Mapping fuel moisture content in upland vegetation using airborne hyperspectral imagery

This paper tests the application of airborne hyperspectral image data for estimating live vegetation fuel moisture content (FMC) in a Calluna vulgaris-dominated semi-natural upland area in the United Kingdom. Airborne hyperspectral imagery was collected over a north/south flight line covering the study site in May and July 2008. Ground data on live FMC were collected concurrently with the flights for ten study plots. Radiance values for the study plots were extracted from the airborne imagery and calibrated to reflectance using spectral measurements from reference targets measured on the ground at the time of the overflights. First derivatives, and a number of vegetation indices (VI), were calculated and correlated with field measured live FMC collected at the study plots. Vegetation FMC maps were produced for the study site for both dates. The results showed that live FMC exhibited spatial and temporal variations that affect the spectral reflectance measured by the airborne hyperspectral instrument, particularly in the near infrared and shortwave infrared regions. Using the first derivative and specific VI improved the correlation between the hyperspectral data and live FMC, but the simple two-wavelength Moisture Stress Index, based on measurements in the near infrared and shortwave infrared, was shown to be effective for FMC estimation. Live FMC was estimated with a root mean square error of 16.8% for all vegetation plots and 10.0% when considering plots composed only of C. vulgaris. The results point to the prospect of FMC mapping for improved modelling of fire risk in UK uplands using remotely sensed data.

► Vegetation fuel moisture content is estimated using airborne remote sensing data.
► Ratio-based vegetation indices are strongly correlated with fuel moisture content.
► Fuel moisture content in C. vulgaris stands is estimated with a RMSE of 10%.
► The results signpost the potential of remote sensing as input to fire risk models.