Downscaling: An Introduction (Regionalisation)

1
Downscaling An Introduction(Regionalisation)
Why do we need to downscale?
2
Because there is a mismatch of scales between
what climate models can supply and what
environmental impact models require.
Impact models require ...
Global Climate Models supply...
3
Downscaling Using GCMs

• GCM output is generally the starting point of
  any regionalisation technique, so
• GCMs should perform well in simulating
  circulation and climatic features affecting
  regional climates, e.g., jet streams, storm
  tracks
• it is better to use variables where sub-grid
  scale variations are weak, e.g., mean sea level
  pressure
• Main advantage of using GCMs is that
• internal physical consistency is maintained

4
A variety of methods and techniques have been
developed to address this scale problem

1. High resolution and variable resolution AGCM
   time-slice experiments - numerical modelling
2. Regional Climate Models (RCMs) - dynamic
   downscaling
3. Empirical/statistical and statistical/dynamical
   models - statistical downscaling

5
But the very simplest approach is the
interpolation of grid box outputs

• Overcomes problems of discontinuities in change
  between adjacent sites in different grid boxes
• But
• introduces a false geographical precision to the
  estimates

6
Interpolation
CGCM1 GHG only, Winter, Maximum temperature
change (C), 2020s
7

• Main downscaling approaches
• higher resolution experiments
• or
• empirical/statistical or statistical/dynamical
  downscaling processes

8
High Resolution Models

• Numerical models at high resolution over region
  of interest
• GCM time-slice experiments
• variable resolution GCMs
• high resolution limited area models (regional
  climate models - RCMs)

9
REGIONAL CLIMATE MODELS

1. Driven by initial conditions, time-dependent
   lateral meteorological conditions and surface
   boundary conditions which are derived from GCMs
   (or analyses of observations)
2. Account for sub-grid scale forcings (e.g. complex
   topographical features and land cover
   inhomogeneity) in a physically-based way
3. Enhance the simulation of atmospheric
   circulations and climatic variables at finer
   spatial scales

10
Comparison of detail in precipitation patterns
over western Canada as simulated by CGCM1 and
CRCM.
Source G. Flato, in Climate Change Digest
Projections for Canadas Climate Future, H.G.
Hengeveld.
11
The Canadian RCM - CRCM
12
Screen Temperature (ºC) 5-year mean Winter
Validation work in progressRuns are underway
13
Precipitation rate (mm/day)5-year mean Winter
Validation work in progressRuns are underway
14
High Resolution Models

• ADVANTAGES
• are able to account for important local forcing
  factors, e.g., surface type elevation

• DISADVANTAGES
• dependent on a GCM to drive models
• computationally demanding
• few experiments
• may be locked into a single scenario, therefore
  difficult to explore scenario uncertainty, risk
  analyses

15
Effect of scenario resolution on impact outcome
Spatial Scale of Scenarios
Source IPCC, WGI, Chapter 13
16
Empirical/Statistical, Statistical/Dynamical
Methods

• PREDICTAND PREDICTORS
• Sub-grid scale climate ? f(larger-scale
  climate)
• Transfer functions - calculated between
  large-area and/or large-scale upper air data and
  local surface climates
• Weather typing - relationships calculated
  between atmospheric circulation types and local
  weather
• Weather generator parameters can be conditioned
  upon the large-scale state

17
Main Assumptions

• Predictors are variables of relevance to the
  local climate variable being derived (the
  predictand) and are realistically modelled by the
  GCM
• The transfer function is valid under altered
  climatic conditions
• The predictors fully represent the climate change
  signal

18
Transfer Functions
Grid Box
Extract predictor variables from GCM output
Predictor variables e.g., MSLP, 500, 700 hPa
geopotential heights, zonal/meridional components
of flow, areal TP
Select predictor variables
Transfer function e.g., Multiple linear
regression, principal components analysis,
canonical correlation analysis, artificial neural
networks
Calibrate and verify model
Drive model
Site variables for future, e.g., 2050
Observed station data for predictand
19
Transfer Functions

• Fundamental Assumption
• the observed statistical relationships will
  continue to be valid under future radiative
  forcing

