Variable-Rate Herbicide Application
INTRODUCTION
With current cultural practices (e.g., no-till planting), large quantities of herbicides are applied to U.S. agricultural lands. U.S. farms apply approximately three billion pounds of agricultural pesticides annually (Compressed Air, 1988) at a total cost of $26 billion. Of these chemicals nearly 45% are herbicides. This continuing level of chemical application has the potential to increase production cost, reduce food quality, and create potential environmental contamination. These problems have drawn the attention of the public to the use of agricultural chemicals in crop production. On the other hand, weeds pose a threat to food and fiber production. Losses in both yield and quality of crops due to weeds, including the cost of weed control, constitute a tremendous economic problem in agricultural production. The average annual economic loss in the U.S. caused by weeds with current control strategies has been estimated at $4.1 billion. If herbicides were not used, this loss might be as much as $19.6 billion, nearly a fivefold increase (Bridges, 1992).
Enabling technologies that may be useful in resolving conflicting production goals are the Geographic Information System (GIS), Global Positioning System (GPS) and direct injection. With direct injection sprayer systems the chemical concentrate is maintained in a reservoir separate from the water carrier. The chemical concentrate is metered into the carrier and mixed ahead of the spray nozzles. Together, these technologies will facilitate field management on a "site-specific" basis instead of using the "field average" methods that are common place today. Farmers will be able to manage fields in accordance with the variability in yield potential, soil properties, and a host of other agronomic factors. This new control strategy, site-specific management of herbicide application, shows promise for the future of agriculture. Sprayer performance will ultimately determine the success or failure of the control strategy.
The main focus of this investigation was on assessing the feasibility of direct injection for variable rate herbicide application. Specific objectives were: 1) to simulate a direct injection system used for site-specific application of preemergence herbicides in corn; and 2) to investigate the effects of injection location, spray tip size, hose size, system pressure, travel speed, and software correction on the overall system performance.
BACKGROUND
Recently, researchers have focused attention on site-specific management of agricultural production. The inputs, for example fertilizer and herbicide, are applied based on variation in soil properties, yield potential, or distribution of weeds. Weber et al. (1987) developed algorithms for herbicide application rates based on soil properties. They stated organic matter content was highly correlated with herbicide bioactivity. Blumhorst et al. (1990) investigated the efficacy of selected herbicides as influenced by soil properties. They recognized that herbicide activity was highly correlated to soil organic matter content, and suggested that herbicide application rates should be determined in accordance with soil properties. Herbicide rate equations were developed as a function of soil organic matter content and percentage clay content. Mulla (1991) suggested using geostatistics and GIS to manage spatial patterns in soil fertility. Ollila et al. (1990) developed a liquid chemical sprayer system with a direct injection module interfaced with a laptop microcomputer. This system was capable of applying specific chemical rates to specific locations based on a predetermined rate map. Qiu et al. (1994) developed site-specific management strategies for herbicide application based on soil properties, weed competition, and yield potential of the field. Application rates of herbicides were determined based on soil properties and the economics of application at each location. Up to one-half of the preemergence herbicides applied in corn could be saved by using site-specific application practices. According to Qiu et al. (1994), potential savings would come from reducing or eliminating herbicide application in areas of a field that have low organic matter content and where weed control has not been a problem historically. Increased yield would be possible in areas of a field where ineffective weed control could be offset by increased application rates.
With the development of new technologies, the traditional pesticide application system, a boom sprayer, can be modified to accommodate these new control strategies. Traditional sprayers limit the range of application rates and are not suitable for site-specific control strategies. This limitation is imposed as a result of relatively narrow pressure operating ranges. As pressures drop below a specified level, the spray pattern becomes distorted and application uniformity is sacrificed. When nozzles are operated above the recommended pressure range, too many small droplets (diftable fines) are generated. Within the recommended pressure range, discharge increases of less than 50 percent from the lower to upper pressure limits are common. Direct injection systems have been developed during the past decade to eliminate traditional application problems. Advantages of these systems include: elimination of wasted chemical, limited operator exposure, automatic on-the-go selection of herbicide application rates, and control of multiple injection modules with different herbicides. One disadvantage of a direct injection system is that the dynamic characteristics of the sprayer are affected by the time required for the chemical to travel from the injection point to the spray tip.
