TL;DR
The ESA Fire_cci project aims to create a consistent, stable, and error-characterized global burned area (BA) dataset using European satellite sensors. This initiative addresses the critical need for accurate fire disturbance data to improve climate and atmospheric models, which are currently hampered by uncertainties in existing BA estimations.
What Problem Does This Paper Address?
Accurately quantifying global fire activity is crucial for understanding its impact on climate, atmospheric chemistry, and vegetation dynamics. Biomass burning is a significant contributor to global CO2 emissions, and its role in the Earth's system is complex and not fully understood. Current estimates of burned area worldwide, often derived from satellite data, suffer from high uncertainty and variability. This makes it challenging to accurately model the amount of biomass consumed by fires, their associated greenhouse gas emissions, and the broader implications for land-use change and climate feedback loops. The ESA Fire_cci project directly tackles this by developing a robust methodology to produce reliable BA information from multiple European satellite sources.
Methods
The Fire_cci project employs a multi-phase approach. It begins with a thorough user requirement survey to understand the specific needs of climate, atmospheric, and ecosystem modelers. This informs the product specifications, defining the desired spatial and temporal resolution, accuracy, and formats for the BA data. A key methodological step is the pre-processing of data from the ATSR, VEGETATION, and MERIS sensors. This involves advanced geometric correction, radiometric calibration, and atmospheric/topographic correction using algorithms like ATCOR. Crucially, masks for water, clouds, snow, and cloud shadows are developed to mitigate potential confusion with burned areas.
The core of the project involves developing and testing burned area detection and merging algorithms. A "Round-Robin" exercise was conducted to evaluate existing algorithms, with the goal of selecting the best performing one for global BA production. The final product will be a merging of BA information derived from the three sensors to create a synthetic, consistent dataset. Validation is a critical component, utilizing reference fire perimeters derived from Landsat TM/ETM+ data to assess spatial and temporal accuracy and stability.
Datasets
- Input Satellite Data:
- ATSR (Advanced Along-Track Scanning Radiometer)
- VEGETATION
- MERIS (MEdium Resolution Imaging Spectrometer)
- Validation Data:
- Landsat TM/ETM+ multitemporal images for generating reference fire perimeters.
- Background Data for Study Site Visualization:
- GFED_V3 (Global Fire Emissions Database version 3) emissions data.
The project aims to produce BA maps at two resolutions: a pixel-level product at the best potential resolution and a grid-level product at 0.5-degree resolution.
Key Insights and Ideas
- The Importance of Multi-Sensor Synergy: By combining data from ATSR, VEGETATION, and MERIS, the project aims to overcome the limitations of individual sensors and create a more comprehensive and robust BA product.
- User-Driven Product Development: The explicit inclusion of user requirement surveys ensures that the generated BA data will be directly applicable and useful for the target modeling communities.
- Rigorous Pre-processing is Paramount: The paper emphasizes that accurate atmospheric and topographic corrections, along with robust masking of confounding features, are essential for reliable BA detection.
- The Need for Standardized Validation: The development of a standard protocol for validation using Landsat data is crucial for ensuring the quality and comparability of the Fire_cci BA products.
Key Results
The paper is presented as a progress report on a project that is currently running. Therefore, final quantitative results for the global BA product are not yet available. However, the following progress has been made:
- Pre-processing: Corrected reflectances and masks for water, cloud, snow, and cloud shadows have been generated for ten study sites for three years (2005, 2006, and 2008). The full temporal series (1995-2009) is being processed for the Australia site.
- Validation Datasets: Reference fire perimeters from Landsat data for temporal validation are ready, and a dataset for spatial validation is being developed.
- Algorithm Development: The project is in the process of developing and testing BA detection and merging algorithms, with a Round-Robin exercise conducted to select the best performing algorithm.
The target accuracy values for the BA product are: 85% user and producer accuracy, <15% omission and commission errors, <1000 m geo-location accuracy, ±3.5 days temporal reporting accuracy, and 15% temporal stability.
Limitations and Future Work
- Current Status: The project is still in progress, and the final global BA outputs are not yet available. The focus is currently on developing and testing the algorithms.
- Algorithm Selection: The selection of the "best performing" algorithm is a critical step that will influence the final product.
- Computational Demands: The pre-processing of input data is described as computationally demanding.
- Future Work: The project will continue with the full temporal series processing, the development and implementation of the merging algorithm, and the extensive validation of the BA products. A key future step will be testing the BA information within atmospheric and carbon cycle models to analyze its advantages and limitations.
Summary updated