Regional estimation of woodland moisture content by inverting Radiative Transfer Models

We inverted the PROSPECT and GEOSAIL Radiative Transfer Models (RTM) using Moderate Resolution Imaging Spectrometer (MODIS) data to retrieve Live Fuel Moisture Content (LFMC) in woodlands located in the peninsular territory of Spain. Ecological rules were used to parameterize the RTM. This approach reduces the probability of an ill-posed problem in the inversion of the selected RTMs, by rejecting unrealistic combinations of input parameters. Three species representatives of each region were used to derive the ecological rules: Quercus ilex L., Quercus faginea L., and Pinus halepensis Mill. for the Mediterranean region, and Fagus sylvatica L., Quercus robur L. and Eucalyptus globulus Labill for the Eurosiberian region. Equivalent Water Thickness, Dry Matter and Chlorophyll content were taken from several data sources to separately parameterize both the Mediterranean (water-limited) and Eurosiberian (energy-limited) ecoregions of Spain. GEOSAIL was parameterized using a restricted range of Leaf Area Index (LAI) and specific canopy cover values, keeping other parameters fixed. The inversion was based on the Look Up Table technique using the minimum spectral angle as merit function. Several models were tested by using different inputs from standard MODIS products, as well as the fractional cover product developed by Guerschman et al. (2009). The model based on the reflectance bands and the Normalized Difference Infrared Index computed from the Nadir Bidirectional Reflectance Distribution Function-Adjusted Reflectance product (MCD43A4) provided the most accurate results, with a LFMC's Root Mean Square Error (RMSE) of 27.7% (RMSE = 27.3% for the Mediterranean and 28.7% for the Eurosiberian woodland). The estimation of LFMC was performed within the framework of a fire risk assessment system.

► PROSPECT and GEOSAIL were used to derive LFMC in woodlands from MODIS products. ► Ecological criteria were used in the parameterization of the RTM. ► The MOD09, MCD43, MOD44B and the MOD15 products were considered in the inversion. ► Slopes close to 1 and determination coefficients of 0.5 were obtained from the MCD43. ► The LFMC product can be integrated with other variables in a fire risk index.