In addition to numerical model-derived TC precipitation forecasts, previous attempts to forecast TC precipitation have been based on passive microwave satellite retrievals. The most popular of these methods are the Rainfall-Climatology and Persistence (R-CLIPER) forecast, the Parametric Hurricane Rainfall Model (PHRaM), and the Tropical Rainfall Potential (TRaP) technique. Each uses an instantaneous precipitation field to provide a persistence forecast. These techniques usually assume that precipitation exponentially decays away from the radius of maximum precipitation. While these are well-reasoned methods of producing TC rainfall guidance, precipitation fields within a TC can change rapidly due to factors such as topography and changing vertical wind shear. Previous research (e.g., Chen et al., 2006) and our own have shown that precipitation for TCs making landfall is primarily asymmetric with a downshear-left maximum. This maximum is the result of cyclonically rotating convection and further cyclonic advection of precipitation around the core of the storm. While some of these previously developed persistence models correct for shear and topography, the assumption of persistence can lead to serious errors. Therefore, numerical weather prediction models can provide a more accurate forecast since they are not limited by assuming the precipitation field around the storm.
The primary goal of this paper is to investigate precipitation patterns and amounts associated with TCs that make landfall, or come within 300 km without making landfall, along the U.S. Atlantic and Gulf of Mexico Coasts using National Centers for Environmental Prediction Stage IV (STIV) and Global Forecast System (GFS) data. The coasts are divided into ten geographical regions determined by locations of landfall, or closest pass by land within 300 km without landfall, during the years 2004–2013. This paper first will present preliminary results investigating TC precipitation variability between these geographic regions using STIV data as a function of storm motion, vertical shear, and storm speed. Second, we focus on comparing observed STIV rainfall with forecast TC precipitation from the GFS. Similar to the STIV analysis, GFS TC precipitation variability within the geographic regions will be presented as a function of storm motion, vertical shear, storm speed, and forecast hour. We seek to determine spatial and temporal biases that can be used to "correct' the numerical guidance. This research is part of a broader study aiming to provide a more valuable TC precipitation forecast product than currently exists. Future research will seek to produce corrections to the GFS-derived predictions so that improved predictions of rainfall can be made.