APPORT DE MODELES NUMERIQUES DE HAUTEUR A L'AMELIORATION DE LA PRECISION D'INVENTAIRES STATISTIQUES FORESTIERS.

Our forest information needs are growing with high requirements in terms of accuracy, spatial resolution and costs. Thus, forest inventories innovation is a constantly evolving field. In this study, we evaluated the contribution of digital height models (DHM) as auxiliary data to improve the accuracy of inventory estimates. Two study areas were selected. The first covers the Nièvre 'department', where a recent campaign of aerial images (winter 2011) was used to generate a DHM. The second covers the Vosges area of the Haut-Rhin 'department', and was described by a LiDAR acquisition (1360 km², in spring 2011) which allowed the production of a state of the art DHM. Other forest information layers such as the main ecological regions, the « sylvoécorégions » as well as maps of forest types produced by IGN were also used as auxiliary variables. The results showed that for two key forest parameters examined (i.e. forest area and volume), the relative efficiency was improved by a factor 1 to 8 using DHM and forest type maps. This efficiency can be interpreted as the factor by which the initial inventory sample should be multiplied to obtain the same precision than when using the relevant auxiliary variables. This efficiency, however, varies depending on the study area. For an area mainly composed of deciduous forests like the Nièvre 'département', winter images did not lead to a quality DHM, making it impossible to quantify the real contribution of photogrammetric DHM. However, in areas of pure conifer stands, the calculated efficiency in volume estimates ranged from 1.2 to 1.7, which is similar to the results obtained in the Vosges, where the LiDAR DHM allowed a gain of 1.8 in efficiency for volume estimates. In general, the auxiliary variables improve the accuracy of inventory estimates, provided they are correlated to the parameters of interest. DHM are thus certainly a possible source of accuracy improvement for forest inventories that must be carefully considered. [ABSTRACT FROM AUTHOR]