Reviews and syntheses: Trends in primary production in the Bay of Bengal – is it at a tipping point?

Ocean primary production is the basis of the marine food web, sustaining life in the ocean via photosynthesis, and removing carbon dioxide from the atmosphere. Recently, a small but significant decrease of global marine primary 10 production has been reported based on ocean color data, which was mostly ascribed to decreases in primary production in the northern Indian Ocean, particularly in the Bay of Bengal. Available reports on primary production from the Bay of Bengal (BoB) are limited, and due their spatial and temporal variability difficult to interpret. Primary production in the BoB has historically been described to be driven by diatom and chlorophyte clades, while only more recent datasets also show an abundance of smaller, visually difficult to detect 15 cyanobacterial primary producers. The different character of the available datasets, i.e. direct counts, metagenomic and biogeochemical data, and satellite-based ocean color observations, make it difficult to derive a consistent pattern. However, making use of the most highly resolved dataset based on satellite imaging a shift in community composition of primary producers is visible in the BoB over the last two decades. This shift is driven by a decrease in chlorophyte abundance, and a coinciding increase in cyanobacterial abundance, despite stable concentrations of total chlorophyll. A similar but somewhat 20 weaker trend is visible in the Arabian Sea, where satellite imaging points towards decreasing abundances of chlorophytes in the North and increasing abundances of cyanobacteria in the eastern parts. Statistical analysis indicated a correlation of this community change in the BoB to decreasing nitrate concentrations, which may provide an explanation for both, the decrease of eukaryotic nitrate-dependent primary producers and the increase of small unicellular cyanobacteria related to Prochlorococcus, which have a comparably higher affinity to nitrate. Changes in community composition of primary 25 producers and an overall decrease of system productivity would strongly impact oxygen concentrations of the BoB’s low oxygen intermediate waters. Assuming decreasing nitrate concentrations and concurrent decreasing biomass production, export and respiration, oxygen concentrations within the oxygen minimum zone would not be expected to further decrease, an effect which would be enhanced by stronger stratification as a result of future warming. Therefore, given a decrease in primary production, the BoB may not be at a tipping point for becoming anoxic, unless external nutrient inputs increase. 30 1 The role of the Bay of Bengal for primary production in the global Ocean – a historical perspective Marine primary producers contribute around 50% to global net primary production (Behrenfeld et al. 2001), leading to a carbon flux from the atmosphere into the ocean of 4550 Tg C per year (Longhurst et al. 1995). Changes in ocean primary https://doi.org/10.5194/bg-2020-494 Preprint. Discussion started: 26 January 2021 c © Author(s) 2021. CC BY 4.0 License.

production exert an important control on atmospheric carbon dioxide (CO2) concentrations, and thus on global climate (Falkowski, Barber, and Smetacek 1998). The BoB has often been described as an area of low primary production compared 35 to the Arabian Sea. This low productivity has classically been ascribed to a strongly stratified water column as a result of increased surface water temperatures (Kumar et al. 2004) in combination with lowered surface water salinity due to monsoon-governed episodes of massive rainfall and river discharge with maximum freshwater inputs in September (e.g., Mahadevan (2016)). The stratification extends through large parts of the BoB basin (Subramanian 1993), restricting nutrient fluxes to the surface and eventually limiting primary production. In coastal areas, nutrient inputs from the major rivers have 40 been described to stimulate primary production, however, rapid consumption as well as a ballasting effect with lithogenic particles and subsequent sedimentation of organic matter prevent offshore transport (Singh, Gandhi, and Ramesh 2012;Singh and Ramesh 2011;Krishna et al. 2016;Kumar et al. 2004;Ittekkot 1993). Open waters therefore appear low in macronutrients, exhibiting at least temporarily a slight nitrogen undersaturation (Bristow et al. 2017;Löscher et al. 2020).
