Precision Farming: The IHARF Experience

Vaughn Johnson¹, Guy Lafond², Paul Bullock³, Alan Moulin⁴, Judy McKell⁵, and Yann Pelcat⁶

¹Montmartre, SK

²Indian Head Research Farm, Indian Head, SK

³Noetix Research Inc, Winnipeg, MB

⁴Brandon Research Centre, Brandon, MB

⁵Saskatchewan Agriculture and Food, Indian Head, SK,

⁶Indian Head Agricultural Research Foundation, Indian Head, SK,

1.0 Background

Precision Farming (PF), including Global Positioning Systems (GPS), Geographical Information Systems (GIS), yield monitors and variable rate controllers, allows the producer to vary one or more inputs across the field according to a pre-determined prescription as well as determine the variability in crop yields across his field or verify easily and accurately the impact of specific production technologies. The goal with PF is to allow more efficient use of inputs, whether they are fertilizers, pesticides, management or labor, across the landscape. The end result is to maximize financial advantage and minimize production risks while at the same time ensuring environmentally sound production practices. The challenge for the producer is to assign management units cost effectively and to develop a prescription for those units.

The Indian Head Agricultural Research Foundation (IHARF) and the Saskatchewan Soil Conservation Association (SSCA) have developed a partnership to facilitate research and demonstrations on PF. IHARF has purchased 307 acres to dedicate to this research and make these activities possible. The field site has been sub-divided into 8 equal fields and the crops present on the site include field pea, canola and spring wheat. With the help of SSCA, industry partners were secured to provide equipment, crop inputs and assist with the financing of the land purchase.

The current practice is to assign management units based on soil characteristics derived from extensive soil sampling, usually following a grid sampling procedure. Some of the soil variables measured include depth of topsoil, soil texture, organic matter, salinity, pH, CEC, residual N-P-K-S & micronutrients. From these results, a prescription is prepared. The problem with this approach is that it is very costly and time consuming, making it more or less impractical on a large farm-scale basis.

2.0 Objectives

The objectives of this project are to demonstrate the effectiveness of other potential methods of assigning management units within a field relative to the standard method of extensive soil sampling and characterization using a grid-mapping approach. Some of the methods proposed include a general soils map, a detailed elevation map, aerial photos, infra-red photos, remote sensing images, depth of topsoil, soil map unit using an EM38 or other similar technologies and yield map from a combine equipped with yield monitor and GPS unit.

The intent is to provide producers with cost effective alternatives to extensive soil sampling and analysis. This project will also familiarize producers with technologies applicable to crop monitoring, such as remote sensing, soil salinity mapping, infra-red photographs etc., by providing concrete examples on a farm scale basis. Information about this project and information generated from this project is available on the web site dedicated specifically to this project and for disseminating information about Precision Farming in general. The web site address is: http://paridss.usask.ca/precisionfarm

3.0 Anticipated Results:

It is anticipated that after four years, producers will have a much better understanding of alternative tools available for assigning management units within a field along with their relative effectiveness. With the trend towards larger farms, farm managers will find it more and more difficult to monitor their fields on a continual basis because of time constraints. It is hoped that farm managers will have access to real time satellite images in the near future thereby giving them effective tools for monitoring any negative changes in their crop over very large areas. It will direct them to very specific areas within their fields. It is hoped that this project will familiarize producers with this new technology in anticipation of more extensive use in the near future.

4.0 Description of Study

4.1 Site

The total area is 307 acres subdivided into 8 more less equal fields in size. A crop sequence involving spring wheat-canola-spring wheat-field pea has been established and this sequence is replicated twice on the site. In the spring of 1998, a one-acre grid was established across the entire site. This involved burying iron pins at designated locations and geo-referencing them in

order to ensure that we are always back at the same location. These permanent markers ensure that we are always collecting the information at the same location all the time.

4.2. Measurements

Equipment: The site was seeded with a 27' Flexi-Coil 5000 air wagon equipped with double shoot capabilities and the side-banding stealth opener. The crop was sprayed with a Wilmar high clearance sprayer. The crop was harvested with a Ford-New Holland TR98 equipped with a yield monitor and GPS unit.

