WP4: Finalisation of D12 and related work

1
WP4 Finalisation of D12 and related work

• What has the completed work told us?
• What additional analyses could/should be
  undertaken of D12 results?
• What papers can be produced based on D12 work?

2
Principles of verification

• Predictor dataset NCEP reanalysis
• Predictand datasets FIC dataset and regional
  sets
• Study regions
• Stations within regions
• Core indices
• Verification period 1979-1994 (for compatibility
  with ERA15-driven regional models)
• Training period 1958-1978 1995-2000
• Statistics RMSE, SPEARMAN-RANK-CORR for each
  station/index

3
Study Regions
The FIC dataset
Germany 10 stations
Alps 10 stations
UK 6 stations
Greece 8 stations
Iberia 16 stations
Italy 7 stations
4
Partners/regions
5
Core indices
6
Questions

• Is the performance consistent among stations for
  a given method, region, core index and season?
• Is there any difference in performance of the
  methods between winter and summer?
• Is there any systematic difference in performance
  between regions?
• Is there any systematic difference between
• Direct method The seasonal core indices are
  downscaled using seasonal circulation indices
• Indirect method Daily precipitation/temperature
  values are generated and the seasonal core
  indices are calculated from these.
• Can a best method be selected at this stage?

7
UK 90th percentile rainday amounts
8
Iberia 90th percentile rainday amounts
9
Greece 90th percentile rainday amounts
10
Greece Tmax 90th percentile
11
Answers to questions

• Is the performance consistent among stations for
  a given method, region, core index and season?
  Large scatter between stations
• Is there any difference in performance of the
  methods between winter and summer? Generally
  better performance in winter in areas affected by
  Atlantic weather systems
• Is there any systematic difference in performance
  between regions? Generally better performance in
  the regions affected by the westerlies
• Is there any systematic difference between
• Direct method The seasonal core indices are
  downscaled using seasonal circulation indices
• Indirect method Daily precipitation/temperature
  values are generated and the seasonal core
  indices are calculated from these.
• No obvious systematic difference between direct
  and indirect methods
• Can a best method be selected at this stage?
  Requires further work

12
Reasons?

• Inhomogeneity of station data
• Too short verification and calibration period
• Missing predictors
• Stochasticity

13
Additional analyses

• Closer inspection to find bad stations
• More detailed regional analysis

14
Papers

• No common core paper from D12 - but may be later
  from D16
• Regional papers