1. Introduction
Sea surface temperature (SST) is an important environmental variable used to study weather, climate, and ocean processes. Reynolds et al. (2007) introduced two types of daily optimum interpolation SST analyses (DOISST), produced at the National Climatic Data Center, that blend in situ and satellite data on a regular grid and use statistical techniques to fill in gaps. One analysis uses in situ data from the International Comprehensive Ocean–Atmosphere dataset (ICOADS) (Woodruff et al. 2011) and infrared satellite data from the Advanced Very High Resolution Radiometer (AVHRR), which is referred to here as “AVHRR only.” The other analysis includes additional microwave satellite data from the Advanced Microwave Scanning Radiometer (AMSR). The short-lived Advanced Earth Observing Satellite-II (ADEOS-II) carried the first AMSR instrument, but in this paper AMSR is used to refer only to its successor, AMSR-E [i.e., AMSR for the Earth Observing System (EOS) on board the Aqua satellite] and the related DOISST product is called AMSR+AVHRR. Both daily analyses are done on a ¼° spatial grid but have different advantages and applications. The AMSR has better all-weather coverage than AVHRR because microwave retrievals are restricted only by precipitation and proximity to land, while AVHRR retrievals can only be made in cloud-free conditions. Given this difference, the AMSR+AVHRR analysis has higher signal variance than the AVHRR-only analysis in regions with clouds. On the other hand, the AVHRR-only time series is longer because AVHRR SST data are available beginning in 1981, while the AMSR dataset begins in 2002. Currently, the AMSR+AVHRR record may be too short to examine long-term trends, but as the time series is extended it will become increasingly relevant at climate time scales. In fact, the follow-up microwave instrument to AMSR, called AMSR2, was launched in May 2012 on the Japanese Global Change Observation Mission-Water 1 (GCOM-W1) satellite. The SST data from AMSR2 are expected to be released after May 2013.
Unfortunately, in October 2011, the AMSR antenna had stopped rotating because of excessive torque probably from aging lubricant (http://www.jaxa.jp/press/2011/10/20111004_amsr-e_e.html), thereby ending the instrument’s capability to collect daily global SST data. The AMSR+AVHRR analysis could no longer be produced. Another satellite-based microwave instrument, the WindSat Polarimetric Radiometer (WSAT), designed primarily to measure surface vector winds and launched in 2003 (Gaiser et al. 2004), could be an alternative source of SSTs in the interim. The purpose here is to evaluate if WSAT can bridge the temporal gap between the two AMSR instruments. This was done by producing four analyses using the same DOISST procedures. All analyses used the same in situ observations from ships and buoys and proxy SSTs derived from sea ice coverage to better define the SSTs in the sea ice margins. The four analyses used either AMSR data with and without in situ bias correction or WSAT data with and without in situ bias correction. In a departure from Reynolds et al. (2007), AVHRR was not included in any of the analyses so that differences in the microwave data were more apparent. Also, Reynolds et al. (2007) used input data only corresponding to the day analyzed. Here, to increase the temporal smoothness and stability of the product, three consecutive days of data were used, but the impact of the off days (the days before and after the day being analyzed) was reduced by doubling their noise-to-signal ratio (standard deviation). This is the standard version-2 methodology used for current DOISST products (Reynolds 2009).
2. Data
The in situ observations, AMSR data, and proxy SSTs generated from sea ice have been described previously (Reynolds et al. 2007), so, for brevity, only updates are mentioned here. Ship and buoy SSTs were extracted from the most recent ICOADS release 2.5, which extends up to 2007 (Woodruff et al. 2011). From 2008 onward, an interim in situ dataset, also available at the ICOADS website, was used.
Daily AMSR and WindSat SST datasets on a global 25-km grid were downloaded online (
The WSAT is the primary payload on the U.S. Department of Defense’s Coriolis mission (http://www.nrl.navy.mil/WindSat/). Because WSAT is not an operational civilian satellite, the data record had some large temporal gaps, extending from weeks to months, during the early years of the mission (Fig. 1). The longest gaps lasted 4 months (14 February–16 June) in 2005 and almost 3 months (9 June–6 August) in 2007. In contrast, until the antenna failure in 2011, AMSR experienced few data dropouts and for only short duration (one to several days). The long periods with missing data indicated that WSAT would be less preferable to AMSR for making an SST analysis before 2008.
