Ukrainian agro-giant Kernel, in association with its partners, runs one of the biggest grain farms in the world. The operation spans 6,000 fields and more than 500,000 hectares of land, making Kernel the largest producer and exporter of sunflower oil in the world. The company operates a mixed fleet of hundreds of combines and relies on a suite of enterprise-grade farm management systems to maximize yield and profitability every season.

FarmTRX first partnered with Kernel back in 2020 when the Ukrainian company started to look for a reliable yield monitoring solution in support of their multi-billion-dollar mega-farm operation. Today, FarmTRX is at the heart of Kernel’s sophisticated yield monitoring and mapping process.

We recently sat down with Valerii Datsenko, Head of Kernel’s Precision Agriculture Department, to discuss how it all started, and how FarmTRX fits into the company’s precision agriculture strategy today.

This transcript has been lightly edited for clarity and length.

How did you measure crop yield before implementing FarmTRX yield monitors?

We simply measured yield at the elevator after unloading trucks from the field, using an average across the field. We did have some yield mapping systems before FarmTRX, but they were strictly factory-installed, proprietary systems that couldn’t be used across our other brands or with our rental partners’ combines of older brands and models. It wasn’t scalable.

Did you evaluate other yield monitoring solutions?

We evaluated solutions from the big brands, e.g., John Deere, Precision Planting, Trimble, and Claas Telematics—but all were deemed too expensive and couldn’t easily be implemented. Many of these also had constant problems with moisture sensors and data quality, requiring frequent recalibrations.

Finally, we looked at a Ukrainian manufacturer, but they didn’t meet our needs with regards to product reliability and API integration capabilities.

What were the most important considerations for you when choosing a yield monitoring solution?

The most important aspects were scalability, reliability, speed of installation, affordability, and the ability to use yield monitors with displays from multiple brands. We also required the solution to operate independently in the background, regardless of the machine operator’s actions. For example, operators used to have to constantly start the monitor, select fields, calibrate, and turn on recording. If they forgot, or missed any part of the sequence, mapping was impossible or required time-consuming post-processing of the actual data.

Valerii Datsenko, Head of Kernel’s Precision Agriculture Department

Precision agriculture is a data-driven and fine-grained approach to farming that aims to lower costs and maximize yields by considering local variances across a field. This process depends on one critical foundation: accurate yield maps that provide detailed information about how much each field produced at harvest time, acre by acre.

Before accurate yield maps were widely available, farmers evaluated their fields based on averages, or intuition, and they often ended up using the same amount of seed, fertilizer, and nutrients across their field. Without detailed yield maps they wasted resources in low-producing areas and missed the opportunity to improve yields in high-performance zones.

Stats show that most small farms and many mid-size farms in North America have an operating margin of 10% or less which puts them into the “significant financial risk” category. Building detailed yield maps, and the precision agriculture operations they enable, is one of the most reliable ways to increase productivity and profitability of your farm.

What is a yield map?

Yield maps are created by combining data from yield sensors, moisture sensors, header height sensors, and GPS. These maps can show precisely how much crop was produced at different locations across a field during harvest.

As the combine moves through the field, the system records yield and moisture data in real time, building a precise, spatial view of performance across every acre. The most sophisticated yield monitoring systems can take advantage of error-correction technologies like PPP HAS and RTK, and record positions with an accuracy of down to a few centimeters on the horizontal plane.

Header height sensors help improve the accuracy of yield maps by telling the system to collect yield data only when the header is in the working position. These sensors can also improve the in-cab calculation of harvested area and provide real-time cut height feedback to the operator.

Once all the raw data from each sensor has been collected, sophisticated algorithms are applied to account for areas that could introduce errors if left uncorrected. For instance, around the edges of a field it is common for a combine to do a partial swath where the remaining crops only cover half the width of the combine, resulting in only half the amount of grain being harvested compared to a full row. Without corrections, the yield map might show this area as a low-performance part of the field, although it might be the opposite.

