STSM Topics

If you're interested in one of the topics proposed below please contact the host personally and discuss the possibility of handing in an STSM application for this topic.
 

1. Evaluation of  Rglimclim for "hard" downscaling applications

Host: Richard Chandler, University College London

Proposed length: 3 months

Description: Generalised linear models (GLMs) are increasingly becoming accepted as one of the leading methodologies for statistical downscaling. However, their effective use requires a good level of statistical awareness, so they are typically used only for "hard" applications in which it is really important to reproduce adequately a wide range of weather characteristics including timing, spatial organisation and rare (e.g. 100-year) extreme events. The Rglimclim software package uses GLMs to generate daily  multisite, multivariate weather sequences and can be used for downscaling applications. The multivariate functionality is new and relatively untested, however. It is therefore of interest to examine its performance. One possibility would be for the visitor to build models for their own data using Rglimclim and to compare the results with those already obtained using other methods. Another would be to participate in existing projects using Rglimclim, to gain experience with the software and insight into the kinds of applications where it is being used. The HydEF project would be particularly relevant, as this has partially funded the development of Rglimclim - the visitor would have the opportunity to interact with a varied team of climatologists, hydrologists and hydrogeologists as well as learning some statistics.

 

2. Including meteorological predictors for stochastic bias correction

Host: Douglas Maraun, GEOMAR Helmholtz Centre for Ocean Research, Kiel

Proposed length: 2-3 months

Description: Within the PLEIADES project, we have proposed a stochastic bias correction approach. Instead of transforming RCM simulated precipitation into a single corrected precipitation value, as done in classical bias correction methods, we developed a method that predicts a full corrected distribution for each day. Currently, the only predictor used in this model is simulated precipitation. Yet it is anticipated that predictors carrying additional information on regional scale circulation and humidity might improve the downscaling. The aim of the proposed STSM is therefore to identify potential meteorological predictors, include them into our model and validate the performance of this extended model compared to our standard model.