expression of RMSE for the log-transformed data of three different sensors concerning in situ values depicted the overall uncertainty and observed highest for MODIS (0.411), VIIRS (0.284), and lower for OCM2 (0.126) ( Table 2).

Discussion
The satellite-derived maps provide a unique temporal and spatial picture of the Chl-a at a global scale (McClain et al., 1998a(McClain et al., , 1998b. The OBPG has developed methodologies to validate many historical and operational ocean color missions and the datasets, in such cases a high-quality data set covering a multi-year period is a must and the results can be used for global climate studies (Bailey & Werdell, 2006). The global accuracy in the case of clear and natural waters of modern sensors for the Chlorophyll-a was defined in the range of 5% to 35%, (Hooker et al., 1992). Marina et al. (2006) have reported that the surface  Chlorophyll-a concentration (C a , mg m -3 ) in the Southern Ocean was significantly lower than in situ water samples and also revealed the percentage overestimation of in situ data would increase with the decrease in concentrations. However, satellite observations are limited to the ocean surface, and the error was calculated by matchup analysis of concurrent satellite and in situ measurements, which were evaluated to vary around ± 35% in the open ocean (Moore et al., 2009). Antoine et al. (1996) classify the geographic diversity into three trophic regimes based on the Chl-a concentration, i.e., oligotrophic (C a-B 0.1 mg m -3 ), mesotrophic (0.1 \ C a B 1.0 mg m -3 ) and eutrophic (C a [ 1.0 mg m -3 ). The in situ Chl-a measurements represented the northeastern Arabian Sea falls under the two spans, i.e., mesotrophic and eutrophic (0.1 [ C a [ 1.0 mg m -3 ), which explains the region with better productivity). Previous research investigations stated that the northeastern part of the Arabian Sea is one of the highly productive zones and sustains above-average biological productivity (Chauhan et al., 2001;de Sousa et al., 1996;Dwivedi et al., 2006;Shanmugam, 2011;Smitha et al., 2022). The present study region is located amid two gulfs, i.e., the Gulf of Kutch and the Gulf of Kambath, which supports better primary productivity because of the vertical mixing of nutrients. The possible reasons for the high productivity due to the coastal upwelling by monsoonal winds (Swallow, 1984) and convective mixing of dense north waters by northeast monsoons (Kumar & Prasad, 1996).
In the study region, the range of Chl-a concentration was, i.e., in situ (0.50 to 2.96 mg m -3 ), MODIS (0.28 to 18.75 mg m -3 ), VIIRS (0.36 to 11.82 mg m -3 ) and OCM2 (0.57 to 2.38 mg m -3 ), which are in similar ranges with the previous studies. The Chl-a data were acquired from encompassing, which is ranging from 0.012 to 72.12 mg m -3 (Werdell & Bailey, 2005). The OC2v2 algorithm was used in the marginal west sea of the Antarctic Peninsula, for which the SeaWiFS imageries underestimate Chlorophyll-a (0.7-43 mg m -3 ) concentrations by 60% (Dierssen & Smith, 2000). Along the South Georgia region (54.5 0 S, 37.0 0 W) underestimation of 87% was found in the lower Chl-a concentration range (0-1.0 mg m -3 ), but a 30% underestimation was recorded for the Chl-a concentration value of [ 5 mg m -3 (Kwok & Comiso, 2002). The Chl-a concentration ranges from 0.5 to 7.1 mg m -3 , with an average value of 2 mg m -3 in the Godavari estuarine waters of India (Latha et al., 2013). The Chl-a concentration of satellite data with in situ values, the OCM2, had a similar range of Chl-a concentration in the region.