Post-harvest data corrections will use the defined field borders and precise GPS locations to identify all the various turns and edge rows that a combine moves through and make adjustments as needed.

Premium yield mapping solutions can seamlessly consolidate data from multiple combines working the same field, or multiple fields, resulting in detailed, accurate, color-coded map of your fields, divided into High (green), Medium (yellow), and Low (red)-performing zones.

How are yield maps created from multiple combines in the same field?

If you have multiple combines running on your farm at harvest, you can get unified yield maps for the whole operation by adding FarmTRX yield monitors. Regardless of the make, ages and models of the combines in field, data collection can be simplified and standardized with the farmer getting seamless yield data and maps all from one add-on system.

The FarmTRX harvest monitoring kit is a universal, retrofit yield monitor and periphery sensors that can be installed on any combine with a clean grain elevator and 12V switched power, equipping the harvester for precision yield monitoring and automated yield map creation in every field. When FarmTRX yield data is synced to the farmer’s online Web App account, yield maps will begin to generate immediately for all harvested area covered by the yield monitors. Differences between multiple combines in the same field (like calibration factors, operator involvement, header values etc.) are interpreted and resolved in the Web App processor and generated into yield maps. The process of yield data clean up and processing into maps is automatic and fields are completed within minutes of data sync. What does the farmer end up with? Unified, cohesive yield maps for the whole farm that are ready to support agronomic review and decision making.

Here is how yield data from multiple combines is treated in the Web App:

When multiple harvesters are similarly calibrated during harvest using weigh wagons, a grain cart with scales or even a simple volume measurement (like the known volume of a combine grain tank or hopper-full), not much adjustment is required in the Web App to normalize the collected yield data between different combines. Regardless, a multi-combine correction check is undertaken in the Web App automatically.

Multi-combines with different calibrations

Consider some of the factors that may lead to differing calibrations for multi-combines in the same field: an operator missed performing an in-field calibration for the crop type altogether, one combine may be running at a different threshing speed than the others, several different calibrations were performed for the same combine all within the same field, or simply – there are different combine makes and models running in one field. In the Web App, you’ll notice a map layer called “Raw Points” and “Raw Map”. These layers reveal what that map would look like without Web App correction and processing. Segments of the map are distinctly (visually) separated between harvesters, ultimately rendering a harvest report card that’s unhelpful for comparing true yield performance across an entire field or entire farm. Check out this example:

Transformation from raw points to Corrected Maps (automatic)

Multi-combine checks and corrections are part of the processing of yield data when it is synced to the Web App, so the farmer gets a suite of corrected maps for each field that reveals true yield trends that surpass day-to-day harvest conditions (combine unloading in the field, breakdowns, plugged threshers, multiple combines with different or missing calibrations).

Without saying too much, here’s part of the clean-up process and why it works:

To address different combines harvesting in the same field, sophisticated algorithms compare combines working in close proximity to each other and determine how each can be statistically re-calibrated to produce an accurate picture of yield over a wider area. This addresses visual striping that is evident in the raw (un-corrected) yield map – called the Raw Map. This process helps produce accurate yield trends and sums when no calibration can be found for the other combine. If the farmer later adds the known field total post-harvest, the individual data from each combine is reprocessed to sum to the new known totals. Post-harvest calibration tools in the Web App allow the farmer (or the farmer’s agronomy team) to get accurate, unified yield maps after harvest, from uncalibrated Yield Monitors.

How many combines can the Web App handle?

Fortunately, there’s no limit. As long as FarmTRX Yield Monitors are added to the machines, seamless yield maps will be created for all combines running in the same field. How many is too many? Check out this example from a FarmTRX customer farming in Ukraine, where 9 combines in 520 ac./210 ha. fields are the norm. So far, no problem for the FarmTRX Web App. Note the way the yield trends reflect the soil properties of old stream beds, regardless of which combine is providing the data.

If you’ve been running with FarmTRX Yield Monitors for a few seasons or have the system installed across multiple harvesters, you may be wondering: Which model of FarmTRX Yield Monitor do I actually have?

