NUMERCIAL WEATHER PREDICTION MODEL ENSEMBLES

What are forecast ensembles?

Observation networks have limited spatial and temporal resolution, especially over large bodies of water such as the oceans (exposure error), which introduces uncertainty into the true initial state of the atmosphere (initialization). To account for this uncertainty, stochastic or “ensemble” forecasting is used, involving multiple forecasts created with different model systems, different physical parameterizations, or varying initial conditions. The ensemble forecast is usually evaluated in terms of the ensemble mean of a forecast variable, and the ensemble spread, which represents the degree of agreement between various forecasts in the ensemble system, known as ensemble members. The ensemble spread may consist up to 100 different computer runs, each with slightly different starting conditions or model assumptions, are combined into a weather forecast. Along with statistical techniques, ensembles can provide accurate statements about the uncertainty in daily and seasonal forecasting.

A look at the model output shows:

GFS 00z Ensemble Members

Shown above are 12 different ensemble members for the GFS. Each member varies  (in some circumstances significantly) from one to the next. Finding the extremes can help significantly when forecasting. This is how the Hurricane Center derives their “cone of uncertainty” for tropical storm paths.

When forecasting medium range (3-7 days) events, whether it be a hurricane or a Nor’eastor’s track, using the ensemble mean normally verifies better than the operational model run. When we “combine” all of the ensemble members, a product that looks like this is produced.

GFS 00z Ensemble Mean