Illumination effects on the differenced Normalized Burn Ratio's optimality for assessing fire severity

Abstract

The influence of illumination effects on the optimality of the dNBR (differenced Normalized Burn Ratio) was evaluated for the case of the 2007 Peloponnese (Greece) wildfires using a pre/post-fire Landsat TM (Thematic Mapper) image couple. Well-illuminated pixels (south and south-east facing slopes) exhibited more optimal displacements in the bi-spectral feature space than more shaded pixels (north and north-west exposed slopes). Moreover, pixels experiencing a small image-to-image difference in illumination obtained a higher optimality than pixels with a relatively large difference in illumination. To correct for illumination effects, the c-correction method and a modified c-correction technique were applied. The resulting median dNBR optimality of uncorrected, c-corrected and modified c-correction data was respectively 0.58, 0.60 and 0.71 (differences significant for p < 0.001). The original c-correction method improved the optimality of badly illuminated pixels while deteriorating the optimality of well-illuminated pixels. In contrast, the modified c-correction technique improved the optimality of all the pixels while retaining the prime characteristic of topographic correction techniques, i.e. detrending the illumination–reflectance relationship. For a minority of the data, for shaded pixels and/or pixels with a high image-to-image difference in illumination, the original c-correction outperformed the modified c-correction technique. In this study conducted in rugged terrain and with a bi-temporal image acquisition scheme that deviated up to two months from the ideal anniversary date scheme the modified c-correction technique resulted in a more reliable change detection.

Introduction

Wildfires play an important role in many ecosystems (Dwyer et al., 1999, Pausas, 2004, Riano et al., 2007) as they partially or completely remove the vegetation layer and affect post-fire vegetation composition (Epting and Verbyla, 2005, Lentile et al., 2005, Telesca and Lasaponara, 2006). The fire impact can be described as (i) the amount of damage (Hammill and Bradstock, 2006, Gonzalez-Alonso et al., 2007, Chafer, 2008), (ii) the physical, chemical and biological changes (Landmann, 2003, Chafer et al., 2004, Cocke et al., 2005, Stow et al., 2007, Lee et al., 2008) or (iii) the degree of alteration (Brewer et al., 2005, Eidenshink et al., 2007) that fire causes to an ecosystem and is quantified as the severity of fire. In this context, the terms fire and burn severity are often interchangeably used. Fire severity estimates the short-term fire effect in the immediate post-fire environment (Lentile et al., 2006). An immediate post-fire assessment minimizes the interference with ecosystem's response processes (such as vegetation recovery) and it is best suited for emergency rehabilitation plans (Key and Benson, 2005, Eidenshink et al., 2007). Burn severity, on the other hand, quantifies both short- and long-term impacts as it includes response processes (Key and Benson, 2005), although the distinction between fire and burn severity has recently become subject for discussion (Keeley, 2009). In this study we focus on fire severity.

Even though a considerable amount of remote sensing studies have focused on the use of the Normalized Difference Vegetation Index (NDVI) for assessing fire severity (Isaev et al., 2002, Diaz-Delgado et al., 2003, Ruiz-Gallardo et al., 2004, Chafer et al., 2004, Hammill and Bradstock, 2006, Hudak et al., 2007), the Normalized Burn Ratio (NBR) has become accepted as the standard spectral index to estimate the severity of fire (e.g. Lopez-Garcia and Caselles, 1991, Epting et al., 2005, Key and Benson, 2005, Bisson et al., 2008). The NBR relates to vegetation structure and moisture by combining near infrared (NIR) and mid infrared (MIR) reflectance and is defined as:

$$\text{NBR}=\frac{\text{TM}4-\text{TM}7}{\text{TM}4+\text{TM}7}$$

where TM4 and TM7 are respectively the NIR and MIR reflectance of Landsat Thematic Mapper (TM) imagery. Since fire effects on vegetation produce a reflectance increase in the MIR spectral region and a NIR reflectance drop (Pereira et al., 1999), bi-temporal image differencing is frequently applied on pre- and post-fire NBR images resulting in the differenced Normalized Burn Ratio (dNBR) (Key and Benson, 2005). Apart from the correlation with field data (Key and Benson, 2005, De Santis and Chuvieco, 2009), the performance of bi-spectral indices can be evaluated by assessing a pixel's shift in the bi-spectral feature space. As such, a pixel-based optimality measure, originating from the spectral index theory (Verstraete and Pinty, 1996), has been developed by Roy et al. (2006). They used the optimality concept to question the dNBR method as an optimal fire severity approach. As the optimality approach is pixel-based, it does not suffer from field sampling constraints and, as a consequence, can be applied on all burned pixels. The optimality value varies between zero (not at all optimal) and one (fully optimal). An optimal fire severity spectral index needs to be very sensitive to fire-induced vegetation changes and insensitive to perturbing factors such as atmospheric and illumination effects.

These illumination effects are initiated by both topography and solar position at the moment of image acquisition and influence an object's reflectance behavior. Differences in solar illumination due to topographic position cause a high variation in reflectance response for similar terrain features: well-illuminated areas show higher reflectance values than expected, whereas in shaded areas reflectance is typically lower (Leprieur et al., 1988). Topographic effects consequently become more obvious in more rugged terrain. They influence any image processing technique based on individual band reflectance, such as Land Use Land Cover (LULC) classifications (Bishop and Colby, 2002, Riano et al., 2003, Mitri and Gitas, 2004). Hence, a range of topographic normalization techniques have been developed with the prime goal to detrend the illumination–reflectance relationship (Teillet et al., 1982, Civco, 1989, Vincini and Frazzi, 2003, Kobayashi and Sanga-Ngoie, 2008). Topographic effects in ratio-images based analysis, however, are assumed to be minimal (Song and Woodcock, 2003) and are not considered in most studies using the NBR to assess fire severity. Key (2005) stated that poor illumination and increased shadow decreases the definition of fire effects and sharpness of dNBR images. Verbyla et al. (2008) showed that topography clearly affects both NBR and dNBR values. They simulated the incoming solar radiation and found a decreasing trend in post-fire NBR while insolation increased, whereas NIR reflectance, MIR reflectance and dNBR values increased with increasing insolation. Therefore, the focus of this research is to quantitatively evaluate illumination effects due to topography using the pixel-based optimality measure and to propose a topographic correction approach to become a more reliable fire severity assessment using the dNBR. This general objective is fulfilled by (i) disclosing the effect of illumination conditions on index performance in topographically uncorrected images, (ii) evaluating the effect of topographic correction on individual band reflectance, and (iii) evaluating the effect of topographic correction on the dNBR optimality.