Impact of common SST anomalies on global drought and pluvial frequency

1
Impact of common SST anomalies on global drought
and pluvial frequency

• Kirsten Findell and Tom Delworth
• Geophysical Fluid Dynamics Laboratory
• Princeton, NJ
• 33rd Climate Diagnostics and Prediction
  Workshop/CLIVAR Drought Workshop (CDPW)
• October 2008

2
SST forcing patterns used for Clivar Drought
Working Group Study

• Climatology derived from Hadley Centre Monthly
  SST data, 1901 to 2004 (HADISST1, Rayner et al
  2003)
• Rotated EOF analysis first three modes
• Linear trend pattern (27.2), scaled by 1s
• Pacific pattern (20.5), scaled by 2s
• North Atlantic pattern (5.8), scaled by 2s

3
SST forcing patterns used for Clivar Drought
Working Group Study
REOF 3
REOF 2
REOF 1
All runs fixed SSTs, 50 years long Control run
Climatological SSTs
4
Outline

• SST forcing patterns
• Drought index used SDDI
• Extensive look at Cold Pacific experiment
• GFDL model to demonstrate the methodology
• Multi-model analysis (GFDL, NSIPP, CAM)
• The three models tend to yield similar results in
  terms of drought/pluvial frequency and intensity
• Pac/Atl combination experiments, Multi-model
  means
• Positive/negative forcings tend to yield opposite
  results
• Trend and Trend/Pac/Atl combinations
• What do they tell us about the relative strengths
  of these different forcings in the present day?
• Trend impacts are generally overwhelmed by
  Pac/Atl impacts
• Additional experiments to look at future
  scenarios
• Trend impacts should not be overwhelmed by
  Pac/Atl impacts in the future
• Conclusions

5
Supply-Demand Drought Index (SDDI)

• From Rind et al. (JGR, 1990)
• Difference between moisture supply (precip) and
  demand (potential evap)
• Similar in construct to PDSI
• dSDDI P Ep (P Ep)clim
• ZSDDI d/s
• Y(i) 0.897 Y(i-1) Z(i)
• (P Ep)clim, s seasonal cycle (monthly values)
  determined from Control run
• Benefits
• Tied to soil moisture (through evaporative
  demand), but since soil moisture is very model
  dependent (e.g., depth saved) this index is good
  to use when looking at multiple models
• No grid-specific empirical coefficients to
  estimate (might these change if climate changes?)

6
Cold Pacific Changes in Mean SDDI
GFDL model Annual mean SDDI differences, where
significantly different from control (95 level,
modified t-test)
7
Cold Pacific Drought Frequency
Average number of months/year in drought (SDDI lt
-2.0)
GFDL model differences in mean
8
Cold Pacific Changes in Drought Frequency
Difference from Ctl, months/year in drought (SDDI
lt -2.0)
GFDL model differences in mean
Locations with reductions in mean SDDI also show
increased drought frequency. There are some
additional locations with minor increases in
drought frequency, despite a lack of significant
change in the mean SDDI (Central Asia, Central
Africa).
9
Cold Pacific Pluvial Frequency
Average number of months/year with SDDI gt 2.0
GFDL model differences in mean
Locations with increased mean SDDI also show
increased pluvial frequency. There are some
additional locations with minor increases in
pluvial frequency, despite a lack of significant
change in the mean SDDI (high latitudes).
10
Cold Pacific Drought Frequency
Multi-model mean
11
Cold Pacific Drought Intensity
Multi-model mean
12
Cold Pacific Pluvial Frequency
Multi-model mean
13
Cold Pacific Pluvial Intensity
Multi-model mean
14
Model agreement in Cold Pacific run

• Drought is
• Much more frequent in Continental US and southern
  South America
• Slightly more frequent in eastern Canada, from
  35-50N in Europe and Asia, and along the east
  coast of Asia
• Pluvials are
• Much more frequent in Central America, northern
  South America, and Oceania
• Somewhat more frequent in Arabia and Australia

15
Model differences in Cold Pacific run

• CAM is much more sensitive than GFDL and NSIPP
  throughout Africa (greater tendency towards
  pluvials in this run)
• African pluvial pattern in CAM run extends into
  Mediterranean region, pushes European drought
  region further north
• GFDL model more sensitive in high latitudes (more
  pluvials in this run)
• The three models differ in the central location
  and intensity of the drought response in the US

16

Drought Frequency, MMM, Pac/Atl runs
Cold Atlantic
Cold Pacific
Warm Pacific
Warm Atlantic
17

