1. Introduction
Sea Surface Temperature (SST) is an essential parameter for understanding various oceanic phenomena and climate variability. It plays a fundamental role in the exchange of heat, momentum and water vapor between the atmosphere and the ocean. The distribution and characteristics of SST have an important impact on marine ecosystems and SST data have been widely used in many environmental applications such as climate change monitoring, weather forecasting, and ocean data assimilation (Yan et al., 2015; Dai, 2016).
The conventional method for measuring SST was to directly observe the surface by using the oceanic buoy or the ship. However, the distribution of this in-situ measurements is very sparse due to the spatial constraints, making it very inefficient to obtain SST data over a large area (Banzon et al., 2010). In comparison, SST observations through satellites are useful for analyzing the spatial distribution of SST and for supplementation the coverage of in-situ measurements, as it allow the retrieval of continuous data over the global area. The satellite SST can be either derived from the infrared or from the microwave spectral region. Infrared (IR)sensors in the 3.7 and 10-12 μm spectral range have been used to retrieve SST for more than 20 years and the IR SST has high spatial resolution (~1 km) and comparatively high accuracy (0.4 - 0.6°C). However, it cannot be retrieved in the presence of clouds or aerosols, limiting its useulness in all weather conditions (Harris and Saunders, 1996; Donlon et al., 2007; Walton, 2016), and large uncertainties occur due to incomplete removal of cloud contamination (Emery et al., 1994; McClain, 1989). On the other hand, in the microwave region, it is possible to acquire data even in the presence of clouds, and the effects of the atmosphere can be corrected relatively easily, so that continuous SST data with much lesser space can be obtained (Wentz et al., 2000). The properties of being able to penetrate clouds and not being highly contaminated with aerosols allows the microwave SST to fill the gaps in the IR SST, thus providing valuable means for the production of spatially-complete merged SST maps (Donlon et al., 2004; Maturi et al., 2017). However, the use of microwave SST may be limited by such issues as decreased sensitivity at low water temperatures and relatively low spatial resolution of about 50 km.
Early microwave radiometers were restricted in their usefulness due to the poor calibration systems and the absence of low frequency channels sensitive to sea surface parameters. High quality microwave SST data was available from observations at the 10.7 GHz frequency of the TropicalRainfall Microwave Mission (TRMM) Microwave Imager (TMI) launched in November 1997. Afterwards, the SST retrieved from microwave sensors such as Advanced Earth Observing Satellite-II (ADEOS-II) / Advanced Microwave Scanning Radiometer (AMSR) and EOS-Aqua / AMSREarth Observig System (AMSR-E) have been widely used. Currently, WindSat and Global Change Observation Mission-Water1 (GCOM-W1)/ AMSR2, Global Precipitation Measurement (GPM) / GPM Microwave Sensor (GMI) are in operation.
GPM launched in February 2014 by the National Aeronautics and Space Administration (NASA) and the Japan Aerospace and Exploration Agency (JAXA) have an orbit inclined with an angle of 65 degrees, which has an improved temporal resolution. Data can be collected for all the local time and the whole area of the earth can be observed in approximately two weeks. The GMI sensor on this satellite is designed to undergo a more rigorous calibration process than any previous microwave satellite sensor, enabling the acquisition of data with better precision (Draper et al., 2015). Microwave SST is typically retrieved using bands of 6.9 GHz and 10.7 GHz, with GPM/GMI SST being produced at 10.7 GHz.
Many researches have been carried out to evaluate the accuracy of the microwave SST and to analyze the error characteristics. The accuracy of microwave SST is currently estimated to be about 0.4°C to 0.9°C, and errors may be caused by various environmental factors such as wind speed, land, and thermal fronts (Gentemann et al., 2004; Dong et al., 2006; Hihara et al., 2015; Gentemann and Hilburn, 2015; Kim et al., 2016). However, most of the validation studies for microwave SST have been conducted on the global area, which did not reflect regional characteristics. Inthe calcuation of SST, unevenly distributed in-situ temperature measurements make it difficult to calibrate for local areas, and errors can occur in the regions where in-situ measurements are sparse (Dong et al., 2006). Thus, the accuracy of the retrieved SST can vary greatly depending on various regional characteristics, and unfortunately, the accuracy and error characteristics of the GMI SST have yet to be revealed for the Northwest Pacific. The seas around Korea and the Northwest Pacific have a wide range of SST (Fig. 1). Especially, there are thermal fronts and meso-scale eddies in the East Sea and the Kuroshio region, hence the SST fluctuates widely both in time and space (Park et al., 2015). In the areas where the SST varies greatly, the satellite SST may be significantly different from the in-situ SST due to various oceanic and atmospheric conditions. Therefore, in this study, microwave SST of the GPM/GMI are compared with in-situ temperature measurements, to quantitatively evaluate the accuracy and to analyze the characteristics of the error for the Northwest Pacific during the 3 years of 2014 to 2016.
