Field Evaluation of a Spinner Disc Variable-Rate Fertilizer Applicator
J.P.
Fulton, S.A. Shearer, G. Chabra and
S.F. Higgins
Fulton, J.P., S.A. Shearer, G. Chabra and S,G. Higgins. 1999. Field Evaluation of a Spinner Disc Variable-Rate Fertilizer Applicator. ASAE Paper No. 991101. Annual International Meeting, Sheraton Centre, Toronto, Canada, July 18-21.
Abstract
The popularity of spinner spreaders for application of granular fertilizers and agricultural lime along with increased interests in variable-rate technology has raised concern about application accuracy and distribution of these spreaders. This investigation analyzes a variable-rate spinner spreader, equipped with DGPS and a variable rate control system, to assess its distribution accuracy using a 13 by 13 matrix of collection pans and following test procedures outlined in ASAE Standard 341.2. Uniform and variable rate tests where performed to characterize the application variability of the spreader and test the effect of rate changes via GPS control. From the collected data, uniform and variable rate application models where developed. A sigmoidal function was used to describe the rate change from a 56 kg/ha to 168 kg/ha application rate while the average transverse distribution pattern was used to model the uniform 56 and 168 kg/ha rates. The resulting models were then compared to the actual collected distributions. It was found that the models did a good job of projecting the actual application rates for uniform and variable-rate application. The intent is to use this model to simulate variable-rate application for estimation of application errors.
Introduction
Economic and environmental pressures are causing the agricultural production sector to seek more competitive methods of producing food and fiber products. With the development of the global positioning system (GPS) and variable-rate technology (VRT), precision farming is now a common place on many farms. The combination of GPS and VRT has the potential to improve productivity and profitability while conserving and protecting our natural resource base.
Traditional methods of fertilizer and chemical application tend to treat all areas of the field the same, regardless of variation. With the traditional approach, soil cores are pulled at random throughout a field and mixed into a single composite sample. These samples are then analyzed and fertility recommendations made for single nutrient application levels. Nutrients (phosphorus, potash and nitrogen) or agricultural lime are then applied in broadcast fashion to the entire field. Prior to the development of precision agriculture, the variability within many fields was accepted as a factor that producers could do little to correct. And, in fact this variability was quantified for the sake of increasing recommended fertilizer applications rates to account for this variability, and to insure that the crop is not limited by these nutrients.
Searcy (1995) defined Site-Specific Crop Management as "the use of local soil and crop parameters to make precise application of production inputs to small areas within similar characteristics." Spatial variability occurs with the respect to many parameters, such as soil type, fertility, slope, etc., that affect crop production. There is the potential to vary production inputs (fertilizer or crop seed) as a function of location in a field by using VRT. Site-specific management utilizes intensive soil sampling, geographic information system (GIS), computers, GPS and VRT.
In the 1920's, Linsley et al. (1929) outlined a practice to intensively soil sample fields for the purpose of mapping soil pH variation and determining areas for variable lime application. The methodology was not as evolved as today but, it demonstrated the concept of site-specific management. Today, technology provides the potential to make site-specific management a reality. However, concerns still exist with regard to whether site-specific management is profitable when compared with traditional field-average application practices. Application distribution uniformity and accuracy are important properties to assess VRT spinner disc fertilizer spreaders. Therefore, tests must be conducted to assess and characterize spread variability of a spinner spreader and then mathematically model uniform and variable-rate application of dry fertilizer and agricultural lime. The model can then be used to predict overall application efficiency and estimate the deviation from the desired spread pattern of granular materials by a particular spinner spreader during field operation. This will help establish and refine the acceptable management grid resolution based on the truck's limitations.
Objectives
Background
Site-specific management has allowed farmers to start managing their fields on a much smaller resolution rather than on a whole field basis. This approach to managing nutrients has three advantages over traditional approaches; agronomic, economic and environmental. The most visible precision farming tools are the VR controllers on application equipment. As with any equipment, the question always arises about how well are nutrients being really applied in the field? With variable rate equipment more complexity is introduced due to rate changes when covering a field. Therefore, methods need to be designed to test the accuracy of variable rate equipment.