However, nitrogen fixation has been described low to non-existent (Saxena et al. 2020;Löscher et al. 2020), therefore not 45 compensating the nitrogen deficit. The available geological record suggests that nitrogen turnover is generally absent since the last glacial maximum where isotope records lastly were enriched for delta 15 N, indicative for nitrogen fixation (Contreras-Rosales et al. 2016;Shetye et al. 2014), as reflected by low primary production suggested from deep time records of total organic carbon (TOC, Fig. 1A) on a geological time scale.
While a decrease in primary production has been derived in models for the last decades ( Fig. 1B; (Gregg and Rousseaux 50 2019;Roxy et al. 2016)) in the Indian ocean, shorter historical records of primary production in the BoB are not too abundant. However, records of direct rate measurements go back to the RV Galathea and RV Anton Bruun expeditions in the early 1950ies, followed by the International Indian Ocean Expedition (IIOE) from 1959 to 1965 (Snider 1961). Those earliest records report primary production rates of 0.1 -2.16 mg C m -2 d -1 for the shelf regions and 0.1 -0.3 mg C m -2 d -1 for open ocean waters of the BoB. Comparably higher rates were reported from an expedition with the Russian RV Vityaz from 55 1956 to 1960, with rates between 70 to 3600 mg C m -2 d -1 , from a record from 1970 with a rate of 190 mg C m -2 d -1 (Nair et al. 1973), and from a summer monsoon situation in 1976 with rates between 130 and 330 mg C m -2 d -1 (Radhakrishna et al. 1978). Some of those earlier measurements were, however, likely biased as a result of trace metal contamination before trace metal clean techniques were available, a problem identified by calculating primary production to chlorophyll ratios, which turned out to be extremely high (250-2500 compared to an average of 23 ± 13 in later data presented in Table 1; 60 Madhupratap et al. (2003)). Later reports show a high variability of primary production rates ranging between 0.3 and 936 mg C m -2 d -1 (Gomes, Goes, and Saino 2000;Murty et al. 2000;Balachandran et al. 2008;Madhupratap et al. 2003;Gauns et al. 2005;Kumar et al. 2010;Mohanty, Pramanik, and Dash 2014;Subha Anand et al. 2017;Löscher et al. 2020), and extremes of 2200 mg C m -2 d -1 (Bhattathiri, Devassy, and Radhakrishna 1980), with generally higher rates in shelf regions as compared to the open ocean, which were combined into average rates of 500 and 300 mg C m -2 d -1 for shelf and open ocean, 65 respectively, to obtain a carbon flux budget (Naqvi et al. 2010). These average rates are quantitatively comparable to the studies presented in Table1, however, for instance mesoscale water mass dynamics have been observed to promote primary https://doi.org/10.5194/bg-2020-494 Preprint. Discussion started: 26 January 2021 c Author(s) 2021. CC BY 4.0 License. production in the BoB beyond those ranges up to 920 mg C m -2 d -1 , likely because of eddy-related decreases in stratification and pumping of nutrients into otherwise nutrient-exhausted photic surface waters (Sarma and Udaya Bhaskar 2018).
Additional variation results from the strong influence of the three monsoon governed seasons on primary production rates 70 ( (Gomes, Goes, and Saino 2000;Jyothibabu et al. 2018;Madhu et al. 2002;Gauns et al. 2005); Table 1). Based on the presented data, a current estimate of primary production rates would be in the range of 361 ±145 and 236 ± 121 mg C m -2 d -1 for coastal and open ocean regions, respectively, which is one to two orders of magnitude below the Arabian Sea, depending on the region and time of the year (Naqvi et al. 2010).

