A satellite-based burned area dataset for the northern boreal region from 1982 to 2020

SPECIAL ISSUE |RESEARCH PAPER https://doi.org/10.1071/WF22102 A satellite-based burned area dataset for the northern boreal region from 1982 to 2020 José-Andrés Moreno-Ruiz A , José-Rafael García-Lázaro A , Manuel Arbelo B,* and Pedro A. Hernández-Leal B ABSTRACT Background. Fires in the boreal forest occur with natural frequencies and patterns. Burned area (BA) is an essential variable in assessing the impact of climate change in boreal regions. Aims. Spatial wildfire occurrence data since the 1950s are available for North America. However, there are no reliable data for Eurasia, mainly for Siberia, during the 1980s and 1990s. Methods. A Bayesiannetwork algorithm was applied to the Long-Term Data Record (LTDR) Version 5 to generate a BA DataSet (BA-LTDR-DS) for the Boreal region from 1982 to 2020, validated using official reference data and compared with the MODIS MCD64A1 product. Key results. A high correlation (>93%) with all the reference BA datasets was found. BA-LTDR-DS data grouped by decades estimated a linear increase in BA of 4.47 million ha/decade. This trend provides evidence of how global warming affects fire activity in these boreal forests. Conclusions. BA-LTDR-DS constitutes a unique data source for the pre-MODIS era, and becomes a reliable source when other products with higher spatial/spectral resolution are not available. Implications. The BA-LTDR-DS dataset constitutes the longest time series developed for the boreal region at this spatial resolution. BA-LTDR-DS could be used as input in global climate models, helping improve wildfire prediction capabilities and understand the interactions between fire, climate and vegetation dynamics. Keywords: AVHRR, Bayesian network algorithm, boreal forest, burned area mapping, Eurasia, LTDR, MODIS, North America, remote sensing, Siberia, time series analysis. Introduction The boreal forest is the most extensive terrestrial biome; it occupies ~14% of land on Earth in a circumpolar belt surrounding the subarctic regions of the northern hemisphere. Two-thirds of these forests are in Eurasia (Scandinavia and Russia), the remaining third in North America (Canada and Alaska). Boreal forests play a critical role in regulating climate and the global carbon cycle in the Earth–atmosphere system (Chapin et al. 2000; Kasischke et al. 2005). Boreal forests have been considered for years as a carbon sink (Jobbágy and Jackson 2000; Ciais et al. 2010; Pan et al. 2011). In fact, more carbon is stored in the boreal forest regions than in any other region of the planet, possibly up to twice as much carbon as is stored in tropical forests (Bradshaw and Warkentin 2015). However, recent studies show that boreal forest carbon sinks could even be becoming a net source of emissions (Bonan 2008; Kurz et al. 2008; Bradshaw and Warkentin 2015; Portier et al. 2019; Eckdahl et al. 2022). The rapid warming that the boreal and Arctic regions have experienced over the past 30 years is modifying the dynamics of natural disturbances that were historically dominated by fire (Walsh 2014). This anomalous increase in temperature alters the humidity of the fuels and, therefore, modifies the severity, the natural regime of boreal forest fires and the burned area, which leads, in turn, to possible feedback effects on climate change (Goldammer and Furyaev 1996; Flannigan et al. 2005; Balshi et al. 2009a, 2009b; Tchebakova et al. 2009; Georgiadi et al. 2010; de Groot et al. 2013; Kelly et al. 2013; Coffield et al. 2019). The monitoring of these changes and the analysis of future scenarios are of vital importance to implement For full list of author affiliations and declarations see end of paper *Correspondence to: Manuel Arbelo Departamento de Física, Universidad de La Laguna, 38200 San Cristóbal de La Laguna, Spain. Email: [email protected] Received: 22 June 2022 Accepted: 15 April 2023 Published: 4 May 2023 Cite this: Moreno-Ruiz J-A et al. (2023) International Journal of Wildland Fire 32(6), 854–871. doi:10.1071/WF22102 © 2023 The Author(s) (or their employer(s)). Published by CSIRO Publishing on behalf of IAWF. This is an open access article distributed under the Creative Commons AttributionNonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND) OPEN ACCESS management policies on climate change that protect both the boreal forest and the carbon it stores (Bonan et al. 1992; Kasischke et al. 1995; Fuchs et al. 2009; Shuman et al. 2011; Loboda et al. 2012; Krylov et al. 2014; Ponomarev et al. 2016). Satellite remote sensing has become an effective technique for the identification and spatio-temporal characterisation of forest fires and burned areas owing to its coverage on both a global and regional scale (Andreae 1991; Cooke et al. 1996; Dwyer et al. 2000; Duncan 2003; Chu and Guo 2015). Currently, different platforms provide satellite images that are used in algorithms to map burned area (BA) (Vivchar 2011; Mouillot et al. 2014; Campagnolo et al. 2016; Chen et al. 2016a, 2016b). The ModerateResolution Imaging Spectroradiometer (MODIS) on board NASA’s Terra and Aqua satellites and the Advanced Very High Resolution Radiometer (AVHRR) on board the series of National Oceanic and Atmospheric Administration (NOAA) satellites are examples of such sensors. They provide daily imagery and higher-level land and atmosphere products for global mapping. For reliability and period covered, the most frequently used BA products are MODIS Collection 6 (MCD64A1) (Giglio et al. 2018; Boschetti et al. 2019) from the year 2000 onwards, the Global Fire Emissions Database version 4 (GFED-4) BA available from mid-1995 through to the present (Giglio et al. 2013), and the most recent MODIS product, Fire_cci v5.1, available for the years 2001–2020 (Chuvieco et al. 2018). As Chuvieco et al. (2008) argue, ‘Longer time series data are required to acquire a better understanding of fire regimes, and their mutual relationships with global warming.’ In spite of the fact that current satellite remote sensing systems (and their derived fire products) have enhanced temporal, spatial and spectral resolutions, the availability of well-built geospatial time series is scarce for Eurasia, especially for Siberia (de Groot et al. 2013; Chen et al. 2016a; Eberle et al. 2016; Ponomarev et al. 2016). In addition, measures from satellite data present strong discrepancies in BA estimations with regard to reported data in the official records (Soja et al. 2004; Sukhinin et al. 2004; Vivchar 2011; Kukavskaya 2013; Chen et al. 2016b), unlike North America, which has been well studied (Kasischke and French 1995; Al-Saadi et al. 2008; Chuvieco et al. 2008; Soja et al. 2009; Kasischke et al. 2011; Moreno Ruiz et al. 2012; Loboda et al. 2013; Moreno-Ruiz et al. 2014a, 2014b, 2019). In this paper, we present the Burned Area Long-Term Data Record DataSet (BA-LTDR-DS), a unique long-term BA product of the boreal forest in the Climate Modelling Grid (CMG) resolution for four decades (1982–2020). The CMG format, in a latitude/longitude geographic projection with a resolution of 0.05°, allows the BA-LTDR-DS dataset to be used as input in global climate models, helping to improve wildfire prediction capabilities and understand the interactions between fire, climate and vegetation dynamics in the northern boreal region. Materials and methods Study region The study region is geographically delimited by the parallels 60°N and 72.5°N, divided in turn into two sub-regions, North America and Eurasia (Fig. 1). The first sub-region includes Alaska and the northern part of the Canadian boreal region and is bounded at the upper right corner of the map at 72.5°N, 168.5°W and the bottom left corner at 60°N 43.5°W. The North American sub-region contains approximately one-third of the entire boreal region of this continent based on the map described by Brandt (2009). The second region stretches from Scandinavia to the Pacific coasts of Siberia and is limited at the upper right corner at 72.5°N 5°E and the bottom left corner at 60°N 180°E. Evergreen coniferous forests (pine and spruce) predominate in the two sub-regions. However, deciduous forests, mainly birch and larch, can also be found, with a spatial distribution influenced by post-fire dynamics (Rogers et al. 2015). The forests of North America tend to have more black spruce, white spruce and pine species with branches lower to the ground, thinner bark and serotinous cones that open after being burned by fire. The Eurasian forests have more fire-resistant species with thick bark, wetter needles and fewer low branches. In addition, between 25 and 30% of the landscape of the boreal forest region is peatland (organic soils) (Gorham 1991; Wieder et al. 2006; Beaulne et al. 2021; Nelson et al. 2021). The presence of different species between North American and Eurasian forests marks a notable difference in fire regimes between the two regions (de Groot et al. 2013). Fire regimes can differ in the same biome, and the boreal forests of North America and Eurasia are an example of this (Haas et al. 2022). Fires in North America tend to be larger and more intense, with higher fuel consumption (crown fires); in contrast, fires in Eurasia tend to be less intense, with lower fuel consumption (surface fires) (Wooster and Zhang 2004; Wirth 2005; de Groot et al. 2013; Sitnov and Mokhov 2018). Reference data The only databases that include reliable information on fires detected in boreal forests for more than four decades are those created and maintained by US and Canadian forestry agencies. For Eurasia, however, there is no reliable BA reference set covering the study period under consideration (1980–2020), with only MODIS sensor-derived BA products available since 2000. The Alaska Fire Service (AFS), at Fort Wainwright, AK, USA, maintains a detailed record of all detected fire events since 1940 (https://fire.ak.blm.gov/). This database, in addition to providing the perimeter of each BA, includes additional fire-related information such as the management office, fire name, geographical coordinates, estimated area, cause and relevant comments. The perimeters are always delineated from the best available data source, which can www.publish.csiro.au/wf International Journal of Wildland Fire 855 include aerial and high spatial resolution satellite imagery (Landsat type), as well as topographic maps. AFS recognises differences in the scale and accuracy of the perimeters depending on the period. Fires larger than 400 ha are included for fires before 1987, those larger than 40 ha from 1987 to 1989, whereas from 1990 onwards, all fires with a BA larger than 4 ha are considered. The Canadian National Fire Database (CNFDB) is compiled and maintained by the Canadian Forest Service (https:// cwfis.cfs.nrcan.gc.ca/ha/nfdb). CNFDB records fire data of all sizes since 1959 that include fire location and perimeter data supplied by Canadian territorial fire management agencies. The information contained in the CNFDB may not be complete or error-free owing to the different mapping techniques used. In addition, the completeness and quality of the data may vary between agencies and between years. The quality of this database and its usefulness have been demonstrated in several publications (Amiro et al. 2001; Stocks et al. 2003; Parisien et al. 2006; Burton et al. 2008; Hanes et al. 2019). Recently, the Canada Centre for Mapping and Earth Observation and the Canadian Forest Service developed a new database called the National Burned Area Composite (NBAC) (Hall et al. 2020; Skakun et al. 2021). NBAC improves the BA determination of the CNFDB database using an automatic method based on hotspots and the Normalized Difference Vegetation Index (NDVI) applied to high spatial resolution (less than 30 m) satellite imagery (Hall et al. 2020). NBAC is available for fires recorded from 1986 onwards (Skakun et al. 2022). The MODIS product selected was the MCD64A1 Collection 6 (Giglio et al. 2018), based on data from the Terra and Aqua satellites and distributed by the Land Processes Distributed Active Archive Center (LP DAAC). MCD64A1 is the NASA official BA product. It is a global monthly gridded product of 500-m spatial resolution. The algorithm used to detect burned pixels considers atmospherically corrected surface reflectances of the shortwave infrared Bands 5 and 7 through a normalised vegetation index in conjunction with active fire data at 1-km resolution. The product contains the estimated burn date, unburned, or even unmapped areas if there were no data to establish burned/unburned status (Giglio et al. 2018). MCD64A1 C6 is currently the product with the highest reliability compared with the other BA products (Padilla et al. 2015; Moreno-Ruiz et al. 2020), and it has the lowest commission and omission errors in the boreal forest region (Boschetti et al. 2019). Pre-processing of the Long-Term Data Record (LTDR) The LTDR funded by the Climate Data Record Program of the NOAA National Climatic Data Center is a consistent long-term dataset at a spatial resolution of 0.05° (~5 km) based on daily data from the AVHRR onboard the NOAA satellites and daily data acquired by MODIS onboard NASA’s Terra and Aqua satellites (Pedelty et al. 2007). The daily global LTDR version 5 (1981–2021) used in the present study was downloaded from https://ltdr.nascom.nasa.gov/cgi-bin/ltdr/ltdrPage.cgi. 130°W 50°W 40°W 30°W 20°W 10°W 0°E 10°E 20°E 30°E 40°E 50°E 140°W 150°W 160°W 170°W 180°W 170°E 160°E 150°E 140°E 130°E 120°E 110°E 100°E 90°E 80°E 70°E 60°E 120°W 110°W 100°W 90°W 80°W 70°W 60°W Fig. 1. The study region (red perimeter) covers the northern boreal region. It is divided into two sub-regions: North America (72.5°N 168.5°W; 60°N 43.5°W) and Eurasia (72.5°N 5°E; 60°N 180°E). Boreal forest (green) is differentiated from all the other