When Social Science meets Computer Science NASA’s Mission Directives are Explored in Innovative Ways |

Student Stories

When Social Science meets Computer Science NASA’s Mission Directives are Explored in Innovative Ways

Dylan Davis: STScI-H-v1833a, Eta Carinae

Written by: Dylan Davis, NASA PA Space Grant Graduate Research Fellow 2019-2020
Featured Image: STScI-H-v1833a, Eta Carinae

When you hear the word “NASA”, you probably don’t think “liberal arts” or “anthropology”. In fact, you probably don’t even think “social science”. Most likely you think “astrophysics” or “computer science” or something along those lines. And yet, social science has a lot in common with NASA.

Dylan Davis: Ground testing uncovered many artifacts that had not been recorded by previous investigation. Here, you see pieces of ceramic (some of which are decorated) and tools made out of marine shells (yellow arrows).

Ground testing uncovered many artifacts that had not been recorded by previous investigation. Here, you see pieces of ceramic (some of which are decorated) and tools made out of marine shells (yellow arrows).

NASA has a number of different missions, including the Earth Science Division directives. The Earth Science Division stresses the importance of technological advances, specifically those which can identify connections between Earth processes and ongoing natural and human-caused changes. Archaeology contributes to this overall directive in two ways: 1) by investigating long-term changes in human land-use and its effects on local ecologies; and 2) by tracing the effects that environmental conditions have on human societies.

My name is Dylan Davis, and I’m a Ph.D. student in the anthropology department at Penn State (University Park) and a 2019-2020 NASA PSGC Fellow. My research is focused on using geospatial technologies – satellites, geographic information systems (GIS), and aerial imagery and machine learning algorithms, a form of artificial intelligence (AI), to detect archaeological features. Remote sensing, a discipline focused on detecting objects using images and other non-invasive techniques, is central to my work. This probably seems more closely aligned with NASA, doesn’t it?

Dylan Davis: The results of the machine learning algorithm. The areas with “high probability” in this image contain many archaeological materials. The lower the probability, the less materials were recovered during ground surveys.

The results of the machine learning algorithm. The areas with “high probability” in this image contain many archaeological materials. The lower the probability, the less materials were recovered during ground surveys.

Machine learning is a branch of AI in which computers are “trained” to recognize specific patterns in datasets and then make decisions about those patterns without human assistance. When machine learning is coupled with satellite image analysis, there is a clear pathway to increasing archaeological knowledge. This work also directly contributes to NASA’s Earth Science Division.

My project focuses on Madagascar, the fourth largest island in the world. Madagascar has a rich – but understudied – archaeological record. The island is also at great risk from the effects of climate change, which threaten the coastal inhabitants of the island (the Vezo) and their history. Coastal archaeological sites are actively eroding and are often difficult to locate because of their subtle appearance on the landscape. However, computer science and remote sensing offer innovative solutions to this problem.

Last year, with support from Penn State’s Institute for Computational and Data Sciences, I developed a machine learning model for detecting archaeological sites in satellite images. The model begins by classifying important environmental resources in the study area of southwest Madagascar. These resources include: coral reefs, offshore islands, mangrove forests, and the Indian Ocean itself. These resources were identified through oral histories of the Vezo people and other anthropological work. The algorithm calculates distances from each of these resources to other parts of the landscape. Following landscape classification and distance calculations, the model generates a probability map showing the most likely locations of potential archaeological sites. This calculation considers the distance from any given spot on the landscape to the different environmental resources mentioned previously.

Over the summer of 2019, The Morombe Archaeology Project (a collaborative archaeological team comprised of archaeologists and local community members in southwest Madagascar) and I surveyed locations that were identified by this predictive algorithm in southwest Madagascar. The team confirmed several dozen new archaeological sites and recovered more than 1000 individual artifacts. The method also reidentified several hundred previously studied archaeological sites along the coastline.

This successful method enables researchers to record cultural history that is at risk of disappearing due to the effects of climate change. Sea levels are rising, and erosion is increasingly damaging archaeological materials around the world. This is especially problematic on Madagascar. The method I developed was able to scan approximately 50 square kilometers (20 square miles) per hour. To put that in perspective, in 1 square mile you can fit over 48,000 football fields. So, in 20 square miles, you are talking about more than 960,000 football fields of space. In an average ground survey, which is how most archaeology is conducted, to cover that much space would take weeks to months, if not longer; it certainly couldn’t be done in an hour!

A map of the study area. Click on each layer for more information about Madagascar and the study region.

This technology is not only useful for archaeology. With the ability to detect human presence from space, researchers can potentially identify similar patterns in other places which could give insight into life on other planets! If we look closer to home, the ability to quickly and methodically record human activity holds promise for important research into human-environmental relationships. Answering questions about how environmental conditions influence human decision-making in ancient contexts first requires a well-studied archaeological record. Using machine learning and satellite images, my work is helping to improve our understanding of past human presence on Madagascar. As such, it will allow future studies of important effects of climate change on human societies, including how such change affects migration.

NASA’s Earth Science Division works to use information to “improve lives and safeguard the future of people all over the world.” Climate change is perhaps one of the greatest threats facing humanity in the 21st century. This focus on the environment’s effects on human settlement and survival are therefore of great importance to the modern world, and helps to advance NASA’s mission. So, coming back to NASA and the social sciences, it really is a perfect fit.

Dylan Davis is a PhD student in the Anthropology Department at Penn State who investigates how machine learning and satellite imagery can reveal previously undiscovered archaeological sites. To learn more about our scholarship program, click here.