Spatial and temporal variability of solar penetration depths in the Bay of Bengal and its impact on SST during the summer monsoon

. Chlorophyll influences regional climate through its effect on solar radiation absorption and thus sea surface temperature (SST). In the Bay of Bengal, the effect of chlorophyll on SST has been demonstrated to have a significant impact on the Indian summer (southwest) monsoon. However, little is known about the drivers and impacts of chlorophyll variability 15 in the Bay of Bengal during the southwest monsoon. Here we use observations of downwelling irradiance measured by an ocean glider and three profiling floats to determine the spatial and temporal variability of solar absorption across the southern Bay of Bengal during the 2016 summer monsoon. A two-band exponential solar absorption scheme is fitted to vertical profiles of photosynthetically active radiation to determine the effective scale depth of blue light. Scale depths of blue light are found to vary from 12 m during the highest (0.3–0.5 mg m -3 ) mixed layer chlorophyll concentrations, to over 25 m when the mixed 20 layer chlorophyll concentrations are below 0.1 mg m -3 . The Southwest Monsoon Current and coastal regions of the Bay of Bengal are observed to have higher mixed layer chlorophyll concentrations and shallower solar penetration depths than other regions of the southern Bay of Bengal. Substantial sub-daily variability in solar radiadion absorption is observed, which highlights the importance of near-surface ocean processes in modulating mixed layer chlorophyll. Simulations using a one-dimensional K-profile parameterisation ocean mixed layer model with observed surface forcing from July 2016 show that a 25 0.3 mg m -3 increase in chlorophyll concentration increases sea surface temperature by 0.35 ° C in one month with SST differences growing rapidly during calm and sunny conditions. This has the potential to influence monsoon rainfall around the Bay of Bengal and its intraseasonal variability.

Simon Laplace climate model (IPSL; Mignot et al., 2013) uses a 3-band model where light is split into red, green and blue wavebands that each have a chlorophyll-dependent attenuation coefficient (Lengaigne et al., 2007). Both GCMs have the capability to assimilate satellite ocean colour measurements. Ocean colour measurements have revolutionized our understanding of how chlorophyll-induced heating affects ocean dynamics and the climate system (Murtugudde et al., 2002; 70 Sweeney et al., 2005;Wetzel et al., 2006). However, the limited spatial and temporal resolution of these GCMs and assimilated ocean colour data mean they inadequately resolve mesoscale and sub-seasonal chlorophyll concentration variability, which might influence the intraseasonal variability of BoB SST and the South Asian summer monsoon.
Across the southern BoB, the seasonal reversal of wind direction during the boreal summer creates conditions conducive for chlorophyll blooms. Southwesterly monsoon winds initiate the Southwest Monsoon Current (SMC), which flows 75 northeastward, advecting cooler, saline water from the Arabian Sea and the western equatorial Indian Ocean around the southernmost point of India and Sri Lanka into the warmer and fresher BoB (Fig. 1b;Jensen, 2003;Sanchez-Franks et al., 2019). The SMC evolves into a shallow, narrow and fast-moving current with surface speeds of up to 0.6 m s -1 and a thickness of up to 550 m (Webber et al., 2018). Large chlorophyll blooms along the southwestern coastal shelf of India, initiated by upwelling nutrients, become entrained in the SMC and are advected around the south of Sri Lanka into the central BoB in 80 summer (Lévy et al., 2007). A tongue of high surface chlorophyll concentrations extends into the central BoB, following the path of the SMC (Fig. 1a). The bloom is sustained east of Sri Lanka in the cyclonic eddy of the Sri Lanka Dome (SLD), identified as a region of lower absolute dynamic topography and cyclonic current vectors in Fig. 1b. Open-ocean Ekman upwelling in the SLD brings nutrients to the near-surface to support the phytoplankton population (Vinayachandran and Yamagata, 1998;Vinayachandran et al., 2004;Thushara et al., 2019). Hence, the high surface chlorophyll concentrations 85 associated with the SMC and SLD are expected to lead to reduced solar penetration depths throughout the summer monsoon period.
The large freshwater flux from river output and rainfall in the BoB during boreal summer creates a barrier layer, where strong salinity stratification forms within the isothermal layer and below the mixed layer (Vinayachandran et al., 2002;Sengupta et al., 2016). The presence of the barrier layer isolates the mixed layer above from cooling by entrainment (Duncan 90 and Han, 2009). Instead, the surface heat flux forcing, such as shortwave radiation and turbulent heat fluxes, primarily controls the warming and cooling phases of the surface ocean (Li et al., 2017). The barrier layer has been found to influence BoB SST (Drushka et al., 2014) and its thickness impacts the summer monsoon intraseasonal oscillation (Li et al., 2017). The additional effects of localised biological heating from surface chlorophyll could amplify the warming in these shallow mixed layers.
Understanding the mesoscale and sub-seasonal solar penetration depth variability and its impact on BoB surface ocean 95 properties would highlight the direct effect of chlorophyll concentration at finer spatial and temporal scales.
In this study, we determine (i) how solar penetration depth varies temporally and spatially across the southern BoB; (ii) how near-surface chlorophyll concentrations affect solar penetration depths; (iii) how chlorophyll concentration directly impacts on SST in the southern BoB. To quantify the influence of chlorophyll on solar penetration depth and SST, an ocean glider and three profiling floats were deployed as part of the joint India-UK Bay of Bengal Boundary Layer Experiment 100 https://doi.org/10.5194/os-2020-125 Preprint. Discussion started: 8 January 2021 c Author(s) 2021. CC BY 4.0 License.
(BoBBLE; Vinayachandran et al., 2018) to measure in situ physical, optical and biogeochemical variables in the upper ocean during July 2016 at high horizontal and temporal resolution. We fit a two-band solar absorption function to observed vertical profiles of photosynthetically active radiation (PAR). PAR is an integral of downwelling irradiance between 400 to 700 nm (blue to red light), allowing us to determine the solar penetration depth.
An overview of the data and methods is presented in Section 2. Section 3 presents an analysis of the temporal and spatial variability of determined h2 (Section 3.1), and a comparison of determined h2 and observed chlorophyll concentration to two previously published parameterisations (Section 3.2). This is then followed by an analysis of five idealised simulations with an imposed solar penetration depth from the h2 observations to investigate the impact of observed chlorophyll on upper ocean radiant heating rate and SST in the southern BoB. The simulations were conducted using the one-dimensional K-profile parameterisation ocean mixed layer model (Section 3.3). Section 4 presents the discussion and conclusions.

