UCLA

As the world has grown increasingly urbanized over the last half-century, management of regional metropolitan landscapes has become more complex. Planning and research efforts alike require an understanding of local conditions and variability within cities. In this study, I examine changing land cover, urban hydrology, and surface temperature within the Los Angeles region. Using a deep convolutional neural network, I disaggregated mixed Landsat pixels to track land cover change over a continuous 35-year record. I found relatively constant urban area but a higher abundance of irrigated tree cover than is commonly detected in long-term remote sensing data. I used the Distributed Hydrology Soil Vegetation Model to examine hydrology in two distinct neighborhoods within Los Angeles, and observed that new urban vegetation in primarily impervious areas reduced precipitation-event runoff but universally increased evapotranspiration more substantially than it did in vegetated areas. Both the hydrologic baseline and the responses to land use change varied substantially across different localities. I assessed the contributions of streets to land surface temperature (LST), and found that localized urban morphology and vegetation, rather than road surface, determine LST. In many cases, large impervious or irrigated pockets dominated neighborhood-scale LST signatures. Together, these analyses demonstrate that disparate vegetation, irrigation strategies, and building footprints across an urban area have unique interactions that can make regional approaches impractical and ineffective. Rather, urban ecosystem services can be strengthened by local consideration and an understanding of scale.