Microwave-infrared, Tropical Rainfall Measuring Mission Microwave Imager and Advanced Microwave Scanning Radiometer – Earth observing system, and WindSat sea surface temperatures are validated in coastal waters of the northern South China Sea using in situ measurements from three anchored buoys from 2010–2013. Validation of satellite sea surface temperatures shows that biases ± standard deviations of Microwave-infrared, Tropical Rainfall Measuring Mission Microwave Imager and Advanced Microwave Scanning Radiometer – Earth observing system, and WindSat daytime and nighttime sea surface temperatures against Maoming buoy sea surface temperatures are 0.18 ± 0.74, 0.60 ± 0.57, 0.76 ± 0.61, and 0.22 ± 0.74°C, respectively, and against Shantou buoy sea surface temperatures are 0.02 ± 0.58, 0.05 ± 0.53, 0.04 ± 0.69, and −0.05 ± 0.71°C, respectively. Because the Shanwei buoy is very close to the coast, the bias of satellite sea surface temperatures against the Shanwei buoy sea surface temperatures is very large, especially in the winter. The mean biases of satellite sea surface temperatures in coastal waters, except for very close to the coast, are roughly equivalent to previous results in the northern South China Sea. The accuracies of satellite sea surface temperatures in coastal waters of the northern South China Sea depend not only on season, wind speed, wind direction, and region, but also on the satellite product. Accuracy and applicability of satellite-derived sea surface temperature also has the potential to further improve, especially in near-coastal areas.

Introduction

The South China Sea (SCS) is a semi-closed marginal sea located on the tropical western Pacific. The mainland around the SCS is often subjected to the influence of disastrous weather from the sea. Sea surface temperature (SST) and its variation influence numerous things, including local weather, climate change, and ecosystem. The SST of the SCS is an important factor in the development and evolution of the local weather and East Asian monsoon (Chu and Chang, 1997; Ose et al., 1997; Lin et al., 2006; Chen et al., 2010; Lin et al., 2012). The northern SCS plays the most important part in water exchange in the SCS with southwestern wind and low cloud cover in summer, and northeastern wind and high cloud cover in winter (Wang et al., 1996). Highly accurate and fine resolution SST data are urgently needed for weather forecasting and climate change studies.

Buoy SSTs do not have the global coverage offered by satellite SSTs. In recent years, there have been a variety of operational and experimental satellite-derived SST products in the SCS. The satellite-derived SST products come from infrared (IR) sensors, microwave (MW) sensors, or visible sensors. Thermal emissions from the sea surface detected using a variety of infrared and microwave radiometers onboard the operational and research satellites offer an effective way for deriving an SST field (Hosoda, 2010). The infrared SSTs are affected by diurnal warming, water vapor attenuation, atmospheric aerosols, and incomplete removal of cloud contamination (Brown et al., 1985; McClain, 1989;Emery et al., 1994). The microwave SSTs are unaffected by aerosols, and can be retrieved in the presence of clouds, simultaneous with wind speed, columnar water vapor, cloud water content, and precipitation rate (Wentz et al., 2000). However, MW SSTs are limited by errors such as decreased sensitivity at high wind speeds and by a relatively poor spatial resolution of 50 km (Gentemann et al., 2004). IR and MW SST technologies are complementary when their individual advantages are considered (Guan and Kawamura, 2004; Castro et al., 2008).

The accuracy of satellite-derived SST remains an issue. The main sources of errors for satellite-derived SSTs are associated with spacecraft navigation, algorithm, and environmental scene errors. Another problem is that the SST retrieval algorithm should be area-dependent (Pearce et al., 1989; Minnett, 1990; Kawai and Kawamura, 1997a,b; Li et al., 2001). The different SST retrieval algorithms have a variety of accuracies for the same satellite (Hosoda and Qin, 2011). The error in SST derived from the same satellite may vary with ocean basin and weather condition (Kawamura et al., 2010).

