Professional Documents
Culture Documents
Lab Report #2
Nice analysis and great report
Thanks for the effort
STAT 656 Ying Wang
25/25
Abstract
This study approaches the topic of global warming by investigating the
relationship between global temperature anomalies and CO2
emissions/solar radiation. We find it is the rise of CO2 emissions that is
linearly correlated with temperature anomalies in the period of 1961-1990,
but not the solar radiation. In addition, temperature anomalies behave
more extremely in the period 1981-1990, which indicates that human
activities might be the major cause because of industry booming, as other
natural forcing that change the composition of atmosphere do not have any
dramatic change in recent decades.
1. Introduction
Climate change has been a controversial issue for decades. The central piece of the
debate has been the cause of the climate change: whether it is attributed to the human
factors or the nature forcing. There is no clear answer to this question yet. For one thing,
this question is complicated enough by its own nature; for another, the findings of climate
change are confounded by the effects of human activity and nature forcing. To tackle this
perplexing problem, we begin by addressing a simpler question: is recent change in earth
temperature more related to solar activities or CO2 levels on earth?
Climate change greatly draws people’s attention in recent decades. Global warming is the
strongest sign. Figure 1 from NASA GISS 1880-2009 Global Temperature shows that
global land-ocean temperature anomaly has been rising fast since the mid-20th century.
-1-
!
Lab Report #2 STAT 656 Ying Wang
Nice reference
2.1.2 Factors
There are many contributing factors to the global warming, such as emissions of
greenhouse gases (e.g. CO2), solar output, volcanic eruptions and continental drift. Here
we focus our study on the effect of the first two factors: CO2 emissions and solar
radiation.
Albeit the sea would discharge more CO2 as the temperature rises, the increase in CO2
emissions is thought to be mainly due to the emission from the industrialization, because
CO2 discharge from the sea due to temperature rise was not so salient in the past, but only
emerging evident after the industry blossom. Probably need a reference here
Solar radiation is known to rise and fall periodically. See figure 2 created by Robert A.
Rohde / Global Warming Art below. Therefore, if the temperature anomaly is caused
solely by the solar radiation, it should also display a periodic pattern.
-2-
!
Lab Report #2 STAT 656 Ying Wang
For these two factors, there exists some evidence supporting the idea that human
activity is the cause but not the solar radiation. Gabriele Hegerl (Hegerl, Gabriele C.; et
al., 2007) pointed out solar output warms the stratosphere; whereas greenhouse gases
cool the stratosphere. Measurements show that the stratosphere has been cooling since
1979 (Randel, William J.; Shine, Keith P.; Austin, John; Barnett, John; Claud, Chantal;
Gillett, Nathan P.; Keckhut, Philippe; Langematz, Ulrike et al., 2009).
-3-
!
Lab Report #2 STAT 656 Ying Wang
2.2 Data
The data we use in this project includes the global yearly temperature anomalies (oC)
from 1901-2000 average, the yearly CO2 concentration (ppm) measured at Mauna Loa
Observatory, Hawaii, and the monthly sun radiation (kWh/m2/day) in Columbus, Ohio,
for the period of 1961-1990. Note that we would first convert the monthly sun radiation
data to yearly average to match up with other observations.
3. Statistical Analysis
3.1 Descriptive analysis
First, we conduct the descriptive analysis with the summary statistics of CO2
level, temperature anomaly, and sun radiation data. The table below includes
mean, sample standard deviation, minimum, maximum, and select quantiles of
each variable.
Summary Statistics
Mean Std. dev. Min 1st Quartile Median 3rd Quartile Max
Table 1 Summary statistics of CO2 level, temperature anomaly, and sun radiation.
3.2 Compare temperature anomaly, CO2 level, and sun radiation before and after 1980
Further, we want to examine the relationship between temperature rise and the two
factors, CO2 and solar output. Also, we would like to take into account of the time effect
for the global temperature anomaly.
Let us take a close look at the period that anomalies are greater than zero. The data of
1961-1990 seem to contain two clusters, one ranging from year 1961 to 1980, and
another ranging from year 1981 to 1990 (1980 is an arbitrary pick to segment the period
according to values of anomaly). We want to test the null hypothesis that anomalies after
1980 are not significantly different than before. The method used here is to draw the 95%
confidence region for the data of 1961-1980, and to measure how much data out of 1981-
1990 would fall into this confidence interval. If most of them fall within the confidence
-4-
!
Lab Report #2 STAT 656 Ying Wang
interval, then we would not have enough evidence to reject the null hypothesis; otherwise,
we would reject the null hypothesis.
See results in figure 4 below. The magenta lines are the upper and lower bounds for the
confidence interval for temperature anomalies in 1961-1980. The blue line is a
segmentation marker at 1980. On its left are the anomalies in 1961-1979 ,while on its
right are the anomalies in 1981-1990. The red line is a linear regression line.