a. Ocean gliders and Argo profiling floats
During the BoBBLE field campaign (Vinayachandran et al., 2018), a Seaglider (SG579) was deployed at 86° E on 30 June 2016 along a transect at 8° N east of Sri Lanka and piloted to 85.3° E by 8 July, where the glider continued to take measurements 115 until 29 July 2016. The glider profiled on a sawtooth trajectory from the surface to 700-1000 m, completing a full dive cycle approximately every 4 hours. The glider was equipped with a Seabird Electronics (SBE) conductivity (salinity), temperature and depth (CTD) sensor, a Wetlabs Triplet Ecopuck measuring chlorophyll-a fluorescence and optical backscatter at wavelengths 470 nm and 700 nm and a Biospherical Instruments quantum scalar irradiance PAR (µE m -2 s -1 ) sensor measuring visible wavelengths between 400 nm and 700 nm. The Wetlabs and PAR sensors sampled to a depth of 300 m with a vertical 120 resolution of ~1 m. Quality control was performed on the entire conductivity-temperature (CT) dataset using Conservative Temperature-Absolute Salinity (IOC et al., 2010) space analysis and further quality control in depth space for individual vertical profiles. Salinity spikes were removed when the glider vertical speed was less than 0.035 m s -1 as the unpumped CT sensor relied on a suitable flow of water for reliable measurements. The ocean glider PAR measurements were factory calibrated. The CT sensor was factory calibrated and was then further calibrated against in situ ship CTD observations.