Key primary producers in BoB waters 75
Compared to records of primary production rates, even less data on the primary producer community are available, and chlorophyll concentrations are often the only parameter presented (Table 1). Typically, coastal chlorophyll concentrations are about an order of magnitude higher compared to those in the central BoB (e.g. (Radhakrishna et al. 1978;Balachandran et al. 2008;Gauns et al. 2005;Kumar et al. 2010  Historically available phytoplankton diversity records have methodological limitations relying mostly on direct or microscopic phytoplankton counts; therefore, small sized phytoplankton and cyanobacteria are likely not correctly represented. There is, however, a general consensus in earlier and newer studies that diatoms dominate the pool of primary producers (Gauns et al. 2005;Madhupratap et al. 2003;Devassy, Bhattathiri, and Radhakrishna 1983), with some historical records being astonishingly detailed, presenting phytoplankton distribution down to the genus and species levels (Nair and 90 Gopinathan 1983) and their results are comparable to more recent studies  showing a diversity of diatoms including Thalassiothrix, Nitzschia, Thalassionema, Skeletonema, Chaetoceros and Coscinodiscus clades being abundant (Devassy, Bhattathiri, and Radhakrishna 1983;. Diversity analysis based on bulk DNA and amplicon sequencing complemented those previously available datasets by adding a higher diversity of eukaryotic phytoplankton, including Pelagophyceae, Haptophyceae, Chrysophyceae, Eustigamatophyceae, Xanthophyceae, 95 Cryptophyceae, Dictyochophyceae, and Pinguiophyceaeadding and importantly by adding small cyanobacteria, which are difficult to count microscopically and were therefore not included into previous records (Löscher et al. 2020;Yuqiu et al. 2020;Bemal, Anil, and Amol 2019;Larkin et al. 2020;Pujari et al. 2019). Those cyanobacteria accounted for up to 60% of the primary producer abundance in sequence datasets in the central BoB (Li et al. 2012), and include Synechococcus and https://doi.org/10.5194/bg-2020-494 Preprint. Discussion started: 26 January 2021 c Author(s) 2021. CC BY 4.0 License.
Prochlorococcus, the former has been detected from the surface down to the chlorophyll maximum, while the latter has been 100 found abundant in the lower margin of the chlorophyll maximum at around 50 -80 m water depth, slightly deeper than the maximum of eukaryotic primary producers (Löscher et al. 2020;Yuqiu et al. 2020). The Prochlorococcus population has been described to consist of several different ecotypes of the HLII clade with their respective abundances being governed by macro-and micronutrient distribution and by temperature (Larkin et al. 2020;Pujari et al. 2019). Similar distributions of Atlantic have previously demonstrated the genetic potential of Prochlorococcus HLII clades to grow on nitrate (Rusch et al. 2007) supporting earlier suggestion that some Prochlorococcus ecotypes thrive at the base of the euphotic zone to acquire 110 nitrate from underlying waters (Vaulot and Partensky 1992;Olson et al. 1990). While a community shift within small cyanobacteria is speculative and may for bulk primary production quantitatively not be chiefly relevant, an overall increase in abundance of small cyanobacteria in concert with a decrease of eukaryotic primary producers would be expected to impact BoB biogeochemistry, especially the spatial expansion and the intensity of the OMZ through modified export production and respiration into low oxygen intermediate waters. 115 Besides those small cyanobacteria, there are reports on nitrogen fixing cyanobacteria of the Trichodesmium clade (Devassy, Bhattathiri, and Radhakrishna 1983;Jyothibabu et al. 2017;Sahu et al. 2017;Hegde et al. 2008;Shetye et al. 2013;Wu et al. 2019); other reports included diatom-diazotroph associations playing a role for BoB nitrogen fixation (Bhaskar et al. 2007).
However, for both types of nitrogen fixing primary producers, datasets are not conclusive and indicate high spatial and temporal variability. Nitrogen fixing microbes have been proposed to be limited by iron, other micronutrients or organic 120 matter in the BoB (Löscher et al. 2020;Saxena et al. 2020;Shetye et al. 2013). While micronutrients would have the potential to also directly limit primary production, a limitation of nitrogen fixers by organic matter would result in a feedback regulation of low primary production limiting nitrogen fixation and resulting low nitrogen availability limiting primary production.

