Active Research:

Climate-resilient machine learning

Climate change shifts natural distributions. This non-stationarity means that our measurements of t he natural world may not accurately communicate climate change to us -- can machine learning tools help us adapt to rapidly changing distributions? This work focuses on increasing the speed of representing spatial patterns in management-relevant models.

Data assimilation

High-quality global snow maps over complex terrain are essential for water resources, hazard management and climate model evaluation. Due to the difficulty of direct measurements over wide regions, these maps must be produced by combining observations of different types via statistically robust methods. Data assimilation is a way to overcome the limitations and uncertainties of current snow observations as it combines, statistically via Bayesian inference, the spatial and temporal strengths of both snow model estimates and observations. In this collaboration with the SNOWDEPTH project at the University of Oslo, we are developing joint assimilation frameworks and experimenting with vegetation and local radar model forcing.

Snow drought

We know that climate change affects hydrological systems, but the details of how different regions will be impacted is still an open question. Snow drought, or a lack of snow water storage in a particular area, is expected to increase in the future due to warming temperatures, but snow drought can also be caused by changing precipitation patterns. This project undertakes an exploration of how predictions of statistical ranges of climate and weather will shift our expected distributions of snow drougths in the future.

Post-fire snow hydrology

In the American West, we rely on snow for a large portion of our fresh water supply. The structure of forests in the places where snow falls has a large impact on what happens to snow -- does it sublimate to the atmosphere, get taken up by surrounding vegetation, infiltrate into the soil, or melt and run downstream to cities and agricultural regions in the spring? Large wildfires alter forest structure in the high mountains where snow falls and we do not fully understand how the types of forest change caused by fires will change what happens to snow in the future. We're using remote sensing, modeling, and field work to track and predict forest-snow interactions with a focus on regions where snow is a major water source for people. I work in two areas: 1) with field work in the Caldor Fire (South Lake Tahoe, California), how does fire change the spatial representativity of monitoring stations? and 2) increasing implementation of ML methods (e.g., LSTMs) to quantify fire's effect on climate-streamflow relationships in snowy basins in the US.

Snow surface observations

Previous Research:

How do wave-current flows impact cohesive sediment transport?

Coastal erosion in the San Francisco Bay threatens the security of shoreline infrastructure. This field campaign from 2018 to 2020 deployed hydrodynamics instrumentation in the South Bay to measure near-bed flow and sediment concentrations. My work focused on phase decomposition of the wave signal in near-bed velocity profiles during a summer deployment. In this project, we developed a method for identifying instantaneous wave phase from a combined wave-current flow using the Hilbert transform and then tracked the timing of at-bed stress and turbulence over the bottom 1.5 cm of the water. We found that the variations in stress and turbulence during the passage of a single wave are high enough that the instantaneous stress can be much higher than reported by a long-term average and that there is a phase lag between stress and turbulence.