Assessing the potential of the differenced Normalized Burn Ratio (dNBR) for estimating burn severity in eastern Canadian boreal forests Jonathan Boucher A,B,C , Andre ´ Beaudoin B,D , Christian He ´bert B , Luc Guindon B and E ´ ric Bauce A A Universite ´ Laval, 2325 de l’Universite ´, Que ´bec, QC, G1V 0A6, Canada. B Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, 1055 du P.E.P.S., PO Box 10380, Sainte-Foy, Que ´bec, QC, G1V 4C7, Canada. C Present address: Socie ´te ´ de Protection des Fore ˆts contre le Feu, 715, 7e Rue de l’Ae ´roport, Que ´bec, QC, G2G 2S7, Canada. D Corresponding author. Email: andre.beaudoin@canada.ca Abstract. There is considerable variation in the degree of burn severity in boreal fires. One approach that has been used to capture this variation from field and remote sensing perspectives for western Canadian boreal forests is the Composite Burn Index (CBI) and differenced Normalized Burn Ratio (dNBR). Of interest was how well these methods may perform for fires in eastern Canada. This study investigated the CBI-dNBR relationship for selected fires in the eastern boreal forests of Canada, with a view towards contributing to the generalisation of a Canada-wide model. Results for the sampled region showed no difference in the CBI-dNBR relationship between black spruce- and jack pine-dominated stands, whereas this relationship was best described by a Generalised Additive Model (GAM). The dNBR-derived maps would also be useful in support of research and post-fire management in burns outside the studied territory and time frame covered by the existing burn severity mapping system already used in this region. The Saturated growth model proposed for the western boreal region also performed well for our eastern boreal region, thus further supporting the development of a national model. Manuscript received 1 July 2015, accepted 20 September 2016, published online 8 November 2016 Introduction The annual burned area has doubled in North American boreal regions in the last 40 years (Kasischke et al. 2008), which has consequences for a wide range of ecosystem processes, as well as new challenges for land managers (French et al. 2008; Sommers et al. 2014). Burn severity has been identified as a key variable that influences most ecological processes (Weber and Stocks 1998; Volney and Hirsch 2005) and as a result, quanti- fying and mapping this attribute is of prime interest for sus- tainable land management (Lentile et al. 2006). Burn severity is the widely used term for describing the degree to which the ecosystem has changed owing to fire (Lentile et al. 2006). Burn severity quantification and mapping help us to better assess the spatial distribution of damage caused by fire to vegetation and timber (Lindenmayer et al. 2008) and to support salvage logging and site rehabilitation (Lindenmayer et al. 2008; Nappi et al. 2011), as well as to estimate the generated carbon emission (Arnett et al. 2015) and the potential for soil erosion and induced water quality issues (Neary et al. 1999; Wood 2011). Moreover, strong arguments were stated for the need to incorporate burn severity into studies of wildlife response to fire (Smucker et al. 2005; Kotliar et al. 2007). Burn severity has typically been evaluated in situ after fire by measuring soil characteristics, such as char depth, organic matter loss and colour, along with aboveground vegetation consumption, scorch, mortality and recovery (Morgan and Tatar 1972; Ryan and Noste 1985). It has often been partitioned into broad classes ranging from low to high severity for practical purposes (Lentile et al. 2006; Miller and Thode 2007). It is a rather subjective measurement and therefore varies according to the issue addressed, creating confusion in its definition and interpretation (Hutto 2006; Safford et al. 2008; Kolden et al. 2015). Such confusion should be reduced by using a common burn severity mapping method for post-fire impact monitoring and planning (Lentile et al. 2006; Kolden et al. 2015). It is also known that the biophysical symptoms of burn severity vary in a continuous manner, e.g. from stands of living trees with few snags, up to stands with 100% mortality (Kennedy and Fontaine 2009). This has led some authors such as Jain et al. (2004) to recommend estimating and mapping burn severity as a continu- ous variable rather than a categorical one. Acquiring in situ burn severity data is often limited by the lack of accessibility and the high costs of surveying large areas. These constraints have necessitated the development of remote CSIRO PUBLISHING International Journal of Wildland Fire 2017, 26, 32–45 http://dx.doi.org/10.1071/WF15122 Journal compilation Ó IAWF 2017 www.publish.csiro.au/journals/ijwf