125
Argo profiling floats 629, 631 and 630 that are part of the international Argo float program were deployed along the 8° N transect at 85.5° E, 87° E and 89° E on the 28 June, 1 July and 4 July respectively, where they sampled to 500 m daily until mid-August and every other day until the end of September. All three floats were equipped with SBE 41N CTD and a Satlantic OCR-504 ICSW radiometer measuring downwelling irradiance at wavelengths 380 nm, 490 nm, 555 nm (µW cm -2 nm -1 ) and PAR (µE m -2 s -1 ). The CTD measurements were factory calibrated and radiometer measurements were factory calibrated with channel-specific coefficients. The vertical resolution on the ascent to the surface was ~1 m for the radiometer and CTD.

b. PAR
All night-time PAR profiles (local solar zenith angles greater than 70°) and low-light PAR profiles (maximum PAR in the top 5 m is less than 100 E m -2 s -1 ) measured by the glider and profiling floats were discarded. Further quality control was 135 carried out to remove the effect of external environmental factors, which include shading by passing clouds that causes sudden fluctuations in light intensity, and wave-focusing that creates sawtooth spiking in the vertical PAR profiles (Zaneveld et al., 2001). All PAR data between 0-5 m depth were removed from the analysis, as noise from wave-focusing obscured the signal of the absorption of solar radiation. A quality-control method using a fourth-degree polynomial, modified from the methodology of Organelli et al. (2016), was used to identify PAR perturbations below the near-surface and remove profiles 140 displaying excessive noise.

c. Chlorophyll
The glider's raw fluorescence voltages were converted into chlorophyll-a concentrations according to the manufacturer calibrations. Since phytoplankton that are exposed to too much sunlight trigger the non-photochemical quenching mechanism 145 to protect themselves from photooxidative damage (Müller et al., 2001), chlorophyll-a fluorescence is suppressed near the surface in the daytime. To correct for quenching, nighttime fluorescence-to-backscatter ratios were used to derive corrected daytime chlorophyll-a fluorescence profiles (Thomalla et al., 2018). The glider fluorescence-derived chlorophyll-a concentrations, after correcting for non-photochemical quenching, showed values that were higher than those derived from the shipboard CTD chlorophyll-a fluorescence sensor. Concentrations were calibrated by applying a scale factor and offset derived 150 using linear regression between the glider and CTD chlorophyll-a profiles.
The profiling floats did not make chlorophyll-a fluorescence measurements, so a novel approach was developed to derive chlorophyll-a concentration from radiometer data alone (see Appendix A for method details). Chlorophyll-a strongly absorbs light at 490 nm wavelength so the vertical gradient of Ed(490), the downwelling radiation flux at 490 nm, was used to derive a proxy for in situ chlorophyll-a concentration to identify the vertical distribution of chlorophyll-a. Vertical profiles of the 155 natural log of Ed(490) were individually corrected for their mean in situ dark count calculated from measurements below 200 m (Organelli et al., 2016). Profiles displaying excessive noise were eliminated using the fourth-degree polynomial method of Organelli et al. (2016). The attenuation coefficient Kd was calculated for each 1 m discretised layer. The attenuation coefficient Kd is the sum of the attenuation of pure seawater (Kw), represented as a constant, and the attenuation due to biology (Kbio), a chlorophyll-a component (Morel and Maritorena, 2001;Xing et al., 2011). Further quality control is applied to vertical profiles 160 of Kbio before chlorophyll-a is calculated using empirically determined coefficients from Morel et al. (2007) (Fig. A1). The chlorophyll-a pigment concentration that was derived from radiometry data or remotely sensed by satellite will henceforth be referred to as "chlorophyll" for convenience.

d. Satellite products
The remotely sensed chlorophyll concentrations used in this study are sourced from the European Space Agency's Ocean Colour -Climate Change Initiative (ESA OC-CCI; Lavender et al., 2015) version 3.1 (available at http://www.esaoceancolour-cci.org). The OC-CCI project involved the merging of remotely sensed chlorophyll concentrations from MODIS,

170
MERIS, SeaWiFS and VIIRS radiance sensors to provide a continuous dataset ranging from 1997-2016 with increased spatial coverage of the global oceans. The radiance sensors on the satellites detect the water-leaving radiance at specific wavelengths to estimate chlorophyll concentration. The thickness of ocean surface layer "seen" by the radiance sensors is approximately one solar penetration depth or the depth where downwelling irradiance decreases to 1/e of the surface irradiance (Gordon and McCluney, 1975), depending on the local chlorophyll concentrations. 8-daily and monthly composites of chlorophyll 175 concentration with spatial resolutions of 4 km have been used to investigate the weekly and monthly variability of chlorophyll concentration influencing solar penetration depths in the deployment region across the southern BoB from July to September

2016.
Satellite-derived absolute geostrophic velocities (meridional and zonal components) and absolute dynamic topography are altimeter products produced by SSALTO/Duacs, distributed by AVISO (https://www.aviso.altimetry.fr) and are available 180 through the Copernicus Marine Environment Monitoring Service (http://marine.copernicus.eu). The daily composites of absolute geostrophic velocities and absolute dynamic topography have a spatial resolution of 0.25° x 0.25° and are used to investigate the surface current velocities that control chlorophyll concentration advection from July to September 2016.

