Back to Contents6.1.1 Assessment of Grain Yield Monitoring Accuracy
Investigators: Scott Shearer, Richard Barnhisel, Sam McNeill, Tom Mueller, Larry Wells and Steve Higgins
Yield monitoring technologies have been introduced to Kentucky grain producers and are now becoming a mainstream tool to help these producers improve their management skills. The goal of yield monitoring activities should be to accurately measure yield variations within a field. It is this variation in yield and the ability to quantify it that has sparked the interest of producers. Once the extent of these variations is known, the next question asked is how will producers manage this variability to their benefit? Increasingly, these same producers rely on yield monitor data to validate management approaches. By accepting yield data as accurate, and generating maps that describe yield variations within a field, producers can identify the scope and magnitude of production problems.
Commercially viable yield monitoring systems utilize impact-based sensors to estimate the mass flow of grain through the combine. The mass flow rates are tied to GPS field position, ground speed, harvest width and grain moisture content to estimate yield. Most attempts to verify the accuracy of yield monitors have focused on integration of the mass flow rate in the combine to determine the total mass of grain harvested. These values are then compared with scale weights of trucks or grain wagons leaving the field. Unfortunately, error terms are also integrated using this approach thereby masking deviations of mass flow sensor readings from actual mass flow rates. As an example one might envision two yield maps where the total mass of grain harvested from either field is the same. However, the range of variation of yield between either map is substantially different, illustrating problems that might arise from calibration and comparison techniques where mass flow errors are integrated.
Laboratory Investigations - Yield monitoring, the process of relating grain mass flow rate and moisture contents to a geographic location, is susceptible to numerous measurement errors. Shearer et al. (1997) investigated and compared yield estimates from nearly identical combines operating within the same field. The researchers found the yield data from one combine to exhibit twice the variation in crop yield as did data obtained from the second combine. It is this variation in yield that producers seek to quantify. In short, proper calibration of the sensing devices is essential if grain producers are to have confidence in the yield maps that are generated using this technology. Some of the physical properties of grain thought to influence calibration accuracy include moisture content, test weight, kernel or grain size, and surface friction characteristics. Just as important are physical features of the combine and field conditions at harvest. An example of a combine feature that may influence data accuracy is wear associated with the clean grain elevator paddles, and their interaction with the impact-style mass flow sensor. Field conditions, such as terrain, may cause the combine to pitch and yaw, thereby influencing acceleration of the grain at the top of the clean grain elevator, and the resulting forces measured at the impact plate. Acceleration variations introduced under these conditions may alter the validity of the calibration.
External influences such as variations in grain properties or field conditions, and their effect on sensing technology, provide the impetus for the development of an indoor yield monitor test facility. Since field conditions cannot be duplicated precisely, it is desirable to simulate the harvest environment to the point where the influence of grain physical properties and field conditions can be quantified. And therefore the intended objective of the proposed test facility is to replicate, to the extent possible, field conditions in a controlled indoor environment. Such a facility permits year-around testing and precise control of the variables of interest.
A laboratory test facility has been constructed at the University of Kentucky. This test facility consists of clean grain elevator components from a late model combine to simulate the movement of grain from the threshing/separating unit to the grain tank. Primary components include the clean grain elevator and bin-loading auger, along with mass flow and moisture sensors. Two 17.5 cubic meter (500-bushel) capacity grain bins were acquired for the project. One bin is equipped with load cells to monitor grain feed and discharge rates from the clean grain elevator. Feed rates to the base of the clean grain elevator will be controlled via a PLC (programmable logic controller). Drag chain conveyors located at the bottom of the supply bin are powered by electric motors with speed control provided by the PLC. Step, ramp, and sinusoidal input functions will be programmed into the PLC to simulate changes in mass flow rates that might occur under actual harvest conditions. A serial port on the PLC permits communications with, and data transfer to, a PC (personal computer). A steel framework has been constructed to support the various combine components. This framework includes a pivoting base to permit the combine components to be tilted fore and aft, or from side to side, to simulate uneven terrain.
The laboratory test facility will be used to investigate the effect of moisture content, grain type, threshing dynamics, and combine dynamics on mass flow accuracy, and the appropriateness of calibration procedures. The brands of yield monitors to be evaluated will include John Deere Company's GreenStar and Ag Leader YM 2000. All testing will be conducted in cooperation with manufacturers of the devices, with the intent of improving the accuracy of data generated with these systems, and not comparison of these products.