Rudolph and Searcy (1994) discussed site-specific management strategies for agricultural chemicals. They suggested methods to reduce the effect of time lag through hardware and software design. They concluded that the practice of site-specific management with a direct injection sprayer was practical. Al-Gaadi and Ayers (1994) investigated the performance and accuracy of a conventional ground-driven boom sprayer when operated with and without an electronic sprayer control system. When ground speeds varied within a range between -7.0 and 4.0% of the desired speed, the application rate errors ranged from -18.0 to 5.0% of desired application rate when operated without the control system. Application errors were reduced to a range of -7.0 to 1.0% of desired application rate when the control system was used. Ghate and Perry (1994) developed a technique to control pesticide application rates in direct proportion to ground speed for a compressed air direct injection sprayer. In this system, a radar sensor and an oval gear flow meter were used to output signals proportional to ground speed and pesticide flow rate, respectively. These signals were used as inputs to a data-logging unit that computed a difference between the desired and actual flow rates. This data-logger activated a stepping motor attached to a pressure regulator which adjusted liquid-reservoir air pressure for obtaining the desired flow rate. Laboratory and field testing results of this unit suggested the desired flow rate of a simulated pesticide was achievable within a reasonable number of control program cycles. Way et al. (1992) simulated the chemical application accuracy of injection sprayers accelerating at 1.6 km/h/s from rest to constant speed. Comparisons were made among injection sprayers with conventional and modified plumbing systems, and conventional tank-mix sprayers without sprayer controllers. Results revealed that the ratio of area receiving unacceptable chemical application rates to the total area sprayed was smaller for the tank-mix sprayer than for any injection sprayer.
Stafford and Miller (1993) developed a system for spatially selective application of herbicide to weed patches in arable crops. The system consisted of four subsystems: weed patch detection, real-time sprayer location, field mapping and spray rate control. They noted that grass weeds in cereal crops occurred in the same location year after year. They also suggested a resolution of 2.0 m was achievable from a differential GPS (DGPS) receiver, and the minimum time delay for the system was 4.3 seconds.
Tompkins, et al. (1990) investigated boom flow characteristics of direct chemical injection systems. Three injection positions were investigated: immediately upstream of the pump, immediately downstream of the pump, and at the individual spray tips. The performance of the direct injection system was dependent upon component selection and system configuration. With the injection point upstream of the pump, approximately 3.5 m (138 in.) ahead of the boom, 26 s were required for the chemical concentration at the outermost spray tip to reach equilibrium. The maximum variation at this spray tip was 2.3% of the time-average chemical concentration. For injection downstream of the pump, chemical concentration equilibrium was reached in about 12 s with a maximum range of variation from 5 to 11% of the average concentration, depending on pump type.
Frost (1990) developed a proportional plus integral plus derivative controller for a chemical metering system. He stated the time taken for the metering system to achieve 90% of a step demand was less than 1.0 s. He also investigated the possibility of reducing time lag at the spray tips through changing hose diameters. It was shown that the overall system delay could be reduced to less than 3.0 s for a practical plumbing arrangement on a typical machine.
Summarizing previous work, the performance of a direct injection system depends on many factors including spray tip size, system pressure, ground speed, injection position, and system configuration. The dynamic characteristics of the direct injection system are dominated by the time it takes for the chemical to travel from the injection point to the spray tip. The objective of this investigation was to model and assess the feasibility of direct injection systems for variable rate herbicide application.
For direct injection sprayers, the delay time for changes in chemical concentration at the spray tips affects the performance of the sprayer. The delay time for a direct injection sprayer was calculated using the method provided by Frost (1990):
where Tdelay is the time lag for a system with n conduits, and Qi is the flow rate through a pipe of diameter Di and length Li.
