Bay of Bengal summer monsoon 10–20 day variability in sea surface temperature using model and observations

Authors


Abstract

[1] Observed sea surface temperature (SST) in the Bay of Bengal exhibits intraseasonal variability during summer monsoon. An ocean general circulation model (OGCM) driven by satellite derived daily winds and heat fluxes during summer of 2000 is able to reproduce aspects of intraseasonal variability (10–20 days) in SST. The intra-seasonal SST simulated using satellite derived shortwave flux compares more favorably with buoy data than that produced using the NCEP reanalysis flux. Analysis of the terms in the heat balance equation of MLD suggests that in the central Bay of Bengal local net heat flux and vertical diffusive mixing changes drive the SST ISOs. Thus, the SST ISOs in this region are governed primarily by one-dimensional processes.

1. Introduction

[2] Observed sea surface temperature (SST) in the Bay of Bengal (BOB) exhibits strong low period oscillations during summer monsoon [Sengupta and Ravichandran, 2001]. These oscillations, known as intraseasonal oscillations (ISOs) are very closely linked to the large scale atmospheric convection associated with the summer monsoon activity [Yasunari, 1979; Sikka and Gadgil, 1980; Webster et al., 1998]. It has been shown that these ISOs are influenced by the air-sea exchanges taking place at the ocean-atmosphere interface. The first ever indication of the variability of this time scale in SST and surface wind in the Bay was reported during the monsoon experiment of 1979. It was further suggested that resultant changes in air-sea fluxes influence intraseasonal variability of the monsoon atmosphere [Krishnamurti et al., 1988]. An important inference made from a recent study [Vecchi and Harrison, 2002] that Northern Bay cooling precedes monsoon breaks by about 1 week suggests of a strong relationship between breaks in the monsoon and sub-seasonal variations in SST. Hence investigation of ISO in the BOB and the role of air-sea fluxes in generating the ISO remains a subject of great interest.

[3] There have been some studies on the ISOs in the Indian Ocean (IO) using OGCMs [Schiller and Godfrey, 2003; Waliser et al., 2004]. Schiller and Godfrey [2003] found that intra-seasonal variability in the IO is controlled by one-dimensional processes. Waliser et al. [2004] show that there are a number of places where variations in entrainment and three-dimensional ocean advection are important in governing the mixed layer heat budget. Feng et al. [2000] found strong signals in horizontal advection in the western equatorial Pacific on intraseasonal scale using data from TOGA-COARE. Schiller and Godfrey [2003] suggest that satellite information can further enhance understanding of these variabilities. SST measured by the Tropical Rainfall Measuring Mission Microwave Imager (TMI) showed a low-period intraseasonal oscillation (ISO) mode of 8–16 days [Parekh et al., 2004]. More recently, Han et al. [2006] studied 10–30 day SST variability in the IO using satellite derived winds. Present work supplements some of the previous studies by including the satellite derived shortwave flux in explaining the genesis of summer SST ISOs (10–20 days) in the Bay of Bengal from OGCM simulations and in-situ observations. We have also studied how the local/nonlocal thermodynamic processes govern these ISOs in SST in this region.

2. Model and Experiments

[4] The model used in the present study is the Modular Ocean Model [Pacanowski and Griffies, 2000] version 3.1 (MOM-3) which has been set up for the global domain excluding polar regions (80°S–80°N) with variable horizontal resolution varying from 0.5 degree in the Indian Ocean to 2 degree in the rest of the oceans. There are 38 levels in the vertical with 8 levels in the upper 40 meters. The bottom topography is based on 1/12° by 1/12° resolution data from the U. S. National Geophysical Data Centre. Wind stress is computed from wind velocity using a wind dependent drag coefficient [Large and Pond, 1982]. The model was spun from rest for 20 years using climatological winds [Hellerman and Rosenstein, 1983] and restoring boundary conditions for SST and sea surface salinity. After this the model was further run with interannually varying daily wind and flux data for 1996–2000 without any restoring. The fresh water flux used was obtained using long term daily mean of precipitation and evaporation from National Centre for Environmental Prediction (NCEP) and monthly climatological river discharge data from UNESCO. Mixed layer depth (MLD) used in the diagnostics has been defined as the depth at which density is greater than or equal to that at the topmost model level (2.5 m) by 0.125 kg m−3.