Soil Sampling: In the spring of 1998, a general soil sample was taken from each field in order to establish the correct levels of fertilizer use for each crop in the first year. In the fall of 1998, soil samples were collected around each pin to a depth of 36 inches. These samples were analyzed for organic carbon, nitrogen, phosphorus, potassium and sulfur. Determinations of pH and conductivity were also done on each sample. These results formed the basis for the fertilizer amounts used in 1999 for each crop. Soil samples were collected again in the fall of 1999 in order to help with the establishment of a variable rate program for nitrogen. In the spring of 1999, soil moisture determinations in the 0-2" layer were done at 78 selected sites across the site along with measurements of crop residue and surface roughness.

Soil Survey: A general soil survey was collected in the fall of 1998 and spring of 1999 in order to characterize the site for general soil characteristics.

EM38 Soil Salinity Survey: EM38 is an instrument used to measure soil salinity. In the absence of soil salinity, it provides a confounded measurement of soil texture and soil moisture. By doing the measurements in the fall of 1998 (dry soil condition) and spring of 1999 (wet soil condition), we were able to get results under both dry and wet conditions. By combining the measurements with a GPS unit, we are able to see if any repeated patterns exists across the site

Satellite Imagery: It has been demonstrated that high-resolution satellite imagery can provide accurate maps of vegetation vigor and soil characteristics at a field scale. The imagery can illustrate the variability in these properties. There are many factors that cause variability in crop and soil conditions. Research and demonstrations on a field scale basis are needed to understand which of these factors can be identified using remote sensing techniques. This study provides an excellent opportunity to further our understanding of how remote sensing can identify the various causes of crop and soil variability at a field level and help with the determination of management units within a field. Noetix Research Inc. has been gathering the remote sensing data for this project. To date, the following data have been collected for this study:

1. SPOT panchromatic acquired 13-May-1998 (spatial resolution - 10 m)

2. LANDSAT TM (bands 2,3,4) acquired 13-July-1998 (spatial resolution - 30 m)

3. RADARSAT S2 acquired 1-May-1999 (spatial resolution - 25 m)

4. IRS-1D acquired 22-July-1999 (spatial resolution - 23.5 m)

Ground-Based Hyperspectral Measurements: In the summer of 1999, a hand held spectrophotometer was used at 78 selected spots across the entire field to encompass all crops. It consisted of measuring the light reflectance off the crop canopy at five different wavelengths. SPAD measurements were also done at same 78 locations across the field.

Aerial InfraRed Photographs: Measuring the reflectance of light over a crop canopy can help distinguish between low and high areas of vegetative growth. Photographs of this sort were taken during both the 1998 and 1999 growing season. Color Infrared Air Photo acquired on 16-July-1998 (spatial resolution - 0.8 m) and on 23-Jul-1999 (spatial resolution - 0.5 m).

Yield Maps: With the use of a combine equipped with a yield monitor and GPS unit, accurate yield maps were obtained allowing us to correlate the yield with the various measurements conducted.

Crop Measurement: In the spring of 1999, 4 plant counts were done around each pin and grain samples were collected around each pin as well in order to examine effects on grain protein but more importantly on overall nitrogen use. Plant tissue samples from the crop for N determinations were also collected at the time of the hyperspectral and SPAD measurements.

Weed Survey: In the spring of 1999, prior to the in-crop spraying operation, 4 weed counts were conducted around each pin. Each count consisted of counting the weeds and identifying them in four separate 0.25 square meter blocks. Maps of weed distribution can be seen on the web site.

Disease Survey: At selected times, disease surveys were conducted around each pin, in order to accurately establish the level of diseases in each crop and field. Maps of plant disease distribution can be seen on the web site.

Deep Nitrate Measurements: Based on the results of recent Green Plan-supported assessments of water quality across the prairies, nitrate contamination of the shallow ground water is the most widespread serious water quality concern. Much of this nitrate contamination has been linked to nitrate leaching from cropland. The amount of nitrate leaching varies greatly within the field but its pattern is predictable. Leaching is greatest in depressions, even if they are slight, where surface runoff accumulates. Nitrate leaching is especially serious where these depressions correspond to areas where soil plus fertilizer nitrate exceeds the crop nitrogen needs such as occurs where the crop is over-fertilized with nitrogen and/or crop growth is poor due to disease or lack of macro-nutrients. Although the soil condition at this site largely prevents severe leaching and resulting nitrate contamination of underlying aquifers, this same slow nitrate movement allows nitrate leaching to be quantified simply and inexpensively by measuring soil nitrate in 2 to 3 metre deep soil cores.

