RT2B: Making climate model projections usable for impact assessment Clare Goodess...

Preview:

Citation preview

RT2B: Making climate model RT2B: Making climate model projections usable for impact projections usable for impact

assessmentassessmentClare Goodess

c.goodess@uea.ac.uk

ENSEMBLES WP6.2 MeetingHelsinki, 26 April 2007

AImetgroup

http://www.meteo.unican.es/ 2

http://www.meteo.unican.es/ensembles

AImetgroup

http://www.meteo.unican.es/ 3

Collaboration with WP6.3 Users.

Fabio MicaleIacopo CerraniGiampiero Genovese

Downscale DEMETER and ENSEMBLES s2d hindcasts to get daily precip, radiation, wind speed, and maximum/minimum temperatures to make crop yield modeling. The goal is to compare the downscaled data to GCM outputs and to estimate seasonal predictability.

Two ongoing research collaborations with s2d users.

Downscale DEMETER and ENSEMBLES s2d hindcasts to get daily maximum and minimum temperatures to make electricity demand forecasts. The goal is to compare the downscaled data to GCM outputs.

Local precipitation forecasts for hydropower production capacities.

ELECTRICITÉ DE FRANCE

Laurent DubusMarta Nogaj

Outline of RT2B approaches: D2B.1 & D2B.2

Preparing datasets

RCM data server – D2B.3Reanalysis – D2B.13Observed – D2B.15GCM-based – D2B.17

Developing/testing models

Statistical: D2B.5, D2B.16Dynamical: D2B.9, D2B.10

Issues and methods

Ensemble averaging: D2B.6Pattern scaling: D2B.7, D2B.25Weighting: D2B.8GCM-RCM matrixRCM quick-look: D2B.21

Interactions with users (RT6)

Web-based downscaling service: D2B.4, D2B.19, D2B.23QuestionnairesDevelopment of tools: D2B.18Preliminary assessment: D2B.20

s2d statistical downscaling: D2B.12

Modification of SDS methods for probabilistic framework: D2B.14

From month 30, the emphasis is on synthesis, application and scenario construction

RT3:ERA@50/25: D3.1.4, D3.1.5 RCM weights: D3.2.2RCM system: D3.3.1

Final RCM system: D3.3.2 GCM/RCM skill/biases: D3.4.1RT3:

25 km scenario runs: D2B.22 mo 36

RCM quick look analysisD2B.24: mo 40

D2B.11: mo 31

Dynamical and statistical downscalingProbabilistic regional scenarios and tools

Applications to case studies• Alps, Mediterranean (D2B.28)…• Storms, CWTs, blocking….• Forestry, water….

Recommendations & guidance on methods for the construction of probabilistic regional climate scenarios: D2B.26 mo 42

s2d dynamical downscalingINM/RCA: MARS

Questions & issuesSources of uncertaintyReducing uncertaintyRobustness of SDS (D2B.27) Synergistic use of SDS/DDS

RT1: Grand PDFsRT2A: stream 1 runs (s2d, ACC)RT5: gridded data set

Mo 36 on:

RT3/RT2B RCM simulations• See table and news on RT3 website• Latest version of the matrix is in D3.3.1

Global modelRegional model

METO-HC

MPIMET IPSL CNRM NERSC Total number

METO-HC 1950-2100 1950-2100 2

MPIMET 1950-2100 1950-2050* 2

CNRM 1950-2050 2

DMI 1950-2100 1950-2050* 2

ETH 1950-2050 1

KNMI 1950-2050 1

ICTP 1950-2050 1

SMHI 1950-2050 1950-2050* 2

UCLM 1950-2050 1

C4I 1950-2050 1

GKSS** 1950-2050* 1

Met.No** 1950-2050* 1

CHMI** 1950-2050* 1

Total (1950-2050) 4 6 2 3 2 17

Some RCM related issues• Good availability of ERA-40 based output• Officially now Dec 2007 for scenario runs• But some earlier?• Quick-look analysis (month 40?)

• ‘Evaluated RCM-system for use in RT2B (choice of RCM-GCM combinations and preliminary RCM weights’)– D3.3.1; Mm3.3– D3.2.2 – describes a set of preliminary weights (PRUDENCE)– Final weights will be based on ERA40@25– Proposing to apply revised REA– ‘Could’ explore Tebaldi et al. Bayesian approach

Refinement of Reliability Ensemble Averaging (REA) method – Filippo Giorgi (D2B.6)

W = F1 x F2 x F3 x F4 x F5

Inverse functions:F1 local mean T biasF2 local mean P biasF3 interannual T Std. Dev. biasF4 interannual P Coeffic. Var. bias

Direct function:F5 correlation obs/sim SLP patterns

D2B.8 recommendations on weighting

• Robust• Informed by processes/expert knowledge• Transparency• Seasonal, range of variables, IAV/trends etc• A common comprehensive/flexible scheme• But some users want ‘tailoring’• Consultation with users• Avoid double counting• Compare weighted/unweighted

Can weighting be used to improve ‘credibility’?Can ENSEMBLES develop a ‘seamless’ approach?

