Impact of Indian Ocean Dipole Events on Phytoplankton Size Classes Distribution in the Arabian Sea

Changes in the environmental condition associated with climatic events could potentially influence the PSC dynamics of the regional marine ecosystem. The Indian Ocean dipole (IOD) is one of the critical ocean–atmosphere interactions that affects the climate of the Arabian Sea, and it could be a potential factor influencing the regional PSC distribution. However, the relationship between PSC and IOD remains unclear and less explored. In this study, using the in-situ database acquired from the Arabian Sea, we reparametrized the three−component abundance−based phytoplankton size class model and applied it to reconstructed satellite−derived chlorophyll−a concentration to extract the fractional contribution of phytoplankton size classes to chlorophyll−a concentration. Further, we investigated the influence of IOD on the changes in the biological–physical properties in the Arabian Sea. The results showed that the biological–physical processes in the Arabian Sea are interlinked and the changes in the IOD mode control the physical variables like sea surface temperature (SST), sea surface height (SSH), and mixed layer depth (MLD), which influence the specific PSC abundance. Unprecedented changes in the PSC distribution and physical properties were observed during the extreme positive and negative IOD events, which clearly indicated the potential role of IOD in altering the PSC distribution in the Arabian Sea. This study highlights the impact of extreme climate events on PSC distribution and the need for a better understanding of the associated physical–biological–climate interactions.


Introduction
Phytoplankton, the basis of the oceanic food web, is an essential indicator of ocean ecosystem health [1,2]. In many studies, the phytoplankton biomass estimation is associated with the chlorophyll−a abundance (chl−a) (proxy of phytoplankton biomass), which is the primary, ubiquitous pigment commonly present in the phytoplankton [3][4][5][6]. However, oversimplifying phytoplankton as a single group does lead to a lack of detailed biological information about the phytoplankton groups and their unique environmental adaptation. The phytoplankton are diverse in nature and can be classified based on their taxonomy, size, and functional properties [7]. Each phytoplankton community is composed of a specific pigment composition that comprises a unique relationship with chl−a associated with a distinct taxonomic type and size through which the phytoplankton groups can be efficiently categorized [8][9][10]. Specifically, the size structure of phytoplankton has unique environmental adaptations and plays a significant role in the ocean's carbon cycle and marine food web [11]. The cell−size based classification of phytoplankton is widely categorized into micro (>20 µm), nano (2-20 µm), and pico (<2 µm) plankton [12].
Recent developments in abundance−based methods facilitated phytoplankton classification based on their size and functional traits from the total chl−a [13][14][15][16][17][18]. Each cell size has a unique pigment composition and a corresponding specific relationship with chl−a, from which the fractional contribution of each cell size to the total chl−a is retrieved [11,19]. In particular, the fraction of microplankton increases monotonically with the total chl−a, The main objective of this study is to investigate the role of IOD on PSC distribution in the Arabian Sea. The main reason for carrying out this study was to understand the PSC sensitivity towards the changing DMI and the influence of IOD on the physical controls in the Arabian Sea. For this study, the regionally reparametrized three−component model found in Reference [13] was applied on 17−year gap−filled chl−a time−series data to extract PSC in the Arabian Sea. The derived datasets provide the synoptic−scale long−time series of the fractional contribution of each PSC to the total chl−a. Further, the DMI values were correlated with the PSC to identify the relationship between each PSC with DMI. Beside this, each PSC response to positive IOD and negative IOD events was then explored. Corresponding to the interannual variability of PSC, the sea surface temperature (SST), sea surface height (SSH), and mixed layer depth (MLD) products were used as the reference to identify the driving mechanism responsible for the PSC distribution in the Arabian Sea that corresponded to the IOD conditions.

