Applying ‘Geospatial First’ Approach to Estimating Heating Loads in Alaska

By: Emily Keller-O’Donnell

Decarbonization efforts to mitigate the impacts of climate change face a unique challenge in Alaska. While about 70 percent of energy use in the state goes to heating, there is a lack of reliable broad-scale data on the portion of energy that is used for heating in buildings. An interdisciplinary team at the University of Washington Data Science for Social Good (DSSG) program is working to create an improved method for estimating thermal energy use in buildings in the Alaska Railbelt and Arctic regions, to support these efforts.

The project, “Heating Loads in Alaska and Beyond,” is one of four projects hosted this year by the eScience Institute. The DSSG program brings together students, stakeholders, data scientists and domain researchers to work on project teams for a 10-week period. The project team consists of four student fellows from universities around the country, a data scientist at the eScience Institute, and the project lead, Erin Trochim, Research Assistant Professor at the Alaska Center for Energy and Power at the University of Alaska Fairbanks.

The project aims to create methods that will support decarbonization planning efforts, which are critical in the Arctic region, where warming is occurring twice as fast as the rest of the world. The project focuses on the Railbelt Transmission Grid, which stretches from Fairbanks through Anchorage to the Kenai Peninsula, and provides about 79 percent of the state’s electrical energy. There are also plans to expand the methods that are under development to the pan-Arctic region.

Data Science Approach

Estimating heating loads for buildings requires the combination of multiple data sets. The team is integrating data sources on building height, size, year built, elevation and construction type, as well as climate data. Challenges include a lack of consistent building codes in Alaska, and the variability of information according to building type and age. One of the primary data sources is the Alaska Retrofit Information System (ARIS), which provides metrics that were gathered before and after the retrofitting of home heating systems through a state-funded program. That data has only recently become widely available.

The team is taking a “geospatial first” approach to the problem by using satellite data in Google Earth Engine to generate variables of interest. Fellow Shamsi Soltani, a doctoral student in the Department of Epidemiology and Population Health at the Stanford University School of Medicine, said, “What’s most novel about this team’s approach is its ‘geospatial first’ lens. This means that to tackle a question like ‘when were the structures in this area built’ from satellite imagery, I have been challenged to think in entirely different ways than in my prior work. Whereas I would usually first look to national survey or administrative data, with this approach we are instead considering how to identify the year that vegetation disappeared in a building footprint, and other strategies leveraging geospatial imagery.”

Nicholas Bolten, a Data Science Postdoctoral Fellow at the eScience Institute, explained some of the data challenges that the team is working to overcome, noting that spatially biased data are “likely to be reflecting underlying socioeconomic biases.” Bolten explained, “For example, our project relies on using polygons that represent the outlines of building footprints, but those are much more available in urban centers than in the more rural areas. At the same time, the more rural areas use a different proportion of sources than urban ones, so it will be interesting to see how we can attempt to mitigate the impacts of these biases, or even account for them as elements of our models.” The team is approaching this issue by “creating spatial joins and overlays to compare these datasets to one another, showing where there was overlap – and where data was missing,” Bolten said.

Stakeholder Engagement

One of the primary aspects of the DSSG program is stakeholder engagement, which begins early in the project to help shape its design. The team held meetings with the Alaska Heating Usage Synthetic Data Group, which includes faculty and researchers from the National Renewable Energy Laboratory and the Alaska Center for Energy and Power at the University of Alaska Fairbanks. The meetings were aimed at gaining insight into topics such as how to counter modeling bias in underlying data sets, how to make the project deliverables usable to individuals and property owners, and options for estimating building heating loads.

Fellow Maddie Gaumer, a master’s student in the Department of Applied Mathematics at the University of Washington, said, “The stakeholder engagement on this project has helped me better understand not only the nuances of the project but also the different ways in which this project could have been approached. In addition, working closely with our stakeholders has enabled us to maximize the usefulness of the outcome of our project.”

Fellow Philippe Schicker, a master’s student in the Heinz College of Information Systems and Public Policy at Carnegie Mellon University, is excited to work on a project with policy implications that are interesting to stakeholders. “Often these researchers are experts in adjacent and overlapping fields and hearing about their work and being able to ask them for input has been immensely valuable. On multiple occasions, stakeholder impact has made us aware of potential challenges or made us think about our problems from a different perspective. It has been particularly helpful to ask stakeholders about environmental justice and Alaska equity aspects,” he said.

Interdisciplinary Learning

The fellows said they are learning a variety of skills including geospatial problem-solving, Google Earth Engine, machine learning prediction tools, and applying coding techniques to environmental issues. Fellow Vidisha Chowdhury, a master’s student at Heinz College of Information Systems and Public Policy at Carnegie Mellon University, said, “My experience in machine learning and data mining has helped me contribute to data exploration, visualization and project pipeline development. In addition, my background in economics will help me address equity issues underlying decarbonization through this project.”

Project lead Erin Trochim described the team’s varied educational backgrounds as an asset to the project. “The interdisciplinary make-up of both our team and stakeholders is a huge asset to addressing our problem of modeling heat in the changing Alaskan environment. Understanding how to go from a single building to entire boroughs to meet the planning needs of decarbonization is a challenge,” she said.

The DSSG final presentations will take place via Zoom and in person on Wednesday, August 17th from 1:00 to 3:00 p.m. PDT. The event is open to the public. Registration is required.