Abstract: Accurately estimating stock volume of plantation forests based on remote sensing images is important for regional forest management. In this study, stepwise multiple regression models were developed to describe the relationships between different remote sensing datasets \[Advanced Land Observing Satellite (ALOS) Phased Array Lband SAR (PALSAR) normalized backscatter data, Landsat-8 Operational Land Imager (OLI) data, Lband ALOS PALSAR data and Landsat-8 OLI data\] and fieldbased stock volume in a Larix principisrupprechtii plantation forest landscape (Saihanba Forest Farm) in North China, and an approach to estimating stock volume of plantation forests was provided in this study. The regression models were assessed based on the reserved samples and the optimal model was selected to estimate the distribution of stock volume of larch plantation. Our results showed that the exponential model regressed by HV normalized backscatter of PALSAR had the highest estimation accuracy (R²=0.67, the relative RMSE=26.78%, P<0.01), and the total stock volume was estimated to be 4.7×10⁶ m³ by using this model. A saturation effect of HV normalized backscatter was observed when the stock volume was greater than 250-300 m³·hm^-2. The stock volume of larch plantation forests increased with stand ages; the stock volumes were 78, 97, 136 and 127 m³·hm^-2 for <20, 20-30, 30-40 and >40 years old stands, respectively.

Key words: sap flux, Quercus liaotungensis, sapwood area, loess hilly region, soil moisture