Aside from the temporal gaps, WSAT has lower spatial coverage than AMSR on a daily basis because the instrument viewing geometries and bands differ. The SST data swath is narrower for WSAT (~1025 km) than for AMSR (~1445 km). Figure 2 shows the daily percent of ¼° ocean grid points containing either WSAT or AMSR observations during 2010. On average, the night- and daytime coverage for WSAT was ~30% and ~26%, respectively, and correspondingly for AMSR was ~41% and ~36% (Figs. 2a,b). The combined night- and daytime coverage, taking into account overlaps, was ~43% for WSAT and ~54% for AMSR (Fig. 2c). Figure 2 also shows that WSAT coverage is more erratic than AMSR, reflecting the frequency of missing scans. Although slightly inferior to AMSR, the WSAT coverage was much higher than that for Pathfinder AVHRR SSTs (i.e., about 12%, 15%, and 23% for daytime, nighttime, and combined daily coverage, respectively) and thus can be expected to yield lower sampling errors than DOISST based on infrared data only.
3. Results
Four analyses using the DOISST procedure were run daily from 1 January 2004 through 31 September 2011 using in situ and microwave satellite data. Two of the analyses used AMSR data; the other two used WSAT. One of the AMSR and one of WSAT analyses included a preliminary large-scale bias correction of the satellite data using in situ data as described in Reynolds et al. (2007); the other two analyses did not have any bias correction.
The zonal monthly average differences between WSAT and AMSR analysis fields with and without bias correction are shown in Fig. 3. The differences without bias correction (Fig. 3a) tended to be almost all positive (i.e., AMSR warmer) and were strongest in middle latitudes during summer, where maximum differences sometimes exceeded ~0.3 K. A separate examination of day- and nighttime average temperatures (not shown) indicated that the overall differences came primarily from daytime differences. Thus, the differences were consistent with summertime diurnal heating and could be attributed to different equatorial crossing times of the two satellites.
Figure 3b shows that differences between the two analyses are mostly eliminated (<0.1 K) by the bias correction using in situ data. The three largest remaining differences occurred in 2005 (March–May), 2007 (June–early August), and 2008 (June), which corresponded to the WSAT major continuous temporal gaps. For these years, the maximum disagreements were located at different latitudes depending on the season in which the gap occurred. In 2007, the AMSR DOISST reflected the maximum temperatures near the Arctic at the height of the boreal summer, while the WSAT DOISST tended to climatology because of missing observations. The similar situation occurred in spring of 2005, when AMSR DOISST showed warmer SSTs from 30°S to 60°N than WSAT DOISST. In 2008, there were no WSAT data for three weeks in June, so the result was similar. Note that, even though WSAT had more full days of data missing in 2006 than 2008 (Fig. 1), the impact on the 2006 analyses was not as evident because the gaps were mostly of 1–2-day duration and spread out over the year.
Fortunately the differences between the two microwave analyses were small from 2009 onward. However, isolated differences still occurred when there were daily changes in sampling. An example of this, in the eastern tropical Pacific, is shown in Fig. 4. The top panels (Figs. 4a,b) show 3 days of combined day- and nighttime satellite coverage for AMSR and WSAT, respectively. Clearly for AMSR, only the rain-affected areas and coastal regions had no data while WSAT had additional large gaps caused by missing orbits. This coverage discrepancy affected the resulting analyses (Figs. 4c,d) and the SST gradients computed from them (Figs. 4e,f). In particular, along the equator, both showed a strong zonal gradient associated with tropical instability waves but WSAT exhibited some weaker false vertical gradients near 125°W where the data dropped out. Moreover, the cool anomaly pattern near the equator tended to be more continuous and narrower in the AMSR DOISST than in the WSAT analysis (Figs. 4c,d). This may be caused by diurnal warming effects under low wind conditions that were detected by AMSR during its afternoon overpass. This was especially evident around 150°W, where AMSR observations along the instability wave were warmer relative to WSAT. This is not seen in WSAT because the evening overpass missed the hottest time of the day.
