A national dataset of 30 m annual urban extent dynamics (1985–2015) in the conterminous United States

Abstract

Abstract. Dynamics of the urban extent at fine spatial and temporal resolutions over large areas are crucial for developing urban growth models and achieving sustainable development goals. However, there are limited practices of mapping urban dynamics with these two merits combined. In this study, we proposed a new method to map urban dynamics from Landsat time series data using the Google Earth Engine (GEE) platform and developed a national dataset of annual urban extent (1985–2015) at a fine spatial resolution (30 m) in the conterminous United States (US). First, we derived the change information of urbanized years in four periods that were determined from the National Land Cover Database (NLCD), using a temporal segmentation approach. Then, we classified urban extents in the beginning (1985) and ending (2015) years at the cluster level through the implementation of a change vector analysis (CVA)-based approach. We also developed a hierarchical strategy to apply the CVA-based approach due to the spatially explicit urban sprawl over large areas. The overall accuracy of mapped urbanized years is around 90 % with the 1-year tolerance strategy. The mapped urbanized areas in the beginning and ending years are reliable, with overall accuracies of 96 % and 88 %, respectively. Our results reveal that the total urban area increased by about 20 % during the period of 1985–2015 in the US, and the annual urban area growth is not linear over the years. Overall, the growth pattern of urban extent in most coastal states is plateaued over the past three decades while the states in the Midwestern US show an accelerated growth pattern. The derived annual urban extents are of great use for relevant urban studies such as urban area projection and urban sprawl modeling over large areas. Moreover, the proposed mapping framework is transferable for developing annual dynamics of urban extent in other regions and even globally. The data are available at https://doi.org/10.6084/m9.figshare.8190920.v2 (Li et al., 2019c).