Saving the Planet (With an Algorithm)

by Mary Martialay on July 9, 2013

Eric Shapiro, Amanda Knight, and Diogo Moitinho de Almeida

Eric Shapiro, Amanda Knight, and Diogo Moitinho de Almeida winners of one of five “Outstanding” prizes in the Interdisciplinary Contest in Modeling

Here’s a knotty problem: There is a relationship between human activity, damage to the environment, and harm to people (greenhouse gases leads to sea level rise which threatens coastal populations), but the cause-and-effect isn’t always clear. And while environmentalists collect stats on damage to the environment (pollutants in air and water, deforestation, species loss), policy makers who can shape human activity think in terms of harm to people (public health crisis, unemployment, threats to infrastructure). So how can we close the loop between human actions, environmental changes, and effects on humanity to save the planet and ourselves?

Three Rensselaer School of Science students have a solution: Skip the middleman (the environment) and draw a straight line between human actions and harm to humans. Their work on a new model called the “Earth Damage Score,” earned Diogo Moitinho de Almeida, Amanda Knight, and Eric Shapiro one of five “Outstanding” prizes in the Interdisciplinary Contest in Modeling. The trio took top prize in a field of 957 entries, and were the only American team among the award-winning top 12 entrants in the contest.

The international contest for high school students and college undergraduates—an extension of the Mathematical Contest in Modeling—draws on various disciplines to explore environmental issues. The 2013 entrants were asked to develop “a dynamic network-based model linking various factors governing the health of the earth, for the purpose of predicting how human activity and government interventions might affect the future quality of the environment and human prosperity.”

In response, the Rensselaer team created the “Earth Damage Score,” which uses an algorithm to estimate economic loss from damage to the environment in a given region based on easily available data on human activity. The algorithm balances a basket of factors – such as population growth, agricultural growth, literacy, GDP, industrial growth – against the influence of geographic proximity and political dialogue between countries to calculate the final score in each country.

Amanda Knight, a junior earning a dual degree in computer science and math, said many environmental models do try to track the relationship in a step by step approach, but they best serve the purposes of environmentalists, not policy makers:

They don’t make it easy for a policy maker to say ‘okay, this is what we can do.’ This paper was about giving lawmakers good models. And for policy makers, it’s more important to look at human actions and human results.

Diogo Moitinho de Almeida, a senior majoring in computer science and math, said the competition allowed them to take a creative approach:

For this competition, it’s four days to solve a real world problem that many people take years to solve. We figured we should explore creative options and see how it works out. We weren’t trying to propose it as the be-all/end-all solution, we said this is another tool that should be in the environmentalists’ tool box.

This is only the second year Rensselaer has competed in Interdisciplinary Contest in Modeling (the Institute has a far longer history competing in the Mathematical Contest in Modeling). Peter Kramer, an associate professor of mathematical sciences and the professor who coordinated the participation of Rensselaer teams, said the Earth Damage Score was exceptionally well executed:

This contest is really pushing students to go fluidly between mathematical thinking, disciplinary thinking, and creative ways to bring that all together and this team did that beautifully. If you look at the paper that this outstanding team put together, it’s really modeled in the style of writing that you would see in premiere journals like Nature and Science. They follow that spirit in that they take a complex question—‘how do I determine the health of the Earth?’—and they find a way to make it precise, they find a way to quantify it, they use available data, and they provide a result that allows us to make statements about the changing state of the Earth. Rather than bombard the readers with math, they present their analysis in a way that a policy maker could begin to understand and make use of.

Eric Shapiro, a sophomore in physics and math, who took the lead on the design and formatting, said the look and feel of the paper was very deliberate:

No one likes sitting down and reading a bunch of equations, so we agreed that we would try to make our paper tell its story through visuals. I had a lot of fun experimenting with different software to create meaningful, aesthetic images that basically explained themselves. I personally wish more papers (and textbooks) were written this way.

To calculate the Earth Damage Score, the students began with a basket of human variables predictive of environmental damage. For example, an increase in population means an increase in needs for shelter, transport, and production of waste; as a measure of economic output, GDP growth can be used to predict the production rate of pollutants; modern agricultural practices mean that agricultural growth can be tied to increases in deforestation, use of pesticides, and water pollution from runoff.

They then used a machine learning algorithm (called Tikhonov Regularization) to determine the relationships between their basket of variables and influences, and the economic loss by country, measured in 2013 U.S. dollars, induced by natural disasters and carbon dioxide damage using data obtained from the World Bank’s Databank and Maplecroft.

The relationship was further refined by adjusting for influences among neighboring countries using data on geographic proximity and political influences between countries (as measured by membership in common political and economic intergovernmental organizations).

The students, who were friends through their membership in the service fraternity, APO, said their individual strengths each played a key role in the final paper. Diogo took the lead in programming. Shapiro investigated real-world models, offering refinements that made their model more robust, and handled the graphic layout of the paper. And Knight crafted “big picture” explanation of their approach for the paper.

Rensselaer students also won awards in this year’s Mathematical Contest in Modeling. Sunli Tang, Vera Axelrod, and Steve Lentine were recognized for their paper on the optimal design of a brownie pan for even heat distribution and loading in the oven. And Jared Salvatore, Tripp Spivey, and Tor E Hagemann were recognized for their paper on designing a mathematical model for water resource strategy for a nation.