Stephen A. Miller
1. Introduction
The Southern Oscillation(SO for short), and its warm phase, called El Nino, has caused
concern due to the much-above-normal precipitation amounts sometimes accompanying this
phase of the SO. During the 1982-1983 period(the strongest El Nino ever measured), the
Jackson, Mississippi, area received precipitation totaling 61.38" during the November through
April period, and 28.51" during the December through February period. The Pearl
River at Jackson reached 39.58' on May 25, 1983, due to the rains, becoming the second
highest level on record. It is the comparison of the current El Nino with this record setting
event that has caused the concern.
The Southern Oscillation is an west to east shifting of warmer than normal waters in the
equatorial Pacific Ocean. During the El Nino phase, the warmer than normal waters shift
eastward, nearer to the coastal waters of Central and South America. During the opposite
phase, called La Nina, the warmer than normal waters shift westward to the western reaches
of the Pacific Ocean. It is the accompanying shift in the global scale air circulations that
either enhances or decreases precipitation over the Pacific Ocean and other, globally close
regions, with the results depending upon the phase of the oscillation.
Sir Gilbert Walker(Walker 1923, 1924, 1928, 1932, 1937) noticed the phenomena, in an
attempt to explain, then predict the cycle of the monsoon rains in India, conducted research in
the 1920's and 1930's. He found a correlation existed between the difference of the average
monthly sea level pressures for Darwin, Australia and Pepeete, Tahiti. It is this difference in
average pressure( with some records dating before 1600) which became the first "index" of
the SO, and (with some adjustments over the years) is still in use today.
Barnston and Livezey(Barnston and Livezey, 1987), explored the various global
circulations, identified thirteen separate teleconnection patterns in the Northern Hemisphere.
One of these teleconnections, the North Atlantic Oscillation(NAO), was found to affect the
mean circulation pattern in the Southeast, throughout the year. Little research has been found
by this author on the interaction between the ENSO, and NAO. At a cursory glance, there
appears a correlation between the relative strengths of the SO and NAO, and the weather in
the Southeast. This will require further research, though, to quantify and verify.
It is during the El Nino phase of the SO in which enhanced precipitation affects the
coastal areas of the Gulf of Mexico, among other regions of the US and world. This paper
explores the various stages of the SO and its effects upon average temperature, total
precipitation, and total snowfall amounts for winter(defined as the period December through
February)in Jackson, MS. Its effects upon temperature, precipitation and snowfall for a period
encompassing the six coolest months of the year(the period including the months from
November though April) also are covered, and attempts to expose any relationships to the
local climatology of the Jackson, Mississippi area.
2. Methodology
The period used in this paper encompasses November through April for the years 1896
through early 1997, and uses climatological sources found on station at the Jackson National
Weather Service Office. Several sources were used to define a period as an El Nino, La Nina,
or neutral (average) year. The first was a listing of El Nino and La Nina years on the
National Oceanographic and Atmospheric Administration (NOAA) Climate Prediction
Center's(CPC) El Nino web page at web address . The second was
taken from University of Massachusetts' El Nino web site from a thesis written by Kimberly
Amaral www.umassd.edu/Public/People/KAmaral/Thesis/ElNinoYears.html>. The third was a
paper by Ropelowski and Jones, 1987. There were some slight differences among the three
sources. Also, the CPC site did not have available the years before 1950. For each year, when
two of the sources agreed upon a year being an El Nino, La Nina or neutral year, that year
was marked as such. No attempt was made to relate the strength of the El Nino or La Nina to
the departures.
The statistics in this paper were calculated using the Minitab statistical software package
and several spreadsheets in Microsoft Excel. Parameters calculated were mean, standard
deviation, maximum, minimum, correlation coefficients, and probabilities using tests for
normal and non-normal distributions. The hypothesis used for the testing was that the El Nino
and La Nina years do not vary significantly from average.
The test used for non-normal distributions was the Student-t test, due to sample sizes for
El Nino(28) and La Nina(17) years being less than the required 30 to assume a normal
distribution. A 95% confidence interval was used for calculating probabilities. For the
correlation coefficients in the following sections, according to Panofksy and Brier, 1958, a
coefficient great than 2.6 times the standard deviation, or 0.2613, is considered significant for
this sample size. All testing compared the sample statistics with those of the total population.
For this study, this includes all the years of recorded data at Jackson.
