About the Forecasts

Show Me Fire produces unique, Missouri-specific fire weather forecasts. By blending high-resolution meteorological data with localized observation networks and machine learning, we provide the most accurate fire danger assessments in the state.

Our Forecasting Process

To generate our daily temperature, wind, and humidity plots, we utilize HRRR (High-Resolution Rapid Refresh) model runs. While the HRRR is a premier tool for short-term forecasting, we actively apply systematic corrections to account for known model biases, ensuring "cleaner" and more reliable data for our users.

The Importance of Fuel Moisture

Predicting fuel moisture is one of the most challenging—yet critical—aspects of fire danger. Our system approaches this through a multi-step integration:

Real-Time Ingestion: We pull morning observation data from RAWS stations across Missouri and neighboring states.
Baseline Analysis: This provides a real-time snapshot of current temperature, wind, and existing fuel moisture levels.
Machine Learning: These variables are fed into our proprietary machine learning model, which translates HRRR atmospheric forecasts into precise fuel moisture predictions.

Continuous Forecast Optimization

Forecast accuracy is refined through a continuous feedback loop. By archiving historical forecast data alongside real-time observations, the system systematically identifies discrepancies to improve the precision of the machine learning model.

During periods of lower fire danger each night, Show Me Fire aggregates the day’s forecast data and compares it against observations collected after peak hours. This automated comparison generates daily performance reports used to monitor model biases and identify specific instances where forecasts may have deviated from actual conditions. Furthermore, 7-day and 30-day rolling averages are tracked to ensure long-term consistency and reliability.

As the local dataset expands, the machine learning model ingests this information to identify and monitor new correlations between atmospheric variables and ground conditions. These insights are then utilized to update the forecast guidance pushed to production, resulting in an increasingly accurate predictive system for Missouri fire weather.

Key Terms & Technology

HRRR (High-Resolution Rapid Refresh): A real-time atmospheric model updated hourly by NOAA. It provides highly detailed weather data at a 3km resolution, making it ideal for predicting short-term weather events that impact fire behavior.
RAWS (Remote Automated Weather Stations): A network of weather stations specifically designed to monitor fire danger. Unlike standard airport stations, RAWS are often located in rural or forested areas and measure specific variables like fuel temperature and fuel moisture.
Fuel Moisture: The amount of water held within living or dead vegetation (fuels), expressed as a percentage. Lower percentages indicate drier fuels that ignite more easily and burn more intensely.
Machine Learning: A subset of data science that uses statistical algorithms to identify patterns in historical weather data. Our model analyzes how specific atmospheric conditions have historically affected Missouri’s vegetation to provide more accurate, data-driven fuel moisture predictions.