Urology Grand Rounds

For the full class schedule, descriptions, and to register, please visit the UConn Recreation website (https://recreation.uconn.edu/group-fitness-schedule/).

It's January and you might be redesigning your spring course. Are you curious about how to use tools like CoPilot or ChatGPT to support your teaching this semester? Some would argue that GenAI can free up course prep time that you can instead spend on enhancing feedback to students. Ideas and resources for using GenAI to edit assignment prompts for clarity, create case studies, generate problem sets, develop remedial modules, and make quizzes, etc., will be shared. In this interactive session, participants are encouraged to share examples, solicit feedback, or ask questions of other attendees. This session is suitable for participants with minimal experience using generative AI but will be most useful if you know how to access either Microsoft CoPilot through your UConn Microsoft 365 login, or ChatGPT (both the free "mini" and paid versions will be demonstrated). Register - https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3336 (https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3336)

Forest fires play a critical role in shaping ecosystems but also pose severe threats to biodiversity, carbon dynamics, and human safety. To address the current challenges, this research focuses on monitoring, characterizing, and predicting burn severity using remote sensing and a machine learning-based data-driven approach. We first demonstrated the feasibility of predicting burn severity using environmental variables through a case study in New South Wales, Australia. By integrating vegetation-specific thresholds, drought indices, and fire weather conditions into machine learning models, we achieved high prediction accuracy of annual burn severity. This foundational work highlighted the potential for scaling the predictive framework globally. Recognizing the lack of a high-resolution global forest burn severity dataset, we developed a 30-m resolution Global Forest Burn Severity (GFBS) dataset using Landsat imagery from 2003 to 2016. This dataset, validated against regional and global benchmarks, provides unprecedented detail in capturing burn severity patterns and ecological impacts. Expanding the dataset to include data through 2023, allowed us to analyze burn severity trends across global ecoregions using the Mann-Kendall test and Sen's slope estimator. Results revealed significant increases in burn severity in tropical and subtropical regions and decreases in boreal zones. For the ecoregions exhibiting significant trends, we developed predictive models using the XGBoost algorithm, incorporating 14 climate variables from the TerraClimate dataset. These models achieved high predictive performance and identified key drivers of burn severity including vegetation water stress and atmospheric dynamics. SHAP (Shapley Additive Explanations) analysis further revealed region-specific factor importance, enabling a detailed understanding of wildfire dynamics. The contributions of this research are threefold: (1) a robust framework for predicting burn severity at local to global scales; (2) the development of a comprehensive, high-resolution dataset of burn severity; and (3) insights into the drivers of wildfire severity, providing a basis for future scenario projections under climate change. By integrating trend analysis, predictive modeling, and climate-driven applications, this study offers critical tools for advancing fire management, enhancing resilience, and mitigating the impacts of wildfires on global carbon cycles and biodiversity.

For the full class schedule, descriptions, and to register, please visit the UConn Recreation website (https://recreation.uconn.edu/group-fitness-schedule/).