3. Discussion and Results
A. Total Precipitation
After calculating the statistics, it was found that for precipitation, both total liquid
equivalent and snowfall, little significant correlation exists between the deviation from average
of the precipitation amount and the type of year. For example, although the average for the
winter precipitation amount is higher for El Nino years and lower for La Nina years, testing
showed a high probability that the sample means are not statistically different from the mean
of the total population. Tables 1 and 2 show the results of tests performed on the winter
precipitation totals. After comparison between deviation from average and type of year,
correlation coefficients of -0.0228 for El Nino, 0.1013 for neutral, -0.1073 for La Nina years,
and -0.0329 for all years were calculated for the winter period. Both are very small
correlations below the 0.2613 cutoff for a significant correlation, suggesting that the relation
between precipitation departure and type of year is from random fluctuations. In Table 2, and
following tables displaying test results from the "Z" and Student-t tests, "Z" value is the
calculated value from the Z normalcy test with "Z - P Value" the associated probability(
percentage divided by 100). The "T - Value" and
Table 1: Winter Precipitation Statistics
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 15.57"
| 4.98"
| 9.12"
| 28.51" |
| Neutral
| 15.27"
| 5.26"
| 6.80"
| 30.19" |
| La Nina
| 15.09"
| 4.22"
| 8.88"
| 22.52" |
Table 2: Winter Precipitation - Z and Student T Test Results
|
| Z Value
| Z - P Value
| T Value
| T - P Value |
| El Nino
| 0.26
| 0.79
| 0.26
| 0.80 |
| Neutral
| -.08
| 0.94
| -.07
| 0.94 |
| La Nina
| -.19
| 0.85
| -0.23
| 0.82 |
"T - P Value" are the same values from the Student-t test. One item that was in evidence for
La Nina years is a small number of above normal precipitation amounts, as shown by Table 1,
and by Figures 1 and 2. Another observation is the smaller
variability of the data for La Nina years, in evidence by the smaller standard deviation. It is
also interesting to note that the two highest winter precipitation totals for 1896 to 1997 are for
neutral, rather than El Nino years.
Expanding the time period to November through April, the statistics show a slightly
different pattern. Average total precipitation, for both El Nino and La Nina years, is less than
for neutral years. With El Nino years, the difference between neutral and El Nino average
precipitation is not significantly different. Figure 3, one can see that the
precipitation total for the 1982-1983 period has become the largest. Still, amounts for El Nino
years are distributed across the entire range of values. With La Nina years, the difference in
average precipitation is almost significant at the level tested for. If the confidence level is
reduced, one could say, that for the November through April period, La Nina years are drier
than average. This is with an 88% confidence level. Also, precipitation totals during La Nina
years are less variable than either El Nino or neutral years, as exhibited by the smaller
standard deviation for these years. Tables 3 and 4 list the statistics calculated for the
November through April period. Figure 4 also helps illustrate these points.
Table 3: November-April Precipitation Statistics
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 29.37"
| 9.51"
| 17.31"
| 61.38" |
| Neutral
| 31.12"
| 8.34"
| 11.49"
| 53.17" |
| La Nina
| 29.96"
| 5.36"
| 19.21"
| 40.82" |
Table 4: November-April Precipitation - Z and Student T Test Results
|
| Z Value
| Z - P Value
| T Value
| T - P Value |
| El Nino
| -0.47
| 0.64
| -0.41
| 0.69 |
| Neutral
| 0.92
| 0.36
| 0.91
| 0.37 |
| La Nina
| -1.07
| 0.29
| -1.65
| 0.12 |
After calculating the correlation coefficients for this period, small numbers were the result;
-0.1144 for El Nino years, 0.1926 for neutral, -0.1190 for La Nina years, and -0.0291 for all
observations compared to type of year, again suggesting the results are from random
fluctuations for El Nino and neutral years.
B. Snowfall
Considering snowfall in Jackson, the key item to remember is that measurable snowfall
has occurred only during 40% of the years or less(37% during the winter months, 40% for the
November through April period), regardless of the type of year. Table 5 shows these
percentages listed according to period and type of year, and with Figures 5 and
6 helping to illustrate. Figures 7 and 8 show the
distribution of measurable snowfall totals. As one can see from these numbers and figures,
measurable snowfall totals are slightly less during El Nino years than in neutral years, more
so during La Nina years. Table 6 includes the basic statistics for snowfall for the winter and
November through April periods for all years. One can see from this table, La Nina years
again have less snow than neutral or El Nino years, and are less variable in the total amounts
for the period when it does occur. Correlation coefficients calculated are summarized in Table
8. Again, the numbers are small and seem to be the result of randomness.
Table 5: Percentages of El Nino, Neutral, and La Nina years with Measurable Snowfalls
|
| Winter
| 6 Cool Months |
| El Nino
| 32%
| 36% |
| Neutral
| 42%
| 44% |
| La Nina
| 24%
| 29% |
Table 6: Winter and November Through April Snowfall Statistics
Winter
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 1.36"
| 2.93"
| 0.00"
| 11.6" |
| Neutral
| 1.44"
| 2.74"
| 0.00"
| 11.7" |
| La Nina
| 0.74"
| 1.74"
| 0.00"
| 5.8" |
November through April
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 1.40"
| 2.91"
| 0.00"
| 11.6" |
| Neutral
| 1.44"
| 2.79"
| 0.00"
| 11.7" |
| La Nina
| 0.77"
| 1.78"
| 0.00"
| 6.0" |
Calculation of the test statistics, shown in Table 7, shows El Nino years are not significantly
differ from neutral years in total snowfall amounts for winter and the November through
April period. La Nina years also do not significantly differ from average at a 95% confidence
level. However, one could say, with a 78% confidence interval, that La Nina years result in
less snow than average. Again, though, correlations are low between snowfall amount for the
November through April period and type of year for La Nina, El Nino, and neutral, and all
years inclusive, suggesting this is due to randomness. The results are again summarized in
Table 8.
