 |
| Titel |
Precipitation nowcasting from geostationary satellite platforms: Neural network methodology exploiting low-Earth-orbit and ground-based data synergy |
| VerfasserIn |
G. Rivolta, M. de Rosa, F. S. Marzano |
| Konferenz |
EGU General Assembly 2009
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 11 (2009) |
| Datensatznummer |
250031253
|
|
|
|
|
|
| Zusammenfassung |
| Many severe meteorological events develop at short time scales. The availability of effective
rain-rate nowcasting techniques is valuable for Civil Protection purposes. Neural network
based nowcasting techniques, exploiting satellite data, have been proven to be more accurate
than conventional techniques. Several rain retrieval techniques have been proposed on the
basis of multi-satellite imagery, exploiting passive sensor measurements acquired by
Geostationary-Earth-Orbit (GEO) and Low Earth Orbit (LEO) platforms. These
approaches tend to overcome some inherent limitations due to the use of satellite
thermal infrared (IR) radiances, which are measurements poorly correlated with
rainfall. In this respect, microwave (MW) radiometric data available from Low Earth
Orbit (LEO) platforms can provide more accurate rain estimates. MW brightness
temperatures are fairly sensitive to liquid and ice hydrometeors since rain clouds are
not optically opaque at microwave frequencies. GEO satellites can ensure Earth
coverage with a high temporal sampling, whereas LEO satellites have the drawback of
low temporal sampling. Therefore, LEO-MW and GEO-IR radiometry are clearly
complementary for monitoring the Earth’s atmosphere and a highly variable phenomenon
such as precipitation. The IR radiances from geostationary images can be properly
calibrated using microwave-based combined algorithms. Microwave data can be
extracted from the microwave imager sensors, but any rain estimation source may be,
in general, foreseen. Ground based meteorological radar reflectivity can also be
exploited.
The objective of this work is to identify guidelines for improving the neural-network
approach successfully applied to the rainfall field nowcast from thermal infrared and
microwave passive-sensor imagery aboard, respectively, Geostationary-Earth-Orbit (GEO)
and Low-Earth-Orbit (LEO) satellites, using infrared (IR) multi-channel data available from
Meteosat Second Generation (MSG) and microwave (MW) data from LEO satellites or
ground based meteorological Radar. The multi-sensor space-time prediction procedure,
being based on the Neural Combined Algorithm for Storm Tracking (NeuCAST),
consists of two consecutive steps: first, the infrared radiance fields measured from
geostationary satellite radiometer (e.g, MSG) are projected ahead in time (e.g., 30
minutes); secondly, the projected radiance field is used to estimate the rainfall field by
means of a MW-IR combined rain retrieval algorithm exploiting GEO-LEO or
GEO-Radar observations. The MSG NeuCAST methodology is here illustrated and
discussed. Its accuracy is quantified by means of quantitative error indexes, evaluated on
selected case studies of rainfall events in Southern Europe between 2003 and 2006. |
|
|
|
|
|
|