• ADVANTAGES
• much less computationally demanding than physical
  downscaling using numerical models
• ensembles of high resolution climate scenarios
  may be produced relatively easily

20
Transfer Functions

• DISADVANTAGES
• large amounts of observational data may be
  required to establish statistical relationships
  for the current climate
• specialist knowledge required to apply the
  techniques correctly
• relationships only valid within the range of the
  data used for calibration - projections for some
  variables may lie outside this range
• may not be possible to derive significant
  relationships for some variables
• a predictor which may not appear as the most
  significant when developing the transfer
  functions under present climate may be critical
  for determining climate change

21
Weather Typing

• Statistically relate observed station or
  area-average meteorological data to a weather
  classification scheme.
• Weather classes may be defined objectively (e.g.
  by PCA, neural networks) or subjectively derived
  (e.g., Lamb weather types UK, European
  Grosswetterlagen)

Pressure fields from GCM
Select classification scheme
Calculate weather types
Identify weather types
Relationships between weather type and local
weather variables
Drive model
Derive
Local weather variables for, say, 2050
Observed weather variables
22
Weather Typing
Fundamental Assumption the relationships
between weather type and local climate variables
will continue to be valid under future radiative
forcing

• ADVANTAGES
• founded on sensible physical linkages between
  climate on the large scale and weather on the
  local scale

23
Weather Typing

• DISADVANTAGES
• the fundamental assumption may not hold -
  differences in relationships between weather type
  and local climate have occurred at some sites
  during the observed record
• scenarios produced are relatively insensitive to
  future climate forcing - using GCM pressure
  fields alone to derive types, and thence local
  climate, does not account for the GCM projected
  changes in, e.g., temperature and precipitation,
  so necessary to include additional variables such
  as large-scale temperature and atmospheric
  humidity

24
Downscaled vs. original GCM
Ex. Animas River Basin (US) with Hydrologic
Model Delta Change HadCM2 results (raw
data)Grey area 20 ensembles with downscaled
climate scenarioSimulated with observed data
Source Hay et al. (1999)
25
Weather Generators
Precipitation Process Occurrence Amount
LARS-WG wet and dry spell length
Non-precipitation variables Maximum
temperature Minimum temperature Solar radiation
Model calibration
Synthetic data generation
Climate scenarios
26
Weather Generators
Spatial Downscaling
Spatial Downscaling
Calibrate weather generator using area-average
weather
Area parameter set
Apply changes in parameters derived from
difference between area and grid box parameter
sets to individual station parameter files
generate synthetic data for scenario
Calibrate weather generator for each individual
station within area
Station parameter set
Calculate changes in parameters from grid box
data
27
Weather Generators
Temporal Downscaling
Observed station data
WG
Parameter file containing statistical
characteristics of observed station data
Monthly scenario information
Generate daily weather data corresponding to
scenario
28
Weather Generators
Fundamental Assumption The statistical
correlations between climatic variables derived
from observed data are assumed to be valid under
a changed climate.

• ADVANTAGES
• the ability to generate time series of unlimited
  length
• opportunity to obtain representative weather time
  series in regions of data sparsity, by
  interpolating observed data
• ability to alter the WGs parameters in
  accordance with scenarios of future climate
  change - changes in variability as well mean
  changes

29
Weather Generators

• DISADVANTAGES
• seldom able to describe all aspects of climate
  accurately, especially persistent events, rare
  events and decadal- or century-scale variations
• designed for use, independently, at individual
  locations and few account for the spatial
  correlation of climate

30
Further Reading

• IPCC TAR(2001) - Chapter 10 13 (www.ipcc.ch)
• Wilby Wigley (1997) Downscaling general
  circulation model output a comparison of
  methods. Progress in Physical Geography 21,
  530-548
• Hewitson Crane (1996) Climate downscaling
  techniques and application. Climate Research 7,
  85-95
• Goodess et al. (2003) The identification
  evaulation of suitable scenario development
  methods for the estimation of future
  probabilities of extreme events,Tyndall Centre,
  Rep. 4. report