Modem numerical weather prediction models must be necessarily specified initial and boundary conditions as basic state vectors. Errors in these conditions affect the model forecast at verification time. An adjoint model efficiently determines where these errors have the greatest impact on the for... Show moreModem numerical weather prediction models must be necessarily specified initial and boundary conditions as basic state vectors. Errors in these conditions affect the model forecast at verification time. An adjoint model efficiently determines where these errors have the greatest impact on the forecast, or, in other words, where the model is most spatially and temporally sensitive. Such a model calculates the sensitivity gradient and total sensitivity of a forecast aspect, or difference in a specified meteorological variable, with respect to the basic Slate vector. This study investigates the sensitivity of a 48-h forecast on 4-6 April 1982 using the Mesoscale Adjoint Modeling System (MAMS) developed at the National Center for Atmospheric Research (NCAR). We determine which basic state variables the nonlinear forecast model (NLM) within MAMS is most sensitive to, show the sensitivity gradient plots that indicate the synoptic prerequisites for cyclogenesis, and prove that the NLM is more sensitive to initial conditions. This research is intended as a model for future sensitivity studies on existing operational regional forecast models. Show less