The satellite-derived SSTs in the northern SCS were validated by using in situ SSTs from the drifting buoy and the hydrographical data collected by the open cruises of the South China Sea Institute of Oceanology (Qiu et al., 2009; Qin et al., 2014). In this study, three anchored buoys from west to east in the coastal waters of the northern SCS have more than three years of data from 2010–2013. The satellite-derived SSTs in the coastal waters of the northern SCS are validated by using in situ SSTs from the anchored buoy over long periods of time.

Data and methods

Satellite data

Tropical Rainfall Measuring Mission Microwave Imager and Advanced Microwave Scanning Radiometer – Earth observing system (TMI_AMSRE) SST daily products at 0.25° resolution are provided by Remote Sensing Systems from June 2002 to the present. The instruments of TMI_AMSRE have the lower frequency channels (6–7 GHz and/or 11 GHz) and the through-cloud capabilities (Wentz et al., 2000; Gentemann et al., 2010).

The 9-km Microwave-infrared (MWIR) optimally interpolated (OI) blended SST daily products combine the through-cloud capabilities of the microwave (TMI, AMSR-E, AMSR2, WindSat) data with the high spatial resolution of the infrared (Terra MODIS, Aqua MODIS) SST data, which are produced by Remote Sensing Systems from December 2005 to the present. To reduce the diurnal contamination of the data set, the diurnal warming was estimated and corrected in the MWIR data (Gentemann et al., 2003).

Daily pass (ascending/descending) SST data are derived from WindSat version 7.0.1 geophysical products, which are mapped to a 0.25° grid and divided into two sets of maps based on ascending and descending passes. The WindSat is a multi-frequency polarimetric microwave radiometer developed by the Naval Research Laboratory (NRL) that was launched on 6 January 2003 aboard the Department of Defense Coriolis satellite. WindSat is mainly designed to measure the ocean surface wind vector from space, but it also provides SST products (Gaiser et al., 2004). The WindSat SSTs are divided into daytime (08:00–20:00 local time) and nighttime (20:00–08:00 local time) according to the WindSat observation time. Solar heating may lead to the formation of a near-surface diurnal warm layer during the daytime, particularly in regions with low wind speed (Gentemann et al., 2003, 2004).

In situ data

The in situ SST data are derived from three anchored buoys, including the Maoming (MM) buoy deployed at (111°39′E, 20°44′N), Shantou (ST) buoy deployed at (117°20′E, 22°19′N), and Shanwei (SW) buoy deployed at (115°33′E, 22°36′N) in coastal waters of the northern SCS (Figure 1). MM buoy SST was active from 11 October 2010 to 31 December 2013; ST buoy SST was active from 2 November 2010 to 31 December 2013; and the SW buoy SST was active from 8 January 2010 to 13 December 2013. All sensors of SST are at about 0.5 m depth. During the above period, both the MM buoy and ST buoy had some missing data. The buoys measure SST every 10 min, and in situ daily mean SST was derived from the average of 10-min SSTs. However, the in situ SST that compared with the ascending/descending WindSat SST was the 10-min SST at the nearest time. Quality control was carried out for the reliability of the in situ SST by first calculating daily mean and standard deviation (SD) of available 10-min SSTs. The SST observation was discarded when an absolute value of 10-min SST subtracting daily mean was greater than four times the SD. This quality control can basically ensure that the large diurnal warming in summer was reserved. The buoys measure meteorological variables such as wind speed and wind direction at 2 m height every 10 min, with winds coming from due north being represented as 0°.

Match-up data

In order to compare the satellite-derived SST with the in situ SST, the satellite-derived SSTs used the spatial averages of 2 × 2 pixels around each buoy location. The in situ daily SSTs were obtained by averaging the in situ 10-min SSTs, which compared with the daily MWIR and TMI_AMSRE SSTs. The WindSat SSTs were divided into daytime (08:00–20:00) and nighttime (20:00–08:00) according to the Local Time, which compared with the nearest 10-min buoy SSTs. The numbers matched-up between satellite SSTs and the in situ buoy SSTs are shown in Figure 2. It was found that the matched-up numbers of MWIR SST or TMI_AMSRE SST were larger than WindSat daytime SST or nighttime SST at MM and ST buoy stations, which may be attributed to the WindSat swath gaps. Because the SW buoy station is very close to the coast, only the MWIR SST is measured in this area, only MWIR SST is compared at this station. The validation results in the next section are based on the matched-up data in Table 1.

