 |
| Titel |
Sampling uncertainties of precipitation estimates from satellites using a climatological radar rainfall data set |
| VerfasserIn |
A. Overeem, H. Leijnse, R. Uijlenhoet |
| Konferenz |
EGU General Assembly 2012
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 14 (2012) |
| Datensatznummer |
250063790
|
|
|
|
|
|
| Zusammenfassung |
| Satellites provide precipitation estimates over large areas and often for areas where no or few
rainfall observations are available, such as oceans and developing countries. The launch of
new satellites, such as the Global Precipitation Measurement (GPM) mission, will
result in precipitation estimates over the tropics and midlatitudes with a revisit time
or sampling time interval of 3 hours and a horizontal resolution of 4-5 km, the
footprint.
It is important to quantify the effect of revisit time, footprint and integration time on
the quality of the precipitation estimates. This can be done by simulating satellite
precipitation images from rain gauges or ground-based weather radar data. The
number of rain gauges is often too small to obtain accurate areal rainfall depths,
especially for subdaily integration times. In contrast, radar data give precipitation
estimates over large areas with high spatial and temporal resolutions. Nevertheless,
radar data may suffer from several (sources of) errors, and long-term, high-quality
radar rainfall data sets with a large number of possible rainfall levels are often not
available.
Two adjustment methods, which use rain gauge data from two networks, were combined to
obtain a high-quality radar rainfall data set suitable for hydrological and climatological
applications. This 11-year data set (1998-2008) has a temporal resolution of 5 min
and covers the entire land surface of the Netherlands (35500 km2). Recent studies
show that this data set is suitable to derive probability distributions of extreme areal
rainfall.
Satellite rainfall depths are simulated from this climatological radar rainfall data set for revisit
times of 15 min to 24 hours, integration times of 1 hour to 1 month, and footprints of
approximately 25 to 1500 km2 and the entire land surface of the Netherlands (35500 km2).
These characteristics are representative for satellite missions such as GPM or Meteosat
Second Generation (MSG).
For instance, a revisit time of 3 hours and an integration time of 24 hours means that 8 5-min
radar rainfall accumulation images are used to compute a 24-hour simulated satellite rainfall
depth for a chosen footprint. Subsequently, the simulated satellite rainfall depths are verified
against the radar rainfall data set, the ground-truth, having a revisit time of 5 min and the
same footprint. Because the same data set is used for both simulation as well as verification,
errors can be solely attributed to sampling uncertainties. Verification results are shown for
different combinations of revisit time, integration time and footprint. For instance, a spatial
verification is given by comparing monthly rainfall depths. Seasonal dependence in
sampling uncertainties is studied, which is possible owing to the long radar rainfall data
set. |
|
|
|
|
|
|