1. Introduction
The variation of globally averaged sea surface temperature (SST) is one of the most-used indicators of Earth’s climate change due to the vast ocean surface area (IPCC 2013, 2019; EPA 2014; Karl et al. 2015; Fyfe et al. 2016). Climate variations over the global oceans are characterized by many SST modes such as El Niño–Southern Oscillation (ENSO), the Pacific decadal variability (PDV), the Atlantic multidecadal oscillation (AMO), the tropical Atlantic SST mode, and the Indian Ocean dipole (IOD; Philander 1990; Latif and Barnett 1994; Schlesinger and Ramankutty 1994; Mehta 1998; Saji et al. 1999). A reliable SST product is critical to many applications in ocean data assimilation, atmospheric simulation, ocean prediction, climate monitoring and assessment, future climate projection, and calibration of satellite observations (Saravanan 1998; Czaja and Frankignoul 1999; Goddard and Mason 2002; Liu et al. 2006; Schubert et al. 2009; Ashfaq et al. 2011; Liang et al. 2019; Iizuka and Nakamura 2019; Dragaud et al. 2019; Aumann et al. 2020; Ciani et al. 2020).
The reliability of SST products strongly depends on the availability of SST observations, among other factors. In situ SST observations are available as early as 1772 in the International Comprehensive Ocean–Atmosphere Dataset (ICOADS; Freeman et al. 2017). At that time, SST observations were used by commercial sailing ships to locate the Gulf Stream (Franklin et al. 1768; Richardson 1980; Emery 2003). However, SST observations made before the 1950s had two shortcomings: 1) large systematic biases and 2) low spatial coverage over the global oceans (Folland and Parker 1995; Kennedy et al. 2011a, 2011b, 2019; Huang et al. 2015a). Since the 1950s, these shortcomings have been greatly reduced.
Since the 1850s, many in situ SST data products have been developed for climate and weather-related research and applications. Examples of well-known in situ SST products include the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed SST (ERSST) in monthly 2° × 2° resolution starting from 1854 (Smith et al. 1996; Smith and Reynolds 2003, 2004; Huang et al. 2015a, 2017, 2020a); the U.K. Met Office Hadley SST (HadSST) in monthly 5° × 5° resolution starting from 1850 (Kennedy et al. 2011a,b, 2019); the Hadley Ice and SST (HadISST) in monthly 1° × 1° resolution starting from 1870 (Rayner et al. 2003); and the Japan Meteorological Office Centennial Observation-Based Estimates of SSTs (COBE-SST) in daily 1° × 1° resolution starting from 1850 (Ishii et al. 2005; Hirahara et al. 2014).
Since the early 1980s, satellite observations have been providing the possibility of global high-resolution SST products in daily 0.25° or finer resolutions. However, satellite-based SST observations may exhibit biases due to instrumental aging and/or contaminations of clouds and atmospheric aerosols (Zhang et al. 2004); biases are generally adjusted using in situ SSTs with various methods (Reynolds et al. 2007; Brasnett 2008; Merchant et al. 2014; Maturi et al. 2017; Good et al. 2020). Examples of well-known satellite-based SST products are the NOAA Daily Optimum Interpolation SST (DOISST) in 0.25° resolution starting from 1981 (Reynolds et al. 2007; Huang et al. 2021), the U.K. Met Office Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) in 0.05° resolution starting from 1981 (Stark et al. 2007; Donlon et al. 2012; Good et al. 2020), and the European Space Agency (ESA) Climate Change Initiative (CCI) SST in 0.05° resolution (Merchant et al. 2014, 2019). Among these products, CCI uses pure satellite-based observations without explicitly blending in situ observations, while the other SST products homogenize the satellite and in situ observations and blend them together.
SST products were assessed commonly by intercomparisons against independent observations such as those from Argo floats or the ensemble median of SST products in regional and global oceans (Barton 2007; Iwasaki et al. 2008; Xie et al. 2008; Martin et al. 2012; Huang et al. 2019; Fiedler et al. 2019; Woo and Park 2020; Yang et al. 2021). The intercomparison system (https://www.star.nesdis.noaa.gov/socd/sst/squam) noted a declined quality of DOISST v2.0 after 2016. DOISST has now been upgraded to v2.1 (Huang et al. 2021) to improve DOISST quality.
This study is to assess the quality of DOISST v2.1 after 2016, particularly its spatial and temporal structures of biases when compared to other similar available SST products. This is important as this topic has been discussed in several GHRSST meetings. Section 2 (data and methods) describes the eight commonly used daily SST products, SST observations from buoys, Argo floats, and buoys specially designed for the Upper Temperature of the Polar Oceans (UpTempO; Steele et al. 2017). Section 3 (intercomparisons) is an assessment of these eight products against buoy and Argo observations and an evaluation of those eight products against independent buoy, Argo, and UpTempO observations over the global oceans from January 2016 to June 2020. Section 4 (discussion) explores the reasons for the resulting differences between the SST products and observations, and the approach needed to provide reliable evaluation of SST products when all or almost all in situ data are ingested. The conclusions of the study are presented in section 5.
2. Data and methods
a. Data from eight SST products
1) DOISST
The NOAA DOISST (Table 1) is a global daily product with a resolution of 0.25° starting from 1981 (Reynolds et al. 2007; Huang et al. 2021). DOISST blends in situ measurements and satellite-derived observations from the Advanced Very High Resolution Radiometer (AVHRR). The AVHRR SSTs are adjusted to the buoy SSTs at the nominal depth of 0.2 m (Reynolds et al. 2007; Huang et al. 2013, 2015b). In ice-covered regions, the SST proxy from ice concentration is blended with SSTs from ships, buoys, and AVHRR, if available.
Daily SST datasets (from January 2016 to July 2020) used in this study (all data were downloaded on 15 August 2020).
DOISST has been updated from v2.0 to v2.1 from January 2016 and onward, while data remain unchanged from 1981 to 2015. The updates include (Huang et al. 2021) the following:
Satellite NOAA-19 is replaced by MetOp-B; MetOp-A remains unchanged (MetOp-A and MetOp-B are European polar-orbiting meteorological satellites).
The SST proxy using the regression between ice concentration and SST is replaced by using the freezing-point temperature in ice-covered oceans (Banzon et al. 2020).
The estimated ship SST bias is reduced from 0.14° to 0.01°C. The biases of 0.14° and 0.01°C are based on observations in periods 1982–2000 and 2016–19, respectively.
Ship and buoy observations from Daily ICOADS (ICOADS-D) Release 3.0.2 (R3.0.2; Liu et al. 2020) are used instead of NOAA National Centers for Environmental Prediction (NCEP) Global Telecommunications System (GTS) receipts.
Argo float observations (Argo 2000; Roemmich et al. 2001) above 5-m depth are included to ensure the best quality of SST production using all available in situ observations.
Note: Argo float observations are first used as independent data to validate the improvements in the updates from steps 1–4; in step 5 they are included in DOISST in operational production.
To assess the quality of DOISST, additional experiments (DOISST_Buoy90 and DOISST_Argo90; Table 1) were designed following Reynolds et al. (2002). DOISST_Buoy90 is the same as DOISST except that 90% of buoy drifters (Buoy90) are randomly selected and ingested and the remaining 10% of buoy drifters (Buoy10) are reserved for independent verification and intercomparison. Similarly, DOISST_Argo90 is the same as DOISST except that 90% of Argo drifters (Argo90) are randomly selected and ingested and the remaining 10% of Argo floats (Argo10) are reserved for verification. It is assumed that the reserved Buoy10 and Argo10 drifters/floats are independent from the Buoy90 and Argo90 used in experiments DOISST_Buoy90 and DOISST_Argo90, respectively.
2) MUR25
The NASA Multiscale Ultrahigh Resolution (MUR) v4.1 analysis is a daily SST product in 0.01° resolution starting from 2002 (Chin et al. 2017). MUR uses wavelets as basis functions in an optimal interpolation approach. MUR v4.1 includes nighttime SSTs derived from AVHRR, Advanced Microwave Scanning Radiometer-EOS (AMSR-EOS), AMSR2, the Moderate Resolution Imaging Spectroradiometers (MODIS), the U.S. Navy microwave WindSat radiometer, and in situ SST observations from the NOAA iQuam project (Xu and Ignatov 2010). The iQuam SSTs include observations from ships, drifting and moored buoys, and Argo floats. Ship and buoy observations in iQuam are from ICOADS (Freeman et al. 2017) and the U.S. Global Ocean Data Assimilation Experiment/Fleet Numerical Meteorology and Oceanography Center (FNMOC). Biases in each satellite sensor are adjusted after the differences between the retrieved and in situ SSTs are assessed. MUR25 v4.1 data are available at https://podaac.jpl.nasa.gov/dataset/MUR-JPL-L4-GLOB-v4.1. For method comparisons, the coarser-resolution (0.25°) version MUR v4.2 (MUR25; Table 1) is used in this study.