Validation of satellite-derived SSTs

Comparisons between satellite-derived sea surface temperatures and in situ buoy sea surface temperatures

The satellite SSTs at the MM buoy station range from 18–30°C, and the match-up numbers have two peaks at about 21°C and 28°C, respectively (Figure 2a). The satellite SST at the ST buoy station range from 20–31°C, and mainly range from 21–29°C (Figure 2b). The MWIR SST at the SW buoy station ranges from 19–30°C, and the match-up number has two peaks at 21°C and 28°C, respectively (Figure 2c). The accuracy of satellite-derived SSTs is indicated by bias (satellite-derived SST minus in situ SST) and SD, which are summarized in Table 1. The satellite SST against the MM buoy SSTs or SW buoy SSTs have significant positive bias, and the bias of MWIR SSTs against the SW buoy SSTs reaches 1.58°C. The large bias of MWIR SSTs against the SW buoy SSTs may be related to the water very close to the coast. However, the biases of four satellite SSTs against the ST buoy SSTs are very small, which shows that the accuracies of satellite-derived SSTs in coastal waters of the northern SCS depend significantly on the region.

Note that the biases and SDs of Moderate Resolution Imaging Spectroradiometer (MODIS) SSTs in the SCS range from −0.19 to −0.34°C and from 0.58 to 0.68°C, respectively (Qin et al., 2014). In the northern SCS, the bias ± SD of Advanced Very High Resolution Radiometer (AVHRR) SSTs against the drifting buoy SSTs is −0.43 ± 0.76°C and −0.33 ± 0.79°C for daytime and nighttime, respectively, and the bias ± SD of TMI SSTs against the drifting buoy SSTs is −0.07 ± 1.11°C and 0.00±0.97°C for daytime and nighttime, respectively (Qiu et al., 2009). The satellite SSTs in coastal waters (except for WindSat nighttime SST against ST buoy SST) are overestimated, which are different from their results. The bias of WindSat SSTs against the MM buoy SSTs is smaller in nighttime than in daytime; this result is consistent with Qiu et al. (2009) where a drifting buoy was used.

Figure 3 shows that MWIR and TMI_AMSRE SSTs against MM buoy SSTs have large positive biases when SST is low, and TMI_AMSRE SSTs and WindSat daytime SSTs obviously have systematic positive biases. The systematic bias may be caused by the difference of temperatures between the surface and the measurement depth (Donlon et al., 2002). In situ SST (such as buoy SST) measures the bulk temperature; however, satellite SST represents the skin temperature. In addition, the bias may be the result of that the SST retrieval algorithm is not very suitable or the diurnal correction is not sufficient in this area. Four satellite-derived SSTs agree well with the ST buoy SSTs (Figure 4), as shown in Table 1. MWIR SSTs against SW buoy SSTs have significant positive bias. In particular, there is a very large positive bias when SST is low (Figure 5).

Temporal variation of satellite-derived sea surface temperature accuracy

Monthly biases and SDs of the satellite-derived SSTs against MM buoy SSTs are shown in Figure 6. Because the MM buoy has some missing data and the samples of WindSat SSTs are small, there are some time series gaps from 2011–2013. The biases of the satellite-derived SSTs against MM buoy SSTs have obvious seasonal variations. The error of MWIR SST is relatively small and has positive and negative oscillations (Figure 6a). The TMI_AMSRE SST is overestimated in all months, especially in winter which corresponds to a low SST (Figure 3b). The error of TMI_AMSRE SST is larger than that of MWIR SST in most months, which is consistent with the results in Table 1. The biases of WindSat SSTs in daytime and nighttime also exhibit periodic variations and the bias is smaller in nighttime than in daytime, which is likely related to insolation (Figure 6b). WindSat SST in daytime has a significantly positive bias in all months.

