Why Did Large Differences Arise in the Sea Surface Temperature Datasets across the Tropical Pacific during 2012?

Boyin Huang * NOAA/National Climatic Data Center, Asheville, North Carolina

Search for other papers by Boyin Huang in
Current site
Google Scholar
PubMed
Close
,
Michelle L’Heureux +NOAA/Climate Prediction Center, College Park, Maryland

Search for other papers by Michelle L’Heureux in
Current site
Google Scholar
PubMed
Close
,
Jay Lawrimore * NOAA/National Climatic Data Center, Asheville, North Carolina

Search for other papers by Jay Lawrimore in
Current site
Google Scholar
PubMed
Close
,
Chunying Liu #ERT, Inc., Laurel, Maryland

Search for other papers by Chunying Liu in
Current site
Google Scholar
PubMed
Close
,
Huai-Min Zhang * NOAA/National Climatic Data Center, Asheville, North Carolina

Search for other papers by Huai-Min Zhang in
Current site
Google Scholar
PubMed
Close
,
Viva Banzon * NOAA/National Climatic Data Center, Asheville, North Carolina

Search for other papers by Viva Banzon in
Current site
Google Scholar
PubMed
Close
,
Zeng-Zhen Hu +NOAA/Climate Prediction Center, College Park, Maryland

Search for other papers by Zeng-Zhen Hu in
Current site
Google Scholar
PubMed
Close
, and
Arun Kumar +NOAA/Climate Prediction Center, College Park, Maryland

Search for other papers by Arun Kumar in
Current site
Google Scholar
PubMed
Close
Full access

We are aware of a technical issue preventing figures and tables from showing in some newly published articles in the full-text HTML view.
While we are resolving the problem, please use the online PDF version of these articles to view figures and tables.

Abstract

During June–November 2012, pronounced differences in tropical Pacific sea surface temperature (SST) anomalies were observed between three widely used SST products: the extended reconstructed SST version 3b (ERSSTv3b), and the optimum interpolation SST version 2 analyses (OISST), produced weekly (OISSTwk) and daily (OISSTdy). During June–August 2012, the Niño-3.4 SST anomaly (SSTA) index was 0.2°–0.3°C lower in ERSSTv3b than in OISSTwk and OISSTdy, while it was 0.3°–0.4°C higher from September to November 2012. Such differences in the Niño-3.4 SSTA index can impact the assessment of the status of the El Niño–Southern Oscillation, which is determined using a threshold of ±0.5°C in the Niño-3.4 SSTA index.

To investigate the reasons for the differences between ERSSTv3b and OISSTdy/OISSTwk, an experimental analysis (called ERSSTsat) is created that is similar to ERSSTv3b but includes satellite-derived SSTs. However, significant differences in the Niño-3.4 SSTA index remained between ERSSTsat and OISSTdy/OISSTwk. Comparisons of ERSSTsat and OISSTdy indicate that their differences are mostly associated with the different schemes for bias adjustment applied to the satellite-based SSTs. It is therefore suggested that the differences in the Niño-3.4 SSTA index between ERSSTv3b and OISSTdy cannot be solely due to the inclusion of but by the bias adjustment methodology of satellite data in OISSTdy.

Finally, the SST products are compared with observations from ships, buoys, and satellites. On the monthly time scale, the area-averaged Niño-3.4 SSTA index in the tropical Pacific is more consistent with in situ observations in ERSSTv3b than in OISSTdy. In contrast, pointwise observations across the tropical Pacific are more consistent with OISSTdy than ERSSTv3b. It is therefore suggested that the differences among SST products are partially due to a structural uncertainty of various SST estimates.

Corresponding author address: Boyin Huang, NOAA/National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801. E-mail: boyin.huang@noaa.gov

Abstract

During June–November 2012, pronounced differences in tropical Pacific sea surface temperature (SST) anomalies were observed between three widely used SST products: the extended reconstructed SST version 3b (ERSSTv3b), and the optimum interpolation SST version 2 analyses (OISST), produced weekly (OISSTwk) and daily (OISSTdy). During June–August 2012, the Niño-3.4 SST anomaly (SSTA) index was 0.2°–0.3°C lower in ERSSTv3b than in OISSTwk and OISSTdy, while it was 0.3°–0.4°C higher from September to November 2012. Such differences in the Niño-3.4 SSTA index can impact the assessment of the status of the El Niño–Southern Oscillation, which is determined using a threshold of ±0.5°C in the Niño-3.4 SSTA index.

To investigate the reasons for the differences between ERSSTv3b and OISSTdy/OISSTwk, an experimental analysis (called ERSSTsat) is created that is similar to ERSSTv3b but includes satellite-derived SSTs. However, significant differences in the Niño-3.4 SSTA index remained between ERSSTsat and OISSTdy/OISSTwk. Comparisons of ERSSTsat and OISSTdy indicate that their differences are mostly associated with the different schemes for bias adjustment applied to the satellite-based SSTs. It is therefore suggested that the differences in the Niño-3.4 SSTA index between ERSSTv3b and OISSTdy cannot be solely due to the inclusion of but by the bias adjustment methodology of satellite data in OISSTdy.