3) GMPE
The Group High Resolution SST (GHRSST) Multi-Product Ensemble (GMPE; Table 1) data are a daily near-real-time SST product with a horizontal resolution of 0.25° in latitude and longitude, starting from 2009. The GMPE selects the median SST from the GHRSST products (http://ghrsst-pp.metoffice.gov.uk/ostia-website/gmpe-monitoring.html; Martin et al. 2012; Dash et al. 2012; Fiedler et al. 2019). The current GHRSST products include CCI, OSTIA, Canada Meteorological Center (CMC) SST (Brasnett 1997, 2008; Brasnett and Colan 2016), NOAA DOISST, HadISST (Rayner et al. 2003; Titchner and Rayner 2014), and Japan Meteorological Agency Merged Global Daily SST (MGDSST; Kurihara et al. 2006). Because selecting the median produces a less-biased SST product, GMPE has frequently been used as a reference to assess the performance of available SST products (Yang et al. 2021, and references therein). GMPE v2 (2016) and v3 (2017–2020) are used for method comparisons in this study. The use of DOISST in GMPE may result in GMPE’s dependence on Argo floats and other in situ observations.
4) GAMSSA
The Bureau of Meteorology (BoM) Global Australian Multi-Sensor SST Analysis v1 (GAMSSA v1; Table 1) is a daily data product produced by optimum interpolation in 0.25° resolution starting from 2008 (Zhong and Beggs 2008; Beggs et al. 2011, 2020). GAMSSA v1 uses SST data derived from AVHRR, the Advanced Along Track Scanning Radiometer (AATSR), the AMSR2, and in situ SST observations from ships as well as drifting and moored buoys from GTS. Biases in AVHRR and AMSR2 SSTs are adjusted using drifting buoy SSTs. The skin SST data derived from AATSR are first converted to the foundation SST (Donlon et al. 2002) and then merged with other SST data.
5) OSTIA
The U.K. Met Office OSTIA v2 (Table 1) produces global daily SST and ice concentration data using an optimum interpolation method in 0.05° resolution starting from 2006 (Stark et al. 2007; Donlon et al. 2012; Good et al. 2020). OSTIA v2 includes satellite SSTs derived from AVHRR, AMSR2, the Visible Infrared Imager Radiometer Suite (VIIRS), the Sea and Land Surface Temperature Radiometer (SLSTR), the Spinning Enhanced Visible and Infrared Imager (SEVIRI), and in situ SSTs from ships as well as drifting and moored buoys. The ship and buoy SSTs are from the World Meteorological Organization’s (WMO) GTS. SSTs from drifting and moored buoys and VIIRS nighttime SSTs are used to adjust the biases in other satellite-derived SSTs using matchups within 25 km and 1 day.
6) GPB
The NOAA Geo-Polar Blended v1 (GPB; Table 1) is a global daily SST product in 0.05° resolution starting from 2014 (Maturi et al. 2017). GPB v1 includes only nighttime SSTs derived from AVHRR, VIIRS, the Geostationary Operational Environmental Satellite (GOES) imager, the Japanese Advanced Meteorological Imager (JAMI), and in situ SSTs from ships and NOAA iQuam drifting and moored buoys (Xu and Ignatov 2010). The ship and buoy observations in iQuam are from ICOADS (Freeman et al. 2017) and the U.S. FNMOC. GBP v1 employs a rigorous, multiscale, optimum interpolation methodology and a data-adaptive correlation length scale to reduce noises. Biases in satellite-derived SSTs are first corrected by regressing them to in situ SSTs, then by the difference between satellite and GPB analysis of the previous day, and finally adjusted by an independent NCEP SST product of Thiébaux et al. (2003) to avoid long-term drift of GPB. It should be noted that the biases in satellite SSTs in Thiébaux et al. (2003) are adjusted by the SST difference within a 7-day running window between satellite and in situ ship and buoy observations, which is similar to that applied in DOISST.
7) CCI
The ESA CCI is a daily SST product in 0.05° resolution (Merchant et al. 2014; 2019). CCI applies a variational assimilation scheme to produce a gap-filled estimate of daily mean SST. CCI v2.0 is available from 1981 to 2019, and v2.1 is available from 1981 to 2016 (http://dap.ceda.ac.uk/neodc/esacci/sst/data/CDR_v2/Analysis/L4/v2.1). In this study, v2.0 data from 2016 to 2019 (Table 1) are used for comparison. The CCI SST provides the mean SST at 0.2 m depth, which is close to the nominal depth of drifting buoy measurements. The CCI SST includes both AVHRR (satellites NOAA-7, NOAA-9, NOAA-11, NOAA-12, and NOAA-14–19) and Along-Track Scanning Radiometer (ATSR) series (ATSR, ATSR2, and AATSR). The biases in satellite observations are adjusted by recalibrating radiances using a reference channel. Therefore, the CCI SST is not explicitly dependent on in situ observations (Merchant et al. 2014). However, numerical weather prediction (NWP) fields from the European Centre for Medium-Range Weather Forecasting (ECMWF) interim reanalysis (ERA-Interim; Dee et al. 2011) are used as auxiliary information for cloud detection and retrieval, which may result in an implicit dependence of CCI SST on in situ observations.
8) CMC
The CMC v3 (Table 1) is a daily SST in 0.1° resolution starting from 2016 to present (Brasnett 1997, 2008; Brasnett and Colan 2016). The early version CMC v2 from 1991 to 2017 is available at https://podaac.jpl.nasa.gov/dataset/CMC0.2deg-CMC-L4-GLOB-v2.0. CMC v3 merges AVHRR SSTs from satellites NOAA-18 and NOAA-19, MetOp-A and MetOp-B, AMSR-EOS, and in situ SSTs from ships and drifting buoys of ICOADS. Biases in satellite observations are estimated from the differences between satellite and in situ pairs within an area of 5° in latitude and 10° in longitude, while the pairs are matched within 25 km. The median difference within the 5° × 10° area is selected to adjust the biases in the satellite observations.
b. In situ data
1) Buoy and Argo SSTs
Drifting and moored buoys at the ocean surface measure the SSTs at depth of 0.2–1.0 m (Castro et al. 2012). The temperature measurements of Argo floats above 5-m depth are used as SST observations in SST analyses and/or evaluations (Roemmich et al. 2015; Huang et al. 2017, 2021). Buoy SSTs are ingested into seven out of the eight products except for CCI, and Argo SSTs are used in DOISST v2.1 and MUR25 (Table 1). In this study, both buoy and Argo SSTs are first used to assess the eight SST products we examine. To further evaluate DOISST v2.1, the Buoy90 and Argo90 SSTs are included in experiments DOISST_Buoy90 and DOISST_Argo90, respectively, which are virtually the same as DOISST. The independent data Buoy10 and Argo10 SSTs are reserved to evaluate DOISST and assess the quality of all eight SST products. Buoy and Argo SSTs are first screened with quality control (QC) procedures that filter out the outliers deviated from the first guess by more than 4 times the standard deviation (STD) as described in Reynolds et al. (2007), and then compared against the eight SST products described in section 2a. The first guess at the current date is the sum of analysis at the previous date and climatological difference between the present and previous dates. The buoy and Argo SSTs are processed into daily 0.25° × 0.25° resolutions and compared against those eight SST products. It should be noted that buoy and Argo SSTs passed the same QC procedures whether they were ingested into the DOISST or used for evaluation purpose, which may give some trivial advantage in validating DOISST over other products. It is assumed that the reserved Buoy10 and Argo10 drifters are independent from the Buoy90 and Argo90, respectively.
