Research

My research focuses on understanding and addressing extreme heat vulnerability through physiological heat stress modeling, climate data analysis, and development of accessible decision-support tools. I work at the intersection of climate science, public health, and adaptation policy to protect vulnerable populations from rising heat extremes.

5
Physiological Models
700+
Indian Districts Covered
4
Work Intensity Levels
8
Countries Analyzed

Extended Heat Index (EHI-N*): Redefining Heat Stress Measurement

India Energy & Climate Center (IECC), UC Berkeley

Standard heat indices treat everyone the same—someone sitting indoors and a construction worker in direct sun get the same "feels like" temperature. The EHI-N* changes that. Built on the JPL/NASA physiological model, it accounts for metabolic heat production (100–350 W/m²), solar radiation, and the body's actual thermoregulatory limits—sweating capacity, vasodilation, and core temperature thresholds.

The framework classifies conditions into five risk zones from Safe (<32°C) through Extreme Danger (>42°C), where the body can no longer shed heat fast enough to prevent organ damage. Applied across 700+ Indian districts, it reveals that heavy outdoor labor pushes workers into danger zones at temperatures traditional indices still call "moderate."

Key finding
At 350 W/m² (heavy work), danger threshold drops by ~8°C compared to resting conditions
Method
Solves coupled heat-balance equations across physiological regions (core, skin, environment)
Impact
Real-time district-level heat alerts now operational for Regions 4/5/6 across India

Future Heat Stress Projections (ML Ensemble)

India Energy & Climate Center (IECC), UC Berkeley

Climate models tell us the planet is warming—but what does that mean for the person working a rice paddy in Bihar at 2 PM? This project bridges that gap using stacked machine learning ensembles trained on CMIP6 climate model outputs to project physiological heat stress decades into the future.

The approach targets the upper tail of the distribution—95th and 99.7th percentile events—because it's the extremes that kill. By combining multiple ML architectures (gradient boosting, neural nets, quantile regression) into a probabilistic ensemble, the model captures uncertainty while identifying which districts face the steepest increases in unworkable days per year.

Status
Active development — ensemble architecture validated, scaling to national coverage
Patent
HEAT RADAR — ML system for future heat projections (listing pending)

Indoor Thermal Resilience of Cooling Shelters

UC Berkeley

Not everyone can escape the heat outdoors. This study investigates how well designated cooling shelters actually perform during extreme heat events—measuring indoor thermal conditions, occupancy patterns, and the gap between intended and actual cooling capacity. The goal is to inform building standards and emergency preparedness for heat-vulnerable urban populations.

Under review — submitted for peer review