Impact of Remote Sensing on Loss Variables
Using quantitative methods like Vegetation Indices, we can continue towards increasing yield and lowering costs on a farm. Some of the more obvious ways are increasing yield by locating problem areas on a weekly (or more frequent) basis, saving money on chemicals by spraying specific problem sections rather than blanket application, watching the patterns of the fields over time to optimize planting in future years, reducing gas and labor costs needed to travel the fields to inspect crops on the ground, and obtaining better estimates of yield earlier in the season.
Unfortunately, extremely limited information on UAV use in agriculture is currently available for us to test these potential ways of saving. However, we can draw on agronomic remote sensing and vegetation index studies using satellites and manned aircraft to show us what we will be able to expect in terms of farm output with UAVs doing the remote sensing work.
Herbicide Variables. One area that has been researched is the economic benefit of site-specific management (SSM) and herbicide use. Studies using satellite imagery have analyzed the benefit of detecting where weeds are in a field and spraying only those sections instead of the entire field. In one example, NDVI satellite images were used to find weeds in a soybean field in Iowa. The images were run through a decision support system to identify what areas needed to be treated, resulting in a recommendation of spraying only 50% of the field. The remaining 50% did not contain any weeds, or did not show enough weed density to be worth spraying (Shaw, 2005). This information suggests that we can save about half of the money spent on herbicides per year. Shaw describes the calculated cost savings below:
“Calculated cost savings using site-specific herbicide applications (SSHA) ranged from $92.24 per hectare to $104.76 per hectare in soybean. Scouting based on sampling at a 10m grid size resulted in an increased net gain of $77.17 compared with a 20m grid-sampling scheme.” (Shaw, 2005)
His observations suggest that there is a large economic advantage to using remote sensing technology in weed detection, and that higher resolutions bring increased savings when sensing for weeds. In his experiment, Shaw found that there is a 50% increase in the amount of money saved on herbicide for a 50% decrease in farm section sizes.
Pesticide Variables. Similarly, variable rate machinery loaded with GIS maps of field conditions proved to reduce pesticide usage by 40% which both saved money and lessened the environmental impact of the chemicals (Dupont, J.K., Campenella, Seal, Willers, & Hood 2000; Pinter, 2003). Since pesticides cost an average of $80 per hectare, we could expect a $32 increase per hectare savings by using remote sensing techniques. Again this data was collected by satellite, and although the resolution was not stated, it was likely higher than 20m per pixel – meaning that there is potential for even higher savings at more accurate resolutions.
In this case, these results suggest two things. First, selective spraying does in fact save farms money in chemicals, and second, that a higher resolution image could lead to even more profitable spraying practices. These numbers were generated from satellite imagery which allowed for survey sections of 10m and 20m.
About the Data. Higher resolutions and smaller sections could allow for early season crop growth assessment, weed and insect monitoring, and fertilizer application monitoring (Shaw, 2005). However, higher resolutions from satellites come at higher costs. Currently, the cost to achieve higher resolution (1-5m) images of a 200 hectare farm (average farm size according to www.agday.org/media/factsheet.php) is $8,000 to $10,000 per new capture. Less accurate images cost less – down to $1,000 to $2,000 at the 20 to 30 meters per pixel accuracy. This means that in order to monitor crops every month, a farmer would have to order a new survey of their land at these prices for each capture. As Shaw acknowledges, the frequency of the monitoring needed to detect and resolve weed issues currently makes satellite imagery cost prohibitive.
Another study by Frank Tenkorang looked at many publically available papers on the economics of remote sensing and found that studies claimed to see benefits of remote sensing ranging from $222 per hectare to $5 a hectare. He conducted a study to estimate a more accurate number based on the currently available studies and found the average benefit of remote sensing to be $31.75 per hectare.
