Object‐based characterisation of reference sites for mine site closure using hyperspatial multispectral UAV imagery

Monitoring schemes to assess the success in mine site rehabilitation typically on data from offsite analogues or pre‐mining baselines (collectively reference sites) that describe the ecosystems of the region. Differences between rehabilitated sites and reference sites are evaluated using environmental indicators. Field‐based methods that sample reference sites are limited in the temporal and spatial scale, due to the time and effort required. Unmanned aerial vehicles (UAVs) for remote sensing acquire imagery at sub‐ decimetre ground sample distance that create orthomosaics and digital surface models (DSMs). These data are of spatial extents far greater than can be achieved using field sampling and consequently, the spatial extent of features and heterogeneity of sites can be more readily characterized. A geographic object‐based framework was used to analyse UAV‐based orthomosaics and DSMs captured over reference sites in Kakadu National Park, northern Australia, surrounding Ranger uranium mine. The UAV sensors provided information at a spatial scale to successfully characterise the reference sites. The analysis provided a number of the key environmental indicators including vegetation structural information, such as canopy cover and height, and fractional ground cover estimates. The methodology established here will form a basis for monitoring the impacts and success of rehabilitation as progress occurs towards mine closure