Plant phenology
Community Phenology: I used a blend of correlation analysis, multivariate linear regression and A/B testing to identify the association between weather and plant phenology patterns in southeastern Madagascar. I discovered that community phenology patterns were linked to both rainfall and temperature, despite low temperature seasonality in the region. Consequently, resource disparity across the year is expected to increase under climate change as the warm-wet season gets wetter leading a period of higher flower and fruit productivity, while the cool-dry season gets drier leading to a period of lower flower and fruit productivity. |
Black-and-white ruffed lemurs (Varecia variegata)
Behavior: I designed, trained and tested recurrent neural networks (i.e., GRU, LSTM, encoder-decoder) to predict the sequential multi-state behavior of wild black-and-white ruffed lemurs. I found an ~80% improvement in the predictive performance of the RNNs based on F1-score compared to models traditionally used to predict animal behavior in ecological and anthropological studies. These lemurs’ prior behaviors were most predictive of their future behaviors and incorporating external features such as weather patterns had no significant effect on prediction outcomes. |
Circular economy
At the United Nations, I worked in a cross-functional team to organize and host the inaugural Great Lakes Circular Economy Forum to catalyze the North American transition to the circular economy. In contrast to the traditional “take, make, waste” linear economy, in a circular economy, products are designed so that their materials are reintegrated back into the economy at the end of their lifecycle instead of tossed. By building strategic partnerships and facilitating dialogue between 120 business, government and technology leaders at the Forum, we developed a vision, roadmap and future agenda for minimizing waster and maximizing resource use efficiencies through the circular economy in North America. |
Mammalian diversity
We used camera traps to quantify mammalian species diversity in Cerro de la Muerte and general linearized models to explain variation in diversity across sites. We uncovered that site-level differences in mammalian diversity was linked to land use patterns, with a greater diversity of mammals found in private lands compared to government lands. |
Salamander genomics
The California tiger salamander (Ambystoma californiense) is a species native to California and threatened by human development and competition from the nonnative barred tiger salamander (Ambystoma mavortium). We synthesized the genomes of >500 specimens collected across California over 30 years to track the movement and hybridization of these salamander species. |
Greater bamboo lemur
I spent 6 months in the rainforests of Madagascar working with indigenous guides to document the behavior, collect the genetic material and monitor the populations of the greater bamboo lemur (Prolemur simus). The greater bamboo lemur is critically endangered, found exclusively in Madagascar and weight only about 1-5 kilograms. Despite their small size, the amount of cyanide they regularly consume from their bamboo diet would be lethal to humans a hundred times their size. |
Solar farms
We assessed 36 competing solar farm development proposals according to their energy output and influence on Mojave Desert tortoise population connectivity using genetics and bioinformatics. We used spatial simulations and nonparametric statistics to test the combined impact of various solar farm configurations and recommended the optimal placement of solar farms to the Desert Renewable Energy Conservation Plan that balanced the California Energy Commission’s business goal and Department of Fish and Wildlife’s conservation goals. |
California reserves
The University of California Natural Reserve System (UCNRS) was founded in 1965 and is a collection of 39 protected areas that span 756,000 acres. We used machine learning (i.e., MaxENT) and counterfactual modeling to model the distribution of 546 imperiled plants and animals in California and quantify the biodiversity found within the UCNRS. Despite the ad hoc formation of the UCNRS, we revealed that these reserves encompassed landscapes that are inhabitable by over 400 imperiled species (>70%), which was more than expected had the reserves been randomly placed. Based on these findings, we recommended new reserve sites that maximized campus proximity and complemented the standing biodiversity coverage to meet the educational and sustainability goals of the UCNRS. |