Characterizing continuous urban growth using composited time-series Landsat data
Wednesday, 17 December 2014
Impervious surfaces are land cover features through which water cannot penetrate into the soil. As an indicator of urban land use, impervious surface cover (ISC) is disproportionally important to human beings−although covering only 0.5% of the Earth’s terrestrial surface, cities support over 50% the Earth’s population. The increasing demand for built-up space by a growing urban population has been driving land use change in urban areas worldwide. An increase in ISC can significantly impact the biophysical characteristics of land surface, such as altering the local surface energy balance, or transforming regional hydrological systems. Remotely sensed data is commonly used as the primary data source for extracting impervious surface information for monitoring urban growth, but current studies often lack the sufficient temporal resolution or thematic detail to reveal the long-term, nonlinear development of impervious surfaces over time. In a previous study (Sexton et al. 2013), we created an annual stack of 30-m percent ISC estimates for the Washington DC-Baltimore metropolitan region from 1984 to 2010 by compositing all available Landsat images in the USGS archive. Here we developed a robust time-series method to detect impervious surface change. The method employs a customized logistic function for every pixel to model the continuous process of urban growth. It quantifies the fractional intensity of ISC change at the sub-pixel level and also characterizes the timing and length (in years) of urban development. The new method detects change based on a sequence of observations before, during and after change and thus is highly resistant to random noises. Our results showed that the DC-Baltimore metropolitan region experienced an accelerated growth pathway from the late 1980s to the late 2000s. The majority of urban and sub-urban development occurred at scales finer than the Landsat resolution (30 m), with a region-wide mean intensity of 46% ISC increase. Our study demonstrates the value of the long-term and fine temporal resolution data offered by the Landsat archive, and also highlights the possible limitations of Landsat’s spatial resolution in characterizing continuous urban development.