Gabriela Rodriguez
MIT Department: Earth, Planetary and Atmospheric Sciences
Faculty Mentor: Prof. Sai Ravela
Research Supervisors: Paul Nicknish
Undergraduate Institution: Williams College
Hometown: Houston, Texas
Website: LinkedIn
Biography
Gabriela Rodriguez is a rising junior Geosciences major at Williams College. While participating in a summer program before her junior year of high school, she first became interested in geosciences after attending a presentation on how rip currents have raised annual swimmer fatality rates worldwide. Having experienced natural hazards when Hurricanes Harvey and Maria respectively hit her Houston home and family in Puerto Rico, she assessed rip current characteristics and corresponding safety signage along her local coast the following year. At Williams, she has conducted research in geochemistry to quantify toxic gas occupancies sampled from local homes in the Berkshires. She hopes to contribute to work modeling the climate at MSRP, and someday link her findings with communities threatened by rapid changes. Outside the lab, she is a board member for the Williams Association for Women in Mathematics and dances with Ritmo, the student-run Afro-Latinx dance team.
Abstract
Understanding scaling behavior for boundary layer winds
Gabriela Rodriguez1, Sai Ravela2
1Department of Geosciences, Williams College
2Earth Signals and Systems Group, Massachusetts Institute of Technology
Sustainability-related decision-making, such as wind energy applications, requires
regional fine-scale details in climate forecasts (e.g., boundary layer winds). However, physics-
based climate models are computationally expensive to simulate at high resolutions and do not
capture the relevant details at low resolutions. In 2022, Saha & Ravela of our group (ESSG)
coupled coarse-scale climate model outputs with simplified physics and statistics in an
adversarial learning framework to obtain a scaling function that closely predicted fine-scale
precipitation with high fidelity. However, a systematic understanding of scaling behavior in both the coarse and fine directions–especially for boundary layer winds and turbulence–is lacking. To gain deeper insights into scaling behavior, we studied an idealized vorticity transport problem controlled by a field of variable roughness. With Machine Learning, we can discover scaling laws that characterize downscaling methods over a dense (continuous) range of roughness scales. A computational procedure for solving the Poisson equation and time-stepping a differential equation are needed to simulate the equations numerically. We investigated three methods for solving the Poisson equation and used a Runge-Kutta method for time-stepping. Results indicate that we can effectively simulate the transport equation at different roughness scales and measure the resulting vorticity spectra.
