Codebase for 8 TCC: Temperature Correlated Conditions Paper over 64 Districts of Bangladesh using Fourier Ensembled ML
Domain:
climateRecord type:
This release contains the complete, deterministic Python codebase for the Temperature-Correlated Conditions (TCC) framework. The pipeline generates uncertainty-aware, 24-month district-level projections across 64 districts in Bangladesh for seven primary thermal and physiological heat-stress indicators.
Key Features & Contents:
-
Fourier & Feature Engineering: Scripts for constructing future-known seasonal features and Fourier components.
-
Machine Learning Pipeline: Gradient-boosting ensemble routines (XGBoost, LightGBM, CatBoost) trained on 2014–2024 meteorological data.
-
Trend & Index Reconstruction: Re-imposition of 1980–2024 observational trends to project primary variables and recompute thermal indices (WBGT, Humidex, Discomfort Index, CDD, etc.).
-
Validation & Analysis: Out-of-sample benchmarking scripts validated against independent 2025 observations.
Note: Raw meteorological inputs from the Visual Crossing Weather API are excluded from the repository due to licensing restrictions.