Changes in rainfall patterns in the
Southeast U.S.A.
G.A. Baigorria, J.W. Jones, J.J. O’Brien
([email protected])
• Baigorria, G.A., J.W. Jones and J.J. O’Brien. 2007. Understanding rainfall
spatial variability in the southeast USA. International Journal of Climatology,
27(6):749-760.
Objective
To characterize the spatial and temporal changes of rainfall
in the Southeastern USA during the last 90 years
Data
Weather station network
Final 523 weather stations
Historical daily rainfall data
record from 1048 weather
stations obtained from the
National Climate Data Center
NOAA-NCDC §
from 1915 to 2004
§
http://nndc.noaa.gov/?home.shtml
Temporal analysis
of the
weather station network
Number of weather stations
180
150
120
90
60
30
0
Number of weather stations
15
350
45
60
75
90
Historical record length (years)
300
250
200
150
100
50
0
1915
30
1930
1945
1960
Years
1975
1990
Methods: Quality control
• Range errors and zeros that had been substituted for replacing
missing values were deleted
• For monthly analyses, months with fewer than 20 days of data
were not considered
• Weather stations beyond their State and County limits for which
coordinates could not be corrected were deleted
Methods: Geographical Trends
• Polynomial equations for each month and year were fitted using
least squares method. The degree of the polynomial equation for
each month and year was selected by comparing the sum of the
squared residuals as a percentage of the original variation
January
July
36
36
35
150.0-160.0
34
140.0-150.0
33
130.0-140.0
35
230.0-240.0
34
220.0-230.0
33
120.0-130.0
32
31
190.0-200.0
100.0-110.0
31
180.0-190.0
170.0-180.0
30
29
70.0-80.0
29
28
60.0-70.0
28
40.0-50.0
30.0-40.0
20.0-30.0
25
0.0-10.0
150.0-160.0
140.0-150.0
27
120.0-130.0
26
110.0-120.0
25
10.0-20.0
24
-89 -88 -87 -86 -85 -84 -83 -82 -81 -80
160.0-170.0
130.0-140.0
50.0-60.0
26
200.0-210.0
32
80.0-90.0
27
210.0-220.0
110.0-120.0
90.0-100.0
30
250.0-260.0
240.0-250.0
160.0-170.0
24
-89 -88 -87 -86 -85 -84 -83 -82 -81 -80
100.0-110.0
90.0-100.0
80.0-90.0
70.0-80.0
60 0 70 0
Rˆ = b0 + b1φ + b2φ λ + b3φλ + b4φλ2 + b5λ3
Rˆ = b0 + b1φ + b2λ
3
φ
= Latitude
λ = Longitude
2
Methods: Interpolation
• After removing the geographical trend, semivariograms were calculated
2
1
(xi − x j )
γ (h ) =
∑
2 N (h ) (i , j ) hij ≈h
• Model fitting by the Stable model
[
γˆ (h ) = δ o + δ1 1 − e
δo = 0
− ( h / A )β
]
γ Semivariance
h Distance
x Rainfall residual
δ o Nugget
δ1 Sill
A Range
Nugget was forced to zero to preserve observed rainfall values
Methods: Interpolation
• Rainfall residuals were interpolated using Ordinary Kriging
(grid cells of 5 km x 5 km)
• Resulting monthly maps were divided into 6 periods of 15 years
each
I
II
III
IV
V
VI
1915
1916
…
…
1929
1930
1931
…
…
1944
1945
1946
…
…
1959
1960
1961
…
…
1974
1975
1976
…
…
1989
1990
1991
…
…
2004
Methods: Statistical Analysis
• Analysis of Variance F-statistics for each grid cell in the
geographical domain was performed
• Duncan’s multiple range test was performed for each grid cell
where significant differences were found (P ≤ 0.05)
• Monthly matrices of maps showing the areas where statistical
differences were detected when comparing each 15-year period
against the other periods were obtained
Results: Differences Between 15-Year Periods
Maps of probability value that the F-test statistics
is at least as large as the observed F-value (α = p).
January
July
Results:
Identifying
differences over
space & time
1945-1959
1945-1959
1930-1944
1930-1944
1960-1974
1960-1974
1990-2004
1990-2004
1915-1929
1915-1929
January
1930-1944
1930-1944
Maps of Duncan’s
multiple range tests
1945-1959
1945-1959
84°0'0"W
81°0'0"W
33°0'0"N
±
33°0'0"N
87°0'0"W
Statistically significance
difference
1975-1989
1975-1989
1960-1974
1960-1974
30°0'0"N
Probability
1.0
0.8
0.6
0.4
0.2
0.0
27°0'0"N
27°0'0"N
30°0'0"N
Non-significant
0
87°0'0"W
200
84°0'0"W
400
km
81°0'0"W
1975-1989
1975-1989
Results:
Identifying
differences over
space & time
1930-1944
1930-1944
1945-1959
1945-1959
1960-1974
1960-1974
1990-2004
1990-2004
1915-1929
1915-1929
July
1930-1944
1930-1944
Maps of Duncan’s
multiple range tests
1945-1959
1945-1959
84°0'0"W
81°0'0"W
33°0'0"N
±
33°0'0"N
87°0'0"W
Statistically significance
difference
1975-1989
1975-1989
1960-1974
1960-1974
30°0'0"N
Probability
1.0
0.8
0.6
0.4
0.2
0.0
27°0'0"N
27°0'0"N
30°0'0"N
Non-significant
0
87°0'0"W
200
84°0'0"W
400
km
81°0'0"W
1975-1989
1975-1989
Results: Temporal pattern in the
geographical trend
0
-100
January
-200
Rˆ = b0 + b1φ + b2λ
φ
= Latitude
λ
1915- 1924- 1933- 1942- 1951- 1960- 1969- 1978- 1987- 19961923 1932 1941 1950 1959 1968 1977 1986 1995 2004
-300
-400
= Longitude
b1
-500
-600
-700
-800
36
35
160.0-170.0
-900
150.0-160.0
34
140.0-150.0
33
130.0-140.0
-1000
12
120.0-130.0
32
31
30
110.0-120.0
100.0-110.0
10
90.0-100.0
80.0-90.0
29
70.0-80.0
28
60.0-70.0
27
26
25
50.0-60.0
40.0-50.0
30.0-40.0
8
b2
6
4
20.0-30.0
10.0-20.0
24
-89 -88 -87 -86 -85 -84 -83 -82 -81 -80
0.0-10.0
2
0
1915- 1924- 1933- 1942- 19511923 1932 1941 1950 1959
1960- 1969- 1978- 1987- 19961968 1977 1986 1995 2004
Conclusions
• Changes in rainfall patterns occurred in some areas but not over
the entire region
• Changes in rainfall were detected across time during the last 90
years, not only during the last 15-year period
• In areas where changes occurred, rainfall tended to increase during
winters and decrease during summers over the 90-year time period
• Isolines of high rainfall amounts in January shifted from the
Southeast to the Northwest in a 27-year cycle