Demarcating Management Zones For Site Specific Farming Using Electromagnetic Induction and Yield Maps in Eastern Oregon

Search PMN

Demarcating Management Zones for Site-Specific Farming Using Electromagnetic Induction and Yield Maps in Eastern Oregon

Stephen Machado, Columbia Basin Agricultural Center (CBARC), P.O. Box 370, Oregon State University, Pendleton 97801

Abstract

Traditional farming practices in eastern Oregon ignore the inherent spatial variability (topography, soils) resulting in inefficient farming and lower net returns. Site-specific farming (SSF), or farming based on requirements of specific areas in a field, has the potential to improve input use efficiency and farm profits. To implement SSF, however, management zones should be demarcated preferably based on factors that most affect yield. The most common method for demarcating management zones is through intensive grid sampling that can be very expensive. The objective of this study was to find cheaper and quicker ways to demarcate management zones in a field in eastern Oregon. This study evaluated the use of electrical conductivity (EC) and yield maps in demarcating management zones. A 28-acre field was characterized for depth and nutrients at a 100-ft grid. Wheat was uniformly seeded throughout the field and harvested at each grid location. Results indicated that the spatial variability of wheat yield was influenced by soil depth. EC data were correlated with both depth and grain yield and can therefore be used to demarcate management zones quicker and cheaper than grid sampling in this field. Information from yield maps was also useful in demarcating management zones for SSF.

Introduction

Farm profits in eastern Oregon are decreasing because of low and stagnant wheat prices and ever-increasing input costs. Managing inputs more efficiently is one way to increase profit margins. Site-specific farming (SSF) provides this opportunity. There is considerable spatial variation in field characteristics (topography, texture, depth, slope, and aspect) (5,14) that often leads to spatial variability of crop yields (1,3,4,6,10,11). Conventional farming ignores this variability. Inputs (seed, cultivars, fertilizer, and pesticides) are applied uniformly throughout the fields at most farms in eastern Oregon. Uniform application of fertilizer can lead to over fertilization in areas with low yield potential, increasing the chances of environmental pollution. Unnecessary pesticide applications may lead to pesticides resistance. Uniform seeding rates may lead to high populations in areas that have low yield potential. Growing of one cultivar on both the south and north-facing slopes may lead to reduced yields on the drier south-facing slope particularly if the cultivar is not drought tolerant. In general, ignoring the spatial variability in field characteristics leads to inefficient use of inputs, reduced profit margins, and increased chances of environmental pollution. These concerns justify SSF.

Implementation of SSF requires information about a field. This information may include slope, aspect, soil depth, and soil chemical, physical, and hydrological characteristics. Information on the spatial variability of grain yield from previous years such as that obtained from yield monitors can also be very helpful. Using site characteristics and information on the spatial variability of grain yield, management zones can be demarcated and different areas in a field can then be managed according to specific requirements. To effectively characterize a field, the grower needs to grid sample the field and send samples for analyses. The cost alone can be prohibitive. Furthermore, such costs can not be justified for low-value crops like dryland wheat, the major cereal crop in the eastern Oregon. However, with the advancement of technology, less costly data acquisition methods have been developed. These methods include electromagnetic induction (EM) and yield mapping. The objective of this study was to characterize a field and demarcate management zones based on factors affecting grain yield the most using EM and yield monitor data.