The monthly biases of satellite-derived SSTs against ST buoy SSTs are quite small in most months and have positive and negative seasonal variations around zero (Figure 7), as is also shown in Table 1 and Figure 4.

MWIR SSTs against SW buoy SSTs have significant positive bias in all months (not shown). The bias has significant annual variation and can reach about +5°C in winter. The accuracy and applicability may be significantly affected because the satellite-derived SST is very close to the coast. MWIR SSTs against SW buoy SSTs have a large positive bias in winter and small bias in summer (not shown). Gentemann et al. (2004) found that the difference between TMI SST and in situ SST shows a seasonal variation. The seasonal biases of others satellite SSTs also exist in the northern SCS (Qiu et al., 2009,,2014). Vastly different air–sea interactions occur in the northern SCS and are associated with prevailing summer and winter monsoons (Hong and Wang, 2006). Seasonal biases of satellite SSTs in the northern SCS may be related to the seasonal variation of cloud (Qiu et al., 2009). In addition, the seasonal biases of satellite SSTs in coastal waters depend on area. SSTs very close to the coast may also bring error, due to the strong mixing processes that occur there (by tides, for example) and land contamination of the raw satellite data. Therefore, the quality of satellite SST should be improved for coastal waters.

Variations of satellite-derived sea surface temperature accuracy with wind speed and wind direction

Similar to the comparison conducted by Kawai and Kawamura (1997b), Donlon et al. (2002), and Qiu et al. (2009), the relationship between buoy wind speed and bias of satellite SSTs against in situ SSTs was examined. The scatter diagrams of the MM buoy wind speed vs. bias of satellite SSTs against MM buoy SSTs are drawn (Figure 8). It shows that the biases of MWIR SSTs and WindSat daytime SSTs have no obvious increasing or decreasing trend when wind speed increases (Figures 8a and c). However, the bias of TMI_AMSRE SSTs increases when wind speed increases below 12 m s−1 (Figure 8b). Bias of WindSat nighttime SSTs mainly increases when wind speed increases below 10 m s−1 and decreases when wind speed increases above 12 m s−1 (Figure 8d). The bias decreasing when wind speed is above 12 m s−1, however, may be related to the small number of samples in this study (Figure 8d). The variation of bias of WindSat daytime SST is different from that of WindSat nighttime SST, which shows that the diurnal correction may need for WindSat SST. The relationship between the bias of satellite SST and the wind speed is inconsistent, which may be due to the influence of the other factors. Figure 9 illustrates the SW buoy wind direction vs. bias of MWIR SSTs against SW buoy SSTs. It shows that the bias of MWIR SSTs against SW buoy SSTs depends on geographical wind direction, this phenomenon is most obvious in spring. However, the variations of biases of satellite SSTs against MM buoy SSTs vs. wind direction are small. The variations of biases of satellite SSTs against ST buoy SSTs vs. wind direction are not obvious (not shown). The reason of the accuracy of satellite SST near coast depending on the wind direction is unclear at present.

The analysis indicates that the differences between satellite SSTs in coastal waters and in situ SSTs depend not only on season, wind speed, region, and satellite product but also on geographical wind direction. The mean biases of satellite SSTs in coastal waters, except for very close to the coast, are nearly equivalent to the previous results in the northern SCS. Because there are complicated local features, such as land contamination, the accuracy and applicability of satellite SSTs may be significantly affected when the water is very close to the coast.

Discussion and conclusions

In this study, the MWIR, TMI_AMSRE, and WindSat-derived SSTs in coastal waters of the northern SCS in 2010–2013 are validated using in situ SSTs from three anchored buoys that are deployed in the western, central, and eastern coastal waters, respectively. A large number of match-up samples are obtained. The validation of satellite SSTs shows that biases±SDs of MWIR, TMI_AMSRE, and WindSat daytime and nighttime SSTs against MM buoy SSTs are 0.18 ± 0.74, 0.60 ± 0.57, 0.76 ± 0.61, and 0.22 ± 0.74°C, respectively, and biases ± SDs of MWIR, TMI_AMSRE, and WindSat daytime and nighttime SSTs against ST buoy SSTs are 0.02 ± 0.58, 0.05 ± 0.53, 0.04 ± 0.69, and −0.05 ± 0.71°C, respectively. The bias ± SD of MWIR SSTs against SW buoy SSTs is 1.58 ± 1.82°C, which may be related to proximity of water to the coast. Most of the satellite-derived SSTs were overestimated. The mean biases of satellite SSTs in coastal waters, except for very close to the coast, are nearly equivalent to the previous results in the northern SCS.