Finally, the SST products are compared with observations from ships, buoys, and satellites. On the monthly time scale, the area-averaged Niño-3.4 SSTA index in the tropical Pacific is more consistent with in situ observations in ERSSTv3b than in OISSTdy. In contrast, pointwise observations across the tropical Pacific are more consistent with OISSTdy than ERSSTv3b. It is therefore suggested that the differences among SST products are partially due to a structural uncertainty of various SST estimates.

Corresponding author address: Boyin Huang, NOAA/National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801. E-mail: boyin.huang@noaa.gov

1. Introduction

The sea surface temperature (SST) datasets from the extended reconstructed SST, version 3b (ERSSTv3b; Smith et al. 2008; Banzon et al. 2010), and optimum interpolation (OI) SST, version 2, produced daily (OISSTdy; Reynolds et al. 2007) and weekly (OISSTwk; Reynolds et al. 2002), are widely used for climate monitoring, prediction, and verification. During boreal winter 2011–12, the SSTs were colder than normal in the equatorial Pacific due to a weak La Niña. The cold SST anomalies (SSTA) began to weaken after January 2012 and became above average in June 2012. Dynamical models initialized with SST analyses in July 2012 predicted development of El Niño during the boreal winter of 2012–13, and statistical model predictions, initialized in August 2012, also suggested a weak-to-moderate El Niño event.

Monitoring and prediction of the El Niño–Southern Oscillation (ENSO) are influenced by dataset differences in tropical Pacific SST and in the Niño-3.4 region in particular (averaged SSTA in 5°S–5°N and 120°–170°W; Barnston et al. 1997). For example, the Niño-3.4 SSTA index was approximately 0.3°C lower in ERSSTv3b than in OISSTdy and OISSTwk from June to August (JJA) 2012 (Fig. 1a). The SSTA difference between ERSSTv3b and OISSTdy/OISSTwk changed sign from September to November (SON) 2012 with the Niño-3.4 SSTA index being approximately 0.4°C higher in ERSSTv3b than in OISSTdy and OISSTwk (Fig. 1a). Such SSTA differences in SST datasets can make the assessment of the ENSO status uncertain, given that part of the National Oceanic and Atmospheric Administration (NOAA)’s operational ENSO definition depends on whether the Niño-3.4 SSTA index anomaly is greater than +0.5°C (El Niño) or less than −0.5°C (La Niña).

Fig. 1.
Fig. 1.

(a) Niño-3.4 SSTA index from January to November of 2012 in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and ERSSTsatR15. (b) Difference of Niño-3.4 SSTA index in OISSTdy, ERSSTv3b, and ERSSTsatR15 with respect to OISSTwk. (c) Numbers of total (left-hand axis in a log scale) and super- observations (right-hand axis). A 5-month running-mean filter is applied in (b).

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

There are several potential factors that may cause these SST differences. These factors include differences in analysis methods, ingested datasets, and/or analysis resolution in space and time. Because ingested in situ data (ships, buoys, and other maritime platforms) and satellite data do not provide complete, gridded coverage across the tropical Pacific, various strategies are employed to interpolate these input data. The monthly 2° × 2° grid ERSSTv3b uses a low-frequency filter and a high-frequency decomposition of empirical orthogonal teleconnections (EOTs; van den Dool et al. 2000) trained with monthly observed SST from 1982 to 2005 (Smith et al. 2008). The daily 0.25° × 0.25° grid OISSTdy and the weekly 1° × 1° grid OISSTwk use an optimum interpolation procedure (Reynolds et al. 2002).

Another important distinguishing feature among these data products is the use of satellite data. OISSTdy and OISSTwk include satellite SSTs, while ERSSTv3b does not. To understand the source of the differences, we conducted two additional experiments (ERSSTsat and ERSSTsatR15) that are similar to ERSSTv3b except that satellite observations are included. These experiments were compared with ERSSTv3b to assess the role of satellite observations and were also compared with OISSTdy to assess the impact of different analysis methods. These datasets are described in section 2. The comparisons of ERSSTv3b/ERSSTsat/ERSSTsatR15 with OISSTdy/OISSTwk are given in section 3, and comparisons of ERSSTv3b/ERSSTsat/ERSSTsatR15 and OISSTdy/OISSTwk with in situ and satellite-based observations are shown in section 4. A summary is provided in section 5.

2. Datasets

There are two operational SST products based on OI: the daily 0.25° grid OISSTdy (Reynolds et al. 2007) and weekly 1° grid OISSTwk (Reynolds et al. 2002). The OISSTdy and OISSTwk use in situ ship and buoy observations from the International Comprehensive Ocean–Atmosphere Dataset (ICOADS R2.1; Worley et al. 2005) from September 1981 to December 1998 and from the Global Teleconnection System (GTS) archives after January 1999 (Table 1).

Table 1.

Datasets and their RMSDs relative to observations from in situ ships and buoys, TAO moorings, ATSR, and AMSR-E. The RMSDs are calculated in the Niño-3.4 region except for those relative to TAO at 140°W. AVHRR adjustment window is defined as the length of AVHRR and in situ SST data used to adjust AVHRR SSTs.

Table 1.

