This project aims to combine remotely sensed and in situ measurements with reanalyses and climate model projections to quantify the changes in the frequency of extreme events.  The projections will be based on multi-model IPCC AR4 data archives in order to assess model structure differences.  The project addresses both extreme high precipitation amounts and persistent low precipitation amounts.  We recognize that model projections and atmospheric models in general do not resolve moist processes well. The simulations of moist processes cannot be relied upon for neither climatology nor extremes. However the models do resolve large-scale (hemispheric) general circulation features that result from the interactions of radiative and dynamical processes.  At the same time extreme high precipitation is generally a local phenomenon that is not resolved in atmospheric models.  We will use statistical conditioning (also known as composites) to find the large-scale dynamical conditions that lead to extremes at local scales. This is done by first conditioning atmospheric reanalysis primitive states on the occurrence of extreme precipitation (as percentile of historical record). These composites provide the large-scale (resolved in models) conditions that lead to local scale extreme precipitation.  This approach has been successfully demonstrated (published results) for Italian Alps region and ECMWF reanalysis by the principal investigator and students.  Then, the frequency of the appearance of this pattern in an AR4 model with current climate is estimated. Finally the changes in the frequency of this atmospheric pattern in a greenhouse-gas forced AR4 model will be estimated.  A mapping of model-to-reanalysis primitive states fields has to be placed in between the steps.  At the other end of extremes, i.e. droughts, the conditioning is no longer on point precipitation gages because droughts are often seasonal in duration and cover large areas. For the droughts we will use merged satellite-based precipitation products as well as merged satellite vegetation index products to define the conditioning state or composite index.  The same analysis of the frequency of the large-scale atmospheric conditions in atmospheric reanalysis and then AR4 model projections will be performed. Again a mapping of reanalysis and climate model is necessary since the two may have biases relative to one-another.  The deliverable and ultimate results of the project are robust estimates of extreme low and high precipitation in a changed climate that uses only the resolved processes of the climate models.