Blending Multiresolution Satellite Data with Application to the Initialization of an Orographic Precipitation Model

Abstract

The use of multisensor, multifrequency satellite data to specify initial conditions for numerical weather prediction (NWP) models offers a unique opportunity to improve the depiction of small-scale processes in the atmosphere through a myriad of data assimilation approaches. The authors previously developed an algorithm to retrieve temperature and dewpoint profiles from a combination of infrared [high-resolution infrared radiation sounder (HIRS), 18?20-km resolution] and microwave [Advanced Microwave Sounding Unit-A (AMSU-A), 48-km resolution] data, using collocated radiosondes. Besides (and separately from) the estimation problem, one key question in the context of model initialization is how to blend multiresolution data to generate fields at the spatial resolution of the NWP model of interest. In this paper, a fractal downscaling technique is proposed to blend multiresolution satellite data and generate brightness temperature fields at 1-km resolution. The downscaled HIRS and AMSU-A data subsequently can be processed by the retrieval algorithm to derive temperature and dewpoint fields at the same resolution. The utility of these products as an initial condition for NWP models was assessed in the context of regional quantitative precipitation forecasting (QPF) applications using a limited-area orographic precipitation model nested with a mesoscale model. Results from the simulation of a wintertime storm in the Pocono Mountains of the mid-Atlantic region show improvement in QPF skill when the satellite-derived initial conditions were used. However, the disparity between the sparse times when the satellite data are available (12-h intervals) vis-a-vis the hourly import of boundary conditions from the host model lessens the impact of improved initial conditions. This result suggests that gains in QPF skill are linked to the availability of relevant remote sensing data at time intervals consistent with the useful memory of initial conditions in NWP models.