The effect of seasonally and spatially varying chlorophyll on Bay of Bengal surface ocean properties and the South Asian Monsoon

Chlorophyll absorbs solar radiation in the upper ocean, increasing mixed-layer radiative heating and sea surface 15 temperatures (SST). Although the influence of chlorophyll distributions in the Arabian Sea on the southwest monsoon has been demonstrated, there is a current knowledge gap in how chlorophyll distributions in the Bay of Bengal influence the southwest monsoon. The solar absorption caused by chlorophyll can be parameterized as an optical parameter, h2, the scale depth of absorption of blue light. Seasonally and spatially varying h2 fields in the Bay of Bengal were imposed in a 30-year simulation using an atmospheric general circulation model coupled to a mixed layer thermodynamic ocean model to investigate 20 the effect of chlorophyll distributions on regional SST, southwest monsoon circulation and precipitation. There are both direct local upper-ocean effects, through changes in solar radiation absorption and indirect remote atmospheric responses. The depth of the mixed layer relative to the perturbed solar penetration depths modulates the response of SST to chlorophyll. The largest SST response of 0.5°C to chlorophyll forcing occurs in coastal regions, where chlorophyll concentrations are high (> 1 mg m3), and when climatological mixed layer depths shoal during the intermonsoon periods. Precipitation increases significantly by 25 up to 3 mm day-1 across coastal Myanmar during the southwest monsoon onset and over northeast India and Bangladesh during the Autumn intermonsoon period, decreasing model biases.


30
The strong coupling of the Indian Ocean to the atmosphere is a major factor in South Asian monsoon seasonal variability (Ju and Slingo, 1995). During the boreal summer, strong southwesterly winds transport heat and moisture from the Indian Ocean surface to sustain deep convection over the Indian subcontinent. The South Asian summer monsoon provides up to 90% of the annual rainfall for the Indian subcontinent (Vecchi and Harrison, 2002), so it is important to accurately predict the seasonal variability of monsoon rainfall given its economic importance to agriculture and other water-intensive industries.

35
The South Asian monsoon is initiated when lower-tropospheric winds, transporting heat and moisture, begin to flow northward from the equator to the Asian continent in response to increasing summer insolation and increasing land-sea thermal and pressure gradients (Grey arrows; Fig. 1; Webster et al., 1998). Mid-tropospheric heating from the elevated Tibetan Plateau increases the land-sea thermal and pressure contrast, further regulating the seasonal reversal of the large-scale circulation (Li and Yanai, 1996). From June to September (JJAS) high climatological precipitation rates (> 20 mm day -1 ), associated with the 40 South Asian southwest monsoon, are anchored to three locations across the Indian subcontinent: the western Ghats of southwest India, the Myanmar coast and from Bangladesh north into the Himalayan foothills ( Fig. 2f-2i). Coupled atmosphereocean general circulation models (GCMs) have improved their representations of the seasonal variability and spatial distribution of South Asian southwest monsoon precipitation, but substantial biases remain. Lin et al. (2008) found that 12 out of 14 coupled GCMs from the Coupled Model Intercomparison Project Phase 3 (CMIP3) captured the South Asian southwest 0.5-1.0°C and increased rainfall by 2 mm day -1 over southwest India during the southwest monsoon onset. Park and Kug (2014) used a biogeochemistry model coupled to an ocean GCM to investigate the biological feedback on the Indian Ocean Dipole (IOD). The response to interactive biology enhanced both warming during a positive IOD (cooling in the eastern Equatorial Indian Ocean) and cooling during a negative IOD (warming in the eastern Equatorial Indian Ocean), thus dampening the IOD magnitude, which could have significant effects on the South Asian summer monsoon.

