Laura Greenstreet

Buses in New York City are slow and unreliable. Through a Cornell Siegel PiTech fellowship, I created a simulation framework using MTA open data to understand what factors drive poor performance and compare interventions. We are now partnering with the MTA to test the simulation internally. While the project is ongoing, simulations and optimization have shown controlling headways at the start of routes can result in signifcant improvements, even compared to strategies using real-time data.

Learn more about the project in this blog post.

Fishery income is highly variable, where low income years can heavily impact small communities and result in millions of dollars of federal support. Diversification in multiple fisheries can reduce income variability, but fisherman may not diversify due to financial costs, difficultly obtaining new skills, or uncertainty about future risk. Working with fisheries ecologist Suresh Sethi, I developed an optimization framework to explore the relative impact of interventions targeted at each challenge. The optimization found there is potential to reduce income variance across the system by over 50% and significant benefits to reducing barriers to obtaining new skills.

Aquaculture, or the farming of aquatic organisms including fish and shellfish, has the potential to create economic growth with lower land-use, freshwater use, and carbon emissions than traditional livestock. Aquaculture is growing rapidly in the Amazon. However, the location, extent, and life-cycle of operations are poorly understood. Working with Dr. Carla Gomes, I helped develop deep learning methods to detect aquaculture ponds from medium resolution satellite data, using temporal information, attention, and contrastive learning to deal with issues including label imbalance, label bias, and generalization to new regions. The model was used by to understand land use change and associated carbon emissions in the Brazilian Amazon.