Resource information
The interactions between wind-blown sand transport, pioneer vegetation and succession vegetation in coastal dune fields play an important role in landform development and determine the balance between stabilization and re-activation of these aeolian landscapes. High-resolution mapping of vegetation communities across a dune field – in particular, the mixture of different functional plant types such as pioneer versus succession species – is critical for the calculation of landscape metrics that enable a rigorous and quantitative testing of numerical simulation models, as well as for informing targeted land management actions that maintain biodiversity and ecological functions. This article presents a method of using maximum likelihood classification (MLC) to inform linear spectral mixture analysis (LSMA) for quantifying sub-pixel abundance of sand, pioneer and succession vegetation in a coastal dune area in Wales, from archived imagery obtained from the Compact Airborne Spectral Imager (CASI) in 1997. LSMA is first applied to derive sub-pixel fractional abundances of soil, green vegetation (GV) and non-GV elements. An MLC is developed separately for automatically identifying pixels believed to contain a mixture of the two functional vegetation types, and this then serves as a basis for applying a transform that interprets the LSMA results in terms of sand and pioneer and succession vegetation communities. Very high resolution (0.1 m pixels) colour aerial photography, taken simultaneously with the CASI data, and field survey data from both 1997 and 2009 were used to aid the MLC and the transform algorithm and were also used for a limited validation exercise. The LSMA abundance maps achieved an overall accuracy of 82.7% (kappa coefficient κ = 0.78). The reduced MLC vegetation maps (four classes) achieved an overall accuracy of 98.2% (kappa coefficient κ = 0.96). Although it was not possible to validate the final pioneer and succession vegetation abundance maps quantitatively, a qualitative review of the results for selected locations within the dune field indicates the viability of applying MLC to help direct a transformation of LSMA abundance maps into functional vegetation abundance data.
