Evaluation of Statistical Methods to Estimate Forest Volume in a Mediterranean Region

The use of three estimation methods is investigated for mapping forest volume all over a complex Mediterranean region (Tuscany, Central Italy). The first two methods are based on the processing of satellite images, specifically a summer Landsat TM scene. These methods exploits the remotely sensed information through a non parametric approach (k-NN) and by means of local locally calibrated regressions. The last method considered, kriging, instead uses only the spatial autocorrelation of tree volume relying on geostatistical principles. The experiments performed demonstrated that, at the original sampling density, the three methods produced similar accuracies. This was no more the case when reducing the sampling density to various levels. While in fact this reduction marginally affected the performances of the two remote sensing based methods, it dramatically degraded that of kriging. Additionally, the investigation demonstrated that per-pixel estimates of error variance are obtainable also by k-NN and locally calibrated regression procedures, in analogy with the same property of kriging. Such estimated error variances can be exploited to optimally integrate the outputs of the methods based on remotely sensed data and spatial autocorrelation. In all cases examined the integrated estimation outperformed the single procedures, which indicates an operational strategy for mapping forest attributes also in complex Mediterranean areas.