Evaluating EM and Yield Maps for Demarcating Management Zones

Site characterization. The study was conducted on a 28-acre field at the Oregon State University Columbia Basin Agricultural Research Center (CBARC), Moro, Oregon (45.5°N, 120.7°W, with elevation of 558 m above sea level). The soil at CBARC, Moro is a coarse, silty, mixed, superactive, mesic Typic Haploxeroll (Walla Walla silt loam). Average annual crop-year (September 1 to August 31) precipitation is about 288 mm. To identify soil factors influencing wheat yields, the field was first characterized. The soil was sampled at 12-inch increments to a depth of 5 ft or to restricting layer at 126 geo-referenced locations (DGPS) on a 100-foot grid (Fig. 1) using a Giddings probe (Giddings Machine Co. Inc., Fort Collins, CO). This procedure resulted in a data density of about 5 points/acre. Locations were geo-referenced with a differential Trimble AgGPS 132 receiver (Sunnyvale, CA). At each depth interval, soil was analyzed for nitrogen, phosphorus, potassium, sulfur, soil organic matter (SOM), and acidity (pH) at the Kuo Testing Laboratories (Othello, WA) to obtain baseline information. No soil texture analysis was done because the soils, having been formed in loess, are fairly uniform (silt loam). The soil probe was set to drill as deep as 5 ft and it was assumed that the soil was more than 5 ft deep in areas where the probe reached 5 ft without going through a restrictive layer. Depth to restricting zone or bedrock was recorded during soil sampling. Information on elevation, slope, and aspect was obtained from GIS data obtained at each DGPS location.

Fig. 1. Digital Elevation Model of the experimental site at CBARC, Moro showing 100-ft DGPS grid locations (red flags) that were sampled.

Yield mapping. To determine the effect of the soil physical and chemical characteristics on wheat yield, soft white spring wheat (Zak) was seeded at a uniform seeding rate of 90 lbs/acre throughout the whole 28-acre field in the spring of 2003. A uniform rate of 30, 45, and 26 lbs N, P, and S per acre, respectively, was applied before seeding. The crop was harvested manually at each DGPS location to determine grain yield and analyzed to determine how grain yield was influenced by site characteristics. At each DGPS point, wheat in four, 3-ft rows was cut by hand, threshed, and cleaned.

Yield maps from a local grower (John McElheran, Wasco County, OR) were also obtained for discussion purposes. A combine, fitted with a grain yield monitor (John Deere Greenstar, Dallas, TX), was used to harvest wheat in 2001 and 2003. Because there were no data available for interpolation, the maps show actual combine passes.

Electromagnetic induction (EM) readings. The EM operation was described by Davis et al. (2). Electromagnetic Induction is a non-invasive, non-destructive sampling method conducted using the EM38 (Geonics Limited, Mississauga, Ontario, Canada) (Fig. 2). The instrument’s transmitting coil induces an electrical current in the soil. The magnetic field created is strongest in the top 15 inches of the soil and has an effective sensing depth of about 5 ft. The induced current creates secondary currents in the field that are sensed by a receiving coil. The relationship between the primary and the secondary currents is related to the electrical conductivity (EC) of the soil. When properly calibrated, the instrument can be used to measure soil depth (12) and soil texture (13). Depth is determined by detecting the restricting clay layer. Sand, silt, and clay have low, medium, and high EC, respectively. The instrument can also measure salt content, soil water, and crop productivity. The instrument is light weight and can be suspended or mounted on a non-metallic cart (Fig. 2).

Fig. 2. EM38 electrical conductivity meter mounted on a non-metallic cart to leave a few inches above ground. The height of the meter above the soil surface can be adjusted.

To take soil EM readings, the EM38 meter was mounted on a non-metallic cart in a vertical dipole mode (upright orientation) (Fig. 2) and dragged across the field at 100-ft intervals corresponding to DGPS locations using an all-terrain vehicle (ATV). In this orientation, the EM38 provides an effective measurement depth of approximately 5 ft. The EM38 was suspended about 3 inches above ground. Data were collected at 3 to 4 second intervals resulting in a data density of about 282 points/acre. Data were read into a PRO4000 field computer (Geonics Limited, Mississauga, Ontario, Canada) using EM38pro software for real-time data logging. Data were geo-referenced (sub-yard resolution) with a differential Trimble AgGPS 132 receiver (Sunnyvale, CA) mounted on the ATV. EC readings can also be obtained by a Veris EC Mapping System (Veris Technologies, Salina, KS). A pair of coulter-electrodes injects electrical current into the soil and the other coulters measure the voltage drop. The Veris and the EM38 methods give very similar results (7).