The OISSTdy and OISSTwk also ingest the Advanced Very High Resolution Radiometer (AVHRR) satellite SSTs. However, AVHRR SSTs exhibit potential biases, which could result from aerosols due to volcanic eruption, windblown dust from continents, and cloud cover (Zhang et al. 2004). Therefore, the SSTs from AVHRR are adjusted to reduce the effects of biases by using in situ SSTs. The bias adjustment procedures in OISSTdy are as follows: (i) original in situ and AVHRR SSTs are first averaged into daily 2° × 2° fields, (ii) a daily bias adjustment field of 2° × 2° is calculated from EOTs that are well sampled by both in situ and AVHRR SSTs within a symmetric 15-day running window, (iii) the bias adjustment field of 2° × 2° is interpolated to individual track locations of AVHRR SSTs, and (iv) individual AVHRR SSTs are adjusted by subtracting the bias adjustment. The bias adjustment in OISSTwk is calculated by solving Poisson equations using weekly in situ SSTs as a boundary condition that enables the system to maintain reasonable large-scale spatial gradients of SST. Reynolds et al. (2002, 2007) provides more details on the bias adjustment procedures.

The spatial (temporal) resolutions are 0.25° (1 day) in OISSTdy and 1° (7 days) in OISSTwk. For monthly comparisons, daily OISSTdy is averaged to monthly, while weekly OISSTwk is first interpolated into daily and then averaged to monthly. Box averaging is used to regrid the finer resolution of OISSTdy and OISSTwk to 2° × 2° that is used for ERSSTv3b. Hereafter, OISSTdy and OISSTwk refer to the monthly, 2° × 2° averages, which are compared against the monthly 2° × 2° ERSSTv3b. All comparisons are made based on the monthly-mean data, and all SSTAs are relative to the same 1971–2000 climatology (Xue et al. 2003).

The ERSSTv3b is based on Smith et al. (2008) but uses in situ data only (no satellite data is used; Table 1) in resolutions of monthly and 2° × 2°. The in situ data are from ICOADS (Woodruff et al. 2011) from 1854 to 2007 and from GTS after 2007. It is important to note that for OISSTdy and OISSTwk, SSTs from ship measurements are adjusted toward those from buoy observations by subtracting 0.15°C from ship observations, while no such adjustment is made in ERSSTv3b. Regardless, the difference of in situ–based SSTs used in ERSSTv3b and OISSTdy/OISSTwk is very small across the tropical Pacific between 1982 and 2012 (not shown), because the in situ SST is calculated based on a weighted mean of ship and buoy observations. The weights of buoy (ship) SSTs are 7 times (1 time) the numbers of buoy (ship) observations based on the data noise levels (Reynolds and Smith 1994). The weights for buoy observations exceed those of ship observations after 1986 due to the increased number of buoy observations.

For this study, and for diagnostic purposes, we also generated an experimental ERSSTsat (Table 1) that is the same as ERSSTv3b except that it incorporates the same AVHRR-based SSTs used in OISSTdy and OISSTwk. However, in ERSSTsat the biases of AVHRR SSTs are adjusted by using EOTs of monthly in situ and AVHRR SSTs. The main difference between bias adjustments of ERSSTsat and OISSTdy is that monthly (a ~30-day window) data are used in ERSSTsat, while a 15-day window is adopted in OISSTdy. The experimental ERSSTsatR15 (Table 1) is the same as ERSSTsat except that the biases of daily AVHRR SSTs are adjusted using a 15-day running window to match OISSTdy, and their monthly averaged AVHRR SSTs are taken as inputs to ERSSTsatR15. The spatial and temporal resolutions of ERSSTsat and ERSSTsatR15 are the same as in ERSSTv3b.

To assess the quality of ERSSTv3b, OISSTdy, and OISSTwk, the following SST data are also used in our assessment: (i) monthly merged in situ–only (ship and buoy) observations in 2° × 2° boxes, and (ii) monthly median Hadley Centre Sea Surface Temperature dataset, version 3 (HadSST3) (Table 1), of 100 realizations in 5° × 5° boxes between 1850 and 2012 from the Met Office (Kennedy et al. 2011). The HadSST3 uses in situ data from ICOADS release 2.5 (Woodruff et al. 2011), and the 5° × 5° analysis was regridded to 2° × 2° for comparison. The regridding is done by first filling a 5° × 5° box datum to 25 1° × 1° boxes and averaging 1° × 1° data to 2° × 2° grid. Also used are the (iii) daily Tropical Atmosphere and Ocean (TAO; McPhaden et al. 1998) SST data at longitudes of 110°W, 140°W, 170°W, 165°E, and 156°E on the equator from 1982 to 2012, which are converted to monthly SSTs for comparisons; (iv) monthly Along-Track Scanning Radiometers (ATSR) SSTs from 1997 to 2011 with a spatial resolution of 0.1° × 0.1° (Merchant et al. 2012); and (v) monthly Advanced Microwave Scanning Radiometer for Earth Observation System (AMSR-E) SSTs from June 2002 to September 2011 with a spatial resolution of 0.25° × 0.25°. The ATSR and AMSR-E data are box averaged to 2° × 2° spatial resolution for comparison.