0.5
0.4
0.3
temperature anomaly
0.2
0.1
-0.1
-0.2
1960 1965 1970 1975 1980 1985 1990
year
The regression line of a positive slope (t=4.641, p=7.41e-05) confirms that the overall
trend for the temperature anomaly is upward. The result that 60% of the data from 1981
to 1990 do not fall into the confidence interval suggests that temperature anomalies are
getting more extreme since the 1980s. Suppose human activity is the main cause for the
temperature anomaly, the possible reason is that global economy began to thrive since the
1960s and some 20 years were like a crash cushion. Hence significant global warming
problem does not show up until after those cushion years.
!
Lab Report #2 STAT 656 Ying Wang
bounds for the confidence interval for CO2 emissions in 1961-1980. The blue line is a
segmentation marker at 1980. On its left are the CO2 emissions in 1961-1979 ,while on
its right are the CO2 emissions in 1981-1990. The red line is a linear regression line.
355
350
345
340
335
CO2
330
325
320
315
310
1960 1965 1970 1975 1980 1985 1990
year
The regression line of a positive slope (t=45.10, p<2e-16) confirms that the overall trend
for the CO2 emissions is upward. Nevertheless, none of the data points in 1981-1990 fall
into the 95% confidence interval of the data of 1961-1980, which echos the observation
that the CO2 emissions level is almost monotonically increasing each year in 1961-1990.
This result is also in line with our findings for the temperature anomalies in the sense that
they are all backed by the prosperation of the economy and the industry in general.
Lastly, we apply linear regression and confidence interval to solar radiation. Results
are shown in figure 6 below. The magenta lines are the upper and lower bounds for the
confidence interval for solar radiation in 1961-1980. The blue line is a segmentation
marker at 1980. On its left are the solar radiation in 1961-1979 ,while on its right are the
solar radiation in 1981-1990. The red line is a linear regression line.
-6-
!
Lab Report #2 STAT 656 Ying Wang
4.8
4.6
sun radiation
4.4
4.2
3.8
1960 1965 1970 1975 1980 1985 1990
year
Figure 6 Solar radiation in 1961-1990.
We can see the solar radiation data oscillates around the regression line, and there is no
clear trend of constant increase over the years. Also, the confidence interval includes
almost all the data. However, the regression line is of low reliability in that the p-value is
not small enough to reject the null hypothesis that the slope is zero (t=-0.482, p=0.634).
This also confirms the cyclical pattern of the solar radiation exhibited in figure 2.
To lend some support to the idea that 1981-1990 made things much worse than 1961-
1980, we construct the 95% confidence region of the samples of the year 1961-1980, to
assess how much data out of the samples from 1981-1990 fall outside. Note that each
sample is a 3-element vector, consisting of temperature anomalies, CO2 emissions level
and solar radiation. Therefore, the confidence region is an ellipsoid. As we can see from
figure 7 below, all observations after 1980 are outside of the 95% confidence region.
Although solar output is not changing abnormally, 1981-1990 is still largely inconsistent
with 1961-1980. Is it expected? even if there is NO change after 1980?
-7-
!
Lab Report #2 STAT 656 Ying Wang
Interesting plot
Figure 7 A 95% confidence region of temperature anomalies, CO2 emissions and solar radiation .
Hence these findings suggest if we use a multivariate normal distribution estimated from
observations before 1981, it would only cast very dim light on the description of the
observations after 1980.
Knowing that both temperature anomalies and CO2 emission have been increasing but
solar radiation has not, the correlation matrix only consolidates the idea that CO2
emission is more correlated to temperature anomalies than solar radiation is. The
correlation matrix is shown below:
Table 2 Correlation matrix of temperature anomalies, CO2 level, and solar radiation.
whereas temperature anomalies and CO2 have a high correlation 0.71 and solar radiation
is not strongly correlated with temperature and CO2 (-0.02 and -0.08 respectively).
-8-
!
Lab Report #2 STAT 656 Ying Wang
Then, we regress temperature anomalies against CO2 and solar radiation and draw the
following graphs.
0.5
0.4
0.3
temprature anomaly
0.2
0.1
-0.1
-0.2
315 320 325 330 335 340 345 350 355
CO2
0.5
0.4
0.3
temprature anomaly
0.2
0.1
-0.1
-0.2
3.9 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7
yearly sun radiation
-9-
!
Lab Report #2 STAT 656 Ying Wang
Table 3 Regression results of temperature anomalies against CO2 level, and solar radiation.
As expected, the coefficient of CO2 level is positive and significant, whereas the
coefficient on solar radiation is not statistically different from 0. Although not in the table,
the adjusted R-squared for this multiple regression is 0.46, i.e. CO2 level “explains” about
half of the variations we observe in temperature anomalies. These results clearly depict
that a linear relationship is more probable between temperature and CO2.