Data analysis. Both traditional and spatial analyses were used to interpret data. Mapping of site characteristics, yield, and EM readings was done using a combination of MapInfo Professional (Version 5.0, MapInfo Corporation, Troy, NY) and SoilRx (Version 1.321, Red Hen Systems, Fort Collins, CO) software. Data were interpolated using the inverse distance technique. Simple correlations (SAS Institute Inc., Cary, NC) were used to describe relationships between measured variables. Factor and regression analyses were used to determine factors that influenced the spatial variability of wheat yield. Regression analysis involving the identified factors is shown in Table 1.

Table 1. Regression equations of common factors affecting grain yields of spring wheat within the field.

Aspect	Regression model	R²	Pr > F
North- south	Y = 20.86 + 2.25depth^b + 0.56aspect + 0.57sulfur	0.34	<0.0001
East- west	Y = 20.86 + 2.01depth^b + 0.63nitrogen + 0.71sulfur^a	0.30	<0.0001

^a significant at 0.10 level of probability.

^b significant at 0.0001 level of probability.

Cheaper and Quicker Demarcation of Management Zones

Factors influencing grain yield. Ideally, management zones should be based on factors that influence grain yield. However, because there are many factors that influence yield (1,3,4,6,8,9,10,11), variable rate application of inputs would be complicated if all factors were included. Therefore management zones should be demarcated based on field characteristics that influence grain yield the most. Data analyses (factor and regression analyses) indicated the spatial variability of wheat grain yield was largely influenced by soil depth (r = 0.47, P < 0.01; Figs. 3 and 4). In dryland agriculture, soil depth has considerable effects on grain yield because of its influence on soil available moisture. Although grain yield increased with increase in soil depth, the correlation was rather weak. The correlation was higher in areas where depth to restricting zone was less that 5 ft (r = 0.66, P < 0.01). In areas that were at least 5 ft deep or more, grain yield was poorly correlated with depth indicating that other factors also influenced grain yields. Observations indicated that there were areas with high incidences of pests (weed infestation and gopher damage) in these deep areas that probably led to low grain yields. Depth also influenced the spatial distribution of N, P, and K. Concentrations of P and K were higher in deeper areas than in shallower areas.

Fig. 3. Spatial distribution of grain yield (Mg/ha) of spring wheat at CBARC, Moro in 2003.

Fig. 4. Depth to restricting layer or bedrock (m) of the experiment site at CBARC, Moro, 2002.

Based on this characterization study, the first layer of management zones should be based on soil depth. Information on depth for this study was obtained through grid sampling using a tractor probe. The process was laborious, expensive, and took four-man hours to complete. It cost $30 per sample (5 depth-samples per location multiplied by $6 per sample) to analyze samples from each of the 126 grid locations for nutrients ($3780 or $135/acre) (Kuo Testing Labs Inc., Othello, WA). Including grid sampling costs (at $10/sample) would increase costs to $360/acre (Kuo Testing Labs, Inc., Orthello, WA). To practice SSF, growers would have to spend similar amounts to demarcate management zones in the first year. In subsequent years, however, the cost will be dramatically reduced as growers take only a representative sample from each management zone. At these initial costs, SSF would be less economical than traditional farming. To this end, cheaper and faster methods using EM readings should be developed or adapted. In this study, the use of EM readings and yield maps to demarcate soil depth zones of this field was evaluated.

Electromagnetic induction readings. It took about 53 min to drag the EM38 throughout the 28-acre field. Electrical conductivity data (Fig. 5) were negatively correlated with depth (Fig. 6) (r = -0.49; P < 0.01). Similar results were obtained in related research (2,12). Although not a strong correlation, these results indicated that soil depth zones in this field could be mapped relatively quicker and cheaper by EC than by manual soil probing. The EC map was probably more accurate (282 data points/acre) than the map derived from manual probing (5 data points/acre). Differences in resolution could have also contributed to the weak correlation between EC and soil depth. As with the depth map, the EC map was significantly correlated to grain yields (Figs. 3 and 5) (r = -0.37, P < 0.01). The lower correlation between EC and yield was either attributed to the low resolution of the grain yield map (5 points per acre) and/or to the probability that EC also mapped other factors that were not strongly correlated to grain yield. Based on these results, however, EC can be used effectively to demarcate management zones based on soil depth in this field. To practice SSF in this field, shallow and deep areas should be managed based on their yield potential. A rough estimate of yield potential can be determined from yield-water-nutrient relationships in a particular area.

