Spatial and temporal variation of forest biomass based on remote sensing and geostatistics

The Daxing’an Mountain, located in Heilongjiang Province of northeastern China, is the largest natural forest area of China. It is of great significance to estimate its forest biomass and further study the characteristics of spatial pattern of forest biomass regionally. Based on the data of TM remote sensing images from Landsat5 series during late 20th century to 2010s, we constructed a remote sensing information model in order to estimate the forest biomass of the Daxing’an Mountain. Morans I coefficient from the direction of 0°, 45°, 90°, 135° and overall direction, and five parameters of semivariagram (namely, nugget value, sill value, variation value, ratio of nugget to sill value, ratio of arch rise to sill) were selected to analyze the heterogeneity and spatial self-correlation of forest biomass separately during three periods of 1994--1997, 2006--2007 and 2010--2011. Results showed that: the integrity of forest biomass in the three periods was good and all continuously distributed without any broken observation. The semivariagram models of biomass of different periods were linear with C/(C0 + C) approaching to 1. Using remote sensing inversion method to estimate forest biomass could provide reliable data source for geostatistics with an estimation precision higher than 75%. Meanwhile, heterogeneity and self-spatial correlation analysis of forest biomass according to geostatistics is a useful supplement to spatial relations of forest biomass by simply using GIS tool. Therefore, combination of remote sensing and geostatistics is confirmed to perform better in spatial-temporal variation analysis of forest biomass.