3. Comparisons of extended reconstructed and optimum interpolation SSTs

Figure 2 shows the SSTA of the tropical Pacific in July 2012 from different SST products, which is an El Niño–like SSTA pattern in OISSTdy (Fig. 2a), OISSTwk (Fig. 2b), and ERSSTv3b (Fig. 2d). It is clear that the amplitude of positive SSTA in the equatorial Pacific east of ~150°W is smaller in ERSSTv3b than in OISSTdy and OISSTwk. However, it is difficult to verify which one is closer to the truth since there are virtually no in situ observations between 140° and 110°W (Fig. 2c). For example, the equatorial TAO buoys at 110°W (155°W) were inoperable starting between May 2012 and March 2013 (after July 2012). The equatorial SSTA near 145°W is about −1°C in the in situ observations (Fig. 2c), while it is about 0.2°, 0.5°, and 0°C in OISSTdy, OISSTwk, and ERSSTv3b (Figs. 2a,b,d), respectively. Also, the equatorial SSTA between 180° and 150°W is 0.5° to 1°C in the in situ observations (Fig. 2c), while it is near 0°C, 0° to 0.5°C, and near 0°C in OISSTdy, OISSTwk, and ERSSTv3b (Figs. 2a,b,d), respectively. In November 2012, the in situ observations indicate equatorial SSTA of 0.5° to 1°C between 180° and 135°W (Fig. 3c), which is well captured by ERSSTv3b (Fig. 3d), but it is weaker in OISSTdy (Fig. 3a) and OISSTwk (Fig. 3b). While between 150° and 170°E, the in situ observations also showed an equatorial SSTA of 1°C (Fig. 3c), which is better captured in OISSTwk (Fig. 3b) than in OISSTdy (Fig. 3a) and ERSSTv3b (Fig. 3d).

Fig. 2.
Fig. 2.

SSTA of July 2012 in (a) OISSTdy, (b) OISSTwk, (c) in situ SSTA, and (d) ERSSTv3b. Shadings and contours are in unit of °C. The green rectangle box indicates the Niño-3.4 region.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Fig. 3.
Fig. 3.

As in Fig. 2, but for November 2012.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Overall, the SSTAs in the central equatorial Pacific in July (November) 2012 in ERSSTv3b are weaker (stronger) than in OISSTdy and OISSTwk in Fig. 2 (Fig. 3). This is consistent with a lower (higher) Niño-3.4 SSTA index value during July (November) in ERSSTv3b than in OISSTdy and OISSTwk, as shown in Fig. 1a. It appears that these SST differences are not clearly associated with the coverage or total number of in situ observations, although the spatial distribution of the observations may matter. The data coverage is quantified using the so-called super-observations, which are defined as the number of 2° grid boxes containing in situ observations (refer to Figs. 2c and 3c). In 2012, the data coverage dropped sharply in the Niño-3.4 region to levels not previously seen (Fig. 1c), which is consistent with the large SSTA differences (Fig. 1a). However, over the whole data period, there is no statistical relationship between low coverage and large SST differences. For example, data coverage is low in 2006, while the difference in the Niño-3.4 SSTA index between ERSSTv3b and OISSTdy/OISSTwk is small (Fig. 1b). In contrast, the data coverage is 20%–30% higher in 1982–84 and 1988 (Fig. 1c), while the differences in the Niño-3.4 SSTA index are large (Fig. 1b). In contrast to the decrease in data coverage, the total number of observations in the Niño-3.4 region increased from 1982 to 1994 and has been nearly constant from 1994 to 2012 (Fig. 1c). The increasing total number of observations is primarily due to the increasing number of buoy observations, while the lower coverage is due to fewer ship observations (not shown). The SSTA difference between OISSTdy and ERSSTv3b reduced from the 1980s to 2010s, which may partly be associated with the overall increased number of buoy observations so that the effect of ship adjustment in OISSTdy becomes small.

One may speculate that the large SSTA differences are caused by the fact that ERSSTv3b does not ingest AVHRR SSTs, while OISSTdy and OISSTwk do. To assess the role of AVHRR satellite data, ERSSTsat is compared with ERSSTv3b. We found that the average differences in the Niño-3.4 SSTA index between ERSSTsat and ERSSTv3b is generally small during 2012 (Fig. 1a) and also during 1982 to 2012 (Fig. 1b). The root-mean-square difference (RMSD) is 0.08°C between ERSSTv3b and ERSSTsat from 1982 to 2012, while the RMSD is 0.22°C between ERSSTv3b and OISSTdy. This suggests that the role of AVHRR SST in ERSSTsat is small (0.08°C), compared to the total difference between ERSSTv3b and OISSTdy (0.22°C). The reason for the small difference between ERSSTv3b and ERSSTsat is that the spatial pattern and magnitude of AVHRR SSTs become similar to those in situ SSTs after the biases of AVHRR SSTs are adjusted. The small difference between ERSSTv3b and ERSSTsat also means that the differences in the Niño-3.4 SSTA index between ERSSTsat and OISSTdy/OISSTwk remain large (near 0.2° to 0.4°C) in 2012.

Looking more closely at why ERSSTsat and OISSTdy/OISSTwk differ even though they use similar in situ SSTs and the same AVHRR SSTs, we found that the differences are due to the distinct bias adjustment schemes used in ERSSTsat, OISSTdy, and OISSTwk. For example, monthly in situ and AVHRR SSTs are used to generate monthly bias adjustments for monthly products of ERSSTsat. In contrast, in situ and AVHRR SSTs within an equal-weight 15-day running window are used to generate daily bias adjustments for OISSTdy. In July 2012, the monthly averaged bias adjustment in the central equatorial Pacific is approximately 0.4°C in OISSTdy (Fig. 4a), while it is approximately 0.8°C in ERSSTsat (Fig. 4b). In contrast, the bias adjustment in the central and eastern equatorial Pacific in November 2012 is −0.4° to −0.6°C in OISSTdy (Fig. 4c), while it is −0.6° to −0.8°C in ERSSTsat (Fig. 4d). Consequently, the Niño-3.4 SSTA index in July (November) 2012 is lower (higher) in ERSSTsat than in OISSTdy. Further investigation is needed to understand why the adjustment to the satellite observations is so large.

