214
C.J. Donlon
toward increasing use of simulations in near-real time from national weather programs to inform the discrimination – either by dynamically calculating thresholds or
as input to a probabilistic calculations (Merchant et al., 2005). Further development
of this approach is expected in the future.
Cloud screening of TIR satellite data is remains a significant challenge and more
effort is required to develop effective systems to minimize the data loss due to inappropriate cloud screening and the increase in error where clouds are not properly
detected.
13.3.3 Improved Treatment of Atmospheric
Aerosol Contamination
The performance of TIR derived SST retrievals is degraded in the presence of atmospheric aerosols (e.g., Saharan dust, volcanic eruptions). This has been a particular
problem for the Meteosat-8 SEVIRI instrument SST retrieval. During the initial
phase of operations the occurrence of Saharan dust outbreaks lead to SST bias errors
of ~1 K. These problems have been mitigated to a certain degree by upgrading the
MSG algorithms to include a Saharan dust index scheme (Merchant et al., 2006)
and the use of ENVISAT AATSR data to derive a bias correction for the aerosol
(and other) contaminated data. There are several aspects to improving atmospheric
aerosol detection and flagging algorithms that will provide increased sensitivity and
performance:
1. in strong SST gradient regions,
2. when sub-pixel clouds and optically thin cirrus are present,
3. when only limited instrument channels are available,
4. when aggregated data are used (e.g. AVHRR GAC),
5. when multi-angle view data are available,
6. based on multi-satellite synergy (e.g. use of geostationary data, (A)ATSR, and
passive microwave sensors),
7. based on probabilistic techniques,
8. based on improved conventional threshold, histogram and spatial coherence
techniques.
It is expected that significant progress will be made in the next decade on these
issues as climate quality SST data sets are derived for a variety of TIR sensors.
13.3.4 Improving Current and Future SST Measurements
Through Better Uncertainty and Error Estimation
A key user request from all user communities (and in particular the SST community)
is the provision of uncertainty estimates to be attached to each pixel in SST products.
A framework has emerged from the GHRSST activity called Single Sensor Error
C.J. Donlon
toward increasing use of simulations in near-real time from national weather programs to inform the discrimination – either by dynamically calculating thresholds or
as input to a probabilistic calculations (Merchant et al., 2005). Further development
of this approach is expected in the future.
Cloud screening of TIR satellite data is remains a significant challenge and more
effort is required to develop effective systems to minimize the data loss due to inappropriate cloud screening and the increase in error where clouds are not properly
detected.
13.3.3 Improved Treatment of Atmospheric
Aerosol Contamination
The performance of TIR derived SST retrievals is degraded in the presence of atmospheric aerosols (e.g., Saharan dust, volcanic eruptions). This has been a particular
problem for the Meteosat-8 SEVIRI instrument SST retrieval. During the initial
phase of operations the occurrence of Saharan dust outbreaks lead to SST bias errors
of ~1 K. These problems have been mitigated to a certain degree by upgrading the
MSG algorithms to include a Saharan dust index scheme (Merchant et al., 2006)
and the use of ENVISAT AATSR data to derive a bias correction for the aerosol
(and other) contaminated data. There are several aspects to improving atmospheric
aerosol detection and flagging algorithms that will provide increased sensitivity and
performance:
1. in strong SST gradient regions,
2. when sub-pixel clouds and optically thin cirrus are present,
3. when only limited instrument channels are available,
4. when aggregated data are used (e.g. AVHRR GAC),
5. when multi-angle view data are available,
6. based on multi-satellite synergy (e.g. use of geostationary data, (A)ATSR, and
passive microwave sensors),
7. based on probabilistic techniques,
8. based on improved conventional threshold, histogram and spatial coherence
techniques.
It is expected that significant progress will be made in the next decade on these
issues as climate quality SST data sets are derived for a variety of TIR sensors.
13.3.4 Improving Current and Future SST Measurements
Through Better Uncertainty and Error Estimation
A key user request from all user communities (and in particular the SST community)
is the provision of uncertainty estimates to be attached to each pixel in SST products.
A framework has emerged from the GHRSST activity called Single Sensor Error