Adrian Albert / Alan Rhoades Contact: aalbert@lbl.gov
Emergent climate change impacts on mountain snowpack pose a serious threat to water resources, especially in the western U.S., with substantial implications to sectors such as agriculture, insurance, and the snow sports industry. Currently, it is very difficult and expensive to measure the amount of water (snow water equivalent, or SWE) across large mountain basins, where ground instrumentation is sparse if at all available, and air monitoring campaigns are expensive and infrequent. As such, current practice is to rely on a fragmented suite of physics-based models that don’t incorporate most observational data available and are computationally expensive to run.
The goal of this project is to build an AI-based emulator of a numeric (physics-based) hydroclimate model. For this, we will investigate the use of modern generative learning approaches such as Generative Adversarial Networks (GANs) that have proven successful in a wide variety of complex scientific and computer vision tasks. A starting point is the recent work on domain translation with conditional GANs. Such an emulator would allow fast inference, easy sensitivity/what- if analysis of simulation output variables with respect to input observational (remote-sensing) data, and scenario analysis via the generative capabilities of GANs.
Desired skills:
– Comfortable with Python programming (object-oriented programming, numpy, matplotlib, pandas etc.)
– Familiarity with modern machine learning frameworks such as PyTorch or TensorFlow
– Knowledge and familiarity with modern machine learning, including deep learning. At a minimum, class work and projects, but previous research work using/developing deep learning methods preferred (ideally having worked with GANs)
-Interest in scientific applications of AI including one or more of climate, energy, materials