| description abstract | The authors address a particular example of the general question of whether high-resolution forecasts provide additional deterministic skill beyond what can be achieved with a coarser-resolution model. To this end, real-time forecasts using the fifth-generation Pennsylvania State University?NCAR Mesoscale Model (MM5) with grid increments of 30 and 3.3 km are compared over a domain centered on the complex terrain region of southern New Mexico during the 1 June 2002 to 1 June 2003 period. The authors use time series of surface data to evaluate the relative ability of the two forecasts to capture significant temporal variations of wind. The authors hypothesize that the additional detail and structure provided by high resolution becomes a ?liability? when the forecasts are scored by traditional verification metrics, because such metrics sharply penalize forecasts with small temporal or spatial errors of predicted features. Thus, a set of verification metrics is designed that is increasingly tolerant of timing errors for temporal changes of wind. The authors find that the barrier-normal (i.e., zonal) wind component over complex terrain reveals the greatest improvement in skill due to increased horizontal resolution for the cases considered here. In addition, the fine-grid forecasts better replicate the cessation of drainage flow or onset of upslope flow near and within complex terrain. The most surprising result is the marginal benefit of the higher resolution over valley locations not in immediate proximity to the mountain slopes, even though the valley is only about 60 km across (east?west). Overall, the gains in forecast accuracy from finer grid spacing are generally incremental, but increase with greater tolerance for timing errors, culminating in the greatest improvement for forecasts of temporal variance. | |
