Sei sulla pagina 1di 8

404 JOURNAL OF CLIMATE VOLUME 11

ENSO Influences on Agriculture in the Southeastern United States*


JAMES W. HANSEN
Agricultural and Biological Engineering, University of Florida, Gainesville, Florida

ALAN W. HODGES
Food and Resource Economics, University of Florida, Gainesville, Florida

JAMES W. JONES
Agricultural and Biological Engineering, University of Florida, Gainesville, Florida
(Manuscript received 19 May 1997, in final form 11 July 1997)

ABSTRACT
The influence of El Niño–Southern Oscillation (ENSO) on crop production in the southeastern United States
was studied to identify crops that are vulnerable to ENSO-related weather variability and therefore likely to
benefit from application of ENSO-based climate forecasts. The historical (1960–95) response of total value and
its components (yield, area harvested, and price) to ENSO phases and quarterly sea surface temperature anomalies
(SST) in the eastern equatorial Pacific was analyzed for six crops (peanut, tomato, cotton, tobacco, corn, and
soybean) in four states (Alabama, Florida, Georgia, and South Carolina). ENSO phase significantly influenced
corn and tobacco yields, the areas of soybean and cotton harvested, and the values of corn, soybean, peanut,
and tobacco. ENSO phases explained an average shift of $212 million or 25.9% of the value of corn. Canonical
correlation analysis identified significant responses of corn, soybean, and cotton yields, and peanut value to SST
across the region; and of peanut and tobacco yields, and tomato and soybean values in particular states.

1. Introduction 1996). Handler showed that historical corn yields were


The uncertainty associated with weather variability is correlated (|R| . 0.36) with Pacific SSTs for the pre-
a perpetual predicament for agriculture. Crop manage- ceding winter months in several midwestern and south-
ment that is optimal in one year may contribute to crop eastern states and with spring and fall SSTs in Florida.
losses, overproduction, or environmental pollution in Besides crop yields, researchers have shown that wheat
another year with different weather. Evidence is growing rust in northern China and the northern Midwest and
that El Niño–Southern Oscillation (ENSO) influences Pacific Northwest in the United States (Scherm and
crop production through its influence on weather pat- Yang 1995), soybean futures prices in the United States
terns in some regions. El Niño events were associated (Keppenne 1995), and the gross value of Australian
with low grain yields in south Asia and Australia and crops (Nicholls 1985) were significantly associated with
high grain yields in the North American prairies (Gar- measures of ENSO.
nett and Khandekar 1992). ENSO-related SSTs were Several research groups are working to develop sta-
strongly related to rainfall and corn yields in Zimbabwe, tistical and numerical models to predict ENSO-related
where SST a full year before planting explained 57% SSTs (Barnston et al. 1994) and their effects on regional
of the variability in yields (Cane et al. 1994). ENSO weather patterns (e.g., Barnett et al. 1994). In many
influences corn yields in the midwestern and south- regions, ENSO-based predictions can now provide
eastern United States (Handler 1990; Carlson et al. weather information with useful skill at lead times of
several months. Researchers expect forecast accuracy to
increase as understanding of the ENSO phenomenon
*Florida Agricultural Experiment Station Journal Series Number
grows and models and data monitoring improve. Sea-
R-05784. sonal forecasts create the possibility of tailoring crop
management to anticipated weather conditions either to
take advantage of favorable conditions or to reduce the
Corresponding author address: Dr. James W. Hansen, Agricultural
and Biological Engineering, University of Florida, P.O. Box 110570,
effects of adverse conditions. At the level of regional
Gainesville, FL 32611-0570. or national agricultural policy, the ability to forecast
E-mail: jhansen@agen.ufl.edu crop production has implications for food staple storage

q 1998 American Meteorological Society


MARCH 1998 HANSEN ET AL. 405

fluence on crops in the southeastern United States is to


identify cropping enterprises in the region that are vul-
nerable to ENSO and to rank them in terms of economic
impact of ENSO. This will provide an indication of
which enterprises are most likely to benefit from ENSO
forecasts. The authors hypothesize that ENSO influenc-
es the value of crop production both directly through
its influence on weather patterns that affect yields, and
indirectly through producer and market response to pre-
vious and expected yields in the region and in other
important production regions.

