Frank Tenkorang and Jess Lowenberg-DeBoer
Introduction
People have been talking about the economics of remote sensing at least since satellite images were first made available for civilian use. For example, in January, 1977, San Jose State University in California organized the “First Conference on the Economics of Remote Sensing Information Systems.” In an effort to understand how thinking has developed on the economics of remote sensing for agriculture we have tried to collect all the publicly available reports on profitability and risk management benefits of using satellite and aerial imagery for making better crop management decisions. This article will focus on our initial observations on the status of the economics of satellite and aerial remote sensing in agriculture.
In this article we are focusing on documents that report dollar values attributed to use of remote sensing in management of field crops. There are literally thousands of journal articles, conference papers, websites and other documents that report on the technical issues related to capturing and interpreting images, but the number that report monetary estimates is small. Some anecdotal evidence suggests that the market for remotely sensed images is growing for use in managing high value vegetable and orchard crops, but we have found no publicly available studies that document the value of remote sensing for these horticultural crops.
Publicly available documents are the focus of our search. This is the information that farmers, crop consultants and other have on which to base decisions. We would like to organize and summarize that publicly available material to facilitate those decisions. Publicly available material includes articles in scientific journals, papers in published conference proceedings, articles in the farm press and websites. It does not include the internal newsletters of remote sensing organizations, reports on contract research done for the National Aeronautics and Space Administration (NASA) or other government agencies, or proprietary marketing research done by firms that hoped to market images to growers.
Remote sensing is usually used in combination with other information to make decisions and with other technologies to implement those decisions, so there is always an “attribution problem” in the economic analysis. For example, when remote sensing is used to define management zones for fertilizer application, should profits be linked to the remote image? Or to the soil testing? Or to the variable rate application (VRA) of fertilizer? We have not tried to unravel this puzzle, but report returns to information and technology packages.
Studies Reporting $ Values
Of the hundreds of remote sensing studies that we have reviewed, ten studies reported dollar values (Table 1). Of those ten only the Carr et al. (1991), Larson et al. (2004);Watermeier (2003); Long (2002); White et al. (2002); White and Gress (2002); and Copenhaver et al. (2002) studies provide enough information to allow us to understand how the estimates were made. Understanding the budgeting methods and assumptions are essential to developing confidence in the estimates.
Table 1. Studies Citing Returns to
use of Remote Sensing in Agriculture
Average
Type
of Input Return
Authors Imagery Crop Managed $/acre
------------------------------------------------------------------------------------------------------------
Zone Determination using Images from Previous Seasons:
Carr et al. (1991) Aerial & Wheat, P & K $0.87
Satellite Barley Fertilizer
OSU (2002) Aerial Cotton Fertilizer, $60.00 -
Insecticide, $150.00*
Growth
Regulator
Larson et al., (2004) Aerial Cotton Fertilizer, -$2.31 to
Insecticide, -$14.96
Growth
Regulator
Seelan, et al. (2003) Satellite Wheat, Nitrogen $98.78*
Sugar Beet
In-Season Management
Copenhaver et al. (2002) Aerial Soybeans Herbicide $1.68
Long (2002) Aerial Wheat Herbicide $0.92
Reynolds & Shaw (2002) Aerial Cotton, Herbicide $27.47 to
Corn $74.65*
Watermeier (2003) Aerial Corn Nitrogen $13.00
White et al. (2002) Aerial Wheat Nitrogen -$1.21
White and Gress (2002) Aerial Corn Nitrogen -$1.06
* No details given on how benefit were estimated. Appears to be gross benefits without subtracting costs of images, analysis and VRA implementation.
The OSU (2002), Reynolds (2002), and Seelan (2003) documents only make passing comments about the economic benefits that might be achieved. Ideally, studies that report economic benefits to remote sensing (and any other technology) should include a budget table and a paragraph describing how the benefits were estimated or cite sources for this information. Key elements in a budget estimating benefits to remote sensing include: crop and input prices, yield and input quantity changes, cost of images and analysis, and the cost of variable rate application or other operation required to implement the decision make based on the remote image. A good study will also deal with the management time required to understand the images and make the decision, as well as the cost of developing the skills to use remote imagery.
The Seelan et al. (2003) study cites remote sensing in sugar beet management as one of the successful examples of using satellite imagery in agriculture. The one paragraph that they devote to the economics of this practice is appropriate given the topic of their article (i.e. a “learning community” approach to using remote sensing), but we hope that they would have given us a reference for the details of their economic estimates. When we contacted the first author directly, he wrote back to say, “The cost figures … were given by the farmer we work with. Unfortunately, we do not have more details than what is in the paper.” It would have been useful to cite the farmer as the source.
Carr et al. (1991) tested soil fertility management by zones based on a combination of soil maps and color infrared aerial photography or satellite imagery in five locations in Montana. Variable rate fertilizer application was based on soil test, yield goals, and Montana State University fertilizer guidelines. The value in Table 1 represents the average return over the five fields when fertilizer rates are determined before planting. This analysis does not include cost of other inputs such as imagery and labor among others. The information and technology package tested would not have been profitable in any of the fields if typical costs for imagery and soil testing were deducted.