2) Arctic buoy SSTs
The buoy SSTs from ICOADS in the Arctic region may be biased because 1) the SST thermistor sensors may be frozen, pushed up, and exposed to the air or 2) ICOADS provides SSTs from the measurements of the topmost thermistor (likely at 0.0-m depth), which may easily be frozen. To assess the SST products in the Arctic region, the SST data from the UpTempO project (Castro et al. 2016; Steele et al. 2017) are used in this study. The UpTempO collects SST measurements from specially designed buoys deployed in the Beaufort Sea and Hudson Bay from January 2016 to January 2019 (Fig. 1). To keep the SSTs from UpTempO observations independent from those of ICOADS and WMO GTS, UpTempO buoy measurements are searched from the second level (mostly at 2.5-m depth) down to 20-m depth and the first measurement from the thermistors submerged completely within water is selected as SST. The maximum depth of 20 m is selected because normal UpTempO observations show that the observed temperature above 20 m in the Arctic is almost uniform. UpTempO SSTs are averaged into daily superobservations on 0.25° × 0.25° grids and then compared against the eight SST products described in section 2a.
Geographic locations of Level 2 UpTempO buoy observations (green dots). Each dot represents a daily average.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
c. Assessment methods
The eight products are first assessed by comparing drifting and moored buoys that are dependent on the eight products except for CCI, and then by comparing Argo floats that are independent from most of the eight products except for DOISST and MUR25. To assess the quality of DOISST using independent observations, 90% of the drifting buoy and Argo floats were ingested into the DOISST experiments while the other 10% were reserved for evaluation purposes as described in section 2a(1). It is assumed that the residual bias of satellite SST and the analysis bias are larger than in situ SST bias, and therefore the biases and uncertainties of measurements and samplings in O are not taken into account in our assessments in Eqs. (1)–(3).
3. Intercomparisons
a. Comparisons against buoy SSTs
The eight daily SST products are compared against buoy SSTs from January 2016 to June 2020. It should be noted the buoy SSTs have been ingested into and are therefore not independent from the eight SST products except for CCI. The SSTs from these eight products are first processed and box-averaged to 0.25° × 0.25° resolution if the original resolution is higher than 0.25° (Table 1). The averaged SST differences (or biases) against buoy SSTs according to Eq. (1) are calculated on 0.25° × 0.25° grids and displayed on 2° × 2° grids for visualization purposes (Fig. 2). Figure 2 shows that SSTs are dominantly cold-biased in the global oceans in most of the eight products except for MUR25, which is warm-biased. The magnitude of these biases is mostly at 0.1°–0.2°C, although biases at magnitude of 0.4°C are found in the region of the Gulf Stream except for DOISST. SSTs are cold-biased in most of the tropical oceans between 20°S and 20°N except for MUR25, in most of the Northern Hemisphere midlatitude (30°–60°N) oceans except for DOISST and MUR25, in the Southern Ocean south of 45°S in CCI, and in the South Pacific south of 45°S in OSTIA and GPB. In the Gulf Stream region, strong warm biases are found except for DOISST, which shows weak warm biases and GAMSSA, which shows strong cold biases. Warm biases dominate over most of the global oceans in MUR25. Warm biases are also found around Australia and the Southern Ocean southeast of Argentina and south of South Africa in all eight products, although the magnitude of these warm biases is relatively small in the Southern Ocean in DOISST. These biases are mostly significant at the 95% confidence level according to Eqs. (8)–(10), indicating a limited capability of these SST analyses in representing observations at local grid scale in those regions.
Average SST biases (°C; January 2016–June 2020) against buoy observations. The biases are stippled when they are significant at the 95% confidence level. (a) DOISST, (b) MUR25, (c) GMPE, (d) GAMSSA, (e) OSTIA, (f) GPB, (g) CCI, and (h) CMC.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
The globally averaged biases range from −0.08° to +0.02°C (Table 2). The bias in DOISST (−0.04°C) is in the middle of the range, indicating a good performance of DOISST in the aspect of biases against buoy SSTs due to the recent revision from v2.0 to v2.1 (Huang et al. 2020a,b). The warm bias in MUR25 is unique among the eight products, which might be associated with the unique use of microwave observations from MODIS.
Globally averaged biases and RMSDs (°C) of SST datasets against buoy and Argo observations. The asterisks indicate the independence of buoy and Argo observation to the SST products.
The performance of the eight SST products is stable during the period from January 2016 to June 2020, which is illustrated by the time series of globally averaged biases (Fig. 3a). However, variations in biases are notable. For example, biases vary from −0.06° to 0.00°C in DOISST, from −0.03° to 0.07°C in MUR25, and from −0.13° to −0.02°C in GAMSSA. The biases are large in the Northern Hemisphere summers (May–July) of 2017–20 but smaller in the summer of 2016, which can be seen clearly from the evolution of GMPE (Fig. 3a, solid black). The stronger cold biases during the summers may result from the biases in satellite measurements due to higher cloudiness and dust aerosols in the tropical oceans (Zhang et al. 2004).
(a) Collocated and globally averaged SST biases of DOISST (solid red), MUR25 (dashed blue), GMPE (solid black), GAMSSA (dotted green), OSTIA (dotted black), GPB (solid light green), CCI (solid purple), and CMC (dotted orange) against buoy observations. (b) As in (a), but against Argo observations. A 15-day filter is applied to all curves for readability.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
The transient variations of biases in Fig. 3a can be quantified by their STDs. These STDs are about 0.02°–0.03°C, which are much smaller (approximately 65% or less) than the mean biases (0.02°–0.07°C) except for MUR25 (approximately 94%). The smaller STDs suggest that the errors in the eight SST products are mostly attributed to the mean or systematic biases rather than transient or random variability. This indicates that reduction of the satellite biases should be the focus in the future improvements of these eight products. Our tests (Table S1 in the online supplemental material) show that the bias differences relative to DOISST and MUR25 according to Eqs. (3) and (8)–(10) are statistically significant at the 95% confidence level. In contrast, the bias differences may not be significant among GMPE, GAMSSA, GPB, CCI, and CMC, which is consistent with the time series shown in Fig. 3a.
Despite the varied spatial distributions of biases in the eight SST products shown in Fig. 2, the spatial distributions of RMSDs according to Eq. (2) are rather similar (Fig. 4). RMSDs are less than 0.4°C in most of the global oceans, particularly the tropical oceans between 20°S and 20°N. However, high RMSDs above 1°C are found along the Gulf Stream and the Kuroshio and their extensions, and in the Southern Ocean southeast of Argentina, south of South Africa, and the sector of the Indian Ocean. The high RMSDs may directly result from mismatches between in situ and satellite observations in these regions. The RMSDs in DOISST are relatively small (<0.4°C) in most of the global oceans. The lower RMSDs in DOISST may indicate 1) the role of Argo in increasing SST quality in DOISST, 2) the role of the algorithm in correcting the biases of satellite SSTs in 2° × 2° grids and 15-day data window as described later in section 4a, and 3) a potential overfitting to the buoy SSTs in DOISST. In contrast, the RMSDs in MUR25, which also ingests Argo SSTs, are higher. The higher RMSDs may result from the facts that 1) the quality-control (QC) procedures applied to Argo and buoy SSTs are the same in DOISST, which may differ from the QC procedures in MUR25, and 2) Argo SSTs in DOISST are defined as the temperatures within a 0–5-m depth, while temperatures closest to the surface were used in MUR25 (Xu and Ignatov 2016). The globally averaged RMSDs are 0.28°–0.41°C (Table 2). The average RMSD in DOISST (0.28°C) lies in the lower end of the range, indicating that performance of DOISST is good among the eight products.
RMSDs of SSTs (°C; January 2016–June 2020) against buoy observations. (a) DOISST, (b) MUR25, (c) GMPE, (d) GAMSSA, (e) OSTIA, (f) GPB, (g) CCI, and (h) CMC.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
It should be noted that the RMSDs described above does not include the biases and uncertainties of measurements and samplings in in situ SSTs. The magnitude of RMSDs may change when the biases and uncertainties of the referenced observations are considered, but their impact to the RMSDs would be the same for all products for a given reference.
b. Comparisons against Argo SSTs
Argo SSTs are independent from most of the eight SST products except for DOISST and MUR25 (Table 1). The comparisons of the eight products against Argo SSTs from January 2016 to June 2020 (Fig. S1) show similar biases to those against buoy SSTs in Fig. 2. Cold biases are found over most of the global oceans in the eight products except for MUR25, which is warm biased. The RMSDs (Fig. S2) remain higher in the regions of the Gulf Stream and Kuroshio and their extensions, and higher in the Southern Ocean southeast of Argentina, south of South Africa, and the Indian Ocean sector. The time series of the globally averaged biases (Fig. 3b) are overall similar to those in Fig. 3a, but the cold biases become stronger in 2018–20 than in 2016–17 in DOISST, GMPE, OSTIA, and GPB, with reasons that are not immediately clear. The biases remain stable throughout the entire period of 2016–20 in MUR25, GAMSSA, CCI, and CMC.
