Particulate Organic Carbon (POC) Estimation and Algorithm Development in Northeast Arabian Sea Coastal Water

In situ sampling and analysis for water quality and bio-optical parameters carried out in the northeast Arabian Sea, in Diu and Daman coastal waters during February and August 2021. The in situ water samples analysed to estimate chlorophyll, nutrients, Particulate Organic Carbon (POC) in the study area. The bio-optical parameters, normalized water leaving radiance (Lwn) and remote sensing reflectance (Rrs) retrieved with the operation of underwater hyperspectral radiometer (Satlantic/Wetlabs Inc). The objective of the study was to develop POC algorithm in coastal waters of northwest Arabian Sea, utilizing the datasets from sampled Gujarat coastal water stations and the use of underwater hyperspectral radiometer measured remote sensing reflectance (Rrs) data and in situ POC data covering different months and seasons, applying suitably optimal wavelengths-based algorithms and its finalization. In Diu, total 11 stations data collected and produced POC range (44–608 mg m−3). These 11 points POC data correlated with Rrs band ratios using 2nd order polynomial function and indicated coefficient of determination (R2 values 0.37, 0.34 and 0.13). This correlation was also made using 3rd order polynomial fit and indicated R2 values 0.41, 0.42 and 0.26. The same datasets were observed with 3-Dimensional correlation between Rrs(490/555), Rrs(443/555), Rrs(510/555) and POC using linear, paraboloid, gaussian and Lorentzian function fit, which showed improved R2 values ranging 0.34 to 0.66. The data from Daman added to Diu data (total 18 points), which resulted with further better correlation (R2 = 0.75, RMSE = 9.12) with higher range of POC upto 2600 mg/m3. This used multiple linear regression (MLR) between blue and green wavelength channels and band ratios (490, 510, 555, 490/555 and 510/555) derived POC versus the in situ POC data. The Modis-Aqua satellite derived POC data for our in situ stations for the sampling dates ranged 392–411 mg m−3 as compared to our in situ data range 44–608 mg m−3. Out of these different POC algorithms inter-comparison and linkage to satellite data, it observed mandatory to have regional POC algorithms with sufficient in situ data points. The coastal water productivity and its variability study is essential in terms of POC mapping and monitoring on monthly and seasonal time scales in regional scale due to the various fluxes, phytoplankton dynamics and sources from rivers, industries, upwelling and coastal ocean processes.


Introduction
The organic carbon pump-strongly influences the CO 2 uptake of the ocean (Volk and Hoffert 1985).There are indications that the CO 2 uptake of the organic carbon pump responds to climate change (DeVries and Deutsch 2014; Duce et al. 2008;Laufkötter et al. 2017).It has been observed that the surface ocean warming will strengthen stratification and favour shifts of the phytoplankton community structure towards clades that increase CO 2 sequestration by the organic carbon pump (DeVries and Deutsch 2014; Teng et al. 2014;Liu et al. 2015;Hu et al. 2016).On the other hand, enhanced respiration at high temperatures is likely to weaken the organic carbon pump by increasing the respiration of organic matter in the upper water column and lowering the share exported into the deep sea (Iversen and Ploug 2013;Marsay et al. 2015;Rixen et al. 2019).Very high annual mean organic carbon fluxes were recorded by sediment traps in the Arabian Sea during the U.S. Joint Global Ocean Flux Study (JGOFS-India) program during 1994-1995 (Honjo et al. (Honjo et al. 1 3 2008; Lee et al. 1998).The concern of global warming is also the suspect for a putative decrease in productivity of the western Indian Ocean since 1998 because it enhances stratification in the upper ocean (Roxy et al. 2016), but other studies report increasing productivity caused by an intensification of upwelling in this region (Goes et al. 2005).Oceanic physical processes can affect the biological and biogeochemical processes with implications for the temporal and spatial variability of organic carbon both in dissolved and in particulate form (Wozniak et al. 2016;Tran et al. 2019).Hence, the studies on particulate organic carbon (POC) has important linkage to the biological pump.
The POC acts as a linkage between surface primary production, the deep ocean and the sediments.The particulate organic matter is defined as both living and non-living matter of biological origin with a size of ≥ 0.2 µm in diameter.The study is initiated with sea water sampling and data analysis is aimed to develop the particulate organic carbon (POC) algorithm in the northeast Arabian Sea.Synchronous Satlantic underwater hyperspectral radiometer has been operated and data collected (Sarangi et al. 2022).Simultaneously, water samples analysed for water quality and biogeochemical parameters and interpreted over the study area.There are few earlier studies using underwater radiometer-based algorithm development, satellite data validation experiments and interpretation have been carried out in Arabian Sea (Chauhan et al. 2002;Sarangi et al. 2008;Dwivedi et al. 2008;Sahay et al. 2014).The forcing by physical parameters such as sea surface temperature (SST), wind speed, etc. and their influence on biological productivity has been studied in the northern Arabian Sea with few works (Kumar et al. 2001;Chaturvedi 2005;Sarangi et al. 2005;Dwivedi et al. 2008).Hence, the current study to develop POC algorithm using under water radiometer data and its linkage to biogeochemical parameter, POC and inter-comparison with different suitable algorithms and satellite data have been the focus of this study.

