Leveraging a timeseries of remotely sensed data to better characterize the dynamics of grassland degradation and restoration on the Mongolian Plateau
As remotely sensed observations of the earth’s surface are becoming more available, they are being increasingly incorporated into a diverse array of ecological studies for the purposes of understanding of land cover, phenology, species distribution and hydrology. One of the advantages of remote sensing imagery is that it can provide multi-year data on land surface characteristics. These products can be used to quanitfy land use/land cover changes over time. However, selecting and processing cloud-free images in order to perform land cover classification can be difficult and time-consuming and thus many analyses of land-cover change are performed over smaller areas.
Recently, several global land cover maps have been produced, each with different purposes, at different spatial scales and utilizing different data for classification training. These global cover datasets can be powerful and useful tools for assessing patterns on large scales. However, these datasets are only available for a one or two years, which makes it difficult to assess land cover change over broader periods of time.
Additionally, these datasets are often created with a focus on global forest cover and have been notably unreliable, and therefore relatively useless, for arid rangeland regions of the world, particularly in Asia. Furthermore, land cover classifications utilize discrete categorizations of cover, but often what we are interested in, when studying land degradation and change, is is at the transition between categories. This is particularly true when studying desertification, the degradation of arid rangelands, wherein gradual and rapid shifts in grass cover across space and time are useful for assessing rangeland condition. Given these limitations, there is a need for a more appropriate approach to using remote sensing data to assess land change dynamics in arid rangelands.
Using the relatively new Google Earth Engine, we have created a new spatial land cover assessment process that focuses explicitly on changes in grassland cover over almost thirty years in northern China. We developed a novel spatio-temporal analysis that combines linear analyses of pixel change over time with classification analyses in order to create a coverage that visualizes the current grassland status, trajectory of change and historic context. These maps allow us a much more nuanced understanding of land use/cover in the region and also allow us to test hypotheses about the underlying mechanisms of change on a landscape scale.