The globally averaged biases and RMSDs against Argo SSTs are overall consistent with those in comparison against buoy SSTs (Table 2), which also show an as good performance of DOISST among the eight products. The differences of averaged biases in reference to DOISST and MUR25 remain significant at the 95% confidence level (Table S2).
One may argue that the good performance of DOISST is associated with using the dependent Argo observations. However, our comparisons show that the spatial distributions of biases and RMSDs are very similar in GAMSSA, OSTIA, GPB, CCI, and CMC when they are compared against the dependent buoy SSTs (Figs. 2 and 4) and the independent Argo SSTs (Figs. S1 and S2). In contrast, the magnitude of biases and RMSDs decreases slightly, which is counterintuitive as one may expect an overall increase of biases and RMSDs against the independent Argo SSTs. The lower biases and RMSDs may suggest that they are largely determined by the large-scale features such as nonlocal bias correction algorithms applied to the satellite observations, and are less determined by whether the reference is dependent.
c. Comparison against independent buoy and Argo SSTs
The performance of DOISST is further assessed using experiments DOISST_Buoy90 and DOISST_Argo90. The Buoy90 and Argo90 SSTs from January 2016 to June 2020 were ingested into experiments DOISST_Buoy90 and DOISST_Argo90 (Table 1), and the independent Buoy10 and Argo10 SSTs were reserved for evaluation purposes, respectively. Comparisons indicate that DOISST_Buoy90 and DOISST_Argo90 are almost identical to DOISST (not shown in figures), and therefore we will simply refer to DOISST_Buoy90 and DOISST_Argo90 as “DOISST” for the convenience of description.
The biases against Buoy10 SSTs (Fig. 5) show spatial patterns similar to those against the full buoy SSTs in Fig. 2. Their spatial pattern correlations according to Eq. (4) are larger than 0.85 in the eight products. The globally averaged biases range from −0.08° to 0.04°C (Table 3). These biases remain close to those against the full buoy SSTs in Table 2, and therefore the differences of biases in reference to DOISST and MUR25 remain significant at 95% confidence level (Table S3).
Average biases of SSTs (°C) against the independent Argo10 SSTs. The biases are stippled when they are significant at the 95% confidence level. (a) DOISST_Argo90, (b) MUR25, (c) GMPE, (d) GAMSSA, (e) OSTIA, (f) GPB, (g) CCI, and (h) CMC.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
Globally averaged biases and RMSDs (°C) of SST datasets against Argo10 SSTs. The asterisks indicate the independence of buoy and Argo observation to the SST products.
In addition to the similarity of biases, the RMSDs against the Buoy10 SSTs (Fig. 6) also show a high similarity to those against the full buoy SSTs in Fig. 4. Their spatial pattern correlation coefficients are greater than 0.90 in the eight products. The globally averaged RMSDs are 0.28°–0.35°C (Table 3). These biases and RMSDs are slightly changed in comparison with those against the full buoy SSTs, probably due to the reduced sampling sizes that may not well represent the global oceans. The exception is that the RMSD in DOISST increases slightly from 0.28° to 0.31°C when the independent Buoy10 SSTs are used for evaluation. Overall, although the buoy SSTs ingested into DOISST is reduced by 10%, the performance of DOISST remains good among the eight products.
RMSDs of SSTs (°C) against the independent Argo10 SSTs. (a) DOISST_Argo90, (b) MUR25, (c) GMPE, (d) GAMSSA, (e) OSTIA, (f) GPB, (g) CCI, and (h) CMC.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
The impact of the sampling size can be seen more clearly when DOISST ingests the Argo90 SSTs in experiment DOISST_Argo90 (Table 1), while the Argo10 SSTs are reserved for independent evaluation. In recent years, the typical number of Argo observations (approximately 1 × 103 per day or 280 surfacing Argo floats) is much less than that of buoy observations (approximately 5 × 104 per day or 1300 drifters) over the global oceans (Huang et al. 2019). Comparisons indicate that the similarity of spatial distributions of biases and RMSDs in reference to Argo10 and the full Argo (Figs. S1 and S3, and Figs. S2 and S4) is low, 0.44–0.56 for biases and 0.62–0.76 for RMSDs according to Eq. (4).
The similarity of RMSDs in DOISST is relatively lower (0.62) due to the higher RMSD in the Southern Ocean when the independent Argo10 SSTs are reserved for evaluation (Figs. S2a and S4a). The higher RMSD in DOISST in the Southern Ocean may be due to the fact that the in situ observations are sparse and therefore DOISST is more sensitive to the reservation of the Argo10 SSTs. This is another persuasive reason why DOISST includes all available observations in the operational production to improve the product quality, in particular for data-sparse regions. Overall, performance of DOISST in reference Argo10 is good among the eight products. The differences of biases in reference to DOISST and MUR25 remain significant except for GMPE (Table S4).
d. Comparison against independent UpTempO buoy SST in the Arctic region
The comparisons in sections 3a–3c do not include the Arctic Ocean because 1) the GMPE SST does not cover the Arctic and 2) buoy observations in ICOADS are from the topmost thermistor and may potentially be biased because the thermistor is exposed to the air by sea ice. Therefore, the independent SSTs from UpTempO project (Steele et al. 2017), collected from specially designed buoys released in the Beaufort Sea from January 2016 to January 2019 (Fig. 1), are used to assess the eight SST products in the Arctic region.
Comparisons of the eight SST products against UpTempO SSTs (Fig. 7) show that the biases are generally small (less than 0.5°C) during the wintertime (from November to May). SSTs in winter are mostly cold biased at the magnitude of −0.2°C. The biases in DOISST are very small, which may largely be due to the application of freezing-point SST proxy (Banzon et al. 2020). However, biases in summer (from June to August) are as large as 3°–4°C in all eight products, which may partly result from using nighttime satellite observations in MUR25 and GPB or using nighttime satellite observations to correct other satellites in OSTIA. Variations of biases are large in the eight products. For example, the magnitude of biases during the summer of 2017 is less than 0.2°C in DOISST but reaches 3°–4°C in OSTIA and GPB.
Collocated SST biases (°C) against independent Level 2 UpTempO buoy observations. (a) DOISST, (b) MUR25, (c) GMPE, (d) GAMSSA, (e) OSTIA, (f) GPB, (g) CCI, and (h) CMC. Each red dot represents one pair of daily averaged SSTs. The total number of pairs are shown in Table 5.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
The small biases during the boreal winter are associated with the constraint of freezing point in these eight products when ice concentration is high. The large biases during the boreal summer result from large variations of SSTs when ice concentration is low (Banzon et al. 2020), which makes it difficult to constrain the SST proxy from ice concentration. The averaged biases range from −0.22° to +0.11°C in the eight products, and the averaged RMSD are 0.42°–0.69°C (Table 4). The performance of DOISST in the Arctic region is good among the eight products in the aspect of bias, although its performance in RMSD is relatively worse. It should be noted that the bias and RMSD of GMPE may not be reliable due to its small sampling size.
Average biases and RMSDs (°C) of SST datasets against UpTempO Level-2 data in the Arctic (Fig. 7). The pair numbers are the counts of collocated data pairs between SST products and UpTempO.
