Direct Verification of Tropical Cyclone Forecasts

The surface (i.e., 10 m) wind field drives the tropical cyclone (TC) warning/watch process of the National Hurricane Center (NHC), but the wind field in the meteorological forecast is represented by a few parameters most notably position (latitude, longitude) and intensity (maximum surface wind speed). Standard verification of TC forecasts, both official and model, involves statistics of position error (PE) and intensity error (IE) separately. However, the radius of gale force winds (R34) and radius of hurricane force winds (R64), and the uncertainty in the analysis and forecast of these radii, determines the setting of watch and warnings. Thus, it seems appropriate to verify the surface wind field directly rather than indirectly through PE/IE.

In this paper, a metric is proposed that directly represents the forecast error (FE) in the surface wind field as a difference in the observed and forecast Integrated Kinetic Energy (IKE) and is formulated as a function of (PE,IE). The main objective is to understand how the IKE-based FE varies with both PE and IE. Assuming that the model or human only has to forecast intensity and position (the radius of maximum wind and the wind radii come from the verifying best track in calculating IKE), we find that FE has, at most, a small (< 10%) dependence on IE.

One implication of the weak dependence of FE on IE is that efforts such as the Hurricane Forecast Improvement Project (HFIP) should emphasize improved track and wind structure forecasts vice intensity. As will be demonstrated, even in the case of Rapid Intensification (RI), track and wind radii still have a dominant role in FE. Special attention will also be given to the big hurricanes of 2017 (HARVEY, IRMA and MARIA) to illustrate how FE is lower in the global models than in the limited-area models despite less skillful intensity forecasts.