Mapping Global Bamboo Forest Distribution Using Multisource Remote Sensing Data

Abstract

Bamboo forest has great potential in climate change mitigation. However, the spatiotemporal pattern of carbon storage of global bamboo forest is still cannot be accurately estimated, because the lack of an accurate global bamboo forest distribution information. In this paper, the global bamboo forest distribution was mapped with the following steps. To begin with, training samples were obtained based on investigation data, statistic data, and literature data. Then, a decision tree was constructed for mapping the global bamboo forest distribution by integrating Landsat 8 OLI and MODIS data. Finally, the global bamboo forest area was estimated using a pixel unmixing algorithm. The constructed decision tree succeeds in extracting global bamboo forest based on remote sensing data, where the overall accuracy of classification was 78.81%. The estimated total global bamboo forest area was 30538.35 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> ha, with a low root-mean-square error of 611.1 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> ha. The estimated bamboo forest area of each province in China and each country were high consistent with the National Forest Inventory in China and Food and Agriculture Organization of the United Nations statistic results (average R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> > 0.9), respectively. Therefore, the global bamboo forest map yielded a satisfactory accuracy in both classification and area estimation, and could provide accurate and significant support for global bamboo forest resource management and carbon cycle research.