land covers (brown), ice (white) and water bodies (blue). J-A Moreno-Ruiz et al. International Journal of Wildland Fire 856 The original files, in hierarchical data format, cover the globe at a 0.05° resolution CMG with 7200 × 3600 cells. The original files were transformed into a binary sequential format (BSQ). Conversion to physical values (surface reflectance and brightness temperature) considered the quality assessment (QA) fields of the Daily Surface Reflectance product (AVH09C1) to remove possible snow-covered pixels and to filter for the presence of clouds using the CLAVR-1 (Clouds from AVHRR-Phase I) algorithm (Stowe et al. 1999). Missing and invalid brightness temperature values found for the NOAA-16 and 18 satellites from 2000 to 2008 were directly replaced by the equivalent values from the MOD09CMG product for that period (https://ladsweb.modaps.eosdis.nasa.gov/ archive/allData/6/MOD09CMG). In addition, the significantly decaying orbit of the NOAA-19 satellite from 2018 (Julien and Sobrino 2021; Giglio and Roy 2022) made it necessary to use the MOD09CMG product for this period as well. Finally, to eliminate residual clouds and cloud shadows that could interfere with the discrimination of burned pixels, the maximum brightness temperature (BT_CH3: 3.55–3.93 µm) criterion constructed 10-day composites from the LTDR-BSQ files (Barbosa et al. 1998). This compositing criterion has proved to be effective in discerning burned from unburned areas (Chuvieco et al. 2005). Temperature values above 350 K were considered as erroneous values. Next, two vegetation indices derived from the original bands were calculated, which were useful for BA discrimination in the boreal regions: Global Environmental Monitoring Index (GEMI) (Pinty and Verstraete 1992) and Burned Boreal Forest Index (BBFI) (Moreno Ruiz et al. 2012). Table 1 describes the band configurations of each 10-day composite file in the BSQ format with floating data type. The burned area detection algorithm The methodology developed by Moreno Ruiz et al. (2012), based on a Bayesian network algorithm (BA-LTDR), was applied to the entire study area to obtain the annual BA maps and their corresponding temporal distribution (Moreno Ruiz et al. 2012). This methodology has been previously applied and successfully validated in different boreal regions and for different periods (Núñez-Casillas et al. 2013; MorenoRuiz et al. 2014a; García-Lázaro et al. 2018). In the current work, the study region was extended to the northern boreal region (above 60°N) using a single algorithm. This approach allowed the coherence in the two sub-regions (North America and Eurasia) and the two study periods (MODIS and pre-MODIS eras) to be assessed jointly, such that the estimates obtained for one sub-region and period can be extrapolated to other boreal sub-regions and periods. The different steps that constitute this methodology are summarised in Fig. 2 and are described in greater depth in previous studies by the same authors (Moreno-Ruiz et al. 2012, 2014b; Núñez-Casillas et al. 2013; García-Lázaro et al. 2018; Guindos-Rojas et al. 2018). The algorithm calculates 12 statistical variables based on the surface reflectance bands ρ 1 and ρ 2 , the brightness temperature T, and the BBFI and GEMI indices, for the 10-day composite of potential fire dates before and after the fire for the year of the fire event, the year before and the year after. In the northern boreal forest, vegetation takes several years to recover, even more than a decade, which is why it was decided to also analyse the year before and after the fire, with a duration of 2 months for the pre-fire and post-fire periods of each year, starting from the hypothetical ignition date determined by the value of the maximum of the BBFI (Moreno Ruiz et al. 2012). The Bayesian network classifier calculated the normalised probability for the unburned and burned classes using a training set based on the perimeters of the BA that was larger than 1000 ha in the NE Siberia region in 2010. These perimeters were generated from 53 pairs of LandsatTM images obtained from the United States Geological Survey (USGS) considering preand post-fire information at 30-m spatial resolution (García-Lázaro et al. 2018). To improve the resulting BA probability maps, the spatial coherence was analysed using a filtering process based on cellular automata theory (Mojaradi et al. 2004; Espinola et al. 2015). Finally, the BA-LTDR algorithm developed was applied to the study region in order to generate a dataset of BA annual maps of the northern boreal forest Table 1. Bands configuration of 10-days composite files. Band name Description Equation ρ 1 Surface reflectance for red channel SREFL_CH1 (0.5–0.7 µm) ρ 2 Surface reflectance for near-infrared channel SREFL_CH2 (0.7–1.0 µm) T Top of atmosphere brightness temperature (K) BT_CH3 (3.55–3.93 µm) GEMI Global Environmental Monitoring Index n n× (1 0.25 × ) 0.125 1 1 1 n=2 × ( ) + 1.5 × + 0.5 × + + 0.5 22122 1 2 1 BBFI Burned Boreal Forest Index +T1 2 2 QA Quality assessment field www.publish.csiro.au/wf International Journal of Wildland Fire 857 (above 60°N) for the 1982–2020 period at 0.05° (~5 km) resolution. We refer to this dataset as the BA-LTDR-DS. Technical validation To assess the accuracy of the BA-LTDR-DS dataset, we split the time series for both sub-regions (Eurasia and North America), into two – pre-MODIS and post-MODIS. Since 2000 (MODIS era), accuracy assessment for the Eurasian sub-region was accomplished using only the MCD64A1 C6 BA product. Although MODIS data are not the most suitable for quantitative validation of the BA-LTDR-DS product, as they are affected by errors due to their spatial resolution of 500 m, their use may be considered according to the protocol of the Committee on Earth Observation Satellites (CEOS). CEOS recommends that an assessment can be made for systematic quality control of a product by statistical comparison with independently obtained BA datasets of better spatial resolution when no other official reference set exists (Morisette et al. 2006; Boschetti et al. 2009). Before 2000 (pre-MODIS era), accuracy assessments could not be evaluated in this way because other BA products with a higher spatial resolution for this study sub-region (Eurasia) were not available. Fortunately, this is not the case for the North American sub-region, for which official reference data are available to compare with the BA-LTDR-DS dataset, as described in Reference data section. The AFS database and the CNFDB have already been used successfully as reference data to assess the accuracy of satellite-derived BA products (Chuvieco et al. 2008; Chang and Song 2009; Giglio et al. 2009; Núñez-Casillas et al. 2013; Moreno-Ruiz et al. 2019, 2020). The new reference database NBAC, available for the period 1986–2020, was used as the reference set for Canada. The first 4 missing years (1982–1985) were completed with data from