Download Slide 1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Regression analysis wikipedia , lookup

Least squares wikipedia , lookup

Time series wikipedia , lookup

Data assimilation wikipedia , lookup

Linear regression wikipedia , lookup

Coefficient of determination wikipedia , lookup

Transcript
Large-scale atmospheric circulation
characteristics and their relations to
local daily precipitation extremes in
Hesse, central Germany
Anahita Amiri
Department of Geography
Justus Liebig University Giessen
Objective

Challenges



Coarse resolution of global climate models
(GCMs)
Doubt about the reliability of some GCM output
variables
scale mismatch between the reliable outputs of
GCM and climate change impact needs
2
Outline




Downscaling
Selecting predictors
Finding statistical relationship between
predictors and predictand
Validating the model
3
Downscaling

What is downscaling?


A method for obtaining high-resolution climate or climate
change information from coarse-resolution GCMs
Downscaling techniques




Regional climate models
Weather classification and re–sampling
Mixtures of stochastic processes, weather generators
Linear and non–linear regression
4
Statistical Downscaling Approaches
5
Practical Considerations





Predictors and Predictand
Selecting best set of Predictors for each
domain (size and location)
Transform Function (or model type)
Seasonal Variability
Calibration and Validation
6
Types of Predictors
1- Synoptic predictors (MSLP, 500 hPa
Geopotential heights)
2- Temperature predictors (T850, Tmax, Tmin)
3- Moisture predictors (specific and relative
humidity, precipitation)
4- Air flow predictors (u, v)
7
Predictor Selection Considerations:

An “ideal” Predictor should be:





Strongly correlated with the Predictand
Physically and/or conceptually sensible
Able to preserve covariance between local
variables
Accurately described by the GCM
Archived at the same temporal resolution as the
local variable(s)
8
Steps for selecting downscaling
predictor variables
1- Calculating correlation between predictors and
daily precipitation monthly /seasonal maxima
2- Calculating PCA1 and average for high correlated
areas
3- Fitting GEV and finding confidence interval for
location parameter
4- Finding correlation between predictors
5- Using AIC to find the best combination of
predictors
9
Correlation between specific humidity at
850hPa and maximum precipitation for winter
10
Correlation between relative vorticity at 850hPa
and maximum precipitation for winter
11
Correlation between temperature at surface
level and maximum precipitation for winter
12
Model Selection



Model Derivation for the statistical relations
between selected set of predictors and the
monthly/seasonal maxima of Percipitation
Calculate AIC for each model
Validate model by cross validation methods
13
Summary
Select
predictand
Station data
Select
predictors
Screen
variables
ERA40
data
Set model
structure
Calibrate
model
ERA40
predictors
Downscale
predictand
Synthesize
Observed data
GCM
predictors
Generate
scenario
Analyse
results
Impact
assessment
14
Asking For your Suggestions
Thank you for your suggestions
15