Fig. 1. (a) Global mean sea surface temperature using climatological data of Multi-scale Ultra-high Resolution Sea Surface Temperature and (b) the locations of the Korean ocean buoy stations used in the study. Box with dotted line represents the study area.
2. Data and Method
1) GPM/GMI Data
The GMI is a conical-scaning passive micrwave sensor mounted on the GPM of a polar orbiting satellite jointly developed by NASA and JAXA. It was launched in February 2014 and data was distributed from March of the same year. The GMI instrument detects radiation emitted from the earth surface at 8 microwave wavelength bands, and consists of 13 channels that are vertically and horizontally polarized for each wavelength except 23.8 and 183.31 GHz. The observations for the same area are repeated in approximately 14 days. Key features and channel information are summarized in Table 1.
Table 1. Specification of GPM/GMI
In the frequency range between 4 and 11 GHz, the vertically polarized brightness temperature (Tb) of the sea surface has an appreciable sensitivity with respect to SST. The Tb also varies depending on the sea surface roughness and the vertical structure of the atmospheric temperature and water vapor. Sea surface roughness can usually be expressed in terms of wind speed and direction. Fortunately, the spectral and polarimetric signals of the sea surface roughness and the atmosphere are clearly different from those of SST, and the effects of these factors can be removed by using multiple frequencies and polarization simultaneously (Meissner et al., 2012).
A commonly applied algorithm for microwave SST retrieval is a physical-based regression, from which the SST can be estimated with the Tb. After estimating the initial SST by calculating the regressioncoefficients determind from the radiative transfer model, a number of localized retrieval algorithms are used for a relatively narrow range of initial SST in the range of -3 to 34°C, taking into account all SST/wind combinations (Wentz and Meissner, 2007). Microwave SST is limited in the presence of rain, sun glint, near land, or when radiofrequency interference (RFI) is present (Gentemann, 2014).
In this study, level 2 (L2) SST data of version 8.2 provided by Remote Sensing Systems (RSS) was collected and used for validation and analysis of error characteristics with in-situ temperature measurements. In order to analyze various error factors, L2 products of wind speed, atmospheric water vapor were also collected and utilized. An analysis was carried out on data from March 2014 to December 2016, selecting the region of the latitude 9-60°N and the longitude 99-177°E in the Northwest Pacific as a study area.
2) in-situ SST measurements
To validate the microwave SST, we have collected a large set of in-situ temperature measurements. The data set consists of satellite-tracked drifter and moored buoy data obtained via the Global Telecommunication System (GTS) database operationally used at Korea Meteorological Administration (KMA). Both data observe the SST at an average depth of 0.2-0.3 m and less than 1 m, respectively. Other in-situ temperature measurements, such as CTD (Conductivity, Temperature, Depth)measurements during ship cruises or RGO (water temperature dat of the Array for Real-time Geostrophic Oceanography) temperatures, were excluded to maintain consistency of observational depth. The temporal resolution of the drifter buoy data is usually 1 hour and the precision is normally known to be about 0.01°C.
We obtained a total of 2,846,337 observations from the GTS drifters in the study area of 2014 to 2016. The data include ancillary information such as the drifter id, observation time, latitude, longitude, as well as some records of the wind speed, wind direction, atmospheric temperature. Fig. 2 shows the tracks of GTS drifters and the spatial distribution of the number of in-situ temperature measurements in the Northwest Pacific from 2014 to 2016.
Fig. 2. Spatial distributions of (a) the tracks of satellite-tracked GTS drifters and (b) the number of the temperature observations in the Northwest Pacific for period 2014-2016.
3) Quality Control of Drifter Data
Observational error may appear in the drifter buoy data due to oceanic conditions, physical environmental factors, and errors of the sensors. For example, some of unqualified in-situ measurements were found to have observed same value for 13 hours near the latitude of 29.54°N and longitude of 152.93°E and to have shown a daily variation of 30°C or more. If these data are directly applied to the validation without quality control, there will be some problems that in-situ measurement error and satellite observationa error are not distinguished from each other, and therefore, the error characteristic in the analysis cannot be regarded as that of the satellite SST. In order to perform an accuracy assessment of satellite SST, it is necessary to remove the erroneous values contained in the in-situ measurements prior to the collocation procedure.