70
The thermal and saline surface properties of the Bay of Bengal (BoB; Fig. 1), in the northeast Indian Ocean, are strongly forced by the monsoonal winds and large freshwater flux. In the north BoB, the large freshwater flux from river discharge and precipitation leads to strong salinity stratification and barrier-layer formation above the thermocline and below the mixed layer (Vinayachandran et al., 2002;Jana et al., 2015;Sengupta et al., 2016). The barrier layer inhibits vertical mixing (Sprintall and Tomczak, 1992;Rao and Sivakumar, 2003) and isolates the mixed layer above from cooling by entrainment (Duncan and Han,75 2009), modulating the seasonal MLD and its temperature (Girishkumar et al., 2011;Shee et al., 2019).
The BoB SST rapidly responds to variations in the net surface heat flux, which in turn are primarily controlled by variations in windspeed (Duncan and Han, 2009). Although BoB SST decreases with increasing windspeed during the southwest monsoon (JJAS), SST remains high enough (> 28°C) to sustain high precipitation rates across the Indian subcontinent, consequently strengthening the salinity stratification and further reinforcing convection across the basin (Shenoi 80 et al., 2002). The salinity stratification is weaker in the southern BoB, allowing monsoonal winds to primarily control the upper-ocean thermal structure (Narvekar and Kumar, 2006). Hence, the southern BoB MLD and SST display larger seasonal variability compared with the northern BoB (Narvekar and Kumar, 2006).
The strong BoB salinity stratification reduces biological productivity by inhibiting the vertical transport of nutrients to the sun-lit surface layers (Kumar et al., 2002;McCreary et al., 2009). Biological productivity during JJAS is also inhibited by cloud cover and by the infiltration of river sediments, which respectively reduce the incoming solar radiation at the ocean surface and the in-water penetration depth of solar radiation (Gomes et al., 2000;Kumar et al., 2010). However, in certain regions of the BoB, localised seasonal physical forcing breaks the strong stratification and increases the vertical transport of nutrients to the sun-lit surface layers, increasing biological productivity. Chlorophyll concentrations in the coastal regions are high (> 1 mg m -3 ; Fig. 1), especially near large rivers such as the Ganges, Brahmaputra, Mahanadi and Irrawaddy, because of 90 nutrients supplied by these rivers during June-October (Amol et al., 2019). Chlorophyll concentrations in the northern coastal region typically peak in October (Fig. 3j;Lévy et al., 2007) when river discharge and nutrients also peak (Rao and Sivakumar, 2003). High chlorophyll concentrations are then transported along the northeast coast of the BoB (Amol et al., 2019).
In the southern BoB, strong southwesterly winds across the southernmost tip of India and Sri Lanka initiate coastal upwelling and thus biological productivity, leading to a maximum in chlorophyll concentration there in August ( Fig. 1; Lévy 95 et al., 2007). The Southwest Monsoon Current (SMC), a shallow, fast current, advects these high chlorophyll concentrations to the southwest BoB ( Fig. 1; Vinayachandran et al., 2004). High chlorophyll concentrations are sustained east of Sri Lanka by the cyclonic (anticlockwise) eddy of the Sri Lanka Dome (SLD), where open ocean Ekman upwelling transfers nutrients to the near surface during JJAS ( Fig. 1; Vinayachandran and Yamagata, 1998;Vinayachandran et al., 2004;Thushara et al., 2019). In the west and southwest BoB in winter, northeasterly winds induce open-ocean Ekman upwelling, leading to increased 100 chlorophyll concentrations peaking in December and January (Fig. 3l-3a;Vinayachandran and Mathew, 2003;Lévy et al., 2007). Chlorophyll concentrations in the open BoB also show sub-seasonal and mesoscale variability. Surface chlorophyll concentrations are periodically enhanced by transient cold-core eddies and post-monsoon cyclones, where the strong salinity stratification is briefly eroded and nutrients are transported to the near-surface in the western and central BoB (Vinayachandran and Mathew, 2003;Kumar et al., 2007;Patra et al., 2007).

105
A few studies have briefly analysed the effect of seasonally varying chlorophyll concentrations on BoB upper ocean dynamics and SST, whilst also speculating how this may affect the South Asian monsoon (Murtugudde et al., 2002;Wetzel et al., 2006). Although the effect of chlorophyll on BoB SST has been demonstrated by these previous studies, the effect of chlorophyll on monsoon rainfall remains a vital knowledge gap. Without this knowledge then missing bio-physical interactions in the BoB could lead to inaccuracies in simulated air-sea interactions that are crucial in representing accurate monsoon 110 behaviour and thus rainfall timing, location and duration over the Indian subcontinent. This study analyses the direct effect of BoB seasonally varying chlorophyll concentrations on the South Asian monsoon in an atmospheric GCM that is coupled to a mixed layer thermodynamic ocean model. A description of the experimental design, model and observed datasets used in this study is presented in Section 2. Section 3 presents the results of the control and chlorophyll-perturbed model outputs. Section 4 discusses and concludes the results from the chlorophyll-perturbed experiment. Atmosphere 7.0 (Walters et al., 2019). The atmospheric and oceanic horizontal resolution is N216 (0.83° longitude x 0.56° latitude), which corresponds to a horizontal grid spacing of approximately 90 km. There are 85 vertical levels in the atmosphere, with approximately 50 vertical levels in the troposphere. MetUM-GOML3.0 is configured similarly to MetUM-GOML2.0 (Peatman and Klingaman, 2018) and MetUM-GOML1.0 (Hirons et al., 2015), except that the atmospheric model 125 is updated to GA7.0 and the air-sea coupling routines are updated to couple the models via the Ocean-Atmosphere-Sea Ice-Soil (OASIS) Model Coupling Toolkit (Valcke, 2013).
MC-KPP consists of a grid of independent one-dimensional columns, with one column positioned under each atmospheric grid point at the same horizontal grid spacing as MetUM GA7.0. The ocean columns are 1000 m with 100 vertical points, with 70 points in the top 300 m; the near-surface resolution is approximately 1 m. This improves the representation of MLD and SST, which has been shown to improve tropical convection and circulation on subseasonal scales when coupled to an atmospheric GCM (Bernie et al., 2005;Bernie et al., 2008;Klingaman et al., 2011). Each column is subject to surface forcing from freshwater, heat and momentum fluxes; vertical mixing is parameterised using the KPP scheme from Large et al. (1994).
The MLD is defined as the depth where the bulk Richardson number equals the critical Richardson number of 0.3 (Large et al., 1994). The coastal region in MC-KPP is represented with columns that are partially ocean and partially land. The surface 135 properties for ocean and land are computed separately in MC-KPP and the mean grid point temperatures are computed in the atmospheric model by combing the ocean and land surface temperatures from MC-KPP.
Solar radiation absorption is represented as a wavelength-dependent penetration depth, with blue wavelengths penetrating deeper than red wavelengths. The decay of solar irradiance through the water column is represented as a simple two-band double-exponential function (Paulson and Simpson, 1977): where I(z) is the solar irradiance at depth z; I0 is the solar irradiance at the ocean surface; R is the ratio of red light to the total visible spectrum; and h1 and h2 are the scale depths of red and blue light, respectively. The scale depth, or e-folding depth, is defined as the depth where solar irradiance is approximately 63% less than its surface value (1 -e -1 ). Paulson and Simpson (1977) determined the optical parameters based on each of the five Jerlov water types that categorise open ocean turbidity 145 (Jerlov, 1968 (Morel, 1988); h1 and h2 are 1 m and 17 m, respectively. Increasing upper-ocean turbidity to water type III, where chlorophyll concentrations exceed 1.5-2.0 mg m -3 (Morel, 1988), yields h1 and h2 of 1.4 m and 7.9 m, respectively. The scale depth for red light (h1 ~ 1 -1.4 m) for all water types is much less than the typical MLD (> 10 m). Hence, all red light is absorbed at the top of the mixed layer. However, the scale depth for blue light (h2 ~ 8 -17 m) is comparable to the typical MLD; a significant 150 fraction of blue light will penetrate below the mixed layer. Hence, the reduction of h2 with increasing turbidity controls the radiant heating of the mixed layer and thus SST (Zaneveld et al., 1981;Lewis et al., 1990;Morel and Antoine, 1994). Paulson and Simpson (1977) (2007). A relaxation timescale of 15 days is optimal to produce temperature and salinity tendency terms that minimise biases in the free-running simulations (Hirons et al., 2015). A mean seasonal cycle of daily temperature and salinity tendencies that is computed from this relaxation simulation is applied to the 30-year chlorophyll perturbation simulations. The spin-up time is small (1 year) as the ocean is adjusted to a mean state after the relaxation simulation. The absence of ocean dynamics means MetUM-GOML does not represent coupled modes of variability (e.g. ENSO or IOD) that rely on a dynamical ocean (Hirons et al., 2015). The benefit of not representing these modes of variability is that the signal from the chlorophyll perturbation experiment will not be obscured by the "noise" of these interannual climate variations. The absence of full ocean dynamics also reduces computational cost and allows the model to be used for climate-length coupled simulations with shorter spin-up periods (Hirons et al., 2015).

