Lecture-9_Statistical_Downscaling_2012

Download Report

Transcript Lecture-9_Statistical_Downscaling_2012

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Change Risk
Management
Lecture-9: Statistical Downscaling
Techniques
Akm Saiful Islam
Institute of Water and Flood Management (IWFM)
Bangladesh University of Engineering and Technology (BUET)
March, 2013
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Topics
Approach of downscaling
 Techniques of downscaling
 Strength and weakness
 Statistical downscaling using SDSM

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
General Approach to Downscaling
Applicable to:
•Sub-grid scales (small
islands, point
processes)
•Complex/
heterogeneous
environments
•Extreme events
•Exotic predictands
•Transient change/
ensembles
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Types of downscaling
Dynamical climate modelling
 Synoptic weather typing
 Stochastic weather generation
 Transfer-function approaches

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Dynamic downscaling

Dynamical downscaling involves the nesting
of a higher resolution Regional Climate Model
(RCM) within a coarser resolution GCM.

The RCM uses the GCM to define time–
varying atmospheric boundary conditions
around a finite domain, within which the
physical dynamics of the atmosphere are
modelled using horizontal grid spacings of
20–50 km.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Limitations of RCM

The main limitation of RCMs is that they
are as computationally demanding as
GCMs (placing constraints on the feasible
domain size, number of experiments and
duration of simulations).

The scenarios produced by RCMs are also
sensitive to the choice of boundary
conditions (such as soil moisture) used to
initiate experiments
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Advantages of RCM

The main advantage of RCMs is that they
can resolve smaller–scale atmospheric
features such as orographic precipitation or
low–level jets better than the host GCM.

Furthermore, RCMs can be used to explore
the relative significance of different external
forcings such as terrestrial–ecosystem or
atmospheric chemistry changes.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Regional Climate Model
Limited area regional models require
meteorological information at their
edges (lateral boundaries)
These data provide the interface
between the regional model’s
domain and the rest of the world
The climate of a region is always
strongly influenced by the global
situation
These data are necessarily provided by
global general circulation models
(GCMs) or from observed datasets
with global coverage
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Weather classification:
LWT scheme to condition daily rainfall
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Weather typing

Weather typing approaches involve grouping
local, meteorological data in relation to
prevailing patterns of atmospheric circulation.
Climate change scenarios are constructed,
either by re–sampling from the observed
data distributions (conditional on the
circulation patterns produced by a GCM), or
by generating synthetic sequences of
weather patterns and then re–sampling from
observed data.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Weather pattern downscaling is founded
on sensible linkages between climate on
the large scale and weather at the local
scale.
 The technique is also valid for a wide
variety of environmental variables as well
as multi–site applications. However,
weather typing schemes ca be parochial, a
poor basis for downscaling rare events,
and entirely dependent on stationary
circulation–to–surface climate
relationships.

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Limitation of Weather typing

Potentially, the most serious limitation is
that precipitation changes produced by
changes in the frequency of weather
patterns are seldom consistent with the
changes produced by the host GCM
(unless additional predictors such as
atmospheric humidity are employed)
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Stochastic weather generators
Stochastic downscaling approaches typically
involve modifying the parameters of
conventional weather generators such as
WGEN, LARS–WG or EARWIG.
 The WGEN model simulates precipitation
occurrence.
 Furthermore, stochastic weather generators
enable the efficient production of large
ensembles of scenarios for risk analysis.

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Weather generator
WGEN model simulates precipitation
occurrence using two–state, first order
Markov chains: precipitation amounts on
wet days using a gamma distribution;
 temperature and radiation components
using first–order trivariate autoregression
that is conditional on precipitation
occurrence.

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Advantages of weather
generator
Climate change scenarios are generated
stochastically using revised parameter
sets scaled in line with the outputs from a
host GCM.
 The main advantage of the technique is
that it can exactly reproduce many
observed climate statistics and has been
widely used, particularly for agricultural
impact assessment.

