Mesoscale Predictability of the “Surprise” Snowstorm of 24–25 January 2000

A mesoscale model is used here to investigate the possible sources of forecast error for the 24–25 January 2000 snowstorm along the east coast of the United States. The primary focus is the quantitative precipitation forecast out to lead times of 36 h. The success of the present high-resolution control forecast shows that the storm could have been well forecasted with conventional data in real time. Various experiments suggest that insufficient model grid resolution and errors in the initial conditions both contributed significantly to problems in the forecast. Other experiments, motivated by the possibility that the forecast errors arose from the operational analysis poorly fitting one or two key soundings, test the effects of withholding single soundings from the control initial conditions. While no single sounding results in forecast changes that are more than a small fraction of the error in the operational forecast, these experiments do reveal that the detailed mesoscale distribution of precipitation in the 24- or 36-h forecast can be significantly altered even by such small changes in the initial conditions. The experiments also reveal that the forecast changes arise from the rapid growth of error at scales below 500 km in association with moist processes. The results presented emphasize the difficulty of forecasting precipitation relative to, say, surface pressure and suggest that the predictability of mesoscale precipitation features in cases of the type studied here may be limited to less than 2–3 days.