Challenge 34 - FireFlux: FRP-based emission calibration from small satellite constellations for emission estimation

[DiegoDomenici]

Challenge 34 - FireFlux: FRP-based emission calibration from small satellite constellations for emission estimation

Stream 3 - Software Development for Earth Sciences Applications

Goal

The overall goal of this challenge is to develop a methodology for estimating biomass burning emissions for given ranges of fire radiative power observations which will be available from small satellite constellations in the coming years.

Mentors

Mark Parrington, Enza Di Tomaso, Miha Razinger (ECMWF)
Dmitry Rashkovetsky, Julia Gottfriedsen, Max Bereczky, Marc Seifert (Orotech)

Skills Required

• Statistical analysis using statistical models and/or machine learning
• Python programming for climate and remote sensing data analysis (xarray, numpy, pandas, seaborn)
• Software engineering
• Remote sensing
• Data Visualization

---

Description

The Copernicus Atmosphere Monitoring Service (CAMS) provides near-real-time estimates of emissions from global biomass burning and wildfires based on observations of Fire Radiative Power (FRP) from the NASA MODIS, and soon NOAA VIIRS, satellite sensors.

An increasing number of small satellite sensors are being deployed to detect and measure wildfires and the OroraTech constellation will provide an FRP product from 2026. The initial focus is on binned FRP products (e.g., low, medium, high intensity), tailored to sensor specifications to provide actionable data soon, while absolute FRP estimates stay the long-term goal.

The aim of this challenge is to develop a robust model to estimate FRP for thermal anomalies detected in OroraTech’s FOREST-2 (but may also extend to Copernicus Sentinel-3 SLSTR and NSMC’s FengYun-3E) Level-1 data and showcase their potential for emission estimation applications. The result is an FRP product that can be derived at two levels of maturity:

1. Binned FRP - At this level of maturity, FRP is reported in categories (bins) without the need for estimating physical quantities (e.g. low, medium, high)
2. Absolute FRP - At this level of maturity, FRP is reported in physical units with uncertainty being optionally reported
   In addition to the FOREST-2, Sentinel-3 and FengYun-3E examples that will be provided by OroraTech, CAMS data will be utilized for potential data augmentation, training, reference, and evaluation of the derived models.

By improving FRP retrieval from small satellite constellations, the challenge directly supports more precise wildfire emissions monitoring, enhancing air quality forecasting and climate impact assessments.

The challenge will be divided into the following stages:

1. Data Collection

• Provision of 10-20 L1 products
• Provision of thermal anomalies (hotspots) detected in L1 products
• Correlation with Emissions:
  • Archived FRP and biomass burning emissions from GFAS
  • Atmospheric constituents related to smoke (e.g., aerosol optical depth, total column carbon monoxide, and surface particulate matter concentrations) from the CAMS global reanalysis.

2. Model development

• Development of physical and/or ML models for FRP estimation based on the provided exemplary data for FOREST-2 and FY3E

3. Data Evaluation

• Statistical analysis of the results compared to CAMS data to identify and categorize FRP in relation to technical specifications for the selected satellite sensors.
• Analysis of how FRP thresholds relate to biomass burning emissions across various biomes and seasons.

4. Prototype Development

• A prototype tool to calculate the potential range of emissions and burnt biomass from user-defined binned FRP thresholds, enabling actionable insights tailored to specific applications.
• Data visualization to compare emissions thresholds with burnt areas and/or air quality metrics.

---

Evaluation Criteria

• Innovative approach
• Easy to maintain / Future-proof approach
• Transferability
• Matching requirements