In this study, the quality control procedures developed by the KMA were carried out in the following order to eliminate errors in drifter buoy data. 1) The drifter buoy with the value below -99 and those with less than 10 observations per day were excluded from the data. 2) If the difference between maximum and minimum values of the day are zero or greater than 4°C, then the data were removed. 3) Data outside the range of m (daily average) ± 3σ (standard deviation) were also regarded as errors and excluded. 4) If the standard deviation is equal to 0°Cor more than 2°Cfor 4 days after the reference date, all data of the corresponding buoy are eliminated during that period. The erroneous values still remained in the drifter buoy data were further discarded using daily SST climatology data. The drifter buoy SST, which differs by more than 8°C from the SST climatology data, was considered outliers(O’Carroll et al., 2006), resulting in the removal of 328,612 drifter SST data.
4) Matchup Procedure
We generated a matchupdatabase between the GMI data and the rifter data within a spatial gap of about 25 km and a temporal interval of 30 minutes. It was considered that enough matchup data could be produced because of the wide spatial gap, so that if multiple drifter buoys are present within the 25 km range, the buoy at the closest distance to the location of satellite was selected as a matchup point. The matchup database also included detailed information such as auxiliary data from drifters (time information, geolocation like latitude and longitude, identification number, wind direction, wind speed, air temperature) and from satellites (time and geolocation information, solar zenith angle, wind speed, and atmospheric water vapor) for analyzing the error characteristics of microwave SST.
3. Results
1) Distribution of Collocated Data
The number of matchup points in the Northwest Pacific was 102,294 for 3 years, from 2014 to 2016. The spatial distribution of the matchup data can vary depending on the satellite orbit, the location of the in-situ measurements, and the observation time. The location information of the matchup points plays an important role in understanding the error of the satellite SST. Fig. 3(a) shows the spatial distributions of the number of collocated points generated in the study area. The matchup points were widely distributed throughout the study area, but the dense distributions were seen in the Northwest Pacific, especially above 40°N, whilethe East Sea, Yellow Sea and Okhotsk Sea reions had a relatively small number of matchup points. This implies that the SST error of satellite observations, which will be discussed later, will reflect the characteristics of the Northwest Pacific relatively more. In the latitude-time plot of Fig. 3(b), most of the matchup points were located in mid-latitude. The numbers of collocated points around 10°N in summer are relatively small since the microwave SST cannot be retrieved in a strong precipitation area (Fig. 3(b)). The small number of matchup data in 2016 (Fig. 3(c)) and the sparse data distribution at high latitudes (Fig. 3(a)) are considered to be based on the temporal and spatial distribution of the in-situ measurements, which may have a significant influence on the numbers of matchup points.
Fig. 3. (a) Spatial distributions of the number of matchup points in the study area for the period 2014-2016. (b) Latitudinal distributions of the number of matchup points and (c) the yearly distributed matchup data according to the month during the period of 2014-2016.
The collocated matchup data showed a wide SST distribution from -1.3°C to 33.5°C, with the most matchup data (36,634 points) in the 25-30°C range, corresponding 36.8% of total distributions. There were 10,497 matchup points below 5°C and 83 points at the extremely low temperature below 0°C. Although the distributional differences over the whole temperature rane are present, it is expected that the accurac of satellite SST in extreme conditions and the error characteristics for various temperature ranges can be analyzed properly.
2) Accuracy of GMI SST
By comparing the drifter temperatures with the satellite SST observed on daytime and nighttime, the accuracies of GMI SST were estimated (Fig. 4(a), (c)). The GMI SST was generally considered to be in good agreement with the drifter temperature, having a linear proportional relationship. Table 2 summarizes the performances of the SST validation trials with the number of matchup points, the root-mean-square (RMS) error, and bias. The RMS error of the GMI SST for all the collocated matchup data, regardless of day and night, was approximately 0.93°C with the bias of about 0.05°C. This was somewhat higher than the values (0.4°C to 0.9°C) presented as the RMS error of microwave SST in previous studies(Gentemann et al., 2010; Hosoda, 2010; Stammer et al., 2003).
