Circulation classification and statistical downscaling

1
Circulation classification and statistical
downscaling the experience of the STARDEX
projectClare Goodess the STARDEX
teamClimatic Research Unit, UEA, Norwich, UK
http//www.cru.uea.ac.uk/projects/stardex/
2
Robustness criteria for statistical downscaling

• Appropriate spatial scale (physics/GCM)
• Data widely/freely available (obs/GCM)

3
Choices to be made

• Surface and/or upper air
• Continuous vs discrete (CTs) predictors
• Circulation only or include atmospheric
  humidity/stability etc
• Spatial domain
• Lags temporal and spatial
• Number of predictors
• Few PC/sEOFs or clusters (e.g., 3-5) vs CT
  classifications (e.g., 12-20 classes)

4
Precipitation/Weather Regimes French Alpes
Maritimes Guy Plaut, CNRS-INLN
Greenland Anticyclone Sole Cyclone 1971-1983
(left) 1983-1995 (right)
5
Fuzzy rule optimisation technique 12 CPs defined
from SLP (Andras Bardossy)
CP02
CP09
6
Probability of precipitation at station 75103
conditioned to wet and dry CPs
Andras Bardossy, USTUTT-IWS
7
Heavy winter rainfall and links with North
Atlantic Oscillation/SLP
CC1 Heavy rainfall (R90N) CC1
mean sea level pressure
Malcolm Haylock, UEA/STARDEX
8
Emilia Romagna, N Italy
NCEP CDD (DJF), 1979-1993
ARPA-SMR
AUTH
9
HadAM3P predictor validation

• UEA and ARPA-SMR
• Principal Components of MSLP, Z500, T850
• Good correspondence in of significant
  components and explained variance (seasonal
  variation).
• Differences in patterns larger in summer.
  (Sampling uncertainty?)

10
HadAM3P predictor validation

• CNRS-INLN
• Daily CPs (Z_at_700), clusters, transition
  probabilities
• Inter-relationships Good correspondence for CPs
  conditional to heavy precipitation. Frequency
  errors (Sampling?).

35
30
35
HadAM3P
37
34
29
NCEP/OBS
11
HadAM3P predictor validation

• U-STUTT
• Lower-tropospheric (westerly) moisture flux
  overestimated in winter and underestimated in
  summer.

DJF
JJA
12

• Will performance be degraded when predictors are
  taken from GCMs?
• How do the statistically-downscaled changes in
  extremes compare with RCM changes?
• Are the observed predictor/ predictand
  relationships reproduced by RCMs - are they
  stationary?

Iberia (16 stations) Spearman correlations for
each of 6 models season averaged across 7
rainfall indices NCEP predictors
http//www.cru.uea.ac.uk/projects/stardex/