State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Hangzhou, China
Corresponding author: G. Wang, State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, 36 Bao Shu Bei Rd., Hangzhou, Zhejiang 310012, China. (email@example.com)
 A winter warm pool off the southwest coast of Hainan Island is uncovered from high resolution satellite measurements and field observations. The warm pool is characterized by warm temperature relative to the surroundings. It forms in October, intensifies from November to next January, and decays in February. Our results show that the wind wake in the northeast winter monsoon due to the orographic blockage by mountains of Hainan Island plays an important role in generating the warm pool by reducing surface latent heat flux. The core temperature of the warm pool is correlated with the El Niño and Southern Oscillation.
 Hainan Island, with an area of about 33,920 km2, is located in the northwest South China Sea (SCS) (Figure 1a). The Qiongzhou Strait separates the island from Mainland China. To the west lies the Beibu Gulf, which is a semi-enclosed shallow bay (Figure 1a). Over the Gulf, the wind is southwesterly in summer and northeasterly in winter. Although the wind direction reverses seasonally, the ocean circulation in the Gulf is generally cyclonic in both summer and winter [Su and Yuan, 2005; Wu et al., 2008].
 Sea surface temperature (SST) is an important variable for studying the ocean dynamics, air-sea interactions, regional climate and biological processes. SST around Hainan Island has been documented in previous studies [e.g., Luo et al., 2003; Tang et al., 2003; Lü et al., 2008; Su and Pohlmann, 2009; Jing et al., 2011]. SST shows large seasonal variability: in summer, the northern Gulf is generally warmer than the south because of bathymetric effect under solar radiation [Su and Yuan, 2005] and the upwelled cold water can appear in the east Gulf [Li, 1990; Chai et al., 2001]; in winter, SST is high in the deep water and low near the coast due to bathymetric effect under surface cooling. The bathymetric effect on the seasonal variability of SST is as follows: the bottom topography in the Beibu Gulf is deep in the middle and shallow in the coastal sides (Figure 1a). Under the same surface cooling (warming), the deep water cools (warms) more slowly than shallow water, hence it stays warmer (colder) in deep water than shallow water [Xie et al., 2002].
 A warm water tongue intruding into the Beibu Gulf has long been considered the main SST pattern in winter [e.g., Chen et al., 2009; Huang et al., 2008]. In this study, we re-examined the winter SST distribution using a 9 km resolution SST climatology based on satellite observations and found a warm pool (or warm patch) southwest of Hainan Island surrounded by cold water (Figure 1b). The SST pattern is very different from the previous knowledge of the SST distribution as shown in Figure 1a. The SST difference between the previous climatology and satellite observations is due to their different sampling: in situ temperature measurements in the past are rather sparse and spatiotemporally irregular while the satellites can observe high spatial and temporary variability of SST in nearly all weather conditions. The 9 km satellite SST data provides an opportunity to observe this relatively small-scale warm pool. To our knowledge, this robust warm pool has not been reported in the literatures. This present study investigates the new climatological feature of the winter warm pool and its seasonal development. We will explore the formation mechanism with a mixed layer model, and examine interannual variability of the winter warm pool.
2. Data and Methodology
2.1. Satellite Data
 Two satellite SST data sets are used here: the Remote Sensing Systems (REMSS) product with 9 km resolution from January 2006 to December 2010, and the National Climate Data Center (NCDC) product on a 0.25° grid from January 1982 to January 2011. The REMSS product are merged by both the infrared sensors from Moderate Resolution Imaging Spectroradiometer (MODIS) and microwave sensors from Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) and Tropical Rainfall Measuring Mission (TRMM) Microwave Images (TMI) [Reynolds and Smith, 1994]. The NCDC SST product is blended by the Advanced Very High Resolution Radiometer (AVHRR) and AMSR-E observations (the product is only from AVHRR before June 1, 2002).
 Two satellite wind data sets are analyzed: the monthly QuikSCAT wind data set on a 0.25° grid from January 2000 to December 2008 [Hoffman and Leidner, 2005], and the monthly blended wind product merged from the microwave radiometers on the Special Sensor Microwave Imager (SSMI) and the TRMM, AMSR-E and the Quick Scatterometer (QuikSCAT). The data is also on a 0.25° grid from January 1988 to December 2010 by NCDC [Zhang et al., 2006].
