UTCI Classification Model
Year: 2023
Course: OEB 137: Experimental Design and Statistics for Ecology
Instructor: Benton Taylor , Calvin Heslop (TF)
Course: Introduction to Generative Artificial Intelligence
Instructor: Sabrina Osmany
This project explores the potential of urban vegetation
in mitigating heat stress in urban
environments using a single-dimensional variable
for thermal comfort, the Universal Thermal Climate
Index (UTCI). This research acknowledges the
increasing concern about climate change and urban
heat island (UHI) effects and their physiological
impacts on humans. UTCI is a function of ambient air
temperature, mean radiant temperature, humidity,
and wind velocity—all factors that interact with
vegetation on a site. As such, the experiment aims to
provide a multi-dimensional understanding of the relationship between trees and their indirect effects on UTCI in a generic site.
Utilizing tools like Ladybug and Grasshopper plugins for Rhinoceros 3D, one thousand simulations were conducted to observe the impact of varying tree
counts on the mean UTCI in a virtual 50x50 meter
plot set in a downtown Manhattan climate. The study
demonstrated a nonlinear
negative correlation between increases in tree counts and decreases in UTCI. Temperatures outside of the comfort range approached the comfortable thresholds with the additions of trees.
A convolutional neural network model
was then successfully trained to classify these simulation images into bins of tree counts,
showing promising results for future applications of machine learning in urban microclimatic research.