Combined estimation of forest parameters from high- and medium resolution satellite imagery and forest inventory data

Pixel-wise estimation of forest parameters from satellite imagery and national forest inventory plots has been demonstrated to provide useful information on the landscape level for forestry- and environmental authorities in the Nordic countries. However, The accuracy at stand level has not been sufficient for operational forest management. In this study it is demonstrated how the estimation accuracy at stand level can be improved by supplementing spectral signatures from medium resolution satellite imagery such as SPOT or Landsat with textural features derived from high resolution imagery (IKONOS) and /or forest inventory records derived by photo interpretation. The method is based on an artificial neural network that is designed to automatically adapt to the specific combination of information sources available in each case.