A 24-ha (59-acre) field at the University of Kentucky's (UK) Woodford County Research Farm was used as a source of data for input to the simulation model. The herbicide application rate map developed by Qiu et al. (1994) was selected as the management strategy for herbicide application in corn. Extrazine II 4L was chosen as a representative herbicide to control weeds in the field. The desired application rates for site-specific management of the field are shown in Fig. 1. The grid size used in this figure is 30 m by 30 m (100 ft. by 100 ft.). Because of system dynamics, delayed changes in chemical concentration at the spray tip create errors as changes in desired application rates are input to the sprayer control system.
For the simulation, a constant pressure system was considered to maintain a constant median droplet diameter at the spray tip. To maintain a constant chemical application rate the chemical formulation injection rate must be proportional to ground speed. According to Way et al. (1992), a relationship for computing the formulation deposition rate for variable rate herbicide application at the ith spray tip is as follows:
where HRi is the
formulation deposit rate for the ith
spray tip at time t, V(t - Tdelayi)
is the speed of sprayer at time (t - Tdelayi),
V(t) is the speed of sprayer at time
t, and HRdes(t - Tdelayi) is the desired formulation
deposition rate at time (t - Tdelayi).
Changes in the ground speed of the sprayer are dependent upon many factors including slope and roughness of terrain, gross vehicle weight, and available power. The topography of the field used in this investigation is shown in Fig. 2. For simulation, the effects of slope on ground speed were considered. Slopes of the field were calculated based on the mapped elevation differentials and travel direction. Relationships between slope and ground speed change were determined in a field experiment. Eighteen different slopes were selected in grass-covered fields on the UK campus. A distance of 61.0 m (200 ft.) was measured for each slope, and the start and end of each slope were marked. An electronic distance measuring (EDM) device was used to survey the elevations for the slope tests. A 52 kW (70-HP) tractor was driven up and down the slopes at a fixed throttle setting that resulted in an unloaded engine speed of 1700 rpm. A 1900-L (500-gal.) trailer-type, boom sprayer was attached to the tractor to simulate actual field loads. The spray tank was filled to half-capacity with water. Travel times up slope and down slope were recorded using a stopwatch. This procedure was replicated five times for each slope and direction of travel, and average times were computed. According to these average times and the distance up and down slope, mean velocities were calculated. Linear regression was used to fit relationships for velocity as a function of slope.
A direct injection sprayer as described by McNeill et al. (1992) was used as the basis of the simulation model. The system is shown in Fig. 3. A boom manifold was constructed from three 2.54 cm (1.0 in.) inside diameter solenoid valves coupled with close nipples. The hose from the injection point to the boom manifold was 3.5 m (134 in.) long with an internal diameter of 2.54 cm (1.0 in.). The hose from the boom manifold to the center of left or right sections of the boom was 4.1 m (161 in.) with an internal diameter of 1.91 cm (0.75 in.). The hose from the boom manifold to the center of the middle section of the boom was 1.0 m in length with an internal diameter of 1.91 cm (0.75 in.). The length of the spray boom was 14.5 m (570 in.), with nine spray tips in each of three sections.
