Forest stand parameters from SPOT-XS data by use of artificial neural networks

Oztopal A., Coskun H.

21st Annual Symposium of the European-Association-of-Remote-Sensing-Laboratories, Paris, France, 14 - 16 May 2001, pp.185-189 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • City: Paris
  • Country: France
  • Page Numbers: pp.185-189
  • Istanbul Technical University Affiliated: No


This study reports on Artificial Neural Networks (ANN) modelling for the forest parameters and the reflectance values SPOT-XS satellite data on the ground control points in the study area. The study area is chosen as dense forest cover in the west of Istanbul. The city is negatively affected by the residential areas as a result of population growth and industrial development. This situation is assessed by satellite digital data and forest parameter measurements, such as the number of trees, basal area, volume, mean diameter, and the stand density of forest. In order to find possible relationships between remote sensing variables such as the reflectance values and forest parameters, an Artificial Neural Networks (ANN) approach is used for refined determination of regression coefficients. Adaptive parameter estimation in the ANN provides efficient computation. These observed reflectances show strong relationships with the all chosen forest parameters. The necessary values are provided in single pixel and average values for each band at the control point in the study area. As the reflectances (in the forest area) in the longer red, near and middle IR increase faster than the reflectances in shorter blue and green wavelengths, forest parameters become positively related to reflectance.