Fig. 5. Electrical conductivity (mS/m) map of the CBARC, Moro field, 2002.

Fig. 6. Wheat yield map of the Paulson field at McElheran’s farm in Wasco Co., 2001.

Yield maps. Yield maps, generated by yield monitors, can effectively be used to demarcate management zones at a lower cost than EC maps. Although yield monitors were expensive at the time they were first introduced, prices have gone down considerably since then. More and more growers are now using yield monitors on a routine basis. Yield maps, taken over a number of seasons, can provide information on demarcating management zones. Maps show yield monitor data in 2001 (Fig. 6) and 2003 (Fig. 7) from the Paulson field at John McElheran’s farm, Wasco Co., OR. It was clear from Fig. 6 that there was considerable spatial variation in grain yield in the Paulson field in 2001 (5 to 40 bu/acre). Low-yielding areas were shallower (less than 2 ft) than high-yielding areas (more than 3 ft deep). Although slightly different (mostly in magnitude), grain yield variability in 2003 (Fig. 7) was remarkably similar to the spatial variability of grain yield in 2001 (Fig. 6). It was drier in 2001 (6.21 inches) than in 2003 (9.29 inches). Based on the information obtained in 2001 and 2003, this field can be demarcated into two major management zones consisting of areas with low yield potential (red to yellow) and areas with high yield potential (green to purple). Similarly, the field at CBARC, Moro can be divided roughly into two management zones – a shallow zone (orange to red) and a deep zone (yellow to green) – using the grain yield map derived from manual harvesting in 2003 (Fig. 3).

Fig. 7. Wheat yield map of the Paulson field at McElheran’s farm in Wasco County, 2003.

Information obtained from yield maps is valuable for fertilizer management. Areas that were low yielding may have high residual nutrients and may require reduced fertilizer rates in the following season. Similarly these areas may require reduced plant stands. Low and high yielding areas can be sampled separately and a representative soil sample from each area would require analysis instead of hundreds of samples obtained from grid sampling. Inputs (fertilizer and seeding rates, etc.) would then be applied accordingly. It may be necessary to fine-tune management zones every year as grain yield variability changes from year to year (Figs. 6 and 7). Although depth was the main factor influencing the spatial variability of grain yield in this field, other factors may interact with depth and change sections of the yield map every year. For instance, high grain yields were produced in the upper most area in 2001 (Fig. 6) but in 2002 (Fig. 7) this area produced low grain yields due to downy brome (Bromus tectorum) infestation.

Conclusions

Results showed that there was considerable variation in wheat yields (5 to 40 bu/acre) within the two fields evaluated in eastern Oregon (Figs. 5, 8, and 9) that justifies SSF. Varying amounts of inputs should be applied to specific areas in a field based on yield potential. To this end, factors influencing grain yield must first be identified. Results indicated that soil depth influenced the spatial variation of wheat yield. In dryland agriculture soil depth has considerable effects on grain yield because of its influence on soil available moisture. In the fields evaluated in this study, shallow and deep areas should be managed differently. More inputs (fertilizer, seeding rates) should be applied in deeper areas with high yield potential and less inputs in shallower with less yield potential. Work to determine the optimum input levels and yield potential of shallow and deeper areas should be conducted. Obtaining EC and yield monitor data was quicker and cheaper than grid sampling. In this study, soil depth, which influenced grain yield the most, was mapped fairly well by EC. Management zones can, therefore, be demarcated with information from EC data in this field. Results also show that yield maps obtained yearly can provide useful information for demarcating management zones. Better information to demarcate management zones can be obtained by combining data from both EC and yield maps. Once the management zones are demarcated, inputs can be applied based on the yield potential of the area.

Acknowledgment

I would like to thank Christopher Humphreys for helping with data collection and John McElheran for providing yield maps. This study was funded by Oregon State University General Research Fund.