Table 2. Root-mean-square (RMS) error, mean bias of daytime/nighttime
Fig. 4. Comparisons between GMI SST and drifter temperature at (a) daytime and (c) nighttime, and the distribution of SST errors (satellite SST – in-situ SST) as a function of the drifter temperature, where the colors represent the number of matchup points. The lines with errorbars are the temperature differences (°C) between GMI and drifter buoy depending on the in-situ SST in b) day and (d) night.
Fig. 4(b) and (d show the error of GMI SST as a function of drifter temperature for day and night. The error values were distributed within the range of -5 to 5°C. Typically, there were large positive mean SST errors below 10°C of water temperature and its variances tended to increase as the water temperature decreased. The degraded sensitivity of the 10.7 GHz SST at colder water temperature (Hollinger and Lo, 1984), which can lead to errors, may account for a portion of these increased errors shown in Fig. 4(b) and (d).
3) Characteristics of GMI SST Errors
(1) Effect of Wind Speed
In order to analyze the effect of wind speed on the GMI SST errors, we used satellite wind speed data. Of course, it is best to utilize the wind speed data from the in-situ measurements in terms of reliability of the data required for analysis, however, there were no wind-recorded drifter data among the matchup points, and only a few wind speed data from the moored buoy of KMA were available. Since the limited number of data were expected to have a great influence on the analysis results, we were constrained to carry out the analysis with satellite wind speed data.
Fig. 5 shows the variations of the day/night SST residuals (satellite SST – in-situ SST) with respect to the wind speed (m/s). In the low wind speed range below 5 m/s, mean SST errors were positive in both day and night data. Increasing positive mea biases with decreasing wind speed were more noticeable at thedaytime, and negative mean biases were observed at wind speed of 7 m/s or higher, less than 13 m/s. The large positive values of mean SST errors at low wind speed, which is a distinctive feature of Fig. 5(a), can be attributed to the over-heating of the sea surface during daytime when the weak wind blows, as shown in previous studies(Gentemann andWentz, 2001; Donlon et al., 2007; Park et al., 2011). The microwave sensor measures the SST by observing the radiant energy from within 1 mm of the sea surface, while the drifter buoy measures the water temperature between the depth of several tens of centimeters. This difference in observation depth between the satellite and the drifter buoy can be an error factor when the vertical change of the water temperature becomes large due to the overheating of the sea surface during the daytime underlow wind speed conditions.
Fig. 5. Distributions of (a) daytime and (b) nighttime SST errors (satellite SST – in-situ SST), as a function of the wind speed (m/s), where the bars represent the mean standard deviations of SST errors for each interval.
The positive mean SST errors were increasing in the high wind speed range above 10 m/s during the night time (Fig. 5(b)). At higher wind speeds, the sea surface emissivity can have larger errors. The surface roughness affecting on the emissivity can be expressed as a function of the wind speed and th wind direction (Meissner and Wentz, 2012), but the correlation btween the emissivity and wind speed cannot be fully defined (Wentz, 1997). This uncertainty of the surface emissivity due to the strong wind speed can be one of the main causes of errors in the microwave SST retrieval.
The microwave SST performance is strongly dependent on both SST and wind speed. To investigate this dependence, we stored the collocated results by wind speed and SST, calculating the mean SST error in the each bin (Fig. 6). During the daytime there were pronounced large positive mean biases at low wind speeds below 6 m/s (Fig. 6(a)), as shown in Fig. 5(a). The mean SST errors at night were particularly notice able in cold water, especially at high wind speed (Fig. 6(b)). These results can be useful for further analysis of error characteristics, allowing consideration of mean SST errors according to water temperature and wind speed simultaneously.
Fig. 6. Distributions of (a) daytime and (b) nighttime SST errors (satellite SST – in-situ SST) collocated by wind speed and in-situ SST (°C), where the colors represent the mean SST errors calculated in the each bin.
(2) Effect of Water Vapor
The relationships between the SST errors and atmospheric water vapor at day and night are shown in Fig. 7(a) and (b). Overall, the lower the atmospheric water vapor, the larger the mean SST errors. The mean biases were positive in the range of less than 0 mm of water vapor. The standard deviation was about 2°C and its vaue decreased with increasing atmospheric water vapor, except where the number of data in the range was small, so that the standard deviation became large. In the microwave spectrum below 100 GHz, atmospheric absorption is attributed to oxygen, water vapor, and liquid droplet particles (Wentz and Meissner, 2000). The reason for the larger errors in lower atmospheric water vapor is probably due to the atmospheric absorption model errors (Wentz and Meissner, 2016).
