Hyperspectral remote sensing and forest mapping, inventory and classification

Accurately mapping the composition of forest communities to different cover categories is crucial for using models with specific parameter configurations to simulate the ecological processes corresponding to each cover category. The structural characteristics of forests and their successional stages are key factors in the construction of forest gap models and the assessment of ecosystem function changes. For example, regenerated tropical forests are considered a significant sink for atmospheric carbon dioxide.

Figure 1 Forest monitoring

Traditional forest mapping typically relies on spectral pattern recognition methods, such as supervised classification methods based on maximum likelihood discrimination theory. Remote sensing spectral data is widely used for forest cover type identification, such as distinguishing between broad cover types like coniferous and broadleaf forests. Major broadband remote sensing instruments used for land cover identification include the Landsat Thematic Mapper (TM), Multispectral Scanner (MSS), and the SPOT High Resolution Visible Spectrometer (HRV). Many studies have utilized these tools to classify forest types at higher species resolution, with varying degrees of success.

In related studies, researchers proposed a set of key spectral bands based on leaf data used to assess forest stand biochemical characteristics. These bands showed statistically significant results in correcting AVIRIS hyperspectral data and estimating leaf nitrogen content and lignin concentration in Harvard Forest and Blackhawk Island plots. The results indicate that hyperspectral data is superior to multispectral data in terms of information content and is more suitable for refined and accurate classification of forest stands.

Figure 2. Average reflectance spectra of six types of conifers (Buddenbaum et al. 2005)

Figure 3. Forest classification map of Ser Nature Park