2. Approach
a. Data
1) CROPS
Historical records of yields, area harvested, price, and
total value of production of six economically important
crops (cotton, corn, peanut, soybean, tomato, and to-
bacco) in a four-state (Florida, Alabama, Georgia, and
South Carolina) region (Table 1) were examined. His-
torical state-level crop production, areas planted and
FIG. 1. Raw and smoothed time series (a) and resulting anomalies harvested, and price statistics are from the United States
(b) of Florida corn yields, 1960–95. Department of Agriculture (USDA) National Agricul-
tural Statistics Service (NASS). Data cover 1960–95
except for tomato areas, prices and values, and all to-
capacity, trade agreements to offset local shortages in bacco data, which were not available for 1995 at the
food supplies, and planning to provide assistance to ad- time of analysis. Alabama stopped reporting tobacco
versely affected producers and populations. data in 1981. Mean yields for each state were calculated
In the southeastern U.S. coastal plain, El Niño years from total production and areas harvested. Where crop
tend to be cool and La Niña years tend to be warm values were reported for individual states, the average
between October and April (Ropelewski and Halpert price each year was calculated as the total value divided
1986, 1996; Kiladis and Diaz 1989; Sittel 1994a). Al- by the total production of the four-state region. Other-
though the influence on rainfall is spatially less consis- wise, national average prices were used, and total value
tent, El Niño years tend to be wet and La Niña years was calculated from national prices and reported pro-
dry during these months. Although the ENSO signal in duction for each state. Prices were adjusted to a 1995
the region is strongest in the fall and winter months, basis using annual values of the U.S. urban consumer
some evidence exists that La Niña summers tend to be price index.
slightly wetter than normal (Sittel 1994a,b). ENSO in-
fluence on rainfall has a lagged effect on streamflows
2) ENSO
in the region, as Zorn and Waylen (1996) showed for
the Santa Fe River in northern Florida. This study analyzed responses of crop variables to
The objective of this exploratory study of ENSO in- both categorical and continuous measures of ENSO. Sea

TABLE 1. Recent crop production and value by state a, mean of 1991–95.

Production (Gg) Value (million $)


Crop AL FL GA SC Total AL FL GA SC Total
Peanut 239 98 790 14 1141 169 64 544 10 787
Tomato 22 796 47 50 915 11 653 35 42 741
Cotton 122 21 250 70 463 180 32 383 106 701
Tobaccoc 0b 8 41 50 99 0b 31 161 201 393
Corn 507 163 1280 587 2537 51 16 128 60 255
Soybean 203 32 361 351 947 47 7 82 81 217
a
AL 5 Alabama, FL 5 Florida, GA 5 Georgia, SC 5 South Carolina.
b
Not reported after 1980.
c
Mean of 1991–94.
406 JOURNAL OF CLIMATE VOLUME 11

TABLE 2. Significancea of response of anomalies of crop variables to ENSO phase.

Yield Area harvested Price Value


Effect Lag 0 Lag 1 Lag 0 Lag 1b Lag 2 Lag 0b Lag 1b Lag 2b Lag 0 Lag 1 Lag 2
ENSO phase *** *** ** n.s. n.s. n.s. n.s. n.s. ** *** ***
Phase 3 state n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s.
Phase 3 crop * n.s. ** n.s. *** n.s. n.s. n.s. n.s. ** n.s.
Peanut n.s. n.s. n.s. n.s. n.s. * n.s.
Tomato n.s. n.s. n.s. n.s. n.s. n.s. n.s.
Cotton n.s. n.s. n.s. ** n.s. n.s. n.s.
Tobacco n.s. ** * n.s. n.s. * n.s.
Corn ** * n.s. n.s. * * ***
Soybean n.s. n.s. *** ** n.s. * **
a
Asterisks indicate significance at 5% (*), 1% (**), and 0.1% (***) level; n.s. indicates not significant.
b
Individual crops were not analyzed because ENSO phase and phase 3 crop effects were not significant.