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Limitations weather generator

The key disadvantages relate to the low
skill at reproducing inter-annual to decadal
climate variability, and to the unanticipated
effects that changes to precipitation
occurrence may have on secondary
variables such as temperature.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Transfer functions

Transfer-function downscaling methods
rely on empirical relationships between
local scale predictands and regional scale
predictor(s). Individual downscaling
schemes differ according to the choice of
mathematical transfer function, predictor
variables or statistical fitting procedure.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Types of transfer functions

To date, linear and non–linear regression,
artificial neural networks, canonical
correlation and principal components
analyses have all been used to derive
predictor–predictand relationships.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Strength and weakness of
transfer function

The main strength of transfer function
downscaling is the relative ease of
application, coupled with their use of
observable trans–scale relationships.

The main weakness is that the models
often explain only a fraction of the
observed climate variability (especially in
precipitation series).
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
SDSM
Developed by Loughborogh university, UK
 www.sdsm.org.uk

Data can be downloaded from Canadian
Climate Change Scenario network (CCSN)
 http://www.cccsn.ca/?page=dst-sdi
 SDSM is best described as a hybrid of the
stochastic weather generator and transfer
function methods.

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
SDSM- Statistical Downscaling
Model
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
SDSM Algorithm

Optimisation Algorithm: SDSM 4.2
provides two means of optimising the
model – Dual Simplex (as in earlier
versions of SDSM) and Ordinary Least
Squares. Although both approaches give
comparable results, ordinary Least
Squares is much faster.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
The User can also select a Stepwise
Regression model by ticking the
appropriate box.
 Stepwise regression: works by
progressively adding all parameters into
the model and selecting the model which
models the predictand most strongly
according to one of two criteria: either
AIC(Akaike information criterion) or
BIC(Bayesian information criterion).

WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam



The Akaike information criterion is a measure of the
relative goodness of fit of a statistical model.
In statistics, the Bayesian information criterion (BIC)
or Schwarz criterion (also SBC, SBIC) is a criterion for
model selection among a finite set of models.
When fitting models, it is possible to increase the
likelihood by adding parameters, but doing so may result
in overfitting. The BIC resolves this problem by
introducing a penalty term for the number of parameters
in the model. The penalty term is larger in BIC than in
AIC.
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Patuakhali Tmin(1961-2001):
Ncepmslpas (mean sea
level pressure)
Partial r
0.763
ncepp500as (500 hpa
geopotential height)
0.308
ncepp850as (850 hpa
geopotential height)
0.61
ncepr850as (relative
humidity at 850 hpa)
0.383
Mean E%
34.2
Mean SE
1.461
Monthly variations in the percentage of explained
variance (Patuakhali)
Explained Variance (%)
Predictor variable
60
50
40
30
20
Tmin (1961-2001)
10
0
Month
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Validation of downscaled monthly Temperature in
patuakhali
30
Tmin (deg-C)
25
20
15
modeled data (1988-2000)
10
observed data (1988-2000)
5
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
0
Month
Validation of downscaled monthly Temperature in
patuakhali
35
25
20
15
modeled Data (1988-2000)
10
observed data (1988-2000)
5
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Tmax (deg-C)
30
Month
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Validation of downscaled mean dry spell length at
Patuakhali
25
20
15
modeled data (1988-2000)
10
observed data (1988-2000)
5
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Dry spell length (days)
30
Month
6
5
4
3
modeled data (1988-2000)
2
observed data (1988-2000)
1
0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Dry spell length (days)
Validation of downscaled mean wet spell length at
Patuakhali
month
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
Predicted minimum daily mean temperature in three
different time period (Patuakhali)
35
Tmin (deg. C)
30
25
20
15
10
5
0
predicted Tmin (20112040)
predicted Tmin (20412070)
predicted Tmin (20712099)
Tmax (deg. C)
Predicted maximum daily mean temperature in three
different time period (Patuakhali)
40
35
30
25
20
15
10
5
0
predicted Tmax (20112040)
predicted Tmax (20412070)
predicted Tmax (20712099)
WFM 6311: Climate Risk Management © Dr. Akm Saiful Islam
forcasted dry spell length at different time period
(Patuakhali)
Dry spell length (days)
14
12
10
predicted dry spell (20112040)
8
6
predicted dry spell (20412070)
4
2
January
February
March
April
May
June
July
August
September
October
November
December
0
predicted dry spell (20712099)
6
5
4
3
2
1
0
redicted wet spell (20112040)
December
November
October
September
July
Month
August
June
May
April
March
February
predicted wet spell (20412070)
January
wet-spell length (days)
Forcasted wet-spell length at different time period
(satkhira)
predicted wet spell (20712099)