Ontario, Canada – Farmers can bring yield data from any combine directly into Trimble Ag Software using FarmTRX.

Raw yield data points measured by the FarmTRX yield monitor will_­ always contain some error points – points that, although accurate to what was measured by the sensors, aren’t a true representation of how the crop yielded for that location in the field. Why is this?

Answer: the reality of yield monitoring at harvest – machine design of combine harvesters, the necessary sequence of getting the crop off the field, variation in field topography and, of course, humans are part of the process. This means calibration errors, multiple combines with different calibrations in the same field, running the combine at a throttle different to what was used in calibration, and the general unpredictable nature of harvest, all just happen.

Here lies the task of the FarmTRX Web App, to generate usable yield maps that tell an accurate story of performance across the field from uploaded data points. To create a valuable yield map, errors in points must be automatically detected based on many criteria, and corrected or removed. The automated, smart process of detection and correction is what transforms the log of raw data point into precision yield and moisture maps that are actionable and enable farmers and agronomists to make decisions impacting return on investment.

The process of taking points to maps can be completed by an agronomist, data scientist or those comfortable working in GIS software. With FarmTRX, the process is automated and part of the yield monitoring package, offering farmers an easy and near real-time solution to getting reliable yield and moisture maps for their fields.

Error point detection and correction is one algorithmic task out of many of the Web App processor. The goal of this article is to help our audience understand what sort of data points need correction by the Web App and how they are addressed and removed to make maps. In the FarmTRX Web App, the process begins automatically once data is synced, so farmers will start to see maps in their account within just a few minutes to hours of upload.

Raw points to Trend Map

But first, what is the FarmTRX Web App?

The Web App is FarmTRX’s online platform where synced yield data from the yield monitor and moisture sensor is sorted into field boundaries and can be viewed, exported or connected to other farm management platforms like John Deere Ops Center and Trimble Ag Software. For the first year after purchase of the yield monitor, every farmer has free, unlimited access to Premium-level service in the Web App. Premium includes automated yield and moisture mapping for every field, post-harvest calibration features and data integrations. After one year, farmers can choose whether to renew Premium service – an annual fee based on harvested acres in the account (ranging from 0.50 down to 0.10 cents/acre), or use the Free version of the Web App.

If a farmer is using the Free version of the App, they will see 2 map layers per field in their Web App account instead of 9: raw yield points and a field boundary. Raw points can always be viewed and exported as shp files at any time.

What data points are removed during yield map processing?

Start and End Pass Delays

A processing delay is the time it takes for the crop to enter at the header and fill the clean grain elevator. For most grain harvesters the processing delay is 10-12 seconds. FarmTRX automatically adjusts for processing delay for all points, including raw points, so farmers don’t have to worry about a lag time in their raw points.

Start and end pass delays occur when the first and last points recorded in a harvest pass are measuring grain flow that has not yet stabilized as it gradually fills up/empties the elevator.

The start and end pass delay effect means that the 3 or 4 yield points recorded at the start and end of each row show a gradual increasing/tapering off in yield value – for example, the yield recorded at the start of a pass may increase from zero to 15, 40, 75, 100 bu/ac within just a few seconds. In reality, all these points are likely close to the 100 bu/ac mark.

Start and end pass points are detected by the Web App processor based on their rate of change of speed, location, geometry, heading and yield. To fix these points, it’s important to note that instead of simply taking them out from the data, the processor temporarily removes the points and then reinserts the points with new, legitimate values by interpreting the surrounding “good” points. By re-inserting the points with updated values the total area and yield calculations remain accurate and swath maps may be generated without holes.

Headlands/Pre-Harvested Area

What about when the combine covers land that it has already harvested, like headlands?

Below you’ll see an example of raw points recorded by the yield monitor in a field. The points show continuous overlap as the combine turned in the headland. In this field, the operator was harvesting both east to west and north to south, so at the intersection where the combine turned, many additional points are recorded. While these points are accurate measurements of yield at that time, they aren’t helpful in a yield map.

