|
Titel |
Bayesian inverse modeling for quantitative precipitation estimation |
VerfasserIn |
Katharina Schinagl, Christian Rieger, Clemens Simmer, Xinxin Xie, Petra Friederichs |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250143204
|
Publikation (Nr.) |
EGU/EGU2017-6906.pdf |
|
|
|
Zusammenfassung |
Polarimetric radars provide us with a richness of precipitation related measurements.
Especially the high spatial and temporal resolution make the data an important
information, e.g. for hydrological modeling. However, uncertainties in the precipitation
estimates are large. Their systematic assessment and quantification is thus of great
importance.
Polarimetric radar observables like horizontal and vertical reflectivity ZH and ZV ,
cross-correlation coefficient ρHV and specific differential phase KDP are related to the drop
size distribution (DSD) in the scan. This relation is described by forward operators which are
integrals over the DSD and scattering terms. Given the polarimetric observables, the
respective forward operators and assumptions about the measurement errors, we investigate
the uncertainty in the DSD parameter estimation and based on it the uncertainty of
precipitation estimates.
We assume that the DSD follows a Gamma model, N(D) = N0Dμ exp(−ΛD), where
all three parameters are variable. This model allows us to account for the high variability of
the DSD. We employ the framework of Bayesian inverse methods to derive the posterior
distribution of the DSD parameters. The inverse problem is investigated in a simulated
environment (SE) using the COSMO-DE numerical weather prediction model. The
advantage of the SE is that - unlike in a real world application - we know the parameters
we want to estimate. Thus, building the inverse model into the SE gives us the
opportunity of verifying our results against the COSMO-simulated DSD-values. |
|
|
|
|
|