In a classic 1969 paper, Lorenz proved that the predictability in a simple (frictionless and unheated) model of the earth's atmosphere was, in principle, limited to about five days. Specifically, this meant that the forecast error would exceed the variability (say of wind, or pressure) on the actual map at the verification time of the forecast, after some forecast period. A better model and/or a more complete and accurate set of observations at initial time might reach this time of "useless" forecasts later, but even the best combination of these would have to reach it in about five days. The predictability, by this measure, in an actual operational forecast model now used in the ECMWF has gone from five days in 1980 to eight days in 1990. This operational model does include the effects of heating, and frictional and topographic interference with the flow caused by the earth's surface. With these complications it is not amenable to the type of theoretical treatment used by Lorenz. However, it is possible that precisely these complicating processes somehow enhance the predictability of the operational model. Thus the time limit on useful forecasts has yet to be determined, and ways to enhance the predictability deserve extended exploration. The objective of the research supported under this award is to estimate the attainable enhancement of predictability by: i) Improving the model's treatment of the larger-scale flow components, which have the most obvious forecast errors, and which are most directly linked to pole to equator and continent- ocean contrasts in heating and frictional/topographic interference; ii) Fitting a research version of the model with more detailed formulas for the calculation of momentum, heat, and water exchanges with the underlaying surface, and iii) Studying the sensitivity of the global-scale forecast error growth rate to simulated observational errors in selected regions which are presumed to require fine mesh specification of the initial state. The PI's prior research in Numerical Weather Prediction,and tropical, boundary layer, and mountain meteorology is especially relevant to the proposed work. These studies will increase our understanding of the factors affecting predictability, and are likely to contribute to further improvements in Numerical Weather Prediction over the next decade.
