Effects of forest cover, terrain, and scale on timber volume estimation with Thematic Mapper data in a rocky mountain site

Abstract

The biophysical basis of the utility of Landsat Thematic Mapper (TM) data to estimate coniferous timber volume was assessed for a mountainous, mixed-conifer species site in southeast British Columbia. Classification of forest inventory polygons with TM data was found to be 78% accurate to within 25% of the inventory volume for stand volumes between 150 m 3 ha −1 and 300 m 3 ha −1, compared to an accuracy of 44% for polygons with stand volumes less than 150 m 3 ha −1. In polygons with timber volumes less than 150 m 3 ha −1, variations in stand density and background reflectance, identified from MEIS-3 meter imagery, impeded volume classification from TM data. Forest ground call samples and corresponding terrain data, derived from a digital terrain model, were classified according to spatially homogeneous TM data, and the resulting classes were used to analyze the sensitivity of TM data to stand characteristics. From correlation and analysis of variance of classes, TM Bands 4 and 5 were found to be strongly inversely related to stand timber volume ( P = 0. 01) with TM Band 5 having the greatest dynamic range. Timber volume was strongly related to the bulk reflectance characteristics of the stand including stand height, age, species composition, slope, and aspect, and could thus be estimated from the remote sensing data. A geometric-optical reflectance simulation was used to estimate the relative contributions of forest cover and terrain to stand reflectance. The effects of terrain on the relationship between TM data and volume were small relative to the effects of forest cover. The radiometric basis of these relationships was discussed in relation to simulations of stand reflectance. For stand volumes of less than 400 m 3 ha −1, simulated stand reflectance in both TM Bands 4 and 5 was sensitive to the proportions of shadow and exposed background in the stand, but had little sensitivity at volumes greater than 400 m 3 ha −1. TM data may be used to estimate timber volume in other sites that possess similar characteristics to the site assessed here.