Fires and Burned Areas

Biomass burnings (including forest, grassland, peatland and agricultural fires) have important impacts on global terrestrial and atmospheric systems, affecting land cover, surface albedo, and the atmospheric concentration of greenhouse gases, chemically reactive species and aerosols. Several products have been generated in the last years to estimate total burned area, but uncertainties remain, particularly those associated to small and low intensity fires. Impact of climate and societal changes modify traditional fire regimes, extending fire seasons, increasing fire severity or introducing fire in sensitive areas.

The Fire_cci project of the European Space Agency Climate Change Initiative aims to generate consistent time series of burned area products to assess the extents of biomass burnings, as well as their spatial and temporal characteristics. Fire impacts on atmospheric and terrestrial processes are assessed, with particular attention to CO₂, CO and CH₄ emissions, modifications of vegetation patterns and biomass availability.

The global burned area products of the Fire_cci program are being developed from MERIS and MODIS sensors (300 m and 250m of spatial resolution respectively), complemented with a small-fire databased generated from medium resolution sensors (Sentinel-2’s MSI, 10-20 m of spatial resolution, and Proba-V, 100-300 m resolution) for the African continent. BA algorithms for new Sentinel-3 sensors (OLCI and SLSTR) will also be developed. Validation of the global products is based on a statistical sampling design of Landsat frames. For these sites, fire reference perimeters are generated based on multitemporal analysis of TM/ETM+/OLI images.

Fire plays a key role in the exchange of carbon between the terrestrial biosphere and the atmosphere for a variety of ecosystems. During a fire disturbance vast quantities of carbon are released from the vegetation, litter and soil directly into the atmosphere. This violent restructuring of carbon pools triggers a range of critical processes (post-fire dynamics) such as vegetation succession, changes in surface albedo and hydrology, permafrost thawing and degradation.

Despite its significance, even the most complex Dynamic Vegetation Models (DVMs) embedded in the IPCC General Circulation Models (GCMs) fail to properly represent fire behaviour and post-fire dynamics, a fact that remains unstated. As DVMs operate on a deterministic, gridcell-by-gridcell basis they are unable to simulate a host of important fire characteristics such as propagation, magnitude of area burned and stochastic nature. As a consequence, we are unable to use DVMs and therefor GCMs to determine fire feedbacks and dynamics in the current or a changing climate.

Here we describe a model-independent methodology which assimilates a burned area satellite product into a DVM to achieve for the first time realistic fire simulations. During the 1^st stage of the approach we use image analysis techniques and specifically 3-dimenstional Connected Component Labelling to identify individual fires in Global Fire Emissions Database (GFED) burned area images. We extract statistical information on a number of key fire characteristics such as area burned, fire extend and duration and compare against field data in order to verify our approach. The 2^nd stage involves assimilating this information in a DVM; with minimum model restructuring we retain the Fire Return Interval produced by the model whilst assigning pragmatic characteristics to its fire outputs and allowing realistic simulations of fire-related processes.

We focus our simulations in the Arctic and specifically Canada and Russia and we offer a snippet of how this approach permits models to engage in post-fire dynamics hitherto absent from any other model regardless of complexity.

Fire is a key component of the carbon cycle, related to greenhouse gases and aerosol emissions to the atmosphere. Its interaction with climate has also been shown in a variety of studies. Due to its relevance in the global system, the GCOS (Global Climate Observing System) identified fire disturbance as one of the Essential Climate Variables (ECV). The European Space Agency’s Climate Change Initiative (ESA-CCI) selected fire as one of the target ECVs to include in the CCI programme. In the first phase of the fire_cci project a burned area (BA) detection algorithm was developed for the MERIS sensor, obtaining global BA maps for years 2006 to 2008. Global validation and user assessment showed the potential of this product to contribute to a better characterization of burned area maps and emission estimates.

On its second phase (started in September 2015), the fire_cci project aims to process the full MERIS archive (2002 to 2012) with an improved version of the BA algorithm developed in phase 1. In addition, this algorithm is being adapted to the MODIS Visible and Near Infrared bands (bands 1 and 2), which have better spatial resolution (250 m) than existing BA products. Experience gained with the MERIS sensor will be used to adapt the BA algorithm to the MODIS VNIR bands, as they are common with the spectral discrimination space used for the MERIS sensor (bands 8 and 10). Better temporal resolution of MODIS VNIR channels is expected to help the identification of burned areas, particularly in cloudy regions. Moreover, the BA database could be extended to the 2000-2015 period. This paper will focus on showing the global results of the improved BA algorithm for MERIS, as well as their adaptation to MODIS VNIR bands for specific study areas, with different fire regimes and biophysical conditions. 