Schueller (1989) described liquid fertilizer mixing and flow control to minimize material transport lag times. He determined that rate and mixture variation are improved by the following conditions: 1) flow control of each system component, 2) quick response times of the pumps and valves involved, 3) connecting hoses should be as short and small in diameter as possible, 4) adequate mixing must be provided, and 5) the mixer should be as close as possible to the nozzles. It was also found that flows could be controlled by varying the pump speeds or the re-circulation flows.
Smith, et al.(1990) interviewed several farmers who have switched to fertilizing based on soil unit needs. One farmer saved over $18,000 on 400 hectares (1,000 acres) of corn when compared to the previous cropping year using conventional fertilizer application practices. Another farmer built his own variable rate fertilizer system for $1000, which has saved him more than 50% in fertilizer costs. A third farmer fertilized his soil on a two-hectare (five-acre) grid, based on soil maps from aerial maps scanned into his computer. He saved 138 to 165 kg/ha (125 to 150 lb/acre) of nitrogen using the precision application system.
Whitney, et al. (1995) developed a spreadsheet simulation to predict the dynamic performance of a variable-rate applicator that applied fertilizer based on on-the-go measurements of nitrogen using the plant-nitrogen-sensor-index (PNSI) method. The authors used this method to predict the error of nitrogen application of the actual machine and the associated costs. The elements evaluated for their contribution to error were quantilization, control valve speed, valve operating time symmetry, ground speed, and field element size. It was found that the weighted binary valves produced greater error than linearly proportional valves due to quantilization error. Valves with asymmetrical on-off times had a high error. The optimal least-cost field element length for 0.76 m nozzle spacing was found to be 0.75 m. The precision method of application was found to save $265.85/ha ($107.40/acre) over conventional N application.
Reichenberger and Russnogle (1989) described a system that simplifies precision fertilizer application. The system utilizes a laptop computer and a fifth wheel to determine placement in the field. An application rate map was stored in laptop memory and machine application rates were controlled with feedback from the fifth wheel. The unit was reported to be ready for modifications that would allow application of liquid fertilizer, chemical injection, planting, and the development of yield maps.
Application accuracy is an important property to assess the VR spinner spreader fertilizer application system. The coefficient of variation (CV) is typically used to characterize the quality of spread distribution by a spinner spreader. The lower the CV the more uniform the distribution pattern. Typically, the CV varies from 5% to 10% for the transverse spread pattern of a spinner spreader, but this variation probably doubles (Parish, 1991) under field conditions due to various field conditions such as terrain irregularities. However, it expected that the CV could increase to 15% to 20% when conducting field tests (Sogaard and Kierkegaard, 1994).
The ASAE standard (ASAE S341.2 Procedure for measuring distribution uniformity and calibrating granular broadcast spreaders in ASAE, 1997) describes a uniform method of determining performance data on broadcast spreaders for granular materials and provides a basic test procedure to compare spreader distribution patterns. Details, such as test setup, collection devices, test procedure, distribution, effective swath width, determination of application rates, are given clear definitions in the test procedure section in the standard.
Olieslagers et al. (1996) described the fertilizer distribution of a spinning disc spreader. Many parameters including orifice position and angular speed of disc impact the distribution pattern of disc spreader. VRT application, accomplished by changing the mass deposition rate on spinner discs, leads to a fluctuating spreader pattern which results in large deviation from the intended application rates. Olieslagers et al. (1997) suggested that continuous change to various spreader adjustments might be needed to maintain a uniform distribution pattern when changing rates on-the-go. They also stated that future work should be concentrated on the dynamic response of the fertilizer spread pattern when changing material rates on-the-go.
