I’ve always thought it’s fun to call up NOAA’s seven-day weather forecast on the computer and watch the green blobs travel over the map and gain hotter colored splotches as the weather shown by the blobs gets worse. But I’ve never given much consideration to the probabilities that come with the forecasts. Today I stumbled upon the information that a prediction like "70% chance of rain" under the picture of the day’s weather is the product of something called ensemble forecasting. This is a form of Monte Carlo simulation that is specially tooled to account for elements of chaos (which makes me wonder if it is used much in the financial sector).
Ensemble forecasting is the pooling of multiple–meaning many,say 50–simulation models run with Monte Carlo software, each of which starts from a slightly different set of conditions. The slightly different initial conditions are intended accommodate the element of chaos, and the pooling of probabilities from each member simulation allows for greater reliability.
Weather is a set of dynamic conditions, and forecasting a future set of these conditions goes far beyond the standard operations research puzzle. It is a legendary challenge in environmental risk analysis that only becomes more challenging the farther into the future it extends. Even with ensemble forecasting this is still true, because a small error in the initial conditions will grow with the lead time. This is the reason that, to the despair of the UK’s Natural Environmental Research Council, "we can forecast major weather patterns reasonably well up to about three days ahead. Beyond that, the uncertainties in the forecasts become so large that the forecast is no longer meaningful."
For now, I’ll look up the green blobs for only the next three days. After that, it all just depends on the weather.