The permissible errors in the log-transformed C a data during the algorithm development were about 0.2 or less, and nearly 60% root mean square (RMS) relative error (O'Reilly et al., 2000). In the global validation of C a , the error for most of the ocean basins is about 0.3 (Gregg and Casey 2004), but in the case of the Southern Ocean, the reported errors are significantly larger. According to Raman (2013), the RMSE for the log-transformed data of satellite-derived versus in situ Chl-a value was higher for OC2 (23.50%), OC4 (20.90%), and OC-OCM (16.10%). The present results, i.e., log bias and log RMSE percentage, are coherent with the previous studies for different algorithms MODIS (log bias is 0.348 and log RMSE is 41.10%); VIIRS (log bias is 0.200 and log RMSE is 28.40%) and OCM2 (log bias is -0.031 and log RMSE is 12.60%). The OCM2 sensor performed better for the estimation of Chl-a in the region with lower log RMSE and log bias values than the other two sensors, i.e., MODIS and VIIRS. A comparative analysis of Chl-a during bloom and non-bloom periods of phytoplankton was studied along Ardley Cove, King George Island, and observed as maximum Chlorophyll-a reached 9.87 mg m -3 during the bloom period. But, the average Chlorophyll-a concentration observed was less than 2 mg m -3 from 1991 to 2009. In the case of high latitude and high Chlorophyll-a regions, the fluorescence approach ensures a better correlation at 59.46% (Chen et al., 2017). The satellite measurements still have large errors in estimating phytoplankton biomass (Marrari et al., 2006) and the global Chlorophyll-a satellite algorithm typically underestimates Chlorophyll-a in the Southern Ocean (Friedrichs et al., 2009;Gregg & Casey, 2004).
In many cases, it remains obvious to underestimate ([ 2.0 mg m -3 Chlorophyll-a) and overestimates (\ 2.0 mg m -3 Chlorophyll-a), but the RMS was surpassed by 60% for both OC3M and OC3V algorithms. The Chlorophyll-a data from MODIS and VIIRS sensors showed a weak correlation with in situ Chl-a. However, the VIIRS sensor still performs better than MODIS in the estimation of Chl-a with a better regression coefficient. The little differences in band ratio were estimated by using a sensitivity test while declaring the selection of both MODIS and VIIRS (Chen et al., 2016). The RMSE percentage measures the similarity/dissimilarity between the in situ and satellite-derived Chl-a value, the OCM2 had lower RMSE (49.01%) than the other algorithms, i.e., VIIRS (212.72%) and MODIS (427.81%). Various empirical algorithms have been developed successfully using regional data sets by considering the local geophysical conditions (Darecki and Stramski 2004;Garcia et al., 2005;Gohin et al., 2002). The current research attempt is to strengthen the in situ database in support of the above objective to develop practical algorithms.
The regression analysis showed significant variability in comparison with in situ data, i.e., MODIS (R 2 = 0.36 and slope = 3.73), followed by OCM2 (R 2 = 0.32 and slope = 0.40) and VIIRS (R 2 = 0.19 and slope = 1.25). Werdell et al. (2003), attempted the validation trials for the Sea-WiFS global data set with in situ measurements and recorded error and regression parameters as a percentage difference (27.3%), RMSE (0.567), slope ± SE (1.034 ± 0.025), and R 2 (0.849). Chen et al. (2017) reported the OC3 global band ratio algorithm Chlorophylla estimation showed a nonlinear relationship with in situ values with an 11.43% correlation (y = -0.1094x ? 1.9846; R 2 = 0.1143) along the Southern Ocean. The algorithms which estimate Chlorophyll-a range from 0.1 to 10 mg m -3 , have a similar range with in situ Chlorophyll-a. In the Antarctica Peninsula region, the OC3 algorithm coefficients showed better performance (r 2 = 0.64) (Gabric et al., 2005). In the present study, the variability between the satellite sensor and in situ data sets is meliorate explicated. In contrast, the MODIS data set explains the highest of 36.0% of the variability of in situ Chl-a, followed by OCM2 (32.0%) and VIIRS (19.0%). The coefficient of determination of R 2 of satellite estimated versus in situ dataset was highest for OC-OCM (R 2 = 0.76; slope = 0.89), followed by OC4 (R 2 = 0.59; slope = 0.56) and OC2 (R 2 = 0.45; slope = 0.43) (Raman, 2013). The validation trials of in situ Chlorophyll-a and OCM-2 (OC2 and OC4V4 algorithms) showed a significant regression relationship (R 2 = 0.60) along the estuarine waters of the Bay of Bengal and also revealed overestimation at low Chla concentration and underestimation at high Chl-a concentration (Latha et al., 2013). The coefficient of determination computed for the OCM2 in the current study was lower in comparison with the previous studies (Latha et al., 2013;Raman, 2013).
