1 Mean height and variability of height derived from Lidar data and

The mean height and standard deviation of the height of the forest canopy, derived from lidar data show to be important variables to summarize forest structure. However lidar data has a limited spatial extent and very high economic cost. Landsat data provide useful structural information in the horizontal plane and have easy access. The integration of both data sources is an interesting goal for sustainable forest management. Different spectral indices (NDVI and Tasseled Cap) were obtained from 3 Landsat scenes (March 2000, June 2001 and September 2001). In addition, mean and standard deviation of lidar height werecalculated in 30x30m blocks. Correlation and forward stepwise regression analysis was applied between these two variables sets. Best correlation coefficients are achieved among mean lidar height versus NDVI and wetness for the three dates (range between 0.65 to-0.73). Others authors indicate that wetness is one of the best spectral indices to characterize forest structure. Best regression models include NDVI and wetness of June and September as dependent variables (adjusted r2: 0.55 – 0.62). These results show that lidar data can be useful for training Landsat to map forest structure but it should be interesting to optimize this approach