National and Kapodistrian University of Athens
Ανάπτυξη μεθοδολογίας για την αξιολόγηση προγνωστικών πεδίων βροχόπτωσης υψηλής χωρικής ανάλυσης σε πολύπλοκο ανάγλυφο με τη χρήση ρεαλιστικών πλεγματικών παρατηρήσεων
Abstract
dc:descriptionVerification of precipitation forecasts is a very challenging task. High-resolution models are able to produce more detailed precipitation structures, but their primary benefit lies in delivering more realistic fields rather than the information provided for a specific grid point. Spatial verification methods generally reward closeness or resemblance by relaxing the requirement for exact matches between forecasts and observations that is the cause of double penalty errors in traditional approaches. Spatial precipitation fields derived from numerical weather prediction (NWP) model output are produced in the form of grids. In order to assess their accuracy with spatial verification methods, it is necessary to transpose any available observations (ground stations, radar products, satellite estimates) to the same grid. The challenge is to overcome the lack of density of observations in both time and space to accurately represent the highly variable nature of precipitation. In this thesis, the MISH (Meteorological Interpolation based on Surface Homogenized Data Basis) method is used to estimate the distribution of precipitation on both hourly and daily timescales. The idea behind the MISH interpolation scheme stems from the principles that gridded data can be created at higher quality when correlated with certain climate statistical parameters (geophysical characteristics), which can be modelled using long climate data series. The modelling of the statistical parameters for a given location is based on the long-term homogenized monthly data of neighbouring stations. In order to better exploit the information from stations neighbouring each grid point, remote sensing precipitation estimates are added as background information, which is evaluated in combination with the interpolated value. Subsequently, the precipitation forecasts are assessed using various spatial methods in an effort to identify the ability of each verification approach to provide distinctly useful information on model performance. The methods chosen represent different approaches in model evaluation strategy, such as that of neighborhood or fuzzy, scale separation and object-oriented techniques. Model output of the limited area COSMO model at two different resolutions is compared to the gridded precipitation fields created by MISH for four different test cases over Greece in an effort to obtain answers to questions related to their performance at different scales, to their location errors, intensity errors and distributions and to the precipitation forecast structure errors. Finally, the development of an integrated verification system for precipitation forecasts is proposed that meets the requirements of an operational system. The evaluation framework can flexibly incorporate all available observation sources and meet the needs of a wide range of users based on the various statistical approaches it applies. The system is adapted to the specificities of the Greek territory, consequently it can provide sufficient feedback on the performance of very high resolution NWP models in areas with extreme topography where in situ observations are too limited to accurately represent the phenomena.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Gofa, Flora
- Γκόφα, Φλώρα
Subjects
dc:subject × 12Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/48074