The better MAPD was observed with the OCM2 sensor (35.01%), followed by VIIRS (142.91%) and MODIS (192.39%). The MAPD for OC2, OC4, and OC-OCM datasets was observed at 47.17%, 30.87%, and 22.34% with in situ Chl-a concentration (Raman, 2013), estimated MAPD of the OCM2 sensor in the present study is similar to previous results. The global ocean color missions aimed to set the RMSE error at 0.35 (IOCCG, 2000), and the OCM2 showed the lowest RMSE error 0.126, which is comparatively lower than the reference error limit. The RMSE computed for VIIRS and MODIS is 0.284 and 0.411, respectively. However, the OCM2 algorithm has a lower log bias (-0.03) than the other two algorithms, i.e., VIIRS (0.20) and MODIS (0.35). The OCM2 product has negligible bias and better performance than the other two, i.e., VIIRS and MODIS.
In general, most of the widely used global algorithms provide acceptable results for the estimation of Chlorophyll-a, when developed using a cohesive global data set (e.g., Maritorena et al., 2002). Research gaps existed in the estimation of Chl-a using satellite sensors and periodic validation of remotely sensed products in comparison with in situ values. The accuracy performance of different sensors studied in the region, the OCM2 showed better overall statistical indicators, i.e., MPD, RMSE, Bias, log bias, and log RMSE and regression coefficients than the other sensors (MODIS and VIIRS). The global sensors (MODIS and VIIRS) overestimate in situ Chl-a concentration (mg m -3 ), while the OCM2 algorithm is in tune with the in situ values in the northeastern Arabian Sea. Overall, the satellite sensors (MODIS and VIIRS) systematically overestimated Chl-a concentration in comparison with in situ data, mainly due to the significant influence of sediments, CDOM, and land discharges. The validation exercise in the region suggests better accuracy indicators for the OCM2 algorithm and is more suitable for the region's bio-optical characteristics.

Conclusions
The Northern Arabian Sea is considered the most productive region in the globe and with unique bio-geochemical and oceanographical coupling processes. Because of the region's productivity identity and scarcity of in situ data, the current study was initiated to generate in situ Chla datasets and also to evaluate the performance of ocean color satellite sensors in the region. In obvious, the validation process of the satellite datasets is coupled with many limitations, i.e., atmospheric correction due to cloud coverage and sun glint, lack of large in situ dataset, the disparity in the in situ data analysis methodologies, coverage of satellites in spatial and temporal scales and frequency of matchup points, etc. Among three satellite sensors (MODIS, VIIRS, and OCM2) evaluated in the region, the OCM2 sensor dataset performed better than the others (the ratio is 0.97 ± 0.27 mg m -3 ; bias is 34.37%; RMSE is 49.01%; slope 0.40 and R 2 is 0.32 (p \ 0.001)). In relation to in situ data, the MODIS (2.53 mg m -3 ) and VIIRS (1.75 mg m -3 ) sensors indicate the overestimation of Chla concentration in the region, while the OCM2 (0.97 mg m -3 ) sensor match with the ground truth data. The error bias in the case of the OCM2 sensor is comparatively lower than the global error accuracy limit (35%) and is suggested to use in the region. However, large in situ datasets are required to calibrate regional-specific algorithms to estimate Chl-a more accurately in the northeastern Arabian Sea. In consideration of the field-level practical difficulties, the in situ sampling protocols (Standard Operating Procedures, SOPs) need to be formulated as per the international standards (JGOFS) among the diverse organizations engaged in the field of validation experiments.