Single Prescription Optimized Crop Fertility Considering Variability Within the Field

Variability within a specific field can be a challenge when it comes to devising a single prescription, non-variable, crop fertility plan. Client discussions often arise about how such variability should play into the fertilizer plan for a given field. A common concern in these discussions is that since fields contain areas such as salinity, or sand ridges, etc, that the “optimized” fertility plan does not consider these poor yielding areas of a field (which are intentionally avoided when sampling). Often the suggestion is that planning for 55 or 60 bushel canola yields should be scaled back accordingly in consideration of sub-optimal areas which dilute the “true” yield expectations down to 40 or 45 bushels. It is a valid concern and worthy of consideration in determining the ultimate fertility plan for a field.

How should one deal with such a challenge? Fortunately with Western Ag, we possess a tool that we can use to help us create simulations of this type of variability. Then we can compare the bottom lines generated under different crop fertility management ‘philosophies’ to see what type of strategy is most favorable.

Often the first instinct of a farmer client is to think in terms of managing fertility based on average productivity and yield expectations. In other words, the view is that, say, 60 bushel canola yields are unrealistic for certain fields and therefore fertilizer plans should be scaled back to reflect a more ‘realistic’ yield expectation of 45 bushels (or whatever). Is this the correct philosophy to determine the most profitable or lowest risk fertilizer plan for such a field?

When we sample a field, we intentionally avoid any saline areas which tend to get seeded. And low productivity areas like sand ridges would also be avoided unless such areas were typical of the field. So when we run the CropCast model to devise an optimized fertility plan the yields generated by the model do not represent an average across all areas of the field. We are estimating a yield which excludes the poor productivity areas. If we go with such a plan then the philosophy would be to ignore the yield diluting effects of low productivity areas such as saline areas and sand ridges and plan for 60 bushels.

So we have now two competing philosophies in setting the crop fertility plan for a field which we can use to run simulations and hopefully clear up which strategy is likely to be most profitable and/or lowest risk. The alternative management philosophies are:
1. Set a plan based on yield which excludes low productivity areas like salinity and sand ridges – ie: 60 bushels.
2. Set a plan based on the overall average yield for a field – ie: 45 bushels.

Once the lab analysis comes back and the CropCast model is set up, we are ready to run the simulations of the competing philosophies and select the one which is best for our field. Using philosophy #1 the model does not consider low productivity areas and an optimized plan is generated with an expectation of 60 bushel yield (Figure 1 below). However we know that the field consists of salinity and sand ridges, which means that the true yield expectation of the field with this type of fertility is 45 bushels not 60 bushels. In this case, from discussions with the client we understand that about 60% of the field could be expected to yield 60 bushels per acre (in line with the CropCast model) while the other 40% would have an expected yield of only 22 bushels per acre with the lower yields attributable to the marginal saline and sand ridge areas. The overall average yield for the field is then expected to be 45 bushels. From the model run, the applied fertility for the field is 99-33-0-24 at a cost of $79.32/acre. If the price of canola is $9.50 per bushel and all other non-fertilizer costs are $250 per acre then the profit margin for the field calculates as follows:
• 45 bushels x $9.50 per bushel = gross of $427.50 per acre
• $427.50 – $79.32 (fertilizer costs) – $250 (all other costs) = $98.18

Figure 1: CropCast simulation excluding low productivity areas of field

Figure 1: CropCast simulation which excludes low productivity areas of field.

Now for the alternate plan (philosophy #2) where fertility is set according to average yield expectation of 45 bushels (Figure 2 below). Here we see that the CropCast model has computed a fertility plan of 61-21-0-16 to generate this yield (note that this field has some N carryover supply of about 30 lbs already in the soil). Before we consider the yield diluting effects of salinity and sand, the CropCast summary shows the following:
• 45 bushels x $9.50 per bushel = gross of $427.50 per acre
• $427.50 – $50.01 (fertilizer costs) – $250 (all other costs) = $127.49
So at first glance the strategy of fertilizing to a set ‘realistic’ yield target of 45 bushels averaging across good and poor productivity areas would appear to be the more profitable and lower cost strategy. However this analysis has not yet accounted for the low yielding saline and sand ridge areas of the field. Even if we make the (generous) assumption that the yield in these low productivity area is 22 bushels per acre – same yield as with the higher fertility scenario – the overall average yield would be 35 bushels and not 45. The profit margin would then calculate out as follows:
• 35 bushels x $9.50 per bushel = gross of $332.50 per acre
• 332.50 – $50.01 (fertilizer cost) – $250 (all other costs) = $32.49

Figure 2:

Figure 2: CropCast simulation that sets yield target by averaging across low productivity areas of field.

We see from this exercise that it is a mistake to devise a fertility plan for this field that averages across the poor yielding saline and sandy areas of the field. We wind up shooting ourselves in the foot and missing out on $98.18 – $32.49 = $65.69 profit per acre. While it might seem counter intuitive, the best fertility plan for this field is to stick with the model optimization which excludes such areas. Real life yield expectations can still be set straight after running the model by accounting for low productivity areas and adjusting true yield expectations accordingly.