In Situ Database
The database of PSC and chl−a were gathered from field observations and various sources in the Arabian Sea. Both pigment and size−fractioned chl−a in situ datasets were used for this study. This includes the following datasets: Tara Ocean Expedition (TOE)−SEABASS database, SATCORE programme from ESSO−INCOIS, Sagar Sampada cruise (SS2009, SS2010, and SS2011), Sagar Nidhi cruise (SN 128 and SN137), and Sagar Kanya cruise (SN358). The dataset information and sampling locations are shown in Figure 1 and Table 1. All the acquired PSC datasets are bound to the northern and eastern parts of the Arabian Sea due to the piracy issue along the western part of the Arabian Sea. As shown in Table 1, the PSC datasets were obtained from various sources. The datasets used in this study were composed of different measurement methods: HPLC, spectrophotometer, and fluorometer. The Tara Ocean Expedition (TOE) PSC datasets were obtained from the NASA SEABASS/NOMAD database, and the High−Performance Liquid Chromatography (HPLC) was used to extract phytoplankton pigment from the water samples. With the obtained pigment database, the PSC was measured using diagnostic pigment analysis (DPA) adapted from the work by [16] and [17]. For the Sagar Nidhi cruise, Sagar Sampada, and SATCORE database, the total chl−a and size−specific chl−a were collected using the filtration method. The total chl−a and size fractioned chl−a were calculated using a fluorometer for the samples from the Sagar Sampada cruise and a spectrophotometer for the SATCORE database and Sagar Nidhi cruise.  Considering the total chl−a and PSC database obtained from different teams, the quality assurance processes were completed in the following way: Individual pigment and PSC data were visually checked, and the data of low quality (e.g., continuously repeated values, typically low values, values out of limit of measurement) were removed. Only surface samples were used to retune the three−component model, as this study intended to implement the model on satellite observations. Of 188 data points, 156 observations were scrutinized to reparametrize the model, from which ~10% (16) of datasets were retained for validation.

Model Reparameterization
In this study, the three−component model from Reference [13] was reparametrized based on the in situ PSC and total chl−a datasets obtained in the Arabian Sea and further applied to the satellite−observed chl−a to retrieve continuous PSC time−series. The in situ data collected from the Arabian Sea is the primary source for the model reparameterization. The nonlinear least square regression method was used to fit the 140 PSC in situ sample observation against the in situ chl−a to reparametrize the relationship between the PSC and the total chl−a derived from Reference [13]. The reparametrized three−component model estimates the size−specific chl−a (microplankton−Cm, nanoplankton−Cn, and picoplankton−Cp) to the total chl−a. In this model, the total chl−a is the sum of the Cm, Cn, and Cp. The following was therefore obtained: Cp,n = Cp,n max [1 − exp(−Sp,n C)] Cn = Cp,n − Cp (4) Considering the total chl−a and PSC database obtained from different teams, the quality assurance processes were completed in the following way: Individual pigment and PSC data were visually checked, and the data of low quality (e.g., continuously repeated values, typically low values, values out of limit of measurement) were removed. Only surface samples were used to retune the three−component model, as this study intended to implement the model on satellite observations. Of 188 data points, 156 observations were scrutinized to reparametrize the model, from which~10% (16) of datasets were retained for validation.

Model Reparameterization
In this study, the three−component model from Reference [13] was reparametrized based on the in situ PSC and total chl−a datasets obtained in the Arabian Sea and further applied to the satellite−observed chl−a to retrieve continuous PSC time−series. The in situ data collected from the Arabian Sea is the primary source for the model reparameterization. The nonlinear least square regression method was used to fit the 140 PSC in situ sample observation against the in situ chl−a to reparametrize the relationship between the PSC and the total chl−a derived from Reference [13]. The reparametrized three−component model estimates the size−specific chl−a (microplankton−C m , nanoplankton−C n , and picoplankton−C p ) to the total chl−a. In this model, the total chl−a is the sum of the C m , C n, and C p . The following was therefore obtained: C = C m + C n + C p (1) C p,n = C p,n max [1 − exp(−S p,n C)] C n = C p,n − C p (4) and Oceans 2022, 3