185
The penetration depth of solar radiation is frequency dependent, with higher frequencies ("blue" light) having a much larger penetration depth than lower frequencies ("red" light). A two-wavelength double-exponential function (Paulson and Simpson, 1977) can be used to parameterise this behaviour in ocean models (e.g., Sweeney et al., 2005) and in mixed-layer heat budget studies (e.g., Vialard et al., 2008;Girishkumar et al., 2017). For this study we use a double-exponential function of the form: where Q (W m -2 ) is the irradiance at depth z (m). Surface irradiance just below the ocean surface for blue light is denoted by q2 (W m -2 ). The scale depths, h1 and h2, represent the e-folding depths (m) of absorption of red and blue light respectively. The parameter R is the ratio of the flux of red light to the total and is a measure of the partition of the solar flux into the arbitrary red and blue bands. An offset d (W m -2 ) has also been introduced to allow for a non-zero instrument response at zero radiation flux, when applied to a radiometer. In practice, d is very small (compared with q2).
195 Paulson and Simpson (1977) determined the optical parameters (R, h1 and h2) for each of the five Jerlov water types, which represent the range of turbidity observed in open ocean water (Jerlov, 1968). Optical parameter R is 0.58 in water type I and 0.78 in water type III. The two-band model is an arbitrary approximation of the full solar spectrum, and there is no a priori definition of the value of the cut-off frequency between the red and blue bands. Hence, the parameter R is allowed to vary, along with h2, to maximise the fit of the two-band model to the data. The variation of R should not be interpreted as a physical change in the fraction of red light at the surface, which of course is 205 independent of the ocean conditions below. Instead, the variation of R should be interpreted as a degree of freedom in fitting a simple two-band scheme to model the full solar spectrum.
In this study we use Eq.
(1) to fit to profiles of PAR reported in terms of moles of photons ( E m -2 s -1 ) instead of units of energy (W m -2 ). For PAR measurements, the conversion of units from E m -2 s -1 to W m -2 can only be an approximation as the PAR instrument measures photons across a range of visible wavelengths, but the exact spectrum across that range is 210 unknown at any particular time (Sager and McFarlane, 1997). Although the absolute values of PAR change with unit conversion, the attenuation rate of visible light with depth and thus the value of h2 is independent of the unit conversion of PAR. Hence, in practice we fit Eq.
(1) to profiles of PAR with units of E m -2 s -1 to determine values of h2 and avoid PAR conversion uncertainty.
From the excessively noisy 1-m vertical resolution PAR measurements close to the surface we are unable to determine the 215 transmission of red light (values of R and h1). We assume Jerlov water type IB (Paulson and Simpson, 1977) to be applicable to our region, based upon initial determinations of h2 ~17 m from fitting Eq. (1). We therefore constrain R to be 0.67 and h1 to be 1 m and thus fit PAR profiles between 5 m and 100 m to the transmission of blue light with depth (h2) using Eq. (1) (Fig.   2a). The same fit plotted in log space ( Fig. 2b) results in a near straight line below 5 m, demonstrating that the decrease in PAR can be approximately represented with a single exponential below this depth. The contribution of the fixed parameters 220 used for the fit was estimated by varying R and h1 between Jerlov water type I to III from Paulson and Simpson (1977) and varying the depth of removed near-surface PAR between 3-7 m. We combine uncertainties associated with the varied parameters and associated with fitting, to produce the overall uncertainty in the derived values of h2.

