Tropical cyclone statistical rainfall climatology and experimental guidance derived from NCEP Stage IV data

Despite intensive research to improve tropical cyclone (TC) intensity and track forecasts, it is well-known that TC precipitation-induced flooding is the primary cause of deaths, at least between the years 1970-2000 (Rappaport 2000). Flooding caused by intense precipitation also is a major cause of TC-related destruction. In comparison, there is less research on TC precipitation forecasts which motivates the present study whose goal is to develop a reliable statistical operational TC precipitation forecast guidance product. This product is developed using National Centers for Environmental Prediction's Global Forecast System (GFS) operational forecast fields and Stage IV (STIV) precipitation data coupled with the National Hurricane Center's Best Track TC fixes.

Prior research conducted by Jones (1995) and Frank and Ritchie (1999) revealed the non-linear relationships between environmental vertical shear, TC motion, and storm structure . Since these publications, numerous studies using satellite data have focused on these relationships and the resultant asymmetric rain field (e.g., Lonfat et al. 2004; Chen et al. 2006; Wingo and Cecil 2010). Therefore, precipitation data from both STIV and GFS forecasts during the years 2004-2013 were compiled into earth-relative, storm motion-relative, and 200-850 hPa shear-relative datasets on a matching spatial grid. This facilitated comparisons of precipitation variance between observed and forecast environmental conditions unique to TCs. Precipitation fields from the GFS TC forecasts were based on its track and intensity forecasts, regardless of their accuracy, thereby assuming that even if the GFS provided an inaccurate track or intensity forecast for a specific storm, its precipitation forecast was not necessarily bad. Similar to the aforementioned studies, the observed and forecast datasets of TC rainfall were categorized into weak, moderate, and strong vertical shear environments and slow, moderate, and fast TC translational speeds. The effect of TC intensity on precipitation structure also was investigated. Considerations beyond the effects of vertical shear, storm motion, and intensity, such as terrain, location along the U.S. coast, and baroclinicity, were also explored to help explain additional precipitation variance. Parameters from each dataset, as well as criteria for vertical shear, storm motion, and intensity, were combined statistically to create storm- and environment-specific TC precipitation forecast guidance. Results of these statistical forecasts then were compared to the benchmark Rainfall Climatology and Persistence (R-CLIPER) model's forecasts to determine if this rainfall guidance product is superior to climatology. The forecasts also were tested for accuracy against test cases using metrics similar to Marchok et al. (2007). We quantify the ability of our forecast models to 1) match the observed rainfall pattern, 2) match the mean rainfall and the distribution of rain volume, and 3) produce extreme TC-related precipitation amounts. This presentation will describe our procedures and results in much greater detail.