MC-KPP uses the
We directly impose a seasonally varying h2 value (representative of chlorophyll concentration) to selected columns within Furthermore, the absence of biological and physical feedbacks on chlorophyll development means that a consistent seasonally varying h2 value is directly imposed on columns within the BoB throughout the simulation.

Chlorophyll-a data
To produce a temporally and spatially varying field of h2 for MC-KPP, a monthly climatology of chlorophyll-a concentration, measured from the Moderate Resolution Imaging Spectroradiometer (MODIS) on the Aqua satellite, was used.

185
Chlorophyll-a retrievals from 0.25 to 0.3 mg m -3 were calculated by merging the CI and OC3 algorithms to create the Ocean Color Index (OCI) algorithm (Wang and Son, 2016;Hu et al., 2019). Chlorophyll-a concentration retrievals above 5 mg m -3 reduce the effectiveness of the OC3 algorithm (Morel et al., 2007). Organic and terrestrial material, introduced by rivers or mixed by tidal currents in coastal regions, change the scattering of visible light, affecting the water-leaving radiances (Boss et al., 2009) and leading to an overestimate in chlorophyll-a concentration (Morel et al., 2007). Hence, remotely sensed chlorophyll-a concentrations were not determined in the eutrophic coastal regions of the Ganges and Irrawady river deltas because of the large amount of suspended organic and terrestrial material (grey shading; Fig. 1; Tilstone et al., 2011). MODIS sensor degradation on the Aqua satellite has been small (Franz et al., 2008) and all ocean color products have since been corrected and improved after cross-calibration with the SeaWiFS climatology (Meister and Franz, 2014). Chlorophyll-a will henceforth be referred to as "chlorophyll" for convenience.

Experiment set-up
To investigate the impact of the seasonal and spatial variability of chlorophyll-induced heating in the BoB, two 30-year simulations were completed, with differing prescribed h2 (representative of chlorophyll concentrations): a control run using h2 = 17 m globally and a perturbation run using an annual cycle of h2 at daily resolution for the BoB region (defined below) and 200 h2 = 17 m over the rest of the global ocean. In both simulations, R and h1 were kept constant, at 0.67 and 1.0 m respectively, representative of water type IB. The first year of both simulations was discarded due to spin up; the analysis was carried out on the remaining 29 years.
The control simulation used an effective constant global chlorophyll concentration of ~0.15 mg m -3 , which corresponds to h2 = 17 m (Jerlov water type IB; Morel, 1988). Previous studies have used control simulations with zero chlorophyll 205 concentrations to see the full impact of chlorophyll on physical and dynamical processes (e.g. Gnanadesikan and Anderson, 2009), whilst other studies have used constant scale depths determined from parameterizations of the lowest chlorophyll concentrations encountered (e.g. Shell et al., 2003;Turner et al., 2012). Satellite observations show that the global average chlorophyll concentration for oceans deeper than 1 km is 0.19 mg m -3 , similar to the value in our control simulation.

