Hurricane Prediction

The first real-time forecasts using the Model for Prediction Across Scales (MPAS) were conducted in 2013 and focused on tropical cyclones across the Northern Hemisphere. A distinguishing character of MPAS is its variable horizontal resolution and smoothly adapting mesh structure that allows regional resolution refinement. The physical parameterizations and numerical algorithms in MPAS are somewhat similar to those in the Advanced Research Weather Research and Forecast (WRF) model. MPAS is used for (1) predictions involving many scales of motion with clear societal relevance such as tropical cyclones and their global connections; and (2) exploration of practical predictability beyond roughly three to four days, i.e., the effective limit of utility for limited-area models.  MPAS has the potential to expand the ability of the research community to conduct fine-scale modeling of moist convection on the globe, and atmospheric motions on scales from regional (e.g., severe thunderstorms and hurricanes) to global (e.g., the Madden-Julian Oscillation).  In addition, MPAS offers the capability of providing greater lead time in tropical cyclone forecasts than can currently be achieved using regional models.

In FY 2015, forecasts were conducted with MPAS using both a uniform resolution and variable resolution for tropical cyclones during the period August 1 to November 3, 2014. The variable resolution was chosen to be finest over the Eastern Pacific basin because of the much greater amount of tropical cyclone activity – including both weak and intense storms – in that basin as compared to the Atlantic.   MMM staff developed software to identify and track tropical cyclones in MPAS and a new verification method was developed to account for false alarms and missed events.  This new method was needed because standard hurricane evaluation methods assume that the storm is present at the start of the forecast.

It was found that the variable and uniform-resolution MPAS produce similar skill in predicting tropical cyclones over the eastern Pacific in 2014.  This was likely due to the relatively slow growth of differences between variable and uniform-resolution MPAS over the tropical eastern Pacific, shown in the figure below.  The tropical cyclone environment in both configurations was similar for 6-7 days, and this allowed similar development and movement of the TCs in both configurations. 

Graph
Figure: Root mean squared differences between variable and uniform resolution MPAS in the tropics (red) and extratropics (blue). Differences are normalized by the average spatial standard deviation in their respective regions. The plot shows that, where the resolution of the two configurations is similar, differences grow more slowly in the tropics between days 2 and 7 than differences in mid-latitudes.