Radar-derived standwise forest inventory

The application of remote sensing methods in the estimation of forest stand characteristics, especially biomass and stem volume, has been intensively investigated during the last few years. The new methods, however, have not been accurate enough for operational standwise inventory with a required accuracy typically of 15% for main stand characteristics (stem volume, basal area, and mean height). The present work demonstrates the feasibility of a nonimaging helicopter-borne ranging scatterometer for standwise forest inventory. The radar-derived stand profiles were compared with the standwise field inventory data by applying multivariate data analysis methods. The 1300 ha Teijo test site, locating 130 km west of Helsinki, was divided into 18 parallel radar flight lines with a 150 m spacing. A total of 28 radar variables, including profile information and ground and crown backscatter contributions at 5.4 and 9.8 GHz (polarizations VV, HV, and HH), were used in regression model development. The capability of a ranging radar to classify development class, land use class, bog type and fertility (site) class was demonstrated for the first time. The accuracy of the radar-derived estimates for mean height was 1.6 m (13%) meeting the requirement of operational use. The obtained stem volume accuracy of 31 m/sup 3//ha (26%) was slightly better than has been obtained by aerial photographs. The accuracy of stem volume estimation could be easily improved by decreasing the space between flight lines. However, this leads to considerable increase in flight costs, and, therefore, a scanning ranging radar capable of producing three-dimensional (3D) images of forests would be a better alternative.