Prediction of stem volume in complex temperate forest stands using TanDEM-X SAR data

Abstract

Abstract Reliable estimations of stem volume are important for sustainable forest management planning as well as for monitoring of global changes. However, the derivation of stem volume in cubic meters per hectare based on traditional sampling-based forest inventories (usually with a repetition rate of ten years) is very expensive, labor-intensive and only available for the minority of the forest areas worldwide. Thus, spaceborne synthetic aperture radar (SAR) data can provide estimations of forest parameters with sufficient spatial and temporal resolution for large areas. Height information extracted from two interferometric dual-polarized TanDEM-X data sets were used to investigate the potential of polarimetric interferometric X-band SAR data for stem volume estimation in the complex forest stands of the Traunstein forest in Southeast Bavaria, Germany. In contrast to other studies of forest parameter estimation from X-band SAR data carried out in boreal or tropical forest stands, the current study investigated stem volume estimation based on X-band SAR data in complex temperate forest stands. A linear regression model based on the allometric relationship of forest height (estimated from SAR data combined with an airborne LiDAR-based Digital Terrain Model) and stem volume per unit area (deduced from traditional forest inventory) was derived. Moreover, the model was extended and thus improved by integrating novel parameters derived from the co-occurrence matrix as surrogates for horizontal forest structure. This linear regression model predicted stem volume at plot (circular plots of 500 m2) level with a coefficient of determination of R2 = 69% and a root mean square error of RMSE = 155 m3 ha− 1 and stand (areas of 1.5 to 6.4 ha) level with R2 = 94% and RMSE = 44 m3 ha− 1 respectively.