Automated Quality Assurance and Quality Control (QA/QC) in Developing Decadal Global Lake Dynamics Products using Landsat-7 and 8

Abstract:

Inland lakes, important water resources, play a crucial role in the global water cycle and are sensitive to climate change and human activities. There clearly is a pressing need to understand temporal and spatial variations of lakes at global and continental scales. To inventory dynamics of global lakes is rather challenging due to their high abundance and low accessibility. With its broad spatial coverage and monitoring capability, satellite remote sensing is the only feasible approach to inventory global lake dynamics, but requires tens of thousands of high-resolution satellite images, automated mapping algorithms, and more importantly tedious yet essential quality assurance and quality control (QA/QC) procedures. Millions of lakes have been mapped out using over 20 thousands Landsat images acquired at lake-stable seasons. Even though a set of highly replicable and automated lake mapping algorithms and tools have been developed, commission and omission errors still exist and some lake boundaries may not be adequately delineated. These errors need to be identified and fixed through intensive QA/QC processes. However, QA/QC of such a huge quantity of lakes remains a great challenge, and the currently available lake datasets from remote sensing were not produced through a rigorous QA/QC process. We have developed two QA/QC strategies with automation. Automated QA requires mapping the Earth twice in the same seasons for lakes and identifies “inconsistent” lakes for further QA/QC. Other lakes without significant changes are considered quality assured, and labor-intensive QA/QC is only limited to those “inconsistent” lakes. We have also developed semi-automated QC tools to further reduce the workload for manpower, and have produced a high-resolution systematically-generated circa-2000 global lake database with adequate QA/QC using Landsat-7. The recent operation of Landsat 8 extends the unprecedented Landsat record to over 40 years, allowing long-term, large-scale lake dynamics mapping. A circa-2015 dataset of lake extents and distributions is being produced using multi-temporal Landsat-8 images acquired in lake-stable seasons. As such, seasonal and inter-annual variability is addressed in the circa-2015 product, significantly reducing the QA/QC workload.