Regardless of whether bias correction was performed, there was a residual difference between the two analyses at high latitudes north of 65°N and south of 50°S, with AMSR tending to be cooler than WSAT by about 0.2 K (Fig. 3). This highlights the limitations of bias adjustment approach where in situ data are sparse. In addition, the differences exhibited seasonality because most of the near-polar areas were masked by ice in winter. The negative difference became most pronounced at the end of the AMSR lifetime above 65°N (Fig. 3). From July to September 2011, the proximity flags for AMSR were more severe, tending to exclude more data in this region than WSAT. The reason for the increased proximity flag differences warrants further investigation.
4. Conclusions
A comparison of two microwave satellite-based SST datasets showed that AMSR offers better spatial coverage than WSAT, because the narrower WSAT swaths cover less ocean than AMSR on a daily basis. Moreover, WSAT data gaps tended to occur more frequently, lasting on the order of months in 2005 and 2007. Despite these issues, WSAT was found to be a suitable candidate for the National Oceanic and Atmospheric Administration’s microwave-based daily optimum interpolation sea surface temperature analysis because its spatial coverage was still far superior to that of AVHRR and data dropouts in recent years were less frequent. While unadjusted WSAT and AMSR SST analyses were found to differ, the use of an in situ–based bias correction produced analyses that were very similar to each other. These results are very encouraging and clearly show that WSAT data can be a useful substitute for AMSR until AMSR2 data become available in 2013. An analysis using WSAT will be generated for the period when an AMSR+AVHRR analysis is not possible because of the lack of AMSR data. However, this analysis will not be produced operationally because WSAT data may not always be available in real time.
Acknowledgments
The AMSR-E and WindSAT SSTs are produced by Remote Sensing Systems under the sponsorship of the NASA Earth Science MEaSUREs DISCOVER Project, the AMSR-E Science Team, and the NASA Earth Science Physical Oceanography Program. The WindSat temperature data record is provided to RSS by Peter Gaiser at the Naval Research Laboratory in Washington, D.C. The ICOADS data are from the Research Data Archive (RDA), which is maintained by the Computational and Information Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR). NCAR is sponsored by the National Science Foundation (NSF). The original ICOADS data are available from the research data archive (http://rda.ucar.edu) in dataset ds540.0. The comments of Dr. Deborah Smith at Remote Sensing Systems and two anonymous reviewers helped improve the manuscript with their excellent comments.
REFERENCES
Donlon, C., and Coauthors, 2007: The Global Ocean Data Assimilation Experiment High-Resolution Sea Surface Temperature Pilot Project. Bull. Amer. Meteor. Soc., 88, 1197–1213.
Gaiser, P. W., and Coauthors, 2004: The WindSat spaceborne polarimetric microwave radiometer: Sensor description and early orbit performance. IEEE Trans. Geosci. Remote Sens., 42, 2347–2361.
Gentemann, C. L., cited 2011: Microwave sea surface temperatures for climate. [Available online at http://www.wcrp-climate.org/conference2011/posters/C14/C14_Gentemann_T45B.pdf.]
Gentemann, C. L., C. J. Donlon, A. Stuart-Menteth, and F. J. Wentz, 2003: Diurnal signals in satellite sea surface temperature measurements. Geophys. Res. Lett., 30, 1140, doi:10.1029/2002GL016291.
Reynolds, R. W., 2009: What's new in version 2. NOAA/NCDC Rep., 10 pp. [Available online at http://www.ncdc.noaa.gov/oa/climate/research/sst/papers/oisst_daily_v02r00_version2-features.pdf.]
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.
Woodruff, S. D., and Coauthors, 2011: ICOADS release 2.5: Extensions and enhancements to the surface marine meteorological archive. Int. J. Climatol., 31, 951–967.