Medium-Range Prediction of an Extratropical Oceanic Cyclone: Impact of Initial State

Abstract

Significant progress has been made in the short-range (1?2 days) prediction of east coast cyclogenesis over the past decade. This is the result of improved model resolution, physical parameterization, and good analysis of the upstream conditions often well sampled by the high-density North America observing network. The prediction of cyclogenesis over the eastern Pacific Ocean, or a longer-range forecast of east coast cyclones, does not share the same degree of success, largely due to the fact that the upstream conditions fall over the data-void regions of the Pacific Ocean. In this paper, the authors study the prediction of the ERICA IOP-4 storm using a 120-km hemispheric version of the Penn State/NCAR Mesoscale Model MM5, with forecast duration ranging from 36 to 120 h. Specifically, the impact of uncertainties in the initial conditions on the 5-day forecastof this cyclogenesis event was examined. Initial perturbations were then introduced to the original analysis at initial time based on the 12-h forecast errors. Results from the numerical experiments led to the following conclusions. The skill of the model forecasts degraded as the forecast duration was lengthened. The model was able to capture the cyclogenesis up to 4.5 days in advance. Significant degradation occurred between day 4.5 and day 5, and the 5-day forecast failed to predict a major cyclone over the western Atlantic Ocean. Initial perturbations, determined by minimizing the errors of the initial 12-h forecast and introduced to the original analysis, were shown to improve the 5-day forecast substantially. Most remarkably, the 5-day forecast using the perturbed initial condition performed better than a forecast that was initialized 12 h later. Analysis of the derived initial perturbation showed that the main uncertainties in the initial condition were related to 1) the lower-tropospheric temperature analysis over the southern Rocky Mountain and Mexico regions and 2) the description of the upper-level potential vorticity (PV) anomaly over the Gulf of Alaska. The latter is partially related to the model?s systematic bias errors over that region. Sensitivity experiments carried out by adding the initial perturbation only to limited regions on selected model variables showed that the modification of the upper-level PV anomaly in itself was not sufficient to improve the 5-day forecast substantially. In contrast, the perturbation made to the low-tropospheric temperature field was critical for forecast improvement, by changing the patterns of thickness advection and structure of a shortwave trough involved in the cyclone development. The results presented in this paper suggest that it may be possible to improve the skill of medium-range forecasts of certain types of east coast cyclogenesis (such as the IOP-4 storm studied in this paper) if both (i) the quality of model initial conditions over key data-sparse regions (such as the eastern Pacific and Rocky Mountain regions) and (ii) the accuracy of short-term forecasting can be improved.