surface temperature anomalies (SST, 8C) averaged over includes October–December. The period from October
the region of the eastern equatorial Pacific known as 1957 to September 1995 includes 10 El Niño (beginning
NINO3 (58N to 58S, 1508 to 908W) are commonly used 1957, 1963, 1965, 1969, 1972, 1976, 1982, 1986, 1987,
as a continuous measure of ENSO strength. Monthly and 1991) and 7 La Niña events (beginning 1964, 1967,
SST values obtained from the National Oceanic and 1970, 1971, 1973, 1975, and 1988).
Atmospheric Administration (NOAA) Climate Predic-
tion Center were aggregated into quarterly means.
b. Data preparation
ENSO phases (El Niño, neutral and La Niña) serve
as a categorical measure of ENSO activity. This study The analyses required us to separate each historical
used the COAPS (Center for Ocean–Atmospheric Pre- yield series into a trend that reflects changes in man-
diction Studies) classification based on monthly SST agement and anomalies about the trend due to weather
anomalies in the region from 1508 to 908W and 48N to variability. Crop management changes in response to
48S, smoothed into 5-month running means (Sittel price fluctuations and the availability of new technology
1994b). A year (October–September) is classified as El in a way that is difficult to fit to a simple regression
Niño (La Niña) if SST anomalies are at least 10.58C function; there is no theoretical basis for selecting one
(#20.58C for La Niña) for at least 6 consecutive months functional form over another. A harmonic smoothing
and if this 6-month period starts before October and technique was therefore applied to separate higher-fre-

FIG. 2. Spectral densities of raw ( ) and smoothed (– – –) data and anomalies (········) of
Florida corn yields, 1960–95, smoothed with a five-point Hamming window.
MARCH 1998 HANSEN ET AL. 407

of anomalies (y, i.e., observed minus smoothed values)


about the trend, where y is the mean and s is the standard
deviation of the series of anomalies.

c. Analyses
1) RESPONSE TO ENSO PHASE

For each crop variable (yield, area harvested, price


and value), the hypothesis that ENSO phase influences
anomalies about the trend was tested by analysis of
variance (ANOVA) (Steel and Torrie 1980). Stratifi-
cation of standardized anomalies of yields, areas, or
values among the six crops, four states, and three ENSO
phases produced a three-way (ENSO phase, crop and
state) factorial sampling design, with generally 36 yr
observations per time series for each crop and state
combination. Although the design permits tests of the
influence of all three factors and their interactions on
each crop variable, the use of standardized anomalies
FIG. 3. Box plots showing 0, 25, 50, 75, and 100th percentiles and
means (solid line) of yield anomalies with a significant response to
gave each crop and state a mean value of zero; crop
ENSO phase, pooled across states, weighted by areas harvested in and state main effects and their interaction have no
each state, 1960–95. Means of ENSO phases with no common letters meaning. Therefore, only the ENSO phase effect and
below their respective box-and-whiskers are significantly different (P its interactions with crop and state were considered.
5 0.05). Crop prices were combined across states, resulting in a
two-way (ENSO phase and crop) factorial design. When
ANOVA indicated a significant ENSO phase effect, the
quency anomalies from lower-frequency trends. The response of actual anomalies for each crop to ENSO
technique (Press et al. 1989) removes any linear trend phase was analyzed by ANOVA, using pooled within-
from a data series, transforms the data to the frequency sample and ENSO phase-by-state mean squared devi-
domain using Fourier analysis, applies a low-pass filter ations as the denominator for F-tests. Crop yields were
covering a specified period, transforms back to the time weighted by areas harvested in each state. To identify
domain, then reinserts the linear trend. Spectral analysis which ENSO phases differed in their effect, Duncan’s
shows that rainfall in various regions of Florida oscil- multiple range test was applied for each crop that
lates with periods from 5.2 to 6.3 yr (Hanson and Maul showed a significant response to ENSO phase.
1991) and that El Niño events (1525–1988) recurred
with a fundamental period of 6.7 yr (Hanson et al. 1989).
Therefore, a filtering period of 7.0 yr (frequency of 2) RESPONSE TO SST
0.143 yr21 ) was used to smooth all crop data. Figure 1
For yield and value, we considered SST values for
shows raw and smoothed Florida corn yields for 1960–
the 2 yr (eight quarters) ending with the winter (De-
95, and anomalies about the smoothed trend. Spectral
cember–February) before the reported harvest (i.e., Feb-
densities (Fig. 2) show that smoothing effectively sep-
ruary of the production calendar year). For prices and
arated the series into a lower-frequency trend and high-
areas harvested, we considered the 2 yr ending February
er-frequency signals. Subsequent analyses were done on
of the calendar year preceding the harvest, based on the
standardized values,
assumption that the impact of weather variability on
z 5 (y 2 y)/s, (1) yields could influence prices the year after harvest.

