Seasonal to intraseasonal variability of the upper ocean mixed layer in the Gulf of Oman

,


58
The circulation, upper ocean stratification, and associated biogeochemical cycles in the 59 The aim of this study is to describe the evolution of MLD and stratification in the GoO. 141 We use high-resolution underwater glider data, collected over both winter and spring,  The latent heat flux (QL) can be estimated from wind speed and air-sea humidity 223 differences using the following bulk parameterization (Yu, 2009;Kumar et al., 2017): where Le is the latent heat of vaporization and is a function of sea surface temperature,

236
The MLD is defined using the threshold method with a finite difference criterion for where 0(z) is the potential density at depth z and 0(10 m) is the potential density at 10 241 m depth (Suga et al., 2004    where bx is defined as in Equation 8 and y is the along-front component of wind stress.

311
Wind stress is temporally collocated to the gridded glider data, and following du Plessis  However, it may also be due to spatial differences in regional dynamics such as waters before the spring restratification (∑z N 2 , Figures 4a-b). N 2 presents periodic 419 peaks when the glider transits over the shelf (Figures 4a-b). There is a noticeable to intensify throughout spring (Figures 4a-b). Evaporation dominates over precipitation 430 (Figures 4c-d), which is enhanced by intense wind episodes throughout winter and present a distribution that is higher than the climatological yearly mean (Figure 5b).

542
However, these winds have a higher humidity rate and therefore do not cause significant given little evidence for temporally-induced variability. Conversely, during spring, the 567 stronger diurnal cycle signal and shallow MLD provoque a temporally-dominated horizontal buoyancy gradients (∂b/dt >> ∂b/∂x). Seasonally, the weakest bx are found in 569 the winter ML with an enhancement occurring in spring ML (Figures 3i-j). Expectedly, 570 bx are amplified at the base of the ML, which is likely an artifact of internal wave 571 processes vertically displacing the pycnocline and the glider sampling the pycnocline at 572 somewhat variable depths over space and time (Figures 3i-j). The ML distribution of the 573 bx indicates an overall seasonality (Figure 6a). The upper limit of the winter bx 574 distribution is lower than the spring one, even after accounting for the glider sampling 575 underestimation of 69% (see Section 2.4) (Figure 6a). Only 16% of the horizontal 576 buoyancy gradients exceed 10 -7 s -2 during winter, compared to 38% of the profiles 577 during spring (Figure 6a).

579
The horizontal Turner angle (Tu) is computed to quantify the relative effect of temporal 580 and spatial variations (horizontal gradients) of ML temperature and salinity on the 581 horizontal density (buoyancy) gradients (see Section 2.4, Figure 6b). The distribution of 582 Tu determines that horizontal temperature gradients have a major impact on density 583 fronts than salinity gradients (distributions shifted to ± 2 ). We observe more frequent and    would suggest a lesser impact of QSMS on surface ML. As QSMS are mostly restratifying the upper ocean in this region, our hypothesis would align with many of the regional

Introduction
The supporting information contains three additional figures. We provide a supporting figure to validate the glider data management ( Figure S1). Seaglider 579 was deployed in March 2015 until the end of May 2015 (91 days) during the spring intermonsoon and Seaglider 510 was deployed in mid-December 2015 and recovered at the end of March 2016 (108 days) during the winter NW monsoon. The data shows the bias between the up and downcast of the corrected glider data profiles for each season. There is an evident deviation during both seasons in the measurements at the first meters of the downcast profiles, more prominent during spring. The temperature bias is caused by the warming of the sensors during the communication phase at the surface between dives. Strong solar radiation warmed the glider and its sensors, causing an artificial rise in potential temperature. The bias in the downcast profiles produces fictitious results when observing lateral gradients, hence only climb profiles are used in this study. We provide a figure to show the little spatial variability of the atmospheric variables (ERA5 products) ( Figure S2) and a figure to validate the election of the ERA5 reanalysis product over TropFlux ( Figure S3).  Freshwater flux, (E-P, black line) for winter (e) and spring (f). In red shading the standard deviation due to averaging the four ERA5 grid cells collocated on the glider transect. Notice that the standard deviation in the wind speed is very small and therefore not visible in panels c and d. Figure S3. Comparison between ERA5 and Tropflux atmospheric forcing. ERA5 variables have been resampled to a daily resolution to compare to Tropflux products. Top panels compare the timeseries of net heat flux (Qnet) and wind speed (U) for ERA5 (red) and Tropflux (blue) for winter (fist column) and spring (second column). The shading shows the standard deviation. Daily mean biases between the products are (63 ± 48) W·m -2 in Qnet and (0.6 ± 1.3) m·s -1 in U. Bottom panels compare the ERA5 vs. Tropflux for Qnet (left) and U (right) for winter (black) and spring (orange). The error bars mark the standard deviation for each value. The dotted line shows the 1 to 1 relation between data sources. Correlation values (r 2 ) for all the data are displayed in the bottom panels.