Glider and profiling float observations
The SLD is a prominent feature in the southwest BoB during the summer monsoon and is typically associated with high surface chlorophyll concentrations (Thushara et al., 2019). At the start of July 2016, the SLD is centred around 85-86° E and 5-10° N to the west of the SMC (Fig. 1b). Glider SG579 is located inside the SLD from 30 June and observes the weakening of this cyclonic eddy after 2 July, remaining in a localised region between 85-86° E ( Fig. 1c; black diamond). The average 230 mixed layer salinity and temperature are 34 g kg -1 and 28°C respectively ( Fig. 3a and 3b). Chlorophyll concentrations peak on values of h2 decrease from an average of 16 m on 30 June to 13 m on 1 July, as the average 0-30 m chlorophyll concentration increases from 0.2 mg m -3 to 0.5 mg m -3 in one day ( Fig. 3d; black circles).
After 2 July, the SLD weakens and shifts towards the northwest, but the SMC continues to flow into the south-central BoB.

235
Patches of surface chlorophyll, with concentrations of 0.1-0.4 mg m -3 (Fig. 1d), continue to be advected by the SMC into the glider SG579 deployment region (85-86° E) until 19 July. Within the SMC, the mixed layer warms to 29°C and freshens to 33.3 g kg -1 ( Fig. 3a and 3b). Chlorophyll concentrations below the mixed layer remain around 0.5 mg m -3 forming a deep chlorophyll maximum between 30-50 m depth (Fig. 3d). Meanwhile average 0-30 m chlorophyll concentrations decrease to less than 0.2 mg m -3 (Fig. 3d) and the corresponding average values of h2 increase to more than 20 m until 16 July ( Fig. 4a; 240 dashed black line). The position and velocity of the SMC relative to the biologically productive southern coast of Sri Lanka determines how much surface chlorophyll is entrained and advected into the south-central BoB (Vinayachandran et al., 2004).
Throughout most of July the SMC is too far south to intercept the high surface chlorophyll concentrations along the southern coast of Sri Lanka (Fig. 1d), explaining why in situ surface chlorophyll concentrations are relatively low after 2 July (Fig. 3d).
The variability of h2 in the SMC is large (Fig. 4a). Values ranged between 15-31 m from 4 July onwards, which we partly 245 attribute to sub-daily temporal variability in the mixed layer and surface chlorophyll concentrations. However, the derived h2 values from glider SG579 are associated with relatively high uncertainty (typically ±2 m) due to the fitting of the double exponential function to noisy vertical PAR profiles, which may contribute to this apparent variability.
The profiling float dataset allows us to extend the glider dataset temporally and spatially, providing daily measurements of solar penetration depths until mid-August and then measurements every 2 days until the end of September, spanning much of 250 the southern BoB. The vertical profiles of downwelling irradiance measured from the profiling floats are less noisy than those measured from the glider. Hence, the profiling floats display lower uncertainty in determined values of h2 when compared with the glider ( Fig. 4; a-d). As the SMC flows northeastward into the south-central BoB during early July, the surface current bifurcates. The main branch flows northward towards 10° N and the smaller branch flows eastward towards 90° E (Fig. 1d). Daily variations in salinity of 0.2 g kg -1 are observed by float 630 during 6-12 July, with the highest salinity recorded at 34.4 g kg -1 in the mixed layer and the barrier layer on 10 July (Fig. 8b), possibly due to eddies shearing off from the main SMC flow (Fig. 1d). September, the SMC influence at 89° E reduces and the current shifts to the western side of the basin (Fig. 1f), consistent with climatological observations (Webber et al., 2018). Consequently, at 89° E a southeastward flow containing water from the eastern side of the basin along with some recirculated surface water from the SMC is observed (Fig. 1e and 1f). Float 631 yields h2 values greater than 20 m (Fig. 6d), possibly indicating that the southeastward flow advects low surface chlorophyll concentrations from the biologically unproductive eastern side of the BoB. We hypothesise that the displacement of the SMC 285 to the western BoB would lead to reduced solar penetration depth in the west and increased solar penetration depth in the east during the summer.

