Measuring fire-induced change in the understorey of an Australian dry sclerophyll forest using remote sensing

摘要

This research investigates the use of remote sensing technologies for measuring and mapping the changes in the forest understorey in response to prescribed burning. Remote sensing has been used extensively to map the burn areas in fire-affected landscapes, but less work has been done focussing on beneath the canopy. Sub-canopy vegetation layers are important for habitat and for understanding the fuel hazards they may pose to the risk of wildfire. Accordingly, instruments and approaches must be able to perform in both pre- and post-burn environments, and be able to provide meaningful measures of change. Wildfires are increasing in intensity and frequency, and in response prescribed burning is used to mitigate threats posed by them. Quantifying post-fire effects is important for burn severity, ecosystem recovery and post-fire hazard assessments. This information will allow land managers and scientists to understand fires in their environmental, economic and social contexts and help formulate responses and policies accordingly. However, measures of fire effects and fuel hazards which are done via visual assessments are known to be subjective and inconsistent between assessors and over time. What is needed is an improvement in the reporting procedures around quantification of fire effects which are both repeatable and quantifiable. In this research two remote sensing technologies were used to measure, map and track changes in the understorey of an Australian dry sclerophyll forest. Terrestrial Laser Scanning (TLS) was used to derive vegetation structure variables and HSR (HyperSpectral Radiometry) was used to derive vegetation physiological variables. The study site was located in St. Andrews, Victoria, Australia within which a control plot and three fire treatment plots were set-up and monitored over a two year period, before and after a prescribed fire event conducted in autumn 2012. The datasets collected were used for statistical and spatial analysis of changes in understorey vegetation, and to assess those metrics best suited for describing different vegetation responses to fire effects. The first part of this research examined the potential of TLS to detect fire-induced change in the forest understorey. From TLS point clouds a total of 18 metrics were extracted which were tested against accuracy and reliability criteria. Three metrics; mean AGHchange (Above Ground Height), median AGHchange and point countchange were shortlisted. To report different post-fire changes in burnt understorey, mean AGHchange metric was used. This metric was able to report fire effects such as total burn area, measures of patchiness, spatial distribution of burnt and unburnt areas, fuel accumulation and prescribed burn efficiency across various temporal scales. The second part of this research analysed hyperspectral data of the near-surface (grass) and surface fuel layer (litter). Spectral changes in the near-surface fuel layer were observed in Visible (550nm), Near-Infrared (680-750nm) and Middle Infrared (970nm, 1220nm, 1550nm) domains of the electromagnetic spectrum. For the surface fuel layer (litter) changes were observed in the Middle Infrared domain (1140nm, 1225nm and 1700nm). The greatest difference from pre-burn levels for both the fuel layers occurred within the first two weeks post-burn. Spectral indices corresponding to the above determined broad spectral bands were tested to ascertain which were best at characterising burnt from unburnt targets whilst also tracking recovery. Indices such as NDVI, NBR and D720 were found to be the most suitable for near-surface fuel layer whilst D1230 for surface fuel layer. A preliminary investigation into comparing the change detected by the two remote sensing technologies suggested that physiological change detected by HSR, recorded vegetation recovery as early as six weeks post-burn. Structural change detected by TLS even after two years post-burn was recorded as being close to two weeks post-burn levels. This finding matched well with visual assessments of structural measures (plant cover and height). The findings of this study suggest that improvements in reporting procedures around quantification of fire effects can be achieved using TLS and HSR technology. TLS-derived structural metric, mean AGHchange can accurately detect quantified measures of fire-induced change in forest understorey that can be validated with field assessments. It can also report post-fire effects at various temporal scales including area burnt, burn patchiness, fuel load accumulation and prescribed burn efficiency. Spectral indices such as NDVI, NBR and D720 were able to accurately detect both vegetation loss and recovery. There is merit in further investigating TLS and HSR in conjunction for quantified and robust reporting of fire effects. The change detected by these technologies can be linked to inform both vegetation recovery and fuel accumulation.