Table 7: November - April and Winter Snowfall Departures - Z
and Student T Test Results
Winter
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 0.26"
| 0.79"
| 0.23"
| 0.82" |
| Neutral
| 0.59"
| 0.55"
| 0.56"
| 0.58" |
| La Nina
| -0.79"
| 0.43"
| -1.16"
| 0.26" |
November - April
|
| Z Value
| Z - P Value
| T Value
| T - P Value |
| El Nino
| 0.17
| 0.86
| 0.16
| 0.88 |
| Neutral
| 0.35
| 0.73
| 0.33
| 0.74 |
| La Nina
| -0.85
| 0.39
| -0.85
| 0.22 |
Table 8:Winter and November - April Snowfall Correlations
|
| Winter
| Novmeber - April |
| El Nino
| -0.0589
| -0.0202 |
| Neutral
| 0.0870
| 0.1141 |
| La Nina
| -0.0451
| -0.0586 |
| All Years
| -0.0702
| -0.0669 |
C. Average Temperature
In this study, average temperature showed the most correlation between departure from
average and type of year; warm for La Nina, cooler for El Nino. For winter, the correlation
coefficient was 0.2890, and for November through April, it was 0.3220. Both are modest, yet
significant, positive correlations. The coefficients are summarized in Table 13. One can also
see, the correlation for El Nino years for the November through April period is above the
0.2613 significance level, with the significance level taken as an absolute number.
Tables 9 and 10 summarize the basic mean temperature statistics with Tables 11 and 12
summarizing the test results. The average of all El Nino winters is 1.4 degrees below the
population average of 48.6 degrees, and 0.9 degrees below the November through April
average of 59.4 degrees. La Nina years were 1.4 degrees above for winter and 1.3 degrees for
November through April. Figures 9 and 10
help to illustrate these statistics.
After calculating the statistics for the test of means, there was a
point two to four percent chance(depending on testing method and of the data) that the data
for El Nino years could be part of the average. Thus, one can state that El Nino years are
warmer than the average, with 96% to 98% confidence, for winter and November through
April periods. Figures 11 and 12 help
to illustrate this; one can see that El Nino years tend to cluster at the cooler end of the
figure with La Nina years clustering near warmer end.
With La Nina, period of the data is important. With winter data, an 89% to
90% confidence level of La Nina years being warmer than average was calculated. For the
November through April period, the confidence level rises to over 98%.
Table 9 :Winter Temperature Statistics(numbers in degrees F)
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 47.2
| 2.818
| 41.2
| 54.1 |
| Neutral
| 49.0
| 3.318
| 43.4
| 57.4 |
| La Nina
| 50.0
| 3.198
| 41.5
| 54.7 |
Table 10 :November - April Temperature Statistics(numbers in degrees F)
|
| Mean
| Standard Deviation
| Minimum
| Maximum |
| El Nino
| 52.9
| 1.457
| 49.8
| 55.2 |
| Neutral
| 53.9
| 2.559
| 45.3
| 59.4 |
| La Nina
| 55.1
| 2.023
| 50.6
| 58.2 |
Table 11: Winter Average Temperature - Z and Student T Test Results
|
| Z Value
| Z - P Value
| T Value
| T - P Value |
| El Nino
| -2.28
| 0.023
| -2.64
| 0.013 |
| Neutral
| 0.72
| 0.47
| 0.71
| 0.48 |
| La Nina
| 1.65
| 0.10
| 1.69
| 0.11 |
Table 12: April - November Average Temperature - Z and Student T Test Results
|
| Z Value
| Z - P Value
| T Value
| T - P Value |
| El Nino
| -2.10
| 0.036
| -3.32
| 0.0026 |
| Neutral
| 0.19
| 0.85
| 0.17
| 0.87 |
| La Nina
| 2.37
| 0.018
| 2.70
| 0.16 |
November - April and Winter Temperature Correlations
|
| Winter
| Novmeber - April |
| El Nino
| -0.1936
| -0.2917 |
| Neutral
| -0.0017
| 0.1412 |
| La Nina
| 0.2332
| 0.1625 |
| All Years
| 0.2890
| 0.3220 |
4. Summary
As the preceding section illustrated, the effects of the El Nino Phase of the Southern
Oscillation upon precipitation in the Jackson area are erratic, thus hard to predict. It is this
variability over the entire range of observations that keeps the average for El Nino years from
deviating significantly from the average of all years. With temperature, what is considered the
"usual effects" of El Nino occur. El Nino years result in cooler than average temperatures for
the winter and November through April periods in Jackson with a 96% to 98% confidence
interval.
The La Nina phase, on the other hand, tends to be less variable with respect to
precipitation, thus easier to quantify and predict. This holds true for melted liquid
precipitation, and snowfall. For liquid and solid precipitation, the tested 95% confidence level
wasn't achieved, but not by much. Confidence levels ranged for all precipitation ranged from
80% to almost 90%.
With respect to temperature, again, what is considered "usual" is confirmed. La Nina
years tend to result in warmer than usual winters. This is also extended to the November
through April period.
|
weather.gov 