Consistent with findings in related work, the plumbing design of boom sprayers plays an important role in determining the lag times for changes in chemical concentration at the spray tip. According to Tompkins et al. (1990), injection upstream of the pump produced uniform chemical concentration from improved mixing, but longer transient time. To reduce system lag time, a hydraulically-driven sprayer pump can be used to replace the original PTO driven pump to facilitate closer mounting to the boom manifold. The injection point can then be moved closer to the boom manifold for injection upstream of the pump. Optimal systems minimize lag period and maintain uniform chemical application rates along the boom. Altering the on-boom hose diameters from the center boom section to the furthest spray tip was considered in the analysis. A computer program was written to compute the pressure loss and corresponding changes in flow rate at each spray tip; time lag was found for the outermost spray tip given spray tip size, spacing, system pressure, hose length, and diameter. Pressure loss along a straight hose was calculated by the following equation:
where P is pressure loss, l and d are the length and diameter of the hose, r is the density of the liquid, v is fluid velocity, and f is a friction factor that depends on Reynold's number. Approximations for the friction factor f are as follows:
where NR (the Reynold's number of flow in the individual hoses) is computed by:
where Kv is the kinematic viscosity of the fluid. For the tee fitting at the boom manifold the pressure loss was computed using the above equation and an equivalent length of 60 m (197 ft.) (Fox and McDonald, 1992). The pressure loss due to the change of diameters in a straight hose was calculated by the following equations (Keller, 1974):
where DPd is pressure loss due to the diameter change; Di, diameter of pipe; vi, fluid speed at section i; and D(i-1), the diameter of hose at section (i-1) of the pipe. All the factors that affect system pressure were considered in the program. The spray tip flow rate was calculated based on the following equation:
where Qi and Pi are the flow rate and pressure at spray tip i, and K is an experimentally determined coefficient. Various plumbing configurations with three different spray tips, numbers DG80015VS, DG8003VS, and DG8005VS (Spraying Systems Company); three system pressures, 207, 276, and 414 kPa (30, 45 and 60 psi); and five different hose diameters, 0.64, 0.95, 1.27, 1.91 and 2.54 cm (0.25, 0.375, 0.50, 0.75, and 1.00 in.) for the system were considered. The program computed flow rate errors, pressure losses, and delay times for each spray tip. Based on the results, and limitation of spray tip discharge rate variations of less than 5%, the most suitable plumbing configuration for reducing the time delay was selected.
By
knowing the direction of travel and anticipating changes in desired chemical
application rates, software corrections can reduce the effect of time delay
(Rudolph and Searcy, 1994). The
control system can anticipate changes in application rate to attenuate the
effect of the plumbing volume and reduce the time delay of rate changes at the
nozzle tip. In the simulation
model, software correction was implemented using a least-square-error-fit for
reducing this delay. The dead
reckoning method was used to determine the position of the sprayer based on the
boom width and travel distance for the purposes of simulation. The simulation model was also used to assess the performance
of the direct injection sprayer at various herbicide application rates, ground
speeds, variations in positioning system errors, and traditional versus modified
plumbing configurations. For this
simulation model, the following assumptions were made: 1) the application
procedure is continuous, and 2) sprayer deposition is uniform across the boom.
Based on these assumptions, a C++ program was written to simulate the
performance of the sprayer in the field.
RESULTS
Ground speed was determined using the mean sprayer speed on level ground plus the change due to the slope of the field. The relationship between speed change and slope is shown in Fig. 4. Linear equations were obtained to describe the relationship of ground speed variation to slope. The equation for up slope velocity was
and the equation for down slope travel was:
where Sp is the slope of the position. The correlation coefficient ( R2) of the linear regression for up-slope was 0.69 and the down-slope was 0.83. One problem associated with the linear equations was that the intersection point for up-slope did not overlay with the starting point at zero-slope. Perhaps field roughness or errors in recording travel times had an effect on this result. The slope coefficient was used to modify ground speed, ignoring the offset at zero-slope.
The injection point was moved from the original position, at the injection module located at the front of the sprayer, to a position immediately ahead of the boom manifold. Simulation results are summarized in Figs. 5 and 6 for three spray tip sizes. Fig. 5 shows the relationship between maximum delay time and spray tip size for the sprayer with the original plumbing configuration. Time delays were reduced with increased spray tip size or system pressure as in either case the discharge rate for the spray tips increased. Fig. 6 represents the relationship between maximum flow rate error of the system and spray tip size. The maximum flow rate error increased when spray tip size became larger. When spray tip size was not changed, the change of system pressure had little effect on the maximum flow error because of low flow rates for the small spray tips. The maximum time delay for each spray tip was 52.6, 26.3, and 15.9 s for original sprayer configuration using spray tips DG80015VS, DG8003VS, and DG8005VS, respectively. The maximum flow rate errors for each tip size were 0.88%, 3.02%, and 7.46%, respectively. The results showed that the maximum flow rate errors for DG80015VS tips were greater than 5% at all conditions.