CNFDB. NBAC significantly improves the CNFDB BA polygons by including small fires and some fires in remote locations not previously considered in CNFDB, and by removing unburned islands and water bodies within those polygons (Hall et al. 2020; Skakun et al. 2021, 2022). The polygons of all the fires registered in the AFS and CNFDB + NBAC between 1982 and 2020 were used to produce the annual vector layers for ground-truth verification. Next, these vector layers were reprojected to a geographic projection with a pixel size of 0.005° (~500 m) to generate annual ground-truth maps. To determine how the pixel was assigned a burned/non-burned value, the method of maximum area within the pixel was used (Arnone et al. 2016). However, to compute spatial and temporal accuracy, the annual BA maps from the BA-LTDR-DS and the MCD64A1 C6 datasets were clipped to the North American (72.5°N, 168.5°W, 60°N, 141°W) and Eurasian (72.5°N, 5°E, 60°N, 180°E) boreal sub-regions. All reference data maps were resized to a geographic projection with a pixel size of 5 × 5 km by pixel aggregation (an aggregated pixel represents the percentage of BA at the subpixel level). Daily LTDR-BSQ (Northern boreal region) MOD09CMG 10-day composites (ρ, T, GEMI, BBFI) preand post-fire North America Eurasia AFS + (CNFDB + NBAC) (1982–2020) MCD64A1 C6 (2001–2020) MCD64A1 C6 (2001–2020) Accuracy assessment Bayesian network algorithm Burned probability maps Spatial coherence analysis 160°W 70°N 65°N 60°N 150°W Daily global LTDR version 5 (1981–2021) Fig. 2. Flowchart of the process to obtain annual maps of the BA-LTDR-DS (Burned Area Long-Term Data-Record Dataset) in the northern boreal region for 1982–2020 and to assess their accuracy against reference data ( Moreno Ruiz et al. 2012; NúñezCasillas et al. 2013; Moreno-Ruiz et al. 2014b; García-Lázaro et al. 2018; Guindos-Rojas et al. 2018). J-A Moreno-Ruiz et al. International Journal of Wildland Fire 858 The temporal accuracy of the BA-LTDR-DS product in each sub-region was assessed considering the total calculated annual BA. A timing distribution of the BA-LTDR-DS product was represented on a chart together with the time series of reference BAs, and a correlation analysis was carried out. The relative percentages of the annual BA of the BA-LTDR-DS product were calculated with respect to the reference data for the common years when available. For the spatial accuracy assessment, scatter plots of the annual BA proportions on 50 × 50 km grids distributed uniformly for the BA-LTDR-DS dataset against the reference data in each sub-region were constructed and a linear regression analysis was performed. Next, a detailed analysis of the spatial accuracy for the BA-LTDR-DS dataset was made based on error matrixes versus the reference maps at the pixel level on an annual basis, calculating commission and omission errors of the burned class (Stehman 1997). Omission errors were calculated as the ratio of burned pixels classified as unburned to the total burned pixels in the BA reference map, while commission errors were calculated as the ratio of unburned pixels classified as burned to the total burned pixels in each BA product under analysis. We considered a pixel size of 50 km to prevent errors derived from geo-referencing of the images due to the difference in spatial resolution (Moreno-Ruiz et al. 2014a). Results Annual burned area maps for the northern boreal region (LTDR-BA-DB) Fig. 3 presents the grouping by decade of the annual maps of BAs for the two boreal sub-regions considered (North America and Eurasia) obtained from the LTDR-BA-DS product with the burned pixels represented in red. Fig. 4 shows a composition with the four decades of the BA detected by BALTDR-DS for the northern boreal region considered between the parallels 60°N and 72.5°N. Annual distribution of the burned area estimates Fig. 5 shows the estimated annual distribution of BA in the northern boreal region for the period 1982–2020 from the BA-LTDR-DS. A non-uniform pattern was observed with years where strong fire activity was detected (BA > 3 million ha) and other years where barely 0.5 million ha BA was detected. On average, for the entire region and period, ~1.84 million ha burned per year, but with high variability. For example, in 2014 (the year with the greatest BA detected), 5.41 million ha burned whereas in 1992 (the year with the least BA detected), ~0.26 million ha burned. For the entire period analysed, North America had 35.9% of the BA compared with 64.1% in Eurasia. However, this average contribution of each subregion seems to have no statistical significance owing to the large annual fluctuations. It should be noted, for example, that the greatest imbalances occurred in the year 2020, with a contribution of 0.8% from North America compared with 99.2% from Eurasia, or the opposite case for 2004, where North America contributed 96.7% to the total compared with 3.3% from the rest. Temporal accuracy North America In the North America boreal sub-region, the total BA registered by the AFS and CNFDB + NBAC databases in the 1982–2020 period was 41.23 million ha, with an irregular annual distribution. The highest fire activity (4.79 million ha) occurred in 2004, and the lowest in 1984, when only 0.10 million ha burned. Fig. 6 shows the annual distribution of BA from the reference AFS and CNFDB + NBAC databases, the estimated BA for the BA-LTDR-DS product for the period 1982–2020 and estimates for the MCD64A1 C6 product for the period 2000–2020. The BA-LTDR-DS detected 62% of the reference BA, underestimating the BA in all years in the time series. There is a strong correlation (0.93) between the LTDRBA-DS data and the reference data. If we divide the time series into two parts, pre-MODIS (1982–1999) and MODIS (2000–2020), the BA-LTDR-DS presents almost homogeneous behaviour, both in the percentage of the estimate of BA (66% vs 60%) and in the correlation coefficient (0.97 vs 0.93) with respect to the reference data. For its part, the MCD64A1 C6 product underestimated BA in North America by approximately 69% with a correlation coefficient of 0.99 with respect to the reference data. Making an inter-comparison of the BA-LTDR-DS with MCD64A1 C6 for the common period (2000–2020), a correlation of 0.93 was found. Eurasia For the time accuracy assessment in the Eurasian subregion, a correlation analysis between the time series of the BA-LTDR-DS and the MCD64A1 C6 product was conducted (Fig. 7). MCD64A1 C6 was used as the reference data for the common years (2000–2020). MCD64A1 C6 estimated a BA of 47.10 million ha for these years and the BA-LTDR-DS ~66% of that value. The BA-LTDR-DS underestimated the BA in all common years except for 2000. However, its temporal pattern fits remarkably well with that of the reference, yielding a correlation coefficient between MCD64A1 C6 and BA-LTDR-DS of 0.95. Spatial accuracy North America Table 2 shows the results of the linear