U.S. Geological Survey Open-File Report 2010-1251
Moderate-Resolution Sea Surface Temperature Data for the Nearshore North Pacific
AVHRR SST Data
The Advanced Very High Resolution Radiometer (AVHRR) is a multi-generational, 6-channel instrument that has been flown on board a series of NOAA polar-orbiting satellites, and has been returning reliable SST data since 1981. Sea surface temperature is not measured directly; rather, it is the result of a differentiation of brightness temperatures recorded by two of the sensor's infrared (IR) channels (channels 4 and 5). It has been shown that atmospheric aerosols and particulates (for example, volcanic ash or Saharan dust) are sources for error in IR brightness temperature measurements, and potentially introduce a cold bias (Vazquez-Cuervo and others, 2004). However, one advantage of using satellite IR retrievals to derive SST is that the long IR wavelengths are less affected by atmospheric contaminants (such as cloud or fog). Another benefit of IR wavelength measurements is that daylight is not a requirement for data collection. And while nighttime AVHRR data have been preferred for some SST analyses, it has been shown that nighttime-only AVHRR Pathfinder data do not necessarily produce superior results with respect to reproducing in situ temperature measurements (Casey, 2002). Therefore, we chose to use the daytime series of the AVHRR Pathfinder 4 km dataset based on data availability. We found daytime data to be more abundant in our near-coastal regions of interest, especially in areas that experience frequent evening ground fog, such as the coasts of the states of Washington and Oregon.
The AVHRR Pathfinder v.5 (and later) datasets were developed jointly by the University of Miami Rosenstiel School of Marine and Atmospheric Science and the NOAA National Oceanographic Data Center (NODC) as a more accurate, downscaled (9.28 to 4 km) version of a previous global AVHRR dataset (Vazquez-Cuervo and others, 1998). We obtained the version 5 (PFSST V50) raster data from the NASA JPL Physical Oceanography Distributed Active Archive Center (PO.DAAC) (ftp://podaac.jpl.nasa.gov/pub/sea_surface_temperature/avhrr/pathfinder/data_v5/). AVHRR products (PFSST V50 and PFSST V51) are also available through the NODC (ftp://data.nodc.noaa.gov/pub/data.nodc/pathfinder). The source data were provided in HDF-SDS (Hierarchical Data Format Scientific Data Set) format. For a detailed description of the AVHRR and Pathfinder algorithms, we direct the reader to Vazquez-Cuervo and others (1998; 2010) and Kilpatrick and others (2001).
AVHRR Data Quality
The PFSST data include a quality flag product in which each SST pixel is designated a value ranging from 0 (worst quality) to 7 (best quality). These quality flags convey the level of confidence attributed to the SST value calculated for each pixel location. The level of confidence is evaluated on pixel-by-pixel performance with respect to a number of tests that estimate validity and consistency of brightness temperature readings, sun angle effects, and cloudiness, which are combined to establish an overall quality rating. The version 4 Pathfinder release of the SST dataset (PFSST V41) included a standard product called "best SST," or "BSST." BSST data includes pixels with quality flags greater than 3 (Kilpatrick and others, 2001). We generated an analogous SST product from the PFSST data by disregarding SST values with corresponding quality flag values of less than 4. Despite the rigors of the flagging algorithms, a small number of pixels with illegitimate jumps in SST gradient have been detected (Evans and others, 2009). These jumps in the SST gradient must be detected and removed by the user depending on their data requirements.
Currently the only reports on accuracy of Pathfinder SST values are linked to specific studies across a variety of spatial and temporal resolutions, pathfinder versions, and quality flag thresholds. In general, the temperature values are reported to have RMS errors between 0.1 and 1.0 °C (Vazquez-Cuervo and others, 2010; Xu and Ignatov, 2010, Kearns and others, 2000; Kirkpatrick and others, 2001) when compared to in situ temperatures. However, it must be cautioned that in situ temperature data derived from multiple sources, such as moored buoys and shipboard observations, are prone to large random errors and rarely have excellent agreement amongst them (e.g. Kearns and others, 2000). Furthermore, in situ measurements are of bulk temperatures (typically taken between 1 – 3 meters depth) rather than true sea surface temperature. Lastly, comparison to in situ data sources such as buoys and the National Centers for Environmental Prediction (NCEP; Reynolds and others, 2002) are somewhat contrived as these data (Reynolds and others, 2002) provide critical calibration and validation for the PFSST data product (Kilpatrick and others, 2001; Vazquez-Cuervo and others, 2010). In practice, when the scaling algorithm provided with the AVHRR data is applied, the resulting SST values imply a 1/1000 of a degree precision, a level of precision that is beyond the limitations of the original processing algorithm. Based on these considerations, the accuracy of the nearshore SST data is approximately 0.5 °C (personal communication, Jorge Vazquez-Cuervo, February 3, 2011).
Coastal SST Product
We selected AVHRR data for their global coverage at moderate resolution (4 km), their long data record (~30 years) relative to other satellite missions , and their substantial level of processing, including extensive calibration and atmospheric correction. The data represent a time-series of monthly mean sea surface temperatures over the last 29 years. The data presented here encompass the nearshore region of the entire North Pacific, from the coast to approximately 16 km from the shoreline, as defined by the Global Self-consistent, Hierarchical, High-resolution Shoreline (GSHHS) version 1.3 (Wessel and Smith, 1996). In some places, the data appear jagged, or "moth-eaten," immediately adjacent to the coast. There are two reasons for this. The first is that satellite data for some regions of the coast are missing or are unusable due to low quality. The second reason for the jagged appearance of the data is that the Pathfinder developers used a MOD12Q1 land mask, which has a 1-km resolution rather than the 0.2-km resolution of the GSHHS. As detailed below, we processed the data using a number of tools to create several nearshore SST products, including georeferenced North Pacific SST shape and raster files, along with their metadata, and Access 2003 Databases.