Fig. 4.
Fig. 4.

Monthly averaged biases of AVHRR SST in (a) OISSTdy in July 2012, (b) ERSSTsat in July 2012, (c) OISSTdy in November 2012, and (d) ERSSTsat in November of 2012. Contour intervals are 0.2°C. The green rectangle box indicates the Niño-3.4 region.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Reduction of data in OISSTdy may mean the bias adjustment will not capture the EOTs well, which could result in a less reliable bias adjustment. In contrast, when there are more data included in the bias adjustment procedure, such as in ERSSTsat, systematic and random errors from in situ data may contaminate the bias adjustment when these errors are large. Therefore, a bias adjustment with low uncertainty is essential in effectively including satellite observations in SST data products. The varying bias adjustments applied in OISSTdy and OISSTwk can also explain large SST departures (Fig. 1b) between OISSTdy and OISSTwk with an RMSD of 0.18°C.

An additional experiment, ERSSTsatR15, is designed to verify whether the differences in bias adjustments of AVHRR SST indeed result in the differences between ERSSTsat and OISSTdy. Therefore, bias-adjusted daily AVHRR SSTs of OISSTdy are used in ERSSTsatR15. The analysis shows that the Niño-3.4 SSTA index in ERSSTsatR15 is almost identical to that in OISSTdy in 2012 (Fig. 1a) and is also very close to OISSTdy from 1982 to 2012 (Fig. 1b). The RMSD is 0.07°C between ERSSTsatR15 and OISSTdy from 1982 to 2012, which is much smaller than those between ERSSTsat and OISSTdy (0.19°C) and between ERSSTv3b and OISSTdy (0.22°C). The small RMSD between ERSSTsatR15 and OISSTdy implies that the data reconstruction methods do not play a key role in contributing to the SSTA differences between ERSSTv3b and OISSTdy. Therefore, we conclude that the SSTA differences in the tropical Pacific between ERSSTsat/ERSSTv3b and OISSTdy are primarily due to the satellite bias adjustments.

4. Comparisons with in situ, ATSR, and AMSR-E SSTs

To further assess the SST quality in the central tropical Pacific, OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 are compared with observations from in situ ships and buoys (which include the TAO mooring), the TAO mooring only, and ATSR and AMSR-E satellites over the Niño-3.4 region. Figure 5a shows comparisons of SSTs relative to in situ observations in 2012. Here, the in situ SSTs are calculated by including ship SST adjustment when OISSTdy/OISSTwk is compared, since the ship SSTs are adjusted in producing OISSTdy/OISSTwk. In contrast, the in situ SSTs are built without ship SST adjustment when ERSSTv3b/ERSSTsat/ERSSTsatR15/HadSST3 is compared, since the ship SSTs are not adjusted in producing ERSSTv3b/ERSSTsat/ERSSTsatR15/HadSST3.

Fig. 5.
Fig. 5.

Niño-3.4 region (5°S–5°N, 120°–170°W) averaged SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to in situ SSTA. (a) January–November 2012 and (b) 1982–2012. A 5-month running-mean filter is applied in (b).

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Figure 5a indicates the deviations of ERSSTv3b and ERSSTsat from the in situ observations are generally smaller than those of OISSTdy and OISSTwk from January to November 2012 (except for June 2012). Interestingly, the deviations for the OISSTdy/OISSTwk became larger starting around June 2012; this coincides with the missing equatorial TAO buoys at 155° and 110°W, as previously mentioned. One can speculate that the coarser resolution ERSSTv3b may still capture the area-averaged signals. The finer-resolution OISSTdy/OISSTwk may be more sensitive to the lack of in situ observations and is therefore unable to accurately correct the satellite biases. Furthermore, the deviations of ERSSTv3b and ERSSTsat are mostly smaller than those in OISSTdy from 1982 to 2012 (Fig. 5b). The RMSD in 1982–2012 is 0.16°, 0.15°, 0.09°, 0.09°, and 0.07°C in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3, respectively (Table 1). Therefore, it appears that the Niño-3.4 SSTA index from ERSSTv3b and ERSSTsat more faithfully represents the box-averaged in situ SSTA than OISSTdy and OISSTwk.

Comparisons with the pointwise TAO-only mooring observations at 110°W, 140°W, 170°W, 165°E, and 156°E at the equator, however, indicate that overall deviations of ERSSTv3b and ERSSTsat are larger than those of OISSTdy and OISSTwk. Figure 6 shows an example of the deviations at 140°W within the Niño-3.4 region. The RMSD is 0.27°, 0.25°, 0.64°, 0.58°, and 0.73°C in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3, respectively (Table 1). The coarser (finer) spatial and temporal resolutions of SSTs originally reconstructed in ERSSTv3b/ERSSTsat/HadSST3 (OISSTdy/OISSTwk) likely lead to larger (smaller) SST deviations from the pointwise TAO buoy observations.

