Forest Vegetation Classification and Biomass Estimation Based on Landsat TM Data in a Mountainous Region of West Japan

Abstract

Landsat Thematic Mapper (TM) data corrected topographically with the aid of digital terrain data were applied to the classification and mapping of forest vegetation and the estimation of its biomass in a mountainous region of Hiroshima Prefecture in west Japan. Topographic correction was made based on the relationship between cos r (the solar incidence angle relative to the local terrain slope) and practical radiance response (digital number of TM). The forest vegetation was well classified into three forest types, including deciduous broadleaf forest (dominated by Quercus serrata, Castanea crenata and Carpinus laxiflora), pine forest(Pinus densiflora) and Japanese cedar plantation (Cryptomeria japonica with cypress, Chamaecyparis obtusa), and into two non-forest types, clearcut area and cultivated land area, using the decision tree classification method based on corrected TM data. Classification accuracies for each vegetation type increased by 8-11% when corrected TM data were used instead of uncorrected data. Four vegetation indices were evaluated. Linear relationships were observed between two vegetation indices and forest biomass. However, the coefficient values of these relationships were not identical among the vegetation types. The correlation coefficient (r) between the Normalized Difference Vegetation Index (NDVI) and biomass for the pine forest was 0.85; correlation coefficients between the Differential Vegetation Index (DVI, Band 5-Band 7) and biomass for the Japanese cedar plantation and the deciduous broadleaf forest were -0.83 and 0.80, respectively. Based on the linear relationships, above-ground biomass for all vegetation types was estimated and mapped. Mean biomass for the pine, Japanese cedar and deciduous broadleaf forests was estimated to be about 143, 135 and 121 t ha-1, respectively, and the mean and total biomass of forest vegetation within the study area (2040 ha) were estimated to be about 133 t ha-1 and 275.0x103 t, respectively.