TABLE 3. Canonical correlationa between crop yield anomalies and quarterly SST.

Crop Alabama Florida Georgia South Carolina Region


Peanut 0.585 n.s. 0.566 n.s. 0.651* 0.636* 0.674 n.s.
Tomato 0.451 n.s. 0.436 n.s. 0.472 n.s. 0.563 n.s. 0.633 n.s.
Cotton 0.519 n.s. 0.556 n.s. 0.441 n.s. 0.597 n.s. 0.739*
Tobacco 0.532 n.s. 0.664* 0.535 n.s. 0.487 n.s. 0.670 n.s.
Corn 0.569 n.s. 0.633* 0.497 n.s. 0.558 n.s. 0.760*
Soybean 0.594 n.s. 0.468 n.s. 0.487 n.s. 0.482 n.s. 0.799*
All 0.741 n.s. 0.871** 0.772 n.s. 0.822* 0.984*
a
Asterisks indicate significance at 5% (*), 1% (**), and 0.1% (***) level; n.s. indicates not significant.
408 JOURNAL OF CLIMATE VOLUME 11

TABLE 4. Effect of ENSO phase on mean annual crop value anom-


alies, total for Alabama, Florida, Georgia, and South Carolina. Value
is based on prices deflated to December 1995 basis.

Phase(s) Change in value


Lag Low High
Crop (yr) value value Million $ Percent*
Corn 2 neutral Niña 212.0 25.9
Soybean 2 neutral Niña 132.6 18.4
Peanut 1 Niña, neutral Niño 86.3 10.0
Tobacco 1 Niña, neutral Niño 56.5 7.4

* Percent of the long-term mean.

groups. The strength of the linear association between


the two groups is expressed by the canonical correlation
coefficient (R) defined as the linear correlation between
the canonical variates resulting from the first transfor-
mation. The canonical R thus expresses the maximum
association that linear combinations of variables in each
set can account for. A x 2 test is used to test the signif-
icance of the canonical R (Stevens 1992). When com-
paring a single predictand variable with a group of pre-
dictor variables, the canonical R is equivalent to the
multiple R returned by multiple linear regression. For
each of the four crop variables (yield, area, price, and
value), the linear relationship between anomalies for the
set of combinations of crop specie and state and quar-
FIG. 4. Box plots showing 0, 25, 50, 75, and 100th percentiles and terly SST series was analyzed by CCA. If the overall
means (solid line) of harvested area anomalies with a significant relationship was significant for one of the crop variables,
response to ENSO phase, summed across states, 1960–95. Means of
ENSO phases with no common letters below their respective box- the correlation of each crop, state, and crop–state com-
and-whiskers are significantly different (P 5 0.05). bination series with the set of lagged quarterly SST
series was then analyzed. Multiple linear regression was
used to confirm the multiple correlation coefficients and
Analysis of yields, areas harvested and value con- significance tests returned by CCA.
sisted of comparing six crops in four states, or 24 crop
series, with eight quarterly SST series, yielding 192
comparisons. Critical values for statistical significance 3. Results
of Pearson’s coefficient of linear correlation for com- a. Yields
paring an individual pair of time series are not valid
when comparing many pairs of variables due to the Crop yields showed a significant response to ENSO
increasing probability of spurious correlations with an phase in both the current and the previous year (Table
increasing number of comparisons. Because of the mul- 2) and to its interaction with crop specie in the current
tiplicity problem that results from many comparisons, year. The lack of significant interaction with state sug-
confidence in a relationship between a particular crop gests that crop response to ENSO phase was consistent
and a particular SST time series (individual or com- across the region. Crop yields tended to be higher than
parisonwise significance) depends on first establishing the trend in La Niña years and lower than the trend in
the relationship between the set of crops and the set of years immediately following La Niña events (i.e., lag
SST series (field or experimentwise significance) (Liv- 5 1 yr). ANOVA of individual yield anomalies pooled
ezey and Chen 1983). across states showed significant response to ENSO
Canonical correlation analysis (CCA) is used to ad- phase only for corn and tobacco (Table 2, Fig. 3). Mean
dress the multiplicity problem in this study. CCA mea- yield anomalies of corn were 499 kg ha21 , or 13.9% of
sures the linear association between two groups of vari- the long-term average, higher in La Niña than in neutral
ables (Nicholls 1987; Stevens 1992). The procedure it- and El Niño years, and 426 kg ha21 (11.8%) lower in
eratively transforms the data to find the orthogonal lin- the years following La Niña events. ENSO phase had
ear combinations (canonical variates) of variables in a smaller effect on tobacco yields (84.2 kg ha21 , or
each group that account for the maximum possible as- 3.6%, higher in years following El Niño events).
sociation between the two groups of variables, using CCA showed a significant overall influence of SST
eigenanalysis on the intercorrelation matrix for the two on crop yield anomalies in the region (Table 3). These
MARCH 1998 HANSEN ET AL. 409