484
The total input chl−a is expressed as C [mg m −3 ]; the subscripts m, n, and p refer to micro−, nano−, and pico plankton. The reparametrized C p max = 0.253 and C p,n max = 0.998, are the asymptotic maximum values of in situ C p and C p,n , respectively, and S p is 2.692 and S p,n = 0.987, are the initial slope−corresponding relationship between C p , C p,n and chl−a. The fraction of microplankton (F m ), nanoplankton (F n ), and picoplankton (F p ) to chl−a are obtained by dividing the C m, C n, and C p by chl−a.

Satellite and Reanalysis Data
The present study is conducted based on a chl−a [mg m −3 ] MODIS−Aqua Level−3 products, obtained from the NASA ocean color at daily and four−km resolution from the website (https://oceancolor.gsfc.nasa.gov/, accessed on 13 June 2020). The analyzed series covers the period from 2003 to 2019 for the Arabian Sea (43 • E to 79 • E and 0 • N to 31 • N). The satellite−based data products are generally prone to missing data values due to overcast conditions. The reconstruction of missing values is possible through the Data Interpolation Empirical Orthogonal Function (DINEOF), which was developed by the authors of Reference [46].
The DINEOF method is based on the Empirical Orthogonal Function (EOF), which reconstructs the missing data values with the number of EOF modes on an iterative basis. The DINEOF technique interpolates the missing values in the geophysical data using the singular vector decomposition (SVD) method. Initially, the time series datasets are transformed into the matrix, the values are demeaned, and the missing values are set to zero. Following this, 10% of the original chl−a satellite values were retained for cross−validation purposes. Subsequently, the input data matrix is decomposed through SVD iteratively with EOF modes until RMS converges. The same procedure continues with the number of EOFs modes until the optimum RMS converges. DINEOF is a freely available method that can be obtained from http://modb.oce.ulg.ac.be/mediawiki/index.php/DINEOF (accessed on 7 August 2019). Many studies have incorporated this method and successfully reconstructed oceanographic datasets like chl−a [3,26,47].
For comparison with physical variables other than SST, monthly−mean multi−mission observed sea surface height (SSH) data products and mixed layer depth (MLD) covering the period from 2003 to 2019 were acquired from the Copernicus Marine and Environment Monitoring Service (CMEMS) website (https://marine.copernicus.eu/, accessed on 8 July 2020). In addition, all the physical variables are spatially regridded corresponding to the PSC to efficiently compare the relationships on a pixel−by−pixel basis. The Indian Ocean Dipole Mode Index (DMI) data were downloaded from the NOAA ESRL Physical Sciences Laboratory website (https://psl.noaa.gov/gcos_wgsp/Timeseries/DMI/, accessed on 8 July 2020). The monthly PSC and physical properties anomalies were obtained from the 17−year climatology time series. Further, the changes in PSC and physical factors were investigated for the extreme positive (October 2019) and negative (July 2016) IOD events.

Re−Parameterized Three−Component PSC Model
In this study, in situ phytoplankton measurements for the Arabian Sea from various sources and databases were used to reparametrize the three−component model to improve the retrieval of the regional−based size−specific chl−a in the Arabian Sea. Figure 2 shows the size−specific chl−a (C pn , C p, and C m ) relationship with the total chl−a. The reparametrized model efficiently estimates the relationship between the retrieved PSC and the total chl−a for the Arabian Sea ( Figure 2). Figure 3 indicates the distinct relationship between each PSC and the chl−a. The fractional contribution of the microplankton (F m ) monotonically increases with the increasing chl−a (Figure 3a). The fractional contribution of nanoplankton (F n ) shows a unimodal relationship with the chl−a, whereas the fractional contribution of picoplankton (F p ) monotonically decreases with increasing chl−a. For the testing performance of the reparametrized model for the Arabian Sea, the similar PSC models from References [15,27,28] were compared using the mean absolute error (MAE), Pearson correlation coefficient (r), and bias and root mean square error (RMSE) as evaluation metrics to compare in situ and model values. Table 2 shows the performance of the reparametrized PSC model for the Arabian Sea and similar PSC models using 10% of the datasets kept for validation (16 measurements).
Oceans 2022, 3 485 contribution of picoplankton (Fp) monotonically decreases with increasing chl−a. For the testing performance of the reparametrized model for the Arabian Sea, the similar PSC models from References [15,27,28] were compared using the mean absolute error (MAE), Pearson correlation coefficient (r), and bias and root mean square error (RMSE) as evaluation metrics to compare in situ and model values. Table 2 shows the performance of the reparametrized PSC model for the Arabian Sea and similar PSC models using 10% of the datasets kept for validation (16 measurements).

