The 24-25 January 2000 winter storm was noteworthy as operational numerical weather prediction (NWP) model guidance for the cyclone was particularly poor. For shorter forecast lead times, while model guidance regarding the surface cyclone position and intensity on 25 January improved, guidance for the precipitation remained extremely poor. With lead times as short as 24 hours, forecasts of an intense mesoscale precipitation band remained too far to the east.

A sensitivity study allows one to evaluate how a particular function of a NWP model forecast state changes with respect to changes in the model control variables. The calculation of forecast sensitivities allows one to evaluate the sensitivity of a variety of response functions with respect to the model state at some prescribed time. In this study, forecast sensitivities are used to diagnose why improvements in the cyclone position and intensity were not necessarily associated with improvements in the precipitation forecast. These sensitivities are most efficiently evaluated using the adjoint of a forecast model.

In this presentation, we calculate the sensitivities of a variety of response functions for the January 2000 case using the adjoint of the Penn State/NCAR MM5 version two model. The response functions chosen for this study include 'energy weighted' forecast error, lower tropospheric circulation, horizontal frontogenesis, and vertical motion. We interpret similarities and differences in the gradients of the various response functions with respect to the model state vector as well as with respect to derived quantities including relative vorticity and potential vorticity.