4. Discussion
a. Causes for SST biases in DOISST
To track the source of biases in DOISST described in sections 3a to 3c, the satellite SSTs (MetOp-A and MetOp-B of daytime and nighttime) are compared against buoy SSTs from January 2016 to June 2020 (Figs. 8a–d). The comparisons show that the satellite SSTs exhibit warm biases north of 45°N and south of 40°S. The warm biases are larger in MetOp-B (0.4°C) than MetOp-A (0.2°C) for both daytime and nighttime. The warm biases in nighttime MetOp-A extend more broadly in the Southern Hemisphere oceans. Between 40°S and 45°N, the satellite SSTs are cold biased. The cold biases are larger in MetOp-B (−0.6°C) than MetOp-A (−0.2°C) for both the daytime and nighttime. However, the cold biases in nighttime MetOp-A are confined in the northwest Indian Ocean and tropical Atlantic, and are very weak in the tropical Pacific. On a global average, the magnitude of biases is much smaller due to cancellations of cold and warm biases in different regions. The globally averaged biases range from −0.11° to +0.02°C (Table 5).
Average biases of satellite SSTs (°C) against buoy observations: (a) daytime MetOp-A, (b) nighttime MetOp-A, (c) daytime MetOp-B, and (d) nighttime MetOp-B. (e)–(h) As in (a)–(d), but for the bias-adjusted satellite SSTs.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
Globally averaged biases and RMSDs (°C) of original and bias-adjusted daytime and nighttime satellite observations in comparison with buoy SSTs.
These biases in AVHRR SSTs are adjusted by in situ observations from ships, buoys, and Argo floats in DOISST via the following four steps (Reynolds et al. 2007; Huang et al. 2015b; Huang et al. 2017):
Daily AVHRR and in situ SSTs are separately bin-averaged to 2° × 2° grids, which increases the area coverages of in situ observations over the global oceans.
AVHRR and in situ SSTs are separately filtered by the decomposition of 130 empirical orthogonal teleconnection modes (EOTs). The EOTs are the localized empirical orthogonal functions that are damped to zero 3000/5000 km away from the mode centers in the latitude/longitude direction. The EOT decomposition is critically dependent on the area coverage of data (Huang et al. 2020b, 2021). EOT decomposition is used to filter out small spatial scale noises.
Differences between the decomposed AVHRR and in situ SSTs are calculated within a 15-day running window, which is to filter out noises in a short-time period.
The differences in step 3 are defined as AVHRR biases, which are interpolated back to DOISST 0.25° × 0.25° grids and then subtracted from AVHRR SSTs.
The accuracy of the bias adjustments described above is evaluated by comparing the adjusted AVHRR SSTs against buoy SSTs (Figs. 8e–h). It is clear that the warm biases decrease from 0.4° to 0.2°C in MetOp-B for both the daytime and nighttime south of 40°S and north of 45°N, and in nighttime MetOp-A in the Southern Hemisphere oceans. The cold biases between 40°S and 45°N decrease from −0.6° to −0.2°C in daytime MetOp-B and from −0.2° to −0.1°C in nighttime MetOp-B and daytime MetOp-A. However, the warm biases in the North Pacific north of 40°N increase slightly in MetOp-A for both the daytime and nighttime. The warm biases in the North Atlantic in nighttime MetOp-A are overadjusted and become cold biases, indicating the limitation of bias correction algorithms in DOISST. The globally averaged biases range from −0.04°C to −0.02°C (Table 5). The improvement in globally averaged biases is clear in daytime MetOp-B but very slight in daytime MetOp-A. The globally averaged biases even become slightly larger in nighttime MetOp-A and MetOp-B.
These remaining residual biases in the adjusted AVHRR SSTs have the following two characteristics: 1) the magnitude of the globally averaged biases is small (from −0.02° to −0.04°C; Table 5), which matches with the final DOISST bias (−0.04°C; Table 2), and 2) the spatial distributions of the AVHRR biases in Figs. 8e–h are similar to that of DOISST in Fig. 2a. The spatial correlation coefficients between Figs. 8e–h and 2a range from 0.50 to 0.58. These features suggest that the residual biases from the adjusted AVHRR SSTs are a major source contributing to the final DOISST biases. Therefore, the future development of DOISST should focus on removing these residual biases by improving the bias-adjustment algorithms.
The contribution of the residual biases of the adjusted AVHRR SSTs to the final DOISST biases can also be seen from the RMSDs against buoy SSTs (Fig. 9). Figures 9a–d show that the RMSDs are large (0.6°C) in MetOp-A and MetOp-B for both daytime and nighttime in the regions of the Gulf Stream and the Kuroshio and their extensions, and the Southern Ocean south of 40°S. The RMSDs are also large (from 0.4° to 0.6°C) in daytime MetOp-A and MetOp-B in the tropical oceans. The globally averaged RMSDs range from 0.52° to 0.62°C (Table 5).
RMSDs of satellite SSTs (°C) against buoy observations: (a) daytime MetOp-A, (b) nighttime MetOp-A, (c) daytime MetOp-B, and (d) nighttime MetOp-B. (e)–(h) As in (a)–(d), but for the bias-adjusted satellite SSTs.
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
The RMSDs of AVHRR SSTs mostly remain after the bias adjustment (Figs. 9e–h), except that the RMSD of the adjusted AVHRR SST from daytime MetOp-B decreases substantially in the tropical oceans and Southern Ocean. The spatial distributions of these RMSDs are very similar to that of the final DOISST shown in Fig. 4a. The spatial correlation coefficients between Figs. 9e–h and 4a are 0.68–0.74, which indicate that the residual biases from the adjusted AVHRR SSTs contribute to the final DOISST biases. The globally averaged RMSDs of the adjusted AVHRR SSTs range from 0.54° to 0.58°C, practically showing no improvements over those of the original AVHRRs (Table 5). Nevertheless, these RMSDs are much higher than that of final DOISST (0.28°C; Table 2). The large contrast between the RMSDs in the adjusted AVHRR SSTs and final DOISST indicates that the noises in the adjusted AVHRR SSTs have been damped by in situ SSTs, which is another reason to include all available in situ SSTs such as Argo SSTs to improve the quality of SST products. Although these residual biases of satellite observations may also exist in other SST products, we were unable to assess this as these intermediate data are generally unavailable to the public.
Our analyses indicate that the spatial patterns and magnitude of biases and RMSD do not change much when these products are compared with buoy and Argo, or the 10% of reserved buoys or Argo floats. The difference of biases and RMSD among products are clearly seen. These results suggest that the biases in these products may directly be associated with the algorithms correcting the biases of satellite SSTs as indicated by our earlier studies (Huang et al. 2013, 2015b, 2016).
b. Independent observations
One of the challenges in assessing the performance of the eight SST products is the availability of independent in situ observations. It should be noted that the in situ observations were neither perfect in quality nor always consistent among different platforms. The observations must be checked by QC procedures regardless of whether they are to be ingested into or to validate the products. However, the QC procedures may differ among products and impact the number and area coverage of the observations, and their roles may differ among products.
On the one hand, we want to reserve independent observations for evaluations. For example, Argo observations have been reserved to independently evaluate SST productions in GAMSSA, OSTIA, GPB, CCI, and CMC. On the other hand, we want to use as many observations as possible to increase the reliability of SST products. For example, Argo observations are ingested into DOISST and MUR25 to best represent the SST analyses. However, the spatial distributions of biases and RMSDs against Argo observations are similar to those against buoy observations. The magnitude of biases against Argo is similar to that against buoy, while the magnitude of RMSD decreases slightly. These features are exhibited not only in GMPE, DOISST, and MUR25 where Argo observations are dependent, but also in GAMSSA, OSTIA, GPB, CCI, and CMC where Argo observations are independent. In other words, the smaller biases and RMSD in DOISST, MUR25, and GMPE may not necessarily result from comparing against dependent buoy and Argo SSTs.
The similar biases and RMSDs in GAMSSA, OSTIA, GPB, CCI, and CMC suggest that it is not necessary to reserve Argo observations purely for evaluation purposes. Inclusion of all high-quality in situ data (including Argo SSTs) is important to increase the quality of the SST products that utilize both in situ and satellite observations. The addition of Argo sampling can improve the coverage of in situ SSTs in some regions, which is important for satellite bias adjustments.
Our studies indicated that the DOISST biases and RMSDs mainly result from algorithms used for bias adjustment of satellite observations, while impacts are small from methods of blending in situ and satellite observations and from methods of interpolations (Huang et al. 2013, 2015b, 2016). Consistent with previous studies, our study indicates that the residual biases in the adjusted satellite-derived SSTs are the main contributor to the final biases in DOISST, which may also be true for other SST products. The residual biases are caused by imperfect matchups in most products or large-scale differences between in situ and satellite observations in DOISST.