The validation of satellite SSTs also shows that the accuracies of satellite SSTs in coastal waters of the northern SCS depend not only on season, wind speed, wind direction, and region but also on satellite product. The accuracy and applicability of satellite SSTs are very poor when the water is very close to the coast. The reason of the accuracy of satellite SST near coast depending on wind direction is worthy of further analysis. The SST retrieval algorithm should be improved for coastal waters.

Acknowledgements

Many thanks to the Guangdong Meteorological Bureau for providing data from the three buoys, and to NASA for providing the satellite-derived SST products. We are grateful to Dr. Huiling Qin of the South China Sea Institute of Oceanology, Chinese Academy of Sciences, for her useful suggestions on this study.

Funding

This work was supported by the National Basic Research Program of China (973 Program) under Grant 2011CB403504, NSFC under Grants 91215302 and 41275053, and the Strategic Priority Research Program of the Chinese Academy of Sciences under Grant XDA11010403, and the Clean Development Mechanism Foundation of China under Grant 1212014.

The text of this article is only available as a PDF.

References

Brown, O.B., Brown, J.W., Evans, R.H.,
1985
.
Calibration of advanced very high resolution radiometer infrared observations
.
Journal of Geophysical Research
90
,
1667
1677
.
Castro, S.L., Wick, G.A., Jackson, D.L., Emery, W.J.,
2008
.
Error characterization of infrared and microwave satellite sea surface temperature products for merging and analysis
.
Journal of Geophysical Research
113
,
C03010
. DOI:.
Chen, Y., Xie, Q., Meng, W., Yuan, J., Wang, D.,
2010
.
A numerical study of the influence of sea surface temperatures with different temporal resolutions on typhoon DUJUAN over the South China Sea
.
Journal of Tropical Meteorology
16
(
2
),
195
200
.
Chu, P.C., Chang, C.P.,
1997
.
South China Sea warm pool in boreal spring
.
Advances in Atmospheric Sciences
14
,
195
206
.
Donlon, C.J., Minnett, P.J., Gentemann, C., Nightingale, T.J., Barton, I.J., Ward, B., Murray, M.J.,
2002
.
Toward improved validation of satellite sea surface skin temperature measurements for climate research
.
Journal of Climate
15
,
353
369
.
Emery, W.J., Yu, Y., Wick, G.A., Schluessel, P., Reynolds, R.W.,
1994
.
Correcting infrared satellite estimates of sea surface temperature for atmospheric water vapor attenuation
.
Journal of Geophysical Research
99
,
5219
5236
.
Gaiser, P.W., Poe, G.A., Jones, W.L., Bevilacqua, R.M., Chang, P.S.,
2004
.
The WindSat space borne polarimetric microwave radiometer: sensor description and early orbit performance
.
IEEE Transactions on Geoscience and Remote Sensing
42
(
11
),
2347
2361
.
Gentemann, C.L., Donlon, C.J., Stuart-Menteth, A., Wentz, F.J.,
2003
.
Diurnal signals in satellite sea surface temperature measurements
.
Geophysical Research Letters
30
(
3
),
1140
. DOI:.
Gentemann, C.L., Meissner, T., Wentz, F.J.,
2010
.
Accuracy of satellite sea surface temperatures at 7 and 11 GHz
.
IEEE Transactions on Geoscience and Remote Sensing
48
(
3
),
1009
1018
.
Gentemann, C.L., Wentz, F.J., Mears, C.A., Smith, D.K.,
2004
.
In situ validation of Tropical Rainfall Measuring Mission microwave sea surface temperatures
.
Journal of Geophysical Research
109
,
C04021
. DOI:.
