Why clean your yield data? Find out why this step is so important…
April 25, 2017
We had a great comment/question come across our agchat from Agri-Coach Andrew Clements, co-owner of Premium Ag Solutions in the Lethbridge AB area, referring to the interpolated yield map shown in Fig 1. “I see these all the time in the crop and you can also see them in Google Earth. Now I am seeing it in the yield data. Originally, I thought, it was due to swathing crops for so many years but not sure, why this would be so yield debilitating. Now I am wondering, what is causing this?”
Denis Vermette, Agri-Coach and owner of Ag Success Strategies near Winnipeg MB suggested, “Many years of discer ridges is my best guess. I see them all the time too but they usually have higher fertility because the farmers used to go up and down the corners when they were done the field to cover up the misses, which would result in higher seed and fertilizer rates”.
New Precision Support Specialist Jason Steffen offered his take on the situation – Having looked at many yield maps, my first question would be, “Was this yield data cleaned prior to use as a yield map?”
The other Agri-Coaches are correct in that these patterns can occur from past tillage practices. I’ve seen this pattern in fields where farmers used a moldboard plow to plow a field in a circle for many years. This caused the X pattern you see, as that is the point where the farmer would come back and clean up the turns.
However, that X pattern could also be an artifact of the how the yield monitor recorded the data. As the combine goes around the corner, the header may become fuller/emptier and the yield data recorded for those points won’t be directly related to the actual yield on the ground. If that was the case, those points should be cleaned from the data prior to bringing them in to use as a yield map. Without seeing the raw point data, this is difficult to determine what is really going on.
We all know that harvest is a period of great time pressure, with a just “get ‘er done” attitude. Nevertheless, time taken to ensure the GPS, yield monitor, and data storage device are working correctly and the combines are calibrated can make a tremendous difference to the quality and value of the data at the end of the season. Even after taking the time to set up the equipment properly, it is also important for the driver to follow certain procedures to reduce the amount of “bad” data generated. They must remember to set the monitor to the right Field and Load, set the correct Crop, drive at a consistent speed, keep the header full, and make sure the harvester is set properly and grain is not going out the back of the harvester.
The driver can do everything perfectly, yet harvesters will still encounter inherent errors that are specific to each machine, header, or crop. Because of the dynamic nature of grain passing through the harvester to the yield sensor, it is almost impossible for the system to accurately estimate the yield at any given point. Before using yield data, it is important to remove as many of the errors present in the data as possible.
The first error that needs to be resolved is the Flow Delay. When the GPS on the machine records its position, it provides the position of the machine as the crop is entering the head of the machine. However, it takes time for the grain to move from the head, into the harvester, through the separator, up the clean grain elevator, and then to the yield sensor where it finally measures yield. The consequence is that the yields are not mapped in the correct spot, and each pass going in the opposite direction create a saw-toothed pattern.
To correct this, we must enter a Flow Delay value that shifts the yield data back to the position it was “X” seconds back. This Flow Delay value is dependent on the harvester, head, GPS location, and even crop. Once the proper Flow Delay is entered in the display or the post processing software, the yield map will show yield transitions much more smoothly.
While the harvester is starting a pass, it takes time for the grain flow at the yield sensor to reach a steady state. Alternatively, at the end of a pass, it takes time for the grain flow to return to zero. We need to remove the points for the Start Pass Delay and the End Pass Delay. This is just a fixed number of seconds of data at each end of the pass that we just throw out.
Other points that we need to remove include those recorded when the harvester is traveling too fast, traveling too slow, or changing speed too quickly. Data that is harvested during these times is just not reliable. We need to remove unreasonably high and low values. “Reasonable” values can be determined using statistical analysis. Finally, we need to remove the data points recorded with partial swath widths or at angled headlands. Since the head is not harvesting a full swath in these situations, the yield sensor underestimates the yields. Fig 4 shows the yield map after these errors were removed from the raw data.
Another type of error can occur when two or more machines are operating in the same field and the data is combined. The distinct streaks indicate that one or both combines were not calibrated or there was big difference in how the two operators set and ran their machines. This map is not useful in its current state as the differences in the data from the two harvesters mask any spatial difference the yield data might show us. Before using this data, it is important to adjust the data so that all machines show yields in the same relative ranges. All of these errors contribute to create noise in the yield data and unless cleaned out, they reduce the value of the yield maps. This is especially true if you want to use the information for further decision-making processes.
Yield data is an incredible source of information about your fields. Ensuring that your harvesters are set-up correctly and they are collecting good data can be an important way to provide you useful information help you improve your future crop production plans. With good, clean yield data, we can:
• Get a report card on how your management practices did for the year
• Determine CPUP (Cost per Unit of Production) by Zone
• Create Profit Maps to determine where your management program is making/losing money
• Refine estimated yield goals to better match actual values in the field
• Calculate nutrient removals
• Determine better estimates of nutrient use efficiencies
The best practices to get accurate yield maps are as follows:
• Set up the yield monitor with proper farm, field, and crop information prior to starting harvest
• Perform factory recommended calibrations for each crop on each machine
• Confirm settings on display before starting each field
• Keep intake head full, adjust display if swath width changes
• Run the data through cleaning software after harvest. Your Agri-Coach can help you with this.
Jason Steffen has joined our Knowledge Team from Trimble as a Precision Support Specialist. He is based out of California. He is very knowledgeable about Yield Data and the value it can bring to an operation… providing it is cleaned and as accurate as possible.