Remote sensing studies investigating landscape patterns and changes have historically looked at discrete changes over a small collection of images or have been spatially confined to local-scale analyses. Scientific compromises have been largely driven by storage and computation limitations. Google Earth Engine and other cloud computing environments have ushered in a new era where the land surface can be evaluated continuously, over large spatial areas and long periods of time.
How are USGS scientists using the cloud to conduct research? This session will use a series of case studies to explain how USGS researchers are moving past desktop-driven computing and adopting cloud-based solutions for scientific analyses. By sharing use cases and discussing preferred workflows, the session will aim to facilitate greater use of emerging cloud-computing technologies (like Google Earth Engine) and enhance access to geospatial analyses without paying for traditional GIS/Remote Sensing software.
The session will include presentations from four USGS scientists that will share research methods and results that rely on Google Earth Engine (GEE) cloud computing. Presentations will cover 10 different remote sensing applications.
The first presentation (Chris Soulard) will chronicle four GEE efforts to better understand vegetation change dynamics, summarizing research on the effect of fire on vegetation in mountain meadows, the effect of climate on meadow greenness throughout the Sierra Nevada, the recovery of vegetation after oil and gas development in the Colorado Plateau, and the role of energy development and climate on annual grass invasion in the Mojave Desert.
The second presentation (Gabriel Senay) will describe how the GEE cloud computing platform is also being used to process and create CONUS-wide evapotranspiration (ET) water use information at the field-scale. The Operational Simplified Surface Energy Balance (SSEBop) model was applied to over 16,000 Landsat satellite images along with associated weather data layers to produce a seamless annual ET map in less than one week. Creating such a nationwide-wide annual ET product at the Landsat resolution is the first of its kind which would have been nearly impossible without these transformative remote sensing resources.
The third presentation (Jessica Walker) will demonstrate the utility of GEE for broad-scale temporal and spatial analyses on the basis of two ongoing investigations related to disturbance phenomena. The adaptation of the Landsat-based Dynamic Surface Water Extent algorithm to the GEE platform has allowed researcher to map inundation history across California's Central Valley from 1985 to 2015. GEE also provided the computing power for a shorter but more spatially expansive study of wildfire intensity across the state of Alaska as determined through MODIS Fire Radiative Power values from 2002 to 2018.
The final presentation in the session (Roy Petrakis) will cover the application of GEE scripting to produce multi-temporal vegetation and landscape indices using Landsat imagery, with the ability to produce these outputs for multiple fires by simply changing the fire boundary and the study dates. With these data sets, researchers were able to assess post-fire changes in the vegetation for different fuel treatment boundaries following the 2013 Creek fire in Arizona. Additional applications of GEE have produced multi-temporal raster and tabular datasets which have been used for monitoring of vegetation and landscape properties, using various sensors and other datasets that are freely available in GEE.