The Web App automatically detects points in areas previously harvested using several factors, not just location, as GPS drift and other location accuracy concerns can introduce overlap errors.

Headlands point cleanup from Raw to Corrected

Stops and Starts in the Field

The raw points will reflect when the combine stopped harvesting and was stationary in the field, for example when stopping to unload the grain tank, do repairs, or to make sure the crop is properly feeding through the threshing system to avoid plugging. At these times, the yield monitor will be recording false high yields as the combine slows but grain flow remains the same. As the elevator empties, false low yields will be recorded.

While these points are accurate measurements of crop flow, they are not a true representation of what the area yielded.

Below you’ll see an example where error points are detected in a field with swathed canola clumps – times where the combine operator had to slow down to a stop to harvest a clump. The Web App identifies these error points and performs the process of removing from the log to make the correction, then intelligently reinserting with new values based on their surrounding points. This cleanup process gives farmers a clearer report on yield performance without the data noise from harvesting conditions.

Raw points to cleaned in clumps

One of nine map outputs in the Web App is the Swath Map. We include the swath map so farmers can see corrected yield points across the combine header width as it travels the field. The swath map is a great tool for seeing pass-to-pass differences or evaluating side by side trials, and lets the viewer easily see yield results for every pass in the field.

Swath map output

Detecting and addressing errors in data points is an important bridging step to taking data from yield points to a yield map. Understanding the nature of harvesting equipment, harvest behavior and conditions informs how the FarmTRX processing engine works. FarmTRX removes the noise in yield data to help farmers get usable, accurate maps and keeps it easy to get started.

Ontario, Canada – FarmTRX  is now connected with the John Deere Operations Center through the John Deere API Services.

With a FarmTRX yield monitor, farmers with any make or model of combine can get yield maps for their fields. FarmTRX simplifies the process of getting yield data off the combine, generated into usable maps and prepped for export. Seamless yield data transfer from FarmTRX to Operations Center allows more farmers to unify their farm data and get reliable yield data and maps for their fields. Here’s the video tutorial on how to set up FarmTRX to Operations Center connection.

The connection between FarmTRX and John Deere Operations Center is a two-way transfer through API integration, easily managed in the FarmTRX Web Application. Farmers using Operations Center to manage their farm data can feed corrected yield data points from FarmTRX into Operations Center with just a few clicks. Connection between the two products makes it easier for farmers to unify and build the complete picture of their farm data, regardless of the make, model or year of the combine/s.

If you already have an active Operations Center account, getting started with FarmTRX becomes even simpler as you can seamlessly transfer Client, Farm, Field and Boundary information into the FarmTRX Web App as part of the account setup process.  

“Today, we’re seeing more farmers with the latest in seeding, spraying and auto-steer technology on the farm, the investment priority doesn’t often lean towards combines because of their cost and restricted operational time”, remarks Kyryll Druzhynin, Senior Technical Product Manager at FarmTRX.

“FarmTRX offers the solution of enabling older combines for yield mapping without the substantial investment. The product embodies simplicity – a value we’re seeing requested more from farmers regarding their farm data. Integration with John Deere (Operations Center) is a step towards completing the agronomic data loop for many. Adding yield data to complement the tools in Ops Center lets more farmers compare, analyze and quantify their field data”.

Kyryll notes that “FarmTRX not only showcases yield data but also provides insights into moisture levels, operator details, speed, header specifics, elevator, time and machine data. Users gain info not only on the field but on specific performance zones within the field”. This marks the beginning. We appreciate that yield is just one piece of the agronomic puzzle and making sure yield data remains easy to get and easy to share is a clear priority for the farmers we serve. Therefore, we’re working on growing our integrations list in the near term.

The FarmTRX Web Application is our cloud-based platform where farmers can view, query, interact with and export their collected yield data. The farmer registers their account by entering information about their operation such as preferred units of measurement, headers and crop types (including any custom varieties) that will be harvested. The Web App is the place where all recorded yield data will sort and land once it is synced from the Yield Monitor.