Trends in primary production in the BoB
Over the last two decades, primary production in the global ocean has decreased (Gregg et al. 2003;Behrenfeld et al. 2006), and recent studies deducted a decrease in ocean primary production of 2.1% per decade associated largely to a decrease of chlorophytes in the marine photic realm (Gregg, Rousseaux, and Franz 2017;Gregg and Rousseaux 2019). Decreasing rates of primary production have been associated with high latitude regions (Gregg et al. 2003), but also with the Northern and 130 Equatorial Indian Ocean with a decrease of 9.7 and 17.2 % per decade, respectively (Gregg and Rousseaux 2019). These estimates, based on satellite imaging, were explained by a decrease in diatom and chlorophyte primary production of 15.4 https://doi.org/10.5194/bg-2020-494 Preprint. Discussion started: 26 January 2021 c Author(s) 2021. CC BY 4.0 License. and 24.8 % per decade, respectively, for both the BoB and its sister basin, the Arabian Sea (Fig. 3). This decrease has been connected to decreasing nitrate and silicate concentrations of 32.4% and 22.8% per decade in those waters limiting those larger, fast-growing primary producer groups (Gregg and Rousseaux 2019), with nitrate rather than silicate being limiting 135 primary production if assuming Redfield stoichiometry (Kumar et al. 2010;Radhakrishna et al. 1978). At the same time, an increase in small cyanobacterial primary producers, Prochlorococcus and Synechococcus, was described in this region, with an increase in cyanobacterial primary production of 16.7 % per decade (Gregg and Rousseaux 2019). Satellite-based imaging indeed showed a southward expansion and increase in abundance of cyanobacteria in the Bay of Bengal and through the Southern Arabian Sea (Fig. 3). Molecular genetic data showed, however, that not only Prochlorococcus is expanding but 140 that mostly certain ecotypes of high light Prochlorococcus increased in abundance and extended their habitat (Larkin et al. 2020). Given the decrease in both nitrate and silicate, a decrease in the silicate correlated ecotypes currently dominant in the northern BoB would be expected and those may be replaced by an open ocean ecotype sensitive to increasing iron concentrations in those waters. The overall increase in cyanobacteria derived from satellite monitoring is, however, not provable by direct measurements due to the lack of counts in the earlier reports, and further doesn't seem to impact the 145 overall prediction on primary production decrease.
Qualitatively consistent with the short term trend of decreasing primary production between 1998 and 2015, a pronounced decrease of up to 20% in phytoplankton in the Western Indian Ocean over the past six decades has been ascribed to increasing ocean stratification as a consequence of rapid warming in the Indian Ocean, which suppresses nutrient mixing from subsurface layers (Roxy et al. 2016). This result is indeed consistent with a long term trend with decreasing 150 productivity since the last glacial maximum (Contreras-Rosales et al. 2016;Shetye et al. 2014), therefore challenging the often discussed situation of the BoB at a tipping point (Bristow et al. 2017;Canfield et al. 2019;Rixen et al. 2020) to developing severe anoxia.

Possible scenarios in response to changes in primary production on the BoB OMZ 155
Reports of decreasing primary production in the BoB available from geological records, Earth system modelling, and satellite imaging are consistent over different time scales. But explanations on why primary production and chlorophyll concentrations decrease differ. Proposed important parameters include iron stress with iron concentrations having decreased in the geological record over the last 5000 years (Shetye et al. 2014), a decrease in nitrate and silicate availability directly 160 impacting primary producer growth (Gregg and Rousseaux 2019), and a rapid temperature increase of 0.6˚C over the last six decades, which may directly or indirectly via increased stratification decrease primary production (Roxy et al. 2016). These considerations cannot clearly be compared and evaluated using the few direct measurements available, as those expose a high temporal and spatial variability. inputs may increase and enter the ocean via rivers but would also at the same time be removed quickly and close to the coast as currently happening in the BoB (Naqvi et al. 2010), and temperatures would increase.
