A simple variable-rate prescription algorithm for your fields

Sunday, November 8, 2015

In the fourth of a series of articles, the authors outline how to develop a variable-rate algorithm which may not save you money but should result in better-fed crops and higher yield

by MIKE DUNCAN, RICK WILLEMSE and SARAH LEPP

On Sept. 10, we made a presentation at ZoneSmart, a precision farming event, to what was a very appreciative audience. The day itself was meant to be about how to define management zones, but most of the people there seemed to want to skip to the end and see what a variable-rate solution might look like.

There's a lot of buzz surrounding the idea of variable rate, whether it makes sense to do and whether it is profitable. In this series of articles, we've been slowly moving towards the goal of defining a variable-rate farming algorithm.

At ZoneSmart, there were presentations on a variety of methods to define management zones that would lead to variable-rate prescriptions. One of the simplest variable-prescription algorithms avoids zones altogether and takes the pattern of yield in the field to modulate the blanket fertilizer application. This is likely the simplest algorithm for a set of prescription maps that is possible, but it only works if the field shows stability of the yield pattern from year to year for a given crop.

Many years ago, Doug Aspinall, a researcher at the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA), noticed a strong correlation between yield and the topographical features of a farm field. Doug began to assemble the measurement and software tools necessary to explore this relationship. The most important discovery that he made is that yield patterns seem to remain stable from year to year.

Our group at Niagara College has assembled a large volume of data to verify this and discovered that these patterns are easily visible in crops like corn, wheat, and often soybeans, and that these patterns correlate very well with the shape of the field. In my first article in Better Farming, in March 2015, A simple exercise that will change your views on variable rate fertilizing I outlined the Yield Probability Index (YPI) algorithm. The presence of zones on a YPI map is due to this yield stability. YPI provides proof that Aspinall's observation was correct and can be exploited.  

Just to emphasize the notion of yield stability, any of Doug's many presentations from the last 20 years shows the effect. The most remarkable aspect of Doug's maps, and the similar analyses that we've done at Niagara College, is that, regardless of management strategy (till or no-till, different fertilizer strategies, different seed varieties, single and multi-trait), the yield patterns remain. In one of our example fields in Parkhill, the drainage was blocked in part of the field one year (see Figure 1, year 2000), creating a blot on the yield patterns. Once the problem was fixed, the blot was gone and the pattern returned.

What this means is that we can, in most cases, point out exactly where a field is higher in production and where it is lower and that, based on the historical data of the field, the pattern will stay that way. This knowledge of highs and lows is easily exploitable as the basis of a prescription algorithm, because it represents an exact modulation that needs to be applied to a blanket fertilizer application to make it match the pattern on a farm field.

So, the first part of the algorithm to do prescription maps in your farm field is to:
Take your historical data (or start collecting yield data ASAP);
Clean it (get rid of ridiculously high and low values);
Normalize each year to one (divide each cell in your map by the average yield for that field that year; the range of normalized maps is typically between 0.2 and 1.6);
Add the normalized maps together, and divide by the number of maps to get the Yield Index (YI) map that Doug Aspinall has been talking about all these years. (This can also be done by using the Niagara College Crop Portal at globalfarmmanager.ca.)

The YI map in this exercise is the pattern for your prescription. It is also the end of where the math can do all the work. The next step involves a guess by the farmer as to what next year's yield might be. There exists no commercial algorithm to do this for the farmer (this algorithm must be used field by field and year by year). All we have to do now is figure out what your average yield might be this year and multiply each cell in the map by the expected average yield to get your potential yield map. Unfortunately, the only help math can be in figuring out the yield for this year is to average your historical maps and see what the number is, which might be a helpful start. At this point, the farmer has to decide to raise or lower the value of the yield target.

The new map that results from multiplying the YI map cell by cell by the selected yield target number will be in units of bushels per acre. The next step is to turn this map into a map of application rates. We do this by multiplying this map by the uptake rate for the crop being planted. The uptake values can be found in the OMAFRA soils guide, or on the web.

The final step is to balance the nutrients present in the various fertilizers available to Ontario farmers. For example, when balancing N, the farmer needs to know how much N has already been applied in terms of popup and side-dressing, and in the formulation of other fertilizers.

The resulting set of maps, one for each nutrient, are your prescription maps. Systems like John Deere's Apex can take these maps and guide the equipment to apply fertilizers according to the patterns in the maps.

Doing all that math and making one educated guess results in a prescription map that will meet your field's needs better than a blanket prescription. The reason this is true is that the YI map is higher where the yield in your field is higher and lower where the yield is lower, hence the prescription map follows your pattern of yield.

The big question, as always, is "Where's the money?" This method won't save a dime on inputs; however, it should result in a better-fed crop, which often results in higher yields. The problem with blanket applications of fertilizers is that they under-feed the high-producing parts of the field and over-feed the lower-producing parts of the field, hence the blanket technique can be accused of wasting fertilizer and starving the field's high producing areas.

Of course, this isn't the end of the story, it's just the beginning. One of us, Rick Willemse, began developing a whole farm management system based on these ideas in 2008. Rick's algorithm can achieve true savings, but involves many more decisions by the farmer. The basic algorithm presented here is the start of that.

However, as was found in the 1990s, which is the last time variable-rate prescriptions were in vogue, the realization came that the process cannot be fully automated. Indeed, the algorithm here needs the input of the farmer or Certified Crop Adviser to specify a target yield. Beyond that comes the realization that every field is different and must be treated as such.

Since the efforts in the 1990s were to achieve "automatic prescriptions," the effort fell apart and is only now being resurrected with the realization that the prescription map is just part of a whole farm optimization and management system. BF

Rick Willemse is a cash cropper from Parkhill. Dr. Mike R. Duncan, PhD, is Natural Sciences and Engineering Research Council of Canada, Industrial Research Chair for Colleges in Precision Agriculture and Environmental Technologies, Niagara College Research and Innovation. Sarah Lepp is a senior research associate.