Integrating Unmanned Aerial Systems Into the Crop Production System Through On-Farm Research

Precision agriculture strives to manage variations in the field in order to increase yield while adapting input factors to preserve resources and decrease production costs. Unmanned aerial systems (UAS) are advancing precision agriculture by allowing for nondestructive and convenient, as well as cost and time efficient mapping of spatial variation in fields with higher spatial resolution than previous methods. However, while there is much anticipation regarding the potential role for UAS in precision agriculture, their role still requires additional application-based testing. The objective of this work was to explore how growers best integrate the UAS product into their farm workflow. Two on-farm investigations were undertaken with vegetable growers for the duration of a growing season. Combinations of two unique unmanned aircraft (UA) platforms fitted with two different multispectral sensors were used to gather spectral reflectance data. The investigations found that the UAS product enabled the growers to optimize their field management practices, while overcoming a labor shortage, and create a more sustainable operation.


Introduction
Growers are increasingly being tasked with producing more food with fewer resources while reducing their impact on the environment.The goal of meeting society's food needs without compromising the ability of future generations to meet their own is the aim of sustainable agriculture (Feenstra et al., 2018).In order to meet this goal, in conjunction with continuous population growth, innovation in agriculture must continue.
While the use of remote sensing platforms, such as inhabited aircraft and satellites, have been used in precision agriculture for some time, the implementation of UAS into precision agriculture holds unique promise.Unmanned aircraft (UA) enable aerial imagery with greater resolution, a higher temporal frequency, and reduced costs.The ability of a UAS to obtain a spatial resolution on the order of centimeters and fly daily is in stark contrast to low resolution satellite imagery and manned aircraft overflights a few times a growing season.The availability of this nascent high spatial and temporal resolution data empowers the grower to more proactively manage crop health.
UA were first used in agriculture as far back as 1990 for the spraying of rice paddies (Sato, 2003).However, subsequent access to high quality global positioning system (GPS) signals and cost effective GPS receivers, advancements in integrated circuits and battery technology, and the miniaturization of aircraft systems has now led to the proliferation of small and affordable UAS for agriculture.UA, equipped with a wide variety of sensors, have more recently been shown to be effective in agriculture to manage irrigation (Gonzalez-Dugo, Goldhamer, Zarco-Tejada, & Fereres, 2015), ascertain soil moisture (Hassan-Esfahani, Torres-Rua, Jensen, & Mckee, 2017) discriminate plowing techniques (Tripicchio, Satler, Dabisias, Ruffaldi, & Avizzano, 2015), investigate the microclimate over crops (Adkins & Sescu, 2017), estimate yield (Geipel, Link, & Claupein, 2014), and monitor crop stress (Stanton et al., 2017), amongst other uses.
While there is much anticipation regarding the potential role for UAS in precision agriculture, their use still requires additional investigation into how they can be most effectively implemented by commercial growers, as much of the research to date has taken place on research farms (Thomasson & Valasek, 2016), (Elston, 2016), (Komp, 2018), (Bendig, Yu, & Aasen, 2015), (Nebiker, Lack, Abächerli, & Läderach, 2016).Consideration needs to be given as to how a grower best implements the technology into their farm and crop management workflow.The employment of new technology by a farmer in the absence of thorough on-farm evaluation, or knowledge of how best to integrate the technology into the farm workflow, presents a high risk.The main goal of this work was to investigate, in conjunction with the grower, how to best integrate UA-based multispectral imagery into crop management decisions.This investigation took place on two farms.Each grew an assortment of vegetables that included mustard greens, cabbage (white, Shanghai (bok choy), and wawa), and Chinese white radishes.