Guan, L., Kawamura, H.,
2004
.
Merging Satellite Infrared and Microwave SSTs: Methodology and Evaluation of the New SST
.
Journal of Oceanography
60
(
5
),
905
912
.
Hong, B., Wang, D.,
2006
.
Diagnostic analysis on the northern South China Sea winter counter-wind current
.
Chinese Scientific Bulletin
51
(
B12
),
9
16
.
Hosoda, K.,
2010
.
A review of satellite-based microwave observations of sea surface temperatures
.
Journal of Oceanography
66
(
4
),
439
473
.
Hosoda, K., Qin, H.,
2011
.
Algorithm for estimating sea surface temperatures based on AQUA/MODIS global ocean data. 1. Development and validation of the algorithm
.
Journal of Oceanography
67
(
1
),
135
145
.
Kawai, Y., Kawamura, H.,
1997a
.
Seasonal and diurnal variability of differences between satellite-derived and in situ sea surface temperatures in the south of the Sea of Okhotske
.
Journal of Oceanography
53
(
4
),
343
354
.
Kawai, Y., Kawamura, H.,
1997b
.
Characteristics of the satellite-derived sea surface temperature in the oceans around Japan
.
Journal of Oceanography
53
,
161
172
.
Kawamura, H., Qin, H., Hosoda, K., Sakaida, F., Qiu, C.,
2010
.
Advanced sea surface temperature retrieval using the Japanese geostationary satellite, Himawari-6
.
Journal of Oceanography
66
(
6
),
855
864
.
Li, X., Pichel, W., Clemente-Colon, P., Krasnopolsky, V., Sapper, J.,
2001
.
Validation of coastal sea and lake surface temperature measurements derived from NOAA/AVHRR data
.
International Journal of Remote Sensing
22
(
7
),
1285
1303
.
Lin, A., Liang, J., Li, C.,
2006
.
Characteristics of frequency spectrum variation of intraseasonal oscillation of convection during South China Sea summer monsoon
.
Journal of Tropical Meteorology
12
(
1
),
34
40
.
Lin, A., Li, C., Gu, D., Zheng, B.,
2012
.
Variation and causes of persistent drought events in Guangdong province
.
Journal of Tropical Meteorology
18
(
1
),
54
64
.
McClain, E.P.,
1989
.
Global sea surface temperatures and cloud clearing for aerosol optical depth estimates
.
International Journal of Remote Sensing
10
,
763
769
.
Minnett, P.J.,
1990
.
The regional optimization of infrared measurements of sea surface temperature from space
.
Journal of Geophysical Research
95
(
C8
),
13497
13510
.
Ose, T., Song, Y., Kitoh, A.,
1997
.
Sea surface temperature in the South China Sea–an index for the Asian monsoon and ENSO system
.
Journal of the Meteorological Society of Japan
75
,
1091
1107
.
Pearce, A.F., Prata, A.J., Manning, C.R.,
1989
.
Comparison of NOAA/AVHRR-2 sea surface temperatures with surface measurements in coastal waters
.
International Journal of Remote Sensing
10
(
1
),
37
52
.
Qin, H., Chen, G., Wang, W., Wang, D., Zeng, L.,
2014
.
Validation and application of MODIS-derived SST in the South China Sea
.
International Journal of Remote Sensing
35
,
4315
4328
.
Qiu, C., Wang, D., Kawamura, H., Guan, L., Qin, H.,
2009
.
Validation of AVHRR and TMI-derived sea surface temperature in the northern South China Sea
.
Continental Shelf Research
29
(
20
),
2358
2366
.
Wang, D., Zhou, F., Li, Y.,
1996
.
On the features in the annual cycle of SST and surface heat budget in the South China Sea
.
Acta Oceanologica Sinica
15
(
11
),
111
125
.
Wentz, F.J., Gentemann, C.L., Smith, D.K., Chelton, D.B.,
2000
.
Satellite measurements of sea-surface temperature through clouds
.
Science
288
(
5467
),
847
850
.