 Heat fluxes come from two products: latent heat and sensible heat fluxes from the Japanese Ocean Flux data set with Use of Remote Sensing Observations version 2 (J-OFURO2) [Tomita et al., 2010]; surface long wave radiation flux and shortwave radiation flux from the Objectively Analyzed air-sea Fluxes (OAFlux) produced by the Woods Hole Oceanographic Institution (WHOI). The J-OFURO2 data set is on a 0.25° grid while the OAFlux product on a 1.0° grid. Both cover the period from January 2002 to December 2007.
2.2. Hydrographical Data
 The monthly temperature and salinity climatology with a horizontal resolution of 0.25° × 0.25° is from U.S. Navy Generalized Digital Environment Model (GDEM-Version 3.0), derived from the temperature and salinity profiles extracted from the Master Oceanographic Observational Data Set edited at the Naval Research Laboratory (NRL) [Canes, 2009].
 Two hydrographic data sets are used to verify our findings. One is from World Ocean Database 2009 (WOD09), 89 historical temperature profiles are from a cruise (WOD Cruise Reference: SU000050) operated by the Pacific Oceanological Institute (Vladivostok) of the former Soviet Union. The cruise was from January 16, 1960 to February 9, 1960. The stations of temperature observations are shown in Figure 2a; the other is from the cruise operated by 908-01-ST09 project of China. Totally 139 CTD profiles shown in Figure 2b are obtained during the period from December 25, 2006 to January 22, 2007. The accuracy of temperature is 0.01°C and 0.005°C for WOD cruise and 908 cruise, respectively.
where Tl, Sl, Vl is the mixed layer temperature, salinity, horizontal velocity, respectively, Q is the net air-sea heat flux (sum of longwave radiation, shortwave radiation, latent heat flux and sensible heat flux), ρr is the reference density (1.025 g cm−3), Cp is the heat capacity of seawater, δT, δS, δV is the difference in T, S, V across the interface, respectively, E − P is evaporation minus precipitation, f is the Coriolis parameter, τ is wind stress vector, h is the mixed layer depth (thickness), and ∂h/∂t = −We (We is entrainment velocity). Density is calculated from a linear state equation:
where α = −3.0 × 10−4 °C−1 and β = 7.6 × 10−4 ‰−1.
 In PWP, the latent heat flux and wind stress are two independent inputs. The former belongs to the heat budget module while the latter in the mixing module [Price et al., 1986]. In our simulation, the vertical resolution is set to be 1 m in order to resolve the shallow mixed layer depths. The GDEM temperature and salinity in September are set as the initial condition. The surface forcings are winter average net heat flux, wind stress and freshwater flux from November to January. The net heat flux includes surface sensible heat flux, latent heat flux, longwave radiation flux and shortwave radiation flux. The PWP model integrated for 90 days.
Figure 1b shows the REMSS climatological winter SST averaged from November to January. The most conspicuous feature is warm water surrounded by cold water off the southwest coast of Hainan Island. The warm water covers a region from 107.5°E to 109°E in the zonal direction and from 17.5°N to 18.5°N in the meridional direction. The warm pool is defined here in a relative sense, as the region with SST warmer than the surroundings. The average temperature in the core of the relative warm pool is about 25.5°C, which is about 0.5°C–1.0°C higher than its surroundings.
 Hydrographic observations in the Beibu Gulf carried out from January 16 to February 9, 1960 by the former Soviet Union capture the relative warm pool (Figure 2a) surrounded by cold water. The SST pattern from 89 temperature stations shows that the temperature in the warm water core is about 24.5°C that year. 908 field experiment also demonstrated the relative warm pool (Figure 2b), although it was deployed only in the west coast of Beibu Gulf. Generally the SST pattern is similar between hydrographic and satellite observations. A drifter (ID: 56662) also captures the relative warm pool from January 7 to February 28, 2006. The temperature at 15 m depth observed by the drifter is warmer in than outside the warm pool, consistent with the SST patterns from satellite and hydrographic observations.