The Tennessee study (Larson et al., 2004) return estimate in Table 1 is negative because of a $20/acre consulting fee. If the grower could handle the analysis and development of recommendation maps without additional cost, the return would be $5.04 to $17.69/acre. The range is due to differences in per acre equipment costs due to cotton acreage. The lower return is on a small farm with only 356 acres of cotton. The upper end of the range is on a larger operation where equipment costs can be spread over ten times as many acres (i.e. 3560 acres of cotton). Costs might also be reduced if the variable rate controller and other precision agriculture equipment could be used for other crops on the farm.
Based on airborne imagery, Copenhaver et al. (2002) classified a soybean field into three weed pressure categories; severe, moderate, and sparse. In their “full boom variable rate” treatment herbicide use was reduced by 30% compared to conventional uniform application. They estimate the value of the herbicide savings at $3.45/acre. No yield difference was detected between uniform or variable rate treatments. The cost of the images averaged $1.77/are, resulting in a “net return to machinery” of $1.68/acre. The cost of image analysis was not deducted. The cost of developing on-farm skills in image analysis and operating a variable rate sprayer, or the cost of hiring someone with these skills was not taken into account. They point out that whether a grower could justify a variable rate herbicide sprayer depends on the number of acres covered each year.
Long (2002) did research on use of aerial photos to identify wild oats in wheat for patch spraying. He compared the cost of aerial photos to ground based scouting and uniform application of herbicides without scouting. In the cost estimate he includes the cost of the aerial photo, the herbicide and application, and lost yield in areas incorrectly identified as uninfested. The photos are analyzed visually by two individuals, but Long does not mention any compensation for their time and effort. The herbicide application is apparently charged only on those patches sprayed; this is appropriate for patch spraying, but with variable rate herbicide application the application charge would more likely be incurred on the whole field.
Long does not provide an average return to using aerial photography for wild oats management. He does note in his summary that the environmental benefit from lower herbicide use may be offset by losses due to inaccuracies in identifying infested areas. The average return to remote sensing for the Long study in Table 1 is calculated as the difference between the uniform application and the average with patch spraying based on the two image analysts. It should be noted that the average is positive because of a large benefit on one of the three field studied. On two of the three fields, the cost of the aerial imagery approach was higher than that of the uniform application.
The cost estimate approach used by Long avoids the problem of measuring yields, though he muddies the waters a bit by including a yield loss for areas that are incorrectly identified as uninfested. The cost estimate approach is good for preliminary estimates, but in the longer run it is better to have yield estimates. For example, the Long study implicitly assumes that infested patches that are sprayed have the same yield as uninfested areas, but it is possible that treated areas might yield less because the herbicide was ineffective or the wild oats had already caused some yield loss before the herbicide was applied.
Watermeier (2003) looked at the use of remote sensing for midseason application of nitrogen on corn. In the remote sensing based approach, nitrogen was sidedressed at half the recommended rate and remote sensing was used to identify the areas that needed additional nitrogen shortly before tasseling, The control was sidedress nitrogen at the recommended rate. When the remote images were analyzed, it was discovered that the areas requiring additional nitrogen were a small part of the field and only about 45 lbs/a were applied, so the decision was made not to apply any additional nitrogen. The yields on the full rate and half rate strips were almost the same. All the benefit of not having applied the additional nitrogen was attributed to remote sensing.
Watermeier includes the cost of the images and additional nitrogen, but not the analysis of the image or variable rate application cost. This information is a good start. Several years of data would be required to understand the variability of results over time.
The goal of the White et al. (2002) study was to test the effectiveness of remote sensing in identifying nitrogen variability within a winter wheat field, and develop an image-based technique for site-specific nitrogen fertilizer management. Treatments included a conventional uniform rate application, uniform rate nitrogen application based on aerial imagery and VRA based on that imagery. They conclude that yields are about the same with all treatments, but that the imagery based applications use slightly less nitrogen overall. With the constant rate based on imagery, the reduced nitrogen about paid for the imagery cost resulting net returns that were very similar to whole field management. With variable rate application based on imagery, there was a loss of slightly over $1/acre largely because of the VRA custom application charge and a slightly lower yield. It is not clear if image analysis was included in the VRA charge.
White and Gress (2002) assessed the potential of using airborne imagery to guide in-season VRA of nitrogen. They tested the technology on one field in Champaign County, IL. VRA based on imagery resulted in a much lower nitrogen application, an average of 92 lbs/acre for the VRA compared to the 160 lbs/acre in the conventional whole field treatment. Yield for the VRA treatment was almost 14 bu/acre lower than the uniform 160 lbs./acre application. An imagery cost of $1.86/acre and cost of VRA of $3/acre were included in the economic analysis. It is not clear if imagery cost included analysis. The net return in the VRA treatment was slight more than $1/acre lower than that of the 160 lbs/acre uniform rate application.
