Improved Telescopic Nesting and Accurate Storm Tracking in the NCEP Operational HWRF Model

The goal of the NCEP operational HWRF model is to produce a high quality five day forecast of the track and intensity of every storm requested by the National Hurricane Center every six hours- typically over a thousand simulations per year, while fitting within a tight time window and limited computational resources in operations. Starting with the 2012 hurricane season, the HWRF model was upgraded to a telescopic nesting model with 3km and 9km storm-following computational domains within a large 27km stationary domain. This model was successful in many ways, but was found to have several problems related to nest movement and associated interpolation procedures. This talk discusses the issues, suggested solutions implemented for 2013 HWRF, and future enhancements geared towards implementation for 2014 hurricane season.

The mechanics of finding the storm, following it, and passing mass and energy between moving domains is far more complex than it sounds. Terrain height is different in the three domains, sometimes by more than a kilometer, necessitating mass adjustment procedures to maintain numerical stability. This must be done in different ways depending on physics chosen, especially if one is to fit the model within the 80 minute window for running the operational HWRF model. Another matter is where to move the nests: clearly they must follow the storm, but where is the storm?

Tracking a hurricane in a 3km domain is a fairly easy task regardless of the method chosen, but tracking a decaying remnant of a tropical depression that never even merged into a single vortex is another matter altogether. As we found in the 2012 HWRF, the nest motion algorithm has to track even the hardest cases, otherwise the 9km or 27km domains are the only ones seeing the storm, and they are not sufficient to resolve the storm structure necessary for accurate intensity forecasts. The new 2013 and 2014 algorithms incorporate aspects of the GFDL Vortex Tracker and the Unified Post Processor, directly into the WRF-NMM model, but parallelize them to the point where they are no longer an impact on model runtime. The result is a nest motion algorithm that can track even the most extreme fringe cases.

The nest motion algorithm and nest-to-parent interpolation algorithms require some inputs that are relatively expensive to calculate, such as pressure level data, column maximum data and so on. Rather than discarding that data, it is now recycled, producing high temporal frequency information about the storm location, intensity and structure, allowing new types of experimental forecast and research products.

The net result is that the 2013 and 2014 HWRF models meet the expectations of the 3km HWRF of providing a true high resolution operational forecasting model, while also providing new types of products for analyzing the model and provide enhanced guidance to the forecasters.