FIG. 5. Box plots showing 0, 25, 50, 75, and 100th percentiles and means (solid line) of crop
value anomalies with a significant response to ENSO phase, summed across states, 1960–95.
Means of ENSO phases with no common letters below their respective box-and-whiskers are
significantly different (P 5 0.05).

results identified ENSO signals in soybean and cotton cotton, and peanut harvested (Table 2). Anomalies of
yields that ANOVA did not identify, but did not confirm areas of soybean harvested averaged 96 123 ha (8.3%)
the ENSO signal in tobacco yields. The results suggest higher in neutral than in La Niña years (Fig. 4). ENSO
that SST had a strong influence on yields of the six phase also accounted for substantial shifts in areas har-
crops in Florida (R 5 0.871) and a weaker influence in vested for tobacco (2777 ha, or 5.3% higher in neutral
South Carolina (R 5 0.822). than La Niña years) and cotton (40 861 ha, or 8.7%
higher in years following neutral than following El Niño
phases). Because CCA did not show a significant overall
b. Prices
relationship between harvested area anomalies and quar-
The results of ANOVA (Table 2) and CCA did not terly SST series, relationships of individual crops and
support the hypothesis that ENSO influences prices of states to SST were not analyzed.
the six crops studied. Therefore, price response to SST
was not analyzed for individual crops.
d. Total value
Crop value is the product of yield, area harvested,
c. Areas harvested
and price. ANOVA results suggest that ENSO phase
ANOVA showed that the ENSO phase 2 yr prior to influenced the value of all crops except tomato and cot-
harvest influenced anomalies of the areas of soybean, ton (Table 2, Fig. 5). The rainfed grain crops (corn,
410 JOURNAL OF CLIMATE VOLUME 11

TABLE 5. Canonical correlationa between crop value anomalies and quarterly SST.

Crop Alabama Florida Georgia South Carolina Region


Peanut 0.612 n.s. 0.718** 0.712** 0.556 n.s. 0.741**
Tomato 0.332 n.s. 0.679* 0.423 n.s. 0.457 n.s. 0.687 n.s.
Cotton 0.524 n.s. 0.468 n.s. 0.541 n.s. 0.379 n.s. 0.673 n.s.
Tobacco 0.677 n.s. 0.608 n.s. 0.554 n.s. 0.508 n.s. 0.646 n.s.
Corn 0.370 n.s. 0.468 n.s. 0.486 n.s. 0.549 n.s. 0.701 n.s.
Soybean 0.463 n.s. 0.679* 0.537 n.s. 0.570 n.s. 0.713 n.s.
All 0.713 n.s. 0.874*** 0.791 n.s. 0.678 n.s. 0.992***
a
Asterisks indicate significance at 5% (*), 1% (**), and 0.1% (***) level; n.s. indicates not significant.