Assuming a limitation of primary production could by nitrogen availability, we would expect a niche for nitrogen fixation developing in the BoB. Until now, nitrogen fixation rates have shown to be low (Löscher et al. 2020;Saxena et al. 2020) and while there were reports on local blooms of the efficient nitrogen fixer Trichodesmium (Shetye et al. 2013), the nitrogen 170 fixer community is dominated by typically less active heterotrophic bacteria Turk-Kubo et al. 2014).
Nitrogen fixers in general have a high requirement for iron, therefore an iron limitation could ultimately limit nitrogen fixation and indirectly primary production, as discussed earlier (Löscher et al. 2020). A further decrease in iron would intensify this limitation and progressive decrease the productivity in the BoB. In addition, a decrease in silicate would limit diatom growth which need silica to form their shells. Therefore, a combined decrease of iron, nitrate and silicate 175 concentrations will lower primary production of various groups of primary producers at the same time, which may not only explain the trend visible from satellite imaging (Gregg and Rousseaux 2019) but may allow to predict a future trend for the BoB biogeochemistry.
Our earlier studies presented possible feedback cycles able to explain the persistent nanomolar levels of oxygen in the BoB OMZ (Canfield et al. 2019;Löscher et al. 2020). One approach included low mixing levels, or permanent stratification 180 limiting euphotic zone nutrient concentrations to an extent that new production is persistently low and organic matter recycling will not support the organic carbon requirement of the detected heterotrophic nitrogen fixer community. This we suggested to lead the system to being locked in a low productivity and increasingly nitrogen limited scenario with the OMZ increasingly weakening. Applying lower concentrations of nitrate and iron ( Fig. 4; silicate is not parametrized in our model but would lead to a similar effect if it would be limiting primary production), we observe that the OMZ respiration will 185 lower, low oxygen concentrations will be maintained, and denitrification will only occur if nutrients are imported into the OMZ from land, via rivers, from the atmosphere, or by increased upwelling (Fig. 4). This would mean the BoB may not be at a tipping point towards anoxia but is a system with a weakening OMZ in its open waters, with progressive warming stabilizing this trend by increasing stratification in the photic zone and cutting this part of the water column off any nutrient supply. 190 Comparing scenarios of primary production and its impact on the BoB OMZ, the last glacial maximum signifies an episode of high productivity in the geological record (Contreras-Rosales et al. (2016) , Fig. 5A). Higher land runoff and riverine inputs led to both higher nutrient imports, but also increased loads of terrigenous material leading facilitating organic carbon export from the productive zone to the sediments through ballasting. This effect is currently also seen, with nutrients being imported and consumed close to the shelf and organic material exported out of the photic zone (Fig. 5B), leading to a carbon 195 pump with similar export rates like in the Arabian Sea (Singh and Ramesh 2015). A scenario with production being enhanced would strongly depend on external nutrient inputs, those could come from land and riverine inflow, and could for example result from deforestation, enhanced monsoon events, increasing atmospheric input, or enhanced upwelling, which has been described to enhance primary production in mesoscale eddies (Sarma and Udaya Bhaskar (2018) Table 1: Historical record of water column integrated chlorophyll a concentration, surface chlorophyll a concentration, and primary production rates   leading to an anoxic OMZ. The latter would require higher nutrient fluxes from either the atmosphere, from upwelling or from rivers and land. Fluxes from land may increase e.g. from deforestation, or enhanced rainfalls, however, higher terrigenous particle load would likely accompany increased nutrient loads and therefore, even if coastal primary production would increase, export production would increase via ballasting, too. The effect on the OMZ would then be rather small. 420 Upwelling-dependent nutrient pumping is unlikely to increase due to warming and enhanced stratification but may occur especially in eddy systems, i.e. as eddy pumping. Grey triangles depict nutrient fluxes from land and rivers, from upwelling along the shelf and from the atmosphere, brown arrows depict terrigenous particle fluxes as imported from rivers and leading to ballasting and enhanced organic carbon export, green arrows depict organic material exported from the photic zone primary production (green bubble, PP) into the OMZ (purple bubble). Sizes of triangles, arrows, and bubbles qualitatively 425 indicate proportions.