Our analyses show that the residual biases are critically dependent on the coverages of in situ superposed observations (superobservations) (Fig. 10), which is attributed to the bias correction algorithms using EOTs. The coverages of in situ data (Fig. 10) are defined as a ratio between the counts of days with superobservations and days with or without superobservations from 1 June 2016 to 31 June 2020. The coverages are calculated on 2° × 2° grids, since biases of satellite observations are estimated on 2° × 2° grids. Figure 10f shows the difference between the coverages of blended ship + buoy + Argo and ship + buoy observations. The coverage difference highlights the role of Argo observations when they are ingested into SST analysis systems. The figure indicates an increase of coverage by 0.2–0.3 in the Southern Ocean, which is about 100% of ship + buoy coverage (Fig. 10d). Therefore, we can speculate that the matchups and therefore the overall performance would be notably improved in GAMSSA, OSTIA, GPB, and CMC if Argo observations were ingested, particularly in the Southern Ocean.
Coverage by observations of (a) ship, (b) buoy, (c) Argo, (d) ship + buoy, and (e) ship + buoy + Argo, using of 2° × 2° spatial resolution and 15-day data windows, January 2016–June 2020. (f) Coverage difference between (e) and (d).
Citation: Journal of Climate 34, 18; 10.1175/JCLI-D-21-0001.1
In DOISST, biases in satellite SSTs are estimated by the large-scale difference between in situ and satellite observations on 2° × 2° grids within the 15-day data window (Reynolds et al. 2007). Large-scale patterns of in situ and satellite SSTs are based on EOTs, which is sensitive to the coverage of in situ observation (Huang et al. 2019). The coverage of buoy SSTs is low in the Southern Ocean (Fig. 10b), and the total coverage of in situ SSTs (Figs. 10d,e) is sensitive to the addition of Argo observations. As a result, the estimation of biases in satellite SSTs and therefore the residual biases in the adjusted satellite SSTs are sensitive to the Argo10 SSTs. This may explain why the final biases and RMSDs become larger in the Southern Ocean when the Argo10 SSTs are reserved as evaluation data.
We want to note that the results presented in this study may differ from previous studies due to factors such as using different time periods, validation metrics, and validation datasets in assessments. Martin et al. (2012) showed that DOISST has a smaller mean bias but its standard deviation is large, which is consistent with our assessment. Fiedler et al. (2019) showed a large bias in DOISST v2.0, which is consistent with Huang et al. (2021), and a smaller bias in CMC, CCI, and GMPE, which is different from our assessment. Yang et al. (2021) showed an overall good performance of CCI and OSTIA and an intermediate performance of DOISST v2.1 and MUR25, which is different from our assessment.
5. Conclusions
Our assessments of the eight SST products indicate that DOISST v2.1 has a good performance in global-averaged biases and RMSDs in reference to buoy and Argo observations, as well as in reference to the independent Buoy10 and Argo10 SSTs. MUR25 has warm biases, while other seven products have cold biases. The differences of biases in reference to DOISST and MUR25 are statistically significant, while the differences of biases among GMPE, GAMSSA, OSTIA, GPB, CCI, and CMC are less significant. Our comparisons indicate that the quality of SST products may be improved if all in situ observations are included. This is consistent with developments of DOISST from v2.0 to v2.1, as after the inclusion of Argo data the differences relative to Argo data have decreased, and the fitting to regional structures of in situ data resulted in the higher similarity of DOISST spatial structures to those of in situ data. For methodology and product development research, one may resort to reserving some observations (such as Argo floats) as independent evaluation sets. However, for products that depend on in situ observation for satellite SST bias corrections (such as in DOISST), the operational production should utilize all good quality data to provide best quality product to users, in particular as there are still data-sparse regions as of today (e.g., the Southern Ocean region). This is akin to the manufacturing industry—manufacturers use the best available materials to produce the best quality products for customers, not just reserving the best materials for product evaluation purpose (as was done at an earlier experimental stage).
Acknowledgments
The authors thank four anonymous reviewers for their constructive comments that greatly improved the manuscript. Authors appreciate comments from Anthony Arguez and Xuepeng Zhao and English proofreading from Andrea Anderson in improving the manuscript. The sources of the eight SST products are listed in Table 1 and cited in the main text. ICOADS-D R3.0.2 data are provided at NOAA/NCEI (https://doi.org/10.7289/V5CZ3562), access date: 1 August 2020. MetOp-A and MetOp-B SST were provided by NOAA/NCEI Common Ingest system (
REFERENCES
Argo, 2000: Argo float data and metadata from Global Data Assembly Centre (Argo GDAC). SEANOE, accessed 1 August 2020, https://doi.org/10.17882/42182.
Ashfaq, M., C. B. Skinner, and N. S. Diffenbaugh, 2011: Influence of SST biases on future climate change projections. Climate Dyn., 36, 1303–1319, https://doi.org/10.1007/s00382-010-0875-2.
Aumann, H. H., S. E. Broberg, E. M. Manning, T. S. Pagano, and R. C. Wilson, 2020: 2020: Evaluating the absolute calibration accuracy and stability of AIRS using the CMC SST. Remote Sens., 12, 2743, https://doi.org/10.3390/rs12172743.
Banzon, V., T. M. Smith, M. Steele, B. Huang, H.-M. Zhang, 2020: Improved estimation of proxy sea surface temperature in the Arctic. J. Atmos. Oceanic Technol., 37, 341–349, https://doi.org/10.1175/JTECH-D-19-0177.1.
Barton, I. J., 2007: Comparison of in situ and satellite-derived sea surface temperatures in the Gulf of Carpentaria. J. Atmos. Oceanic Technol., 24, 1773–1784, https://doi.org/10.1175/JTECH2084.1.
Beggs, H., A. Zhong, G. Warren, O. Alves, G. Brassington, and T. Pugh, 2011: RAMSSA—An operational, high-resolution, regional Australian multi-sensor sea surface temperature analysis over the Australian region. Aust. Meteor. Oceanogr. J., 61, 1–22, https://doi.org/10.22499/2.6101.001.
Beggs, H., L. Qi, P. Govekar, and C. Griffin, 2020: Ingesting VIIRS SST into the Bureau of Meteorology’s Operational SST analyses. Proc. 21st GHRSST Science Team Meeting, virtual meeting, EUMETSAT, in press.
Brasnett, B., 1997: A global analysis of sea surface temperature for numerical weather prediction. J. Atmos. Oceanic Technol., 14, 925–937, https://doi.org/10.1175/1520-0426(1997)014<0925:AGAOSS>2.0.CO;2.
Brasnett, B., 2008: The impact of satellite retrievals in a global sea-surface-temperature analysis. Quart. J. Roy. Meteor. Soc., 134, 1745–1760, https://doi.org/10.1002/qj.319.
Brasnett, B., and D. S. Colan, 2016: Assimilating retrievals of sea surface temperature from VIIRS and AMSR2. J. Atmos. Oceanic Technol., 33, 361–375, https://doi.org/10.1175/JTECH-D-15-0093.1.
Castro, S. L., G. A. Wick, and W. J. Emery, 2012: Evaluation of the relative performance of sea surface temperature measurements from different types of drifting and moored buoys using satellite-derived reference products. J. Geophys. Res. Oceans, 117, C02029, https://doi.org/10.1029/2011JC007472.
Castro, S. L., G. A. Wick, and M. Steele, 2016: Validation of satellite sea surface temperature analyses in the Beaufort Sea using UpTempO buoys. Remote Sens. Environ., 187, 458–475, https://doi.org/10.1016/j.rse.2016.10.035.
Chin, T. M., J. Vazquez-Cuervo, and E. M. Armstrong, 2017: A multi-scale high-resolution analysis of global sea surface temperature. Remote Sens. Environ., 200, 154–169, https://doi.org/10.1016/j.rse.2017.07.029.
Ciani, D., M.-H. Rio, B. B. Nardelli, H. Etienne, and T. Santoleri, 2020: Improving the altimeter-derived surface currents using sea surface temperature (SST) data: A sensitivity study to SST products. Remote Sens., 12, 1601, https://doi.org/10.3390/rs12101601.
Czaja, A., and C. Frankignoul, 1999: Influence of the North Atlantic SST on the atmospheric circulation. Geophys. Res. Lett., 26, 2969–2972, https://doi.org/10.1029/1999GL900613.