Validation of Reconstructed Satellite Chl−a and PSC
To test the quality of the reconstructed chl−a for the appropriate input of the PSC model, we assessed it by comparing it with one−third of the in-situ database (i.e., 52) (Figure 4). From this evaluation, it was found that the efficacy of the DINEOF method was better at reconstructing the chl−a, which aided a gap−free chl−a as an input to perform the PSC model. Using the randomly selected (500 points) gap−filled satellite estimates of chl−a contribution of picoplankton (Fp) monotonically decreases with increasing chl−a. For the testing performance of the reparametrized model for the Arabian Sea, the similar PSC models from References [15,27,28] were compared using the mean absolute error (MAE), Pearson correlation coefficient (r), and bias and root mean square error (RMSE) as evaluation metrics to compare in situ and model values. Table 2 shows the performance of the reparametrized PSC model for the Arabian Sea and similar PSC models using 10% of the datasets kept for validation (16 measurements).

Validation of Reconstructed Satellite Chl−a and PSC
To test the quality of the reconstructed chl−a for the appropriate input of the PSC model, we assessed it by comparing it with one−third of the in-situ database (i.e., 52) (Figure 4). From this evaluation, it was found that the efficacy of the DINEOF method was better at reconstructing the chl−a, which aided a gap−free chl−a as an input to perform the PSC model. Using the randomly selected (500 points) gap−filled satellite estimates of chl−a

Validation of Reconstructed Satellite Chl−a and PSC
To test the quality of the reconstructed chl−a for the appropriate input of the PSC model, we assessed it by comparing it with one−third of the in-situ database (i.e., 52) ( Figure 4). From this evaluation, it was found that the efficacy of the DINEOF method was better at reconstructing the chl−a, which aided a gap−free chl−a as an input to perform the PSC model. Using the randomly selected (500 points) gap−filled satellite estimates of chl−a and fractional contribution of PSC (F m , F n, and F p ) obtained from the reparametrized three−component model, the relationships between the chl−a and F m , F n and F p were observed to check whether the satellite estimates achieved the same relationship as in situ PSC and chl−a, as shown in Figure 3. Figure 5a-c shows the relationship between the reconstructed satellite chl−a and satellite retrieved PSC from the reparametrized model for the Arabian Sea.
Oceans 2022, 3 486 and fractional contribution of PSC (Fm, Fn, and Fp) obtained from the reparametrized three−component model, the relationships between the chl−a and Fm, Fn and Fp were observed to check whether the satellite estimates achieved the same relationship as in situ PSC and chl−a, as shown in Figure 3. Figure 5a-c shows the relationship between the reconstructed satellite chl−a and satellite retrieved PSC from the reparametrized model for the Arabian Sea.

