In preparation for my Principles of Crop Production class on what limits our crop yields, I listed two major limitations, climate and soils with some other factors like pests also having an influence.
When I looked at my notes there was a lot of general information about the climate on the Prairies but I wondered how I could make it more interesting and applicable to the students.
So I did an analysis of Manitoba’s average yields of wheat and tried to correlate that yield to the rainfall. Surely, I thought, you could predict the yield of a crop by the amount and timing of rainfall.
Unfor tunately when you use averages over the entire province there is not much of a correlation. The best predictor of wheat yields in Manitoba over the last 100 years is actually YEAR. The only other factor that comes up as a small influence is JULY rainfall.
Then I thought of my dad’s records. My dad loves keeping track of everything on his farm. He has recorded the crops he has grown, the variety, the seeding date, the monthly rainfall and the yields since 1949 and his information is site specific.
So I meticulously transferred his paper records to a spreadsheet and my wife Patricia (an expert in statistical analysis) and I analyzed the data using a statistical software program called NCSS.
I was gleeful when we found two factors that influence wheat yields for my dad’s farm. We initially tried a number of factors: monthly rainfall from April to August, the year and the seeding date. The only two that had a significant impact on wheat yields were the rainfall in MAY and in JULY.
My dad, Abe Martens, farms on Red River and Osborne clay near the town of Kleefeld, about 15 miles east of the Red River in southeastern Manitoba. The formula for predicting wheat yields applies to his location and I do not know if this formula extrapolates to other areas of Manitoba.
The predicted wheat yield for my dad’s location is:
Wheat yield = 51. 8 2.9xMay rainfall (in inches) 2.7xJuly rainfall (in inches.
This formula has an R2 of .57 which means that 57 per cent of the wheat yield can be predicted by this formula. Obviously, there are other factors not in this formula that also affect the yield. Things like which variety he used, how much fertilizer he used and other agronomic factors. This formula is very specific to my dad’s situation. He tends to use a similar amount of fertilizer each year and his practices are fairly consistent over the years.
Theoretically, this formula predicts that if my dad had no rain in May and no rain in July he could get 51.8 bushels of wheat per acre. The rainfall in May and July are both negatively correlated to the yield. That is, the more rainfall in May and/ or July the lower the yield of wheat.
In 2004, he received 7.1 inches of rain in May and another 4.2 inches in July. When you plug those numbers into the formula you get a wheat yield of:
51.8 (2.9×7.1) (2.7×4.2) = 51.8 20.6 11.3 = 19.9 bushels per acre.
His actual yield in 2004 was 20 bushels per acre.
No formula is perfect. It will only give us an approximation. When I did this calculation to find the predicted wheat yield from 1982 to 2009, I found that the difference between the predicted yield and the actual yield was 15.5 per cent. That is, I was able to predict the actual yield quite well.
This formula may be different for each farm. If you have the monthly rainfall data (May-August), seeding date and yield records per year from your farm you can create your own formula using a statistical analysis called multiple regression.
To get a useful formula you need enough data points, probably 30 years. If that is beyond your abilities or those of your children or friends I would be happy to calculate a formula for your farm. Your data would be treated confidentially.
Send data in an Excel spreadsheet to [email protected]
(Gary Martens is an agronomy instructor at the
University of Manitoba.)