Transcript WG2_report_Trieste

VALUE WG2
Benchmark data set &
pseudo-reality (year 1-2)
Report, Trieste Meeting Sep.12
Sven Kotlarski, José Gutiérrez
13 Members
Chair / co-Chair: Kotlarski (CH)
Gutiérrez (E)
WG Members:
Bärring (S)
Bosshard (CH)
Brands (E)
Brzóska (PL)
J.H. Christensen (DK)
Coppola (I)
Jaczewski (PL)
Jones (UK)
Maraun (D)
Pagé (FR)
Pianko (PL)
Slide 2 of 13
Overview (MoU)
“Observational benchmark data sets and pseudo realities will be set up for the validation by WG2.
These will comprise climatically distinct regions, e.g., maritime, continental, alpine and
Mediterranean climates. The daily station data set will build upon the ECA&D data set publicly
available from the Dutch weather service KNMI. Sub-daily station data will be provided by the Swedish
Meteorological and Hydrological Institute SMHI for Sweden, MeteoSwiss for the Alps and the Regional
Agency for Prevention and Environment ARPA of the Emilia Romagna for a Mediterranean climate.
Station data may require the user to individually sign an agreement with the data provider to use the
data for scientific purposes only. Additionally, the benchmark data set will contain daily gridded
observations for the validation of SDS and in particular RCMs. The latter simulate area averages and
are not directly comparable with station observations. As gridded observations, the EOBS data set will
be employed, which is the gridded version of the ECA&D station data set developed within the
ENSEMBLES project. Several researchers have noted potential weaknesses of the EOBS data set in
regions of sparse data (e.g., Herrera et al, 2010), such that the validation will be restricted to selected
regions, where a high data quality is ensured. Currently, a gridded precipitation data set with subdaily resolution for the Alps, based on a combination of station and radar data, is under development
by MeteoSwiss and might be used in a later stage of VALUE. Further data sets which might be
available in the future or by further partners joining VALUE at a later stage will be considered
additionally.
The pseudo reality …”
Slide 3 of 13
Tasks, Milestones, Deliverables
T5: compile benchmark data set; upload benchmark data set to website
1st half of year 2
T6: compile pseudo reality; upload pseudo reality to website
1st half of year 2
D3: observational benchmark data set and a pseudo reality for the
validation (upload to website)
1st half of year 2
M3: observational benchmark data set and pseudo reality compiled
1st half of year 2
Slide 4 of 13
Observational Benchmark Dataset
2
(1)
Regions
1
4
3
Validation SD
Validation SD and DD
Station-based obs. data
(e.g. ECA&D)
Gridded obs. data
(e.g. EOBS)
Parameters and required temporal
resolution depend on target
(i.e. on WG)
Parameters and required temporal
resolution depend on target
(i.e. on WG)
Slide 5 of 13
Observational Benchmark Dataset
(2)
Open issues (to be discussed this meeting)
• Which pilot regions?
Example: maritime -> UK, continental -> PL, alpine -> CH, Mediterrenean -> E / IP
• Is quality of observational data appropriate?
• Is temporal resolution appropriate?
• Is spatial coverage appropriate?
Depends on
target / WG
Concerns especially network density for gridded observations
• Are all required parameters available?
• Is length of observed period appropriate?
30 years at least?
Slide 6 of 13
Dataset inventory
• Compiled by WG2 with input from further WGs
• But: Not much feedback, inventory probably not complete yet
(especially wrt. regional/national station data sets)
• Please add data from your country!
• Data availability and terms of use still need to be checked
Slide 7 of 13
E-OBS (1)
• Standard reference data set for RCM evaluation
• Entire European land surface, 1950 - present, daily resolution,
•
available on standard RCM grids
Would allow for a “fair” comparison of SD and DD performance
BUT
• Poor data quality in regions of low network density
(over-smoothing of the spatial field)
• Smoothing particularly affects tails of dictribution
(reduction of daily extremes)
• Temporal inhomogeneities due to changes in contributing station
network
Hofstra et al. 2009, Hofstra et al. 2010, Kysely and Plavcova 2010,
Maraun et al. 2012, Herrera et al. 2012
Slide 8 of 13
E-OBS (2)
Slide 9 of 13
E-OBS (3)
?
?
?
?
Restrict validation exercise to regions of high network density or
apply different regional / national gridded datasets
Slide 10 of 13
Suggestions for pilot regions
• Covers both atlantic and mediterranean climates
• Publicly available high-quality datasets at daily resolution
Spain / IB
Spain02: daily P, Tmax, Tmin at 0.2° for 1950-2008
• Several downscaling studies already exist, including validation of
E-OBS and ENSEMBLES RCMs
• Besides daily data: hourly data available for some stations
Switzerland
• Alpine climate
• High-quality gridded (2 km) and station-based data available
• Gridded hourly precipitation up from 2003, sub-daily station data
available
• Large number of previous downscaling studies (SD and DD)
Poland
• Continental climate
• Gridded 25km daily data, quality probably superior to E-OBS
France (?)
• Maritime to alpine climate
• Safran analysis available (8 km res., 1958-2010, hourly to daily)
Slide 11 of 13
Further Issues
• Selection of predictors (ERA40? Identical for each SD method? Include
predictors with high uncertainty, e.g. humidity?)
• Data distribution / data server
Santander group set up a THREDDS data
server that could host observational reference
data and predictor datasets
http://www.meteo.unican.es/thredds/catalog/VALUE/
• Involvement of WGs 3 to 5
Selection of reference data and selection of predictors: This meeting?
Further discussions: One / two representatives of each WG?
• Scenarios
− Downscale CMIP5
projections?
− Data server could host
GCM predictors
− GCM selection -> Paper
Brands et al.
Median of the absolute SLP bias/std values along
the lateral boundaries of the Euro-CORDEX domains
(1 Re-analysis and 7 CMIP5 GCMs)
Slide 12 of 13
Thanks …
Suggestion for pilot region Spain / IP
• Publicly available high-quality datasets at daily resolution
(> 100 stations, Spain02 / IB02 gridded datasets)
• Spain02: daily P, Tmax, Tmin at 0.2° for 1950-2008
• Covers both atlantic and mediterranean climates
• Several downscaling studies already exist, including validation of
E-OBS and ENSEMBLES RCMs
• Besides daily data: hourly data available for some stations
Herrera et al. 2012
Slide 14 of 13
E-OBS: Temporal coverage
Slide 15 of 13
E-OBS: Distribution of T stations (1)
Slide 16 of 13
E-OBS: Distribution of T stations (2)
Slide 17 of 13