Wednesday, March 28, 2012

Have Kentucky’s Corn Yields Become More Variable With Time?

J.H. Grove1, E.M. Pena-Yewtukhiw2, and A.A. Marchi3

            We often hear that Kentucky’s climate is quite variable. “If you don’t like today’s weather, just wait until tomorrow” is commonly said. But the current climate debate has serious implications for rainfed agriculture – corn, wheat and soybean need water to mitigate heat stress – and a changing climate could impact Kentucky’s air temperature, rainfall or both.
            There have been numerous statements that climate has become more variable, subject to more extreme events, statements made in the local restaurant, the university classroom, and over the home dinner table. Recently, a grower asked: Have Kentucky’s corn yields become more variable with time?” The answer to that question is: “It depends.” What follows is the answer to that question.
            We acquired planted corn area and yield data for the 43-year period (1969 to 2011) from the Kentucky Agricultural Statistics Service. The data were compiled for each of the state’s six agricultural districts (Figure 1), and for the state as a whole.

Figure 1. Kentucky's six agricultural districts. 

            Kentucky’s annual planted corn area has changed over the 1969 to 2011 time period (Table 1), rising to a maximum of 1.74 million acres in 1985, and ending with 1.38 million acres in 2011. Districts 1 and 2 have steadily increased in planted acreage over the time period, and now dominate Kentucky corn production. Other districts have generally declined in planted area, especially Districts 4, 5 and 6, where the area planted to corn has fallen by half since 1969. District 3’s planted area peaked in 1985, and the district was second only to District 2 in corn production importance at that time. Since 2009, the planted corn area in Districts 3, 4, 5 and 6 has risen by almost 100,000 acres, constituting two-thirds of the increase in Kentucky’s corn production area that has occurred in the past two years.

Table 1. Kentucky’s annual planted corn area, by district and for the state.


1969
1979
1989
1999
2009
District
planted corn area (acres)






District 1
131,300
161,000
208,000
259,000
253,000
District 2
451,300
505,000
563,000
651,100
622,000
District 3
255,300
427,000
309,000
240,300
232,000
District 4
37,400
57,000
41,000
23,100
17,000
District 5
142,800
210,000
150,000
106,100
69,400
District 6
68,900
80,000
59,000
40,400
29.600






Kentucky
1,087,000
1,440,000
1,330,000
1,320,000
1,220,000







            The annual state average corn grain yield was plotted against year (open squares, Figure 2, below), and a linear model (trendline) best described the relationship. The slope of the trendline indicates that Kentucky’s average annual yield has increased about 1.73 bushels per acre per year over the 43-year period (explaining some of the need for more grain storage). Trendline yield has doubled, rising from 70 bushels per acre in 1969 to 142 bushels per acre in 2011. The trendline R2 value indicates that 67% of the variation in state average yield was “predictable” by the simple linear progression of time – improvements in corn management and genetics. Some shifting of corn acres to more productive soils, fields, and farms, may also explain an unknown portion of the observed progress in corn yield.
            The remaining 33% of the variation in yield, the ups and downs (deviations above and below the model trendline) over the 43-year period, were likely due to seasonal weather/climate. These annual deviations were both positive (better than trendline average) and negative (worse than trendline average). To determine whether the yield variation represented by these deviations from the yield trendline had become greater with time, all deviations were converted to positive (absolute) numbers, plotted against year (closed squares, Figure 2), and again fitted with a linear model.
            The R2 for the deviation versus year model was very low and indicates that only about 0.2% of the variability in state yield deviation from trendline was explained by the simple linear progression of time. The deviation versus year relationship was not statistically significant (Table 2), meaning that the expected yield deviation would be constant, about 11.9 bushels per acre, across the 43-year time period. Further, this means that yield deviation, as a proportion of the expected state average yield, has dropped. In 1969, the deviation was 17% (11.9/70=17%) of the trendline yield, but only 8.4% (11.9/142=8.4%) in 2011. Kentucky’s average annual corn yields have not become more variable with time.

Figure 2. Kentucky's average annual corn grain yield from 1969 to 2011. 