Chaplin et al. (1995) investigated the distribution of dry material during field application. They described a methodology based on American Society of Agricultural Engineers (ASAE, 1997), and did testing for a single-disk mounted fertilizer spreader. Pettersen et al. (1991) investigated how the distribution pattern of a twin-disc spreader was influenced by fertilizer particle size. They provided a detailed test method for collecting fertilizer samples and used interpolation techniques and computer graphics to get continuous distribution patterns.
Overview of Spreader and Control System
The Biosystems and Agricultural Engineering Department at the University of Kentucky currently maintains a custom built variable-rate fertilizer truck that uses spinner discs to apply granular products. The spreader box was modeled after a Newton Crouch spreader. It was fabricated in-house and has the capability of varying the application of two products at once. The are two separate compartments intended to contain two different nutrient sources or fertilizer blends. However, for this investigation, only the rear compartment will be used since applying only potash. The rear compartment feeds material by a traditional apron chain to the two spinners as one would see on a Newton Crouch bed. A gerotor motor powers the apron chain drive. Flow to this motor is controlled using a Source Fluid Power motorized control valve along with additional valves to control the portion of fluid by-passed for speed control. The truck is equipped with a Midwest Technologies Inc., TASC 6200 console controller automatically controls the opening and closing of the motorized control valve for varying the speed of the metering apron chain and thus varying material flow from the rear compartment.
The TASC 6200 is interfaced to a laptop computer through an RS-232 serial data link. The drive motor is equipped with a speed sensor as part of the feedback controls for the system. Agris's FieldLink is used for communicating the required application rate to the TASC 6200 through the data link. For position determinations, an Omnistar 7000 receiver using C-band correction is mounted on top of the truck and is linked to the laptop via another RS-232 serial port. FieldLink provides the ability to record the test site boundary and establish a rate change line or two polygons for testing rate changes from low to high or vise versa.
Spreader Testing
Field tests were conducted to evaluate the deposition from the variable-rate spinner disc spreader applying granular potash fertilizer. Application rate, distribution and the effect of rate changes via DGPS were evaluated by modifying ASAE Standard S341.2 to include a two-dimensional array of collection pans. Test cases to investigate the application of potash included: 1) fixed-rate application at a low rate; 2) fixed-rate application at a high rate; 3) variable-rate application from a low to high rate; 4) variable-rate application of a single nutrient from a high to low rate. Single pass tests were performed for each of these cases to assess application accuracy and characterize the rate change. The effect of overlap on spread uniformity was evaluated by performing multiple parallel passes for each test.
Tests were conducted in-situ at uniform and variable application rates, spreading murate of potash to evaluate the spreader truck's distribution uniformity and assess the accuracy of application. All field tests were conducted on days when sustained wind speeds were less than 8 kilometers per hour at a height of 2.5 m (5 ft) above the ground and the slope of the testing site was less than 2 % (ASAE Standard: ASAE S341.2). All tests were run with the hopper filled to approximately 40% to 50% capacity as defined as ASAE S341.2.
ASAE S341.2 was followed to fabricate aluminum collection pans for testing the spreader. The pans measured 40.6 cm (16 in.) wide, 50.8 cm (20 in.) long and 10.2 cm (4 in.) in height. An aluminum divider with a 10.2 cm by 10.2cm (5.1 cm (2 in.) height) grid was also fabricated to place inside each tray to reduce material from ricocheting out of the tray. The test procedure included a two-dimensional array of collection trays to evaluate the effect of rate changes via DGPS control. The size of the array normal to the direction of travel will include 13 collection pans, spaced evenly across the anticipated distribution pass. Along the axis of travel will be 13 evenly spaced row of collection trays. Figure 1 and Figure 2 show the two different 13 by 13 pan layouts along with the pan spacing used for testing the applicator. Spacing was determined as a function of response speed of the controller and ground speed of the applicator so when performing the rate change test, the full change was captured before exiting the pans. The positions of the four corner trays was measured out first and recorded for orientation. The remaining trays were referenced from the positions of these corner trays. For the multiple pass tests, two longitudinal rows of trays were omitted to allow area for the spinner spreader truck to pass over the pans as shown in Figure 2.