210
For the perturbation simulation, the BoB region was defined as the area 77-99.5° E and 2.5-24° N (black dashed box; Fig. 3). The region extends far enough south and west to incorporate the high surface chlorophyll around the southernmost tip of India and Sri Lanka, but excludes the relatively low near-equatorial surface chlorophyll concentrations ( Fig. 3f-3j). The Isthmus of Thailand and Myanmar to the east, and India and Bangladesh to the north and west, form a natural boundary to the defined BoB region (Fig. 1). An annual cycle of daily chlorophyll concentration for MetUM-GOML was derived by linearly ocean. The relationship shows scale depth varying as a power law function of surface chlorophyll concentration with the largest variability of scale depth (> 18 m) at the lowest concentrations (< 0.1 mg m -3 ).
Missing h2 data were common in regions such as the Ganges River delta due to undetermined remotely sensed chlorophyll concentrations from highly turbid coastal waters. Missing h2 data in this delta extend further out onto the continental shelf during JJAS as floodwaters drain into the BoB transporting finer silt and clay further offshore (Kuehl et al., 1997). The missing 225 h2 data were typically associated with regions where the land fraction was less than 1, which includes the narrow Isthmus of Thailand and the low-lying land of the Ganges delta (black hatching; Fig. 1). A minimum of two h2 values from two neighbouring data points were required to find an average h2 value to fill in the missing data point. At the boundary of the BoB domain, to avoid sharp gradients the seasonally varying h2 values within the BoB domain were smoothly transitioned (linearly) to the constant h2 = 17 m outside the BoB domain, over a buffer region of three grid points.

230
Vertically integrated moisture fluxes (VIMF) were used to evaluate the water vapour transport sourced from the chlorophyll-forced BoB to the surrounding Indian subcontinent. The VIMF was calculated as where ,⃗ is the horizontal wind velocity, q is the specific humidity, g is the acceleration due to gravity, p is pressure and the integration was between 1000 and 100 hPa. Note that ,⃗q was output directly by the model as monthly mean values. VIMF 235 divergence was used to evaluate the precipitation rate changes that are due to changes in water vapour divergence. The VIMF divergence was calculated as where the integration was between 1000 and 100 hPa.
The observed monthly 18-year (1998-2015) climatological precipitation rate measured from the Tropical Rainfall

240
Measuring Mission (TRMM) 3B42 satellite product (Huffman et al., 2007) was used to diagnose the bias in the model precipitation rate. An area-weighted re-gridding scheme was used to reduce the 0.25° horizontal resolution of the observed monthly climatological precipitation rate to match the horizontal resolution of MetUM-GOML.

Southwest monsoon onset (April to June)
The BoB surface ocean responds to the imposed annual cycle of h2 in the perturbation run during the onset of the The imposed annual cycle of h2 directly affects coastal SST. During April the increase in solar absorption by chlorophyll along the northern and western coastal regions significantly (at 5% level) increases monthly average SST by 0.5°C ( Fig. 4d). Correspondingly, the monthly average 1.5 m air temperature increases by 0.5°C in the perturbation run (Fig. 4g).
The strengthening alongshore wind over the warmer western coast results in a large increase in upward latent heat flux of 20 W m -2 (Fig. 4j). This increase in atmospheric moisture leads to an anomaly in the VIMF of 30 kg m -1 s -1 (Fig. 5a) that is in the same direction as the mean VIMF in the control run (Fig. 6a). The increase in VIMF converges over northeast India and Bangladesh as shown by the negative VIMF divergence (Fig. 5a), supplying extra moisture needed for the increase in precipitation rate of 2 mm day -1 (significant at the 5% level; Fig. 4m).
The increase in solar absorption in the mixed layer by high chlorophyll concentrations persists during May and June along the coasts ( Fig. 4b and 4c). Low h2 along the northern and western BoB coastal regions acts to increase monthly average 265 SST by 0.5°C ( Fig. 4e and 4f). The upward latent heat flux increases ( Fig. 4k and 4l) due to an increase in the specific humidity at the surface, which is associated with the higher SST. This increase in SST is therefore offset by the negative feedback from the latent heat flux increase.
In June, the precipitation rate over the Myanmar coast increases by 3 mm day -1 (significant at the 5% level; Fig. 4o).
The monthly average 1.5 m air temperature increases by 0.4°C (Fig. 4i), which is caused by an increase in SST (Fig. 4f) where 270 h2 along the western BoB is low (Fig. 4c). The upward latent heat flux increases by 10 W m -2 (Fig. 4l) and the VIMF increases by 20 kg m -1 s -1 (Fig. 5b) in addition to a strengthening southwesterly moisture transport during the southwest monsoon onset (Fig. 6b). The enhanced convergence of VIMF over the Myanmar coast ( Fig. 5b) supplies the moisture for the increase in precipitation rate (Fig. 4o). We have thus demonstrated that high coastal chlorophyll concentrations perturb the absorption of solar radiation that increases air temperature and SST, which significantly increases spring intermonsoon precipitation rates in 275 the northern and eastern BoB.

