Callicoon Field Work: Drone Mapping and Imaging

Spatial Analysis for Sustainable Development (SDEVUN3450) field trip to Callicoon, New York; April 4-5, 2022.

Introduction

Our Goals

As part of the Spatial Analysis for Sustainable Development (SDEVUN3450) class, we traveled to Callicoon, New York for "Flight Week" – the dedicated time for us to fly drones in the culmination of learning the physical piloting skills and analytical programming skills we have been acquiring and practicing for the duration of the spring 2022 semester.

Our course goals included gaining enough knowledge and practical experience with Unmanned Aircraft Systems to be capable of passing the U.S. Federal Aviation Administration's (FAA) Part 107 exam. Passing this exam grants an individual licensure to fly small unmanned aircrafts, including drones. Attaining the skills to program and operate drones, in conjunction with the spatial analysis skills at the foundation of this course, provides countless opportunities for real-world application—whether in further research or in a future career.

Our goals for this trip were to explore the land surrounding the Villa Roma Resort, getting hands-on practice with flying drones and collecting raw aerial imagery data on a drone mission, and to mimic what it might be like to work as a drone pilot as an occupation.

Our Experience

Our entire class of eleven students, plus our professor and teaching assistant (TA), were able to attend the fieldwork trip. The students and our TA convened on Columbia's Morningside Heights campus, where we loaded our belongings into a van and began the two-hour road trip to our final destination.

Upon arrival at the Resort, we were able to settle into our individual rooms and decompress from the drive. Then, our class gathered in the mission control room to review FAA weather reports and forecasts, review our standard operating procedures, and complete preflight checklists to test and prepare each of the three drones for flight: the Mavic 2 Enterprise, the Mavic 2 Pro with add-on Survey3 MAPIR infrared camera, and the Skydio 2 Enterprise.

Top Left: Skydio 2, Top Right: DJI Mavic 2 Enterprise, Bottom Left: DJI Mavic 2 Pro, Bottom Right: Survey3 MAPIR Camera

After dinner together in the Resort's dining room, we set out to conduct our nighttime drone flight, hoping to capture and model light emanation as a proxy for the presence of human populations. Our launch point had to be at a higher elevation than the resort, so we found an elevated section of a parking lot from which to launch our mission.

Day 2 of our trip consisted of the majority of our drone missions. After breakfast, we set out across the golf course to access the Villa Roma clubhouse, a structurally complex building with intricate roofing features, to practice our 3D mapping skills using the Skydio 2 drone.

Following lunch, we hiked up a small ski hill to gain a better vantage point for drone launch and to access different airspace and land cover views than had been explored around the golf course and clubhouse. The ski hill missions included both the vegetation and terrain data, whose results can be seen below.

Processed Data and GIS Outputs

Nighttime Lights

On the evening of April 4, we headed to a relatively clear and elevated area to collect nighttime lights imagery.

While we were staring up into the sky at the drone, oblivious, a family of deer approached our launch site!

We utilized the DJI Enterprise with its anti-collision lighting so that the mission complied with FAA requirements for safety.

A sample of raw images from the drone before processing. The team can be seen in the bottom right photo.

In order to convert these raw drone images into GIS data, we added the nighttime images to ESRI's ArcGIS Drone2Map software, which made it easy to create an orthomosaic and dynamic tile layer, as pictured below, and shared them publicly as a service on ArcGIS Online.

Orthomosaic of nighttime lights images on the right, with underlying built features on the left.

3D Model

It was a cloudless morning as we set out on a short hike to the Villa Roma Clubhouse, on the second day of our trip. After checking out the structure on foot, we debriefed in the parking lot. The drone we used for this project was the Skydio 2 Enterprise, an autonomous unmanned aircraft that can make decisions on the fly. It can even lock onto moving objects and follow them!

A sample of raw images from the drone before processing.

Once the Skydio took off, it scanned the building, using its proprietary algorithm to calculate the best, most energy-efficient path to create a 3D rendering of the building. Next, the drone began its rendering analysis, snapping well over a thousand images of the building from all angles. To make sure no accidents occurred, we spread ourselves out around the building as visual observers. Watching the drone fly autonomously was both fun and a little nerve-wracking, especially when it made sudden movements around trees or came close to the clubhouse balcony.

The initial output of all these images once processed was a 3D point cloud that we classified as different surface types. In the image below, red indicates building surface, while brown and green indicate ground and vegetation respectively.

The Villa Roma Clubhouse as a 3D point layer.

Using this 3D point cloud, a mesh was created. Though not perfect, the drone was able to capture an amazing amount of detail. It even picked up generators on each side of the building!

The Villa Roma Clubhouse as 3D mesh layer.

NDVI Vegetation Mapping

After lunch and taking some time to regroup in the mission control center, we climbed the ski hill next to Villa Roma where we would begin our final drone missions. While walking to our destination at the top of the hill, we used the ArcGIS Collector app to record ground control points for georeferencing.

We used the DJI Mavic 2 Pro with a Survey3 MAPIR near-infrared camera to collect data on vegetation in the area, allowing us to calculate a normalized difference vegetation index (NDVI). We also produced an orthomosaic of the infrared images.

A sample of raw images captured by the Survey3 MAPIR near-infrared camera before processing

Orthomosaic of infrared images on the left, symbolized on the right by the NDVI scale. Greener spaces represent more vegetated areas, and red areas represent built, less vegetated areas.

Terrain Mapping

Digital Terrain Model

We used the DJI Mavic 2 Enterprise to gather images to be used for mapping elevations and deriving information on how surface runoff could traverse the terrain.

A sample of raw images from the drone before processing.

After the drone gathered the imagery, it was processed to create an orthomosaic and digital terrain model. Below is the digital terrain model visualized in two different ways:

Digital Terrain Model with 1 meter contours (left) and hillshade (right). Redder and whiter color indicates higher elevation and greener color indicates lower elevation.

Watershed Delineation

After the digital terrain model was created, we used this data to create a watershed delineation. This involved preconditioning the data by filling any erroneous outliers. Then the watershed was delineated identifying a watershed area using a pour point, and determining flow accumulation.

Watershed Delineation, symbolized by flow direction (left) and watershed boundary based on selected pour point in red (right). The delineated watershed represents all locations on the surface which drain into the selected pour point.

Bonding and Learning As Peers

Photos of the Trip

Views of the area surrounding the Villa Roma Resort, and our class eating dinner.

Top row: distant and near view of the Villa Roma clubhouse adjacent to their golf course which we 3D-modeled using a drone. Bottom row: view of the stationary ski lift, view of the ski hill with the drone flying above, and two students (Madeline Liberman and Lauren Khame) enjoying the ski lift.

Videos of the Trip

As was stated earlier, the Skydio Enterprise 2 can recognize and track objects, including people! On our walk back to the resort, it flew back with us, dodging trees to keep us in its line of sight. Below are some clips that Charlotte Burger captured of the drone in action along with the full video of it tracking our hike.

Student's operate the Skydio 2 until it is placed in autonomous mode while it tracks the short hike.

Analysis Attribution

The entire class contributed to the collection and processing of the spatial data in this StoryMap. Special thanks to our professor, Kytt MacManus, and our TA, Juan Nathaniel.

Nighttime Lights

Madeline Liberman and Lauren Kahme

NDVI Vegetation Mapping

Owen Fitzgerald, Grace Tulinsky, Michael Higgins, and Alex Contreras

Terrain Mapping

Emme Fraser, Yumtso Bhum, and Ethan Rankin

3D Model

Camilla Green and Charlotte Burger

While we were staring up into the sky at the drone, oblivious, a family of deer approached our launch site!