SIGN-IN

Publication: Extension of Geostatisical Output Perturbation (GOP) Method for Probabilistic Weather Forecasting of Surface Temperature

All || By Area || By Year

Title Extension of Geostatisical Output Perturbation (GOP) Method for Probabilistic Weather Forecasting of Surface Temperature
Authors/Editors* Erika Kramer and Yulia Gel
Where published* The 1st TIES North American Regional Meeting, Seattle, WA, USA. (TIES stands for The International Environmetrics Society.)
How published* Proceedings
Year* 2007
Volume
Number
Pages
Publisher
Keywords spatio-temporal modelling, probabilistic weather forecasting, CART
Link
Abstract
Our project focuses on further development of Geostatistical Output Perturbation (GOP) method for probabilistic mesoscale weather forecasting of surface temperature. In particular, in order to more accurately capture spatio-temporal non-stationarity of surface temperature, we apply GOP to local subdomains in the US Pacific North-West. The subdomains are selected by Classification and Regression Trees (CART) for different seasons. The subdomain selection remains consistent through different periods of time. We develop a hierarchical GOP approach by modeling sills of variograms as a spatio-temporal random process rather than a deterministic quantity and then generating GOP ensembles with a randomly selected sill. The resulting prediction intervals appear to be calibrated and shorter than from the usual GOP.
Go to Simulation and Modeling
Back to page 65 of list