Agriculture constantly strives for improved efficiencies. Agricultural application or swathing efficiency is one area of interest given the potential of new technologies in precision farming. Tools and techniques are needed to assess application efficiency so in-field guidance techniques can be compared and quantified. One measure of application efficiency is the occurrence of overlap in applications. An analysis method using ArcView and the Spatial Analyst extension was developed to evaluate the application efficiency of ground-based applications using different guidance systems. This GIS method was compared to an alternative method based on Trimble software called TrimMap. The TrimMap and ArcView methods gave similar answers for all treatments. The guidance techniques or "treatments" investigated were a traditional form of guidance based on a foam marker system, a sub-meter accurate Differential GPS receiver with lightbar guidance and a centimeter accurate GPS receiver with lightbar guidance. Overlap area was highest with the foam marker, was less with sub-meter GPS and lowest for the centimeter GPS guidance treatment.

Precision agriculture technologies are providing the integration of tools to better understand and manage in-field variability effectively (National Research Council 1997). At the same time, agriculture strives towards improved management of farm resources and efficiencies to maximize crop production and economic returns, while ideally mitigating environmental impacts and maintaining long term farm sustainability. Improved efficiencies can be sought in many ways. Application coverage efficiency is one area that potentially could improve farm efficiencies particularly when multiple applications or treatments are being made to a field each year. In the past, the ability to assess application or swathing efficiency has been difficult to achieve, often done by ground-based point observations to estimate overlap or skip occurrence as spray units or spreader vehicles make consecutive passes in the field. This is tedious and in a large field can require a significant sample size of points in order to control the typically high variability of the this data.

Application or swathing efficiency could be represented by one or more key variables, including time to carry out the application, ground speed, distance off a pre-generated set of swath lines, volume of active ingredient applied per swath and per field application, and occurrence of double application (overlap areas) or missed application (skip areas). This paper focuses on the assessment of application efficiency in terms of occurrence of overlaps and skips in field applications.

The theoretical goal of most agricultural applications is to maximize coverage of chemical across the field while ideally minimizing both the occurrence of overlap and skip areas. In reality, many operators intentionally apply to minimize skips particularly with agrichemicals that can leave a "striping" effect in the crop if areas do not receive adequate chemical. Since skip can be visible (i.e,. measurable by appearance of "stripes" in crops), there is a tendency for operators to make extra overlap at the expense of application efficiency. The ability to actually manage both overlap and skip simultaneously while applying is not easy with the accuracy of many of the guidance systems to date. Limitations of the current guidance methods have not allowed farmers and custom applicators to achieve both minimum skip and minimum overlap in all their applications. It has also been difficult if not impossible to manage these variables in order to match specific crop and chemical combinations (e.g., minimize overlap when double doses of active ingredients can kill certain crop plants).

Trimble Navigation has been evaluating tools to assist with the assessment of swathing efficiency. The goal has been to better understand application efficiency, how to manage for this in specific crop-chemical situations and to develop a methodology that easily quantifies and demonstrates swathing efficiency. The integration and application of high accuracy GPS technology with the ArcView Geographic Information System (GIS) and Spatial Analyst extension (Esri 1997 a and b) is one method used to evaluate the degree of application overlap between adjacent swaths. A preliminary study was conducted in an intensive cropping region of New Zealand, on a 1000 acre farm close to the New Zealand office of Trimble Navigation. Comparisons were made using a number of guidance techniques including a traditional foam marker system and two types of GPS-based guidance.

1. Field Data Collection

A large spray vehicle mounted with a 78ft (24m) spray boom and spray equipment (AirTech Spray Systems) and an in-cab spray application controller (RDS) was used for the study.

Figure 1 - Photo of spray vehicle applying to the field with two GPS antennae and one radio antenna (communication to the RTK base station) shown on top of the vehicle cab.

One of the guidance techniques was to provide the driver guidance by an installed foam marker system. The two other guidance techniques used GPS and a lightbar device for guidance; one technique interfaced the lightbar (Figure 2a) to a sub-meter accurate Differential GPS receiver (Figure 2b) and the other interfaced the lightbar to a centimeter accurate RTK (Real Time Kinematic) GPS receiver (Figure 3).

 

Figure 2a - Lightbar used for guidance with the AgGPS132 sub-meter GPS receiver. This guidance system is sold as the Trimble AgGPS Parallel Swathing System

 

Figure 2b - Sub-meter GPS receiver (AgGPS132) used with a lightbar for the sub-meter GPS guidance treatment. This receiver and lightbar are sold as the Trimble AgGPS Parallel Swathing System.

