Detecting Human Disturbances to Land Cover Using a Bayesian Ensemble Model

Tongxi Hu
Category: 
Graduate (PhD)
Advisor: 
Kaiguang Zhao
Department: 
Environment and Natural Resources
Abstract: 

Human-directed disturbances to the landscape have increased spatially and temporally in response to rapid industrialization. These disturbances could cause habitat fragmentation for some species and contribute to a loss in biodiversity. But sometimes, human-caused disturbances are difficult to monitor because their sizes are spatially minor, for example, the development of right-of-ways for utility transmission lines. Here we provide a case study to show how to detect human disturbances to land cover leveraging an advanced change-detection algorithm Bayesian Estimator of Abrupt change, Seasonality, and Trend (BEAST) and high-density Landsat time-series data from Google Earth Engine. This study was conducted in a watershed that has been highly disturbed by various human activities and infrastructure including surface mining and utility transmission lines. We analyzed patterns and magnitudes of disturbances in the watershed caused by these human activities. The greatest loss of vegetated area was due to surface mining activities with nearly 3479 ha, followed by the development of utility pipelines (power lines and gas pipelines) with about 1100 ha. These activities disturbed 9.7% (including both inactive and active activities) of the land cover within the watershed over 18 years. Although the disturbed area is relatively small, the disturbances are discontinuous with various patterns (e.g., lines, patches) and not concentrated in particular areas but spread over the entire watershed. The disturbances that were captured by BEAST provide considerable scope for future research regarding their impacts on the environment, such as the destruction of wildlife habitat, at a local level.