Forest structure parameter extraction using SPOT-7 satellite data by object- and pixel-based classification methods

Area of Science:

• Forestry
• Remote Sensing
• Geospatial Analysis

Background:

• Accurate forest structure mapping is crucial for effective forest management.
• Satellite data offers a scalable solution for large-area forest assessment.
• Hyrcanian forests present unique challenges and opportunities for remote sensing applications.

Purpose of the Study:

• To extract key forest structure parameters (species, density, canopy, gaps) using SPOT-7 satellite data.
• To assess the accuracy of remote sensing methods for forest structural parameter extraction.
• To evaluate the effectiveness of combined pixel-based and object-based approaches.

Main Methods:

• Field inventory of 12 plots (100m x 100m) with detailed tree data and DGPS coordinates.
• Extraction of spectral transformations, vegetation indices, and ratios from SPOT-7 data.
• Application of supervised pixel-based classification (Support Vector Machine) and object-based classification for canopy and gap delineation.

Main Results:

• Forest type classification achieved 95% accuracy (Kappa: 0.8).
• Canopy and gap coverage accuracy was 91% (Kappa: 0.7).
• Tree density estimation was, on average, 47 trees/hectare lower than ground truth.

Conclusions:

• SPOT-7 satellite data, combined with pixel- and object-based methods, enables accurate forest structural parameter extraction.
• The study demonstrates the potential of integrated remote sensing techniques for forest inventory and management.
• Further refinement is needed to improve tree density estimation accuracy.