PSC and Physical Drivers vs. IOD
The continuous−time series of PSC anomaly datasets were computed by subtracting the PSC monthly time series from the PSC monthly climatology. Further, the PSC anomalies were assessed to examine the influence of IOD on the PSC distribution in the Arabian Sea, including the extreme and negative events of IOD indicated by the Dipole Mode Index (DMI). Figure 6 shows the monthly DMI for the period from 2003 to 2019 (17 years), where the positive value denotes the pIOD condition (positive IOD), and the negative value represents the nIOD condition (negative IOD). To understand the influence of DMI values on PSC distribution in the Arabian Sea, the DMI values were spatially correlated against the PSC anomalies over a 17−year time series. A significant correlation was observed between the specific PSC (micro and pico plankton) and the DMI in the Arabian Sea. Figure 7 shows the relationship between PSC, physical properties (SST, SSH, and MLD), and DMI. The correlation between the PSC versus DMI over the 17−year time series in the Arabian Sea indicates that a large distribution of negative correlation was observed between Fm and DMI.
On the contrary, a large distribution of positive correlation was observed between Fp and DMI. In comparison, no significant correlation was noticed between Fn and DMI in and fractional contribution of PSC (Fm, Fn, and Fp) obtained from the reparametrized three−component model, the relationships between the chl−a and Fm, Fn and Fp were observed to check whether the satellite estimates achieved the same relationship as in situ PSC and chl−a, as shown in Figure 3. Figure 5a-c shows the relationship between the reconstructed satellite chl−a and satellite retrieved PSC from the reparametrized model for the Arabian Sea.

PSC and Physical Drivers vs. IOD
The continuous−time series of PSC anomaly datasets were computed by subtracting the PSC monthly time series from the PSC monthly climatology. Further, the PSC anomalies were assessed to examine the influence of IOD on the PSC distribution in the Arabian Sea, including the extreme and negative events of IOD indicated by the Dipole Mode Index (DMI). Figure 6 shows the monthly DMI for the period from 2003 to 2019 (17 years), where the positive value denotes the pIOD condition (positive IOD), and the negative value represents the nIOD condition (negative IOD). To understand the influence of DMI values on PSC distribution in the Arabian Sea, the DMI values were spatially correlated against the PSC anomalies over a 17−year time series. A significant correlation was observed between the specific PSC (micro and pico plankton) and the DMI in the Arabian Sea. Figure 7 shows the relationship between PSC, physical properties (SST, SSH, and MLD), and DMI. The correlation between the PSC versus DMI over the 17−year time series in the Arabian Sea indicates that a large distribution of negative correlation was observed between Fm and DMI.
On the contrary, a large distribution of positive correlation was observed between Fp and DMI. In comparison, no significant correlation was noticed between Fn and DMI in Figure 5.
(a-c) shows the satellite−based PSC data extracted using a reparametrized three−component model plotted against the reconstructed satellite chl−a.