Relationship between scale depth and chlorophyll concentration
Visible radiation in the upper ocean decreases by approximately 63% (1 -e -1 ) from the surface to a depth equal to one scale 290 depth. Glider observations show that over 80% of PAR is absorbed to a depth of 30 m (Fig. 3c). The chlorophyll concentration of the surface layer, where the majority of visible radiation is absorbed, is a key control on the amount of visible radiation absorbed and thus on the radiant heating rate of the surface layer. We examine the relationship between the average chlorophyll concentration in the surface layer and h2, both observed by the glider. The average mixed layer depth in the glider time series  (490)).
As expected, h2 is inversely related to chlorophyll concentration (Fig. 9). Observed average chlorophyll concentrations The observations compare well with two commonly used double exponential parameterisations in ocean GCMs relating light absorption to chlorophyll concentration ( Fig. 9; Table 1), from Morel and Antoine (1994) [MA94] and Ohlmann (2003) [O03]. We assume for the O03 two-band solar absorption scheme that the incident angle of solar radiation on the ocean surface

Implications of chlorophyll concentration on BoB SST
The determined values of h2 for each glider and float timeseries varies by a factor of two ( Fig. 4; where we specify H = 30 m to represent the average mixed layer depth from the glider, = 1021 kg m -3 to represent the average density of seawater in the upper 30 m from the glider dataset and cp = 3850 J kg -1 K -1 to represent the specific heat capacity of sea water. The daily average solar irradiance absorbed in this mixed layer is calculated by taking the difference between the daily average solar irradiance incident on the ocean surface, 01 (0), and daily average solar irradiance at the base of the mixed 325 layer, 01 (H). At depths greater than 5 m, we assume all red light is absorbed and 01 (H) is then the blue light radiation flux that penetrates the base of the mixed layer.
The daily average solar irradiance incident on the column surface is estimated to be 280 W m -2 based on solar irradiance measurements during clear sky conditions during the observation period (Vinayachandran et al., 2018). For the purposes of this calculation, we ignore the effects of advection, entrainment and mixing, as well as any atmospheric feedbacks from The BoBBLE campaign took place during a suppressed period of convection or a break phase in the South Asian monsoon.
The South Asian monsoon is subject to active-break cycles on subseasonal timescales (10 to 30 days) driven by the Boreal Summer Intraseasonal Oscillation (BSISO; Wang and Xie, 1997), which are strongly influenced by air-sea interactions (Sengupta et al., 2001). Associated with this break phase, no precipitation is recorded, and solar shortwave flux remains high 365 during the campaign between 4-15 July ( Fig. 10b and 10c), allowing for strong diurnal heating of the ocean surface during this period. By 15 July, precipitation increases (Fig. 10c) as deep atmospheric convection enters the campaign region marking the transition into an active phase of the BSISO.
The KPP experiments demonstrate that changing h2 from 26 m to 14 m leads to an increase in daily average SST of 0.35°C by the end of July 2016 (black line; Fig. 10e). The average mixed layer depth is 34 m and remains relatively constant during July. Hence, the previous idealised calculation was a good approximation as we estimated a similar amount of radiant heating for a mixed layer of comparable thickness. Decreasing h2 from 26 m to 17 m, 19 m and 21 m, leads to progressively smaller increases in daily average SST from 0.25°C, 0.18°C and 0.14°C by the end of July 2016, respectively (Fig. 10e). The maximum diurnal change in SST for the h2 = 14 m simulation is 1.0°C, compared with 0.62°C for the h2 = 26 m simulation (Fig. 10d).
From 1-15 July the SST from the h2 = 14 m simulation warms at the greatest rate of 0.04°C day -1 , compared with 0.02°C day -375 1 for the h2 = 26 m simulation (Fig. 10d). From 15 July onwards, during an active phase of the BSISO, SST warming for the h2 = 14 m simulation is just 0.01°C day -1 , compared with the slight SST cooling in the h2 = 26 m simulation (Fig. 10d).
Decreased solar penetration depth leads to increased absorption of solar radiation over a shallower depth of ocean. Hence, the mixed layer warms and the water below the mixed layer cools as less solar radiation penetrates deeper in the water column ( Fig. 10h).