[5] Two experiments differing in the nature of their forcings have been performed using the OGCM for the year 2000. The first experiment (R1) is one in which the model is driven at the surface by QuikSCAT scatterometer daily wind product from Florida State University and surface air temperature, specific humidity, net long wave and net shortwave radiation from NCEP. Latent and Sensible heat components of the net heat flux are computed using model SST. Scatterometer wind speed is validated against National Institute of Ocean Technology (NIOT) data buoy winds at three locations in the Arabian Sea and Bay of Bengal for the entire period. The buoy winds have been scaled to 10 m height. Root Mean Square Errors (RMSE) at these three locations are 1.4, 1.8 and 1.4 m s−1 respectively. In the other experiment (R2), the NCEP net short wave radiation (SWR) has been replaced by satellite derived radiation, keeping the other forcings same. Satellite-derived radiation source is of W. G. Large and S. Yeager (LY) following the earlier approach of Large and Nurser [2001], in which the NCEP surface radiation and precipitation are replaced by satellite-based estimates. A comparison of NCEP and LY SWR with SWR from surface mooring deployed by Woods Hole Oceanographic Institution at 15.5°N and 61.5°E [Weller et al., 1998] during the period 1994–1995 is carried out. The RMS errors are found to be 61 W m−2 and 37 W m−2 for NCEP and LY SWR respectively. Also a comparison of SWR obtained using an OLR based empirical relationship used by Sengupta and Ravichandran [2001] with the same buoy data resulted in RMSE of 45 W m−2.

3. Comparison of Simulated SST With Indian Ocean Buoy Data

[6] The measured SST anomaly observations [Rao and Premkumar, 1998; Premkumar et al., 1999] at the ocean buoy DS3 (location: 12.2°N and 90.8°E) and the corresponding anomalies from both the runs are shown in Figure 1. The anomalies are obtained by subtracting the mean of 123 days (starting 15 June, 2000). A strong correspondence between the SST anomaly in R2 and SST anomaly from buoy is observed at intra-seasonal (10–20 days) time scales. The ISO signals in R1 are rather weak. This point towards the significant role played by shortwave radiation in the generation of ISOs in SST and hence suggests the use of better shortwave estimates for studying such oscillations.

Figure 1.

Time-series plot of SST anomaly at DS3 location (12.2°N and 90.8°E) from R1 (solid line), R2 (dashed line), and observations (dotted line) from buoy.

4. Intraseasonal Variability of the SST and Processes Governing the Evolution

[7] It was mentioned in the previous section that the ISO in SST is better generated in R2. Two sets of strong intraseasonal events at the DS3 location, one during July–August and another during September–October can be clearly observed. The rate of change of SST denoted by dsst/dt and net heat flux (Figure 2) suggest that the SST variability at intraseasonal time scales is driven by the net heat flux in agreement with Sengupta and Ravichandran [2001]. However, in their study the shortwave component of the net heat flux was computed using an OLR based empirical relation, whereas in the present study satellite based estimate of SWR has been used. It has been mentioned [Stammer et al., 2004] that over a 17-year period the global average of LY heat flux into the ocean is only 2 W m−2 pointing to the high accuracy of the flux products used in run R2.

Figure 2.

Time evolution of model rate of change of SST (°C/day) and net heat flux.

[8] A clear lag in the peaks in dsst/dt has been observed with respect to the peaks in SWR (figure not shown). Time series of SWR showed smaller fluctuations of low period oscillations riding over a much longer period. In order to confirm this, we carried out Fourier power spectrum analysis. Analysis resulted in two clear harmonics in SST, one with 40-days cycle and another with 10–20 days cycle. Similar modes are obtained in SWR too, with the exception that Fourier spectrum of SWR shows multiple peaks in the period of 10–20 days. Further check of inter-relationships between SWR and SST was performed using the technique of lagged correlation. It was found that there is a lead time of 3-day in SWR as against SST indicating low period fluctuations in SWR to be the cause of ISOs in SST. To determine the power of each dominant mode, Morlet wavelet transform [Torrence and Compo, 1998] of SWR and SST was performed. The result suggests that low period (<20 day) oscillation in SWR is spread throughout the summer monsoon season giving rise to similar oscillations in SST. Another dominant mode is obtained with 40 days periodicity in both SST and SWR. However, power of the former mode is more than the 40 day mode.

