Creating a large area landcover dataset for public land monitoring and reporting

Enhanced up-to-date information on the extent and quality of woody vegetation systems is required to inform sustainable land management. Government and land management agencies require tools to generate consistent and scientifically robust large area (i.e. landscape level) woody vegetation spatial features – with which to conduct environmental monitoring and assessment, under a number of regulatory environmental monitoring frameworks (e.g. UN Framework Convention on Climate Change, State of the Forest reporting - Victoria, National Forest Inventory – Australia). This paper presents a method for the creation of a partial sample land cover baseline to complement existing on-ground forest monitoring plots and the total sample mapping using moderate-resolution remote sensing imagery. The land cover baseline was created via non-stereo, on-screen aerial photographic interpretation (API). Land cover maps were produced from 790 2x2 km digital (georectified) high-resolution (≤ 50 cm) colour aerial photographs (photoplots) located across a large area sampling grid (stratified by IBRA1 Bioregion), with an average of 72 photoplots in each Bioregion. Forest type and structural components (based on the National Vegetation Information System – NVIS) and land cover information in non-forest areas (based on the UN FAO Land Cover Classification System - LCCS) were derived using a semi-automated API (air photo interpretation) methodology. Initially, landscape elements (image objects) were delineated using automated segmentation algorithms and these were tuned to ensure resultant image objects approximated the land cover classes of interest given the stated minimum mapping unit (0.5 ha). Image objects were subsequently verified and classified, using operator knowledge, into hierarchical land cover categories (including dominant species, height, canopy cover, LCCS class - and where appropriate, whether the area had been fire affected). Modelling to a baseline year was required to account for landcover change events and non-synchronous image capture. Continuous improvement and validation of the dataset was built into the methodology so as to provide on-going feedback to the API team and report accuracy across all bioregions. This case study demonstrates the operational capability of semi-automated API to support large area (state-wide) environmental monitoring and assessment