soybean, and peanut) showed the greatest response to prices. This would seem to contradict the results of Kep-
ENSO phase (Table 4). penne (1995) who presented evidence of a strong re-
Canonical correlation analysis showed a highly sig- lationship between monthly SOI (Southern Oscillation
nificant overall response to SST (Table 5). Peanut was index) and U.S. soybean futures prices, both prefiltered
the only crop that showed a significant response to SST using singular spectrum analysis (an adaptation of prin-
for the whole region. The significant response of Florida cipal component analysis applied to time-series data).
crop value to SST suggests that, for the six crops stud- The price (futures versus farmgate) and ENSO (SOI
ied, it is the most vulnerable of the four states to ENSO- versus SST and phases) data, and the methods used were
related climate variability. different than in this study. However, ANOVA would
indicate a significant 2-yr lag price response to ENSO
phase if only soybean prices were considered, with the
4. Discussion
average price $17.15 Mg21 higher in the second year
Results of this study support the hypothesis that following La Niña than neutral or El Niño years (Fig.
ENSO influences the total value of crops in the south- 7). This illustrates the multiplicity problem that arises
eastern Unites States. The influence of ENSO on yields when examining response to ENSO for several crops
can be explained more easily than its influence on areas across several locations. In this study, appropriate mea-
harvested. Corn yields tended to be high in La Niña sures to avoid spurious results for the set of crop prices
years and low in the years that follow La Niñas (Fig. resulted in a loss of power to identify an ENSO signal
3). We attribute this largely to ENSO influence on June in a particular crop price.
rainfall. Corn yield is most susceptible to water stress One possible explanation for the significant response
at tasseling, which occurs in mid-June through most of of areas of soybean harvested to ENSO phase relates
the study region. June rain tends to be high in La Niña to the fishing industry off the southwest coast of South
years (Fig. 6a) and in the years that follow El Niño America. Abnormally high sea surface temperatures as-
events (Fig. 6c) in the main crop-growing region of the sociated with El Niño events lead to reduced harvests
Georgia coastal plain. The high yields following neutral of anchovy and tuna, thus increasing the demand and
years relative to years following La Niñas (Fig. 3) may price for substitutes for fish protein for livestock feed
be due to greater residual soil moisture from relatively (Philander 1990; Keppenne 1995). The increased soy-
high May rainfall (Fig. 6b). bean area observed 2 yr after El Niño events (Fig. 4)
Results of this study did not support the hypothesis may be a lagged response to the resulting increase in
that ENSO influences crop values by influencing their demand.

FIG. 6. Box plots showing 0, 25, 50, 75, and 100th percentiles and means (solid line) of (a)
June rainfall by current ENSO phase, (b) May rainfall by preceding ENSO phase, and (c) June
rainfall by preceding year ENSO phase, Tifton, Georgia, 1960–95.
MARCH 1998 HANSEN ET AL. 411