Fig. 7. Distributions of (a) daytime and (b) nighttime SST errors (satellite SST– in-situSST),as a function of the atmospheric water vapor (mm), where the bars represent the mean standard deviations of SST errors for each interval.
(3) Land Contamination
Due to the low spatial resolution of the microwave sensor, errors may occur caused by the influence of land during coastal observations. In the microwave spectral region, land has a mean emissivity of 0.85, which is higher than the mean emissivity of the ocean, 0.4.When the microwave sensor passes along the coast, the microwave signals emitted from the land can be observed together. Then, the Tb values are calculated to be higher than the actual Tb, causing the positive bias to remain in the microwave SST (Ricciardulli and Wentz, 2004). RSS does not retrieve the GMI SST within 100 km from the coast in consideration of such land contamination.
Fig. 8 is a plot of the mean SST errors depending on the distance from the cost. Within 200 km from the coast, the SST biases increased approaching to the coast with a maximum bias of 1.16°C. Despite of the exclusion of the data within 100 km from the coastline, errors can occur due to the influence of the islands around the coast not being considered in the retrieval of microwave SST (Gentemann et al., 2004).
Fig. 8. SST errors (satellite SST – in-situ SST) as a function of distance from the coast (km), where the bars represent the mean standard deviations of SST errors for each interval.
(4) Seasonal Variations of SST Errors
Fig. 9 shows the seasonal distribution of the large SST errors over ±2°C. The period for each season was divided into spring (March to May), summer (June to August), autumn (September to November), and winter (December to February). All four seasons had large error values concentrated in the vicinity of the Kuroshio region above 30°N. The densely distributed area of large error values corresponded to a cold region with a mean SST of about 15°C (Fig. 1(a)), and one of the reasons for this error distribution is that the sensitivity to the variation in the SST degrades over cold water in the case of the 10.7 GHz band where the GMI SST is retrieved (Gentemann et al., 2010).
Fig. 9. The spatial distribution of large difference (>±2°C) between satellie SST and in-situ temperature measurements divided into (a) spring, (b) summer, (c) auumn, and (d) winter. The colors represent the values of the GMI SST errors.
Large SST errors were shown in winter with the largest number of distributions and the least in summer. In addition, there were predominantly positive bias values in winter, and in the other seasons, negative bias values were more dominant. As a factor of seasonal variations in the GMI SST errors, the wind intensity can be taken into account. The microwave emitted from the sea surface has different emissivity depending on the sea surface roughness, and as the sea surface roughness becomes larger, so does the emissivity (Stogryn, 1967; Wentz, 1983). Since the sea surface roughness is affected by the wind speed, wind direction, and wind drift distance (Yoshimori et al., 1994), the high wind speed increases the emissivity at sea surface, thereby increasing the Tb measured by the satellite. An error may occur if these effects are not well corrected in the retrieval of microwave SST. The low water temperature, as noted in the previous section as an error factor of microwave SST, and the strong wind blowing in winter may contribute to the large numbers of positive bias values.
(5) Spatial Distribution of Probability of SST Errors
Given RMS error of the GMI SST as a function of latitude, the RMS error increased with higher latitudes, and in particular, it increased sharply from the latitue of 30°N (Fig. 10(a)). This is also evident in Fig. 10(b), which illustrates the probability map of the large SST errors. Each colored bin in Fig. 10(b) represents the ratio of the large SST errors over ±2°C among the number of the matchup data generated in the grid. The error ratio values were generally high in latitudes above 30°N, compared with the low latitudes below 20°N. The large error ratios of more than 15% were mainly observed in coastal regions, and some of them showed error rates of even more than 30%.It can be considered that the error ratios which appeared largely in the local waters such as Kamchatka, the Sea of Okhotsk and the Yellow Sea are due to a number of complex factors, such as environmental conditions, the distribution of the drifter data, and the moored buoy observation. Using this error percentage map, which gives the probability of occurrence of large error values, the area where the reliability of the data is guaranteed can be firstly selected and applied for further researches with the GMI SST data, especially in SST composite process. For example, if the researchers only want to select data with an error rate of less than 10%, they may be able to use the data except near coastal and high latitude regions for their study.
Fig. 10. (a) Root-mean-square (RMS) error (°C) of GMI SST as a function of latitude (°N), where the bars represent the mean standard deviations of SST errors for each interval. (b)The spatial distribution of the probability of SST errors over ±2°C in the Northwest acific for the period of 2014 to 2016.