Southwest monsoon (July to October)
The values of h2 continue to decrease in the BoB open ocean into July and August ( Fig. 7a and 7b), as high chlorophyll BoB surface ocean and regional climate respond to the above changes in h2 during JJAS. Higher coastal SSTs (significant at the 10% level) are co-located with the high coastal chlorophyll concentrations, whereas open-ocean SST is largely unchanged by BoB chlorophyll forcing ( Fig. 7e-7g). In July, an increase in SST and a slight increase in alongshore windspeed over the west BoB increases the upward latent heat flux (Fig. 7m), but this does not significantly change 290 precipitation rate (Fig. 7q). In August, a further increase in the alongshore windspeed increases the magnitude and spatial extent of the upward latent heat flux across the northern BoB (Fig. 7n). During September an increase in windspeeds over the northern Myanmar coast increases surface ocean evaporation (Fig. 7o). The VIMF increases in magnitude and remains approximately in the same direction as the mean VIMF in the control run ( Fig. 5c and 6c). Negative VIMF divergence over the northern Myanmar and Bangladeshi coast in the perturbation run (Fig. 5c) supplies moisture for the increase in precipitation 295 rate in this region (significant at the 5% level; Fig. 7s).
By October the combined atmospheric moisture sourced from the warmer western BoB and Andaman Sea leads to an increase in precipitation rate of up to 3 mm day -1 over west Bangladesh and northeast India (significant at the 5% level; Fig. 7t). The spatial extent of the increased precipitation rate is considerably larger than previous months, extending further west over the Indo-Gangetic plain and encompassing megacities such as Kolkata and Dhaka. An area-weighted 29-year monthly average precipitation rate over west Bangladesh and northeast India (20-25° N, 85-90° E; black dashed box in Fig.   7t) shows a rainfall maximum in August in both simulations (Fig. 8a). The precipitation rate differences gradually increase from July to August and peak in October at 2 mm day -1 (Fig. 8b). Alongshore winds over the warmer Isthmus of Thailand and the coast of Myanmar further increase atmospheric moisture transport to the northern BoB (Fig. 6d). The upward latent heat flux increases by 13 W m -2 (Fig. 7p) and the VIMF increases by 30 kg m -1 s -1 over the coast of Myanmar (Fig. 5d). The negative 305 VIMF divergence over west Bangladesh and northeast India supplies moisture for the increase in precipitation rates in this region (Fig. 5d). As in the spring intermonsoon, the increase in precipitation rate during autumn intermonsoon in the northern BoB is primarily attributed to high coastal chlorophyll concentrations and increased SST extending from the Andaman Sea to the Ganges river delta along the chlorophyll-perturbed BoB coastal region.
The enhanced convective activity over west Bangladesh and northeast India during October is associated with an 310 increase the vertical wind velocity at the 500 hPa pressure level (Fig. 9a). At the 200 hPa pressure level enhanced westerly winds converge over eastern China (Fig. 9b), which leads to increased subsidence (Fig. 9a). This subsidence reduces precipitation and increases surface temperature over eastern China (significant at the 5% level; Fig. 9c and 9d). This indirect remote response resembles the effect of the "Silk Road" pattern; a stationary Eurasian-Pacific Rossby wave train that occurs during the Northern Hemisphere summer and produces significant air temperature and rainfall anomalies in east Asia (Ding 315 and Wang, 2005).

Mixed layer radiant heating and SST modulation
The hypothesised direct link between a change in h2 and a resultant change in SST is examined in more detail in this 320 subsection. The radiant heating rate of the mixed layer, and resultant change in SST, depends not only on h2, but also on changes in the surface flux of shortwave radiation, which is dependent on cloud cover, and changes in the depth of the mixed layer. Here, we assess which of these three factors is primarily responsible for the changes in the radiant heating rate of the mixed layer.
We assume that the red-light radiative flux is absorbed within approximately the top 1 m and entirely within the where T is the temperature of the mixed layer; t is time; :; (0) is the monthly 29-year average downward shortwave radiation flux incident at the ocean surface; = 1025 kg m -3 is the density of the mixed layer; < = 3850 J kg -1 K -1 is the specific heat 330 capacity of sea water; R = 0.67 is the ratio of red light to total visible light for Jerlov water type IB; H is the monthly 29-year average MLD; and h2 is the monthly average h2 that was imposed in the control and perturbation run.
Within the BoB, the largest imposed change in h2 is 13 m. Assuming that the other variables remain constant, a change in h2 of 13 m changes radiant heating rates by 0.4°C month -1 . The largest model change in downward shortwave radiation is 14 W m -2 , which changes radiant heating rates by 0.3°C month -1 , comparable to the change from h2 variations. The largest model 335 MLD change is 3 m, which changes radiant heating rates by 0.4°C month -1 , also comparable to the change from h2 variations.
We compare the mixed layer radiant heating rates of the control and perturbation runs during June and October ( Fig. 10a and 10b). We focus on two regions: the open ocean region of the SMC (83-86° E, 5-8° N; black boxes in Fig. 10) and the coastal region of the Irrawaddy Delta (95-98° E, 14-17° N; black boxes in Fig. 10). The two regions are an important source of heat and moisture for the June and October precipitation rate perturbations and display distinctive chlorophyll regimes. The SMC is an open ocean region that displays large seasonal changes in h2, whilst the Irrawaddy Delta is a coastal region that displays continuously low h2. In June and October, coastal regions have the highest radiant heating rate difference between the control and perturbation runs ( Fig. 10a and 10b). In June, the area-weighted mean radiant heating rate in the coastal region of the Irrawaddy Delta increases by 0.4°C month -1 in the perturbation run (Fig. 10a). An h2 decrease of 9 m has the largest contribution to the radiant heating rate increase of 0.7°C month -1 (Fig. 11a), compared with an MLD decrease of 0.2 m (Fig.  10e), which contributes to less than 0.1°C month -1 (Fig. 11e). A decrease in downward shortwave radiation flux of 8 W m -2 ( Fig. 10c), associated with an increase in monsoon cloud cover, cools the region by 0.3°C month -1 (Fig. 11c). In October, the radiant heating rate difference in the Irrawaddy Delta increases by 1.5°C month -1 in the perturbation run (Fig. 10b). The radiant heating rate difference is larger than June because of an increase in monthly average downward shortwave radiation flux and a shallower MLD in both the control and perturbation runs. A decrease in h2 of 9 m has the largest contribution to the radiant 350 heating rate increase of 1.4°C month -1 (Fig. 11a), whereas, a decrease in the MLD of 0.1 m (Fig. 10f) and an increase in downward shortwave radiation flux of 1 W m -2 (Fig. 10d) only contribute to less than 0.1°C month -1 of the increase in radiant heating rate respectively ( Fig. 11d and 11f). The changes in h2 are more influential on mixed layer radiant heating rates and SSTs compared with small changes in MLD and downward shortwave radiation flux in the Irrawaddy Delta during June and October.

