Table 1.
Methodology for Computing Heat Index (HI): Two-stage Calculation Using Steadman’s Formula and Rothfusz Regression Model with Empirical Adjustments [28].
Table 2.
Classification of HI values and associated health risk levels [29].
Fig 1.
Time series of 3-hourly Heat Index (HI) in Dhaka from 2014 to 2023, decomposed using STL into original values (orange), seasonal-adjusted series (red), and long-term trend (green).
Fig 2.
Heatmap of diurnal variation in average HI by year (2014–2023).
Colors indicate HI magnitude (°C).
Fig 3.
Monthly distribution of 3-hourly HI values (2014–2023) shown as boxplots, indicating median, interquartile range, and extremes.
Fig 4.
Hourly–monthly heatmap of mean HI values (2014–2023).
Each cell represents the average HI for a given hour and month.
Fig 5.
Ridgeline density plots of 3-hourly HI distributions by hour for each year (2014–2023).
Fig 6.
Seasonal distribution of HI risk categories (Normal, Caution, Extreme Caution, Danger, Extreme Danger) for 2014–2023.
Fig 7.
Monthly mean HI values plotted against HI risk categories (2014–2023).
Table 3.
Test-period performance comparison across five modeling approaches for 3-hourly HI estimation based on multiple accuracy metrics.
Fig 8.
Observed 3-hourly Heat Index (HI) during the test period (2022–2023) and scenario-based projections for 2024–2027 using the Random Forest Regressor.
Colored lines represent mean projections under optimistic, moderate, and pessimistic assumptions. Shaded regions denote 95% empirical prediction intervals derived from ensemble tree variability. The vertical dashed line indicates the start of the projection period.
Fig 9.
Scenario-based projected distribution of 3-hourly HI risk categories in Dhaka for 2024–2027, highlighting the predominance of sustained “Extreme Caution” conditions relevant for public health preparedness.