4. Summary and Conclusion
Although the accuracy of the microwave SST can vary widely depending on various regional characteristics, the validation of the SST observed by GPM/GMI, a microwave sensor jointly operated by NOAA and JAXA, has not yet been performed. Therefore, in this study, the accuracy assessment of GMI SST in the Northwest Pacific was conducted and the error characteristics were quantitatively analyzed. We produced a total of 102,294 collocated matchup database between GPM/GMI satellite and drifter data for the period of 2014-2016.
Overall, the RMS error and bias of the GMI SST were about 0.93°C and about 0.05°C, respectively, which are slightly below the accuracy levels of the microwave SST reported in previous studies. Microwave SST is affected by various environmental factors, causing an error in the data. The GMI SST tended to be higher than the drifter buoy temperature at low wind speed (<6 m/s) during the daytime, and the positive bias became larger with increasing wind speed at nighttime.In addition, the less the atmospheric water vapor, the greater the mean SST errors and the standard deviation. The sensitivity degradation at low water temperatures and data contamination by high land emissivity also affected an increase in the error of GMI SST.
The spatial distributio of the matchup points with the error of ±2°C or more shows that the large error values are clustered around the Kuroshio region in the latitudes above 30°N. The seasonal distribution of large SST errors indicates the largest number of errors in winter and a significant positive biases. The large positive bias that occurs frequently in winter can be explained in relation to wind speed and the water temperature. As the wind speed increases, the sea surface emissivity becomes larger, resulting in an increase in the brightness temperature measured by the satellite. Since these errors may be reflected in the retrieval of SST, a large positive bias appears in winter with strong wind and low water temperature. The spatial distribution of the probability of SST errors over ±2°C showed an increasing error rate near coastal regions, in the latitudes above 30°N, and in some local waters. The results of this analysis can be used to obtain more accurate data in further studies using GMI SST.
Acknowledgment
This work was supported by ‘Development of Scene and Surface Analysis Algorithms’ project, funded by ETRI, which is a subproject of ‘Development of Geostationary Meteorological Satellite Ground Segment (NMSC-2017-01)’ program funded by NMSC(National Meteorological Satellite Center) of KMA (Korea Meteorological Administration).
References
- Banzon, V.F., R.W. Reynolds, and T.M. Smith, 2010. The Role of Satellite Data in Extended Reconstruction of Sea Surface Temperatures, Proc. of 2010 Oceans from Space Symposium, Venice, Apr. 26-Apr. 30, pp. 27-28.
- Chelton, D.B. and F.J. Wentz, 2005. Global microwave satellite observations of sea surface temperature for numerical weather prediction and climate research, Bulletin of the American Meteorological Society, 86(8): 1097-1115. https://doi.org/10.1175/BAMS-86-8-1097
- Dai, A., 2016. Future Warming Patterns Linked to Today's Climate Variability, Scientific Reports, 6: 19110. https://doi.org/10.1038/srep19110
- Dong, S., S.T. Gille, J. Sprintall, and C. Gentemann, 2006. Validation of the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) sea surface temperature in the Southern Ocean, Journal of Geophysical Research: Oceans, 111(C4).
- Donlon, C.J., L. Nykjaer, and C. Gentemann, 2004. Using sea surface temperature measurements from microwave and infrared satellite measurements, International Journal of Remote Sensing, 25(7-8): 1331-1336. https://doi.org/10.1080/01431160310001592256
- Donlon, C., N. Rayner, I. Robinson, D.J.S. Poulter, K.S. Casey, J. Vazquez-Cuervo, E. Armstrong, A. Bingham, O. Arino, and C. Gentemann, 2007. The Global Ocean Data Assimilation Experiment High-Resolution Sea Surface Temperature Pilot Project, Bulletin of the American Meteorological Society, 88(8): 1197-1213. https://doi.org/10.1175/BAMS-88-8-1197
- Draper, D.W., D. Newell, F.J. Wentz, S. Krimchansky, and G.M. Skofronick-Jackson, 2015. The Global Precipitation Measurement (GPM) Microwave Imager (GMI): Instrument overview and early on-orbit performance, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(7): 3452-3462. https://doi.org/10.1109/JSTARS.2015.2403303
- Emery, W.J., Y. Yu, G.A. Wick, P. Schluessel, and R.W. Reynolds, 1994. Correcting infrared satellite estimates of sea surface temperature for atmospheric water vapor attenuation, Journal of Geophysical Research: Oceans, 99(C3): 5219-5236. https://doi.org/10.1029/93JC03215
- Gentemann, C.L. and F.J. Wentz, 2001. Satellite microwave SST: Accuracy, comparisons to AVHRR and Reynolds SST, and measurement of diurnal thermocline variability, Proc. of 2001 Geoscience and Remote Sensing Symposium, Sydney, Jul. 9-Jul. 13, vol. 1, pp. 246-248.
