Extended Abstract (1.2M)JP8.1
Web-based global quality control and monitoring of NESDIS AVHRR SST products for long term stability and cross-platform consistency in near real-time
Prasanjit Dash, NOAA, Camp Springs, MD; and A. Ignatov, Y. Kihai, J. Sapper, and X. Liang
Since the early 1980s, sea surface temperature (SST) products have been operationally generated by NESDIS from the Advanced Very High Resolution Radiometers (AVHRR) onboard NOAA satellites. Ensuring their quality, stability, and cross-platform consistency in near real-time is critically important for many SST applications.
The AVHRR SST products generated by the heritage Main Unit Task (MUT) system are validated against collocated in-situ SSTs approximately once a month. This validation methodology remains the “gold standard” used to measure the accuracy and precision of AVHRR SST. However, this customary approach has its own limitations. In-situ data are sparse and geographically biased, which may cause the validation statistics to lack global representation. Furthermore, the accuracy of in-situ SSTs may be suboptimal and highly non-uniform as data with different measurement protocols originate from different sensors manufactured and maintained by different countries and programs. Although the number of in-situ match-ups is now quite large, it still requires up to a month to collect enough collocated data points, perform reliable quality control of in-situ data, and generate trustworthy validation statistics. Still, the quality of SST product in remote oceanic areas of the globe may remain uncertain.
To monitor global satellite SST products for quality, stability, and cross-platform consistency in near-real time, another approach is needed that is global in nature. In this presentation, we explore a methodology based on statistical analysis of anomalies in AVHRR SST (TS) with respect to global reference SST fields (TR), available from either optimally interpolated blended satellite/in-situ analysis SST or from a climatological SST. The underlying assumption is that the probability density function of global anomaly is close to a Gaussian distribution (although the distributions of TS and TR are highly asymmetric). In this work, SST products from NOAA 16, 17, 18, and MetOp-A from 2004 to the present are compared against the following SST reference fields: weekly Reynolds-Smith OI.v2 SST, two daily Reynolds OI SSTs (AVHRR-based and AVHRR+AMSR-E based), RTG low and high resolution, OSTIA, ODYSEA, and Bauer-Robinson climatology. The four satellite products are also evaluated for cross-platform consistency.
Histograms of SST anomalies are plotted, and their conventional and robust statistics are calculated. In addition, the median and a robust standard deviation are used to identify and remove outliers for SST global quality control, and empirical histograms are analyzed for proximity to a Gaussian shape. The first four moments of the empirical distribution are calculated both before and after outliers are removed. Subsequently, time series of the mean and median, conventional and robust standard deviations, skewness, kurtosis, and fraction of outliers are over-plotted for four platforms. Although the absolute values of the moments do depend upon a given reference SST, this does not affect the monitoring of their stability and cross-platform consistency in the long-term time series. Overall, all products show high degree of stability and cross-platform consistency, except for SST from NOAA16, whose AVHRR is known to have scan mirror problems.
In addition to these time series, global maps of anomalies are generated to facilitate analyses of geographical distribution of the respective anomalies. Also, SST anomalies are plotted as a function of retrieval conditions (e.g., view zenith angle, solar zenith angle, number of retrievals per cell in gridded data, local time, scattering angle, and glint angle) to ensure self-consistency of the product in the full retrieval domain. The maps and artefactual trend-plots are also animated to separate out persistent artifacts from regular features in the data. Such animated plots are helpful in identifying and fixing systematic and random errors related to SST algorithm, cloud mask, or sensor malfunctioning.
The global QC tool has been implemented for the heritage MUT SST products and is also currently being tested for the newer AVHRR Clear-Sky Processor over Oceans (ACSPO) currently operational in NESDIS. The implementation is based on IDL codes and UNIX scripts and the results of diagnostics are automatically posted on the web at www.star.nesdis.noaa.gov/sod/sst/gqt in near-real time. This tool can be easily adopted for other platforms/sensors products. We are currently adopting it to analyze the MSG/SEVIRI data. In the future, we also plan to employ it to monitor the SST products produced from NPOESS/VIIRS and GOES-R/ABI.
Supplementary URL: http://www.star.nesdis.noaa.gov/sod/sst/squam/
Joint Poster Session 8, Operational Products and Transition from Research to Operations
Wednesday, 14 January 2009, 2:30 PM-4:00 PM, Hall 5
Next paper