REFERENCES

Barnett, T. P., and Coauthors, 1994: Forecasting global ENSO-related


climate anomalies. Tellus, 46A, 381–397.
Barnston, A. G., and Coauthors, 1994: Long-lead seasonal forecasts—
Where do we stand? Bull. Amer. Meteor. Soc., 75, 2097–2114.
Cane, M. A., G. Eshel, and R. W. Buckland, 1994: Forecasting Zim-
babwean maize yield using eastern equatorial Pacific sea surface
temperature. Nature, 370, 204–205.
Carlson, R. E., D. P. Todey, and S. E. Taylor, 1996: Midwestern corn
yield and weather in relation to extremes of the southern oscil-
lation. J. Prod. Agric. 9, 347–352.
Cherry, S., 1996: Singular value decomposition analysis and canon-
ical correlation analysis. J. Climate, 9, 2003–2009.
Garnett, E. R., and M. L. Khandekar, 1992: The impact of large-scale
atmospheric circulations and anomalies on Indian monsoon
droughts and floods and on world grain yields—A statistical
analysis. Agric. For. Meteor., 61, 113–128.
Handler, P., 1990: USA corn yields, the El Niño and agricultural
drought: 1867–1988. Int. J. Climatol., 10, 819–828.
FIG. 7. Box plot showing 0, 25, 50, 75, and 100th percentiles and
Hanson, K., and G. A. Maul, 1991: Florida precipitation and the
means (solid line) of U.S. soybean price anomalies by ENSO phase
Pacific El Niño, 1985–1989. Fla. Sci., 54, 160–168.
2 yr prior, 1960–95.
, G. W. Brier, and G. A. Maul, 1989: Evidence of significant
nonrandom behavior in the recurrence of strong El Niño between
1525 and 1988. Geophys. Res. Lett., 16, 1181–1184.
Keppenne, C. L., 1995: An ENSO signal in soybean futures prices.
Although CCA generally confirmed the results of J. Climate, 8, 1685–1689.
ANOVA, the two methods identified ENSO responses Kiladis, G. N., and H. F. Diaz, 1989: Global climatic anomalies as-
for different sets of crops. The two methods consider sociated with extremes in the Southern Oscillation. J. Climate,
2, 1069–1090.
different information. CCA considers the magnitude and Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance
timing of SST anomalies. Much of that information is and its determination by Monte Carlo techniques. Mon. Wea.
lost when converting SST into ENSO phases, suggesting Rev., 111, 46–59.
that ANOVA might be less sensitive to real ENSO ef- Nicholls, N., 1985: Impact of the Southern Oscillation on Australian
crops. J. Climatol., 5, 553–560.
fects. On the other hand, CCA assumes a linear rela- , 1987: The use of canonical correlations to study teleconnec-
tionship between SST and crop variables. However, both tions. Mon. Wea. Rev., 115, 393–399.
crop response to weather and the response of regional Philander, S. G. H., 1990: El Niño, La Niña, and the Southern Os-
weather to ENSO events can be quite asymmetrical, cillation. Academic Press, 293 pp.
thus, response to an El Niño can be qualitatively dif- Press, W. H., B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling,
1989: Numerical Recipes: The Art of Scientific Computing. Cam-
ferent from a La Niña. CCA cannot be expected to iden- bridge University Press, 702 pp.
tify ENSO effects correctly when this is the case. Fur- Ropelewski, C. F., and M. S. Halpert, 1986: North American precip-
thermore, CCA is more conservative and less robust itation and temperature patterns associated with the El Niño
than ANOVA with respect to the small sample sizes Southern Oscillation (ENSO). Mon. Wea. Rev., 114, 2352–2362.
, and , 1996: Quantifying Southern Oscillation–precipitation
relative to the number of crops and states considered relationships. J. Climate, 9, 1043–1059.
(Stevens 1992; Cherry 1996). For these reasons, CCA Scherm, H., and X. B. Yang, 1995: Interannual variations in wheat
results should be seen as secondary to ANOVA results rust development in China and the United States in relation to
in this study, and should be interpreted with caution. the El Niño/southern oscillation. Phytopathology, 85, 970–976.
Sittel, M. C., 1994a: Marginal probabilities of the extremes of ENSO
These findings have important practical implications. events for temperature and precipitation in the Southeastern
Crop value serves as an index of the aggregate economic United States. Tech. Rep. 94-1, Center for Ocean–Atmospheric
impact of ENSO in the region. The magnitude of the Studies, The Florida State University, Tallahassee, FL, 155 pp.
influence of ENSO phase on crop value was quite large: [Available from Center for Ocean–Atmosphere Prediction Stud-
more than $200 million or 25.9% of the long-term av- ies, The Florida State University, Tallahassee, FL 32306.]
, 1994b: Differences in the means of ENSO extremes for max-
erage value for corn (Table 4, Fig. 5). We believe that imum temperature and precipitation in the United States. Tech.
such a strong influence of ENSO on the value of crop Rep. 94-2, Center for Ocean–Atmospheric Studies, The Florida
production in the southeast United States justifies efforts State University, Tallahassee, FL, 50 pp. [Available from Center
to further characterize, predict, and find ways to exploit for Ocean–Atmosphere Prediction Studies, The Florida State
University, Tallahassee, FL 32306.]
this effect. By knowing the ENSO phase, farmers could Steel, R. G. D., and J. H. Torrie, 1980: Principles and Procedures
modify practices such as cultivar selection, planting of Statistics: A Biometrical Approach. 2d ed. McGraw-Hill, 633
dates, and fertilizer or irrigation schedules to reduce pp.
losses or take advantage of favorable conditions. More Stevens, J., 1992: Applied Multivariate Statistics for the Social Sci-
ences. 2d ed. Lawrence Erlbaum, 629 pp.
study is needed to characterize ENSO influence on Zorn, M. R., and P. R. Waylen, 1996: Seasonal response of mean
weather and crops and to identify optimum management monthly streamflow to El Niño/southern oscillation in north cen-
practices for each ENSO phase. tral Florida. Prof. Geographer, 49, 51–62.

Potrebbero piacerti anche