380
On the 25 July, high precipitation rates of 4 mm day -1 freshen the ocean surface (Fig. 10g), which contributes to an increase in mixed layer salinity stratification and a reduction in the maximum mixed layer depth in all five simulations (Fig. 10f). A reduction in wind stress also partly contributes to the reduction in the maximum mixed layer depth, as wind-driven turbulent mixing is reduced (Fig. 10c). The mixed layer in the h2 = 26 m simulation shoals to a maximum depth of 30 m and recovers to a previous depth of 34 m a day later (Fig. 10f). Conversely, the mixed layer in the h2 = 14 m simulation shoals to a maximum 385 depth of 23 m and recovers to a previous depth of 34 m five days later (Fig. 10f). Decreased solar penetration depth and increased solar radiation absorption further increase mixed layer thermal stratification and stability, which amplifies and prolongs the vertical and temporal change in mixed layer depth. Shoaling the mixed layer to a depth comparable to the solar penetration depth increases the sensitivity of SST to changes in chlorophyll concentration (Turner et al., 2012;Giddings et al., 2020). Hence, freshwater input through precipitation and additional biological warming through the occurrence of high chlorophyll concentrations in the SMC and SLD region would enhance SST increase during an active BSISO phase, which would potentially have a positive impact on atmospheric convection. The relationship between determined h2 and observed chlorophyll concentrations measured from the glider has limitations.

Discussion and clonclusions
Determined h2 not only represent the attenuation of blue light due to chlorophyll-a concentration, but also the attenuation of blue light due to other biological constituents and other suspended particles. Furthermore, the observed chlorophyll 420 concentration is only a proxy for actual chlorophyll-a concentration. Hence, determined h2 values potentially overestimate blue light attenuation due to chlorophyll-a pigments, affecting the relationship between determined h2 and average observed chlorophyll concentration and the fit of MA94 and O03 shown in Section 3.2. Future climate modelling studies should consider different types and concentrations of biological constituents that affect h2, such as coloured dissolved organic matter (e.g., Kim et al., 2018).

425
Relatively low blue-light scale depths are likely to occur within the SMC and SLD due to the higher surface chlorophyll concentrations that will in turn lead to locally enhanced warming. The width of the SMC is approximately 300 km (Webber et al., 2018) and surface chlorophyll begins to increase in April and typically peaks in July (Lévy et al., 2007) resulting in a considerable area and duration of enhanced biological surface warming. Likewise, the eastern and western BoB coastal regions also display smaller solar penetration depths, further widening the region impacted by biological surface warming.

430
The additional biological warming is likely to be non-uniform across the basin and subject to variability during the summer to the break periods of the BSISO is an additional factor to consider when modelling intraseasonal convective events.

Appendix A: Determining in-situ chlorophyll-a concentration from downwelling irradiance
This Appendix provides a description of the method, key assumptions and quality control process used to derive a proxy for in situ chlorophyll-a concentration from downwelling irradiance at 490 nm measured by the Argo profiling floats.

465
Downward irradiance, Ed, at wavelength, l, decays approximately exponentially as it penetrates through the water column.
The irradiance just below the surface, Ed(l, -0), decays with depth, z, at each discretized layer, dz, from the surface, 0, to z.
The rate of decay of irradiance, defined as the diffuse attenuation coefficient, Kd(l, z), is allowed to vary in each discretized layer of 1 m thickness. The function is given as: where Kd(l, z) is defined as the sum of the attenuation of pure seawater (Kw) and the attenuation due to biological material (Kbio). For each discretised layer, Kw is assumed to remain constant, but Kbio is allowed to vary in order to derive depth-varying chlorophyll-a concentration profiles. Kbio varies as a non-linear power law function of chlorophyll-a concentration, [Chl-a] 475 (Morel, 1988;Morel and Maritorena, 2001) meaning Kd(l) is defined as: where c(l) and e(l) are empirically determined coefficients. Eq. (A1) then becomes, The relationship between light absorption and chlorophyll-a concentration is assumed to be constant and open ocean water in the southern BoB is assumed to be categorised as "Case 1" waters, where optical properties are affected by chlorophyll 495 pigments and detrital organic matter (Morel, 1988). The BoB surface ocean mainly consists of chlorophyll-a pigments, as shown from in situ water samples (Madhu et al., 2006), chlorophyll-a fluorescence measurements and remotely sensed satellite measurements (Thushara et al., 2019). Hence, Eq. (A3) and the empirically determined coefficients are suitable to determine chlorophyll-a concentration profiles from Ed(490) measured by the floats.

Data Availability
The satellite chlorophyll-a products were produced by the Ocean Colour project European Space Agency Ocean Colour -