 

Figure 3 - Centimeter Real Time Kinematic (RTK) GPS receiver which is sold as the Trimble 7400Msi survey grade GPS receiver. This was interfaced to the lightbar to provide centimeter GPS guidance. The same type of receiver was used to track the positions for all the three guidance treatments in the swathing study.

With all three guidance techniques (or treatments) an RTK centimeter GPS receiver was used to track the actual path that the spray vehicle traveled. A 486MHz computer on-board the vehicle was interfaced to the RTK GPS receiver to collect data at 1 second intervals for key variables such as distance offline, vehicle ground speed and actual vehicle position. The operator of the spray vehicle was a farmer who is accustomed to both foam marker guidance and lightbar guidance having had 10 years of foam marker experience prior to changing to lightbar guidance over the last 12 months.

GPS guidance in the study was carried out by defining an A point and a B point approximately 500 meters apart in a field, from which an A-B line is created. Based on the spray boom width programmed into the GPS guidance systems, parallel lines are then created out from the A-B line at the swath width spacing. When exactly on line, the center three LEDs illuminate on the lightbar. As the operator drives offline (i.e., to the left or right of swath center), the illumination band moves out from the center LEDs. The operator steers to bring the illuminated LEDs back to center. The lightbar is placed in the operator's peripheral vision. Approximately ten swaths were sprayed for each of the three guidance treatments.

 

2. Data Analysis

Overlap and skip area analysis between adjacent swaths was carried out using two different techniques. One method was based on Esri's ArcView GIS and Spatial Analyst extension which was used to determine overlap between adjacent swaths using a raster or cell based GIS analysis technique. The other technique used was based on software called TrimMap (Trimble 1997a) using a vector based process. Figure 4 shows the theoretical overlap and skip occurrence between adjacent swaths which were being derived from these analyses

Application overlap and skip were calculated for each pair of adjacent swaths as well as for the whole field using the following equations:

% Overlap between swaths = S (overlap areas b/t Swath _N and Swath _N+1) x 100%

% Skip between swaths = S (skip areas b/t Swath _N and Swath _N+1) x 100%

% Overlap for field = S (overlap areas b/t all swath pairs) x 100%

% Skip for field = S (overlap areas b/t all swath pairs) x 100%

Data was collected in the on-board 486 computer which was running a real-time mapping program that generated DXF plots of the field data being collected. Each guidance treatment was tracked with a centimeter RTK (Real Time Kinematic) GPS receiver recording the vehicle's path. The DXF maps were created using Northing and Easting values (based on WGS84 data collected in the field) and these were loaded into the ArcView GIS.

ArcView and the Spatial Analyst extension to ArcView were used to determine the percentage of overlap. Percentage overlap was investigated initially for every pair of adjacent swaths made (Figure 4) and then a total value obtained across the entire field (see equations above). Percentage skip was not determined with the GIS at the time in this GIS method at the time this work was reported. This was due to limited time being available to repeat the entire analysis on the full data set (i.e., to do both the TrimMap and the GIS method for both skip and overlap calculations). Also, the primary purpose of the GIS method was to compare it to the TrimMap method to see if the GIS was an accurate and reliable way to analyse results more efficiently. Deriving the overlap values was considered adequate to achieve this.

The following steps were carried out on the data to determine overlap:

1. The data was filtered in ArcView as follows:

A rectangular subset of the data was used which was covered the majority of the application logging but was just within the length of the original A-B line that was driven. This served to remove secondary logging, turning data at the ends of swaths.

Any GPS fix points collected when there was not the highest RTK accuracy status of Fixed Integer RTK were also removed (i.e., lower accuracy positions such as that under Floating Integer RTK were not included in the analysis).

