What is it about?
We used 40 years of weather data and 128 models to forecast heat discomfort in Rajshahi, Bangladesh. The seasonally-adjusted ML model, STL-TBATS-LSTM, was more accurate in predicting 40% 'High Discomfort' days by 2025-2027. This helps cities issue heat warnings and protect vulnerable people.
Featured Image
Photo by Anik Roy on Unsplash
Why is it important?
This work is unique because it is the first to apply a seasonally-adjusted, machine-learning hybrid model specifically to forecast thermal discomfort in a heat-vulnerable region of Bangladesh. Unlike simpler models, our STL-TBATS-LSTM approach captures both long-term warming trends and complex seasonal humidity patterns. The results are timely: climate change is accelerating, and Rajshahi already experiences extreme pre-monsoon and summer heat. Our projections show a clear increase in high-discomfort days over the next three years, giving public health officials and urban planners actionable lead time. This matters because heat stress is not just uncomfortable — it kills, reduces labor productivity, and strains hospitals. By providing more accurate, locally relevant forecasts, this work can directly support early warning systems, adaptive work policies, and climate-resilient infrastructure in low-income, tropical countries.
Perspectives
As researchers based in Bangladesh, we live with the reality of extreme heat. Rajshahi is known as one of the country's hottest cities, and every year we see more people struggling with heat exhaustion, missed workdays, and sleepless nights. What motivated us was the gap between knowing "it's getting hotter" and being able to say when and how much. Building this model felt personal — we wanted something that local authorities could actually use, not just another academic paper. The hardest part was testing 128 model combinations, but seeing the STL-TBATS-LSTM model outperform everything else was worth it. We hope this work moves beyond forecasting and into action: heat warnings, shaded bus stops, cooler school classrooms, and safer construction sites. Science should help people breathe a little easier — literally.
Dr. Rumana Rois
Jahangirnagar University
Read the Original
This page is a summary of: Application of seasonal-adjusted hybrid models for forecasting Discomfort Index in a heat-prone region of Bangladesh, PLOS One, March 2026, PLOS,
DOI: 10.1371/journal.pone.0344556.
You can read the full text:
Contributors
The following have contributed to this page