Figures 3a and 3b show the vertical structure of the warm pool in a transection for WOD cruise and 908 cruise, respectively. The transections shown in Figures 2a and 2b include 8 and 12 hydrographic stations, respectively. Both of them show the similar vertical structure for temperature: water is well mixed from the surface to bottom with little stratification in winter. Water in the warm pool is warmer than on both sides throughout the water column. The well mixed nature of the Beibu Gulf in winter is associated with tides and surface flux cooling [Hu et al., 2003; Su and Yuan, 2005].
Figure 4 shows the seasonal evolution of the relative warm pool from REMSS SST product. The warm pool begins to form in October, and comes to a mature stage in November, December and January. After that, the relative warm structure begins to decay in February. Note that the color scale in the figure varies for different months to offset the background seasonal cycle. To give a quantitative estimation, we define three 1° × 1° boxes as shown in Figure 1b to calculate the temperature difference between the relative warm pool and the surroundings. The temperature difference between the relative warm pool and water to the east is the largest (around 0.6°C) in November and December, while the difference with water on the west side reaches the maximum value (around 1.1°C) in December and January. The relative warm pool is most pronounced from November to January as measured by spatial differences in SST.
3.2. Warm Pool Formation
Figure 1b shows the winter-mean QuikSCAT wind velocity vectors and speed. Wind speed southwest of Hainan Island is small but large in other regions. The wind speed pattern is due to mountains on Hainan Island. In Figure 1b, elevations higher than 300 m are shaded in brown. The northeast winds are blocked by the mountains and slow down over the ocean to the southwest. Meanwhile, the winds are accelerated over the ocean on the northwestern and southeastern flanks. The warm pool southwest of the mountain range corresponds to the area of small wind speed. Such a good correlation in space implies that wind plays an important role in the generation of the warm pool.
Figure 5a shows the temperature distribution from the PWP model simulation forced with the observed heat flux, wind stress and freshwater flux. The control run reproduces the warm pool quite well southwest of Hainan Island over the region of small wind speed. The cold tongue east of the warm pool in the simulation is weaker than observations (Figure 1b) because the model does not include thermal advection [Liu et al., 2004].
 We carried out six additional experiments using different forcings. In the sensitivity tests, we set latent heat, sensible heat, longwave radiation, shortwave radiation, wind stress and precipitation to be an area-averaged value of the Beibu Gulf for experiment a-f, respectively. For the independent experiment a-f, the averaged values are set to 120 w/m2, 20 w/m2, 40 w/m2, 100 w/m2, −0.06 N/m2 for zonal wind and −0.05 N/m2 for meridional wind, and 4.48 × 10−8 m/s, respectively. As the latent heat flux is set to the area averaged value (120 w/m2), the spatial structure of the warm pool such as the warm water core around 108.5°E, 18.25°N cannot be reproduced (Figure 6a). The resultant SST structure is somehow similar to the SST distribution in Figure 1a, with a broad warm water tongue penetrating into the Beibu Gulf, a pattern indicative of the bathymetric effect. All the other experiments can produce the warm pool structure in space when the model includes latent heat flux but set other heat fluxes, wind stress or precipitation to be an area-averaged value (Figures 6b–6f). We conclude that the latent heat flux is very important for the spatial structure of the relative warm pool. These results suggest the following conceptual model. The northeast winds slow down over the ocean southwest of Hainan due to the mountain blockage. The low wind speed reduces latent heat flux (Figure 5b), resulting in high SST in the lee of the island. The relative warm pool disappears as the northeast monsoon switches the direction.
3.3. Correlation With ENSO
 The SCS and its climate are strongly affected by ENSO [e.g., Xie et al., 2003; Liu et al., 2004; Fang et al., 2006; Wang et al., 2006]. This section investigates interannual variability of the warm pool. We define its core temperature of the relative warm pool by averaging SST in the region of 108°E–109°E, 17.5°N–18.5°N from November to January. The region is chosen because the core of warm pool is there in climatology, and the period is selected because the warm pool is strong during the 3 months.