2. The View / Properties dialog was selected to ensure that the Map Units were set to meters and the Distance Units set to meters. ArcView's Geographic projection was used with the Northing/Easting DXF plots that were loaded into ArcView. Once the units were specified in this dialog with the ArcView Geographic projection, some checks on distance and area measurements on the themes were made and were very consistent and accurate to the real values. For example, inter-swath distance was exactly 24 m for the pre-generated swath lines, matching the swath width set into the GPS guidance systems. Various distances and areas in ArcView were found equivalent to that measured when the same data set was loaded into TrimMap.
3. Each swathline driven by the spray vehicle (i.e., was tracked by the RTK centimeter GPS and computer) was separated into independent ArcView Shape files. These themes (layers) were turned on in the ArcView project.
4. A buffer was created about adjacent pairs of these line features using the Analysis / Find Distance function, specifying the cell size to be as small as possible. Due to memory limitations (i.e., office computer being used and possible memory leaks in the pre-release version of Spatial Analyst) the minimal acceptable cell size was 0.25 m for these analyses. The buffering process creates a raster or grid layer which can be set to 1 classification category, of 12 m (equal to half a swath) so the display of the buffer indicates the actual swath that was driven (Figure 5).

Figure 5 - Swath line centers driven by the vehicle shown as the black lines. Two adjacent swath lines are shown buffered by 12 metres either side of the swath line centers using the ArcView-Spatial Analyst data analysis method

5. Comparison between 2 adjacent buffer grids were made by doing an Analysis / Map Query to look for the areas between the 2 lines where overlap occurred. For example, between swath line 1 and 2, the query was:

Find areas with distance to Swath Line 1 � 12.0 m AND the distance to Swath Line 2 � 12.0 m

 

The resulting grid layer from this query between Swath 1 and 2 which were both less than 12 m from their swath center lines, so represented area common to both swaths and therefore overlap areas.

6. The area of these overlaps were determined using the Analysis / Tabulate Areas function.
7. This process, numbers 3 - 6 were repeated for each pair of adjacent swaths for all treatments.
8. Screen snaps were captured for the overlap areas between swaths for each treatment, ensuring these were at the same scale in ArcView's View.

The recorded WGS84 latitude and longitude positions along each swath line center were loaded into TrimMap as a CSV (comma separated data file) and displayed using the New Zealand Map Grid projection.

1. The following data filtering was done using the graphical map display in TrimMap:

A rectangular subset of the data was used which was located just within the original A-B line that was driven to remove secondary logging and turning effects

Any additional secondary logging fixes were removed.

Any points along the swath line where fix numbers doubled back on themselves, as this caused problems in Trimmap when generating parallel lines to the original swath line.

2. The logged positions in TrimMap are numbered in the order that they were collected. These positions were then joined by a line in this logging order using the CAD / Add Lines function in TrimMap.
3. Parallel offset lines were then made to the swath center lines at 12 m to the left and 12 m to the right of the swath center line, using the Graphical Calculations / Coordinates / Parallel offset function. This defined the swath width about the swath center line.
4. Steps 2 - 3 were performed on each swath line so each pair of adjacent swaths could be studied in terms of their degree of cross-over (i.e., intersections or change over points from overlap to skip). Steps 5 - 6 below were used to determine areas of overlap and skip:
5. Areas or polygons were defined between adjacent swaths by determining where the lines intersected. A new point was placed at the cross-over or intersection points of the two closest (outermost) swath lines using the Graphical Calculations / Intersections / Split Lines.
6. The areas within each polygon created were defined using the Graphical Calculations / Areas & Lots / Define a Lot functions within TrimMap.

The GIS analysis produced graphical displays of the overlap areas in ArcView (Figures 6, 7, 8). Overlap occurrence in the field application was highest with the foam marker guidance technique

Figure 7 - Overlap areas in red that were found between each swath for the sub-meter GPS guidance treatment, as derived with ArcView + Spatial Analyst. The initial line shown is the A-B line defined at the start of the application.

Sub-meter GPS guidance overlap was less than foam and appeared to be greater than the centimeter GPS guidance techniques (Figures 7 & 8).

As these ArcView plots were only qualitative estimates of overlap, actual percentages of overlap were then determined.

Figure 9 shows all the inter-swath overlap percentages calculated using both the ArcView - Spatial Analyst and the TrimMap analysis methods. The TrimMap results showed similar patterns to the ArcView results for differences between guidance techniques although the actual overlap percentages with TrimMap were always slightly lower than those seen from the ArcView analysis. This consistent difference between analysis methods is due to the limitation of the 0.25 m cell size that was able to be used in the GIS method which would cause an exaggeration of the extent of overlap (note - extra values or bars are shown with the TrimMap Sub-meter GPS treatment - this was due to an additional replicate of data being analyzed with TrimMap method but was not done with GIS method)

