in situ SSTs are critical in satellite SST Cal/Val. In SST community, calibration means to derive coefficients for regression algorithm, and validation means to monitor and evaluate quality of satellite product. However, the quality of in situ SST are highly non-uniform due to several reasons. First, in situ data are measured by a large variety of sensors on different platforms, e.g. ships, drifting and mooring buoys. Second, extra errors are induced during transmission and processing. For buoys, they are often operated in hostile environment and some may be left unattended for long time. For Ships, the measurments may be impacted by human operations. Therefore, quality control (QC) is absolutely necessary before in situ data can be further used in Cal/Val.
A few erroneous measurements could corrupt the whole data set and thus significantly affect Cal/Val results. A demonstration is given below: the Standard Deviation (SD) of in situ SST minus Pathfinder SST (from satellite), before QC (top) and after QC (bottom). It clearly shows that removing only 2~4% report will significantly improve the quality of the data. Without QC, the data cannot be used in Cal/Val.
Currently in remote sensing community, QC is usually overly simplistic and non-objective, e.g. ±2K wrt. climatology. On the other hand, QC can be much more sophisticated as usually done in meteorological Community. Such advanced QC is necessary because different types of errors should be handled in different ways and QC should be more robust to inaccurucies in the reference SST. QC algorithm employed by iQuam is described in algorithm tab.
iQuam is a system developed in NESDIS/STAR to perform near real time QC of in situ measurements and monitor the statistics. Ships and buoys (drifters and moorings) are included. Source data are available from NCEP GTS in every 12 hours. QC are then performed and quality information are appended. QCed in situ data in self-described HDF format are available via FTP. Currently, all processing of the previously month data are done in the 5th day. Near real time (daily or hourly) updates for currently month data will be provided in later versions.
The system is orgnized as follows.
Algorithm and configrations used in QC processing are described in algorithm tab.
Explanations and tips of web interfaces of iQuam are described in monitor interface tab.
QC algorithm implemented in iQuam consists of different steps of prcessing which can be categorized into five groups:
In addition to some basic processing steps, 5 major steps are duplicate removal (DR), platform track check (TC), SST spike check (SC), reference check (RC) and cross-platform check (XC). Algorithms for the reference check and the cross-platform check are employed from the currently existed Bayesian QC method [Lorenc and Hammon, 1988; Ingleby and Huddleston, 2007] with minor modifications. Details can be found in publications:
Flowchart of QC algorithm is as follows.
Monitoring of statistics of both QC and SST anomalies are also provided in iQuam. It includes monthly global map, monthly QC/SST statistics, long-term time-series and individual platform statistics (track map, SST time series and performance history).
Interfaces are shown below.