Profile
My research: The focus of my project is the combination of Machine Learning (ML) algorithms with numerical simulation models, what we call hybrid modelling, to predict and understand geoscientific processes.
Current work: Recently I have been working on improving predictions of Snow Water Equivalent - in essence, the amount of snow - in the Northern Hemisphere based on meteorological data, crucial for managing the water resulting from snowmelt in downstream communities and ecosystems. We have seen that training not only with the few observations available but also with numerical simulations can improve the performance of a ML model for extrapolation to stations without historical observations. Beyond that, I have also previously applied ML algorithms to understand the dynamics of soil and vegetation in semiarid hillslope ecosystems, which could help prevent their desertification. Here we used ML to simplify the outputs of spatially-detailed simulation models, reducing the bias introduced by modelers when creating analytical alternatives.
Future plans: The continuation of my work will involve upscaling the use case on Snow Water Equivalent and applying hybrid models to other case studies, e.g. wind speed modelling or soil subsidence parameterization.