New methods for evaluating rainfall forecasts from operational models for landfalling tropical cyclones

Over the past few decades, significant research has been conducted into improving tropical cyclone track and intensity forecasts, while relatively little work has been done to improve tropical cyclone rainfall forecasts. This is at least partly due to a lack of rainfall forecast validation schemes designed specifically for landfalling tropical cyclones. We have addressed this issue by developing a set of validation techniques and objective skill metrics specific for landfalling tropical cyclones that provide a baseline measure of quantitative precipitation forecasting (QPF) skill.

In this presentation, we will cover the following topics:

* Development of new TC QPF validation techniques: In order to capture the unique characteristics of TC rainfall, we have developed validation techniques that address these four elements of TC rainfall: (1) A model's ability to match the large-scale rainfall pattern; (2) A model's ability to match the mean rainfall and the distribution of rain volume; (3) A model's ability to produce the extreme rainfall amounts often observed in TCs; and (4) The impact of a model's track forecast error on its QPF skill.

* Development of objective QPF skill metrics: Based on the above validation techniques, QPF skill metrics are derived that allow for objective comparison of QPF performance among dynamical models and against the benchmark R-CLIPER model.

* Operational model QPF verification for landfalling U.S. storms, for (1) the 1998-2004 seasons; and (2) the 2005 season: Using the newly developed techniques, validations will be presented for rainfall forecasts from the operational versions of the GFDL, GFS, NAM and R-CLIPER models. Results from the initial 7-year sample will be compared to those from the active 2005 Atlantic hurricane season.