A New Method for the Spatialization of Forest Cover by Fusing Forest Inventory and MODIS Data

Abstract:

The acquisition of accurate spatial and temporal data on forest cover is the foundation for the sustainable management and utilization of forest resources. Although forest inventory data can provide accurate statistical information about forest type, such data do not give the specific spatial distribution. Remote sensing data provide accurate spatial information, and vegetation indices provide measures of land surface vegetation cover and growth conditions. By fusing these two sources of data, specific information about the spatial distribution of different types of forest can be obtained. Here, in a case study of Heilongjiang Province, we obtained forest dominant species area from the sixth and seventh national forest inventories and MODIS composite remote sensing data for the same periods to study forest cover by developing a spatialization method. Based on pixel features (such as NDVI and near-infrared reflectance) and their relationships with forest types, thresholds between different forest types in the remote sensing information were set according to the statistical data, which allowed the two sets of data to be fused. As a result, we generated forest cover maps for 2000 and 2005 that show the distribution of four forest types. Taking vegetation map of China as reference data, an error matrix analysis shows that the overall classification consistency reaches 76.7%, but only 70% for evergreen needleleaf forest and mixed forest. This study paves the way for further research on improving the accuracy of forest cover classification accuracy, on expanding the spatial and temporal scales of interest, and on quantifying forest dynamics