PSC and Physical Drivers vs. IOD
The continuous−time series of PSC anomaly datasets were computed by subtracting the PSC monthly time series from the PSC monthly climatology. Further, the PSC anomalies were assessed to examine the influence of IOD on the PSC distribution in the Arabian Sea, including the extreme and negative events of IOD indicated by the Dipole Mode Index (DMI). Figure 6 shows the monthly DMI for the period from 2003 to 2019 (17 years), where the positive value denotes the pIOD condition (positive IOD), and the negative value represents the nIOD condition (negative IOD). To understand the influence of DMI values on PSC distribution in the Arabian Sea, the DMI values were spatially correlated against the PSC anomalies over a 17−year time series. A significant correlation was observed between the specific PSC (micro and pico plankton) and the DMI in the Arabian Sea. Figure 7 shows the relationship between PSC, physical properties (SST, SSH, and MLD), and DMI. The correlation between the PSC versus DMI over the 17−year time series in the Arabian Sea indicates that a large distribution of negative correlation was observed between F m and DMI.
the Arabian Sea. Further, the DMI values were correlated against the physical variables to understand the SST and SSH response to DMI. The DMI versus physical variables shows a strong relationship, and a high distribution of positive correlation was observed between SST and SSH versus the DMI. In contrast, a high distribution of negative correlation was noticed between the MLD and DMI in the Arabian Sea.  To identify the influence of physical drivers on the PSC distribution, we further investigated the connection between the PSC and the physical variables like SST, MLD, and SSH in the Arabian Sea region, and the mean time−series anomalies of each PSC (Fm, Fn, and Fp) corresponding to the SST, MLD, and SSH was calculated. The biological and physical factors appeared to be closely correlated in the Arabian Sea. Figure 8 indicates that the inter−annual anomalies of Fp are largely positively correlated with the SST and SSH. This shows an inverse relationship with MLD, which coincided with the environmental adaptation of picoplankton. In contrast, the Fm showed a negative correlation with the SST and SST, and it had a positive relationship with MLD, which also coincided with favorable conditions for micro plankton. No correlation was observed between Fn and physical variables (SST, SSH, and MLD) as nanoplankton are ubiquitous. In Figure 8, the microplankton, picoplankton and physical drivers appeared to be coupled throughout the Arabian Sea. the Arabian Sea. Further, the DMI values were correlated against the physical variables to understand the SST and SSH response to DMI. The DMI versus physical variables shows a strong relationship, and a high distribution of positive correlation was observed between SST and SSH versus the DMI. In contrast, a high distribution of negative correlation was noticed between the MLD and DMI in the Arabian Sea.  To identify the influence of physical drivers on the PSC distribution, we further investigated the connection between the PSC and the physical variables like SST, MLD, and SSH in the Arabian Sea region, and the mean time−series anomalies of each PSC (Fm, Fn, and Fp) corresponding to the SST, MLD, and SSH was calculated. The biological and physical factors appeared to be closely correlated in the Arabian Sea. Figure 8 indicates that the inter−annual anomalies of Fp are largely positively correlated with the SST and SSH. This shows an inverse relationship with MLD, which coincided with the environmental adaptation of picoplankton. In contrast, the Fm showed a negative correlation with the SST and SST, and it had a positive relationship with MLD, which also coincided with favorable conditions for micro plankton. No correlation was observed between Fn and physical variables (SST, SSH, and MLD) as nanoplankton are ubiquitous. In Figure 8, the microplankton, picoplankton and physical drivers appeared to be coupled throughout the Arabian Sea. On the contrary, a large distribution of positive correlation was observed between F p and DMI. In comparison, no significant correlation was noticed between F n and DMI in the Arabian Sea. Further, the DMI values were correlated against the physical variables to understand the SST and SSH response to DMI. The DMI versus physical variables shows a strong relationship, and a high distribution of positive correlation was observed between SST and SSH versus the DMI. In contrast, a high distribution of negative correlation was noticed between the MLD and DMI in the Arabian Sea.
To identify the influence of physical drivers on the PSC distribution, we further investigated the connection between the PSC and the physical variables like SST, MLD, and SSH in the Arabian Sea region, and the mean time−series anomalies of each PSC (F m , F n, and F p ) corresponding to the SST, MLD, and SSH was calculated. The biological and physical factors appeared to be closely correlated in the Arabian Sea. Figure 8 indicates that the inter−annual anomalies of F p are largely positively correlated with the SST and SSH. This shows an inverse relationship with MLD, which coincided with the environmental adaptation of picoplankton. In contrast, the F m showed a negative correlation with the SST and SST, and it had a positive relationship with MLD, which also coincided with favorable conditions for micro plankton. No correlation was observed between F n and physical variables (SST, SSH, and MLD) as nanoplankton are ubiquitous. In Figure 8, the microplankton, picoplankton and physical drivers appeared to be coupled throughout the Arabian Sea.