Dash, P., and Coauthors, 2012: Group for high resolution sea surface temperature (GHRSST) analysis fields inter-comparisons—Part 2: Near real time web-based level 4 SST Quality Monitor (L4-SQUAM). Deep-Sea Res. II, 77–80, 31–43, https://doi.org/10.1016/j.dsr2.2012.04.002.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Donlon, C. J., P. Minnett, C. Gentemann, T. J. Nightingale, I. J. Barton, B. Ward, and J. Murray, 2002: Towards improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15, 353–369, https://doi.org/10.1175/1520-0442(2002)015<0353:TIVOSS>2.0.CO;2.
Donlon, C. J., M. Martin, J. Stark, J. Roberts-Jones, E. Fiedler, and W. Wimmer, 2012: The Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) system. Remote Sens. Environ., 116, 140–158, https://doi.org/10.1016/j.rse.2010.10.017.
Dragaud, I. C. D. V., M. Soares da Silva, L. P. de Freitas Assad, M. Cataldi, L. Landau, R. Nascimento Elias, and L. Pimentel, 2019: The impact of SST on the wind and air temperature simulations: A case study for the coastal region of the Rio de Janeiro state. Meteor. Atmos. Phys., 131, 1083–1097, https://doi.org/10.1007/s00703-018-0622-5.
Emery, W. J., 2003. Air–sea interaction/sea surface temperature. Encyclopedia of Atmospheric Sciences. R. H. James, Ed, Academic Press, 100–109, https://doi.org/10.1016/B0-12-227090-8/00065-8.
EPA, 2014: Climate change indicators in the United States, 2014. 3rd ed. Environmental Protection Agency Rep. EPA 430-R-14-004, 112 pp., https://www.epa.gov/sites/production/files/2016-07/documents/climateindicators-full-2014.pdf.
Fiedler, E. K., and Coauthors, 2019: Intercomparison of long-term sea surface temperature analyses using the GHRSST Multi-Product Ensemble (GMPE) system. Remote Sens. Environ., 222, 18–33, https://doi.org/10.1016/j.rse.2018.12.015.
Folland, C. K., and D. E. Parker, 1995: Correction of instrumental biases in historical sea surface temperature data. Quart. J. Roy. Meteor. Soc., 121, 319–367, https://doi.org/10.1002/qj.49712152206.
Franklin, B., T. Folger, E. Wright, E. Halley, H. Moll, J. Mount, and T. Page, 1768: Franklin-Folger chart of the Gulf Stream (London: sold by Jno. Mount and Tho. Page; map retrieved from the Library of Congress). https://www.loc.gov/item/88696412/.
Freeman, E., and Coauthors, 2017: ICOADS release 3.0: A major update to the historical marine climate record. Int. J. Climatol., 37, 2211–2232, https://doi.org/10.1002/joc.4775.
Fyfe, J. C., and Coauthors, 2016: Making sense of the early-2000s warming slowdown. Nat. Climate Change, 6, 224–228, https://doi.org/10.1038/nclimate2938.
Goddard, L., and S. Mason, 2002: Sensitivity of seasonal climate forecasts to persisted SST anomalies. Climate Dyn., 19, 619–631, https://doi.org/10.1007/s00382-002-0251-y.
Good, S., and Coauthors, 2020: The current configuration of the OSTIA system for operational production of foundation sea surface temperature and ice concentration analyses. Remote Sens., 12, 720, https://doi.org/10.3390/rs12040720.
Hirahara, S., M. Ishii, and Y. Fukuda, 2014: Centennial-scale sea surface temperature analysis and its uncertainty. J. Climate, 27, 57–75, https://doi.org/10.1175/JCLI-D-12-00837.1.
Huang, B., M. L’Heureux, J. Lawrimore, C. Liu, V. Banzon, Z.-Z. Hu, and A. Kumar, 2013: Why did large differences arise in the sea surface temperature datasets across the tropical Pacific during 2012? J. Atmos. Oceanic Technol., 30, 2944–2953, https://doi.org/10.1175/JTECH-D-13-00034.1.
Huang, B., and Coauthors, 2015a: Extended Reconstructed Sea Surface Temperature version 4 (ERSST.v4), Part I. Upgrades and intercomparisons. J. Climate, 28, 911–930, https://doi.org/10.1175/JCLI-D-14-00006.1.
Huang, B., W. Wang, C. Liu, V. F. Banzon, J. Lawrimore, and H.-M. Zhang, 2015b: Bias adjustment of AVHRR SST and its impacts on two SST analyses. J. Atmos. Oceanic Technol., 32, 372–387, https://doi.org/10.1175/JTECH-D-14-00121.1.
Huang, B., C. Liu, V. F. Banzon, H.-M. Zhang, T. R. Karl, J. H. Lawrimore, and R. S. Vose, 2016: Assessing the impact of satellite-based observations in sea surface temperature trends. Geophys. Res. Lett., 43, 3431–3437, https://doi.org/10.1002/2016GL068757.
Huang, B., and Coauthors, 2017: Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5), Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Huang, B., C. Liu, G. Ren, H.-M. Zhang, and L. Zhang, 2019: The role of buoy and Argo observations in two SST analyses in the global and tropical Pacific oceans. J. Climate, 32, 2517–2535, https://doi.org/10.1175/JCLI-D-18-0368.1.
Huang, B., and Coauthors, 2020a: Uncertainty estimates for sea surface temperature and land surface air temperature in NOAAGlobalTemp version 5. J. Climate, 33, 1351–1379, https://doi.org/10.1175/JCLI-D-19-0395.1.
Huang, B., M. L’Heureux, Z.-Z. Hu, X. Yin, and H.-M. Zhang, 2020b: How significant was the 1877–78 El Niño? J. Climate, 33, 4853–4869, https://doi.org/10.1175/JCLI-D-19-0650.1.
Huang, B., C. Liu, V. Banzon, E. Freeman, G. Graham, B. Hankins, T. Smith, and H.-M. Zhang, 2021: Improvements of the daily optimum interpolation sea surface temperature (DOISST) version 2.1. J. Climate, 34, 2923–2939, https://doi.org/10.1175/JCLI-D-20-0166.1.
Iizuka, S., and H. Nakamura, 2019: Sensitivity of mid-latitude heavy precipitation to SST: A case study in the area of Japan area on 9 August 2013. J. Geophys. Res., 124, 4365–4381, https://doi.org/10.1029/2018JD029503.
IPCC, 2013: Climate Change 2013: The Physical Science Basis. T. F. Stocker et al., Eds., Cambridge University Press, 1535 pp.
IPCC, 2019: Technical summary. IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. H.-O. Pörtner et al., Eds., in press.
Ishii, M., A. Shouji, S. Sugimoto, and T. Matsumoto, 2005: Objective analyses of sea-surface temperature and marine meteorological variables for the 20th century using ICOADS and the Kobe Collection. Int. J. Climatol., 25, 865–879, https://doi.org/10.1002/joc.1169.
Iwasaki, S., M. Kubota, and H. Tomita, 2008: Inter-comparison and evaluation of global sea surface temperature products. Int. J. Remote Sens., 29, 6263–6280, https://doi.org/10.1080/01431160802175363.
Karl, T. R., and Coauthors, 2015: Possible artifacts of data biases in the recent global surface warming hiatus. Science, 348, 1469–1472, https://doi.org/10.1126/science.aaa5632.
Kennedy, J. J., N. A. Rayner, R. O. Smith, D. E. Parker, and M. Saunby, 2011a: Reassessing biases and other uncertainties in sea surface temperature observations measured in situ since 1850: 1. Measurement and sampling errors. J. Geophys. Res., 116, D14103, https://doi.org/10.1029/2010JD015218.
Kennedy, J. J., N. A. Rayner, R. O. Smith, D. E. Parker, and M. Saunby, 2011b: Reassessing biases and other uncertainties in sea surface temperature observations measured in situ since 1850: 2. Biases and homogenization. J. Geophys. Res., 116, D14104, https://doi.org/10.1029/2010JD015220.
Kennedy, J. J., N. A. Rayner, C. P. Atkinson, and R. E. Killick, 2019: An ensemble data set of sea surface temperature change from 1850: The Met Office Hadley Centre HadSST.4.0.0.0 data set. J. Geophys. Res., 124, 7719–7763, https://doi.org/10.1029/2018JD029867.