[9] The model simulated MLD and other parameters like vertical velocity at the base of MLD (entrainment) and surface currents have been diagnosed in order to investigate their response/role in generating low period variability in SST. The entrainment rate did not show any appreciable variation during this period. Model simulated MLD (Figure 3) is around 35–40 m, but during these two ISO events, MLD became quite shallow (5–10 m). This time around the surface water at this location is normally fresh due to strong river discharge [Shetye and Gouveia, 1998]. In fact, Sengupta and Ravichandran [2001] proposed that the MLD should be of the order of 5 meters. It is remarkable that our model simulation leads to this same MLD. There was also a significant drop in the wind speed (seen in in situ observations) from 8 m s−1 to 2 m s−1 during these periods. This reduction in the wind speed drastically reduces the available kinetic energy and suppresses the turbulent mixing, resulting in shallow MLD. The u-component of surface current at this location also shows a sudden drop (Figure 3) in its magnitude due to low wind speed. The entire sequence of events can be summarized like this: Large insolation combined with low wind speed led to SST warming and strong stratification leading to shallow MLD. The warming phase is maintained as long as the winds are weak. Once, the wind speed became high again (8 m s−1), more turbulent kinetic energy was available for mixing upper layer. As a result, MLD deepened and the surface water cooled. It can be understood that the generation of oceanic ISO of SST is entirely a local effect (weak advection) primarily driven by the ISO in surface solar radiation under low wind condition. We confirm this in the next section giving a complete heat balance of the MLD and emphasizing the various mechanisms responsible for the observed SST variability.

Figure 3.

Time series of model simulated mixed layer depth (solid line) and zonal surface current (dashed line) at the DS3 location.

5. Heat Budget of the MLD

[10] A simple heat balance (rate of change of SST being proportional to net surface heat flux) discussed by Sengupta and Ravichandran [2001] during monsoon of 1998 in the Bay of Bengal was found to be valid to the first approximation. However, in order to understand the mechanism that generates the SST ISO's, we investigated various terms in the heat budget equation averaged over the mixed layer. The rate of change of temperature is given by

equation image

In the above equation, diffusive and advective terms include both horizontal and vertical components. Source term is the net heat flux, a combination of turbulent and radiative fluxes, which warms the layer. These terms (dT/dt, source and vertical diffusion added together and advective terms) averaged over the depth of mixed layer (computed using model density) on daily basis for a period of 123 days (starting 15 June 2000) for the location 12.2°N and 90.8°E are shown in Figure 4. The large heating associated with the 10–20 day ISOs is mainly due to the local heating (net heat flux) and vertical diffusion. These two terms (added together) dominate the heat budget in the mixed layer. It can be seen from Figure 4 that the horizontal advection is quite small. Other terms like, horizontal diffusion and vertical advection are also small in magnitude. This essentially points toward the one-dimensional processes at work during such events in this region.

Figure 4.

Model mixed layer depth temperature tendency (solid line) at the location 12.2°N and 90.8°E. The source term and the vertical diffusion terms added together (dashed line) and the horizontal advection (dotted line) are also shown for the same location.

6. Concluding Remarks

[11] The aim of this study was to examine the role of atmospheric forcing in generating low period oceanic ISO (10–20 days) using an OGCM in the Bay of Bengal. Another objective was to obtain a comprehensive view of the oceanic mechanisms underlying this variability. We have shown that the origin of the low period ISO of SST in the Bay of Bengal is largely a one-dimensional process induced by diffusive vertical mixing of the positive net heat gain at the ocean surface under low wind condition. The findings of this study suggest the usage of satellite-derived flux in forcing the model for understanding the low period variability. Results of the present study elucidate the coupled (ocean-atmosphere) nature of the ISO problem and suggest treatment of such phenomenon using a coupled model.

Acknowledgments

[12] Authors are extremely grateful to the Director, Space Applications Centre (SAC) and Deputy Director, Remote Sensing and Image Processing Area, SAC, for their encouragement. MOM-3 code has been acquired from Geophysical Fluid Dynamics Laboratory (NOAA). R. A. Weller of Woods Hole Oceanographic Institute is thankfully acknowledged for providing the central Arabian Sea Shortwave radiation data from buoy. Authors thank Debasis Sengupta from CAOS, Indian Institute of Science (Bangalore), and C.M. Kishtawal from SAC for many useful discussions. They also thank Chris Reason and two anonymous reviewers for their constructive comments and suggestions.

Ancillary