Conclusions
Out of the hundreds of remote sensing in agriculture documents reviewed, only 10 reported economic benefit estimates. Three of those documents do not provide details on how the economic benefit was estimated. Clues in the reports and the fact that the numbers are much larger than those for detailed studies suggest that those three are reporting gross benefits without deducting the cost of images, analysis, VRA and other expenses related to site-specific management. Based on the studies that we have found, we suggest skepticism when a study of remote sensing in field crop management reports benefits of over $50/acre. It is possible in higher value crops (e.g. sugar beets, cotton), but unlikely for grains and oilseeds. The highest return in any study which provided estimation details was $14/acre (Watermaier et al, 2003) and this was without taking into account the cost of image analysis.
Because the seven studies that provide budgeting information vary widely in what is being managed and how the remote image is used, it is difficult to see any pattern in the profitability estimates. One study uses aerial images in developing management zones for P & K application in Montana. Two studies focus on site-specific weed management, one in wheat in North Dakota and the other in soybean in Illinios. One study tests use of aerial images in determining nitrogen application rates on wheat in Kentucky. The Tennessee study was of an relatively integrated system managing seeding rates, fertilizer, insecticide and growth regulators on cotton.
Two studies focus on mid-season nitrogen application in corn, one in Ohio and the other in Illinois. The Ohio study reports a $13/acre benefit because the image led researchers to forego a mid-season nitrogen application in a dry year. The Illinois VRA study was also done in a dry year. The authors argue that VRA of nitrogen would be more profitable in a year with average or higher rainfall. Multiple years of testing the same approach to using remote images in the same area would be needed to document benefits.
The Tennessee study includes image analysis in the consulting fee. It is not clear how the other six studies that reported budget details deducted the cost of image analysis. None of the seven explicitly deducted the cost of developing on-farm skills in image analysis or operating VRA equipment, or hiring that skill. In the White et al (2002) and White & Cress (2002) papers, the cost of this skill may have been implicit in the custom service fees.
All seven of the studies appear to have data only for one year at any given site. The Tennessee report was part of a multiyear study, but it is not clear from Larson et al (2004) if more than one year of data was used for the economic analysis. The Carr et al. (1991) study has two years of data, but none of it appears to be at the same site. Assessing the variability of returns to remote sensing will require multiple years of data at the same location.
It should be noted that this is a preliminary report on an on-going study. Readers are urged to bring additional studies of remote sensing economics in field crops to our attention (email either ftenkora@purdue.edu or lowenbej@purdue.edu).
More Information
Carr P.P. G.R. Carlson, J.S. Jacobsen, G.A Nielsen, and E.O. Skogley. Farming soils, not fields: a strategy for increasing fertilizer profitability. Journal of Production Agriculture, Vol. 4, no. 1, 1991, p. 57-61.
Copenhaver K., T. Gress, C. Sprague, and D. Alderks. Variable rate application of post-emergence herbicide to soybeans using remotely sensed imagery. In Proceedings of the Sixth International Conference on Precision Agriculture. ASA-CSSA-SSSA, Madison, Wisconsin, 2002.
Larson, J.A., R.K. Roberts, B.C. English, R.L. Cochrane and T. Sharp, “A Case Study Analysis of a Precision Farming System for Cotton,” paper presented at the Beltwide Cotton Conference, San Antonio, TX, 2004.
Long, Dan,
“Large-Scale Aerial Remote Sensing to Improve Management of Agricultural
Chemicals,” report prepared for the Chouteau County Conservation District,
July, 2002 (ag.montana.edu/narc/PDF%20Doc/Final%20Report.pdf).
Oklahoma State University, “Reducing Cotton Production Costs Using Remote Sensing and Spatially Variable Insecticide/Defoliation (SVI/SVD) Technologies,” 2003 (www.cotton.org/cf/projects/general-prof-precision-ag.cfm).
Reynolds, D. B. and D.R. Shaw, “Detection and Site-Specific Control of Weeds through Remote Sensing,” (www.rstc.msstate.edu/publications/ 99-01/rstcofr01-010.pdf)
Seelan, Santhosh, Soizik Laguette, Grant Casady and George Seielstad, “Remote Sensing Applications for Precision Agriculture: A Learning Community Approach,” Remote Sensing & Environment, 88 (2003), p. 157-169.
Watermeier, N. (2003). In-season variable rate application of nitrogen in corn-based on remotely sensed imagery. Ohio Geospatial Technologies Conference Agriculture and Natural Resources, Columbus, Ohio March 24-26, 2003 (http://geospatial.osu.edu/conference/proceedings/index.html#link5).
White, S.E., M. Bethel, and T. Gress. The use of remotely sensed imagery to make nitrogen recommendations on winter wheat in Western Kentucky. In Proceedings of the Sixth International Conference on Precision Agriculture. ASA-CSSA-SSSA, Madison, WI., 2002.
White, S.E. and T. Gress. The use of remotely sensed imagery to make in-season nitrogen recommendations for corn. In Proceedings of the Sixth International Conference on Precision Agriculture. ASA-CSSA-SSSA, Madison, WI, 2002.