A number of global and continental coarse resolution burned area products have been developed, including the NASA MODIS, the ESA Fire CCI, and the EU Copernicus burned area products.  Several of these are candidate Essential Climate Variable (ECV) products that have been validated by comparison with reference burned area maps derived by visual interpretation of Landsat data.  The potential research, policy and management applications of ECV products place a high priority on  rigorous, quantitative assessment of ECV product accuracy and precision. This requires the adoption of design-based validation methods, where the reference data are selected via a probability sampling, allowing for unbiased estimation of accuracy metrics.  Design based techniques have been used for annual land cover and land cover change product validation, but have not been widely used for burned area product validation, or for the validation of global products that are highly variable in time and space.

Design based techniques developed for land cover validation typically use probability sampling to determine the location of the reference data: we demonstrate that when these techniques are adopted for burned area validation, they result in a significant bias in the accuracy metrics, due to the impermanent nature of fire effects.  As a solution, we present a tri-dimensional sampling grid that allows for probability sampling of Landsat reference data in both space and time. To sample the globe in the spatial domain with non-overlapping sampling units, the Thiessen Scene Area (TSA) tessellation of the Landsat path/row geometry is used. The TSA grid is combined with the 16-day Landsat acquisition calendar to provide tri-dimensonal elements (voxels). This allows for implementation of stratified random sampling designs where not only the location but also the time interval of the reference data are explicitly drawn by probability sampling. The statistical properties of the proposed approach are demonstrated by sampling simulations using the MODIS global fire products.

This new validation approach, used for the validation with Landsat data of the NASA MODIS and forthcoming VIIRS global burned area products, is a general one and can be applied to different burned area products e.g. CCI, Copernicus and future ones derived from Sentinel-2 data stream. 

Vegetation fires are a major disturbance in the Earth System. Fires change the biophysical properties and dynamics of ecosystems and alter terrestrial carbon pools. By altering the atmosphere’s composition, fire emissions exert a significant climate forcing.
To realistically model past and future changes of the Earth System, fire disturbances must be taken into account. Related modelling efforts require consistent global burned area observations covering at least 10 to 20 years. Guided by the specific requirements of a wide range of end users, the ESA fire_cci project is currently computing a new global burned area dataset. It applies a newly developed spectral change detection algorithm upon the full ENVISAT-MERIS archive (2002 to 2012). The algorithm relies on MODIS active fire information as “seed”. A first, formally validated version has been released for the period 2006 to 2008. It comprises a pixel burned area product (spatial resolution of 333 m) with date detection information and a biweekly grid product at 0.5 degree spatial resolution.
We compare fire_cci burned area with other global burned area products (MCD64, GFED4(s), GEOLAND) and a set of active fires data (hotspots from MODIS, TRMM, AATSR and fire radiative power from GFAS). Output from the ongoing processing of the full MERIS timeseries will be incorporated into the study, as far as available.
The analysis of patterns of agreement and disagreement between fire_cci and other products provides a better understanding of product characteristics and uncertainties.
The intercomparison of the 2006-2008 fire_cci time series shows a close agreement with GFED4 data in terms of global burned area and the general spatial and temporal patterns. Pronounced differences, however, emerge for specific regions or fire events. Burned area mapped by fire_cci tends to be notably higher in regions where small agricultural fires predominate. The improved detection of small agricultural fires by fire_cci can be related to the increased spatial resolution of the MERIS sensor (333 m compared to 500 in MODIS). This is illustrated in detail using the example of the extreme 2006 spring fires in Eastern Europe.
We further analysed the occurrence of active fire pixels detected by TRMM and AATSR, respectively, which are not picked up as burned by fire_cci.  The results indicate that the fires missed by fire_cci are primarily very small and that omission of burned area is rather related to the lower detection limit of the MERIS sensor than to omission errors in MODIS active fire product itself, which is used as mapping “seed”.  We also show that fire_cci maps a lower spatial extent of fire when compared to the recently released GFED4s. The latter complements GFED4 burned area (i.e. MCD64 burned maps) with indirect estimates of omitted small fires.
We identify and discuss strengths and weaknesses of each fire product. The cross-comparison results may provide guidance to users on which product to select for a given research question.