Study Area
The water sampling using fishing boat initiated in the northeast Arabian Sea in Diu coastal water in the month of February 2021 in Diu and during August 2021 in Daman coastal water.The water samples analyzed for water quality parameters for 18 coastal water stations.The study area map with station location has been displayed in Fig. 1.

Data and Methodology
In situ measurements play a significant role in model/algorithm development and subsequent validation.As a part of the TDP project (POC estimation), in situ data collection was planned and executed in February and August 2021 around Diu and Daman coastal waters.Field data include optical parameters (radiance and irradiance data), biogeochemical parameter (POC).
During February and August 2021, field trip was conducted and the above parameters measured over the chosen station locations.The sampling location is displayed in Fig. 1.The depth of all station point is around 10-20 m and highly optically complex in terms of chemical and biological components (high sediment load, high chlorophyll biomass, high CDOM).The bio-optical data were measured using underwater hyperspectral profiler (HyperOCR) provided with three sensor components for the measurement of (1) upwelling radiance (L u ), (2) Downwelling Irradiance (E d ), (3) Downwelling irradiance at the surface (E s ).The sun light interacts with the particles in the complex water and in turn absorbed and reflected by the medium.After the necessary connections, the radiometer was deployed in the seawater (Station-1, Fig. 1) and measurements were made for more than three casting and stored as raw file in data logger connected by PC.The same procedure was applied to the remaining 12 stations and simultaneously, water sample was collected for the estimation of water quality parameters and other biological components.Water leaving radiance (L w ) was calculated as per the following equation, L w (0 + , λ) -The radiance leaving the surface after interaction with the particle in the medium (captured by L u sensor, just above the surface water).The ρ(λ,θ), Fresnel reflectance index of seawater and ηw(λ) is refractive index of seawater (1.345) has been considered as per the radiometer data analysis software (Prosoft 7.7 Product Manual 2017) for calculation of water leaving radiance, Lw and this radiometer derived data utilized in many research wok on development of ocean colour algorithms and applications (Chauhan et al. 2002;Sarangi et al. 2008).
The collected L w were normalized by irradiance data at the surface (E s sensor) and termed as Normalized Water Leaving Radiance (L wn ). (1) F o (λ) -the mean extraterrestrial solar irradiance (Neckel and Labs 1984), Es(λ) -downwelling spectral irradiance at surface, L w (λ) -upwelling radiance propagated through the surface.Another important ocean color component is Remote Sensing Reflectance (R rs ) which is defined as the ratio of upwelling spectral radiance to the downwelling spectral irradiance and measured in per steradian.R rs and L wn are crucial parameters in determining the water type and useful in the algorithm development for the estimation of particle concentration (from the chemical/physical/biological/geological sources).
(2) The following statistical approaches have been used to test the algorithm over the northern Arabian Sea water.

Results and Discussion
The Satlantic hyperspectral underwater radiometer measured datasets are stored as raw files, and were processed using Prosoft software (designed for HyperOCR data processing) providing necessary calibration files of the instrument.The processed level 4 (L4 -spectrum of L wn and R rs ) is displayed in Fig. 2. It is clearly observed that the spectrum shown in Fig. 2 is a typical representation of case-2 water with strong reflectance (peak) in green channel and strong absorption in blue channels.The peak radiance (L wn ) is centered around 555 -560 nm and ranges between 1-2.5 (μW cm −2 nm −1 sr −1 ).Blue region absorbs almost the incoming solar radiation and reflects in the range of 0.2-0.5 μW cm −2 nm −1 sr −1 .This strong absorption is manifested as high load of colored dissolved organic matter (CDOM) and chlorophyll biomass.The same trend has been conferred with   (Stramski et al. 2008).Level-3 POC product of MODIS Aqua acquired over Diu coastal waters on 20 February is displayed in Fig. 3 (generated using Stramski et al. 2008 algorithm).POC has been observed high in coastal areas (400-450 mg m −3 ) and low in offshore regions (1-50 mg m −3 ) as per the Fig. 3. Satellite based POC values has been extracted for all stations and furnished in Table 1 against in situ POC data.The spatial distance between sampling stations is approximately confined with the gap of 1-2 km.As the spatial resolution of MODIS Aqua POC is 4 km, same values of POC falls for nearby stations.However, the trend of POC distribution is similar to chlorophyll concentration (Fig. 1), as both the products are derived using blue-green ratio.
Blue-green ratio has been calculated from in situ R rs of all stations and regressed against in situ POC values.This has been performed to investigate the relationship between global Blue-green ratio and POC values in Indian coastal waters.Significant relationship was not observed in linear and fit as observed by Stramski et al. (2008).Though, better coefficient of determination was obtained using polynomial function (R 2 = 0.41 (3rd order), R 2 = 0.37 (2nd order)).In addition, 3D regression analysis has been made using three independent variables (R rs (443)/ R rs (555), R rs (490)/ R rs (555) and R rs (510)/ R rs ( 555)) against in situ POC.The 3D plot was analysed by fitting linear, paraboloid, gaussian and lorentzian curves.Significant correlation obtained using the linear, paraboloid, gaussian and Lorentzian functions.The R 2 ranged 0.34 to 0.66 with standard Error of Estimate (SEE) value range 137.37 to 182.16 (Fig. 5).The detailed regression relations with respective equations and coefficients are furnished in Tables 2, 3 and 4.

