Drone ecology

A scientific consultancy for UAS

Why I left a tenure-track faculty job for the drone industry. Part 1

Part 107 – Commercial drone rules are out.

Some notes on mapping with the Sequoia camera on 3DR Solo

To begin mapping with Sequoia, you are first going to need power (5v, 2.4 amp). MicaSense has an adapter cable that will power Sequoia out from the Solo gimbal bay (as well as Phantom 3). It can be found here.  It has been working well for me.

Micasense also sells a downward facing mount, which I recommend picking up. I have found that the fixed GoPro  mount that comes with Solo adds some vibration to the RGB imagery and you have to stick the camera halfway out in order to plug in the Sequoia cables.  It’s not as clean as the downward facing MicaSense mount and I found jello in my RGB imagery.


From there, I have just been mounting the external GPS to the center of Solo using velcro for easy swapping of batteries. I use a cheap velcro cinch strap that wraps around the vehicle for cable management. Next, I use Tower to plan the mission on my Android tablet mounted on the standard Solo controller mount.  I design a standard Survey in Tower in the Editor.  As they have not updated Tower with the Sequoia settings yet (still in the works), I select the Canon S110 at 80% front and side lap.  It’s probably  overkill but the stitches have been great, so  I stopped experimenting with other cameras.

Sequoia works best at higher altitude flights.  So I tend to fly between 40 and 100 m in my surveys, depending on the resolution I am looking for. What still remains to be seen is how fast you need to set the waypoints navigation (you edit these when logged into a powered-up Solo under Parameters and scroll down to W in the alphabetical listing) speed in Tower.

I fly at 500cm per sec at 130 meters and it seems to work well. At 30 m, I fly fairly slow at 100 -150 cm per sec.  The primary concern with speed is the RGB, as it is on a rolling shutter, whereas the 4 bands are on a global shutter.

Next, you need to set Sequoia parameters.  You need to power up Solo so Sequoia is also powered on. Instead of switching back and forth on the between Tower and Sequoia on the Android tablet wi-fi, I use my phone to log into the wifi on the Sequoia camera, to start and stop triggering, and set the distance I want the photos trigger at for 80% overlap at a given altitude.

I mainly use the GPS to trigger the Sequoia camera by distance. This distance is set in the camera app (there is an easy calculator), but you can also use intervalometer and single capture. Triggering by GPS though optimizes the number of photos taken (and there are a LOT with Sequoia).

When you have your mission designed and your camera parameters are set, then hit “Start Capture” on your phone for Sequoia, launch your mission on Tower using the Android tablet, and you are in the multispectral mapping business.

When the copter lands, log back into Sequoia wi-fi with your phone and hit ‘Stop Capture’.  I tend to check the thumbnails in the Sequoia Gallery at that point to make sure my photos are there.  If it is a really important dataset, I will download the photos to my laptop before I fly again just to be sure I have a backup in my data (e.g. if there is some malfunction in the drone at a later point).

In terms of post processing, Pix4D is necessary for stitching at the moment. Parrot worked closely with Pix4D on the Sequoia project. Pix4D can automatically handle the photos from the four bands, as well as a calibration panel and Ground Control Points if you want to use those.  I have not used Photoscan or Agisoft, but I suspect the stitching won’t be great.

Pix4D will only stitch the RGB or the 4 bands within a single process, so you have to pick either/or for a given job. As I stated, for the 16 megapixel  RGB camera, there is a rolling shutter.  Pix4D is set up to deal with correcting the impacts of the rolling shutter.

I was impressed with how easy it was though to map with Solo and have that distance based triggering in Sequoia.  Here is an example dataset from a forestry area in northern Montana (clearcut next to intact second-growth).

Shoot me an email if you have further questions.

So you want to build a drone program?

Some answers to common questions I get in the drone industry from educators:

Drone technology changes so quickly, shouldn’t I wait?

No.  You should absolutely not wait.

If your goal is to prepare your students for an emerging industry, then know you are not alone. More schools are adopting drone technology every day, some with substantial funding from grants and forward-thinking administrators.

The longer you wait for the dust to settle in drone technology, which all indications suggest won’t happen soon, the further behind you will be. My advice is to start with a basic platform like Parrot minidrones or the Bebop 2 and begin to understand what the technology is capable of today.

Then, plan to pivot. You know the technology is changing quickly, so create a plan to adapt over time as new vehicles come out with new sensors and software.

How do I keep up with such a rapidly moving industry?

You don’t.  You let us do it.

The best thing you can do is partner with those of us firmly embedded in the space. We will let you know how to stay a step ahead. Let us follow the industry for you and keep you posted while you focus on instruction and innovation in the classroom or field.

Why go with a single drone company for teaching?

I  hear from folks who are excited to introduce students to all the different drones on the market. Certainly you can do this, but it’s expensive and not scalable for an entire program. For an introductory computer class, instructors typically don’t buy every Mac, PC, Linux, Rasperry Pi, desktop, laptop, mobile, etc…to teach with. Similarly, photography teachers often recommend one camera as a standard learning platform.