regression analysis of the BA percentages of the BA-LTDR-DS and the MCD64A1 C6 products versus the reference data. The average determination coefficient (R 2 ) over the 1982–2020 period for the BA-LTDR-DS product was 0.78, with a slope of 0.73, with no www.publish.csiro.au/wf International Journal of Wildland Fire 859 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 70°N 65°N 60°N 70°N 65°N 60°N 60°W 50°W 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 60°W 50°W (a) North America 1980s 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 70°N 65°N 60°N 70°N 65°N 60°N 60°W 50°W 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 60°W 50°W (b) North America 1990s 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 70°N 65°N 60°N 70°N 65°N 60°N 60°W 50°W 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 60°W 50°W (c) North America 2000s 160°W 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 70°N 65°N 60°N 70°N 65°N 60°N 70°N 65°N 60°N 70°N 65°N 60°N 60°W 50°W 160°W 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E 150°W 140°W 130°W 120°W 110°W 100°W 90°W 80°W 70°W 60°W 50°W (d) North America 2010s (e) Eurasia 1980s 70°N 65°N 60°N 70°N 65°N 60°N 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E (f) Eurasia 1990s 70°N 65°N 60°N 70°N 65°N 60°N 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E (f) Eurasia 1990s 70°N 65°N 60°N 70°N 65°N 60°N 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E 10°E 20°E 30°E 40°E 50°E 60°E 70°E 80°E 90°E 100°E 110°E 120°E 130°E 140°E 150°E 160°E 170°E 180°E (h) Eurasia 2010s Fig. 3. Decadal burned area maps for North America (a–d), and Eurasia (e–h) sub-regions from the Burned Area Long-Term Data Record DataSet (BA-LTDR-DS). Red, burned area; blue, water; green, non-burned. J-A Moreno-Ruiz et al. International Journal of Wildland Fire 860 significant differences between the two time sub-intervals (pre and MODIS eras). The average determination coefficient (R 2 ) over the 2000–2020 period for MCD64A1 C6 and the BA-LTDR-DS products was 0.88 and 0.78, with a slope of 0.69 and 0.69, respectively. A detailed analysis of the spatial accuracy for the North American sub-region on an annual basis for the BA products BA-LTDR-DS and MCD64A1 C6 is shown in Table 3. The average commission and omission errors for the BA-LTDRDS are 0.14 and 0.47 respectively for the entire study period whereas for the MODIS era, the MCD64A1 C6 product presents a commission error of 0.09 and omission error of 0.37. Eurasia Table 4 shows the results of the accuracy of the BA estimate obtained from the linear regression analysis of the BA percentages from the BA-LTDR-DS product versus MCD64A1 C6 using 50 × 50 km grids. The average determination coefficient (R 2 ) over the 2001–2020 period for the BA-LTDR-DS product was 0.78, with a slope of 0.81. Table 5 shows the commission and omission errors derived from the error matrix for each year, taking as reference the MCD64A1 C6 product (only available from the year 2000). 140°W 160°W 160°E 140°E 130°E 110°E 90°E 70°E 50°E 130°W 110°W 90°W 70°W 50°W 40°W 1980s 1990s 2000s 2010s 20°W 20°E 40°E0° 0° Fig. 4. Mapping of burned areas detected by BA-LTDR-DS (Burned Area Long-Term Data-Record Dataset) from 1982 to 2020 for the northern boreal region between 60°N and 72.5°N. Colours correspond to burned areas by decade. AFS + (CNFDB _NBAC) BA-LTDR-DS MCD64A1 C6 0 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 1 2 3 Burned area (million. ha) 4 5 Year Fig. 6. Annual distribution of burned area estimate (ha) in the North American boreal sub-region from reference data of AFS (Alaska Fire Service) and CNFDB + NBAC (Canadian National Fire Database + National Burned Area Composite), and the BA-LTDR-DS (Burned Area Long-Term Data-Record Dataset) and the MODIS Collection 6 MCD64A1 C6 burned area products. 0 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 1 2 3 Burned area (million. ha) 4 5Eurasia North America Year 6 Fig. 5. Annual distribution of burned area estimate (ha) in the northern boreal region from the BA-LTDR-DS (Burned Area Long-Term Data-Record Dataset). www.publish.csiro.au/wf International Journal of Wildland Fire 861 For this period, the average commission and omission errors for the BA-LTDR product were 0.21 and 0.47, respectively. Discussion In this paper, we describe the BA-LTDR-DS dataset, which provides annual BA at a spatial resolution of 0.05° from 1982 to 2020 for the boreal region between 60°N and 72.5°N. We built this dataset from the latest released Version 5 of the LTDR dataset. The LTDR Version 5 incorporates improvements of the Bidirectional Reflectance Distribution Function (BRDF) correction, the calibration of AVHRR/3 data on board platforms NOAA-16, 18 and 19, compositing atmospheric corrections and the QAs (https://landweb.modaps.eosdis. nasa.gov/cgi-bin/ltdr/ltdr/ltdrPage.cgi?fileName=LTDR_ update). However, as has been discussed in previous studies (Otón et al. 2019, 2021; Giglio and Roy 2022), these effects cannot be eliminated. Indeed, in this new version, we detected missing and wrong values for the TOA brightness temperature bands from the NOAA-16 and 18 satellites (from 2000 to 2008) in the northern boreal region. To preserve the coherence of the temperature bands in the LTDR dataset, we replaced the Band T3 with the Band 20 Brightness Temperature (3.360–3.840 μm) of the MOD09CMG product. In addition, the significantly decaying orbit of the NOAA-19 satellite from 2018 (Julien and Sobrino 2021; Giglio and Roy 2022) made it necessary to use the MOD09CMG product for this period. Finally, the QA bit was updated properly. Using this new modified version of the LTDR, we have generated the longest BA time series yet built at a spatial resolution of 0.05° in the CMG for the northern boreal region. To do this, a machine learning algorithm based on a Bayesian network was used, developed specifically for the detection of BA in that region. The BA-LTDR-DS extends by more than 10 years the time interval of the Global Fire Emissions Database (GFED4), which is from 1995 to the present (Giglio et al. 2013), and improves its spatial resolution by up to five times (from 0.25° to 0.05°) as well as the beta long-term BA dataset (FireCCILT1.0) developed by the Climate Change Initiative (CCI) program of the European Space Agency. In a recent version of the CCI program, Otón et al. (2021) obtained a new product (FireCCILT11) at the same spatial resolution as the BA-LTDR-DS. This product uses a random forest algorithm that calculates the percentages of BA for each pixel (soft classification), unlike BA-LTDR-DS, which only determines whether the pixel is completely burned or not (hard classification). The main problem encountered when comparing BA-LTDR_DS with FIRECCILT11 relates to the years composing both time series: FireCCILT11 is 2 years shorter (ending in 2018) and does not include the year 1994, which considerably distorts the comparison as 