2024 JIFRESSE Summer Internship Program (JSIP) Achievements
The 2024 JIFRESSE Summer Internship Program (JSIP), conducted from July 1 to September 6, 2024, provided six UCLA undergraduate students with the opportunity to engage in cutting-edge research under the mentorship of experts from the Jet Propulsion Laboratory (JPL) and UCLA. This program aims to strengthen collaborations between JPL and UCLA while fostering the next generation of scientists and engineers. The following summarizes the achievements of the six student interns and their respective projects.
Project 1: 'Hollywood (Hills) Nights'—Understanding the Geography and Dynamics that Shape Heatwave Patterns Across the Los Angeles Area
Intern: Ningweizhi (Jason) Ge
Mentors: Dr. Colin Raymond (JPL), Prof. Gang Chen (UCLA)
Overview:
This project focused on the increasing frequency and intensity of extreme heat events in the Los Angeles region, characterized by complex terrain and varying land-cover types. The study aimed to understand the spatial variability of heatwaves (HWs), particularly the differences between high coastal anomaly (HCA) and low coastal anomaly (LCA) events, and the phenomenon of elevation-dependent warming (EDW).
Achievements:
• Data Analysis: Utilized ERA5 reanalysis data and weather station data from 2004 to 2023 to analyze temperature anomalies and large-scale weather dynamics affecting HWs.
• Heatwave Classification: Classified HWs into HCA and LCA cases based on temperature gradients between coastal and inland areas.
• Elevation-Dependent Warming: Investigated the characteristics of EDW events and their correlation with HWs, finding that EDW is more prevalent during LCA events.
• Findings: Discovered that large-scale circulation patterns and lower tropospheric conditions significantly influence HW spatial variability and EDW occurrences.
Project 2: Deep Self-Supervised Global Disturbance Mapping with Sentinel-1 OPERA RTC Synthetic Aperture Radar
Intern: Harris Hardiman-Mostow
Mentors: Dr. Charles Marshak (JPL), Dr. Alexander Handwerger (JPL)
Overview:
The project aimed to develop a globally applicable, self-supervised machine learning model for mapping natural disaster damage using synthetic aperture radar (SAR) data. The focus was on creating a model that requires no labeled data for training, enhancing its utility across different regions and disaster types.
Achievements:
• Model Development: Created a self-supervised transformer model trained on global SAR data from the OPERA RTC Sentinel-1 dataset.
• Validation: Successfully validated the model on three distinct damage events—a landslide in Papua New Guinea, fires in Chile, and flooding in Bangladesh.
• Performance: Demonstrated that the model outperforms previous approaches, including recurrent neural networks and classical log-ratio methods, in accuracy and applicability.
• Global Applicability: Showcased the model's effectiveness across various terrains and disaster scenarios without the need for hand-labeled training data.
Project 3: Understanding the Effects of Sub-Pixel Heterogeneity on Atmospheric Sounding
Intern: Erin O'Neil
Mentors: Evan Fishbein (JPL), Prof. Yu Gu (UCLA), Tianhao Zhang (UCLA)
Overview:
This study addressed the challenges in cloud classification due to sub-pixel heterogeneity. By using K-Means clustering on MODIS satellite data, the project aimed to improve the accuracy of atmospheric sounding and cloud property analyses.
Achievements:
• Data Utilization: Employed MODIS data on cloud top height, optical thickness, phase, and effective radius to inform clustering.
• Clustering Analysis: Implemented K-Means clustering to classify heterogeneous cloud types, determining the optimal number of clusters using the Elbow Method.
• Correlation Length Evaluation: Analyzed spatial autocorrelation before and after clustering to assess the impact on predictive capabilities.
• Findings: Found that accounting for heterogeneity increases the correlation length, indicating improved predictive accuracy, especially in expansive cloud regions with minimal missing data.
Project 4: Non-Keplerian Starshade Station-Keeping for Hybrid Exoplanet Observations
Intern: Thomas Maxfield
Mentors: Prof. Artur Davoyan (UCLA), Dr. Dan Goebel (JPL)
Overview:
This project explored a hybrid ground-space observatory concept for exoplanet observation, combining ground-based telescopes with a space-based Starshade to enhance direct imaging capabilities of Earth-like planets.
Achievements:
• Station-Keeping Methodology: Developed an improved method for calculating stationkeeping costs for the Starshade's non-Keplerian orbit.
• Δv Savings: Demonstrated potential delta-v savings of up to 7 m/s per observation compared to previous analytical methods, leading to significant propellant reductions over multiple observations.
• Propulsion Trade Study: Conducted a comprehensive analysis comparing chemical, solar electric, arcjet, and solar thermal propulsion systems.
• Optimal Propulsion System: Concluded that a hybrid chemical/solar electric propulsion system is the most suitable choice for the Starshade mission, balancing efficiency and feasibility.
Project 5: Effective Cloud Seeding in California
Intern: Nathan Chen
Mentors: Dr. Jonathan Jiang (JPL), Dr. Longtao Wu (JPL), Dr. Yu Gu (UCLA)
Overview:
The project aimed to optimize cloud seeding operations in California by developing a machine learning model to predict precipitation patterns using aerosol and weather data.
Achievements:
• Model Development: Designed a Convolutional Neural Network (CNN) to forecast 12-hour precipitation patterns.
• Data Integration: Utilized data from NASA's MERRA-2 and IMERG datasets, along with NOAA's Global Forecast System, covering regions across California, Nevada, and Oregon.
• Operational Tool: Demonstrated the model's accuracy in detecting precipitation patterns, providing a valuable tool for real-time decision-making in cloud seeding operations.
• Future Work: Proposed enhancements to include real-time forecasting pipelines to further improve operational efficacy.
Project 6: A Machine Learning Approach to Analyze the Impact of Atmospheric Circulation and Land Surface Feedbacks on Drought in the US Great Plains
Intern: Annie Rosen
Mentors: Yizhou Zhuang (UCLA), Dr. Longtao Wu (JPL), Prof. Rong Fu (UCLA)
Overview:
This study investigated how large-scale atmospheric circulation patterns influence land-atmosphere interactions and soil-moisture-precipitation (SMP) feedbacks in the Southern Great Plains during the warm season.
Achievements:
• Methodology: Applied a Self-Organizing Map approach to identify prevalent atmospheric circulation patterns associated with SMP feedbacks.
• Key Findings: Discovered that specific geopotential height patterns—Western-Low/Eastern-High over dry soils and Dominant-Low over wet soils—are precursors to strong afternoon precipitation rates.
• Lower Tropospheric Humidity: Found that lower tropospheric humidity significantly enhances precipitation over dry soils by influencing convective buoyancy.
• Implications: Highlighted the importance of considering lower tropospheric humidity in SMP studies and its role in modulating drought conditions.
Conclusion
The 2024 JSIP has been highly successful in achieving its goals of fostering collaboration between JPL and UCLA and providing undergraduate students with invaluable research experience. The interns have made significant contributions to their respective fields, addressing critical scientific challenges ranging from climate phenomena and atmospheric science to advanced observational technologies and machine learning applications.
Their work not only advances scientific understanding but also has practical implications for environmental management, disaster response, and space exploration. We commend the students for their hard work and dedication and extend our gratitude to the mentors for their guidance and support throughout the program.