Many factors directly affect fertilizer distribution and accuracy of application such as systematic errors associated with machine calibration and metering efficiency. To minimize the combined effect of these factors and achieve accurate fertilizer distribution, the spreader truck was calibrated prior to performing the tests. The best distribution was achieved by adjusting the rear divider forward and backwards until a uniform transverse distribution was achieved for an average application rate of potash. A one-dimensional array of 13 pans was during this process. Spinner speeds were set at 650 rpm with the gate opening positioned at 4.4 cm (1.75 in) above the floor of the bed. The truck was operated in second gear and at 1800 rpm which results in an 20.4 km/hr (12.7 mph) field speed.
Test application was made at 25% and 75% (ASAE 341.2) of the maximum application rate as recommended by the University of Kentucky’s Lime and Fertilizer Recommendations for murate of potash (AGR-1, 1998). AGR-1 recommends a maximum application of 134 kg/ha (120 lb/ac) for potash (K2O). Murate of potash is 60% K2O (0-0-60) therefore 58.0 kg/ha (50 lb/ac) of material represents the low rate and 168.1 kg/ha (150 lb/ac) was used for the high rate. The center of each swath was flagged so that the driver had a visual guide when traversing the test site. Potash was collected within the swath width of the spreader using the trays for each pass by the spreader. Particle samples were collected, bagged and labeled for each of the field tests. All samples were then weighed back in the lab and recorded. Distribution plots along with surface plots, using the program Surfer, were generated for all test cases.
Results and Discussion
Figure 3 and Figure 4 present the uniform single application surfaces for the 56.0 and 168.1 kg/ha. The 56.0 kg/ha application surface appears rather uniform with the exception of a few irregularities. This would be expected from a spinner spreader since they are known for their nonuniformity of spread. The apron chain does not perfectly meter potash continuously. Instead, it tends to fall off the apron chain in small clumps. Both Figure 3 and 4 shows evidence of this when looking at longitudinal cross sections. The surface plot in Figure 3 does hint that a W-shape pattern (less material at the center of the pattern) is occurring. In fact, the plot of the mean transverse application rate for the uniform 56.0 kg/ha, presented in Figure 5, shows a slight decrease in material at the center of the pattern. For the most part the spreader does an acceptable job at the low rate.
Figure 4 shows the same type of irregularities as Figure 3, but tends to show a shift in pattern distribution. The pattern moves to a W-shaped pattern where more material is being applied at the pattern's center then on either side of the center. Figure 6 exhibits the plot of the mean transverse application pattern and demonstrates this W-pattern. Both the high and low tests were performed with the same settings on the truck. The shift in pattern coincides with the conclusions of Olieslagers et al. (1997) that continuous changes in the spreader settings (divider position, spinner speed, ect) are needed to maintain a uniform distribution.
The mean transverse distribution application pattern for the 56.0 and 168.1 kg/ha uniform single pass are presented in Table 1 along with the standard deviation and coefficient of variation. The CV is lower at the center of the pattern, but gets larger towards the edges. This can be contributed to the very small amount of material collected at the outside pans. Usually these pans would only have a few particles in them so an additional particle or two can quickly increase the CV. This explains the higher variation in the low rate test compared to the high rate test. Looking at only the center pan and three on either side (represents the effective swath width), the CV appears acceptable with the majority of the CV's for each test right around 20% with a few being higher. The test area was an actual hay field and some depressions across it. Therefore, the tests closely represent what the spreader truck would encounter in Central of Kentucky. Sogaard and Kiekegaard (1994) stated that the CV would definitely increase under field tests from the desired 5-10% to 15-20%.
The mean transverse application rate for the uniform low and high rates were used to model the application of both and are shown in Figures 7 and 8. These should do a good job when wanting to generate a field application model for the desired and actual application of potash. The strength of the relationship for both uniform rates will be discussed later in this section.