4. Conclusion
This study investigated the relationship between temperature anomalies and CO2
emissions/solar radiation. It is found that temperature anomalies and CO2 emissions have
been rising in the period of 1961-1990 and are highly correlated in a linear relationship.
However, solar radiation does not show any departure from its usual cyclic activities
during this period and does not affect temperature anomalies in the way that CO2 does.
In conclusion, in line with other findings, this study discloses a high correlation
between global warming and CO2 emissions. As most of CO2 is from fuel burning,
human activity is probably the major cause for climate change.
-10-
!
Lab Report #2 STAT 656 Ying Wang
5. References
[1] “Global Surface Temperature Anomalies,” last modified August 12, 2010,
http://www.ncdc.noaa.gov/cmb-faq/anomalies.html.
[2] “NASA GISS 1880-2009 Global Temperature,” last modified February 10, 2011,
http://data.giss.nasa.gov/gistemp/graphs/
[3] ”Solar Cycle Variations,” image created by Robert A. Rohde / Global Warming Art,
http://www.globalwarmingart.com/wiki/File:Solar_Cycle_Variations_png
[4] ”Climate Change Attribution,” image created by Robert A. Rohde / Global Warming
Art, http://www.globalwarmingart.com/wiki/File:Climate_Change_Attribution_png
[5] Hegerl, Gabriele C.; et al. (2007). "Understanding and Attributing Climate Change".
Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to
the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. IPCC.
[6] Randel, William J.; Shine, Keith P.; Austin, John; Barnett, John; Claud, Chantal;
Gillett, Nathan P.; Keckhut, Philippe; Langematz, Ulrike et al. (2009). "An update of
observed stratospheric temperature trends". Journal of Geophysical Research 114:
D02107.
-11-
!
Lab Report #2 STAT 656 Ying Wang
library(plotrix)
attach(temperature)
attach(radiation)
year<-c(1961:1990)
a<-matrix(c(1,1))
for (i in 1:30)
{a[i]<-Year[12*(i-1)+1]}
avg_rad<-matrix(c(1,1))
for (i in 1:30)
{avg_rad[i]<-sum(SunRadiation[(12*(i-1)+1):(12*i)])/12}
plot(year,CO2,main="CO2 in 1961-1990")
-12-
!
Lab Report #2 STAT 656 Ying Wang
All<-cbind(Temp,CO2,avg_rad)
m<-c(mean(Temp[1:20]),mean(CO2[1:20]),mean(avg_rad[1:20]))
#covariance matrix
cm<-cov(All[1:20,1:3])
##test if the last 10 years fall into the confidence region of first 20 years
F<-10.7186
n<-20
invcm<-solve(cm)
yes<-matrix(c(1,1))
for (i in 1:10)
{yes[i]<-1}
else
{yes[i]<-0}
sumyes<-sum(yes)/10
##confidence intervals
#temperature
T<-1.7291
-13-
!
Lab Report #2 STAT 656 Ying Wang
for (i in 1:10)
{yes[i]<-1}
else
{yes[i]<-0}
sumTyes<-sum(yes)/10
#CO2
for (i in 1:10)
{yes[i]<-1}
else
{yes[i]<-0}
sumCyes<-sum(yes)/10
#sun radiation
for (i in 1:10)
{yes[i]<-1}
else
{yes[i]<-0}
-14-
!
Lab Report #2 STAT 656 Ying Wang
sumSyes<-sum(yes)/10
reg1<-glm(Temp~CO2)
plot(Temp,CO2)
reg.line(glm(Temp~CO2),col="red")
reg2<-glm(Temp~avg_rad)
# Confidence region
library(rgl)
m.1 <-m[2:4]
ellipse3dCR<-ellipse3d(cm,centre=m.1,subdivide=3,t=sqrt(qchisq(0.95,3)), smooth=F,
which=1:3)
x <- after801[,1]
y <- after801[,2]
z <- after801[,3]
points3d(x,y,z,col='green',size=6,pch=20)
year=1961:1990
T=1.7291
len=T*sqrt(42.7746976)
figure(2)
plot(year,CO2,'-o')
hold on
Y= -2205.778 + 1.285*year
plot(year, Y,'r-','linewidth',2)
-15-
!
Lab Report #2 STAT 656 Ying Wang
hold on
hold off
xlabel('year')
ylabel('CO2')
year=1961:1990
T=1.7291
len=T*sqrt(0.024241776)
figure(2)
plot(year,rad,'-o')
hold on
hold on
Y= 7.242316 -0.001529*year
plot(year, Y,'r-','linewidth',2)
hold off
xlabel('year')
ylabel('sun radiation')
-16-