engleski

Sensor specific error statistics: a case study of the AVHRR-derived Adriatic SST

The sea surface temperature (SST) derived from the Advanced Very High Resolution Radiometer (AVHRR) sensors is well investigated parameter but due to the empirical nature of the regression coefficients, SST algorithms need calibration and validation with in situ measurements. In the era of operational monitoring and global forecasting of the ocean there is further emphasis on accuracy of the SST products (less then 0.4 K), hard to obtain with just one-sensor measurements. The recently initiated Global High-Resolution Sea Surface Temperature Pilot Project (GHRSST-PP) aims to combine several independent SST measurements into a single product with well defined error bounds. With such requirements in mind, we have carried out a detailed Adriatic centered investigation of sensor specific error statistics (SSES) for the AVHRR sensors aboard NOAA 16 and NOAA 17, for both daytime and nighttime passes. Results are compared with SSES statistics for MODIS aboard AQUA and TERRA. Various parameters were investigated to assess the magnitude of the SSES changes in time and space. Two different in situ databases (drifter buoys and fixed-platform SST measurements), and two sets of algorithms were used in the analysis. Traditional split-window multi-channel algorithms and NOAA/NESDIS operational algorithms (non-linear split for day and multi-channel triple for night (“ Andy” )) were used to derive daytime and nighttime SST estimates. To investigate a possible Adriatic sub-domain SSES variability the Adriatic Sea was divided in different domains: the open sea, the coastal zone and the Po-river influenced coastal zone. The third consideration was given to the influence of wind, derived from LAMI atmospheric model output (at 7 km resolution). Wind data were interpolated at matchup points and SSES were derived for low winds (below 2.5 m/s), medium winds (between 2.5 and 6 m/s) and high winds (higher than 6 m/s). Results exhibit lowered nighttime scatter (~0.3 K) for algorithms using short IR channels, for all sensors and satellites and regardless of algorithm form. Daytime results for both algorithms were masked by diurnal warming effect, exhibiting constant positive bias (0.3 K) with higher scatter (0.7 K) in summer season for both satellites. The comparison of Adriatic domains suggests a small, but increased bias for coastal regions, for afternoon passes and the downstream of the Po delta in particular.

Sensor-specific error ; AVHRR ; Adriatic SST

nije evidentirano