Upload
jessica-williamson
View
214
Download
1
Embed Size (px)
Citation preview
AImetgroup
http://www.meteo.unican.es/ 1
José Manuel GutiérrezDaniel San MartínDpto. Matemática Aplicada y Ciencias de la Computación
Applied Meteorology Group
Univ. Cantabria – INM (Santander, Spain)
Bartolomé OrfilaAntonio S. CofiñoInstituto Nacional de Meteorología
AImetgroup
http://www.meteo.unican.es/ 1
http://www.meteo.unican.es
http://www.meteo.unican.es/ensembles
Statistical Downscaling Portal for ENSEMBLES
(RT2B)
AImetgroup
http://www.meteo.unican.es/ 2
Use of the Portal by Example: Crop Yield in Italy
Example: To input a crop-yield model, the Joint Research Center (JRC) needs to obtain seasonal forecasts of several surface variables: maximum and minimum temperature, mean daily rainfall, daily global radiation, evapotranspiration.....in a high-resolution 50kmx50km grid over Italy.
GOAL: Daily values for june-august 2006 in a suitable format (e.g., text file, or Excel file).
This downscaling process can be performed in 4 steps (Perfect Prog):1. Select a set of predictors from a reanalysis database (ERA40,…) 2. Select simultaneous observations in the desired grid (e.g., JRC gridded observations).
3. Fit a downscaling model (e.g., a weather clustering scheme, or analogs)
4. Apply it to the output of some seasonal GCM (e.g., DEMETER, STREAM1, EURO-SIP).
AImetgroup
http://www.meteo.unican.es/ 3
PrecipitationTemperature
(T(1ooo mb),..., T(500 mb); Z(1ooo mb),..., Z(500 mb);
H(1ooo mb),..., H(500 mb))
Xn
Regres., CCA, …
Yn = WT XnYn
Features and Structure of the Downscaling Portal
ERA40 and NCEP reanalysis (fields over Europe).
Observations in a 0.5x0.5 grid over Europe provided by JRCObservations in 1000 local points provided by ECA.
S2d outputs from DEMETER models (seven models from 1958-2001) and ENSEMBLES STREAM1 (three models from 1991-2001)
Predictors PredictandsDownscaling Model
AImetgroup
http://www.meteo.unican.es/ 4
S2D Algorithms Implemented in the Downscaling Portal
1. Weather-Clustering Analog Method based on Self-Organizing Maps
A generalization of the analog method which uses a pre-classification of the reanalysis patterns to obtain “weather classes” from where probabilistic forecasts according to the observed climatolgies within groups are obtained. The clustering method used is a self-organizing map (SOM) which provides a lattice of “weather classes” which is the support of the resulting PDF.
Clustering methods for statistical downscaling in short-range weather forecast J.M. Gutiérrez , R. Cano, A.S. Cofiño, and M.A. Rodríguez Monthly Weather Review, 132(9), 2169 - 2183 (2004).
Analysis and downscaling multi-model seasonal forecasts in Perú using self-organizing maps A.S. Cofiño, J.M. Gutiérrez, and R. Cano Tellus A, 57, 435-447 (2005). 2. Weather Generator
Stochastic generation of daily precipitation conditioned on predictions of the probability of a wet day in the season and daily persistence. The method uses SVD of model output and observations to obtain ensemble-mean seasonal means to feed the stochastic model.
A method for statistical downscaling of seasonal ensemble predictions H. Feddersen and U. Andersen Tellus A, 57, 398 - 405 (2005).
Soon available
Other methods from other partners to be include.
AImetgroup
http://www.meteo.unican.es/ 5
Current Registered Users are from ENSEMBLES partners.
AImetgroup
http://www.meteo.unican.es/ 6
AImetgroup
http://www.meteo.unican.es/ 7
AImetgroup
http://www.meteo.unican.es/ 8
AImetgroup
http://www.meteo.unican.es/ 9
AImetgroup
http://www.meteo.unican.es/ 10
http://www.meteo.unican.es/ensembles