Is the RdNBR a better estimator of wildfire burn severity than the dNBR? A discussion and case study in southeast China

Abstract

A growing number of wildfires around the world have been resulting in severe post-fire effects. Understanding the accurate distribution of the wildfire burn severity is particularly important for post-fire management. The differenced Normalized Burn Ratio (dNBR) and its relative dNBR (RdNBR) are the two main indices for burn severity assessment. In China, little information is available on the capabilities of the dNBR and RdNBR for the burn severity assessment. Moreover, differing results of the detection capabilities of the dNBR and RdNBR for the burn severity have been reported worldwide. This study aimed to assess the potential of the dNBR and RdNBR for the wildfire burn severity estimation in southeast China to influence the post-fire management. Landsat satellite imageries on 15 August 2017 (pre-fire) and 15 August 2019 (post-fire) and field-based burn severity surveys were used to develop effective burn severity models using regression methods. The results showed that the dNBR is a better and more reliable index for burn severity mapping than the RdNBR. Using the RdNBR to estimate burn severity likely leads to an underestimation. RdNBR was designed to normalize for variations in pre-fire vegetation and therefore should most useful in heterogenous landscapes, such as the field area, which has areas of grassland and areas of forest. However, we found the opposite pattern: for predicting burn severity, the dNBR resulted in higher R² and smaller RMSE values than the RdNBR for forested areas alone, and for combined forested and grassland areas (i.e. the entire field area). For grassland areas alone, the dNBR had similar R² and RMSE to that of the RdNBR. In brief, the inconsistent results of the detection capabilities of the dNBR and RdNBR for burn severity around the world are closely related to diverse local conditions. The RdNBR must be used with caution in estimating the burn severity at a specific site. Our study enriches the knowledge of the wildfire burn severity assessment in the southeast coast of China.

Keywords:

• Burn severity
• dNBR
• RdNBR
• linear and quadratic models
• remote sensing

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Science Foundation of China (41807502).