Kurihara, Y., T. Sakurai, and T. Kuragano, 2006: Global daily sea surface temperature analysis using data from satellite microwave radiometer, satellite infrared radiometer and in-situ observations (in Japanese). Wea. Service Bull., 73, 1–18.
Latif, M., and T. P. Barnett, 1994: Causes of decadal climate variability over the North Pacific and North America. Science, 266, 634–637, https://doi.org/10.1126/science.266.5185.634.
Liang, X., M. Losch, L. Nerger, L. Mu, Q. Yang, and C. Liu, 2019: Using sea surface temperature observations to constrain upper ocean properties in an Arctic sea ice-ocean data assimilation system. J. Geophys. Res. Oceans, 124, 4727–4743, https://doi.org/10.1029/2019JC015073.
Liu, C., and Coauthors, 2020: Blending BUFR and TAC marine in situ data for ICOADS near-real-time release 3.0.2. [These are raw data that may be updated.] ftp://ftp.ncei.noaa.gov/pub/data/cmb/ersst/v5/tmp/icoads.3.0.2.
Liu, Q., N. Wen, and Z. Liu, 2006: An observational study of the impact of the North Pacific SST on the atmosphere. Geophys. Res. Lett., 33, L18611, https://doi.org/10.1029/2006GL026082.
Martin, B., and Coauthors, 2012: Group for high resolution sea surface temperature (GHRSST) analysis fields inter-comparisons. Part 1: A GHRSST multi-product ensemble (GMPE). Deep-Sea Res. II, 77–80, 21–30, https://doi.org/10.1016/j.dsr2.2012.04.013.
Maturi, E., A. Harris, J. Mittaz, J. Sapper, G. Wick, X. Zhu, P. Dash, and P. Koner, 2017: A new high-resolution sea surface temperature blended analysis. Bull. Amer. Meteor. Soc., 98, 1015–1026, https://doi.org/10.1175/BAMS-D-15-00002.1.
Mehta, V. M., 1998: Variability of the tropical ocean surface temperatures at decadal–multidecadal timescales. Part I: The Atlantic Ocean. J. Climate, 11, 2351–2375, https://doi.org/10.1175/1520-0442(1998)011<2351:VOTTOS>2.0.CO;2.
Merchant, C. J., and Coauthors, 2014: Sea surface temperature datasets for climate applications from Phase 1 of the European Space Agency Climate Change Initiative (SST CCI) starting 1981. Geosci. Data J., 1, 179–191, https://doi.org/10.1002/gdj3.20.
Merchant, C. J., and Coauthors, 2019: Satellite-based time-series of sea-surface temperature since 1981 for climate applications. Nature Sci. Data, 6, 223, https://doi.org/10.1038/s41597-019-0236-x.
Philander, S. G., 1990: El Niño, La Niña, and the Southern Oscillation. Academic Press, 293 pp.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.
Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609–1625, https://doi.org/10.1175/1520-0442(2002)015,1609:AIISAS.2.0.CO;2.
Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, and M. G. Schlax, 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 5473–5496, https://doi.org/10.1175/2007JCLI1824.1.
Richardson, P. L., 1980: The Benjamin Franklin and Timothy Folger Charts of the Gulf Stream. Oceanography: The Past. M. Sears and D. Merriman, Eds., Springer, 703–717.
Roemmich, D., and Coauthors, 2001: The global array of profiling floats. Observing the Ocean in the 21st Century, C. J. Koblinsky and N. R. Smith, Eds., Australian Bureau of Meteorology, 248–258.
Roemmich, D., J. Church, J. Gilson, D. Monsellesan, P. Sutton, and S. Wijffels, 2015: Unabated planetary warming and its ocean structure since 2006. Nat. Climate Change, 5, 240–245, https://doi.org/10.1038/nclimate2513.
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363, https://doi.org/10.1038/43854.
Saravanan, R., 1998: Atmospheric low-frequency variability and its relationship to mid-latitude SST variability: Studies using the NCAR Climate System Model. J. Climate, 11, 1386–1404, https://doi.org/10.1175/1520-0442(1998)011<1386:ALFVAI>2.0.CO;2.
Schlesinger, M. E., and N. Ramankutty, 1994: An oscillation in the global climate system of period 65–70 yr. Nature, 367, 723–726, https://doi.org/10.1038/367723a0.
Schubert, S., and Coauthors, 2009: A U.S. CLIVAR project to assess and compare the responses of global climate models to drought-related SST forcing patterns: Overview and results. J. Climate, 22, 5251–5272, https://doi.org/10.1175/2009JCLI3060.1.
Smith, T., and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperature based on COADS data (1854–1997). J. Climate, 16, 1495–1510, https://doi.org/10.1175/1520-0442-16.10.1495.
Smith, T., and R. W. Reynolds, 2004: Improved extended reconstruction of SST (1854–1997). J. Climate, 17, 2466–2477, https://doi.org/10.1175/1520-0442(2004)017<2466:IEROS>2.0.CO;2.
Smith, T., R. W. Reynolds, R. E. Livezey, and D. C. Stokes, 1996: Reconstruction of historical sea surface temperatures using empirical orthogonal functions. J. Climate, 9, 1403–1420, https://doi.org/10.1175/1520-0442(1996)009<1403:ROHSST>2.0.CO;2.
Stark, J. D., C. J. Donlon, M. J. Martin, and M. E. McCulloch, 2007: OSTIA: An operational, high resolution, real time, global sea surface temperature analysis system. Proc. Oceans ’07, Aberdeen, Scotland, IEEE, 10.1109/OCEANSE.2007.4302251.
Steele, M., W. Ermold, I. Rigor, 2017: UpTempO buoys deployed in the Arctic Ocean in 2017. Arctic Data Center, accessed 1 August 2020, https://doi.org/10.18739/A2GB1XG6P.
Thiébaux, J., E. Rogers, W. Wang, and B. Katz, 2003: A new high-resolution blended real-time global sea surface temperature analysis. Bull. Amer. Meteor. Soc., 84, 645–656, https://doi.org/10.1175/BAMS-84-5-645.
Titchner, H. A., and N. A. Rayner, 2014: The Met Office Hadley Centre sea ice and sea surface temperature data set, version 2: 1. Sea ice concentrations. J. Geophys. Res., 119, 2864–2889, https://doi.org/10.1002/2013JD020316.
von Storch, H., and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.
Walpole, R. E., R. H. Myers, S. L. Myers, and K. Ye, 2012: Probability and Statistics for Engineers and Scientists. 9th ed. Prentice Hall, 791pp.
Woo, H.-J., and K.-A. Park, 2020: Inter-comparisons of daily sea surface temperatures and in-situ temperatures in the coastal regions. Remote Sens., 12, 1592, https://doi.org/10.3390/rs12101592.
Xie, J., J. Zhu, and Y. Lib, 2008: Assessment and inter-comparison of five high-resolution sea surface temperature products in the shelf and coastal seas around China. Cont. Shelf Res., 28, 1286–1293, https://doi.org/10.1016/j.csr.2008.02.020.
Xu, F., and A. Ignatov, 2010: Evaluation of in situ SSTs for use in the calibration and validation of satellite retrievals. J. Geophys. Res. Oceans, 115, C09022, https://doi.org/10.1029/2010JC006129.
Xu, F., and A. Ignatov, 2016: Error characterization in iQuam SSTs using triple collocations with satellite measurements. Geophys. Res. Lett., 43, 10 826–10 834, https://doi.org/10.1002/2016GL070287.
Yang, C., and Coauthors, 2021: Sea surface temperature intercomparison in the framework of the Copernicus Climate Change Service (C3S). J. Climate, 34, 5257–5283, https://doi.org/10.1175/JCLI-D-20-0793.1.
Zhang, H.-M., R. W. Reynolds, and T. M. Smith, 2004: Bias characteristics in the AVHRR sea surface temperature. Geophys. Res. Lett., 31, L01307, https://doi.org/10.1029/2003GL018804.
Zhong, A., and H. Beggs, 2008: Operational Implementation of Global Australian Multi-Sensor Sea Surface Temperature Analysis. Analysis and Prediction Operations Bulletin No. 77. Bureau of Meteorology, Australia, 2 October 2008, http://cawcr.gov.au/projects/SST/GAMSSA_BoM_Operational_Bulletin_77.pdf.