1994 was the year with the largest BA in North America. Evaluating the spatial and temporal accuracy of the entire time series of BA obtained (BA-LTDR-DS) seems very difficult given the non-existence of another set of reference data for the boreal forest region and the period analysed (from 1982 to 2020) compared with similar products, either obtained from fire records perimeters of official agencies or from products derived from satellite images. That is why we were forced to evaluate our product for the two sub-regions already described (Eurasia and North America) and two different periods marked by the year 2000 when the MODIS sensor was put into operation. For the assessment of the BA-LTDR-DS accuracy in the northern sub-region of North America, the best available baseline data (ground-truth) were used, i.e. the perimeters of BAs recorded by the AFS and Canada (CFSFND + NBAC). Few countries have detailed registries of BA perimeters available to use for a full assessment of products resulting from satellite images. The availability of the above information made it possible to conduct a detailed study of the various products based on the total amount of fires rather than a simple sample (Moreno-Ruiz et al. 2019). These databases, although regularly maintained and updated, may contain errors mainly due to the omission of small fires 0 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 1 2 3 Burned area (million. ha) 4 5BA-LTDR-DS MCD64A1 C6 Year Fig. 7. Annual distribution of burned area estimate (ha) in the Eurasian boreal sub-region from the BA-LTDR-DS (Burned Area Long-Term Data-Record Dataset) and the MODIS Collection 6 MCD64A1 C6 burned area products. J-A Moreno-Ruiz et al. International Journal of Wildland Fire 862 chemistry model. Journal of Geophysical Research: Atmospheres 101, 21051–21065. doi:10.1029/96JD01835 de Groot WJ, Cantin AS, Flannigan MD, Soja AJ, Gowman LM, Newbery A (2013) A comparison of Canadian and Russian boreal forest fire regimes. Forest Ecology and Management 294, 23–34. doi:10.1016/j. foreco.2012.07.033 Duncan BN (2003) Interannual and seasonal variability of biomass burning emissions constrained by satellite observations. Journal of Geophysical Research 108, 4040. doi:10.1029/2002JD002378 Dwyer E, Pinnock S, Gregoire JM, Pereira JMC (2000) Global spatial and temporal distribution of vegetation fire as determined from satellite observations. International Journal of Remote Sensing 21, 1289–1302. doi:10.1080/014311600210182 Eberle J, Urban M, Homolka A, Hüttich C, Schmullius C (2016) MultiSource Data Integration and Analysis for Land Monitoring in Siberia. In ‘Novel Methods for Monitoring and Managing Land and Water Resources in Siberia’. (Eds L Mueller, AK Sheudshen, F Eulenstein) pp. 471–487. (Springer International Publishing: Cham) Eckdahl JA, Kristensen JA, Metcalfe DB (2022) Climatic variation drives loss and restructuring of carbon and nitrogen in boreal forest wildfire. Biogeosciences 19, 2487–2506. doi:10.5194/bg-19-2487-2022 Espinola M, Piedra-Fernandez JA, Ayala R, Iribarne L, Wang JZ (2015) Contextual and hierarchical classification of satellite images based on cellular automata. IEEE Transactions on Geoscience and Remote Sensing 53, 795–809. doi:10.1109/TGRS.2014.2328634 Flannigan MD, Logan KA, Amiro BD, Skinner WR, Stocks BJ (2005) Future area burned in Canada. Climatic Change 72, 1–16. doi:10.1007/ s10584-005-5935-y Flannigan MD, Krawchuk MA, de Groot WJ, Wotton BM, Gowman LM (2009) Implications of changing climate for global wildland fire. International Journal of Wildland Fire 18, 483–507. doi:10.1071/ WF08187 Franquesa M, Rodriguez-Montellano AM, Chuvieco E, Aguado I (2022a) Reference data accuracy impacts burned area product validation: the role of the expert analyst. Remote Sensing 14, 4354. doi:10.3390/ rs14174354 Franquesa M, Stehman SV, Chuvieco E (2022b) Assessment and characterization of sources of error impacting the accuracy of global burned area products. Remote Sensing of Environment 280, 113214. doi:10.1016/j.rse.2022.113214 Fuchs H, Magdon P, Kleinn C, Flessa H (2009) Estimating aboveground carbon in a catchment of the Siberian forest tundra: Combining satellite imagery and field inventory. Remote Sensing of Environment 113, 518–531. doi:10.1016/j.rse.2008.07.017 García-Lázaro JR, Moreno-Ruiz JA, Riaño D, Arbelo M (2018) Estimation of burned area in the northeastern Siberian boreal forest from a Long-Term Data Record (LTDR) 1982–2015 time series. Remote Sensing 10, 940. doi:10.3390/rs10060940 Georgiadi AG, Milyukova IP, Kashutina EA (2010) Environmental change in Siberia. Environmental Change in Siberia: Earth Observation, Field Studies and Modelling 40, 157–169. doi:10.1007/ 978-90-481-8641-9 Giglio L, Roy DP (2022) Assessment of satellite orbit-drift artifacts in the long-term AVHRR FireCCILT11 global burned area data set. Science of Remote Sensing 5, 100044. doi:10.1016/j.srs.2022.100044 Giglio L, Randerson JT, van der Werf GR, Kasibhatla PS, Collatz GJ, Morton DC, DeFries RS (2009) Assessing variability and long-term trends in burned area by merging multiple satellite fire products. Biogeosciences Discussions 6, 11577–11622. doi:10.5194/bgd-6-11577-2009 Giglio L, Randerson JT, van der Werf GR (2013) Analysis of daily, monthly, and annual burned area using the fourth-generation global fire emissions database (GFED4). Journal of Geophysical Research: Biogeosciences 118, 317–328. doi:10.1002/jgrg.20042 Giglio L, Boschetti L, Roy DP, Humber ML, Justice CO (2018) The Collection 6 MODIS burned area mapping algorithm and product. Remote Sensing of Environment 217, 72–85. doi:10.1016/j.rse.2018. 08.005 Goldammer JG, Furyaev V (1996) ‘Fire in ecosystems of boreal Eurasia.’ (Springer Science+Business Media, B.V, Kluwer Academic Publishers) Gorham E (1991) Northern peatlands: Role in the carbon cycle and probable responses to climatic warming. Ecological Applications 1, 182–195. doi:10.2307/1941811 Guindos-Rojas F, Arbelo M, García-Lázaro JR, Moreno-Ruiz JA, Hernández-Leal PA (2018) Evaluation of a Bayesian algorithm to detect burned areas in the Canary Islands’ Dry Woodlands and forests ecoregion using MODIS data. Remote Sensing 10, 789. doi:10.3390/ rs10050789 Haas O, Prentice IC, Harrison SP (2022) Global environmental controls on wildfire burnt area, size, and intensity. Environmental Research Letters 17, 065004. doi:10.1088/1748-9326/ac6a69 Hall RJ, Skakun RS, Metsaranta JM, Landry R, Fraser RH, Raymond D, Gartrell M, Decker V, Little J (2020) Generating annual estimates of forest fire disturbance in Canada: the National Burned Area Composite. International Journal of Wildland Fire 29(10), 878–891. doi:10.1071/WF19201 Hanes CC, Wang X, Jain P, Parisien M-A, Little JM, Flannigan MD (2019) Fire-regime changes in Canada over the last half century. Canadian Journal of Forest Research 49, 256–269. doi:10.1139/cjfr2018-0293 Jobbágy EG, Jackson RB (2000) The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological Applications 10, 423–436. doi:10.1890/1051-0761(2000)010[0423: TVDOSO]2.0.CO;2 Julien Y, Sobrino JA (2021) NOAA-AVHRR orbital drift correction: Validating methods using msg-seviri data as a benchmark dataset. Remote Sensing 13, 925. doi:10.3390/rs13050925 Kasischke ES, French NHF (1995) Locating and estimating the areal extent of wildfires in alaskan boreal forests using multiple-season AVHRR NDVI composite data. Remote Sensing of Environment 51, 263–275. doi:10.1016/0034-4257(93)00074-J Kasischke ES, Christensen Jr NL, Stocks BJ (1995) Fire, global warming, and the carbon balance of boreal forests. Ecological Applications 5, 437–451. doi:10.2307/1942034 Kasischke ES, Hyer EJ, Novelli PC, Bruhwiler LP, French NHF, Sukhinin AI, Hewson JH, Stocks BJ (2005) Influences of boreal fire emissions on Northern Hemisphere atmospheric carbon and carbon monoxide. Global Biogeochemical Cycles 19, GB1012. doi:10.1029/2004GB002300 Kasischke ES, Loboda T, Giglio L, French NHF, Hoy EE, de Jong B, Riano D (2011) Quantifying burned area for North American forests: Implications for direct reduction of carbon stocks. Journal of Geophysical Research: Biogeosciences 116, G04003. doi:10.1029/ 2011JG001707 Kelly R, Chipman ML, Higuera PE, Stefanova I, Brubaker LB, Hu FS (2013) Recent burning of boreal forests exceeds fire regime limits of the past 10,000 years. Proceedings of the National Academy of Sciences 110, 13055–13060. doi:10.1073/pnas.1305069110 Krylov A, McCarty JL, Potapov P, Loboda T, Tyukavina A, Turubanova S, Hansen MC (2014) Remote sensing estimates of stand-replacement fires in Russia, 2002–2011. Environmental Research Letters 9, 105007. doi:10.1088/1748-9326/9/10/105007 Kukavskaya EA, Soja AJ, Petkov AP, Ponomarev EI, Ivanova GA, Conard SG (2013) Fire emissions estimates in Siberia: evaluation of uncertainties in area burned, land cover, and fuel consumption. Canadian Journal of Forest Research 43, 493–506. doi:10.1139/cjfr2012-0367 Kurz WA, Stinson G, Rampley GJ, Dymond CC, Neilson ET (2008) Risk of natural disturbances makes future contribution of Canada’s forests to the global carbon cycle highly uncertain. Proceedings of the National Academy of Sciences 105, 1551–1555. doi:10.1073/pnas. 0708133105 Loboda TV, Zhang Z, O’Neal KJ, Sun G, Csiszar IA, Shugart HH, Sherman NJ (2012) Reconstructing disturbance history using satellite-based assessment of the distribution of land cover in the Russian Far East. Remote Sensing of Environment 118, 241–248. doi:10.1016/j.rse.2011.11.022 Loboda TV, French NHF, Hight-Harf C, Jenkins L, Miller ME (2013) Mapping fire extent and burn severity in Alaskan tussock tundra: An analysis of the spectral response of tundra vegetation to wildland fire. Remote Sensing of Environment 134, 194–209. doi:10.1016/j.rse. 2013.03.003 Melchiorre A, Boschetti L (2018) Global analysis of burned area persistence time with MODIS Data. Remote Sensing 10, 750. doi:10.3390/ rs10050750 Mojaradi B, Lucas C, Varshosaz M (2004) Using learning cellular automata for post classification satellite imagery. International www.publish.csiro.au/wf International Journal of Wildland Fire 869 Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives 35, 991–995. Moreno Ruiz JA, Riaño D, Arbelo M, French NHF, Ustin SL, Whiting ML (2012) Burned area mapping time series in Canada (1984–1999) from NOAA-AVHRR LTDR: A comparison with other remote sensing products and fire perimeters. Remote Sensing of Environment 117, 407–414. doi:10.1016/j.rse.2011.10.017 Moreno-Ruiz JA, García-Lázaro JR, del Águila Cano I, Hernández-Leal P (2014a) Burned area mapping in the North American boreal forest using terra-MODIS LTDR (2001–2011): A comparison with the MCD45A1, MCD64A1 and BA GEOLAND-2 products. Remote Sensing 6, 815–840. doi:10.3390/rs6010815 Moreno-Ruiz JA, Garcia-Lazaro JR, Riano D, Kefauver SC (2014b) The synergy of the 0.05° (~5km) AVHRR long-term data record (LTDR) and landsat TM archive to map large fires in the North American boreal region from 1984 to 1998. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 7, 1157–1166. doi:10.1109/JSTARS.2013.2292853 Moreno-Ruiz JA, García-Lázaro JR, Arbelo M, Riaño D (2019) A comparison of burned area time series in the Alaskan boreal forests from different remote sensing products. Forests 10, 363. doi:10.3390/ f10050363 Moreno-Ruiz JA, García-Lázaro JR, Arbelo M, Cantón-Garbín M (2020) MODIS sensor capability to burned area mapping—assessment of performance and improvements provided by the latest standard products in boreal regions. Sensors 20, 5423. doi:10.3390/s20185423 Morisette JT, Baret F, Privette JL, Myneni RB, Nickeson JE, Garrigues S, Shabanov NV, Weiss M, Fernandes RA, Leblanc SG, Kalacska M, Sanchez-Azofeifa GA, Chubey M, Rivard B, Stenberg P, Rautiainen M, Voipio P, Manninen T, Pilant AN, Lewis TE, Iiames JS, Colombo R, Meroni M, Busetto L, Cohen WB, Turner DP, Warner ED, Petersen GW, Seufert G, Cook R (2006) Validation of Global ModerateResolution LAI Products: A Framework Proposed Within the CEOS Land Product Validation Subgroup. IEEE Transactions on Geoscience and Remote Sensing 44, 1804–1817. doi:10.1109/TGRS.2006.872529 Mouillot F, Schultz MG, Yue C, Cadule P, Tansey K, Ciais P, Chuvieco E (2014) Ten years of global burned area products from spaceborne remote sensing – A review: Analysis of user needs and recommendations for future developments. International Journal of Applied Earth Observation and Geoinformation 26, 64–79. doi:10.1016/j.jag.2013. 05.014 Nelson K, Thompson D, Hopkinson C, Petrone R, Chasmer L (2021) Peatland-fire interactions: A review of wildland fire feedbacks and interactions in Canadian boreal peatlands. Science of The Total Environment 769, 145212. doi:10.1016/j.scitotenv.2021.145212 Núñez-Casillas L, García Lázaro JR, Moreno-Ruiz JA, Arbelo M (2013) A comparative analysis of burned area datasets in Canadian boreal forest in 2000. The Scientific World Journal 2013, 289056. doi:10.1155/2013/289056 Otón G, Ramo R, Lizundia-Loiola J, Chuvieco E (2019) Global detection of long-term (1982–2017) burned area with AVHRR-LTDR data. Remote Sensing 11, 2079. doi:10.3390/rs11182079 Otón G, Lizundia-Loiola J, Pettinari ML, Chuvieco E (2021) Development of a consistent global long-term burned area product (1982–2018) based on AVHRR-LTDR data. International Journal of Applied Earth Observation and Geoinformation 103, 102473. doi:10.1016/j.jag.2021.102473 Padilla M, Stehman SV, Ramo R, Corti D, Hantson S, Oliva P, AlonsoCanas I, Bradley AV, Tansey K, Mota B, Pereira JM, Chuvieco E (2015) Comparing the accuracies of remote sensing global burned area products using stratified random sampling and estimation. Remote Sensing of Environment 160, 114–121. doi:10.1016/j.rse.2015.01.005 Pan Y, Birdsey RA, Fang J, Houghton R, Kauppi PE, Kurz WA, Phillips OL, Shvidenko A, Lewis SL, Canadell JG, Ciais P, Jackson RB, Pacala SW, McGuire AD, Piao S, Rautiainen A, Sitch