PROBABILISTIC WEATHER FORECASTING
Tilmann Gneiting
University of Washington
Probabilistic weather prediction consists of finding joint
probability distributions of future weather quantities or events. Information about the uncertainty of
weather forecasts can be important for decision-makers (e.g., public
transportation authorities, airlines, shipping, and the military), as well as
the public, but currently is routinely provided only for the probability of
precipitation, and not for other weather quantities such as temperature, wind
speed, or the amount of precipitation.
It is typically done by using a numerical weather prediction model,
perturbing the inputs to the model (initial conditions, lateral boundary conditions,
and physics parameters) in various ways, and running the model forwards for
each perturbed set of inputs. The
result is then viewed as an ensemble of forecasts, taken to be a sample from
the joint probability distribution of future weather events. However, forecast ensembles are
typically biased and underdispersive, and there is a need for statistical
postprocessing.
We first consider probabilistic forecasting of a single weather quantity,
such as the temperature at a given place in 48 hours time.
We propose a statistical method for postprocessing ensemble output
that is based on Bayesian Model Averaging (BMA), which is a standard method for
combining predictive distributions from different sources. We next consider
probabilistic forecasting of an entire weather field. We introduce a simple
method, the Geostatistical Output Perturbation (GOP) technique, which breaks
with much previous practice by perturbing the outputs, rather than the inputs,
from the numerical weather prediction model, using a geostatistical
approach. The two techniques can
be combined into the Spatial BMA method.
We apply these methods to obtain probabilistic forecasts of
surface temperature and quantitative precipitation over the Pacific Northwest, with
good verification results. The
forecasts are disseminated to the public in real time via the world wide web,
at probcast.washington.edu.
This is joint work with Adrian E Raftery, Veronica Berrocal, McLean Sloughter, Chris Fraley, Patrick Tewson (University of Washington), Yulia Gel (University of Waterloo), Fadoua Balabdaoui (Paris–Dauphine) and Michael Polakowski (Oregon State University).