Fig. 6.
Fig. 6.

SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to TAO SSTA at 140°W. A 5-month running-mean filter is applied.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

ATSR and AMSR-E satellite data, which are not included in ERSSTv3b/ERSSTsat or OISSTdy/OISSTwk, provide another way for intercomparisons with ERSSTv3b, ERSSTsat, OISSTdy, and OISSTwk. Figure 7 displays averaged deviations from ATSR observations over the Niño-3.4 region. The deviation is largest in OISSTdy, ERSSTv3b, ERSSTsat, and HadSST3, but smaller in OISSTwk. The RMSD is 0.18°, 0.07°, 0.19°, 0.16°, and 0.22°C in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3, respectively (Table 1).

Fig. 7.
Fig. 7.

SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to ATSR SSTA in Niño-3.4 region (5°S–5°N, 120°–170°W). A 5-month running-mean filter is applied.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Relative to the AMSR-E satellite observations (Fig. 8), the deviations of OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 generally become larger than those relative to ATSR in Fig. 7. The RMSD is 0.18°, 0.13°, 0.27°, 0.24°, and 0.30°C in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3, respectively (Table 1). The larger deviations result from colder (approximately 0.09°C) SST observations in AMSR-E than in ATSR (Fig. 8). Just as in the in situ data, errors and biases may exist in satellite observations, although the study by Merchant et al. (2012) suggests that ATSR provides the most accurate satellite SST dataset.

Fig. 8.
Fig. 8.

SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to AMSR-E SSTA in Niño-3.4 region (5°S–5°N, 120°–170°W). A 5-month running-mean filter is applied.

Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-13-00034.1

Overall, the results shown in this section indicate that structural uncertainty in the various SST estimates may arise from a range of factors, such as the input datasets and analysis methods. Therefore, there is no simple answer to which SST dataset provides the best representation of SSTs in the tropical Pacific, and it depends on the specific applications (e.g., for regional averages, such as the Niño-3.4 region, or for pointwise comparisons). For regional averages, coarse-resolution reconstructions, such as the monthly 2° × 2° ERSSTv3b, provide more consistent results. For pointwise or smaller-scale comparisons, higher-resolution reconstructions, such as the daily 0.25° × 0.25° OISSTdy, provide more consistent results, as long as there are enough observations to support the high-resolution reconstructions (Reynolds et al. 2013). In general, a reconstruction has to be supported by sufficient observational data, which remains an active research topic (e.g., Zhang et al. 2006; Vidard et al. 2007; Errico et al. 2013).

5. Summary

In JJA 2012, the Niño-3.4 SSTA index was 0.2° to 0.3°C higher in OISSTdy and OISSTwk than in ERSSTv3b and was 0.3° to 0.4°C lower in SON 2012, which led to an uncertainty in the assessment of ENSO status. Given that these datasets are ingested as initial conditions in various model-based data assimilation systems, it is possible that the prediction of ENSO was also influenced and should be studied further to better understand the consequences of the differences. Our analysis indicates that these SSTA differences were largely due to bias adjustments of AVHRR satellite SSTs in the production of OISSTdy and OISSTwk. This was demonstrated by conducting an experimental ERSSTsat that also ingests AVHRR SSTs. Results showed that the Niño-3.4 SSTA index of ERSSTsat is close to that of ERSSTv3b (using in situ data only), but still differs largely from that of OISSTdy (using both in situ and AVHRR data) by 0.2° to 0.4°C. The Niño-3.4 SSTA index differences between ERSSTsat and OISSTdy also match with the difference in bias adjustments of AVHRR SSTs between ERSSTsat and OISSTdy.

Our analysis indicates that the large differences in AVHRR SST adjustment between ERSSTsat and OISSTdy are critically dependent on how much AVHRR and in situ data are used in the adjustment procedure. AVHRR SSTs were adjusted using the data within approximately 30- and 15-day windows in ERSSTsat and OISSTdy, respectively. It is likely that the bias adjustment is more reliable when more data, particularly more in situ data, are used. These results suggest the importance of maintaining and increasing in situ observations like TAO moorings that can directly impact the bias adjustment of satellite SST observations.

One cannot expect that a single dataset, analysis, or reconstruction is the best for all purposes. In the equatorial Pacific, the Niño-3.4 SSTA based on in situ observations (ships and buoys) is closer to ERSSTv3b, which is constructed with coarse resolution (monthly and 2° × 2°) to capture regional, larger-scale features, than to OISSTdy and OISSTwk, which are constructed with higher resolution (daily 0.25° × 0.25° and weekly 1° × 1°) to resolve smaller-scale features. Thus, OISSTdy and OISSTwk are closer to the TAO moored buoys. ATSR data match well with OISSTwk, but not as well with OISSTdy and ERSSTv3b. Therefore, users of ERSSTv3b, OISSTdy, and OISSTwk should be aware of these structural differences between SST products, particularly the large differences that arise from bias adjustments when the satellite-based SST observations are included. It is important to keep in mind that these SST products are designed for a wide variety of unique applications. For example, OISSTdy was designed to address the need for near-real-time high spatial resolution SST to ingest numerical models, while ERSSTv3b aims to provide a long-term historical perspective.

