A semi-automated approach for quantitative mapping of woody cover from historical time series aerial photography and satellite imagery

Abstract

Savanna landscapes are characterised by a canopy of discontinuous tree cover overlying an understorey of shrubs and continuous grass cover. The distribution of trees (woody cover) is variable both spatially and temporally. Analysis of woody cover dynamics can provide a spatial and temporal envelope encompassing variability is useful for informing mine closure criteria. With the impending closure of Ranger uranium mine, ecologically appropriate closure criteria for ecosystem restoration are being developed through a framework of rehabilitation standards. One such closure criteria is canopy cover and historical woody cover is being used to derive the range in woody cover that can be expected over time once the mine site has been revegetated. This study reports on the development and testing of a technique for extracted woody cover from remotely sensed data (in the form of historical aerial photography and satellite imagery) in the areas adjacent to Ranger uranium mine in the World Heritage listed Kakadu National Park, northern Australia. An object-based image analysis technique was applied to four data sets from four different dates: greyscale, true colour and colour infrared aerial photo mosaics (from 1964, 1976 and 1981 respectively) and a high spatial resolution satellite image (from 2010). Overall accuracies of woody cover from each of the data sets exceeded 94%. In addition, proportional cover derived from this method displays linear relationships to cover derived from visual estimates. Due to the success of the technique, it will be applied to more data sets from different dates over the study area to assess the variability of woody cover over time to inform ecosystem restoration criteria for the mine closure.