With the increasing frequency, intensity, and duration of extreme weather events worldwide, traditional simulation inputs no longer capture the conditions most critical to building resilience. TMY (Typical Meteorological Year) weather files represent averaged conditions and hence do not represent the full spectrum of extremes, which are critical to assessing thermal resilience, peak loads, overheating, and grid reliability under climate stress.

This project adresses that gap by embedding realistic extreme events—both historical and future—into weather files that are fully compatible with building simulation tools like EnergyPlus, ClimateStudio, Rhino/Grasshopper, and more. All data, code, and workflows are open-source, enabling reproducibility and adaptation for global use.

• RMYs (Representative Meteorological Years): Weather files embedded with observed extreme events (heatwaves and cold spells), generated from historical AMYs (Actual Meteorological Years) data using anomaly detection.
• FRMYs (Future Representative Meteorological Years): Morphed weather files embedding future heatwaves and cold spells derived from climate emulator trajectories with annual resolution and embedded uncertainty.
• Open-Source Anomaly Detection: A modular, fully open-source detection framework that identifies climate extremes using an ensemble of methods—static thresholds, Graph Neural Networks (GNNs), and Extreme Value Theory (EVT). The codebase is public and reusable across projects aiming to detect and embed extremes in weather or environmental time-series data.
• Event Integration in Accessible Weather files: Temporal smoothing and seasonal averaging logic that inserts extremes while preserving monthly averages in weather files compatible with a wide range of simulation programs.

This timeline shows detected heatwaves and cold spells across years based on the detection algorithm in a sample city:

This toolkit introduces a multi-method event detection and integration pipeline for constructing Representative Meteorological Year (RMY) and Future RMY (FRMY) weather files. These new formats restore historically observed or projected extremes into standard TMY files to enable realistic simulation for overheating, thermal resilience, and peak demand.

An Ensemble anomaly-detection method is used based on the following:

• Static Thresholding: Identifies extremes based on fixed temperature or percentile thresholds.
• GNN-Based Anomaly Detection: Flags events using graph-based representations of temporal temperature anomalies.
• Extreme Value Theory (EVT): Extracts statistically rare extremes using Peaks Over Threshold (POT) modeling.

Each method is used in a complementary ensemble to identify the most severe year and characteristic events.

For FRMY generation, future extremes are derived from annually-morphed climate emulator outputs that reflect global warming trajectories under different scenarios. These files embed projected heatwaves and cold spells with realistic variability, and are processed using the same anomaly detection and integration methods as RMYs.

1. Detect peak heatwaves and cold spells across 15+ years of EPW files.
2. Match extreme events to base-year dates using overlap logic.
3. Replace those dates with extreme-event days from the most severe year, using smoothing.
4. Rebalance monthly averages by inserting non-extreme days to maintain realism.
5. Output:
   • RMY file with embedded extremes
   • Summary CSVs for heatwaves and cold spells

data/
├── base/               ← base TMY EPW file (1 file only)
├── epws/               ← AMY files for detection
├── RMY/                ← output folder for RMY EPWs
├── FRMY/               ← output folder for FRMY EPWs
├── final/              ← stores output RMY + stats
├── images/             ← visualization assets (map.gif, EEE)

Click below to explore an interactive dashboard where you can download RMY and FRMY weather files for cities worldwide:

👉 Explore the Map here

Install required packages:

pip install -r requirements.txt

Then run the following from the repo root:

from rmy import run_full_rmy_pipeline

Make sure your folder structure matches:

EPWs/base/ → contains the base TMY file (1 file only)
EPWs/epws/ → contains full set of AMY EPWs
final/     → RMY weather file + event summaries will be saved here

As climate conditions continue to shift, extreme events are evolving—not just in frequency, but in duration, intensity, and timing. Future extensions of this project will:

• Track how heatwaves and cold spells are shifting across decades
• Link these shifts to building performance risk and urban equity impacts
• Expand the tool to include dual temperature hazard assessments across cities

Below is a preview of how extreme events are changing over time:

Stay tuned for the next phase: Extremes in Motion — a dynamic, multi-city comparison tool.

Tarkhan, N., Crawley, D., Lawrie, L., & Reinhart, C.
Generation of representative meteorological years through anomaly-based detection of extreme events.
Journal of Building Performance Simulation, 2025.
https://doi.org/10.1080/19401493.2025.2499687

Giani, P., Fiore, A.M., Flierl, G., et al.␣␣
Origin and Limits of Invariant Warming Patterns in Climate Models.
arXiv preprint, 2024.
https://arxiv.org/abs/2411.14183

Tarkhan, N., & Reinhart, C.
Representing Climate Extremes: An Event-driven Approach to Urban Building Performance Assessments.
Comfort at the Extremes Conference, Seville, 2024.
View PDF

This repository is released under the MIT License.