 From 1983 to 2011, there are seven significant El Niño warm events and five significant La Niña events according to the Multivariate ENSO Index (MEI) [Wolter and Timlin, 2011]. The warm pool is strong in the significant El Niño years except 1991–1992, and weak in the significant La Niña years (Figure 7). Note the ocean over the whole SCS basin has a weak response to 1991–1992 event, suggesting the winter monsoon includes some internal atmospheric variability not always a slave of ENSO. The reason still remains unclear at this point [Wang et al., 2006]. The warm pool index is correlated well with December Nino3 SST anomalies with a correlation coefficient of 0.53 (0.374 is 95% confidence level for 26 degrees of freedom based on a t-test). The oscillation period of the warm pool index calculated by Morlet continuous wavelet transform method is 3–5 years, which is also very similar to that of ENSO.
Figures 8a and 8b show the composite winter wind anomalies over the Beibu Gulf for the El Niño years and La Niña years, respectively. The composite maps are obtained by simply averaging the wind anomalies from the climatology. The wind data begin from 1987, so there are five significant El Niño years (1987–1988, 1994–1995, 1997–1998, 2002–2003, 2009–2010), and five significant La Niña years (1988–1989, 1998–1999, 1999–2000, 2007–2008, 2010–2011) for wind composites. The winter monsoonal wind over the Beibu Gulf weakens during El Niño, and strengthens during La Niña. The interannual wind pattern is quite similar to the one over the SCS basin, which is part of atmosphere circulation change that covers the northwestern Pacific [Wang et al., 2000; Xie et al., 2009]. The interannual wind pattern indicates that the SST anomalies in the whole Beibu Gulf may be largely influenced by the ENSO. Figures 8a and 8b also show the composite winter SST anomalies from climatology over the Beibu Gulf for the El Niño years and La Niña years, respectively. Compared with the climatology, the weakened wind in El Niño reduces the release of the surface latent heat flux over the Beibu Gulf, thus warms over the whole Beibu Gulf, and thereby the SST in the warm pool becomes higher. In contrast, the strengthened wind in La Niña intensifies the release of the surface latent heat flux in the Beibu Gulf, thus cools the whole Beibu Gulf, and the SST over the warm pool becomes lower.
4. Summary and Discussion
 A small warm pool off the west coast of Hainan Island in boreal winter was observed and analyzed using a suite of new satellite measurements and field observations. The warm pool exists from the surface to bottom in winter. The relative warm pool, defined as warmer water than the surroundings, begins to develop in October, matures from November to January, and decays quickly in February.
 Our analysis shows that the warm pool is strongly associated with the wind wake behind the mountains of Hainan Island during the northeast monsoon. The dynamic linkage between the warm pool and wind wake can be described as follows: the wind wake behind the mountains of Hainan Island is observed during the northeast wind period. The latent heat flux is weak in the wake of small wind speed southwest of the mountains while strong in the surrounding areas. The pattern of latent heat flux results in the warm pool formation.
 The interannual variability of the warm pool is strongly correlated with ENSO. During El Niño, the northeast wind over the Beibu Gulf weakens, warming the Gulf by reducing the release of surface latent heat flux, thus the SST in the warm pool rises. In contrast, the strengthened wind in La Niña years cools the Beibu Gulf by increasing surface latent heat flux, thus the SST in the warm pool decreases.
 To our knowledge, this is the first time that the warm pool is described and related to the wind pattern and the orographic effect of Hainan mountains. A one-dimensional mixed-layer model captures the gross features of the warm pool when forced by observed surface forcing. Future work needs to evaluate other dynamic effects such as advection. In addition, the implication of the warm pool to the regional climate such as ecosystem response, rainfall and the boundary layer coupling is also an interesting topic for future studies.
 This study was done at the State Key Laboratory of Satellite Ocean Environment Dynamics and was supported by the National Science Foundation of China (NSFC) for Distinguished Young Scholars (41125019), National Basic Research Program of China (2012CB955601), NSFC grant (40976017), and Natural Science Foundation of Zhejiang for Innovative Research Groups (2009R50044). We thank the project “908-01-ST09” from SOA of China for providing the hydrographic data. Comments by anonymous reviewers were helpful to improve the manuscript.