The simulation program was used to optimize the combination of hose diameters for three spray tip sizes and three system pressures. These results are shown in Table 1. From this table it is clear the most suitable hose diameter configuration for the system with the DG 80015VS spray tips was 1.27 cm (0.5 in.) from the injection point to the boom manifold, 0.96 cm (0.375 in.) from the boom manifold to the center of each boom section, 0.64 cm (0.25 in.) between the center of the boom to the 3rd through 7th nozzles in each section, and 0.64 cm (0.25 in.) to the remaining nozzles in each section. With this optimal configuration, the minimum time delay was 16.0 s. For the DG8003VS tips the most suitable combination of hose diameters was 1.91, 1.27, 1.27, and 0.64 cm (0.75, 0.5, 0.5, and 0.25 in.), respectively, for the system (with the exception of an operating pressure of 207 kPa [30 psi]) resulting in a minimum time delay of 10.1 s. For the DG8005VS tips flow rate errors were all over 5% for a mixing point to boom manifold hose diameter of 1.91 cm (0.75 in.). For this tip size this hose diameters should be increased to meet the 5% (minimum) flow rate error limitation. However, it should be noted that this will increase the delay time.
When a “look-ahead” software modification approach was used to reduce the time delay, the hoses that connected the boom manifold and the center of each boom section were assumed to have the same length. The pressure relief was mounted close to the boom center to keep the boom pressure equal to the desired pressure. The injection point position was not an important issue at this condition. The program was run to find the most suitable combination of the hose diameters for the system. A least-squares method was used to determine the optimum software offset time. The results are shown in Table 2. Based on this table, the most suitable combination of hose diameters for the DG80015VS tip was 1.27, 0.64, 0.64, and 0.64 cm (0.5, 0.25, 0.25 and 0.25 in.), except for a system pressure of 207 kPa (30 psi). The time delay for the DG80015VS tip was reduced to 2.13 s. For the DG8003VS tips, the most suitable combination of hose diameters was 1.27, 0.96, 0.96, and 0.64 cm (0.5, 0.375, 0.375 and 0.25 in.) with a corresponding time delay of 1.5 s. For the DG8005VS tips, the most suitable combination of hose diameters was 1.91, 1.27, 1.27, and 0.64 cm (0.75, 0.5, 0.5 and 0.25 in.), and the time delay was reduced to 2.5 s, which was greater than the 1.5 s delay time for the DG8003VS nozzle. When considering all of the factors, obviously the best scenario for reducing time delays for the system was using the DG8003VS tips at a pressure 414 kPa (60 psi).
Figs. 7 through 9 show the effect of “look-ahead” software modification on the time delay for the DG8003VS spray tip at 276 kPa (30 psi). Fig. 7 shows the rate change distance due to the time delay. The largest rate change distance for the current sprayer was about 74.0 m (243 ft.). The software modification applied to the original sprayer configuration resulted in rate change distances as shown in Fig. 8. The maximum distance was 32.0 m. When considering all possible modifications (hose diameter, hose length, injection position, and software modification), the maximum rate change distance was 5.0 m as shown in Fig. 9.
The results of the simulation model for a constant ground speed and a tip size of DG8003VS (for a flow rate of 1.14 L/min at 276 kPa [0.30 GPM at 30 psi]) are shown in Table 3. Based on this table the application error of the original sprayer system was 43.5%. Each factor affected the application rate error. If the injection point was moved closer to the boom manifold, the total area receiving a misapplication was reduced to 37.1%. When changing hose diameters, the misapplication area could be reduced to 31.9%. Software modification alone reduced the misapplication area for the original configuration to 18.6%. If all modifications were used, the misappliction could be as low as 3.7%. The results show that the software modification is the single most important factor for improving the capability of the direct injection system for variable rate herbicide application. Using all the methods to modify the system was necessary to reduce misapplication to less than 5.0% of the desired application coverage as depicted in Fig. 1.
Model results for different pressures and spray tip sizes are summarized in Table 4. This table shows that the application errors are reduced as system pressures are increased and with the selection of larger size spray tips.