Literature Cited

1. Braum, S. M., P. Hinds, G.L. Maltzer, J. Bell, D. Mulla, and P.C. Robert. 1998. Terrain attributes and soil nitrogen: Spatial effects on corn grain yield responses to nitrogen fertilization for a northern, glaciated landscape. In: P. C. Robert, R. H. Rust, W. E. Larson, eds. Proc. of the 4th Int. Conf. on Prec. Agric., St. Paul, MN, July 19–22, 1998. ASA-CSSA-SSSA, Madison, WI.

2. Davis, J. G., Kitchen, N. R., Sudduth, K. A., and Drummond, S. T. 1997. Using electromagnetic induction to characterize soils. Potash and Phosphate Institute. Better Crops with Plant Food 81:6-8.

3. Everett, M. W., and F. J. Pierce. 1996. Variability of corn grain yield and soil profile nitrates in relation to site-specific N management. Pages 43-53 in: P. C. Robert, R. H. Rust, W. E. Larson, eds. Proc. of the 4th Int. Conf. on Prec. Agric., St. Paul, MN, July 19–22, 1998. ASA-CSSA-SSSA, Madison, WI.

4. Jaynes, D. B., Kaspar, T. C., Colvin, T. S., and James, D. E. 2003. Cluster analysis of spatiotemporal corn yield patterns in an Iowa field. Agron. J. 95:574-586.

5. Johnson, D. R., Makinson, A. J., Walker, A. E., Monte, D. K. and Cahoon, S. 1985. Soil survey of Umatilla County area, Oregon. Soil Conserv. Serv. Publ., USDA.

6. Kravchenko, A. N., and Bullock, D. G. 2000. Correlation of corn and soybean grain yield with topographic and soil properties. Agron. J. 92:75-83.

7. Lund, E. D., Colin, P. E., Christy, D., and Drummond, P. E. 1999. Applying soil electrical conductivity technology to precision agriculture. Pages 1089-1100 in: P. C. Robert, R. H. Rust, W. E. Larson, eds. Proc. of the 4th Int. Conf. on Prec. Agric., St. Paul, MN, July 19–22, 1998. ASA-CSSA-SSSA, Madison, WI.

8. Machado, S., Bynum, E. D., Jr., Archer, T. L., Lascano, R. J., Wilson, L. T., Bordosvsky, J., Segarra, E., Bronson, K., Nesmith, D. M., and Xu, W. 2000. Spatial and Temporal Variability of Corn Grain yield: Site-Specific Relationships of Biotic and Abiotic Factors. Prec. Agric. 2:343-360.

9. Machado, S., Bynum, E. D., Jr., Archer, T. L., Lascano, R. J., Wilson, L. T., Bordosvsky, J., Segarra, E., Bronson, K., and Nesmith, D. M. 2002. Spatial and Temporal Variability of Sorghum Grain yield: Influence of Soil, Water, Pests, and Diseases Relationships. Prec. Agric. 3:389-406.

10. Mulla, D. J., and Schepers, J. S. 1997. Key processes for site-specific soil and crop management. Pages 1-18 in: The State of Site-Specific Management for Agriculture. F. J. Pierce and E. J. Sadler, ed. ASA, CSSA, SSSA, Madison, WI.

11. Ping, J. L., Green, C. J., Bronson, K. F., Zartman, R. E., and Dobermann, A. 2004. Identification of relationships between cotton yield, quality, and soil properties. Agron. J. 96:1588-1597.

12. Sudduth, K. A., Kitchen, N. R., Drummond, S. T. 1999. Soil conductivity sensing on claypan soils: Comparison of electromagnetic induction and direct methods. Pages 979–990 in: P. C. Robert, R. H. Rust, W. E. Larson, eds. Proc. of the 4th Int. Conf. on Prec. Agric., St. Paul, MN, July 19–22, 1998. ASA-CSSA-SSSA, Madison, WI.

13. Williams, B., and Hoey, D. 1987. The use of electromagnetic induction to detect the spatial variability of the salt and clay content of soils. Aust. J. Soil Res. 25:21-27.

14. Yang, C., Peterson, C. L., Shropshire, G. J., and Otawa, T. 1998. Spatial variability of field topography and wheat yield in the Palouse region of the Pacific Northwest. T. Am. Soc. Agric. Biol. Eng. 41:17-27.