355
In June, the area-weighted mean radiant heating rate difference in the SMC region decreases by 0.1°C month -1 in the perturbation run (Fig. 10a). A decrease in the downward shortwave radiation flux of 5 W m -2 (Fig. 10c) has the largest contribution to the radiant heating rate decrease of 0.1°C month -1 (Fig. 11c), whereas, a decrease in h2 of 2 m and an increase in MLD of 0.4 m (Fig. 10e) contribute less than 0.1°C month -1 to the radiant heating rate ( Fig. 11a and 11e). In October, the radiant heating rate difference of the SMC region shows an increase of 0.1°C month -1 (Fig. 10b). A decrease in h2 of 3 m has the largest contribution to the radiant heating rate increase of 0.1°C month -1 (Fig. 11b), whereas, a decrease in downward shortwave radiation flux of 1 W m -2 (Fig. 10d) and an increase in MLD of 0.2 m (Fig. 10f) contribute less than 0.1°C month -1 to the radiant heating rate ( Fig. 11d and 11f). In the SMC region, changes in h2 are smaller than those in coastal regions during In the Irrawaddy Delta region during October, the MLD shoals to 9 m (green dashed line; Fig. 12b), which is similar to 370 the perturbed h2 (green dot; Fig. 12b). When the mixed layer is shallow, the increased near-surface radiant heating from reducing h2 is distributed to a shallower depth, increasing the average change in the radiant heating rate by 1.2°C month -1 (∆dT/dt; Fig. 12f). Below 10 m depth radiant heating rates reduce due to reduced h2. There is also no change in MLD in response to reduced h2 in the perturbation run. The increase in local wind speed of 0.8 m s -1 is likely to have de-stratifying effects on the upper ocean that oppose the stratifying effects of increased mixed layer radiant heating. When the MLD deepens 375 below 10 m, the biological-induced effects of the increased radiant heating rates above 10 m and reduced radiant heating rates below 10 m are mixed, reducing the net effect of biological heating on mixed layer temperature. In June, the MLD deepens to 16 m (Fig. 12a), resulting in a smaller average radiant heating rate change of 0.4°C month -1 (Fig. 12e). Consequently, the October SST increases by 0.5°C, compared with a smaller increase of 0.2°C in June. Hence, shoaling the mixed layer to a depth comparable to the perturbed solar penetration depth in October limits the turbulent mixing processes to a depth where 380 chlorophyll perturbs solar radiation absorption, and makes SST more sensitive to chlorophyll concentration changes.
In the SMC region during October, the MLD shoals to 28 m (Fig. 12d), approximately twice the depth of the perturbed h2, resulting in an average change in the mixed layer radiant heating rate of 0.1°C month -1 (Fig. 12h). As in the Irrawaddy Delta region, there is no change in MLD in response to biological warming in the SMC region due to an increase in local wind speed of 0.8 m s -1 , which is likely to oppose the stratifying effects of increased mixed layer radiant heating. During June, the 385 MLD extends to 36 m (Fig. 12c), resulting in an average change in the mixed layer radiant heating rate below 0.1°C month -1 (Fig. 12g). As in the Irrawaddy Delta region, the effect of chlorophyll on upper ocean temperature depends on the MLD in the SMC region, with the shallowest MLD and largest change in radiant heating rate in October. With lower chlorophyll concentrations in the SMC region than the Irrawaddy Delta region, the resultant change in SMC regional average radiant heating rate in the top 10 m is considerably lower. and October respectively (Lévy et al., 2007), whilst the MLD is shallowest across the basin, which results in an increase in mixed layer radiant heating rate and SST in the western BoB in autumn.
The direct changes in h2 in coastal regions are large, and thus more influential on mixed layer radiant heating rate and SST. The resultant increase in the radiant heating rate of the coastal mixed layer and SST during the southwest monsoon onset and autumn intermonsoon increases the latent heat flux and transport of moisture to the Indian subcontinent. Precipitation rates 410 over the Myanmar coast during the southwest monsoon onset increase by 3 mm day -1 . Comparing the monthly average precipitation rate difference ( Fig. 4o) with the control simulation bias (Fig. 13a) shows that the model dry bias of 4 mm day -1 over the Myanmar coast is partly removed in the perturbation run. Precipitation rates over western Bangladesh and northeastern India during the autumn intermonsoon increase by 3 mm day -1 . Comparing the precipitation differences ( Fig. 7t) with the model bias (Fig. 13b) shows that the model dry bias of up to 3 mm day -1 over northeast India is removed in the perturbation 415 run. The reduced model biases after imposing a more accurate representation of chlorophyll further highlights the importance of including chlorophyll in coupled models. concentration region, which reduces radiant heating rates (∆RHR < 0) below the mixed layer. Increasing the mixed layer radiant heating rate increases mixed layer temperature and SST (∆SST > 0). The upward latent heat flux and evaporation increases with increasing SST and strengthening monsoon winds. Convergence of the additional lower-tropospheric moisture that is transported by the monsoon winds increases the precipitation rates to the east.