- Gentemann, C.L., F.J. Wentz, C.A. Mears, and D.K. Smith, 2004. In situ validation of tropical rainfall measuring mission microwave sea surface temperatures, Journal of Geophysical Research: Oceans, 109(C4).
- Gentemann, C.L., T. Meissner, and F.J. Wentz, 2010. Accuracy of satellite sea surface temperature at 7 and 11 GHz, IEEE Transactions on Geoscience and Remote Sensing, 48(3): 1009-1018. https://doi.org/10.1109/TGRS.2009.2030322
- Gentemann, C.L., 2014. Three way validation of MODIS and AMSR-E sea surface temperatures, Journal of Geophysical Research: Oceans, 119: 2583-2598. https://doi.org/10.1002/2013JC009716
- Gentemann, C.L. and K.A. Hilburn, 2015. In situ validation of sea surface temperatures from the GCOM-W1/AMSR2 RSS calibrated brightness temperatures, Journal of Geophysical Research: Oceans, 120(5): 3567-3585. https://doi.org/10.1002/2014JC010574
- Harris, A.R. and M.A. Saunders, 1996. Global validation of the along-track scanning radiometer against drifting buoys, Journal of Geophysical Research: Oceans, 101(C5): 12127-12140. https://doi.org/10.1029/96JC00317
- Hihara, T., M. Kubota, and A. Okuro, 2015. Evaluation of sea surface temperature and wind speed observed by GCOM-W1/AMSR2 using in situ data and global products, Remote Sensing of Environment, 164: 170-178. https://doi.org/10.1016/j.rse.2015.04.005
- Hollinger, J.P. and R.C. Lo, 1984. Low Frequency Microwave Radiometer for N-ROSS, Proc. of 1984 "OCEANS" Conference, Washington, D.C., Sep. 10-12, pp. 167-174.
- Hosoda, K., 2010. A Review of Satellite-Based Microwave Observations of Sea Surface Temperatures, Journal of Oceanography, 66(4): 439-473. https://doi.org/10.1007/s10872-010-0039-3
- Kim, H.Y., K.A. Park, and H.J. Woo, 2016. Validation of Validation of Sea Surface Temperature (SST) from Satellite Passive Microwave Sensor (GPM/GMI) and Causes of SST Errors in the Northwest Pacific GCOM-W1/AMSR2 Sea Surface Temperature and Error Characteristics in the Northwest Pacific, Korean Journal of Remote Sensing, 32(6): 721-732 (in Korean with English abstract). https://doi.org/10.7780/kjrs.2016.32.6.15
- Maturi, E., A. Harris, J. Mittaz, J. Sapper, G. Wick, X. Zhu, and P. Koner, 2017. A New High-Resolution Sea Surface Temperature Blended Analysis, Bulletin of the American Meteorological Society, 98(5): 1015-1026. https://doi.org/10.1175/BAMS-D-15-00002.1
- McClain, E.P., 1989. Global sea surface temperatures and cloud clearing for aerosol optical depth estimates, International Journal of Remote Sensing, 10(4-5): 763-769. https://doi.org/10.1080/01431168908903917
- Meissner, T., F.J. Wentz, and D. Draper, 2012. GMI calibration algorithm and analysis theoretical basis document, Remote Sensing Systems, Santa Rosa, CA, USA.
