Evaluation of Climate Indicators

L. Buja (NCAR) and W. Gutowski (Iowa State U.) L. Kaatz (Denver Water), A. Abatan (Iowa State U.),

CM Ammann, BG Brown, R. Bullock, T. Fowler, E. Gilleland, J. Halley-Gotway, K. Ikeda, D. Yates (NCAR)

Sectoral Perspectives

Standardized Model Diagnostics: CESM Example http://www2.cesm.ucar.edu

TS Trends (DJF)

TS 12-yr running Trends (monthly data)

Weather Forecast Verification Tools offer new and expanding opportunities

Various groups of tools:

• Spatial methods (object-based, neighborhood, field deformation)

• Probabilistic Methods • Extreme Methods

(return intervals, conditional spatio-temporal fields)

• Performance diagrams (multi-diagnostic, Taylor, Roeber-event-based)

from pixels to objects

Object-Based Evaluation of CESM-LE MODE: Method for Object-based Diagnostic Evaluation

(http://www.dtcenter.org/met/users)

• target application-specific

indices with important impact

• automatically id and merge “pixels” into aggregated features (objects)

• match objects between model run and reference (obs, other model) using fuzzy logic

• Capture full statistics of spatial objects, size, orientation, intensity, spatial biases to reference, etc

Ensemble Spread of Precipitation Fields

Large Scale Precipitation

Convective Precipitation

PRECC PRECL

one year DJF precipitation over 30 CESM-LE members

Significant spread across ensemble

little spread

Large Scale Precipitation via Object Attributes

Individual precipitation objects identified by MODE across 30 ensemble members

Little ensemble spread

More ensemble spread

Little ensemble agreement

Intense Precipitation (90th percentile) across 30 ensemble members

(unit: cm / day average for DJF-season)

Global Objects of pr>500mm

Area : Obs often 2x larger than CESM Number : Obs 25% less than CESM >> CESM is more patchy than CRU-TS3.21 90th%ile intensity: ~10% higher in Obs >> CESM has less intense rainfall Observations show larger areas with more uniformly high precipitation

CRU CESM

Obj Areas Obj Counts

90th %ile intensity CRU CESM

Mean Climate Verification with MODE

• 30 year precipitation climatology: 100mm verification • GCM in blue outlines • Observed areas in color

Attributes of the South East US Rain Objects

Area Ratio of GCM and Observation: Median GCM Precipitation: Median Observed Precipitation:

Well represented mean!

Conditional Object Frequency Precipitation DJF - Niño 3.4 > +1

CRU TS3.21 CESM-LE

>100mm

>300mm

Conditional Object Frequency Precipitation DJF - Niño 3.4 > -1

CRU TS3.21 CESM-LE

>100mm

>300mm

Observations CRU TS3.21

Models CESM-LE

Field: pr(DJF)

Freq

50%

25%

75%

100%

0%

El Niño Influence on Occurrence of >900mm DJF Rainfall

CRU TS3.21 pr (djf) CESM-LE pr (djf)

EN- EN+ EN- EN+

Freq

50%

25%

75%

100%

0%

MODE Comparison of SPI and SPEI (36-mo averages, 1950-2012)

SPEI SPI

Object Area

Credit: A. Abatan

Object Intensity: Variability

Next steps to further evaluate climate models • Expand location and process specific performance evaluation:

Focus diagnostics on systematic biases resulting from lack of process representation, spatial displacements, temporal distribution issues

• Need for evaluation of different characteristics of climate: Indices in addition to traditional climate fields

• Conditional analyses using spatio-temporal context of regional multivariate variability

Abitan et al.: US Precipitation 1901-2005 in CRU-gridded and CESM-LE

Thanks!

Improving and Strengthening the Knowledge Chain

Examples of Tools for: • Access to useful data • Efficient evaluation for applications • Approaches to better questions

and appropriate translation

Benefits: • Improved science tools • Better informed decision making • Educational opportunities

Objectives: Relevant Information • Water Sector: Inform management and planning

decisions with relevant weather & climate information (knowledge chain: access, evaluation, translation, good practice)

• Climate Research Community: Understand weather & climate challenges, improve and translate the relevant information (understand challenges at relevant spatial and temporal scales)

• CoDesign Weather & Climate Dashboards with relevant / actionable information (transparent, tied to observations, translated for understanding and context, probabilistic, …)

Relevant Information through a Dashboard

Dashboard: Sorghum in Sudan Sorghum Growth Region

MAP

Climate Change Impact Summary RCP8.5 RCP4.5 RCP2.6

Climate Change Impact Potential

Dashboard: Sorghum in Sudan Sorghum Growth Region

MAP

Climate Change Impact Summary

Future Production Index Change

Cooling Degree Days Trend in Energy Requirement

Future Growth Suitability Index

RCP2.6 RCP4.5 RCP8.5

Seasonal ENSO Teleconnections

Daily Peak Temperatures

RCP8.5 RCP4.5 RCP2.6

Climate Change Impact Potential

RCP8.5

RCP4.5

Drought Occurrence

Explanation of Inter-annual Variability

Dry Period Extremes Storage Loss

Traditional grid-point by grid-point verification

Which simulation result is better? Fig. courtesy of E. E. Ebert

observed models with different resolution