The ground speed errors vary in the field and have some effect on the application error. Ground speed variation results of the simulation model are shown in Table 5. Compared with constant speed results (Table 3), the addition of ground speed variation resulted in an error increase to 63.9% for the original system configuration. When all modifications, including ground speed, were considered the area receiving correct application was 91.0% of the total, with 9.0% of the area receiving either over- or under-application. These results suggest that constant ground speed, or feedback control for correction of injection rate, is important for the variable rate application.
CONCLUSIONS
Based on the results of this investigation, the following conclusions were drawn:
1. The original direct injection sprayer configuration studied was not acceptable for variable- rate herbicide application
2. Software modification can dramatically reduce time delay and the area of misapplication.
3. Optimizing hose diameter significantly reduces application error.
4. All of the noted modifications must be used to reduce the area of misapplication to a less than 5% of the total area when using direct injection for variable-rate application.
5. Ground speed variation is an important consideration for reducing the application error.
Al-Gaadi, K.A. and P.D. Ayers. 1994. Monitoring controller-based field sprayer performance. Applied Engineering in Agriculture. 10(2): 205-208.
Blumhorst, M.R., J.B. Weber, and L.R. Swain. 1990. Efficacy of selected herbicides as influenced by soil properties. Weed Technology. 4: 279-283.
Bridges, D.C.. 1992. Crop losses due to weeds in the United States. Weed Science Society of America, 309 West Clark Street, Champaign, IL 61820.
Compressed Air. 1988. The pied piper is changing his tune. Compressed Air Magazine. 93(11): 24-29.
Frost, A.R. 1990. A pesticide injection metering system for use on agricultural spraying machines. Journal of Agricultural Engineering Research. 46: 55-70.
Fox, R.W. and A.T. McDonald. 1992. Introduction to fluid mechanics. John Wiley & Sons, Inc., New York.
Ghate, S.R. and C.D. Perry. 1994. Ground speed control of pesticide application rates in a compressed air direct injection sprayer. Transactions of the ASAE. 37(1): 33-38.
Keller, G.R. 1985. Hydraulic System Analysis. Hydraulic and Pneumatics Magazine, Cleveland, Ohio.
McNeill, S.G., S.A. Shearer, and L.D. Swetnam. 1992. Demonstration spray rig with direct injection chemical metering. ASAE Paper No. 92-1092. St. Joseph, Michigan: ASAE.
Mulla, D.J.. 1991. Using geostatistics and GIS to manage spatial patterns in soil fertility. Automated Agriculture for the 21st Century. St. Joseph, Michigan: ASAE.
Ollila, D.G., J.A. Schumacher, and D.P. Frochlich. 1990. Integrating field grid sense system with direct injection technology. ASAE Paper No. 90-1628. St. Joseph, Michigan: ASAE.
Qiu, W., S.A. Shearer, and G.A. Watkins. 1994. Modeling of variable-rate herbicide application using GIS. ASAE Paper No. 94-3522. St. Joseph, Michigan: ASAE.
Rudolph, W.W. and S.W. Searcy. 1994. Strategies for prescription application using the chemical injection control system with computer commanded rate changes. ASAE Paper No. 94-1585. St. Joseph, Michigan: ASAE.
Stafford, J. V. and P. C. H. Miller. 1993. Spatially selective application of herbicide to cereal crops. Computers and Electronics in Agriculture. 9: 217-229.
Tompkins, F.D., K.D. Howard, C.R. Mote, and R.S. Freeland. 1990. Boom flow characteristics with direct chemical injection. Transactions of the ASAE. 33(3): 737-743.
Way, T.R., K.V. Bargen, R.D. Grisso, and L.L. Bashford. 1992. Simulation of chemical application accuracy for injection sprayers. Transactions of the ASAE. 35(4): 1141-1149.
Weber, J.B., M.R.
Tucker, and R.A. Isaac. 1987.
Making herbicide rate recommendations based on soil tests.
Weed Technology.
1: 41-45.