Supplement 1. A vector file of polygon fragment boundaries used in this study in KML format.

File List CAO_kipuka_boundaries_20131204.kml (MD5: 9cd3a0afdd95797b21e41515140be36d) A polygon vector GIS layer of the fragment boundaries. Description <b>Description</b> These boundaries were computed using utilities packaged with the GDAL library (http://www.gdal.org) under the following methodology: We used gdalwarp and gdal_translate to stack the computed vegetation height and NDVI images onto the same grid at 2.0m resolution. Cubic spline interpolation as used. We used gdal_calc.py create a binary mask of cells meeting the following thresholds: Canopy height &gt; 3.0 and NDVI &gt; 0.7. We used gdal_sieve.py to groups less than 50 cells (0.02ha) with 8-connectedness. The remaining groups were polygonized using the utility gdal_polygonize.py Finally, the boundaries of these groups were rounded slightly using the -simplify flag of the ogr2ogr utility. Tolerance value (maximum distance segment can move when removing a node) was 2.0.

文件列表：CAO_kipuka_boundaries_20131204.kml（MD5值：9cd3a0afdd95797b21e41515140be36d），该数据集为地块边界的多边形矢量地理信息系统（Geographic Information System, GIS）图层。<b>描述</b> 此类边界通过GDAL库（Geospatial Data Abstraction Library, http://www.gdal.org）自带的工具按如下流程计算得到：首先使用gdalwarp与gdal_translate工具将计算所得的植被高度影像与归一化植被指数（Normalized Difference Vegetation Index, NDVI）影像堆叠至分辨率为2.0米的统一网格，采用三次样条插值法完成重采样。随后通过gdal_calc.py生成满足以下阈值条件的二值掩膜：冠层高度＞3.0，且归一化植被指数＞0.7。接着使用gdal_sieve.py移除包含少于50个单元格（对应面积0.02公顷）的8连通分组。剩余分组通过gdal_polygonize.py工具转换为多边形矢量。最终借助ogr2ogr工具的-simplify参数对上述多边形边界进行轻度平滑处理，设置的容差值（移除节点时线段可移动的最大距离）为2.0。