Team uses SwitchBlade-Elite research drone to measure river ice thickness to predict seasonal flooding.
North America experiences about $250 million in economic costs annually for damaged infrastructure, agricultural losses, and flood recovery due to seasonal river ice jam flooding. Researchers in the Department of Earth Sciences at Montana State University (MSU) are developing methods to predict the timing and magnitude of ice jam floods.
The MSU team employed scientific research drones for part of the project. They selected Vision Aerial’s SwitchBlade-Elite because it has an open architecture, is highly versatile, handles well in harsh environments, is easy to transport, is made in America, and is supported by exceptional customer service.
“In snowy environments there are often problems with camera over exposure. Because the SwitchBlade-Elite supports multiple payloads, we can customize our camera settings for different surface conditions. Therefore we have a lot of tools at our disposal to compensate for potential overexposure.”
Ross P. (Key Researcher)
Research Overview
Background and Goals
An ice jam is a stationary accumulation of ice in a river that restricts water flow in the channel. As rivers begin to swell with spring snowmelt, increased discharge can break up an ice jam very suddenly and send a powerful surge of water and ice downstream. On a large river like the Yellowstone in eastern Montana, these ice jam floods can cause millions of dollars in damage and potential loss of life.
In many rural areas of the United States, the primary method used to get information about river ice extent is through visual observations. Physical access greatly limits these studies. The proximity of a road to a river determines where measurements can be taken.
Current techniques are also lacking in the information they can provide. Visual observations cannot provide quantitative information about ice thickness or volume. Researchers are left with open questions such as:
- How much ice was present when a flood occurred?
- Does ice extent correlate with volume?
- How do extent and volume correlate with ice jam flooding?
The MSU team’s goal is to improve ice jam flood forecasting. To do this, they are measuring current ice conditions and correlating them with projected meteorological variables. As the study progresses, they plan to use artificial intelligence to detect and quantify river ice data in satellite imagery. The final result will be a tool that identifies river ice in satellite imagery, then provides quantitative information that flood forecasters can use to make more accurate predictions.
Challenges with Satellite Imagery
The team needed high resolution images that could be used to define characteristics and categorize different types of river ice. Satellite images are taken from very high altitudes and therefore are not very detailed. A typical satellite image is about 10-30 meters per pixel.
Additionally, ice thickness cannot be measured with satellite images alone. Researchers typically use borehole drills and ground penetrating radar to gather measurements. This is incredibly dangerous because researchers need to walk out on partially frozen, unstable rivers in freezing temperatures.
Drone Photogrammetry as a Solution
With the above challenges in mind, the research team identified drone photogrammetry as a solution.
Low-resolution satellite imagery could be validated with high-resolution photos collected by a drone. Satellite image resolution is typically 10-30 meters per pixel and lacks fine details. On the other hand, drone imagery typically achieves a much higher resolution, around 2 centimeters per pixel. (Learn more about Ground Sample Distance)
The team could also use a drone for determining ice thickness measurements. Digital elevation models produced by drone data could be used to measure ice thickness.
This solution increases access and also improves safety. Researchers are not limited by road proximity, and they don’t have to walk out onto frozen rivers. The next step was selecting the right drone for the job.
Considerations in Research Drone Procurement
Drone Selection Requirements
Challenges that shaped the team’s requirements for selecting a drone for research:
- Collections are often taken in remote locations and require a drone that is easy to transport.
- Data collection about river ice characteristics is performed in harsh environments that are sometimes very windy.
- Ongoing changes in technology and processes require a flexible system that can adapt or use multiple sensors as needed.
- Project received federal grant money, therefore could not use drones made in China.
- Researchers need access to flight data and the ability to adjust other parameters.
The team looked at several unmanned aerial vehicles. On some, it wasn’t easy to swap out payloads. Fixed wing aircraft were ruled out because they generally don’t handle well in the wind. In most cases, drones featured a “black box” construction that didn’t allow researchers to access the flight data or adjust parameters. Many drones came with a fixed camera, where it was difficult (or impossible) to change the settings in order to prevent overexposure when shooting in bright conditions. The team needed an open architecture solution that could be customized. They also needed a company that could provide support for technical challenges as needed.
