Estimation of forest structural information using RapidEye satellite data

Abstract

Forest management plans in Bavaria are generally updated only once every 10 years. However, the increasingly dynamic forest structure due to climatic changes requires more frequent data collection in order to maintain up-to-date information. This study explored the use of RapidEye satellite data to provide more frequent updates to the information database. Forest structural information such as quadratic mean diameter (dq), basal area (BA), stem number (SN) and volume (V) were estimated using multi-seasonal analysis of three RapidEye datasets from 2009. Spectral indices and textural metrics provided additional image feature layers. Forest inventory plots were stratified based on the forest type. A correlation analysis was conducted between terrestrial inventory data and that derived from RapidEye data. A cross-validated stepwise forward regression analysis was performed for each forest type. The coefficient of determination and relative root mean square error (rRMSE) showed that stratification improved the regression models, which obtained determination measures ranging from 0.37 to 0.63 and rRMSE ranging from 25 to 131 per cent. Biases of the regression estimates were small, hence the results obtained from applying the models were of an acceptable level of accuracy. The analysis confirmed the potential of RapidEye data to support forest management.