425
During October, the enhanced precipitation rate and convective activity in the northern BoB perturbs uppertropospheric winds, potentially causing reduced precipitation rates over eastern China, similar to the Silk Road effect. The Silk Road pattern has been found to influence extreme heat waves over eastern China, causing considerable socio-economic devastation (Thompson et al., 2019). Indeed, the model does display significantly warmer surface temperatures in this region at this time (Fig. 9c). The Silk Road pattern dynamics have been previously linked to the South Asian summer monsoon 430 (Stephan et al., 2019). Diverging upper-tropospheric winds caused by precipitation anomalies over the Indian subcontinent interact with midlatitude westerlies, which influences the strength and positioning of the subtropical northwestern Pacific anticyclone over eastern China (Ding and Wang, 2005;Hu et al., 2012a). The effect of chlorophyll on the midlatitude Rossby wave train and its potential impact on East Asian climate needs further investigation. has less biological productivity than the Arabian Sea because of light and nutrient limitation (Kumar et al., 2002), though chlorophyll concentrations in the coastal BoB can be as high as in the Arabian Sea. The BoB is also exposed to the same 440 monsoonal winds as the Arabian Sea. Such localised, physical forcing modulates the MLD, which in turn modulates the biological warming. Hence, the SST increase of 0.5°C in coastal regions of the BoB during the spring and autumn intermonsoons is similar to the increase in SST in the Arabian Sea during the spring intermonsoon.
Previous studies show that the effect of biological warming is amplified due to secondary feedbacks on MLD. In the Arabian Sea, high chlorophyll concentrations increase solar radiation absorption and so increase thermal stratification, which 445 inhibits vertical mixing, shoals the MLD and further increases SST (Nakamoto et al., 2000;Wetzel et al., 2006;Turner et al., 2012). In our study, secondary feedbacks on the MLD are consistent in magnitude with the Arabian Sea studies. The maximum MLD difference is 3 m in the central BoB during June. Coastal MLDs shoaled around the southernmost tip of India and the northern BoB in June by ~1 m and MLDs shoaled around the Isthumus of Thailand in October by ~1 m (Fig. 10e and 10f).
The effect of high chlorophyll concentrations in these localised coastal regions appears to have altered upper-ocean thermal 450 stratification when there is little or no change in windspeed, while in the majority of the BoB, changes in windspeed primarily alter upper-ocean thermal stratification.
In our study, a realistic chlorophyll distribution increased open ocean SST by ~0.1°C and increased coastal SST by ~0.5°C during the intermonsoons and southwest monsoon onset. The simulated increase in open ocean SST is consistent with previous work (Murtugudde et al., 2002;Wetzel et al., 2006). However, the increase in coastal SST, primarily in the eastern BoB coastal Hence, the coastal and open ocean SST responses are more accurately represented here than in previous work.
The derivation of the imposed annual cycle of h2 in coastal regions has limitations. Firstly, the ocean colour algorithms used to determine chlorophyll concentrations from satellite are not completely effective in turbid coastal waters (Morel et al., 2007;Tilstone et al., 2013). Organic and inorganic constituents such as Coloured Dissolved Organic Matter (CDOM) and suspended sediments strongly attenuate blue light and are thus falsely identified as a chlorophyll-a pigment, which typically leads to an overestimation in chlorophyll concentration (Morel et al., 2007). Secondly, the Morel and Antione (1994) chlorophyll parameterisation is not applicable for coastal waters, as the parameterisation is based on the absorption by chlorophyll-a pigments and not by the attenuation of other in-water constituents. Missing h2 values in the Ganges river delta are interpolated from neighbouring h2 values that are likely associated with satellite product and parameterisation uncertainty.
The Ganges coastal region has been found to influence spring intermonsoon SST and precipitation rates in the northern BoB.
Possible positive biases in chlorophyll concentration in the Ganges river delta are likely to lead to an overestimation in the coastal biological warming, SST and precipitation rate increase. Ocean colour algorithms to determine proxy coastal 470 chlorophyll concentrations are still an area of active research (Blondeau-Patissier et al., 2014). Future studies should consider the attenuation of solar radiation from other oceanic constituents in turbid coastal regions to better represent radiant heating in the upper ocean.
CDOM is a common oceanic constituent that perturbs solar penetration depths. The derived values of h2 incorporate the bio-optical property of chlorophyll-a pigment concentration, largely excluding CDOM. CDOM increases the radiant heating 475 rate of nearshore coastal waters of North America (Chang and Dickey, 2004) and in the Arctic (Hill, 2008). Imposing an annual mean of remotely sensed CDOM absorption coefficients in a coupled ocean-atmosphere GCM reduced solar penetration depths and increased coastal SST in the Northern Hemisphere during the summer (Kim et al., 2018). CDOM concentrations are high in the western and northern coastal regions of the BoB at the mouths of major rivers (Pandi et al., 2014). Thus, including the bio-optical properties of CDOM and other biological constituents would likely increase coastal SST in the BoB, with additional 480 implications for regional climate.
The chlorophyll concentration in the BoB upper ocean is not homogeneous with depth. In situ observations show that the vertical depth of chlorophyll maxima varies between 10 and 80 m (Thushara et al., 2019;Pramanik et al., 2020), often occurring at depths undetected by satellite radiometer sensors (Huisman et al., 2006). Variations in the vertical depth of the chlorophyll maxima would vary the vertical depth of enhanced radiant heating. However, if the depth of the chlorophyll maxima occurs at a depth where solar radiation is significantly reduced (e.g., at the euphotic depth where solar radiation is ~1% of its surface value), then the change in local radiant heating at that depth would be negligible (Morel and Antione, 1994).

