The use of adjoint models to improve weather predictions

During the second half of the XX century, the challenge of improving weather predictions with numerical models has received primary attention. The ever increasing demand for more precise initial conditions datasets has led to incorporate complex mathematical bodies, such as the 4D VAR method, to assimilate all kinds of data in the prediction system. The adjoint model is a key player within this context, only developed in some modern weather prediction suites. During recent years, other applications have emerged making use of the sensitivity fields that the adjoint model produces.

The talk will present a simple overview of what an adjoint model is, and will describe two applications recently developed that support the overall improvement of high-impact weather prediction. On the one hand, the adjoint model has provided an opportunity to incorporate the forecaster's knowledge explicitly in the prediction thread. On the other hand, a climatology of sensitivities of high impact weather in the Mediterranean is providing support to European weather services in the current readapting of observing strategies and permanent networks.

After presenting the adjoint model applications, a few complementary slides will show results from the UIB Meteorology group regarding the recent trends in temperature and precipitation over the Balearics. Finally, the high-resolution atmospheric products (http://mm5forecasts.uib.es and http://mm5forecasts.uib.es/OCLIBlocal/) available daily from our group will be described and offered to the interested researchers.