However, although Tenkorang discusses and criticizes the available studies for not explaining all of their methods, his estimate does not specify a certain crop, whether data was collected with a satellite or airplane, or what the resolution of the images were. Information like this is critical to the analysis of the potential benefit of remote sensing both in the costs of data collection and in assessing the augmented gains when using smaller sections. Problems like these likely explain why Tenkorang had difficulties in finding an estimate within a $200 per hectare range.
It is also important to remember that the estimates by Shaw and Pinter are only for the amount of money saved on herbicide and pesticide. Irrigation issues, planting efficiency, and other problems could also increase the total amount of money saved by remote sensing – and it is unclear if Tenkorang attempted to include these other factors in his research.
Most studies like Pinter’s, and Shaw’s use satellites as the primary platforms to collect data on farms. Satellites, while useful for large applications, are hard to use on demand and expensive to use for higher accuracies. Some of the drawbacks of satellites, like resolution, clouds blocking the ground, and frequency of captures, have been resolved by using manned aircraft. While the cost of hiring a pilot and plane is still high, data can be captured in a wider range of conditions, on demand, and with better resolution than a satellite. These advantages come at the cost of not being able to monitor as much area. Furthermore, processing times for manned aerial imaging are roughly 24 hours, compared to the 16 hours of processing needed for satellite images, although both of these time frames are acceptable for the use of weed and pest detection (Shaw, 2005).
UAVs offer a new solution to these problems – and potential for expansion on the actions available with the data. However, since the exact benefit of remote sensing is still unclear because of variation based on type of plant, region, resolution, etc., one good way to dissect this information and determine the usefulness of UAVs is to think of best, worst, and likely case scenarios.
Best, Worst, and Likely Case Scenarios
Best Case Scenario. In the best case scenario, we will take Shaw’s and Pinter’s findings (for herbicide: $100 per hectare savings at 20m sections and $150 savings at 10m sections; for pesticide: $32 per hectare savings at 20m sections) and scale them to resolutions easily attainable with a UAV using the same benefit increase we saw when halving the size of the survey section in herbicide savings.
From the herbicide data above, we see a roughly 50% increase in savings per hectare for every 50% decrease in section size. If we assume that this trend continues all the way down to the one meter per pixel level, then we expect the gain from using one meter per pixel accuracy to be $506 per hectare:
20 meters /2 = 10 meters -> $100 * .5 = $50 + $100 = $150
10 meters /2 = 5 meters -> $150 * .5 = $75 + $150 = $225
5 meters /2 = 2.5 meters -> $225 * .5 = $112.5 + $225 = $337.5
2.5 meters /2 = 1.25 meters -> $337.5 * .5 = $168.75 + $337.5 = $506.25
The process above takes the size of the section, halves it, then adds 50% of the savings at the previous section size to the current total savings to find the estimated savings at the specified section size. We work this logic all the way down to the one meter section size to find the estimated savings at this level. In this case the starting point is $100 from Shaw’s herbicide study. With the same logic, pesticide savings at the one meter per pixel accuracy would be $162 per hectare:
20 meters /2 = 10 meters -> $32 *.5 = $16 + $32 = $48
10 meters /2 = 5 meters -> $48 * .5 = $24 + $48 = $72
5 meters /2 = 2.5 meters -> $72*.5 = $36 + $72 = $108
2.5 meters /2 = 1.25 meters -> $108 * .5 = $54 + $108 = $162
This process is similar to the herbicide savings calculations, but uses a starting point of $32 based on Pinter’s experiment with pesticides. If we combine these two numbers, we get a total of $668 per hectare saved in chemicals. On a 200 hectare farm this is a $133,600 dollar per year estimated savings on chemicals alone.
To add some context to these numbers, we will first look at the costs and benefits of using satellite imagery to carry out our remote sensing missions. Using Pinter’s and Shaw’s estimates on a 200 hectare field, we find it would be cost effective to use low resolution (20 meter sections) satellite imagery ten times to monitor crops assuming each survey cost $2,000. We know this because the savings would be $100 of herbicide savings + $32 of pesticide savings = $132 per hectare. $132 * 200 hectares = a total of $26,400, less $20,000 spent on 10 remote sensing missions would show savings of $6,400 annually.
