Why Field Trials?
Tens of millions of agronomic field trial plots are planted globally every year. It is tremendously expensive and logistically complicated to seed, cultivate, and harvest individual small plots across multiple locations in order to measure how different varieties rank against one another. Yet, there is currently no other way to predict how a given genetic background (cultivar/variety) will perform in a specific environment and across seasons. This variation is known as genotype-by-environment (GxE) interaction or phenotype plasticity. What are some of the environmental variables which differentially impact performance of a given variety?
Seeding rate & row spacing
Irrigated or dryland Farming
Photoperiod across latitudes
Climate & weather (year)
Another reason for conducting field trials is to test crop inputs such as seed treatments, fertilizers, crop protection products, and biologicals. In these experiments, often referred to as GLP field trials (good laboratory practices), only a handful of proven varieties are planted. The variable of interest is rate, dosage, formulation, and so on.
Stand establishment: a critical determinant of yield potential
Crop yield is a function of plant population multiplied by individual plant yield. A key event in setting up potential yield is stand establishment. A variety which germinates quickly, uniformly, and reliably will have the best chance of intercepting the most sunlight, taking up available water, and mining mineral nutrients over the growing season. If germination is spotty, emergence is not uniform, or if a crop canopy does not fill out quickly, then yield potential is diminished and weeds have an opening to compete for resources. Conversely, while breeders have selected for tolerance to crowding -- such as corn with leaves that grow more upright, at a certain point returns on density are diminished. Thus, the upper limits of plant populations for a given variety and location needs fine-tuning as well.
Plant population is determined sometime after full emergence by counting seedlings or young plants. Traditionally, small plot trials have been counted manually: row-by-row, plot-by-plot, and field-by-field. Obviously, this is time consuming, monotonous, and therefore prone to errors. It requires legions of temporary helpers at a time of year when labor is less available. Thus it is often impossible to count every single plant. In many cases only certain sub-rows are counted to save time.
From boots on the ground to drones overhead: stand counts from images
It is no surprise therefore that "drone stand counts" has become a popular search term among breeders and agronomists. The allure of replacing boots on the ground with a UAV overhead has obvious appeal. Only recently, however, has there been a convergence of hardware and software in terms of price, resolution, and usability by non-GIS experts to make counting stands with drones a reality. Now, with an off-the-shelf UAV a licensed pilot with a laptop can complete an automated mapping flight of an average field trial in about twenty minutes. If the pilot is also adept with photography settings, then they have everything needed to perform stand counts on a laptop from their pickup.
The following clip compares counting stands by eye versus UAS-acquired images and Plot Phenix software (view the full video on our YouTube channel):
The UAS plus software approach to stand counting enables additional analyses. For example, is a missing plant indicative of a skip or is it more likely a seed which didn't germinate? Software can count both skips and doubles. If the two events are usually in sequence, then there may have been a planter problem rather than an issue with the seed viability. Few people can keep track of and record a tally of skips, singles, and doubles while walking a field.
(Left: stand counting output from Plot Phenix for a single corn trial plot. Green crosses are singulated plants, blue indicates doubles, and red minuses are skips)
Which crops which can be counted via photogrammetry?
Corn, soybeans, rice, onions, tomatoes, lettuce, and more have been counted from imagery small plot field trials. If it's planted in rows with sufficient space within- and between-rows, we can probably count it.
What is the best time to fly for stand counts?
Algorithms for plant counting are trained to associate clusters of green pixels in an image as a single plant. If plants are beginning to overlap in the overhead view, then it becomes harder to distinguish individuals.
On the other hand, if a seedling has only recently emerged, software may have have difficulty finding enough fully green pixels to correctly identify a plant. The quality and resolution of the image also contributes to how early stand counting with a UAS is feasible.
Therefore, the best time to fly a trial is just before plants have begun to appear to overlap in the images, but also not too soon after emergence. For example, in corn it is recommend to fly around the V2-V3 phenological stage. Furthermore, the flying height required for stand counting is relatively low compared to other metrics extracted from images of larger plants, such as canopy cover.
UAV/drone: unmanned/unoccupied aerial vehicle.
UAS: unmanned/unoccupied aerial system. Includes the UAV airframe, plus the operator and associated systems for control, as well as any payload such as a camera