- O'Carroll, A.G., J.G. Watts, L.A. Horrocks, R.W. Saunders, and N.A. Rayner, 2006. Validation of the AATSR meteo product sea surface temperature, Journal of Atmospheric and Oceanic Technology, 23(5): 711-726. https://doi.org/10.1175/JTECH1876.1
- Park, K.A., E.Y. Lee, S.R. Chung, and E.H. Sohn, 2011. Accuracy assessment of sea surface temperature from NOAA/AVHRR data in the seas around Korea and error characteristics, Korean Journal of Remote Sensing, 27(6): 663-675. https://doi.org/10.7780/kjrs.2011.27.6.663
- Park, K.A., E.Y. Lee, X. Li, S.R. Chung, E.H. Sohn, and S. Hong, 2015. NOAA/AVHRR sea surface temperature accuracy in the East/Japan Sea, International Journal of Digital Earth, 8(10): 784-804. https://doi.org/10.1080/17538947.2014.937363
- Qiu, C. and H. Kawamura, 2012. Study on SST front disappearance in the subtropical North Pacific using microwave SSTs, Journal of Oceanography, 68(3): 417-426. https://doi.org/10.1007/s10872-012-0106-z
- Ricciardulli, L. and F.J. Wentz, 2004. Uncertainties in sea surface temperature retrievals from space: Comparison of microwave and infrared observations from TRMM, Journal of Geophysical Research: Oceans, 109(C12).
- Stammer, D., F. Wentz, and C. Gentemann, 2003. Validation of Microwave Sea Surface Temperature Measurements for Climate Purposes, Journal of Climate, 16(1): 73-87. https://doi.org/10.1175/1520-0442(2003)016<0073:VOMSST>2.0.CO;2
- Stogryn, A., 1967. The apparent temperature of the sea at microwave frequencies, IEEE Transactions on Antennas and Propagation, 15(2): 278-286. https://doi.org/10.1109/TAP.1967.1138900
- Meissner, T. and F.J. Wentz, 2012. The emissivity of the ocean surface between 6 and 90 GHz over a large range of wind speeds and earth incidence angles, IEEE Transactions on Geoscience and Remote Sensing, 50(8): 3004-3026. https://doi.org/10.1109/TGRS.2011.2179662
- Testut, C.-E., P. Brasseur, J.-M. Brankart, and J. Verron., 2003. Assimilation of Sea-Surface Temperature and Altimetric Observations during 1992-1993 into an Eddy Permitting Primitive Equation Model of the North Atlantic Ocean, Journal of Marine Systems, 40: 291-316.
- Walton, C.C., 2016. A Review of Differential Absorption Algorithms Utilized at NOAA for Measuring Sea Surface Temperature with Satellite Radiometers, Remote Sensing of Environment, 187: 434-446. https://doi.org/10.1016/j.rse.2016.10.011
- Wentz. F.J., 1983. A model function for ocean microwave brightness temperature, Journal of Geophysical Research: Oceans, 88(C3): 1892-1908. https://doi.org/10.1029/JC088iC03p01892
- Wentz, F.J., 1997. A well-calibrated ocean algorithm for special sensor microwave/imager, Journal of Geophysical Research: Oceans, 102(C4): 8703-8718. https://doi.org/10.1029/96JC01751
- Wentz, F.J., C. Gentemann, D. Smith, and D. Chelton, 2000. Satellite measurements of sea surface temperature through clouds, Science, 288(5467): 847-850. https://doi.org/10.1126/science.288.5467.847
- Wentz, F.J. and T. Meissner, 2000. Algorithm theoretical basis document (ATBD) version 2 AMSR ocean algorithm, Remote Sensing Systems, Santa Rosa, CA, USA.
- Wentz, F.J. and T. Meissner, 2007. Supplement 1 Algorithm Theoretical Basis Document for AMSR-E Ocena Algorithm, Remote Sensing Systems, Santa Rosa, CA, USA.
- Wentz, F.J. and T. Meissner, 2016. Atmospheric absorption model for dry air and water vapor at microwave frequencies below 100 GHz derived from spaceborne radiometer observations, Radio Science, 51(5): 381-391. https://doi.org/10.1002/2015RS005858
- Yan, Y., A. Barth, J.-M. Beckers, G. Candille, J.-M. Brankart, and P. Brasseur, 2015. Ensemble Assimilation of ARGO Temperature Profile, Sea Surface Temperature, and Altimetric Satellite Data into an Eddy Permitting Primitive Equation Model of the North Atlantic Ocean, Journal of Geophysical Research: Oceans, 120(7): 5134-5157. https://doi.org/10.1002/2014JC010349
- Yoshimori, K., K. Itoh, and Y. Ichioka, 1994. Thermal radiative and reflective characteristics of a wind-roughened water surface, JOSA A, 11(6): 1886-1893. https://doi.org/10.1364/JOSAA.11.001886
Cited by
- 한반도 서해 연안 해역에서의 해양 부이 관측 수온과 위성 마이크로파 관측 해수면온도의 비교 vol.39, pp.6, 2018, https://doi.org/10.5467/jkess.2018.39.6.555