The uniform test plots for multiple passes are presented in Figure 9 and Figure 10. Both surfaces show a wide variation in application distribution with center and outer pans receiving more material that the others. Again, the nonuniformity of application is expected due to the nature of spinner spreaders. Table 2 shows the desired application rate along with the statistical information on the actually applied potash. The actual applied material is close but slightly larger than the desired rate. Both tests show a large range in application rates with a coefficient of variation of 21% for the low rate test and 20% for the high rate test. The spreader is applying the desired rate but not uniformly distributing material.
Figure 11 presents the actual application surface for a rate change from 56.0 kg/ha to 168.1 kg/ha. The zero longitudinal distance denotes the desired transition in the rate change. For this particular test, the delay time in FieldLink was set at zero so that the system latency for the spreader and rate change could be captured when the truck passes over the pans. Determining the latency of the system is not an objective of this paper but, the delay time can always be adjusted later to shift this surface to represent what is happening when applying potash on fields and setting the true time delay in FieldLink.
The rate change surface in Figure 11 demonstrates what was seen in the uniform tests. The spreader does a good job at 56.0 kg/ha of distributing potash but as the rate change occurs, the pattern shifts from a Gaussian shape to a W-shape. The same type of pattern shift occurs when changing from 168.1 kg/ha to 56.0 kg/ha.
The next step was to model the variable rate application process shown in Figure 11. To simplify, symmetry was assumed for transverse distribution. Therefore, equal distant longitudinal rows from the center pans were averaged to create seven longitudinal data sets to represent the rate change dynamics. The center row was used as is and not average with any of the other rows. All pans in the last longitudinal row collected no material and were set to zero application rate. Sigma Plot 4.0 was used to fit a sigmoidal curve to the other six data sets. To fit a curve each data set, a four-parameter sigmoidal function was used:
Figure 12 shows the results for fitting the rate change dynamics to the average of the ± 2.67 m data points. The curve has an R2 equal to .98 which shows a good fit. Table 3 presents the results for all six data sets along with the calculated parameters and R2 values. Constants were calculated for parameters b and xo since all their values were fairly consistent and to simplify the equation. Each equation was then applied to the six longitudinal rows to calculate a predicted application rate for each pan. These predicted values were then used to create a surface plot which models the rate change from 56.0 kg/ha to 168.1 kg/ha (Figure 13). At first glance, Figures 11 and 13 are very similar except that Figure 13 has smoothed out many of the irregularities seen in Figure 11. Using the coefficient of correlation to compare the actual to the predicted data shows a good fit for the model at a .96. Table 4 contains the coefficient of correlation for the rate change as well as comparing the data for each of the single-pass uniform application tests. From the results in Table 4, it appears a model can be developed that does a good job of approximating the actual distribution for this spreader truck. The procedure can be applied to model the rate change from 168.1 kg/ha to 56.0 kg/ha. From this, one can record the spreader truck's traverse over a field using DGPS and model the actual potash application across the field. Similarly, the desired application can be modeled and then compared to the actual modeled application and determine the application error.
Summary
This investigation was conducted to assess the accuracy of a variable rate fertilizer applicator and determine whether uniform and variable rate application of potash could be modeled. Uniform and variable rate tests were performed using a 13 by 13 matrix of collection pans to gather material spread by a spinner spreader truck as it passes over the pans. From the collected data, uniform and variable rate application where modeled. A sigmoidal function was used to describe rate change from 56.0 to 168.1 kg/ha whereas the average transverse spread pattern was used to model the uniform application at 56.0 and 168.1 kg/ha. Comparing the modeled application to the actual material gathered for each test show that the modeled application, of uniform rates of 56.0 and 168.1 kg/ha and changing the rate from 56.0 to 168.1 kg/ha, did a good job of projecting the actual distribution. Further, the comparison between 56.0 and 168.1 kg/ha distribution patterns show that there is a needed change to the spreader adjustments to maintain a uniform pattern. The spreader truck did a good job of applying material at 56.0 kg/ha but at the higher rate of 168.1 kg/ha, the spreader applied material in a W-pattern.