S, Hayes D (2011) A large and persistent carbon sink in the world’s forests. Science 333, 988–993. doi:10.1126/science.1201609 Parisien MA, Peters VS, Wang Y, Little JM, Bosch EM, Stocks BJ (2006) Spatial patterns of forest fires in Canada 1980–1999. International Journal of Wildland Fire 15, 361–374. doi:10.1071/WF06009 Pedelty J, Devadiga S, Masuoka E, Brown M, Pinzon J, Tucker C, Vermote E, Prince S, Nagol J, Justice C, Roy D, Ju J, Schaaf C, Liu J, Privette J, Pinheiro A (2007) Generating a long-term land data record from the AVHRR and MODIS instruments. In ‘International Geoscience and Remote Sensing Symposium (IGARSS)’. pp. 1021–1025. (IEEE) doi:10.1109/IGARSS.2007.4422974 Pinty B, Verstraete MM (1992) GEMI: a non-linear index to monitor global vegetation from satellites. Vegetatio 101, 15–20. doi:10.1007/ BF00031911 Ponomarev EI, Kharuk VI, Ranson KJ (2016) Wildfires dynamics in Siberian larch forests. Forests 7, 125. doi:10.3390/f7060125 Portier J, Gauthier S, Bergeron Y (2019) Spatial distribution of mean fire size and occurrence in eastern Canada: influence of climate, physical environment and lightning strike density. International Journal of Wildland Fire 28, 927–940. doi:10.1071/WF18220 Rogers BM, Soja AJ, Goulden ML, Randerson JT (2015) Influence of tree species on continental differences in boreal fires and climate feedbacks. Nature Geoscience 8, 228–234. doi:10.1038/ngeo2352 Shuman JK, Shugart HH, O’Halloran TL (2011) Sensitivity of Siberian larch forests to climate change. Global Change Biology 17, 2370–2384. doi:10.1111/j.1365-2486.2011.02417.x Sitnov SA, Mokhov II (2018) A Comparative Analysis of the Characteristics of Active Fires in the Boreal Forests of Eurasia and North America Based on Satellite Data. Izvestiya, Atmospheric and Oceanic Physics 54, 966–978. doi:10.1134/S0001433818090347 Skakun R, Whitman E, Little JM, Parisien MA (2021) Area burned adjustments to historical wildland fires in Canada. Environmental Research Letters 16, 064014. doi:10.1088/1748-9326/abfb2c Skakun R, Castilla G, Metsaranta J, Whitman E, Rodrigue S, Little J, Groenewegen K, Coyle M (2022) Extending the National Burned Area Composite Time Series of Wildfires in Canada. Remote Sensing 14, 3050. doi:10.3390/rs14133050 Soja AJ, Cofer WR, Shugart HH, Sukhinin AI, Stackhouse PW, McRae DJ, Conard SG (2004) Estimating fire emissions and disparities in boreal Siberia (1998–2002). Journal of Geophysical Research: Atmospheres 109, D14S06. doi:10.1029/2004JD004570 Soja AJ, Al-Saadi J, Giglio L, Randall D, Kittaka C, Pouliot G, Kordzi JJ, Raffuse S, Pace TG, Pierce TE, Moore T, Roy B, Pierce RB, Szykman JJ (2009) Assessing satellite-based fire data for use in the National Emissions Inventory. Journal of Applied Remote Sensing 3, 031504. doi:10.1117/1.3148859 Stehman SV (1997) Selecting and interpreting measures of thematic classification accuracy. Remote Sensing of Environment 62, 77–89. doi:10.1016/S0034-4257(97)00083-7 Stocks BJ, Mason JA, Todd JB, Bosch EM, Wotton BM, Amiro BD, Flannigan MD, Hirsch KG, Logan KA, Martell DL, Skinner WR (2003) Large forest fires in Canada, 1959–1997. Journal of Geophysical Research 108(D1), 8149. doi:10.1029/2001JD000484 Stowe LL, Davis PA, McClain EP (1999) Scientific Basis and Initial Evaluation of the CLAVR-1 Global Clear/Cloud Classification Algorithm for the Advanced Very High Resolution Radiometer. Journal of Atmospheric and Oceanic Technology 16, 656–681. doi:10.1175/ 1520-0426(1999)016<0656:SBAIEO>2.0.CO;2 Sukhinin AI, French NHF, Kasischke ES, Hewson JH, Soja AJ, Csiszar IA, Hyer EJ, Loboda T, Conrad SG, Romasko VI, Pavlichenko EA, Miskiv SI, Slinkina OA (2004) AVHRR-based mapping of fires in Russia: New products for fire management and carbon cycle studies. Remote Sensing of Environment 93, 546–564. doi:10.1016/j.rse.2004.08.011 Tchebakova NM, Parfenova E, Soja AJ (2009) The effects of climate, permafrost and fire on vegetation change in Siberia in a changing climate. Environmental Research Letters 4, 045013. doi:10.1088/ 1748-9326/4/4/045013 Turetsky MR, Kane ES, Harden JW, Ottmar RD, Manies KL, Hoy E, Kasischke ES (2011) Recent acceleration of biomass burning and carbon losses in Alaskan forests and peatlands. Nature Geoscience 4, 27–31. doi:10.1038/ngeo1027 Vanderhoof MK, Fairaux N, Beal YJG, Hawbaker TJ (2017) Validation of the USGS Landsat burned area essential climate variable (BAECV) across the conterminous United States. Remote Sensing of Environment 198, 393–406. doi:10.1016/j.rse.2017.06.025 Vivchar A (2011) Wildfires in Russia in 2000–2008: Estimates of burnt areas using the satellite MODIS MCD45 data. Remote Sensing Letters 2, 81–90. doi:10.1080/01431161.2010.499138 Walsh JE (2014) Intensified warming of the Arctic: Causes and impacts on middle latitudes. Global and Planetary Change 117, 52–63. doi:10.1016/j.gloplacha.2014.03.003 J-A Moreno-Ruiz et al. International Journal of Wildland Fire 870 Wieder RK, Vitt DH, Benscoter BW (2006) Peatlands and the boreal forest. In ‘Boreal peatland ecosystems’. (Eds RK Wieder, DH Vitt) pp. 1–8. (Springer-Verlag: Berlin, Heidelberg) Wirth C (2005) Fire regime and tree diversity in boreal forests: implications for the carbon cycle Forest Diversity and Function. In ‘Forest Diversity and Function, Ecological Studies. Vol. 176’. (Eds M Scherer-Lorenzen, C Körner, ED Schulze) pp. 309–44. (Springer: Berlin, Heidelberg) Wooster MJ, Zhang YH (2004) Boreal forest fires burn less intensely in Russia than in North America. Geophysical Research Letters 31, L20505. doi:10.1029/2004GL020805 Data availability. The data that support this study are available in the article and accompanying online supplementary material. Conflicts of interest. The authors declare no conflict of interest. Declaration of funding. This research was funded by the Ministerio de Ciencia, Innovación y Universidades (MCIU), the Agencia Estatal de Investigación (AEI) and the Fondo Europeo de Desarrollo Regional (FEDER) through project RTI2018-099171-B-I00. Acknowledgements. The authors wish to thank the LTDR project and its team for making the data available, and Alaska Fire Service, the Canada Centre for Mapping and Earth Observation of Natural Resources Canada, the Canadian Forest Service, NASA, NOAA and USGS for processing and distributing the AFS, CFNDB, NBAC, MCD64A1 and LTDR datasets. Author contributions. J.A.M.-R. and J.R.G.-L. conceived, designed and applied the methodology. All authors obtained, analysed and discussed the results; M.A. and P.A.H.-L. in collaboration with the rest of the authors wrote and contributed to the editing of the manuscript. Author affiliations A Departamento de Informática, Universidad de Almería, 04120 Almería, Spain. Email: [email protected]; [email protected] B Departamento de Física, Universidad de La Laguna, 38200 San Cristóbal de La Laguna, Spain. Email: [email protected]; [email protected] www.publish.csiro.au/wf International Journal of Wildland Fire 871