This Expedition in Computational Sustainability proposed the new research area of Computational Sustainability with the overarching goal of identifying, formalizing, and providing solutions to computational problems for balancing environmental, economic, and societal needs for a sustainable future. A key aspect in achieving this goal is the vision that computing and information science can--and should--play a key role in increasing the efficiency and effectiveness of the way we manage and allocate our natural resources. An important goal of the project was to identify important cross-cutting computational problems in sustainability and help create a community of researchers to solve those problems.

Our team identified a wide range of problems in many fields. In ornithology, we worked with data collected by the eBird project (ebird.org) on bird watcher sightings of birds to develop spatio-temporal models of bird species distribution, which in turn formed the core of the State of the Birds reports for 2009, 2010, 2011, 2013, and 2014. An important issue with such data is that different bird watchers have different skill levels for detecting each bird species; we developed new machine learning algorithms to infer the skill level of each birder and used that information to adjust our species distribution models. We also studied algorithms for compensating for the biased spatial distribution of birder observations; this remains an important problem for future research.

Several sustainability problems involve the management of ecosystems. We developed new models for optimizing the recovery of an endangered species (the Red Cockaded Woodpecker), for managing the spread of invasive species (the Tamarisk), for managing the spread of wildfires in Eastern Oregon, and for designing wildlife corridors for grizzly bears and wolverines  in Montana and Wyoming. An important goal of these methods is to manage the ecosystem to minimize costs and maximize benefits.

Another set of problems involve the collection and interpretation of data. For weather station data, we developed algorithms for the automatic detection of failed or damaged sensors (to ensure data quality). For insect data, we developed computer vision algorithms for determining the species of moths and freshwater macroinvertebrates (insects that live in streams). These data are used to recognize agricultural pests and to assess the health of stream ecosystems. Working with birdsong data collected by our collaborators, we developed algorithms that learn to recognize individual bird species even when multiple birds (from multiple species) are singing simultaneously.

Progress on many sustainability problems (e.g., energy production, fuel cells, carbon capture) would be greatly advanced by the discovery of new materials. One way to create new materials is the “composition spread” approach in which different are sputtered onto a silicon wafer using guns pointed at distinct locations. Each point on the wafer corresponds to a unique combination of the elements, which may constitute a new material. The computational challenge is to analyze the x-ray diffraction data collected from the wafer to discover the regions that share a common crystal structure. To solve this problem, we developed novel algorithms for pattern discovery that incorporate knowledge in the form of constraints.

A final research area focused on sustainable development in Africa. We developed algorithms for mapping poverty and for trying to understand how poor regions can grow or shrink. We studied the migratory behavior of pastoral herders in Kenya as they move their herds from one grazing location to another. Creating such dynamical models is a first step to designing effective methods for designing new policies for poverty reduction and improved agricultural success.

Overall our Expeditions in Computational Sustainability identified several core research themes for maximal potential impact, both in terms of intellectual payoff and societal impact. 

In addition to identifying and solving computational problems, a central goal of the expedition was to create a vibrant research community focused on Computational Sustainability. Our expedition has been very successful in starting and fostering this new field of Computational Sustainability. To that end, we organized three international conferences on Computational Sustainability (Cornell in 2009, MIT in 2010, Copenhagen in 2012), organized four workshops at machine learning conferences, and established a Computational Sustainability track at the annual AAAI Conference on Artificial Intelligence. We trained several PhD students and postdocs who obtained tenure-track positions at several leading institutions including the University of Massachusetts (Amherst), University of Waterloo (Canada), Stanford University, Georgia Tech., University of Warwick (UK), Smith College, and Oregon State University.  Finally, we gave numerous tutorials, keynote speeches, and public lectures on Computational Sustainability. These materials are available through our web site http://computational-sustainability.cis.cornell.edu/.

Last Modified: 11/01/2016
Modified by: Carla Gomes