High resolution (1 meter sections) imagery that cost $10,000 each survey would be cost effective 12 times for the farmer to keep seeing returns on a 200 hectare field. We carry out a similar process to find this. If we save $506 in herbicide and $162 on pesticide at 1 meter section sizes, then we save a total of $668 per hectare. If this number is multiplied by 200 hectares, then we see an annual savings of $133,600 minus the cost of 1 meter resolution satellite imagery. This cost of $10,000 each survey would be profitable up to twelve times or a total cost of $120,000.
While satellite imagery might still be a useful tool, a UAV would allow for even further benefit. If a farmer chose to buy and operate a UAV, then process data in house, the total cost would be around $5,000 for the UAV, then $15 per survey using services like agribotix.com or precisionhawk.com. Assuming we want to survey our 200 hectare field 14 times during the season (one each time a new pest or disease emerges in corn), we will use $210 worth of survey processing a year.
This best case scenario would result in $128,390 in chemical savings for the farmer in the first year and $133,390 in subsequent years, or $667 per hectare. This is calculated as follows: $133,600(estimated benefit at 1 meter sections) – $5000(UAV cost) – $210(processing costs) = $128,390. Each subsequent year the calculation is the same but without the cost of the UAV. This estimate does not account for possible further savings through sensing for optimal irrigation, rotation patterns, sunlight patterns, yield prediction, optimal fertilizer placement, crop height at different stages, etc.
It is important to also mention that the relationship between number of surveys and benefit to yield is not yet known for any crop, so while a UAV would allow for unlimited surveys on demand, best farming practice might require surveys only eight times a season. In this case it would still be more cost effective to use a UAV, as long as the UAV or service costs less than $130,000 a year.
Worst Case Scenario. In the worst case scenario we would assume that the benefit of data does not scale to the image resolution, and that the benefit from remote sensing is as Tenkorang suggests: $32 per hectare. This scenario requires much less math. On a 200 hectare farm, the estimated benefit would be $6,400 annually. We would also assume that this number does indeed capture all potential benefits of remote sensing on farms. Resolution would remain at one meter per pixel – but would not matter since in this scenario we do not assume that lower resolutions and smaller section sizes mean higher gains.
Tenkorang does already subtract the costs in his equation so the $32 number reflects only profit. His average cost per hectare is $5, or on a 200 hectare farm, $1000. We will add this cost back to the total gains since we are calculating our own cost making the total savings $7,400. When we subtract $5,000 for the full price of a UAV and $210 for a year’s worth of surveys, we get $2,190. This is the amount we can expect to save using remote sensing with UAVs in the worst case scenario in the first year. In subsequent years we estimate a savings of $7,190, or $36 per hectare. With more expensive UAV setups this number could easily go negative for the first year, and might take several years to regain the overhead costs.
Likely Case Scenario. The likely scenario involves a bit more guesswork. It is likely that the benefit of remotely sensed data does scale to resolution and section sizes, but the exact diminishing or increasing returns is unclear. It is also likely that the benefit of remote sensing will vary depending on the crop being surveyed – but there is insufficient data on this assumption. So, to be conservative we will assume Tenkorang’s $32 per hectare savings, but add $5 to it to cancel out his assumed cost – leaving us with $37 per hectare. We will take the $37 per hectare number and assume that the benefit of each 50% increase in resolution is met with a 30% increase in savings, and that Tenkorang’s data came from an average resolution of 25 meters per pixel. The 30% multiplier reflects the likely increase in savings with smaller sections, but is a more conservative estimated multiplier than 50%. When we process the savings using the same methods as before, but with $37 as our starting number and .3 as our multiplier, we end with an estimate of $106 per hectare:
25 meters /2 = 12.5 meters -> $37 * .3 = $11.1 + $37 = $48.1
12.5 meters / 2 = 6.25 meters -> $48.1 * .3 = $14.4.3 + $48.1 = $62.53
6.25 meters / 2 = 3.13 meters -> $62.53 * .3 = $18.76 + $62.53 = $81.29
3.125 meters / 2 = 1.56 meters -> $81.29 * .3 = $24.39 + $81.29 = $105.677
On a 200 hectare farm, this is a total benefit of $21,200. Subtract the startup cost of the UAV and data processing and our total is $21,200 – $5,000 – $210 = $15,990 for the first year, and $20,990 for subsequent years, or $105 per hectare. Again, more expensive UAV packages may lower the first year’s gains – possibly to negative numbers, and several years would have to pass to regain the initial cost.
