The year in review

with No Comments

For the first two months of 2020 we planned and worked as though we would be working in Iceland during June; once we realized that the Covid-19 pandemic would prevent that, and much else, we re-organized ourselves and our work in the new normal. This included accommodating people working remotely from a wide range of timezones and reprioritizing tasks based on a new timetable. Whenever possible we tried to make lemonade, for instance we were able to re-organize our data storage structures, something we could not have done without the long timeline available this past year. Here is a summary of the other ground we have covered, future posts will have more details about each of these bits.

Craft research – One of the challenges of doing system integration in this space is the fast pace at which the hardware is evolving. We refined our requirements and resurveyed the market, ultimately deciding to add the Parrott Anafi and the Skydio 2 to our kit. In addition to craft-mounted multi-spectral lens we are also working with a MapIR NIR equipped lightweight camera which we will use as payload on the Anafi.

The Anafi with the MapIR camera payload. Image credit Charlie Peck.

ODM configurations – With time and computational resources we have been able to read and experiment with the ODM options that apply to our analysis workflow, and there are lots of them.

GIS integration – Our workflow now incorporates QGIS, this gives us the ability to use the powerful georeferencing plugin to accurately merge each of the data layers (sensor modes).

Storytelling – For many years we have been taking a fairly ad-hoc and often random approach to getting the word out about our work. We realized that we could now take the time to make a more organized run at storytelling, so a few of us have.

Planning for 2021 – We are now deep into the logistical and science planning for the 2021 field season, as I write this in mid-January we are working with the College and our colleagues in Iceland to plan our time there during June. One of our field sites, the Skalanes Nature Preserve, has experienced a number of mudslides this winter, destroying parts of Seydisfjordur and making travel along the fjord very difficult.

View to the North over the Eider colony at Skalanes. Image credit Charlie Peck.

And we’re off, again, hopefully…

with No Comments

Today is the official re-start of our National Geographic Society supported near-Earth survey work in Iceland (Earlham’s press release). Like many other scientists we have been unable to travel to field sites since the onset of the pandemic, the National Geographic graciously gave us a one year deferral for our grant, and now we are beginning to plan towards working in Iceland this coming June. Our project is based on commonly available UAV technology, our group is designing and building inexpensive, open, software and hardware systems for domain scientists to easily measure a variety of Earth surface parameters. Our first two disciplines are archaeology and sustainability, both of which depend on a variety of sensing modes. A few details about the gear and workflows are below, and in subsequent posts we will describe them in more depth.

Our group is a collection of students, faculty, and professionals primarily based at Earlham College in Richmond, Indiana and at Skalanes, outside of Seydisfjordur, Iceland. Together we cover archaeology, biology, computer science, geography, storytelling, and sustainability. The faculty and professionals are: Emmett Smith, Craig Earley, Rannveig Thorhallsdottir, Olafur Petersun, and Charlie Peck (me). The students we are currently working with are: Dung (Kate) Nguyen, Tamara Blagojevic, Davit Kvartskhava, Pyone Win, and Yujeong Lee. Over the next year you will learn more about all of us as we write posts that describe the specific aspects of the project which we focus on. For the most part we fancy ourselves as generalists, but in reality each of us brings lots of domain knowledge and focus to our work where it is blended into solutions.

Our goal for this cycle is to make it easier and cheaper for archaeologists to locate subterranean points of interest within a known or suspected cultural activity area, and for environmental scientists to quickly survey large areas for e.g. invasive species measurement or erosion. Our approach combines three relatively recent advances in drone, sensor, and machine learning technologies. 

1) Consumer grade drones capable of doing basic field science tasks, at accessible costs. 
2) Significant growth in the types of sensors available, and at low cost. 
3) Powerful, relatively easy to deploy open source machine learning libraries, which can extract deep patterns from large, noisy, multidimensional data sets. 

These three trends can support an approach to subterranean feature detection that is faster, cheaper, and more accessible to a wider range of practitioners than existing methods. Rather than depending on a single very sensitive, often expensive and complicated sensor to detect subterranean features, e.g. satellite or aircraft based LiDAR; our Terrestrial Mapping Platform (TMP) makes it possible to do ground based surveys, in a combination of sensor modes, and then use machine learning algorithms to combine those data sets into a single analysis to detect subterranean anthropogenic features and characterize surface vegetation.

Lunch during a day of soil sampling and aerial surveying at Sólheimajökull, an outlet glacier of the Mýrdalsjökull icecap on the southern coast of Iceland, June 2019. Image credit Porter Libby.