TrimMap and ArcView methods were used to obtain overlap percentage for the entire field application (i.e., for all pairs of adjacent swaths) and for all three guidance treatments (Figure 10)

Within each analysis method, foam exhibited the highest percentage overlap (2.04% with TrimMap and 2.61% for ArcView) while centimeter GPS guidance exhibited the lowest percentage overlap (0.59% for TrimMap and 1.12% for ArcView GIS). Sub-meter GPS guidance resulted in overlap values between the other two treatments (1.00% for TrimMap and 1.53% for ArcView). Analyses of variance (ANOVA) were performed on all the data and at the 95% confidence level the differences between guidance treatments were not found to be statistically significant in this study (Snedecor and Cochran, 1980). This is attributed to the relatively high variability in some of the guidance treatment data sets (e.g., foam and sub-meter GPS). The most significant overlap difference was seen between foam and centimeter GPS guidance using the TrimMap data, which was significant at the lower levels of confidence.

Table 1 shows the overlap differences between guidance techniques for both methods. The differences between each treatment were very similar (e.g., 1.04% for TrimMap vs 1.08 % for ArcView for the difference between foam and meter GPS guidance)

Therefore, the two methods of data analysis showed very similar results, particularly in the overlap patterns or trends. Overall field overlap percentages and differences between treatments are comparable when looking at the two data analysis methods, although the TrimMap results were consistently lower than those obtained with ArcView. The pattern of inter-swath overlap was similar for both analysis methods (Figure 9) where the "high" and "low" patterns in the histograms were comparable. Absolute numbers of overlap were always higher with the GIS method than the Trimmap method. This is likely to result from the raster analysis requiring cells of a specified size and the limitation of 0.25m (25cm) cells when this study was carried out.

What does this mean from an agricultural swathing efficiency perspective? If you had 1000 acres growing crops which required agrichemical applications per year at an average price of $100 per acre for each agrichemical application, the savings in reduced application overlap with meter GPS guidance would be $1,040 per application (i.e,. 1.04% multiplied by 1000 acres and multiplied by $100). An equivalent comparison between foam and the centimeter GPS guidance system would save $1,450 per application.

However, there is another important aspect to agricultural swathing efficiency. An equally important and related variable is skip or missed application areas between swaths. The GIS method was not used for skip analysis but the TrimMap method was used to determine skip. Although the purpose of this report was to compare the GIS and TrimMap analysis methods, a short discussion on skip values derived with TrimMap is warranted. Across all guidance treatments, the field average skip percentages were not statistically significantly different. However the average percentages were highest for sub-meter GPS guidance (1.46%) followed by centimeter GPS guidance (0.77%) and then foam (0.35%).

When one considers the nature of foam markers versus GPS lightbar guidance, the differences in skip and overlap may be due to a couple of factors. Lower average foam percentage skip complements the relatively high percentage overlap observed as is expected (i.e,. if you have more overlap you would expect less skip). Foam marker guidance inherently encourages overlap by the driver who is trying to visualize where to drop foam out at the end of a 30-40+ ft boom. To be sure of field coverage, it is likely that operators adopt a conservative approach. With parallax errors when looking out to the last swath's foam far away from the driver, while trying to ensure no skips in application, the driver is likely to make an exaggerated placement of the foam (i.e., place the foam further over the previous foam thereby increasing overlap). This would also depend on driver fatigue. Many hours of application using foam markers may reduce the likelihood that a driver can "conscientiously" drop the foam and ensure that skip is minimized. However, in this study, there were only ten swaths in the field for each guidance treatment, which was unlikely to cause major driver fatigue. Therefore, the conditions were considered optimal for foam as the day was sunny and without wind, the field was flat and had no undulations (minimizing the boom movements which can cause misplaced foam), the crop residue was low-lying and the color contrasted greatly with the white foam making the foam marker highly visible, plus driver fatigue was likely to be minimal.

With GPS guidance techniques there was a relatively constant amount of overlap to skip. Considering the technique, this is likely to result from the operator being "offline" in either direction of swath center as they follow the lightbar LEDs for guidance. Rather than looking to foam markers in the field out a window, the operator can look straight ahead without needing to focus on the lightbar but rather have this in their peripheral vision. The advantage for GPS guidance is that a user-defined swath width can be set in the GPS guidance system. For example, with a 78 ft (24m) spray boom the driver can specify 77ft (23.5 m) swath width in the GPS guidance system to ensure that skip is minimized for that application. Therefore drivers of spreader trucks and spray units can use GPS guidance to consistently adjust and manage their degree of overlap and skip occurrence for the entire field application. This swath width setting can be adjusted to suit the specific crop and chemical combination (i.e., whether skip should be minimized or double applications of chemicals to some crops should be minimized).

