Drones have recently become an essential device for field observations and aerial phenotyping. In one of our recent articles, we talked about how they help seed breeders and producers screen a large number of plots quickly and efficiently, and generate data at the plant level.
Reaching data accuracy on your experimental fields requires to fly the right drone, at the right time, with the right settings. It is easy, but needs a little bit of preparation. In this guide, we will walk you over the basics of drone configurations, from your optimal GSD based on your desired traits, to the ideal weather conditions as well as other tips to get the perfect scouting maps and analytics.
If you currently fly drones on your experimental fields to measure plants count or other plant characteristics such as plant height or flowering, or if you’re just exploring the drone-based aerial phenotyping world and you’re unsure which type of drone to choose from, then this guide is right for you.
Drone flight recommendations based on the trait measurement you need
GSD is key
Before we dive into the specifics of drone flight configurations, it is important to define GSD. GSD is the real-world size of a pixel in your images. It means Ground Sample Distance. In an orthophotography or any other georeferenced image, it is the distance between two consecutive pixel centers measured on the ground. The lower the GSD, the smaller the image pixel, the more detailed the map.
The type of analytics you need at a given crop stage will determine the GSD. It is crucial to find the best compromise to optimize quality results and work efficiency: flying too high will increase the GSD and reduce the image resolution, which can prevent you from getting an accurate plant count, for example. Flying low will lead to high detailed information, which is great to generate a detailed inventory of a crop, but you should consider the risk on data quality (blurry image and higher risk of weather and crop changes due to increased flight time) as well as a longer flight and data processing time because there will be more images to process.
Your desired trait defines your GSD, thus your drone altitude
Cloud platforms help you extract various analytics than can be used to measure traits along the crop season. A good rule of thumb is to use a target GSD between 6 and 10 cm if you’re planning to do any of the following:
- Measure plant health
- Vectorize your micro-plots to define/geolocate their boundaries
- Measure plant height
- Characterize flowering
- Extract stay green
- Map the fraction of ground covered by green vegetation (FCover)
This range, which corresponds to a flight altitude between 90 and 150m with the UX11AG drone, is sufficient to get you accurate analytics.
The only traits that generally require a much lower GSD are the plant gap & count as well as emergence quality, especially if your plant canopy size is rather small (<12cm) since you have to extract them at an early stage of development
We’ve compiled all phenotyping related analytics in a table with important considerations to keep in mind. Overall, you should always pay attention to the light and weather conditions since the vegetation indices are based on a precise measure of each spectral bands. But we’ll talk about the weather conditions later in this guide.
|Key traits||Optimal GSD||Recommendations|
|Plant health maps||6-10 cm||You can assess plant responses on different topics through multiple scouting maps*: RGB map (bird view) and the vegetation indices: crop vigor (NDVI), green biomass (MCARI2…), chlorophyll (NDRE)|
|Micro-plots vectorization||To create the microplot vector properly, vegetation should be developed enough to see the microplot boundaries but not too spread out either to the point where you have trouble identifying alleys between micro-plots.|
|Plant height||Best practice is to compare the DSM (Digital Surface Model) layer of 2 acquisitions:, ie one survey on bare soil and another one with the vegetation you want to measure. The first flight does not require a low GSD: 9 or 10 cm is enough.In case you missed the bare soil acquisition, extrapolation to estimate crop height is feasible but will lead to lower accuracy of the results.|
|FCOVER||6-10 cm||It is interesting to fly those traits with a bi-weekly frequency to assess the start and the end, the velocity and their temporal evolution.|
|Plant gap and count||Twice smaller than the canopy size||For example, if your plant diameter is 8cm, you can fly your drone with a GSD of 4 cm.
With a multispectral camera, you can target a GSD of 6cm for corn in a 4 leaf stage and for sunflower in a starbud stage. With enhanced resolution (smaller GSD), earlier crop stages can be targeted. It is also important to see bare ground in between two plants on the map.
In addition, we’ve made this infographic that is easy to refer to (and print) as you prepare your flight. It includes recommendations for both fixed-wing drones, such as the UX11 AG and multirotor drones, such as the DJI Phantom 4 Multispectral, among others.
*Scouting map : Aerial map that allows you to visualize details of your fields on different aspects. Scouting maps include the RGB map (bird view) and the vegetation indices: biomass (MCARI2), chlorophyll (NDRE),…
Basic default drone flight configurations
To get the best image quality and radiometric accuracy, right weather conditions during flight time are crucial. Although our UX11 AG and many other professional mapping drones are resistant to moderate rain, we don’t recommend you to fly your drone when it’s raining, even if it’s just a drizzle, because you will get blurred images and low quality data. Likewise, partly cloudy conditions aren’t ideal as the light keeps changing and rolling clouds shadows can end up getting captured by your drone sensor.
The ideal sky cover when you want to fly your mapping drone is clear to ensure a good radiometric calibration over the whole field. Surprisingly, it is absolutely fine if it’s cloudy as well.
As long as the cloud coverage is constant, meaning that there isn’t any sunny spot, or partial cloud clearance there won’t be any issue.