New Algorithm Developed by Applying Rrs 490, Rrs 510, Rrs 555 Bands
The highest coefficient of determination (R 2 = 0.42) found with 3rd order polynomial 2D regression between R rs (510/555) versus in situ POC data (Fig. 4 and Table 2).Similarly, the highest coefficient of determination (R 2 = 0.66) found with 3D paraboloid relation between R rs (510/555), R rs (490/555) with in situ POC data (Fig. 5 and Table 3).However, inconsistency and high bias observed between POC values derived from different algorithms.The blue and green bands have been utilized, significant variation was observed between the algorithm derived POC values.This is possibly due to the variation in the optical properties and constituents in the water column in different water types.The new algorithm with the band ratio inputs have improved the relationship with in situ POC concentration and resulted with R 2 value of 0.75 and RMSE value of 9.12 (Table 5, Fig. 6).The data from Daman added to Diu data (total 18 points), which resulted with better correlation with higher range of POC upto 2600 mg/m 3 .Hence, this algorithm would be useful in future with independent validation datasets.This algorithm covered the broad range of POC in the coastal water of northern Arabian Sea.
The present study inferred that, an algorithm validated over a particular region with sufficient data points may produce inaccurate estimates while applying to another regions.This implies the necessity of regional algorithms as alternate solution.As the Indian coastal waters are complex with high dissolved matters, sediment loads and chlorophyll biomass, a regional algorithm is mandatory for the estimation of biogeochemical components.The present study aimed at collection of more data points in near future for the development of algorithm and its validation.

Conclusion
The POC data interpreted and correlated with in situ radiometer derived remote sensing reflectance (R rs ) data covering 18 data points.Different types of equations implying various functions utilized to develop POC algorithm over northwest Arabian Sea covering two different seasons with broad data range up to 2600 mg m −3 .The two band ratio algorithm showed good relationship with POC; which improved with three band based 3-dimensional regression.Finally the 3 bands R rs (490, 510 and 555 nm) based multiple linear regression with 5 variables including 2-bands R rs ratios (490/555 and 510/555) correlated the best with the in situ POC, with generated 5-different coefficients and intercept, produced the best coefficient of determination (R 2 ) of 0.75 with RMSE of 9.12.We have also compared and validated the in situ POC data with Modis-Aqua derived POC, but Modis data POC observed in limited in situ pixels data with range of 392 -411 mg m −3 .From this preliminary observation, it is inferred that there has been good scope to develop and establish suitable POC algorithm in the study area, coastal water of Gujarat and in northern Arabian Sea.The limited in situ data points has indicated on the biophysical and biogeochemical basis between the interaction of in situ remote sensing reflectance and POC distribution.Hence, the attempted correlation and regression has been able to produce satisfactory correlation, which is expected to improve with further more number field observed data points.The Modis satellite data derived POC found to be of narrow range for the sampled stations in our study area.This study will have linkages to understand the local biogeochemistry and in relations it with the phytoplankton biodiversity and nutrients stoichiometry on time series observations over different months and seasons.

Fig. 1
Fig. 1 Study area map.Level 3 MODIS Aqua Chlorophyll product over Diu coastal waters (both Diu and Daman)

Fig. 2
Fig. 2 Spectrum of normalized water leaving radiance (L wn ) and remote sensing reflectance (R rs ) measured by hyperspectral underwater radiometer

Table 1
Station details with Lat/

Table 2
Statistical results obtained from 2D regression between Rrs ratio and POC

Table 3
Statistical results obtained from 3D regression between Rrs ratios and POC f = POC concentration and x and y are mentioned band ratio variables

Table 4
Remote sensing reflectance and algorithm derived POC