Abstract
We propose a new validation method for land cover maps. Land cover maps are used in the numerical models that estimate ecosystem behavior (such as carbon budget), water cycle (such as river runoff), and climate at the global scale. Because of this wide range of applications, accurate validation of these maps is of crucial importance. Currently, each of the existing land cover maps has been validated with its own validation method, but there is no validation method for land cover maps by using ground-truth with fairly uniform and accurate worldwide distribution. We propose a validation method that can address this shortcoming. Our method employs information gathered by “the Degree Confluence Project (DCP), ” a voluntary-based project that collects on site data from each of the degree confluence points (DCPoints) in the world. DCPoints are located at the intersections of integer level latitude and longitude grid lines. Volunteers with the project visit the DCPoints and collect data in the form of GPS readings, pictures and descriptions of the landscape. We propose to validate land cover maps using this DCP data. We choose Thailand as our test area and reclassified 30 DCPoints into the six landscape categories (classes) defined under the Kyoto Protocol's Land Use, Land Use Change and Forestry (LULUCF) guidelines. These DCP derived classifications were then compared to classifications derived from Landsat Thematic Mapper images. Through this method we were able to obtain validation information superior to that of visual interpretation. We also converted land cover classifications from Global Land Cover 2000 (GLC2000), MODIS Land Cover Dataset (MDO12), Global Land Cover Characterization Version 2.0 (USGS) and AARS Global 4-minute Land Cover Data Set (AARS) into the LULUCF framework. These four land cover maps were then compared in the same way with the DCP data to evaluate their agreements. Agreement of the GLC2000 is 76%, MOD12 is 70%, USGS is 73.3%, and AARS is 33.3%.
