Multi-resolution spectral analysis of wildfire potassium emission signatures using laboratory, airborne and spaceborne remote sensing

Abstract

Thermal remote sensing studies of actively burning wildfires are usually based on the detection of Planckian energy emissions in the MIR (3–5 μm), LWIR (8–14 μm) and/or SWIR (1.0–2.5 μm) spectral regions. However, vegetation also contains a series of trace elements which present unique narrowband spectral emission lines in the visible and near infrared wavelength range when the biomass is heated to high temperatures during the process of flaming combustion. These spectral lines can be discriminated by detector systems that are less costly than the longer wavelength, actively cooled instruments more typically used in EO-based active fire studies. The main trace element resulting in the appearance of spectral emission lines appears to be potassium (K), with features at 766.5 nm and 769.9 nm. Here we study K-emission line spectral signature in laboratory scale fires using a field spectrometer, at a series of moderately-sized woodland and shrubland fires using airborne imagery from a new compact hyperspectral imager (HYPER–SIM.GA) operating at a relatively fine spectral sampling interval (1.2 nm), and at large open wildfires using the EO-1 satellite's Hyperion sensor. We derive a metric based on band differencing of the spectral signal both close to and outside of the K-line region in order to quantify the magnitude of the K-emission signature, and find that variations in this metric appear to track quite well with the commonly used measures of fire radiometric temperature and fire radiative power (FRP). We find that substantial flaming activity is required to generate a potassium emission signature, but that once present this can be detected using airborne remote sensing even through a substantial smoke layer that apparently obscures fire across the remainder of the VIS spectral range. Being specific to flaming combustion, detection of the K-emission line signature could prove useful in refining estimates of the gases released in open wildfires, since trace gas emission factors can vary substantially between flaming and smouldering stages. Finally, we demonstrate the first identification of the K-emission line signature from space using the EO-1 Hyperion instrument, but find it detectable only in certain instances. We conclude that a finer spectral and spatial resolution than that offered by Hyperion is required for improved detection performance. Nevertheless, our results point to the potential effectiveness of airborne and spaceborne K-emission signature detection as a complement to the more common thermal remote sensing approaches to wildfire detection and analysis. Sensors targeting this application should consider careful placement of the measurement wavelengths around the location of the K-line wavelengths, in part to minimise influences from the nearby oxygen A-band features.