Past and Projected Changes in ContinentalScale AgroClimate Indices

  • Slides: 33
Download presentation
Past and Projected Changes in Continental-Scale Agro-Climate Indices Adam Terando NC Cooperative Research Unit

Past and Projected Changes in Continental-Scale Agro-Climate Indices Adam Terando NC Cooperative Research Unit North Carolina State University 2009 NPN RCN Meeting

Motivating Questions • Is the late 20 th century warming found in the surface

Motivating Questions • Is the late 20 th century warming found in the surface temperature record also observable in alternative climate measures that are critical to agricultural production and phenological observations in North America? • Do Global Climate Models (GCMs) have skill in hindcasting the observed trends? • What changes do GCMs predict for the future?

Global Scale Global Mean Temperature over Land & Ocean National Climatic Data Center: 2006

Global Scale Global Mean Temperature over Land & Ocean National Climatic Data Center: 2006

BUT…. . An increase in mean global surface temperature will not necessarily be reflected

BUT…. . An increase in mean global surface temperature will not necessarily be reflected in the same manner for other manifestations of the climate system over the same time period and at different spatial scales.

Meehl et al. 2000

Meehl et al. 2000

A Temperature Example Heat Stress Frost/Freeze Crop Growth

A Temperature Example Heat Stress Frost/Freeze Crop Growth

Agro-Climate Indices • Annual Frost Days (tmin < 0 o. C) • Growing Degree

Agro-Climate Indices • Annual Frost Days (tmin < 0 o. C) • Growing Degree Days (thermal time) for Corn (10 < tavg < 30 o. C) – Strong correlation with crop growth • Heat-Stress Index (tmax > 30 o. C)

US and Canadian Long-term Historical Climate Networks

US and Canadian Long-term Historical Climate Networks

-0. 6 -0. 4 Trend Time Periods • 1956 – 2005: Good data coverage

-0. 6 -0. 4 Trend Time Periods • 1956 – 2005: Good data coverage • Switch in 1970 s • Warming signal detected then on global scale. • Also coincides with phase shift in North American teleconnections (i. e. PDO, NAO) • Most recent data 1880 1900 1920 1940 1956 1960 1976 1975 1980 2005 2000 -0. 2 0. 0 0. 2 0. 4 0. 6

SPATIAL PATTERNS

SPATIAL PATTERNS

Frost Trends (1956 – 2005) -1 Slope (Days/Year) < -0. 5 0 > 0.

Frost Trends (1956 – 2005) -1 Slope (Days/Year) < -0. 5 0 > 0. 5 1

Growing Degree Day Trends (1956 – 2005) 7 Slope (Days/Year) >5 < -5 0

Growing Degree Day Trends (1956 – 2005) 7 Slope (Days/Year) >5 < -5 0 -7

Heat Stress Index Trends (1956 – 2005) 10 Slope (Degree Days Per Year) >

Heat Stress Index Trends (1956 – 2005) 10 Slope (Degree Days Per Year) > 2. 5 0 < -2. 5 -10

Percent Stations with Significant Trends

Percent Stations with Significant Trends

a) b) c) • Trends fairly consistent through time

a) b) c) • Trends fairly consistent through time

GCM Results

GCM Results

GCM Data • 17 GCMs available from Lawrence Livermore National Laboratory • Models used

GCM Data • 17 GCMs available from Lawrence Livermore National Laboratory • Models used in IPCC AR 4 • Fewer years and model runs available for daily data than for monthly data (requires more storage!) • Typically 40 years available for 20 th century (1961 – 2000), and two 20 years periods for 21 st Century (2045 – 2065 and 2081 – 2100)

Questions • Do GCMs have skill in simulating past changes in agro-climate indices? •

Questions • Do GCMs have skill in simulating past changes in agro-climate indices? • What future changes do GCMs predict? • Is the (projected) signal strong with respect to the model noise?

Evaluating GCM Skill

Evaluating GCM Skill

Frost Days r = 0. 52 SLPobs = 0. 04 SLPgcm = 1. 59

Frost Days r = 0. 52 SLPobs = 0. 04 SLPgcm = 1. 59 r = 0. 17 Observations GCM Results GCM Arithmetic Mean SLPobs = 0. 50 SLPgcm = 3. 42 SLPobs = -0. 22 SLPgcm = -0. 21 HSI GDD r = 0. 03 • Poor performance for GDD and HSI evident in trend lines • Good agreement with frost days

Taylor Diagram Correlation Coefficient RMS Error Model Result Standard Deviation Observation or ‘Perfect’ Model

Taylor Diagram Correlation Coefficient RMS Error Model Result Standard Deviation Observation or ‘Perfect’ Model Taylor 2001

Model Weighting GCMs “perfect” model Schneider et al. 2007

Model Weighting GCMs “perfect” model Schneider et al. 2007

Frost Days Correlation Coefficient Standard Deviation Centered RMS Difference Thermal Time Heat Stress Index

Frost Days Correlation Coefficient Standard Deviation Centered RMS Difference Thermal Time Heat Stress Index Standard Deviation Centered RMS Difference

a) Correlation Coefficient b) Correlation Coefficient 16 c) Standard Deviation Centered RMS Difference Heat

a) Correlation Coefficient b) Correlation Coefficient 16 c) Standard Deviation Centered RMS Difference Heat Stress Index Year d) Heat Stress Days Year

Maximum Temperature Correlation Minimum Temperature Negative Standard Deviations Positive Standard Deviations

Maximum Temperature Correlation Minimum Temperature Negative Standard Deviations Positive Standard Deviations

bccr-bcm 2. 0 echam 5 -MPI miroc 3. 2 mri-cgcm 2. 3. 2 observations

bccr-bcm 2. 0 echam 5 -MPI miroc 3. 2 mri-cgcm 2. 3. 2 observations

Projections

Projections

IPCC Emission Scenarios A 2 Scenario

IPCC Emission Scenarios A 2 Scenario

Frost Days Observations GCM Results GCM Arithmetic Mean 2046 -2065 Weighted Mean 2081 -2100

Frost Days Observations GCM Results GCM Arithmetic Mean 2046 -2065 Weighted Mean 2081 -2100 Weighted Mean Heat Stress Index Thermal Time

 • Projected changes large relative to model errors for 20 th century •

• Projected changes large relative to model errors for 20 th century • Largest uncertainties (model spread) around HSI projections

Conclusions • General signal agreement between Tavg and agro-climate indices. • Strong increase in

Conclusions • General signal agreement between Tavg and agro-climate indices. • Strong increase in Thermal Time and decrease in Frost Days that is not seen in HSI. • Still difficult for GCMs to model variables requiring high temporal resolution. • Ensemble mean has greater skill than indiviudal GCMs • Large changes in agro-climate indices predicted by GCMs for A 2 scenario.