A Full Growing Season with SlantRange: Are Population Counts Accurate?
This post was composed by Jesse Badger, owner of Precision Unmanned Aerial.
During the 2017 growing season, Precision Unmanned Aerial, working with Green Thumb Farms in Maine and New Hampshire, began a case study to test the SlantRange 3p multispectral sensor and SlantView analytics software with the goal of determining just how accurate aerial remote sensing with drones can be in correlating early-season population counts with post-harvest yield data. The plan was to test on a small 10-acre corn plot bi-weekly throughout the growing season to collect data during the three main stages of growth: Emergence, Growth, and Maturity. During the study, Precision’s intent is to collect trending data across different stages of the crop, create different data products, and assess the initial accuracy of the population count versus end yield.
Three weeks after planting, the corn had emerged and was roughly V2 stage, 4” tall and 4” leaf tip to leaf tip. None of the plants were touching and from above you could see clear soil spacing between the plants. As such, the corn test plot was in prime condition to collect a SlantView population count map.
After taking a quick measurement of the crop size, the flight plan was optimized to ensure the system had the resolution needed to detect the plants for counting. System setup and flight time of the small test plot took less than 20 minutes. After completing the flight, the data was processed on-site in 10 minutes and SlantView produced a population count map, a plant size map, and a weed location map.
Being able to give the grower something they can use same day was an essential requirement to our business model. Being able to deliver the results and immediately begin ground truthing the data has exceeded our expectations immensely and made our data truly actionable. With data in hand, we set out to examine the results.
According to the agronomist who manages this field, the target planting density was 34,000 plants/acre. The SlantRange population count map and map statistics showed just over 25% of the field with >27,500 plants/acre, 50% between 24,500 and 27,500 plants/acre, and 25% under 24,500 plants/acre.
How Accurate are the SlantRange Population Counts?
The big question was “How accurate are the SlantRange population counts?” The field report gives an easy way to visualize the data and select areas to inspect on foot. With the exported KML maps opened on a tablet with Google Earth, the scout is able to monitor their location on the population count map as they work the field. This is where we find this imagery to provide an immediate ROI. By using the imagery to select areas of interest to collect measurements to corroborate the aerial data, we were able to get an exact picture of the entire field rather than arbitrarily selecting points for ground measurements, which can often result in missing critical problem areas.
The first area surveyed showed an unusually high population count estimate of 34,000 - 39,000 plants/acre in the white area of the population map. However, the corresponding high resolution image in SlantView clearly showed that in this particular region, some plants were classified as corn plants (green overlay), that should have been classified as weeds (red overlay). At the site, a large infestation of weeds was found which confirmed the imagery and explained the excessively high plant count. However, the overly high population area alerted the agronomist to a dense area of weed growth, which was still valuable information.
The next area surveyed on the population count map showed a very low population estimate of 9,000 - 11,000 plants/acre in the grey area. Where we expected corn emergence, large areas void of plants were easily noticed. Using the traditional measuring tape method to count in the void area, 11 plants were counted down the 17.5-foot stretch of corn along the row. This number scales to an estimated 11,000 plants/acre which is within SlantRange’s estimate. The agronomist was not surprised and speculated that the likely cause for the lower population count was sandy soil resulting in improper seed depth and inadequate down force.
The next area surveyed estimated a population count of 32,000 plants/acre and ideal estimated range at 28,000 - 34,000 plants/acre in the dark green area. At the site, you could see full emergence of healthy corn. Using the measuring tape, 32 plants were counted along the 17.5-foot stretch of corn along the row. This scales to an estimated 32,000 plants/acre which was exactly where SlantRange estimated.
Population Count Results
The overall accuracy of the SlantRange population count was impressive. Traditional ground truth methods of population counts in the three areas show the plant counts fell within the estimates given by the population map. The ability to count accurately using aerial imagery gives the grower a more complete picture of field emergence. It also gives the grower a view of the population variability across the whole field. Traditional population count methods provide a glimpse into small areas and then extrapolate that count over a much larger section of the field. The results in this case study show that our data can make your time spent in the field much more valuable and provide much more accurate estimates of your crop emergence and population.
SlantRange Plant Size and Weed Maps
Some other data products that also came from the same dataset and processing are the weed and plant size maps. The weed map was not ground truthed for all locations but the map below shows weed pressure in red and a corresponding high resolution image with obvious weeds.
For the plant size map, we used SlantView to view the high resolution imagery for different areas of the map. In the first example below, the red area showed plants at 0.8 on a normalized scale with 1.0 set to the median plant size, or 20% smaller than the median. In the second example, the map showed plants in the dark green areas as 1.14 or 14% larger than the median. While we did not take physical measurements of the plant sizes in these areas, qualitative analysis of the images appeared to confirm the variability in the plant size map. Variability in plant size at this stage of growth can be due to a variety of causes and can indicate a number of issues to discuss with your agronomist.
Precision plans to continue flying the corn test plot utilizing more of SlantRange’s data products to cover the whole growing season. About three weeks after the population count, the corn is expected to be 18” high and a good time to look for plant stress conditions and additional weed infestations. We plan to use the upcoming data to make decisions on nitrogen and herbicide applications.
Why Should I Use Aerial Agronomy?
Increased efficiency is the goal of every grower and Precision believes aerial agronomy is an essential part of the solution. Nothing can replace ground truthing the data collected by UAVs. With that said, UAV imagery makes scouting your fields vastly more efficient. As we learn what UAVs are capable of in the area of aerial agronomy and find progressive growers like Green Thumb Farms willing to invest in the trending data to make that data even more reliable, we move ever closer to a time when UAVs are the cornerstone of a Precision Agriculture System. To that end with this trial, SlantRange has proven its worth today.
"Nothing can replace ground truthing the data collected by UAVs. With that said, UAV imagery makes scouting your fields vastly more efficient."
Jesse Badger, Precision Unmanned Aerial
About Jesse Badger and Precision Unmanned Aerial
Jesse Badger is the owner of Precision Unmanned Aerial, which has been providing growers with UAV survey services since 2013. He’s long been a believer in UAV use in agriculture and has created a successful business from aerial agronomy. Since inception, Precision Unmanned Aerial has been at the forefront of unmanned aviation technology and has firsthand experience of the pros and cons that come with the use of various spectral cameras, UAV platforms, and data analytics software programs. Precision Unmanned Aerial strives to provide its customers with the best data and works toward creating a true, actionable Precision Agriculture management workflow. During the 2017 growing season, Precision Unmanned Aerial will continue to work with us to employ cutting edge sensors in the field and share their thoughts with us. Follow Precision Unmanned Aerial on Twitter and Facebook or contact them directly at firstname.lastname@example.org for more info on how they put SlantRange technology to work.