Drones are both flying computers and cameras. Keep in mind, they also need accessories (controllers, batteries, propellers, cases, charging cables). They have evolving software that needs to be updated occasionally. If your budget goes to all the different drones, with different interfaces, accessories, and capabilities, then you will spend a lot of course time managing your hardware and software and not on drone applications. I recommend you ensure you have enough vehicles of a single, affordable platform, like Bebop 2, before you start purchasing a range of other drones for comparison learning. Start simple and build from there.

Mapping ancient shrubs in the Mojave

The Mojave desert is a tough place to live by any measure. Receiving less than 13.7 cm (6 in.) of rain per year, it is the driest of all North American deserts. Spanning a range of elevations, the Mojave is prone to extreme temperatures, from sub-freezing conditions at night and in winter up to 49°C (120 °F) or higher in the summer. Without any natural cover, a relentless sun beats down on plants and animals alike.

IMG_1103One species thrives in the Mojave, however: the creosote bush (Larrea tridentata). Creosote is an evergreen shrub that grows 1 to 3 m (3-10 ft) in height and boasts bright yellow flowers. It is one of the most dominant species in the Mojave
landscape, and it provides most of the scenery on the drive from Los Angeles to Las Vegas.

The dark green creosote leaves are full of resins (plant oils), making the plant inedible to all but very specialized insects and contributing to the shrub’s common nickname, ‘greasewood.’ Creosote releases these oils into the air after a precipitation event, and they permeate the desert with a distinct smell that long-time desert dwellers associate with the smell of rain.

A fascinating aspect of older plants is the propensity to form rings. Creosote shrubs can live a long time, and as they age, their oldest, most central branches die off and the crown splits apart into a ring pattern. These rings appear to march, spreading out over the years; however, each new plant is actually the same genetic individual — meaning they’re clones.

The oldest known creosote clones are in the Lucerne Valley, California. Referred to locally as the ‘King Clone,’ this ring has been estimated by scientists to be over 11,000 years old and spans over 40 ft (15 m) in average diameter.

King cloneFigure 1.  The ‘King Clone’ of creosote (Larrea tridentata) in Lucerne Valley, California, is the large ring near the center of the map.  Note the black car on the road shown for scale.

Recently, I headed to the Mojave with the Parrot Sequoia camera, one of the smallest, lightest multispectral sensors on the market. In a single flight, Sequoia captures images across four defined visible and non-visible spectral bands, plus RGB imagery.  I used Sequoia to help address ecological questions about these ancient desert shrubs. Specifically, I was interested in whether ancient rings vary in productivity compared to their younger counterparts? If estimates are correct, the King Clone germinated when wooly mammoths still roamed the earth. Now that’s impressive, for sure, but I also wonder what that kind of aging might mean for the plant itself.

I mounted Sequoia on a 3DR Solo drone and using the free, open source Tower app to plan a fully autonomous mapping mission, I was able to map the King Clone and surrounding shrubs in a 10-minute mission at 30 m altitude. Though it was a single, short flight, the drone-mounted Sequoia captured data for a variety of ring sizes. I then linked up with the Pix4D team in Lausanne to collaborate on data analysis.

Pix4D stitched the Sequoia imagery into a high resolution orthomosaic in both color (RGB) and multispectral data layers. Focusing on a sub-sample of 60 shrubs that ranged from 1.2 to 23.5 m in diameter,we then used the Sequoia’s multispectral data to calculate the average normalized difference vegetation index (NDVI) across the different ring sizes.


Figure 2.  NDVI map showing multispectral data from the Parrot Sequoia camera of the clonal rings of creosote (including the King Clone) in Lucerne Valley, California.

The conclusion? There was no relationship between ring size and NDVI (Figure 1), without or without the King Clone included in the dataset as a statistical outlier (r  = 0.005, P = 0.96).

GraphFigure 3.  Relationship between the diameter (measured at the widest point in meters) of creosote shrub rings and the average (shrub level) NDVI.

Its a a cool result, as it suggests that the physiology of ancient shrubs, or how they function in this extreme environment, may be similar to shrubs several millennia younger. Mark Twain said that age is an issue of mind over matter: If you don’t mind, it doesn’t matter. That certainly seems to be the case with creosote in the Mojave.

Drones have already transformed the way ecosystems are studied and monitored. It’s easier and faster than ever to get a large amount of rich, accurate and actionable data. And with the Sequoia camera in particular, scientists now have an incredibly powerful tool to complement traditional vegetation data collection on the ground.

Journal article on advice for ecologists on the drone industry.

The future of UAVs in ecology: an insider perspective from the Silicon Valley drone industry.

The goal of this article is to provide an inside look at the leading edge of the drone industry from the perspective of an ecologist (and colleagues) working for a Silicon Valley drone company. We intend to do the following: (i) give ecologists a working understanding of existing and emerging technology (remaining as brand-neutral as possible); (ii) advise on a starting place for integrating drones into field research; (iii) advocate for immediate adoption and development of drone technologies specific to ecology; and (iv) encourage “thinking big” about the potential for drones to revolutionize the study of ecosystems.

Up ↑