In the context of economic development, decadal predictions are very worthwhile, since the time horizon of economic planning is often limited to 5-10 years. The MiKlip project PRODEF develops a tool for probabilistic decadal forecasts, which is based on statistical-dynamical downscaling (SDD).
Dynamical downscaling (DD) using atmospheric regional climate models (RCMs) often supports regional interpretation of global climate model (GCM) projections. Because DD is computationally very expensive, the question arises whether simpler methods of statistical or probabilistic downscaling can be used to achieve similar results. To attain resilient conclusions, the applied statistical techniques have to comprise more than relationships between large-scale climate indicators and local observations. SDD utilizes the added value of high resolution meteorological modelling in DD for the statistical and/or probabilistic refinement of GCM output.
The development of a combined SDD and probabilistic forecast tool in PRODEF consists of three steps: The identification of relevant weather clusters in larger scale forcing data, the DD for representative episodes for weather clusters and a construction of climate parameter distributions by “recombining” DD episodes. The forecast tool will be based on Monte-Carlo modelling using transition probabilities for weather types.
SDD will provide a computationally cost efficient tool to produce large numbers of ensembles of probabilistic projections for three different climate-related focal points: i) wind storms ii) wind potential for energy supply and iii) severe rainfall leading to floods. SDD presupposes statistical interpretation of large scale forcing factors and, therefore, implies an evaluation of GCM results with respect to problems of uncertainty and model bias.
PRODEF contributes to MiKlip Module C and has links to validation in Module E. SDD methods are a complementary approach to purely DD approaches and have advantage of being comparatively computationally cheap. PRODEF processes large ensembles of GCM data to produce a larger ensemble of downscaled predictions.
Institut für Geophysik und Meteorologie, Universität zu Köln
PD Dr. Joaquim G. Pinto
Dr. Kai Born
Dr. Mark Reyers
Dipl.-Met. Rabea Haas
Haas, R. | J.G. Pinto, and K. Born
Hueging, H. | R. Haas, K. Born, D. Jacob, J. G. Pinto
Haas, R. | Pinto, J. G.