In the end, the team selected the SwitchBlade-Elite for their research. The chart below outlines how the SwitchBlade-Elite met their requirements.
How the SwitchBlade-Elite Satisfied the Researcher’s Requirements
| Requirement | How Does the SwitchBlade-Elite Satisfy the Requirement? |
|---|---|
| Must be easy to transport to remote locations. | Folds for transport. |
| Must perform well in cold, windy environments. | Can handle wind gusts up to 25 mph. IP52 rating. Operates down to 5°F. |
| Need to be able to adjust camera settings to prevent overexposure in bright conditions. | Cameras are not hard wired to the drone. Settings are exposed to the user and can be adjusted. |
| Must allow multiple sensors and adapt to new technology. | No hard-wired payloads. The Payload Connection System allows payload swapping. Adaptable to future technology. |
| Ability to access flight data and adjust parameters. | SwitchBlade-Elite does not feature “black box” construction. |
| Federal grant required an American made system | Vision Aerial drones are made and supported in the USA. |
Conducting the Research
With a SwitchBlade-Elite in hand, the team just spent the season collecting data for their project. This is an overview of how the research was conducted.
Equipment Used
- Drone: SwitchBlade-Elite tricopter
- Payloads:
- Sony α6000 camera for high-resolution images with fixed nadir mounting.
- Flir Duo Pro R for infrared imagery mounted on a gimbal.
- MicaSense RedEdge for multispectral images. Fixed nadir mounting.
- Accessories:
- Indefinite Flight Package
- RTK
- Ground Control Points
- Software:
- QGroundControl (Mission Planning)
- Agisoft Metashape
Team Member Roles
- Grant Writer(s) – Ross Palomaki, Eric Sproles
- Pilot(s) – Ross Palomaki, Maddie Beck, Andrew Mullen
- Photogrammetry Software Operator – Ross Palomaki, Maddie Beck, Andrew Mullen
- Software Developer
Survey Sites
Yellowstone River in Montana. Total survey area is 250,000 m2.
Field Research Process
Before a day in the field, the pilot creates a flight plan with the mission planning software. Once they arrive at the field, the team sets up ground control points. Setting up ground control points ensures the accuracy of the images. The team also uses RTK to ensure centimeter-level accuracy.
The SwitchBlade-Elite is fitted with a Sony α6000 high-resolution RGB camera. The pilot then deploys the drone and enables the mission. During the autonomous flight, the drone takes hundreds of images of the river ice. Images contain location and camera attributes.
The team relied on Vision Aerial’s expertise to help them configure their system. In order to gather all the data they needed, the team had to access the flight logs, IMU, altitude, and GPS data.
“The service we received in troubleshooting has been very helpful. It’s been incredible to work with the (Vision Aerial) team. The SwitchBlade-Elite is a very capable platform, and the customer support has been the icing on the cake.”
Because of the area covered, flight altitude, and level of overlap, it takes about 2 hours to complete the ice data sampling. In order to maximize flight time, the team uses the Indefinite Flight Package (IFP), which allows them to charge batteries faster than they are consumed.
Field Research Outputs
Back in the office, the team used software to stitch hundreds of images together with the IMU data into an orthophoto. They also generate 3D models in Agisoft Metashape to measure surface roughness and surface height.
In order to determine ice thickness, one elevation model was taken during ice-free conditions. A second model was generated during winter when the river was frozen. Differences in water levels are measured, and then the data was compared between the two models to calculate ice thickness.
Next Steps
Data collection from the field is complete for the first season. Now the team needs to continue processing all the images that were captured. They will develop digital models, and then calculate ice thickness and volume. When the digital models are complete, they will be compared with satellite images. The team will continue to train the neural network to identify key indicators in the satellite imagery that will aid in flood and crop health predictions.
“Who knows what we’ll be using the drone for in 2-3 years? Having access to raw data makes it more versatile for new projects while we’re cooking up the next thing.”