Transcript GLOBAL WARMING AND HURRICANE CORRELATION

GLOBAL WARMING AND
HURRICANE CORRELATION
BY THE SHARK TEAM
Null Hypothesis
• There Is No Correlation
Between Global Warming And
Hurricane Frequency And
Intensity
Global Warming Indicator
• Average global temperature
deviation data from 1899 until
present is used as the global
warming indicator in all
correlations and statistical
analysis.
• We consider the signature of
global warming to be present in the
temperature data from 1973 until
2005.
Hurricane Frequency
• The data shows a relative steady
frequency of hurricanes until a
distinct increase in the last decade
Number of Hurricanes
Number of Hurricanes Per Decade
100
80
60
40
20
0
1940-50
1951-61
1962-72
1973-83
Decades
1984-94
1995-05
Hurricane Pressure
• …and also a substantial decrease
in the average hurricane
pressure
Avg. Pressure of
Hurricanes
Avg. Pressure of Hurricanes Per Decade
990.0
980.0
970.0
960.0
950.0
940.0
930.0
920.0
910.0
900.0
1940-50
1951-61
1962-72
1973-83
Decade
1984-94
1995-05
Analysis Procedures
• ANOVA between the last two
decades of hurricane frequency
looking for a significant
difference
• Correlation between average
global temperature deviation from
1899 and hurricane frequency
• Correlation between temperature
from 1987 until present with the
hurricane frequency
Hurricane Freqency, Strength, and Global Warming
16
0.8
20 yr freq
20 yr ave cat
0.6
Dev 10 yr
12
Frequency
0.4
10
8
0.2
6
0
4
-0.2
2
0
1860
-0.4
1880
1900
1920
1940
Year
1960
1980
2000
2020
Temp Dev from 1800
14
This graph shows…
• Average hurricane strength as
measured by category has not
changed much over this time span.
• However, there is a sharp increase
in hurricane frequency after 1994
after a long period of downward
trend.
ANOVA analysis
• We attempted to quantify this
change in frequency using an
ANOVA between the years ’82’94 and ’95-’05.
ANOVA Results
• ’82-’94 yields an average of 3.6
hurricanes per year. ’95-’05
has an average of 7.85.
• The difference in means was
significant with p=.017.
ANOVA interpretation
• This means that there is a
significant change in
frequency in the last 10 years
compared to the previous 10.
Regression
• Since we are using global mean
temperature as our measure of
global warming, it seems
logical to look for a
correlation between the
temperature and hurricane
frequency.
Regression Analysis
• To this end, we ran two
regressions.
• The first was for the all the
data, the second was from ’73
on.
Hurricane Frequency and Global Warming
16
Hu
rricane Frequency
14
12
10
y = 4.5144x + 10.553
8
6
4
2
0
-0.4
-0.2
0
0.2
0.4
Deviation from 1899 Global Mean
0.6
0.8
First regression
• The first regression yielded an
F-score of 39.1 with 105
degrees of freedom. This yields
a p-value of 9e-9, which is very
highly significant.
But…
• Obviously, there are residuals
about the linear fit that are
non-random, especially a clump
around 0 on the X-axis.
Hurricane Frequency and Global Warming
16
Hurricane Frequency
14
12
10
y = 4.5144x + 10.553
8
6
4
2
0
-0.4
-0.2
0
0.2
0.4
Deviation from 1899 Global Mean
0.6
0.8
Explanation…
• If you look at the first graph, we
can see that hurricane frequency
has a peak that corresponds with
about a fifteen year lag behind the
global temperature.
Hurricane Freqency, Strength, and Global Warming
16
0.8
20 yr freq
20 yr ave cat
0.6
Dev 10 yr
12
Frequency
0.4
10
8
0.2
6
0
4
-0.2
2
0
1860
-0.4
1880
1900
1920
1940
Year
1960
1980
2000
2020
Temp Dev from 1800
14
More explanation…
• This lag means that for any change
in our X value (temperature), there
will be a time of about 15 years
before our Y values change, which
will cause a clump in the data.
This means…
• There is about a fifteen year lag
behind the global warming signal.
• Which means that the system hasn’t
fully responded to the increase in
global temperature.
More meaning…
• The data shows an approximately stable
slope in temperature increase
over the last 20 years. Running a
regression with hurricane frequency
should yield a good linear model with
some predictive power for future
hurricane frequency for the next 15
years.
’87 on Temp/Freq. Reg.
Temperature versus Frequency '87-'05
Frequency
15
10
5
0
0.2
0.3
0.4
0.5
Temperature
0.6
0.7
y = 7.8126x + 8.7406
Prediction
• Using a least squares fit the the
temperature data from ’73-’05, we
get a prediction of .77 degrees
from the 1899 mean temp and a
prediction of about 15 hurricanes
for 2020 up from 14 in 2005.
THE END
• In conclusion, the analysis of the
hurricane data and global
temperature—allows us to reject
our null hypothesis that the two
variables aren’t correlated.
But…
• The correlations also have a high
standard errors in our slope which
when factored in give a 95%
confidence interval for hurricane
frequency of –15 to 52. Obviously,
this range is not physical, which
leads us to conclude:
More but…
• 1) We used a poor proxy for global
warming
• or
• 2) There hasn’t been enough time
for the last uptick of temperature
to show in the frequency of
hurricanes.