            But (there is always a “but”) statewide analysis imposes a “scale” to the answer that may not be appropriate to all parts of the state (unless a grower produces corn across the whole state). The state average yield data are weighted, by county, for corn production area. This is entirely appropriate, but causes District 2 to dominate state average yield (Table 1). Other districts have become more (District 1) or less (Districts 4, 5 and 6) important to state average yield determination since 1969 (Table 1), and the statewide analysis shown in Figure 2 is less likely to reflect what has happened in the latter districts.
            So, the analysis done for state average yield in Figure 2 was also completed for each of the six districts. The results are summarized in Table 2, below. The results show that some districts are “doing better” as regards yield improvement (slope of the trendline) over the 43-year time period. Interestingly, as hypothesized, yield changes with time (deviations from trendline) were not uniform over the state.
            Districts 1, 2 and 3, the western-most three districts, had the most corn acres (Table 1) and exhibited both annual yield and annual yield deviation with time trends very similar to those of the state has a whole (Table 2). Yields have increased at a rate of about 1.7 to 1.8 bushels per acre per year, and the simple linear progression of time explained 62 to 66% of the variation in annual average district grain yield (Table 2). Yield deviation was not significantly linearly related to time in these districts, averaging 13.7, 12.7 and 12.4 bushels per acre for Districts 1, 2 and 3, respectively, across the 43-year time period. And again, because yield about doubled during the period, yield deviation as a proportion of yield has fallen by about half, from 17-20% in 1969 to 9-10% in 2011 (Table 2).
            Districts 4, 5 and 6, the eastern-most three districts, have less corn area (Table 1) and their average annual yield and yield deviation with time trends are not so similar to those for the whole state (Table 2). Yield improvement with time has been more modest in these districts, ranging from about 1.1 (District 5) to 1.4 (District 6) bushels per acre per year (Table 2). And although statistically significant, for all three districts, the yield versus year trendline explained only 35 and 50% of year-to-year yield variation in Districts 5 and 4, respectively.
            The yield deviation versus linear time trend for District 6 was not statistically significant and averaged 9.1 bushels per acre over the 43-year period, a bit less than that for the whole state. Yield deviation still dropped by half (from 13.9 to 7.3%), as a proportion of annual district yield (Table 2).
            In Districts 4 and 5, yield deviations have grown significantly over the 43-year time period, at about 0.17 and 0.29 bushels per acre per year, respectively (Table 2). Though statistically significant, these time trends are also statistically modest – the linear time trend explained only about 4 and 10% of the variability in annual yield deviation within Districts 4 and 5, respectively (Table 2). Still, these trends caused yield deviation, as a proportion of district annual yield, to be maintained (at 11.4 to 12.3%, District 4), or even to increase (from 10.5 to 16.8%, District 5) in these two parts of the state (Table 2).

Table 2. Trends in average yield, and yield deviation, by district – 1969 to 2011.


Yield vs. Year
Yield Deviation vs. Year
Deviation/Yield
District
slope
R2
slope
R2
1969
2011

bu/acre/yr

bu/acre/yr

%
%
District 1
1.79
0.62**
-0.052
  0.004
20.4
9.6
District 2
1.76
0.66**
0.037
  0.003
16.9
8.5
District 3
1.66
0.64**
0.026
  0.001
19.4
9.3
District 4
1.26
0.50**
0.171
  0.043*
11.4
12.3
District 5
1.09
0.35**
0.294
  0.097**
10.5
16.8
District 6
1.41
0.71**
0.037
  0.005
13.9
7.3
Kentucky
1.73
0.67**
0.035
  0.002
17.0
8.4
**Statistically significant at the 95% level of confidence.
*Statistically significant at the 80% level of confidence.
Not statistically significant at the 80% level of confidence.

            In summary, there are three important findings. First, Kentucky corn yields have strongly increased between 1969 and 2011, but that growth in yield has been uneven, across the state. Second, deviations from the yield versus time trendline have been generally constant, for much of the state’s corn production area (Districts 1, 2, 3 and 6), over the 43-year period. This means that yield variation, as a proportion of average yield, has actually fallen in these areas. For other areas, especially Districts 4 and 5, yield deviations from trendline have grown larger – annual yield variation is of greater importance and there is a need for greater understanding. Third, the “scale problem” does not go away with consideration of district yields. Even greater diversity in yield and yield deviation with time trends was observed when individual counties were examined. Growers need to understand that individual soils, fields and farms will also exhibit different trends.
            The answer to the question; “Have Kentucky’s corn yields become more variable with time?” is truly “It depends.” The impact of seasonal weather on these observations is outside the scope of this blog piece, but some sort of a relationship would be expected. And by the way, the trends (lack of trends) shown here are for 1969 to 2011. These relationships inform us about the past, but they do not necessarily predict the future. To paraphrase a famous statement: Past performance is no guarantee of future observation.

1 Professor, Department of Plant and Soil Sciences, University of Kentucky
2 Assoc. Professor, Division of Plant and Soil Sciences, West Virginia University
3 Post-Doctoral Scholar, Department of Plant and Soil Sciences, University of Kentucky

No comments:

Post a Comment