European Climate Assessment

1

European Climate Assessment Dataset Judging
homogeneity of daily series
Fourth seminar for homogenization Budapest, 6-10
October 2003
Janet Wijngaard, KNMI, the Netherlands
2
Topics

• ECAD project
• Approach to homogeneity
• Results, Conclusions

3
ECAD project

• data analysis focusing on observed changes in
  extremes
• gather daily series of observations at
  meteorological stations in
• Europe and the Middle East
• quality control and homogeneity analysis of the
  series
• dissemination of data and analyses results

4
Current participation

• Most data from 1900 up to 2001
• More than 200 stations
• Tmin, Tmax, Tmean,
• precipitation amount, pressure

5
Trend analysis of extremes requires A dense,
high-resolution, accurate and consistent dataset
6
Method
Homogenization of daily series
Instead labelling of series -gt
confidence for trend and variability analysis
7
Two-step approach

1. Four homogeneity tests applied to ECA dataset to
   identify potential inhomogeneities in annual
   resolution testing variables representative for
   the daily resolution

1. Grouping of results -gt overall classification

8
Homogeneity tests and variables

• Tests (absolute)
• SNHT
• Buishand Range
• Pettitt
• Von Neumann Ratio

•  Variables
• precipitation number of wet days
• temperature mDTR and vDTR (annual mean of
  absolute day-to-day
• 
  differences of DTR)

9
vDTR
DTRi Diurnal temperature Range for day i in a
specific year M number of days in the year
10
Classification

• Labels
• Useful (0/1 tests significant)
• Doubtful (2 tests significant)
• Suspect (3/4 tests significant)

 
11
Station Groningen (NL)

 
1948 change of observation hut 1951 relocation
1959 change in sensor height
12
Station Groningen (NL)
Buishand Range, Pettitt and Von Neumann
significant -gt suspect
13
Temperature 1946-1999
mDTR
vDTR
-gt54 suspect, breaks (partly) supported by
metadata
14
Precipitation 1946-1999
Number of wet days
Paper International Journal of
Climatology, May, 2003
-gt 10 suspect
15
Conclusions

•  
• most severe step-wise breaks are detected
• metadata support for detected breaks essential
• no homogenizing of daily series
• labelling system good basis for series selection
  in trend analysis

16
And

•  
• further investigations to test homogeneity on
  daily basis
• MASH method used for homogenization on monthly
  ECAD series

17
More info at http//www.knmi.nl/samenw/eca