Exploiting high resolution multi-seasonal textural measures and spectral information for reedbed mapping

Onojeghuo, Alex Okiemute and Blackburn, George Alan (2016) Exploiting high resolution multi-seasonal textural measures and spectral information for reedbed mapping. Environments, 3 (1): 5. ISSN 2076-3298

Abstract

Reedbeds across the UK are amongst the most important habitats for rare and endangered birds, wildlife and organisms. However, over the past century, this valued wetland habitat has experienced a drastic reduction in quality and spatial coverage due to pressures from human related activities. To this end, conservation organisations across the UK have been charged with the task of conserving and expanding this threatened habitat. With this backdrop, the study aimed to develop a methodology for accurate reedbed mapping through the combined use of multi-seasonal texture measures and spectral information contained in high resolution QuickBird satellite imagery. The key objectives were to determine the most effective single-date (autumn or summer) and multi-seasonal QuickBird imagery suitable for reedbed mapping over the study area; to evaluate the effectiveness of combining multi-seasonal texture measures and spectral information for reedbed mapping using a variety of combinations; and to evaluate the most suitable classification technique for reedbed mapping from three selected classification techniques, namely maximum likelihood classifier, spectral angular mapper and artificial neural network. Using two selected grey-level co-occurrence textural measures (entropy and angular second moment), a series of experiments were conducted using varied combinations of single-date and multi-seasonal QuickBird imagery. Overall, the results indicate the multi-seasonal pansharpened multispectral bands (eight layers) combined with all eight grey level co-occurrence matrix texture measures (entropy and angular second moment computed using windows 3 × 3 and 7 × 7) produced the optimal reedbed (76.5%) and overall classification (78.1%) accuracies using the maximum likelihood classifier technique. Using the optimal 16 layer multi-seasonal pansharpened multispectral and texture combined image dataset, a total reedbed area of 9.8 hectares was successfully mapped over the three study sites. In conclusion, the study has demonstrated the value of utilizing multi-seasonal texture measures and pansharpened multispectral data for reedbed mapping.