Pluvial Frequency, MMM, Pac/Atl runs
Cold Atlantic
Cold Pacific
Warm Pacific
Warm Atlantic
18
SDDI time series in Central US Pac/Atl
combinations GFDL model
19
SDDI time series in Central US Do we see
linearity in the responses?
GFDL model
20
SDDI time series in Central US Do we see
linearity in the responses?
GFDL model
21
Warm Linear Trend Drought Frequency
Multi-model mean
22
Warm Linear Trend Pluvial Frequency
Multi-model mean
23
Model agreement in Warm Trend run

• Modest increases in drought frequency
• 30-45N in North America (Central US) and
• 35-50N in Europe and Asia (Mediterranean region
  and extending eastward)
• Central and Southern Africa
• Modest increases in pluvial frequency
• North of 55N, particularly in Asia
• Central America and NE coast of South America
• Eastern Australia

24
Drought in Trend, Pac, Atl combinations
Cold Trend, Cold Pac, Warm Atl
Cold Trend, Warm Pac, Cold Atl
Cold Trend
Warm Trend, Warm Pac, Cold Atl
Warm Trend
Warm Trend, Cold Pac, Warm Atl
GFDL model
25
Pluvials in Trend, Pac, Atl combinations
Cold Trend, Cold Pac, Warm Atl
Cold Trend, Warm Pac, Cold Atl
Cold Trend
Warm Trend, Warm Pac, Cold Atl
Warm Trend
Warm Trend, Cold Pac, Warm Atl
GFDL model
26
Trend/Pac/Atl combinations

• Impacts of Trend are overwhelmed by
  Pacific/Atlantic impacts
• Since these standard runs had the Trend EOF
  multiplied by 1s, and Pac/Atl multiplied by 2s,
  this result is in part due to the experimental
  design
• Present-day runs composites of the three EOFs
  all multiplied by 1s
• Should indicate if Trend impacts are overwhelmed
  by ENSO/NAO in current world
• Future-scenario runs composites of the Pac/Atl
  EOFs multiplied by 1s, Trend by 2s
• Should indicate if Trend impacts will be
  overwhelmed by ENSO/NAO in the future, assuming
  that ENSO/NAO dont change much in the future

27
Droughts and pluvials in the present
Warm 1s Trend, Cold 1s Pac, Warm 1s Atl
Warm 1s Trend
Warm 1s Trend, Warm 1s Pac, Cold 1s Atl
Drought increases top row
Pluvial increases bottom row
GFDL model
28
Present-day scenario results

• Drought/pluvial frequency is a strong function of
  Pacific and Atlantic condition
• Exceptions
• Central Africa has frequent droughts
• Southern India and SE Australia have frequent
  pluvials
• Calls into question the validity of fixed SST
  experiments for the Indian Ocean

29
Droughts and pluvials in the future
Warm 2s Trend, Cold 1s Pac, Warm 1s Atl
Warm 2s Trend
Warm 2s Trend, Warm 1s Pac, Cold 1s Atl
Drought increases top row
Pluvial increases bottom row
GFDL model
30
Future scenario results

• Droughts become more frequent in the future,
  independent of ENSO/NAO
• Mediterranean Sea region and east past Caspian
  Sea (30-50N in Africa, Europe, and Asia)
• Central Africa
• Pluvials become more frequent in the future
• North of 50N (esp. in Europe and Asia)
• Coastal Asia
• India and Australia (esp. SE)
• Caveat experimental validity in Indian Ocean
  region?
• Drought/pluvial frequency remains a strong
  function of Pacific and Atlantic condition
• Continental US
• Southern and Northern South America

31
Conclusions

• The three models (GFDL, NSIPP, CAM) generally
  yield similar drought/pluvial results
• Cold Pacific
• Increased drought in US, southern SAm., (S
  Europe, SE coast Asia)
• Increased pluvials in Cent. Am., northern SAm.,
  Oceania, (Arabia, Australia)
• Warm Pacific
• In general, opposite of Cold Pacific
• Cold Atlantic
• Increased droughts in Cent. Am., northern SAm.,
  Central Africa, (Oceania)
• Increased pluvials in cent. SAm., (cent. US)
• Warm Atlantic
• In general, opposite of Cold Atlantic
• Warm Trend
• Modest changes that are overwhelmed by Pac/Atl
  impacts in combination runs

32
Conclusions (cont.)

• Areas with an increase (decrease) in the mean
  SDDI tend to also show an increase (decrease) in
  the frequency of extreme SDDI values
• Areas with more frequent extremes also tend to
  show higher intensity extremes
• Present day experiments suggest
• The impact of the background warming trend is
  generally overwhelmed by the ENSO and NAO signals
• Questionable use of fixed SST models in Indian
  Ocean Basin
• Future experiments suggest
• The impact of the trend will be more dominant
  than the ENSO and NAO signals in many areas of
  the globe when the magnitude of the trend has
  doubled, assuming ENSO and NAO characteristics
  remain relatively stable.