Application of a forest reflectance model in estimating leaf area index of Scots pine stands using Landsat-7 ETM reflectance data

We tested the Kuusk-Nilson forest reflectance model in simulating the relationship between leaf area index (LAI) and spectral reflectances of 23 pure Scots pine plots located in southeastern Finland. Landsat-7 enhanced thematic mapper (ETM) reflectance data for channels ETM 3 (red spectral band) and ETM 4 (near-infrared (NIR) band) were used for the analysis. LAI was determined by inversion from canopy gap fraction data measured by the LAI-2000 plant canopy analyzer, using a theoretical forest gap fraction model parameterised for the plots. An independent estimate of LAI was obtained from data on plotwise tree variables using an allometry-based regression developed for Scots pine. The Kuusk-Nilson model can be run in direct and inverse modes. First, the model was applied in direct mode, in which LAI and stand data were used to simulate reflectances in the red (661 nm) and NIR (838 nm) bands. The model-simulated reflectances in the red band were within approximately the same range as the ETM 3 reflectances. In the NIR band, though, a majority of the simulated values were higher than the ETM 4 values. Empirical and simulated relationships between LAI and reflectance in the red and NIR bands showed similar, weakly negative trends in reflectance with increasing LAI. Second, the model was run in inverse mode, i.e., Landsat ETM reflectance data and stand parameters were used for LAI retrieval. The inversion requires as input an initial estimate of LAI and the uncertainty of it. We used as input the allometric estimates of LAI, with two levels of uncertainty (20 and 50%). As the uncertainty increased, correlation between retrieved and initial LAI decreased.