Transcript Issues with the Application of Empirical Mode

Global catchment based comparison
of observed and HadGEM modelled
precipitation, temperature and Köppen
climate type
Murray Peel1, Thomas McMahon1 & Ian
Smith2
1 Civil
& Environmental Engineering, The University of Melbourne,
Victoria, Australia
2 CSIRO Marine and Atmospheric Research, Aspendale, Victoria,
Australia
Outline

Background
– Observed catchment data
– HadGEM data

Comparison
– Mean annual


Precipitation
Temperature
– Köppen climate type

Conclusions
EGU 2008 - Session CL21, Peel et al
Background

Long-term water resources management
will depend upon projections of future
climate change.

How well do CGCMs reproduce past &
present hydroclimate at the catchment
scale?
EGU 2008 - Session CL21, Peel et al
Background
Observed catchment data

Globally 686 catchments.
EGU 2008 - Session CL21, Peel et al
Background
Observed catchment data

Catchment boundary, within 5% of known area
using 1km DEM (HYDRO1k),
 Monthly streamflow data,
– Unregulated for the period of record,

Catchment average monthly Temperature (T) &
Precipitation (P),
– For the period of runoff record,
– Theissen polygon area weighting using GHCN v2 P &
T station data.

Missing P & T infilled from nearby stations with best monthly
correlation
EGU 2008 - Session CL21, Peel et al
Background
Observed catchment data

Elevation correction to Catchment P
– Since station elevation is usually low relative to the
catchment area, catchment average P underestimated,
– Corrected using Budyko like relationship of Fu (Zhang
et al., 2004, WRR 40, W02502).

The 686 catchments are a subset of 900
catchments
– modelled adequately with a monthly rainfall-runoff
model
– Some confidence in the P & T data.
EGU 2008 - Session CL21, Peel et al
Background
HadGEM data

HadGEM a coupled global climate model from the
Hadley Centre, UK.
– IPCC AR4 run
– 20C3M scenario

Monthly P & T for 1/1860 – 12/1999
– Extract catchment average P & T from raw HadGEM
(no downscaling) for the concurrent period of observed
streamflow

Ideally an ensemble average of GCM runs would be better than
the single run used here
EGU 2008 - Session CL21, Peel et al
Background
HadGEM data

Area weighted grid cell to form catchment
average values
EGU 2008 - Session CL21, Peel et al
Comparison

Compare observed and HadGEM catchment
average
– Mean Annual P & T
– Köppen climate type
EGU 2008 - Session CL21, Peel et al
Global catchment mean annual
temperature
GCM has ~1oC cool bias, but overall very
good.
30
GCM mean annual temperature (deg C)

20
10
0
y = 1.001x - 1.191
R2 = 0.914
-10
-20
-20
-10
0
10
Observed mean annual temperature (deg C)
EGU 2008 - Session CL21, Peel et al
20
30
Hemisphere catchment mean
annual temperature
NH better than SH
30
GCM mean annual temperature (deg C)

NH
SH
y = 1.0219x - 1.7672
R2 = 0.9579
20
10
0
y = 0.6319x + 5.205
R2 = 0.4564
-10
-20
-20
-10
0
10
Observed mean annual temperature (deg C)
EGU 2008 - Session CL21, Peel et al
20
30
Global catchment mean annual
precipitation
GCM not capturing the observed range.
10000
GCM mean annual precipitation (mm)

1000
y = 12.7x0.612
R2 = 0.325
100
100
1000
Observed mean annual precipitation (with Fu adjustment) (mm)
EGU 2008 - Session CL21, Peel et al
10000
Hemisphere catchment mean
annual precipitation
Again NH better than SH
10000
GCM mean annual precipitation (mm)

NH
SH
y = 4.2953x0.7738
R2 = 0.5234
1000
y = 76.07x0.346
R2 = 0.1064
100
100
1000
Observed mean annual precipitation (with Fu adjustment) (mm)
EGU 2008 - Session CL21, Peel et al
10000
Köppen climate type

Köppen climate classification rules as used in Peel
et al (2007, HESS, 11: 1633-1644.)
EGU 2008 - Session CL21, Peel et al
Köppen climate type - Tropical

SubClass
Obs
Freq.
Af
22
Am
16
Aw
72
Obs
All
GCM
Freq.
GCM
Freq.
11
110
14
83
58
Cold bias of GCM reduces the number of Tropical
catchments (coldest month ≥ 18oC)
EGU 2008 - Session CL21, Peel et al
Köppen climate type - Arid

SubClass
Obs
Freq.
Obs
Freq.
BWh
0
2
BWk
0
0
BSh
13
21
BSk
16
26
19
GCM
Freq.
GCM
Freq.
49
Errors in NH & SH GCM MAP leads to more Arid
catchments
EGU 2008 - Session CL21, Peel et al
Köppen climate type - Temperate
SubClass
Obs
Freq.
Csa
17
12
Csb
23
29
Csc
1
0
Cwa
21
40
Cwb
36
11
Cfa
83
64
Cfb
153
156
Cfc
6
1
EGU 2008 - Session CL21, Peel et al
Obs
Freq.
340
GCM
Freq.
GCM
Freq.
313
Köppen climate type - Cold
SubClass
Obs
Freq.
Dsa
1
1
Dsb
1
10
Dsc
4
3
Dwa
3
5
Dwb
2
4
Dwc
4
Dwd
0
3
Dfa
3
22
Dfb
105
73
Dfc
85
98
Dfd
3
2
EGU 2008 - Session CL21, Peel et al
Obs
Freq.
211
GCM
Freq.
9
GCM
Freq.
230
Köppen climate type - Polar
Sub- Obs
Class Freq.
ET
6
EGU 2008 - Session CL21, Peel et al
Obs
All
GCM GCM
Freq. Freq.
11
Conclusions

Comparison of observed and raw HadGEM
catchment average P & T.
– Globally MAT good

NH better than SH
– Globally MAP not very good

Again, NH better than SH
– Köppen climate type



Less Tropical & more Arid
Temperate & Cold (mainly NH) good
Useful assessment of GCM performance for later
hydrologic analysis
EGU 2008 - Session CL21, Peel et al
Acknowledgements

The analysis presented forms part of a paper
in currently in preparation.
 Funded by
– Australian Research Council Discovery Grant
EGU 2008 - Session CL21, Peel et al