Future work will focus on the significance of the SST difference between ERSSTv3b and OISSTdy in the tropical or global oceans. An important question to answer is whether the SST difference is within the range of the uncertainty of observed SSTs. Another important goal is to improve the design and application of the satellite adjustments, so in situ and satellite data may be merged more effectively.

Acknowledgments

The authors thank four anonymous reviewers, as well as Richard Reynolds, John Bates, and Thomas Smith for their thoughtful comments and suggestions, which significantly improved the manuscript. Christopher Merchant was helpful in providing us the monthly ATSR dataset. AMSR-E data are produced by Remote Sensing Systems and sponsored by the NASA Earth Science MEaSUREs DISCOVER Project and the AMSR-E Science Team. The AMSR-E data are available online (www.remss.com).

REFERENCES

  • Banzon, V. F., Reynolds R.W. , and Smith T. M. , 2010: The role of satellite data in extended reconstruction of sea surface temperatures. Extended Abstracts, Oceans from Space Symp., Venice, Italy, European Commission, 27–28, doi:10.2788/8394.

  • Barnston, A. G., Chelliah M. , and Goldenberg S. B. , 1997: Documentation of a highly ENSO-related SST region in the equatorial Pacific. Atmos.–Ocean, 35, 367383.

    • Search Google Scholar
    • Export Citation
  • Errico, R. M., Yang R. , Privé N. C. , Tai K.-S. , Todling R. , Sienkiewicz M. E. , and Guo J. , 2013: Development and validation of observing-system simulation experiments at NASA’s Global Modeling and Assimilation Office. Quart. J. Roy. Meteor. Soc., 139, 1162–1178, doi:10.1002/qj.2027.

    • Search Google Scholar
    • Export Citation
  • Kennedy, J. J., Rayner N. A. , Smith R. O. , Parker D. E. , and Saunby M. , 2011: Reassessing biases and other uncertainties in sea surface temperature observations measured in situ since 1850: 2. Biases and homogenization. J. Geophys. Res., 116, D14104, doi:10.1029/2010JD015220.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The tropical ocean-global atmosphere observing system: A decade of progress. J. Geophys. Res., 103 (C7), 14 16914 240.

    • Search Google Scholar
    • Export Citation
  • Merchant, C. J., and Coauthors, 2012: A 20 year independent record of sea surface temperature for climate from Along-Track Scanning Radiometers. J. Geophys. Res., 117, C12013, doi:10.1029/2012JC008400.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and Smith T. M. , 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, 929948.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Rayner N. A. , Smith T. M. , Stokes D. C. , and Wang W. , 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Smith T. M. , Liu C. , Chelton D. B. , Casey K. S. , and Schlax M. G. , 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 54735496.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Chelton D. B. , Roberts-Jones J. , Martin M. J. , Menemenlis D. , and Merchant C. J. , 2013: Objective determination of feature resolution in two sea surface temperature analyses. J. Climate,26, 2514–2533.

  • Smith, T. M., Reynolds R. W. , Peterson T. C. , and Lawrimore J. , 2008: Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21, 22832296.

    • Search Google Scholar
    • Export Citation
  • van den Dool, H. M., Saha S. , and Johansson A. , 2000: Empirical orthogonal teleconnections. J. Climate, 13, 14211435.

  • Vidard, A., Anderson D. L. T. , and Balmaseda M. , 2007: Impact of ocean observation systems on ocean analysis and seasonal forecasts. Mon. Wea. Rev., 135, 409429.

    • Search Google Scholar
    • Export Citation
  • Woodruff, S. D., and Coauthors, 2011: ICOADS release 2.5: Extensions and enhancements to the surface marine meteorological archive. Int. J. Climatol., 31, 951967, doi:10.1002/joc.2103.

    • Search Google Scholar
    • Export Citation
  • Worley, S. J., Woodruff S. D. , Reynolds R. W. , Lubker S. J. , and Lott N. , 2005: ICOADS release 2.1 data and products. Int. J. Climatol., 25, 823842.

    • Search Google Scholar
    • Export Citation
  • Xue, Y., Smith T. M. , and Reynolds R. W. , 2003: Interdecadal changes of 30-yr SST normals during 1871–2000. J. Climate, 16, 16011612.

    • Search Google Scholar
    • Export Citation
  • Zhang, H.-M., Reynolds R. W. , and Smith T. M. , 2004: Bias characteristics in the AVHRR sea surface temperature. Geophys. Res. Lett., 31, L01307, doi:10.1029/2003GL018804.

    • Search Google Scholar
    • Export Citation
  • Zhang, H.-M., Reynolds R. W. , and Smith T. M. , 2006: Adequacy of the in situ observing system in the satellite era for climate SST. J. Atmos. Oceanic Technol., 23, 107120.

    • Search Google Scholar
    • Export Citation
Save
  • Banzon, V. F., Reynolds R.W. , and Smith T. M. , 2010: The role of satellite data in extended reconstruction of sea surface temperatures. Extended Abstracts, Oceans from Space Symp., Venice, Italy, European Commission, 27–28, doi:10.2788/8394.

  • Barnston, A. G., Chelliah M. , and Goldenberg S. B. , 1997: Documentation of a highly ENSO-related SST region in the equatorial Pacific. Atmos.–Ocean, 35, 367383.