490
The mesoscale and sub-mesoscale spatial variability of h2 and associated oceanic processes is inadequately represented in MC-KPP due to its coarse horizontal resolution. The coastal region in MC-KPP is represented by multiple grid points that are partially ocean and partially land at an approximate 90 km horizontal resolution. Such a resolution means that at the coastlines, the mesoscale coastal chlorophyll concentration features and the corresponding solar penetration depths are poorly resolved.
Future studies should consider using a high-resolution, fully dynamical model to accurately resolve the coastline and associated 495 solar penetration depths. The simulated dynamics would improve the representation of mesoscale eddy activity along the coast and open ocean, which increases biological productivity  that in turn increases local solar radiation absorption.
The sub-seasonal temporal variability of h2 is inadequately represented in MC-KPP due to the use of a monthly mean climatological chlorophyll concentration at a reduced horizontal resolution. In reality, the advection of high surface chlorophyll 500 concentrations into the south and central BoB varies with the strength and positioning of the SLD and SMC (Vinayachandran et al., 2004), which is itself further influenced by local wind stress and seasonal Rossby waves (Webber et al., 2018). Surface chlorophyll concentrations are periodically enhanced by transient cold-core eddies and postmonsoon cyclones in the central BoB, which briefly upwell nutrients to the ocean surface (Vinayachandran and Mathew, 2003;Patra et al., 2007). In coastal regions, nutrient concentrations, which affect surface chlorophyll concentrations, vary with river discharge (Kumar et al., 2010). Suspended terrestrial sediment that perturbs solar penetration depths on the continental shelf also depend on river discharge (Kumar et al., 2010;Lotliker et al., 2016). All these factors influence solar penetration depths on timescales of days to weeks and on spatial scales of less than 1 km. By smoothing over the large subseasonal variability of chlorophyll concentration, such variations in solar penetration depth are not represented in the present study.
The limitations of representing ocean dynamics as a mean seasonal cycle means that MC-KPP cannot capture any ocean 510 dynamical response to biologically-induced changes to ocean properties (e.g., changes to ocean temperature and salinity transports). Previous studies have shown large effects of chlorophyll on ocean dynamics in the equatorial Pacific (e.g., Nakamoto et al., 2001;Murtugudde et al., 2002) and in mid-to high-latitude regions (e.g., Manizza et al., 2005;Patara et al., 2012). Modified biological warming at the surface or perhaps modified solar radiation penetration below the mixed layer could affect the dynamics of SMC and SLD in the BoB. Missing modes of variability in MetUM-GOML that rely on a dynamical 515 ocean, such as ENSO and IOD, could combine non-linearly with the ocean anomalies induced by biological warming, with implications for monsoon rainfall. Further research using a fully dynamical coupled ocean-atmosphere GCM is required to show the dynamical changes and feedbacks of biological warming on the BoB oceanic and atmospheric system.
Biological heating has complex physical and dynamical feedbacks in the ocean, which in turn imply similar feedbacks on BoB biological processes. The imposed seasonally and spatially varying h2 in MC-KPP eliminates any biological response to 520 secondary feedbacks in the ocean. A coupled biogeochemistry model linked to an ocean-atmosphere GCM would be needed to further understand secondary feedbacks on phytoplankton productivity. Secondary feedbacks may include changes to cloud cover that affect the incoming shortwave radiation needed for biological productivity; changes to thermal and salinity stratification that affect the vertical mixing of nutrients to the ocean surface; or changes to rainfall that affect river discharge and nutrient availability on the continental shelf that influence biological productivity. The resultant changes to biological 525 productivity could either enhance or deplete chlorophyll concentrations at the surface, with further implications to the spatial and temporal extent of biological heating. It is important that realistic simulations of chlorophyll concentrations are included as an additional Earth system process in high-resolution coupled ocean-atmosphere GCMs, which may improve the simulated seasonality and intraseasonal variability of the South Asian monsoon.         (c) 1.5 m air temperature (°C); (d) precipitation rate (mm day -1 ) and VIMF (kg m -1 s -1 ). The magenta line shows the 10% significance level and the black stippling shows the 5% significance level.     depth of blue light (h2), and the difference in mixed layer radiant heating rate (∆RHR) and SST (∆SST) relative to clear water, which further affects the surface latent heat flux (Qlh) and evaporation (E). The thick red and blue arrows pointing downwards in the mixed layer illustrates the preferential absorption of the shallow penetrating red light and the deep penetrating blue light. The thin blue arrow pointing downwards below the mixed layer shows the small fraction of penetrative blue light below the mixed layer. The dashed black line in the mixed layer represents h2. The three solid black arrows across the ocean surface