Decision Points While Estimating Grain Yield before Harvest
Published: August 16, 2012
A search on the Internet for estimating corn/soybean yields returns more than 750,000 results, so there is obviously much interest in attempting to estimate grain yields before harvest. Reasons for this interest range from making grain marketing plans to just plain curiosity. The interest in an accurate method increases in years like 2012 because drought has reduced yields and farmers are attempting to determine what to do with the damaged crop: wait and harvest for grain, harvest as a forage, or abandon the field. The word “accurate” is highlighted in the previous sentence because accuracy is important to all persons involved in the decision, but regardless of the procedure used, an estimate is only an estimate, and an estimate is almost never accurate.
Yield in the U.S. uses the unit “bushels per acre”. Bushel is actually a volume measurement ( 2150.42 cubic inches). Since it is awkward to use a volume measurement for commerce, sensible people decided to define a bushel as standard weights that differ among crops. Because grain contains moisture and moisture adds weight, they also set standard grain moistures. So, a bushel of corn is 56 pounds at 15.5% moisture and a bushel of soybean is 60 pounds at 13% moisture.
Because weight is the unit of yield, we estimate grain weight in a small land area and use that number to estimate a field average for yield in bushels per acre. Most common methods of estimating grain yields before harvest involve calculating yield components. In their simplest terms, there are only two yield components: number of seeds and weight per seed. Our ability to estimate those two components impact how close a calculated yield estimate will be to actual yield. To estimate seed number, things such as numbers of plants, numbers of pods or ears, and numbers of seeds per pod or kernels per ear are counted. Seeds are usually not weighed to calculate weight per seed because yield estimates are often made before seeds have obtained their full size. Instead, we use an estimation of seed size (weight per seed) to calculate yield.
Procedures used to estimate yield involve compromises between ease of making the estimate and the accuracy of that estimate. Small errors at several steps in the procedure can result in large errors in the final estimate. There are several decision points during most yield estimation procedures. Understanding the impacts from those decisions can help narrow the difference between estimated and the actual yields.
Decision point 1: What should the size and shape of the small areas be?
Most procedures recommend using a sample area equal to 1/1000 acre. This recommendation is purely for convenience because it makes the math easier. Using a larger area would reduce error, but the area would have to be increased considerably to have much effect. In the compromise between ease and accuracy, using 1/1000 acre makes sense. Most grain crops are planted in rows. Row length equal to 1/1000 acre can be calculated as Row Length (in feet) = 43.56/Row Width (in feet). Remember to convert the fraction of a foot to inches for easy measurement. Some examples of row lengths required for 1/1000 acre are: 30-inch rows = 17 feet 5 inches, 15-inch rows = 34 feet 10 inches, and 7.5-inch rows = 69 feet 8 inches.
A single row is easier to mark and count than multiple rows, at least for rows 30 inches apart or greater. However, as row spacing narrows, the row length for 1/1000 acre becomes nearly unworkable. Following a single row of drilled soybean for 69 feet is difficult, at best. Dividing required row length by 2, 3, or 4 and measuring that number of rows will be much easier. Even in 30-inch rows it might be better to count plants in more than one row. Sometimes planter operation may not be uniform across all rows on a planter. Counting plants in more than one row decreases the chances of selecting an odd row related to planter operation.
Decision point 2: How many areas should be sampled?
This is a difficult decision because sampling more areas decreases error, but adding areas greatly increases effort. Clearly, a compromise has to be made. Fields are seldom uniform. The greater the variability for yield in the field the greater the number of sample areas should be used. Drought usually accentuates in-field variability for yield because of variability for compaction, soil texture, and surface or subsurface water movement. If the sample areas are properly selected, increasing the number of areas to greater than 8 will have only a small impact on error.
Decision point 3: How should sample areas be selected?
It may seem appropriate to select areas in a completely random scheme. However, selecting areas at random without restriction might result in a clustering of sample areas, and this could increase the error of the yield estimate. A better alternative is to make sure that areas are selected throughout the entire field. Sampling throughout the field is not easy. It may be difficult to reach distant areas in large fields. But, management decisions may be based on the yield estimate, so it is in the interest of all parties to reduce error as much as possible. If management decisions can be made to separate regions of a field it may make sense to divide large fields into subunits and estimate yield in each subunit.
Human beings often exhibit biases when selecting areas for sampling to estimate yields. It is common to over sample weak or strong areas. A person may have a preconceived idea about the field’s yield potential or even a vested interest in the outcome. Using methods for selecting areas similar to those used during scouting fields for pests will help reduce bias. Two common patterns of crossing a field are zigzag or M-shaped. It is best, but usually not done, to draw a map and select the areas for sampling before arriving at the field.
Decision point 4: What should be counted and what should be ignored when estimating plant population or ear number?
For soybean, the number of plants is counted because that number is required to calculate the number of seeds. Within the sample areas, there may be weak plants, plants with few or no pods, or plants exhibiting disease symptoms. Many procedures recommend ignoring these plants. Counting these plants will inflate yield estimates if weak plants are used to calculate plant population, but ignored when calculating number of pods per plant. There are situations, such as the drought of 2012, when weak plants with few pods are representative of the field. In those instances, these weak plants should be counted. But, if weak plants are counted for estimating population, they must also be selected when counting pods.