4.3 Analysis

All of the measurements collected so far are undergoing both spatial and discriminant analysis in order to establish a variable rate program and determine the merits of indirect measurements relative to soil sampling for assigning management units.

4.4 Information

In order to make it easy for producers to access this information and to follow the results, a comprehensive web site has been put together. The exact color maps of data can be obtained from the web site at the following address: http://paridss.usask.ca/precisionfarm

5.0 Results

Not all results were available at the time this report was put together. We recommend that you visit the web site regularly for updates on the results. The web site address is: http://paridss.usask.ca/precisionfarm

Yield: The average yield amounts over each site can be seen in Table 1. Actual yield maps can be viewed on the web site.

Table 1. Average yield for each field and year.

Satellite Imagery: The objective of this research is to determine techniques which can identify sub-field management zones in a cost-effective, accurate and efficient manner, then evaluate the economics of precision agriculture.

Sub-objectives:

1. Determine cost of data and information sources

2. Evaluate data sets for determination of topographic landscape units

3. Determine statistical relationships between various data sources

4. Design a field layout for the precision agriculture field to evaluate the economics of zone-based versus uniform-rate management

All images have been geo-referenced. The sample point grid has been overlaid on each scene and grid point values have been extracted for each band of image data. In the case of the Radarsat scene, a 5x5 Frost filter and a 3x3 average filter was used to smooth the scene before the grid point data was extracted. No filtering or re-sampling, other than that for the geo-referencing, was done on the other images. Electromagnetic conductivity readings were collected by Saskatchewan Water Corporation using an EM38 with a GPS unit. Transects of EM38 readings were collected in both the fall of 1998 and the spring of 1999. The readings were pixelated into 1 meter pixels, then kridged to create a surface. Grid point EM38 values were then extracted from the surface.

Yield data was collected with an AFS monitor during the harvest of both 1998 and 1999. The Vision software resampled the readings to 5-meter pixel values. Grid point yield values will be extracted from these readings. In addition, the image pixel grid will be overlaid on the yield data and a mean yield per image pixel will also be extracted.

In addition to the above measurements, soil nitrate nitrogen, phosphorus and potassium, hyperspectral reflectance and radiation transmitted through the leaf (650 and 940 nm) were also measured at each grid point. Simultaneous with the RADARSAT overpass on 1-May-1999, soil moisture, surface roughness and surface residue measurements were taken.

Analysis will now be conducted to determine the land management zones suggested purely by each image and by combinations of images. In addition, statistical analysis will be conducted on the grid point samples to determine if there are strong correlations to any of the soil and/or crop samples. The key relationships being sought are those between image channels and crop yield as well as those between image channels and crop nitrogen or soil nitrogen levels.

Soil Fertility: All measurements are in kg/ha. Nitrates and sulphates (0-60 cm) were extracted with CaCl2; PO4-P and K (0-15 cm) extracted with NaHCO3. It would appear there is a significant fertilizer requirement, assuming N targets of 100 kg/ha and P2O5 of 40 kg/ha for cereals, N target of 106 kg/ha and P2O5 target of 31 kg/ha for canola, and P2O5 target of 31 kg/ha for peas.

There is more spatial variability in the P and K measurements than those for nitrate and sulphate (Table 2). Phosphate-P and K seem to vary in long north-south strips, which may be associated with fallow and or surface curvature. Nitrate and sulphate are relatively uniform on the map, the high values for sulphate are located around the gully, while one high nitrate value appears to be located in the approximate area of the old manure pile. Nitrates were lower in 3 of 4 fields seeded to fallow compared to stubble.

Table 2. N-P-K and S levels in the soil based on soil samples taken around each pin in the fall of 1998.

6.0 Conclusions

The objective of this write-up was to provide a general overview of what kind of information is actually being collected on the site and the types of questions that are being asked and where the information can be accessed in order to follow the developments of the project. As the information is being collected, it will be posted on the web site with the following address: http://paridss.usask.ca/precisionfarm