PSC and Physical Drivers Response to Extreme Positive and Negative IOD Events
The extreme negative (July 2016) and positive (October 2019) DMI periods were taken for the investigation to observe, in detail, how the extreme IOD climate events impact the PSC distribution in the Arabian Sea. Figures 9 and 10 shows the monthly anomaly maps of PSC and the physical variables (SST, SSH, and MLD) distribution in the Arabian Sea during the extreme nIOD and pIOD events. The changes in PSC and SST during the nIOD and pIOD periods were observed efficiently using the reconstructed chl−a input datasets. In July 2016 (nIOD), extreme positive anomalies were observed in SST and SSH over the Arabian Sea (Figure 9). A large corresponding distribution of positive anomalies was observed in Fm. In contrast, a wide distribution of negative anomalies was observed in Fp. When the Fn anomalies were noticed, it was understood that the extreme nIOD event did not show any significant connection with the Fn assemblage. In December 2019 (pIOD), extreme negative anomalies were observed in SST and SSH ( Figure 10). A large distribution of negative anomalies was observed in Fm. On the contrary, an extensive distribution of positive anomalies was observed in Fp anomalies. No significant changes were observed in the Arabian Sea on the Fn anomalies.

PSC and Physical Drivers Response to Extreme Positive and Negative IOD Events
The extreme negative (July 2016) and positive (October 2019) DMI periods were taken for the investigation to observe, in detail, how the extreme IOD climate events impact the PSC distribution in the Arabian Sea. Figures 9 and 10 shows the monthly anomaly maps of PSC and the physical variables (SST, SSH, and MLD) distribution in the Arabian Sea during the extreme nIOD and pIOD events. The changes in PSC and SST during the nIOD and pIOD periods were observed efficiently using the reconstructed chl−a input datasets. In July 2016 (nIOD), extreme positive anomalies were observed in SST and SSH over the Arabian Sea ( Figure 9). A large corresponding distribution of positive anomalies was observed in F m . In contrast, a wide distribution of negative anomalies was observed in F p . When the F n anomalies were noticed, it was understood that the extreme nIOD event did not show any significant connection with the F n assemblage. In December 2019 (pIOD), extreme negative anomalies were observed in SST and SSH ( Figure 10). A large distribution of negative anomalies was observed in F m . On the contrary, an extensive distribution of positive anomalies was observed in F p anomalies. No significant changes were observed in the Arabian Sea on the F n anomalies.