For corn, the number of ears is counted and that number is used to calculate number of kernels. Plant population is not needed, but a decision must be made as to which ears to count. Even under good conditions, ear size may vary considerably. Second ears on a plant are usually small, but weather, nutrients, and pests may also affect ear size. Commonly used procedures recommend not counting “nubbins” or small ears. The definition of nubbin is usually left open to interpretation. Ears that are likely to be missed by the combine should not be counted. If small ears are counted while estimating ear number, but ignored when estimating number of kernels per ear, yield estimate will be inflated. Count small ears if they represent the field, and make sure that small ears are selected in the same proportion when ear size is estimated.
Decision point 5: How many plants or ears should be used to determine seed number?
For soybean, the number of pods per plant is required before seed number can be calculated. Pod number per plant is highly variable, even under good weather conditions. Error of the pod number estimate can be reduced by counting more plants. Unfortunately, counting number of pods is tedious, especially if plants are highly branched. Most procedures recommend counting a minimum of 10 plants. A 1/1000 acre may have 150 plants, so 10 plants represent about 7% of the population. This is probably an adequately sized subsample. If plant to plant variation seems large, average pod number for the first 8 plants and then continue to calculate an average after 9 and 10 plants. If the average does not change much from 8 plants to 10 plants, then a 10-plant sample is appropriate. If the average changes too much, then count an additional two plants. Plant variation for pod number will be greater in fields with sparse populations than in fields with dense populations because of branching. A need to increase the number of plants chosen for counting pods is more likely in fields with sparse or uneven populations.
Some procedures recommend randomly selecting the soybean plants for which pods will be counted. It is nearly impossible to be unbiased while selecting plants in this manner. Other procedures recommend counting 10 consecutive plants. Although this method reduces bias while selecting plants, it places all plants in a portion of the row that may not be representative of that row. And, it is easy to be biased when selecting the location in the row to start counting. Consider dividing the number of plants in the small area by 10 (e.g. 150 plants/10 = 15 plants). In this example, count pods on every 15th plant in the row. This will force selection plants from the entire row length. Count pods on the selected plant if you used that plant to determine population. If you did not use it, then count pods on the next plant in the row. Count all pods with seeds large enough not to be lost during harvest. Ignore flat or twisted pods with no seeds.
Decision point 6: Should numbers of seeds per pod and kernels per be counted or estimated?
The number of seeds per pod is also required to calculate seed number. Most procedures recommend using a standard 2.5 seeds per pod instead of actually counting and calculating number of seeds per pod. Study the plants used to count the number of pods. An average of 2.5 seeds per pod means about 60% of the pods possess 3 seeds, 30% of the pods possess 2 seeds, and 10% of the pods posses 1 seed. If plants appear to differ from this ratio, then use a slightly higher or lower number.
For corn, number of kernels per ear is calculated using number of kernel rows and number of kernels per row. Since counting actual numbers is not time consuming, it makes sense to count and not to estimate number of kernels per ear. Several procedures recommend using every 5th ear to estimate size. This is an adequate number, and will also force ear selection from the entire length of row. If the count for the number of ears included the second ear on a plant, then that ear must be one of the 5 ears in the intervals for counting. In other words, one plant might be used for two of the five counts. Count the number of rows of kernels. This number should be an even number. Drought, nutrient deficiency, herbicide damage, or pests may result in atypical row patterns or twisted ears. For those ears, estimate average row number.
Count the number of kernels in several rows. Some procedures recommend that you ignore kernels at the butt and tip ends of the ear. Kernels near the butt are often large and contribute to yield. In most instances, they should be counted even if including them results in an unequal kernel number among rows. Kernels near the tip are often small. Do not count kernels that have obviously aborted or development was not complete. In most instances, these kernels will be less than half of the size of the other kernels on the ear. Stress during pollination may have resulted in unfertilized kernels in the middle of the ear – ignore these when counting. For several reasons, kernel number may vary among the kernel rows, so estimate an average number of kernels per row.
Decision point: What is the appropriate seed size?
Seed size is actually weight per seed and is used to convert seed number per acre to bushels per acre. Both genetics and environment effect seed size. Variation among seeds is large, and may vary more than 40% from smallest to largest seeds. Using the wrong seed size can result in a large error in final yield estimate.
Seed size for soybean is often given in number of seeds per pound. Often, seed size is provided on the seed tag, but that number may only slightly influence seed size in field. Most procedures use a standard number between 2500 and 3000 seeds per pound. Under normal conditions, 2700 or 2800 seeds per pound is appropriate. Most procedures used to estimate yield for corn use kernel number per bushel in the calculations. Under normal conditions, there are about 85,000 kernels in a bushel, which translates into seed size of about 1500 kernels per pound.
Unfortunately, environment, including weather conditions, can have a large effect on seed size. Most procedures recommend using a smaller seed size (more seeds per pound or more kernels per bushel) under poor conditions. But, poor conditions reduce seed size only if they happen during mid to late seed filling. Poor weather conditions during pollination of corn or pod set in soybean may actually increase seed size if stress is relieved during seed filling. Adjust seed size estimate based on actual conditions the field has experienced. Adjusting seed size more than 10% is usually not warranted.
Estimates of yield are estimates. Even when great care is used to select areas for sampling and the plants within those areas, estimates very seldom match actual yield. Errors in opinion polls are often 3 to 4%. Errors in yield estimates are likely to be 20% and may be greater. Careful attention to these decision points will reduce, but not eliminate error. Be judicious when using yield estimates to make crop management decisions.