Cost Benefit Analysis of UAVs vs. Other Platforms
Now the question arises: is it worth it? These scenarios assume that the farmer has enough time, patience, and motivation to assemble a UAV, learn how to operate it, and routinely fly it. Is this hassle worth the potential benefit? The way that the UAV remote sensing industry is beginning to form would suggest that it is not.
Currently, large seed companies like Monsanto and Pioneer are contracting out their farm monitoring remote sensing programs to smaller companies. Those smaller companies will travel to the subscriber’s farm, perform the remote sensing flight, then process the data and provide farmers with code for their variable rate technology equipped machinery and vegetation index maps of their land. Farmers are then charged based on the amount of hectares surveyed (McKinnon, 2015). So, as long as this price per hectare is lower than (from our likely case) $105, the farmer should profit from this service. It would seem reasonable for the cost of surveying and data processing to be far lower than this price point – but the exact prices will need to settle in the market. In our worst case scenario, we would need the price of UAV remote sensing services to be less than $27 per hectare. Even with this number it seems reasonable to assume that UAV remote sensing services would be profitable for a farmer – but again, the market will determine standard pricing.
When we consider the 13 and 14 stages of crop maturation in soybeans and corn respectively, services like this become extremely competitive with satellite and airplane monitoring. If we reflect on the numbers we calculated before on the cost effectiveness of satellite remote sensing images ($26,000 – $2,000 * number of surveys at 20m sections and $133,600 – $10,000 * number of surveys at 1m sections) we notice that it would be a very slim margin of profit (if one exists), to monitor soybeans and corn at each stage of development using satellite imagery. Our analysis above suggests that there would be a much larger profit margin using UAV technology.
Sources:
McKinnon, D. “Misconceptions about UAV-collected NDVI Imagery and the Agribotix Experience in Ground Truthing These Images for Agriculture.” Agribotix. 10 June 2014. Web. 20 Nov. 2014. <http://agribotix.com/blog/2014/6/10/misconceptions-about-uav-collected-ndvi-imagery-and-the-agribotix-experience-in-ground-truthing-these-images-for-agriculture>.
Shaw, David R. “Translation of Remote Sensing Data into Weed Management Decisions.” Weed Science 53.2 (2005): 264-73. Jstor. Web. 10 Dec. 2014. <www.jstor.org/stable/4046967>.
Pinter, Paul J. Jr.; Hatfield, Jerry L.; Schepers, James S.; Barnes, Edward M.; Moran, M. Susan; Daughtry, Craig S. T.; and Upchurch, Dan R., “Remote Sensing for Crop Management” (2003). Publications from USDA-ARS / UNL Faculty. Paper 1372. <http://digitalcommons.unl.edu/usdaarsfacpub/1372>
Moran, M.S., Y. Inoue, and E.N Barnes. “Opportunities and Limitations for Image-based Remote Sensing in Precision Crop Management.” Remote Sensing of Environment 61.3 (1997): 319–346. Web. <http://www.tucson.ars.ag.gov/unit/publications/PDFfiles/MSM_1948.pdf>.