60N
30N
EQ
30S
60S
0
60E
120E
C MAP
C NR M_C M3
180
120W
C S IR O_MK 3.0
G F DL_C M2.0
60W
0
G F DL_C M2.1
MIR OC _3.2
60N
30N
EQ
30S
60S
0
60E
120E
C MAP
MP I_E C HAM5
180
MR I_C G C M2
NC AR _C C S M3
120W
60W
0
NC AR _P C M1
UK MO_HadC M3
Fig. 7. Climatology of precipitation as summarized by the 4 mm day–1 contour line for
JJA for each model (contour colors noted in key) and CMAP observations (blue shading
over 4 mm day–1). The JJA average of 1979–2000 is used because of the availability of
observations.
C NR M_C M3
HadC M3
30N
20N
10N
EQ
10S
20S
0
60E
120E
180
120W
60W
0
C S IR O_MK 3
30S
0
60E
120E
180
120W
60W
0
60W
0
60W
0
60W
0
60W
0
MIR OC _3.2_medres
30N
20N
10N
EQ
10S
20S
0
60E
120E
180
120W
60W
0
E C HAM5
30S
0
60E
120E
180
120W
MR I_C G C M2
30N
20N
10N
EQ
10S
20S
0
60E
120E
180
120W
60W
0
G F DL_C M2.0
30S
0
60E
120E
180
120W
NC AR _C C S M3
30N
20N
10N
EQ
10S
20S
0
60E
120E
180
120W
60W
0
G F DL_C M2.1
30S
0
60E
120E
180
120W
NC AR _P C M1
30N
20N
10N
EQ
10S
20S
0
60E
120E
180
-- -4
120W
---3
60W
0
30S
0
---2 -- -1 .5 -- -1 --0 .5 0.5
60E
1
1.5
120E
2
3
180
120W
4
Fig. 8. Precipitation change in the JJA season for 2070–2099 relative to 1901–1960 (mm day–1) from 10
ocean-atmosphere climate models for the SRES A2 global warming scenario. Values are shown only where
they exceed the 95% confidence interval for a two-sided Student t test. The variance for the test is estimated
at each point in the approximation that 10-year intervals in the control run are independent. One ensemble
member is shown for each model because ensemble size varies from 1 to 5. For these 30-year mean values
ensemble members bear close resemblance; for instance, for NCAR-CCSM3 spatial correlations of one
ensemble member to another all exceed 0.95. The 4 mm day–1 contour interval from the 1901–1960 base
period is shown for each model.
30N
20N
10N
EQ
10S
20S
Inter-model Dry/Wet trend agreement
0
60E
2
120E
3
4
5
6
7
Dry trend number of models
180
120W
2
60W
0
3
4
5
6
7
Wet trend number of models
Fig. 9. Model agreement on the predicted local precipitation trend for JJA from 1979 to 2049. The
number of models at each location that agree on a dry or wet trend are counted as in Fig. 5b, but for
less stringent criteria because of the smaller signal for earlier time. The trend at each point must exceed
95% significance and exceed an amplitude change per century of 20% of the climatology for each
model, and regions with 2 or more of the 10 models agreeing are shaded, with the number given by the
brown (dry) and green (wet) color bars. Points are excluded if any models pass these criteria for a trend
of disagreeing sign. In addition to the Caribbean/Central-American region, a degree of intermodel
agreement on drying trends may be noted in the Southeastern Pacific, eastern equatorial South America,
and southern Africa. These regions have smaller median trend magnitudes than the Caribbean/CentralAmerican region at this time and can be sensitive to the threshold for minimum trend. Wet trends with
local agreement for several models occur in the central Pacific and Indian oceans.
50
(a)
Occurences
40
30
20
10
0
-3
-2
-1
0
1
2
3
2
3
F ifty-year trend [mm/day per 100yr]
100
(b)
P ercent
80
60
40
20
0
-3
-2
-1
0
1
F ifty-year trend [mm/day per 100yr]
Fig. 10. (a) Multimodel histogram of occurrences of 50-year trend values as an estimate of the probability
density function. (b) Resulting cumulative distribution for 50-year trends of an index of Caribbean/CentralAmerican region precipitation, i.e., precipitation averaged over longitudes 91.875–58.125°W and latitudes
11.25–23.75°N. The trends are computed over nonoverlapping 50-year segments from the model control
runs and thus provide an estimate of 50-year trends occurring purely by natural variability. The value of the
observed 50-year trend from VASClimO data for the same averaging region is shown as a red line.