Forest tree species identification and its response to spatial scale based on multispectral and multi-resolution...

Area of Science:

• Forestry
• Remote Sensing
• Geospatial Analysis

Background:

• Spatial scale is critical for accurate forest type identification using multi-resolution imagery.
• The impact of texture information on tree species identification accuracy across different spatial resolutions remains unclear.

Purpose of the Study:

• To investigate the influence of spatial resolution and texture information on forest tree species identification.
• To compare classification results from native and up-scaled remote sensing images at multiple spatial scales.

Main Methods:

• Utilized multi-resolution satellite imagery (1-30m) from GF-1, GF-2, and Landsat-8 over Wangyedian forest farm.
• Employed Support Vector Machine (SVM) classifier for identifying five dominant tree species.
• Analyzed the effect of adding texture information to spectral data.

Main Results:

• Overall classification accuracy was highest at 4m resolution and decreased with coarser resolutions (down to 30m).
• Texture information enhanced classification accuracy by 2.0%-3.6% for resolutions from 1m to 8m.
• Accuracy improvements were not observed for 16m and 30m resolution imagery.

Conclusions:

• Spatial resolution is a key factor influencing tree species identification accuracy.
• Texture information is beneficial for high-resolution imagery but not for medium resolutions.
• Using native high-resolution satellite images is crucial for accurate multi-spatial-scale analysis.