In this study there was high variability found in some of the data sets (e.g., sub-meter GPS overlap, foam overlap shown by size of error bars in Figure 10), which influences the statistical significance of differences between the treatments. The relatively high variability in the sub-meter GPS data has been attributed to a lower accuracy signal for satellite differential GPS that has been recorded in New Zealand compared to North America and Australia. This is contrasted with the very "tight" data (small error bars and low variability) using centimeter RTK from an in-field base station providing the corrections. The relatively high variability in the foam marker overlap percentages is believed to be related to observations seen on the ground where foam blobs were dropped onto existing foam blobs from the previous swath in a quite wide band, approximately 3-6 ft (1-2m). This was under conditions that were considered optimal for a foam marker. For example, the day was sunny without any wind, the field was very flat and without undulations that would increase boom bounce and misplacement of foam blobs, and the field had a low-lying bean crop residue which made it very easy to see the foam marker blobs.

Although the ArcView GIS method showed lower overlap percentages compared to that derived with TrimMap, there were many similarities in the patterns of the data. It is concluded that the ArcView GIS method can be used to analyze the swathing efficiency data for field applications, particularly for treatment comparison purposes (i.e., relative differences between guidance or swathing treatments).

The absolute values of overlap from this GIS method were higher than the TrimMap overlap that occurred due to the minimum cell size being limited to 0.25m in this study. The GIS method could be improved with additional computer processing power, a higher resolution cell size (e.g., 0.01m) and either using the final release version of Spatial Analyst or using a full-featured GIS like ArcInfo. However, the ArcView GIS and Spatial Analyst method is considered a useful means to rapidly compare between treatments.

From an agricultural perspective, sub-meter GPS guidance is at least equivalent to foam in terms of swathing efficiency measured by overlap and skip percentages. Due to less than optimal conditions for the sub-meter GPS guidance method and the close to optimal conditions for the foam marker guidance, this conclusion may be conservative. Centimeter GPS guidance showed the lowest overlap amongst the guidance treatments.

Now that a series of techniques have been established, future research is being planned to repeat swathing efficiency studies and make comparisons with other guidance techniques. Additional work is required to refine the process and further validate findings under a variety of agricultural spraying and spreading situations.

On behalf of the Trimble Navigation Precision Agricultural Systems group, the authors would like to thank David and Roger West of West Brothers Inc., particularly David West who helped to carry out the swathing study and provide equipment. Thanks also to specific Trimble Navigation staff who provided input to the design and data analysis including Phil Jackson, Ian Viney, Greg Price, Eric White and Joan Hollerich.

National Research Council 1997. Precision Agriculture in the 21^st Century. Geospatial and Information Technologies in Crop Management. National Academy Press, Washington D.C. 1997, pp.149

Esri 1997a ArcView version 3.0a software and on-line help.

Esri 1997b Spatial Analyst version 1.1 (pre-release) extension software for ArcView and on-line help

Snedecor, G.W., & Cochran, W.G. 1980. Statistical Methods. 7^th Edition. Iowa State University Press, Ames Iowa, USA. Pp.507.

Trimble 1996 Series 7400 Operation Manual. Part Number 31463-00. Trimble Navigation Ltd., 645 North Mary Avenue, P.O. Box 3642, Sunnyvale, CA94088-3642.

Trimble 1997a TrimMap Survey Software version 6.50. Trimble Navigation Ltd., 645 North Mary Avenue, P.O. Box 3642, Sunnyvale, CA94088-3642.

Trimble 1997b AgGPS132 Operation Manual. Part Number 33301-00. Trimble Navigation Ltd., 645 North Mary Avenue, P.O. Box 3642, Sunnyvale, CA94088-3642. 

Trimble 1997c FieldGuide Software Users Guide. Part Number 70529-00. Trimble Navigation Ltd., 645 North Mary Avenue, P.O. Box 3642, Sunnyvale, CA94088-3642. 

Trimble 1998 AgGPS Parallel Swathing Option Manual Part Number 34900-00. Trimble Navigation Ltd., 645 North Mary Avenue, P.O. Box 3642, Sunnyvale, CA94088-3642.