Home > Research > Research Results > Research Results 2020 > Small forest plots cannot be identified from Global Forest Maps

Update:September 16, 2020

Main content starts here.

Small forest plots cannot be identified from Global Forest Maps

Article title

Causal Analysis of Accuracy Obtained Using High-resolution Global Forest Change Data to Identify Forest Loss in Small Forest Plots

Author (affiliation)

Yusuke Yamada (a), Toshihiro Ohkubo (b), Katsuto Shimizu (a)

(a) Department of Forest Management, FFPRI, Tsukuba, Ibaraki, Japan.

(b) Waseda University, Tokorozawa, Saitama, Japan.

Publication Journal

Remote Sensing, 12(15), 2489, August 2020 DOI:10.3390/rs12152489( External link )

Content introduction

In recent years in Japan, illegal logging and the abandonment of reforestation have become problems, so urgent work is being done to establish a method for identifying areas where forests have been cut. In order to identify cut areas in a wide area, forest loss maps have been created from satellite images, and these can identify cut areas in time series at a relatively low cost.

Against this backdrop, the dataset of Global Forest Change (GFC, Hansen et al, Science 342:850-853, 2013:Global Forest Maps) contains forest loss maps that cover the entire globe at a resolution of 30m. They are created from publicly released data that are generally updated once a year. However, in countries like Japan where cut areas and forest ownership tend to be small in scale, the accuracy of identifying cut areas using GFC data has not been evaluated.

Therefore, in the present study the accuracy of the GFC dataset was examined for a certain municipality on the island of Kyushu. As a result, the GFD dataset could find only 11.1% of the 1480 cut areas. Moreover, it became clear that accuracy decreased with decreasing size of cut areas.

From these findings, we could see that the GFC dataset does not provide sufficient accuracy for managing forests in Japan where the size of cut areas is small. Gaining an accurate and effective understanding of forest cutting is essential for developing a robust forestry industry and for sustainably managing forests. The present study indicates that it is necessary to develop a method that can identify small-scale cut areas that includes the use of higher resolution satellite images.


Figure: Examples of forest-loss polygons
Figure:Examples of forest-loss polygons of (a) the reference dataset, and (b) the GFC dataset with the Landsat composite images after harvesting (2017). The GFC dataset could not identify two locations of small-scale cutting in the lower right of the images. (Part of this figure has been modified from the original figure in the paper).