The maximum wind speed will depend on the type and model of drone you use: for fixed wing drone, you shouldn’t exceed 12.5 m/s (45 km/h) and for multirotor drone, you shouldn’t go faster than 10 m/s (36 km/h).
The optimal flight time of day is within an hour and a half of local solar noon, to avoid deep shadows from the plants which can affect the radiometric accuracy.
Image overlap :
The overlap you set between successives images taken by your drone will determine the success of your map. The side overlap corresponds to the overlap between the 2 images side by side and the front overlap is the overlap between 2 images in front of each other.
A low overlap won’t generate a precise reproduction of your map and plants. A high overlap will give you a faithful reproduction but will also require more processing time due to the large volume of images taken. So, similar to the GSD, you need to find the right compromise before your flight. We recommend you to set up an overlap higher than 75% front and side.
Geolocation accuracy of map
GNSS stands for Global Navigation Satellite System, and is more commonly referred to as GPS. Most drones on the market use a standard GNSS system, delivering basic accuracy of approximately one meter. This is precise enough in some cases but is too limiting for some survey-types uses such as phenotyping operations.
In this case, better accuracy is required from the GNSS system in order to precisely overlay maps from flights throughout the season, and extract key traits within micro-plots at different stages. This is where GNSS data correction comes into play: this technique enables centimeter-grade accuracy and comes in two variants: PPK (Post Processing Kinematic) or RTK (Real Time Kinematic).
These two techniques make use of a base station that we recommend be as close as possible to your survey site (15 km max). Depending on your drone make and model, and location, you would either use a mobile station (provided by your drone manufacturer or another supplier), or a fixed station already available through government or academic agencies. The mobile stations are usually highly portable antenna-like devices, roughly the size of a 15 cm wide cube.
Another more traditional solution to achieve GNSS data correction is to place 5 GCPs (Ground Control Point) across your survey site: one at each corner and one in the center. These GCPs don’t have to be georeferenced if they can stay at the exact same place during the season. This option is more time consuming, and also more risk-prone as an inadvertent displacement of a GCP on the ground could result in inaccurate data.
Other tips for success:
For a perfect orthomosaic, we recommend that you capture a wider zone than the area of interest to ensure a high number of images acquired in the center of the flight, ie over the area of interest. Finally, you should pay attention to the maximum speed of your drone : for a fixed wing drone, the recommended flight speed is about 14m/s and for a multirotor, it is 8m/s.
Multi-rotor or fixed-wing drone?
Opting either for a fixed-wing or multirotor drone can be a complicated choice.
Basically, with a multirotor drone, you can obtain a smaller GSD than with a fixed-wing drone. For some applications, such as counting at an early stage of plant development, it is recommended to opt for a multi-rotor drone. However, the productivity will be well below a fixed-wing drone as the multi-rotor battery life doesn’t usually go over 15 to 20 minutes and the multi-rotor drone flight speed is lower. To get rid of this issue, multi rotor pilot tend to decrease the overlap and sidelap, with impact on the final quality of the data.
With a fixed-wing drone, you will be able to fly your field faster in a single flight, and bigger fields. So if you have multiple fields or if you need to survey multiple times in a single day, then this type of drone is highly recommended. You also lower your chances of getting light changing conditions since the survey will be completed in a shorter amount of time. Finally, the design and flight behavior of fixed-wing drones allow you to fly in windy conditions (up to 12.5 m/s), which is quite common in agriculture fields.
You will find here a tool that allows you to calculate the flight time of the UX11AG fixed-wing drone for a given GSD, overlap and surface.
As a summary, fixed wing drones are ideal drone to deploy at large scale, especially when:
- You need productive drone for fast and frequent mapping, even in windy conditions
- You are looking for a drone that is easy to use with steep adoption curve (check list, security, limited maintenance)
- The GSD is not limiting (>4cm)
On the contrary, multi-rotor drones are ideal for new drone users operating small field, especially when:
- You only need to survey field occasionally
- You need to run specific studies with very high resolution <4cm for example to count plants at very early growth stage
Once you identified your drone based on your needs and environment, it is important to pay attention to several points. Your desired results will impact your flight parameters. Here’s a checklist that you may want to use before each flight:
- Decide if you need precise geolocation of your maps:: post (PPK) or during flight (RTK)
- Check your weather conditions: wind, constant daylight, no rain expected
- Make sure you choose a GSD that is aligned with your desired trait
- Define a side and forward overlap to > 75%
- Set up a flight speed: 14m/s max for the UX11 AG or 8m/s max for a multi-rotor drone
In addition, make sure that you don’t need specific authorisation to fly over your area on the D-day. More generally, it is important to plan your flight program well before your breeding season starts. What we mean by a flight program is a schedule of flights for your entire field trials in the season. It can take the format of an excel file that displays all your experimental fields, their number of flights needed for the season to measure specific traits, flights parameters for each survey, and your estimate flight date. Your flight program will help you assess the total number of flight hours and the amount of human hours needed on the ground. We’ll make sure to publish a blog post about it in the near future.