Laboratory testing will be limited to grains including corn, wheat and soybeans, with the majority of efforts directed at corn because of the range of physical properties and yields possible at harvest. Tests to be conducted will include 1) accumulated mass for fixed mass flow rates, 2) accumulated mass for variable mass flow rates, 3) instantaneous yield for fixed mass flow rates, and 4) instantaneous yields for variable flow rates. Fixed-rate tests will consist of five to seven mass flow-rate increments ranging from 0.0 and 30.0 kilogram per second. Dynamic testing will be conducted using a sinusoidal input with peak to peak mass flow rate amplitudes ranging of 0.0 to 30.0 kilograms per second. All tests will be repeated for corn at angles coinciding with hillside operation of the combine on side-slopes of 0.0, 3.0, 6.0, and 12.0 percent. Sinusoidal input periods of 5.0, 10.0, 15.0, 30.0 and 60.0 seconds will be used. All tests will be conducted for a 5.0-minute duration. All laboratory investigations on corn will be conducted at a grain moisture content of 15.5 + 3.0 percent. Selected testing will be conducted at grain moisture contents ranging from 12.5 to 30.0 percent to quantify and confirm the effect of moisture content on yield prediction.
Data will be downloaded from the yield monitoring devices and compared with weigh-bin masses and grain metering rates to characterize the accuracy of the yield monitoring devices. Dynamic data will be reported using root-means-square errors for 1.0 second sampling intervals of the five-minute duration of laboratory tests. Similarly, actual total grain mass values versus accumulated mass values will be reported. These data will be combined with GPS positioning data obtained under Objective 3 to perform a complete error analysis on yield determination.
Field Investigations – Field investigations of yield monitoring accuracy are needed to help producers gain confidence in the data generated by these devices, and to realize the limitations of the technology. Further, this element of the research activity will enable the field validation of practices learned in the laboratory investigations. Integration of the mass flow rate eliminates time varying errors. Yield monitor calibration must be conducted so that variations in grain yield at some predetermined minimum resolution are preserved. Therefore the purpose and scope of the field-work is to compare variations in yield monitor data versus hand harvested plot yield data for corn. Ideally, it is the measurement of yield variation, or more specifically relative yield, that is important to producers. Once relative yield is measured these values can be scaled to obtain absolute yields. The scale factor between absolute and relative yield is simply the scale weights divided by the integration of the mass flow rate over the total harvested area. The scale weight is obtained by weighing trucks and or wagons as they arrive at the elevator.
During all three years of the project two or more producer fields will be chosen for the investigation. The intent is to select producers and fields such that several makes and models of combines and yield monitors are included in the investigation. Only the harvest of corn crops will be considered as yield monitors are exposed to the greatest range of variation in grain physical properties when compared with wheat and soybean harvests. In addition corn harvests generate the condition of maximum mass flow within the clean grain elevator.
Immediately preceding the harvest, test plots will be hand-harvested at random within the target fields. Hand-harvested plots will consist of an area equivalent in width to one-half the header width of the combine, and of sufficient length to constitute 0.004 ha. Approximately 30 hand-harvested sites will be selected in each field to estimate the mean and standard deviation of the yield distribution. Hand-harvested sample identities will be preserved by assigning GPS coordinates obtained from each plot to the respective samples. Grain samples will be dried on the ear, shelled, graded and weighed. Researchers will work with combine operators prior to harvest to insure that appropriate calibration practices for the yield monitor were followed. The fields will then be harvested without intervention by researchers, with the exception of obtaining scale weights for all grain leaving the field.
Hand-harvested and yield-monitor derived yield estimates will be compared. Estimates of yields surrounding the hand-harvested plots will be obtained by selecting a minimum of 100 yield points at a predetermined radius around the GPS coordinates of the center of the hand-harvested plots. These data will be averaged to obtain a yield monitor estimate for the same site. Means and variances will be compared statistically between the hand-harvested and yield-monitor yield estimates to establish the accuracy of the yield monitoring device. These data will also be used to help establish minimum management grid resolutions as described under Objective 3.
Results from both the field and laboratory portions of the yield monitoring investigations will be published in refereed journals as warranted. In support of extension efforts in Kentucky, bulletins and circulars will be published to guide producers in the establishment of yield monitoring capabilities. The investigations as outlined above seek to resolve the issue of yield monitor accuracy for Kentucky producers. In addition guidelines will be developed to define limits when using yield monitor data for quantifying and comparing crop performance for on-farm research.