Foliage moisture content estimation from one‐dimensional and two‐dimensional spectroradiometry for fire danger assessment

Abstract

Moisture content in live fuels is a critical variable in fire ignition and fire behavior assessment, but it is very difficult to estimate using traditional methods. Remote sensing measurements provide adequate temporal and spatial coverage to be used operationally in fire danger estimation. Before such applications can be routinely undertaken, reflectance changes caused by water content variations must be clearly separable from other factors in reflectance variation (leaf area index, chlorophyll content, dry matter, observation‐illumination angles, etc.). This paper analyzes the effect of these variables in a controlled experiment using young Holm oak (Quercus Ilex) plants, a common Mediterranean species. Spectro‐radiometric measurements were performed systematically in laboratory conditions using a GER 2600 spectroradiometer (point measurements, one‐dimensional (1‐D)) and a Specim hyperspectral camera system (matrix measurements, 2‐D). Leaf and canopy biophysical parameters were sampled to control their influence on reflectance as water stress increased. Equivalent water thickness (EWT) and fuel moisture content (FMC) were correlated against several variations of a new index: the Normalized Difference Spectral Index, (NDSI(λ1,λ2)). Results indicate that the choice of suitable spectral bands is critical to control plant physiological effects on reflectance. Strong correlations were obtained between NDSI and both EWT and FMC by applying operational bands at 1108 and 1242 nm (NDSI(1108,1242): r > 0.93), while the shorter near infrared bands (1000 and 860 nm) show a stronger influence of external factors. A combination of 2‐D pixel selection techniques and first derivative spectroscopy successfully solved absolute reflectance inconsistencies between 2‐D and 1‐D measurements, which may prove to be a helpful means of scaling up from leaf to canopy level, and from the ground to airborne observation.