Comparative analysis of land-cover data accuracy and uncertainty in arid land

LUCC research in arid area showed that population increasing, human activities, and climate change have brought great land changes. As a very important basic geographic data, land cover data reflects anthropic activity and climate change on the environment. The ecological environment in arid area experienced serious degradation. In order to resolve the core issues of sustainable development in the arid area, many models seek quantitative simulation of ecology-hydrological processes. Arid area with scarce rainfall and sparse vegetation, the primary land surface characteristics is landscape fragmentation, which generated huge data inaccuracies and uncertainties. This article set the land-cover data acquired by artificial visual classification with QB (0.61 meters) as actual land surface, and compared it with other land-cover data produced through classification tree with LandSat(30m), SPOT(10 m), and QB(2.5 m), four different resolution images in arid land. Results showed that overall accuracy of land cover maps is TM30m 46%, SPOT10m 73, and QB2.5m 81%, respectively. Coarse resolution image take more errors and uncertainty than fine resolution image.