Here’s what happens to your yield data once it is uploaded to the Web App…

Auto-Yield Correction

FarmTRX leverages years of GIS mapping and tracking expertise to offer farmers an automated data processing engine that actively detects, filters and corrects yield points collected off the combine. After yield data is processed, farmers receive a total of 9 different map options to view their yield data. The Web App processor leaves cleaned, accurate yield data and maps that show true yield performance with high-level fidelity.

Yield maps will begin to process immediately upon data sync to the Web App. It is normal to expect maps to be finished processing within an hour, often times much quicker.

Automated yield mapping and data correction through the Web App is included for the first year after purchasing the Yield Monitor. After one year, the farmer has the option to renew automated yield mapping and other features of the Web App (integrations, post-harvest calibration, etc.) by paying a yearly fee based on harvested acres in the account.

The Web App processor identifies and corrects typical harvest affects such as:

• Headland turns
• Hills
• Multiple combines with different calibrations
• Unloading points
• Sharp starts and stops due to plugging

What If I Don’t Want Yield Maps?

Farmers who don’t want to upkeep their access to all features of the Web App (including automated yield mapping) will always have full access to their raw, unedited yield data recorded by the monitor. Farmers are the sole owners of their yield data and have unlimited ability to view, query and export raw data as .shp files. However, they will have to bring it into another farm management system or agronomist to have it corrected into accurate, agronomy-ready yield maps.

Auto-Correction In Action

On a busy harvest day, FarmTRX farmers don’t need to worry about an incorrect or incomplete calibrations affecting their ability to get accurate yield maps for the field. Rest assured knowing the Yield Monitor is always recording relative differences in grain volume travelling on the elevator paddles while harvesting. Calibration errors can be corrected after the fact via post-harvest calibration.

Here’s an example of what happens to yield data automatically in the Web App with FarmTRX Premium. This farmer in Australia calibrated in-field three separate times. A few user errors with each calibration caused the raw yield map to look chopped up, with total field variability blanketed over (left).

Fortunately, miscalibration does not affect the Yield Monitor’s accurate recording of relative yield in the field. In the Web App, the three calibrations were automatically interpreted and corrected into one true representation of variability across the field (right). By adding the total field takeoff in the Web App, recorded yield variability was associated to real bu/ac. The end result? Accurate, usable yield maps.

IN THE FIELD AT HARVEST

Crop Calibration:

Crop calibrations are done by volume or weight. A volume sample (like harvesting a hopper–full) will get you within +/- 5% bushel accuracy. For greater accuracy, use a larger representative sample and a weight measurement (using weigh wagons or truck/grain cart with scales). The most accurate will be adding the elevator field totals directly into the Web App after harvest and post-harvest calibrating the recorded data.  

1. Set the processing delay
   1. Press “Edit Delay” under Device Calibration in the FarmTRX mobile app 
   2. Enter the processing delay (time from cutting head of the combine to sensor measurement, eg. 10 seconds or time from cutting head to sensor reading/value display). 
2. Complete crop calibration 
   1. When you know you can get a measurable (by volume or weight) sample of the field, navigate to the Crop Calibration page from the main menu of the mobile app 
   2. Ensure all listed conditions are met and press their associated checkbox  
   3. Press “Start Calibration” and begin harvesting 
   4. Press “Pause” as soon as harvest is unloaded. Enter the measured weight and/or volume along with measured moisture (if available/applicable) 

When Perry Casson wanted to begin variable rate trials on his 2,400 acre Saskatchewan farm, he knew he would need a yield monitor to outfit his combines and validate the results. The problem was that new machinery carried a significant price tag and technology that was fast becoming outdated. A yield monitor that was affordable, leveraged the resources he already had and could be added to his various model combines was missing from the precision ag market.

Perry catches up with Angela Kokott from 770 CHQR Global News Calgary to discuss some of the ideas and motivations that led to FarmTRX and why the system is being adopted on farms across the globe.

Listen to the full interview here:

 (Full transcript below.)