Further, field-testing will enable the development of a simulation model for predicting application accuracy by use of the recorded truck's DGPS application traverse. In return, the field investigation allows for characterization and modeling of variable-rate application of granular materials with the ability to assess deviation from the modeled desired application. This will help determine and refine the acceptable management grid resolution based on the spreader truck's limitations for precision agriculture. Determining the spread variability of such equipment furnishes their inherent limitations, therefore lending to the appropriate selection of grid cell size for future soil sampling.
The actual spread model and the calculated application error can also be used to assess an operator's performance after spreading granular fertilizer. Overlap and under-lap can be determined to see if the driver error seriously affects application accuracy. An operator might be consistently driving under the correct swath width, thereby increasing material overlap. Therefore, the width of spread can be adjusted to match the operator's parallel passes in the field and improving application as well as adjusting the grid cell size to fit the operator's performance.
Future work will consist of modeling other granular materials, specifically phosphorus and lime. Once this information is compiled and spread patterns modeled, the spreader truck will be operated on several fields to test the model by recording the driver's traverse during operation. Additionally, two-nutrient application by the two-bin applicator will also tested using a similar procedure as described in this paper. In the same way, modeling the application of two nutrients when applied at the same time will be critical to assess spinner spreader accuracy.
References
AGR-1. 1998. Fertilizer and lime recommendations. University of Kentucky, Lexington, KY.
ASAE Standards, 44th Ed. 1997. 341.2. Procedure for measuring distributionuniformity and calibrating broadcast spreaders. ASAE, St. Joesph, MI.
Chaplin, J. and E. Roytburg. 1995. Measuring the spatial performance of chemical applicators. Site-Specific Management for Agricultural Systems. pp.651-669.
Linsley, C.M. and F.C Bauer. 1929. Test your soil for acidity. Circular 346. Univ. of Ill. Agric. Experiment Station, Urbana, IL.
Olieslagers, R., H. Ramon, and J. De Baerdemaeker. 1996. Calculation of fertilizer distribution patterns from a spinning disc spreader by means of a simulation model. J. Agr. Engr. Res., 63(2): 137-152.
Olieslagers, R., H. Ramon, and J. De Baerdemaeker. 1997. Performance of a continuously controlled spinning disc spreader for precision application of fertilizer. Precision Agriculture 1997. BIOS Publishers Ltd. pp. 661-668.
Parish, R.L. 1991. Effect of rough operating surface on rotary spreader distribution pattern. Applied Engineering in Agriculture ASAE, 7(1):61-63.
Petterson, J.M., J.A. Svendsen, and S. Ovland. 1991. A method of studying the influence of fertilizer particle size on the distribution from a twin-disc spreader. J. Agric. Engr. Res. 50: 291-303.
Reichenberger, L. and J. Russnogle. 1989. Farm by the foot. Farm J. Mid-Mar, 113(6):11-15.
Searcy, S.W. 1995. Engineering system for site-specific management: Opportunities and Limitations. Site-Specific Management for Agricultural System. 603-612.
Schueller, J.K. 1989. Spatially variable fluid fertilizer applicator design concept. http://www.spos.gsfc.nasa.gov/eos_observe/11_12_94/p24.html.
Smith, D., D. Seim and C. Finck. 1990. Micro-managed fields save fertilizer. http://www.sri.bbsrc.ac.uk/webdocs/research/strat_res/preagric/svfo.html.
Sogaard, H.T. and P. Kierkegaard. 1994. Yield reduction resulting from uneven fertilizer distribution. Trans. of ASAE. 37(6): 1749-1752.
Whitney, R.W., D.L. Dludlu, L.J. Swenson, D.G. Hopkins, and W.C. Dahnke. 1993. Variable fertilizer application based on yield goal, fertility, and soil map unit. J. Prod. Agr. 6(2): 255-261.