Discussion
The Indian monsoon and the IOD are the primary drivers controlling physical and biological interactions in the Arabian Sea. As the phytoplankton communities are sensitive by nature and have a unique favorable condition, fluctuations in the IOD can highly impact the phytoplankton distribution. The present study investigated the unique relationship between each PSC and IOD in the Arabian Sea using the reparametrized PSC model. Several studies have shown changes in chlorophyll−a concentration (phytoplankton biomass) influenced by the IOD in the Arabian Sea and the Indian Ocean. Brewin et al. [28] showed that, in the Indian Ocean, the phytoplankton community structure was influenced by the IOD and the relationship between the physical-biological interaction using a continuous 10−year time series of satellite observations. Thushra et al. [32] showed that, in the southeastern Arabian Sea, the unprecedented chl−a bloom occurred during the extreme nIOD event in 2016. Sarma et al. [38] showed the impact of strong pIOD events during 1997-1998 in the Arabian Sea with a significant increase in SST and SSH, which led to a corresponding decrease in chl−a distribution.
Our study investigated the relationship between each PSC and the DMI conditions, highlighting the importance of the regional phytoplankton community−wise responses to climate events. Even though the significant influence of IOD on the physical dynamics and its impact on the chl−a distribution in the Arabian Sea have been well documented, the PSC responses are less explored, so we attempted to investigate this in this study. Satellite−derived PSC obtained using the regionally reparametrized PSC model were closer to the PSC in situ values compared to other similar models for the Arabian Sea. The specific constraint on using the satellite datasets was the missing data values due to cloud conditions that were resolved using the gap−filling method (DINEOF) by reconstructing the missing data values. The gap−filled 17−year continuous chl−a enhanced the spatial coverage of PSC distribution, which facilitated further investigation of the unique relationship between each PSC, physical variables, and IOD in the Arabian Sea.
The observed correlations between the PSC, physical variables, and DMI show the distinct responses of each PSC to the changing environmental conditions. The changes in physical drivers, such as SST, SSH, and MLD corresponding to IOD, showed the influence of IOD on the physical processes and the impact on PSC distribution. The results showed evidence that the physical, biological and chemical processes in the Arabian Sea are interlinked (Figures 7 and 8). Our study examined the fundamental theories of biological−physical processes and ecosystem structures. The microplankton tendency to dominate in mixed waters and under high nutrient conditions is related to colder SST, low SSH, and deeper MLD [11,15,19,48]. The picoplankton prefers to live in a stratified and oligotrophic environment associated with warmer SST, high SSH, and shallow MLD [11,19,48]. In contrast, nanoplankton is ubiquitous and capable of surviving in any conditions [11,19,48].
In the Arabian Sea, the summer and winter monsoon are responsible for the changing physical processes and corresponding seasonal distribution of PSC, which have been discussed in detail in Reference [26]. During the summer monsoon period (June-September), the summer upwelling process in the southwestern part of the Arabian Sea (near the Oman coast) creates optimal conditions for the microplankton to dominate in the specific location. In Figure 9, the extreme nIOD anomaly map showed the widespread positive microplankton anomaly, negative picoplankton anomaly, and associated physical conditions throughout the Arabian Sea, which is an unprecedented change that occurred due to the impact of extreme nIOD. In October, the offset of the summer monsoon leads to the decline of microplankton abundance. In Figure 10, the extreme pIOD anomaly map showed an unusual positive anomaly of picoplankton throughout the Arabian Sea, especially in the north part of the Arabian Sea, which has not been reported elsewhere because picoplankton usually dominates in the southernmost part of the Arabian Sea due to its oligotrophic behavior. The impact of extreme pIOD could be the reason for the unprecedented changes in the distribution of PSC in the Arabian Sea. Nanoplankton anomaly maps indicated the ubiquitous behavior of nanoplankton towards the extreme nIOD and pIOD events. From the observation of the instance of extreme IOD events, it is understood that extreme IOD events influenced the physical drivers, led to changes in the physical properties in the Arabian Sea and impacted the PSC structure.
The present study emphasizes that climatic drivers, such as IOD interaction, influences the physical-biological dynamics of the Arabian Sea. This study showed that each PSC has a distinct response to the IOD, which could likely facilitate a definite ecological province and the structure of the biogeochemical cycles in the Arabian Sea. Our approach is restricted to the size−based classification of phytoplankton, which is the first−order approximation of chlorophyll−a. Developing a method to study phytoplankton species may provide a better understanding of how the phytoplankton species in the regional ecosystem respond to the changing climate.

Conclusions
The present study reparametrized the three−component model for the Arabian Sea, which was applied on a 17−year time series of gap−filled chl−a observations of MODIS−Aqua satellite data from 2003 to 2019 to retrieve PSC continuous datasets. The gap−filled spatial coverage of the chl−a time series was achieved using the DINEOF approach. The continuous spatial and temporal fields of PSC for 17−years were efficiently obtained through this approach.
A novel contribution of this study is the capacity to explore the PSC response to the IOD events and physical processes to get a better understanding of the factors controlling the PSC distribution. The results of the observation of the relationship within DMI, microplankton, picoplankton, SST, SSH, and MLD showed that the changes in the distribution of micro and picoplankton are associated with the physical drivers controlled by the IOD. The present study explored the extreme IOD events' impact on the PSC distributions and observed the unprecedented changes in the biological and physical processes. Thus, the overall study delineates the characteristics of PSC and their response to the changing environmental condition influenced by the IOD in the Arabian Sea. Our findings encourage further studies on phytoplankton community responses to regional and global climatic events.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.