Need broader discussion of these weighting, credibility, reliability issues

• Web forum posting (Jens/Linda/Clare)

• Side event during IAMAS, Italy, 2-13 July

• Next ENSEMBLES GA, November

Scenario generator tools and outputsA scenario generator tool would process dynamically and/or statistically downscaled output for user-specified locations, variables and time periods in a fairly transparent manner – presenting probabilistic regional scenarios in the desired format(s).

Would such a tool be useful to you:Definitely / maybe / no / don’t know

The outputs could be presented in a number of different formats. Please indicate those that would be useful:

Probability density functions Definitely / maybe / no / don’t know

Cumulative density functions Definitely / maybe / no / don’t know

Percentile values (e.g., 10th, 50th, 90th) Definitely / maybe / no / don’t know

Probability of exceeding specified threshold(s) Definitely / maybe / no / don’t know

Response surfaces Definitely / maybe / no / don’t know

Maps Definitely / maybe / no / don’t know

Time series Definitely / maybe / no / don’t know

Joint probabilities (give examples of variables if possible)

Definitely / maybe / no / don’t know

Tailoring of ENSEMBLES regional climate scenario outputs to user needs:

a questionnaire for users, stakeholders and scenario developers

What regional data do users want?• Mainly ‘standard’ surface variables• Daily time series (some sub-daily)• 25/50 km and/or station scale• Indices: blocking, NAO, heatwaves, drought, flooding

• Extremes:– Max 5-day rainfall, Max daily precipitation intensity– Heatwaves, Max wind gust– All kinds!, Will calculate own

• Joint probabilities– Temperature and precipitation– Intense precipitation and wind– Temperature and wind– ??????

Will they/you get what they/you want?

• Majority will use RCM ‘raw’ data

• Willingness to use SDS data

• All seem satisfied! (temporal scale)

What are preferred scenario formats?

• PDFs and time series most popular

• Interest in threshold exceedence

• Also maps and joint probabilities

• Some challenges & contradictions

What tools are available/needed?

• Climate Explorer• Extremes in gridded data sets (D4.3.1)• STARDEX extremes software

• General awareness of tools• Not many users (so far)• Support for better integration with regional scenarios

• Scenario generator tools????????• Lots of potential users for SDS portal

RCM 1

RCM 13

Weather generator

Change in T & P2071-2100

The CRANIUM methodology

Weather generator

100 x 30 yr runs

HIRHAMHIRHAM (ECHAM4) HadRM3PCHRMCLMREMORCAO RCAO (ECHAM4)PROMESRegCMRACMOArpege (HadCM3)Arpege (Arpege)

13 RCM runs from PRUDENCEhttp://prudence.dmi.dk/

10 RCMsForcing from 4 GCMsMost driven by HadAM3All A2 (Medium-high) emissions

• Histograms, PDFs, CDFs etc

• 10 UK case-study locations• Linkoeping, Karlstad• Saentis, Basel• Belgrade, Kaliningrad, Timisoara

• 2080s – Medium-high scenario

• 10 seasonal indices: • means• extremes

e.g., hot days, intense rainfall

39,000

Outline of RT2B approaches: D2B.1 & D2B.2

Preparing datasets

RCM data server – D2B.3Reanalysis – D2B.13Observed – D2B.15GCM-based – D2B.17

Developing/testing models

Statistical: D2B.5, D2B.16Dynamical: D2B.9, D2B.10

Issues and methods

Ensemble averaging: D2B.6Pattern scaling: D2B.7, D2B.25Weighting: D2B.8GCM-RCM matrixRCM quick-look: D2B.21

Interactions with users (RT6)

Web-based downscaling service: D2B.4, D2B.19, D2B.23QuestionnairesDevelopment of tools: D2B.18Preliminary assessment: D2B.20

s2d statistical downscaling: D2B.12

Modification of SDS methods for probabilistic framework: D2B.14

From month 30, the emphasis is on synthesis, application and scenario construction

RT3:ERA@50/25: D3.1.4, D3.1.5 RCM weights: D3.2.2RCM system: D3.3.1

Final RCM system: D3.3.2 GCM/RCM skill/biases: D3.4.1RT3:

25 km scenario runs: D2B.22 mo 36

RCM quick look analysisD2B.24: mo 40

D2B.11: mo 31

Dynamical and statistical downscalingProbabilistic regional scenarios and tools

Applications to case studies• Alps, Mediterranean (D2B.28)…• Storms, CWTs, blocking….• Forestry, water….

Recommendations & guidance on methods for the construction of probabilistic regional climate scenarios: D2B.26 mo 42

s2d dynamical downscalingINM/RCA: MARS

Questions & issuesSources of uncertaintyReducing uncertaintyRobustness of SDS (D2B.27) Synergistic use of SDS/DDS

RT1: Grand PDFsRT2A: stream 1 runs (s2d, ACC)RT5: gridded data set

Mo 36 on:

Recommended