Method
Farmers today have many crop management tools and on-farm research can assist farmers in evolving their management strategies and decisions.This on-farm research utilized various state-of-the-art, commercial off-the-shelf (COTS) UA, sensors, and post-processing software.COTS hardware and software were chosen in order to make the operation viable to a representative farmer.Per the growers' feedback and acceptance, the acquisition, processing, analysis, and application of the data was continually adapted based on lessons learned.This evolution moved operations toward a pragmatic and relevant best practice.The overall generic UAS workflow consisted of flight planning, flight and the capturing of imagery, same day local processing of the data, and the creation of vegetation index maps.The sharing, analysis and ensuing discussion of the results informed the required actions on the part of the grower and established the timeframe for the next flight.All flight operations were conducted under 14 CFR Part 107 of the Federal Aviation Regulations (FAR).Furthermore, all crew members were certificated by the Unmanned Safety Institute (USI) in small UAS safety.

System Description
Commonly referred to as a drone, a UAS is a system comprised of a number of sub-systems to include the air vehicle (often called an unmanned aircraft (UA) or unmanned aerial vehicle (UAV)), the payload (sensors), a control station (CS) (most often a ground control station (GCS)), aircrew, data link, launch and recovery equipment, maintenance and support equipment, and an operational space consisting of rules and regulations (Austin, 2010).To make the operation and results applicable to the typical grower, two different, but widely available, fixed-wing UA were utilized.The majority of the flights were undertaken with the senseFly eBee, Figure 1a, with a few select flights completed with the Parrot Disco-Pro Ag, Figure 1b.Whenever the Disco Pro-Ag was deployed, the flight was immediately followed-up with the eBee platform for consistency.Flight planning for the eBee was accomplished using the accompanying eMotion software on a laptop, which also served as the GCS; planning for the Disco-Pro Ag flights was achieved with the Pix4Dcapture app on a mobile tablet.All data processing was completed using Pix4Dmapper Pro Ag. jas.ccsenet.

Data A
The on-far (Figure 2    inspection.The trade-off between extensive spatial coverage and exhaustive inspection was made especially acute during the associated growing season by the farm labor shortages experienced as the result of the simultaneous conversation in the United States regarding immigration and deportation policy.One of the earliest recognized benefits, by the grower, of the UAS was the comprehensive scouting it enabled in the midst of the labor shortage.The UAS thus enabled comprehensive, high resolution scouting that accounted for the irregular distribution of weeds.Further, analysis of SAVI and NDVI images during this stage of the season alerted the grower to planting and germination issues within the field.

The remo measurem temporal v manageme
The growing season in Florida largely coincides with the region's dry season and the period of time during which the investigation took place was abnormally dry with an absence of any precipitation in the fields for 3 months.In order to avoid excessive soil salinity realized from long-term subsurface irrigation (water table control), an irrigation gun was utilized during critical periods of crop development, and to activate fertilizer following its application.As crops matured, and crop canopy density increased, NDVI-derived maps were more singularly utilized to reveal heterogeneity within the fields.Areas identified as being stressed were specifically targeted for inspection.With the unusually dry growing season, the stress often identified was water stress.The high resolution NDVI map was then subsequently used to guide irrigation.This strategy minimized water usage and decreased the labor effort associated with the irrigation gun.
The ability to monitor temporal variation within the field through regular overflights allowed for the continued observation of plant growth and examination of crop health.In addition to red based NDVI, green based GNDVI images were created and shared with the grower for this purpose.These images helped inform the need for chemical treatment, the required rate of treatment, and allowed for the monitoring of the efficacy of the treatment.While the images readily identified areas needing attention, one benefit unanticipated by the growers was their ability to see the effectiveness of the treatment, specifically following fertilization, prior to being able to visually perceive it in the field through plant color.Figure 5 shows similarly scaled NDVI images taken of two adjacent subplots.The left-hand side of the field contained radish and the right-hand side contained cabbage.From left to right, the images in the figure show an image just prior to fertilization (left image), five days after fertilization (center image), and thirteen days following fertilization (right image).Each field was fertilized on the same day, with the fertilizer activated at the same time by irrigation.No further action was taken on either of the fields during the following two weeks and a uniform visual improvement in plant color was not observed until two weeks elapsed.However, the NDVI images clearly show an improvement in crop health prior to this point in time.The radish field, on the left-hand side, exhibits a swifter improvement than the cabbage field, on the right.The grower attributed this difference to the increased nutrient uptake associated with the increased surface area of this root vegetable.

Figure
Figure 5. S right: a) le