    • Search Google Scholar
    • Export Citation
  • Errico, R. M., Yang R. , Privé N. C. , Tai K.-S. , Todling R. , Sienkiewicz M. E. , and Guo J. , 2013: Development and validation of observing-system simulation experiments at NASA’s Global Modeling and Assimilation Office. Quart. J. Roy. Meteor. Soc., 139, 1162–1178, doi:10.1002/qj.2027.

    • Search Google Scholar
    • Export Citation
  • Kennedy, J. J., Rayner N. A. , Smith R. O. , Parker D. E. , and Saunby M. , 2011: Reassessing biases and other uncertainties in sea surface temperature observations measured in situ since 1850: 2. Biases and homogenization. J. Geophys. Res., 116, D14104, doi:10.1029/2010JD015220.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The tropical ocean-global atmosphere observing system: A decade of progress. J. Geophys. Res., 103 (C7), 14 16914 240.

    • Search Google Scholar
    • Export Citation
  • Merchant, C. J., and Coauthors, 2012: A 20 year independent record of sea surface temperature for climate from Along-Track Scanning Radiometers. J. Geophys. Res., 117, C12013, doi:10.1029/2012JC008400.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and Smith T. M. , 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, 929948.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Rayner N. A. , Smith T. M. , Stokes D. C. , and Wang W. , 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Smith T. M. , Liu C. , Chelton D. B. , Casey K. S. , and Schlax M. G. , 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 54735496.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., Chelton D. B. , Roberts-Jones J. , Martin M. J. , Menemenlis D. , and Merchant C. J. , 2013: Objective determination of feature resolution in two sea surface temperature analyses. J. Climate,26, 2514–2533.

  • Smith, T. M., Reynolds R. W. , Peterson T. C. , and Lawrimore J. , 2008: Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21, 22832296.

    • Search Google Scholar
    • Export Citation
  • van den Dool, H. M., Saha S. , and Johansson A. , 2000: Empirical orthogonal teleconnections. J. Climate, 13, 14211435.

  • Vidard, A., Anderson D. L. T. , and Balmaseda M. , 2007: Impact of ocean observation systems on ocean analysis and seasonal forecasts. Mon. Wea. Rev., 135, 409429.

    • Search Google Scholar
    • Export Citation
  • Woodruff, S. D., and Coauthors, 2011: ICOADS release 2.5: Extensions and enhancements to the surface marine meteorological archive. Int. J. Climatol., 31, 951967, doi:10.1002/joc.2103.

    • Search Google Scholar
    • Export Citation
  • Worley, S. J., Woodruff S. D. , Reynolds R. W. , Lubker S. J. , and Lott N. , 2005: ICOADS release 2.1 data and products. Int. J. Climatol., 25, 823842.

    • Search Google Scholar
    • Export Citation
  • Xue, Y., Smith T. M. , and Reynolds R. W. , 2003: Interdecadal changes of 30-yr SST normals during 1871–2000. J. Climate, 16, 16011612.

    • Search Google Scholar
    • Export Citation
  • Zhang, H.-M., Reynolds R. W. , and Smith T. M. , 2004: Bias characteristics in the AVHRR sea surface temperature. Geophys. Res. Lett., 31, L01307, doi:10.1029/2003GL018804.

    • Search Google Scholar
    • Export Citation
  • Zhang, H.-M., Reynolds R. W. , and Smith T. M. , 2006: Adequacy of the in situ observing system in the satellite era for climate SST. J. Atmos. Oceanic Technol., 23, 107120.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    (a) Niño-3.4 SSTA index from January to November of 2012 in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and ERSSTsatR15. (b) Difference of Niño-3.4 SSTA index in OISSTdy, ERSSTv3b, and ERSSTsatR15 with respect to OISSTwk. (c) Numbers of total (left-hand axis in a log scale) and super- observations (right-hand axis). A 5-month running-mean filter is applied in (b).

  • Fig. 2.

    SSTA of July 2012 in (a) OISSTdy, (b) OISSTwk, (c) in situ SSTA, and (d) ERSSTv3b. Shadings and contours are in unit of °C. The green rectangle box indicates the Niño-3.4 region.

  • Fig. 3.

    As in Fig. 2, but for November 2012.

  • Fig. 4.

    Monthly averaged biases of AVHRR SST in (a) OISSTdy in July 2012, (b) ERSSTsat in July 2012, (c) OISSTdy in November 2012, and (d) ERSSTsat in November of 2012. Contour intervals are 0.2°C. The green rectangle box indicates the Niño-3.4 region.

  • Fig. 5.

    Niño-3.4 region (5°S–5°N, 120°–170°W) averaged SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to in situ SSTA. (a) January–November 2012 and (b) 1982–2012. A 5-month running-mean filter is applied in (b).

  • Fig. 6.

    SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to TAO SSTA at 140°W. A 5-month running-mean filter is applied.

  • Fig. 7.

    SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to ATSR SSTA in Niño-3.4 region (5°S–5°N, 120°–170°W). A 5-month running-mean filter is applied.

  • Fig. 8.

    SSTA difference in OISSTdy, OISSTwk, ERSSTv3b, ERSSTsat, and HadSST3 with respect to AMSR-E SSTA in Niño-3.4 region (5°S–5°N, 120°–170°W). A 5-month running-mean filter is applied.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1118 651 175
PDF Downloads 226 68 8