Discriminating the occurrence of pitch canker infection in Pinus radiata forests using high spatial resolution QuickBird data and artificial neural networks

Pitch canker is causing serious damage to Pinus radiata forests in South Africa. There is an urgent need to find an efficient way to assess the extent and variability of the disease at a broad spatial scale. The aim of this study is to explore the utility of transformed high spatial resolution QuickBird imagery and artificial neural networks, for the detection and mapping of pitch canker disease. Several vegetation indices including the Tasseled Cap Transformation were used to discriminate between healthy and infected P. radiata tree crowns using a feed-forward neural network and a Naive Bayes classifier. The neural network model showed high discriminatory power with an overall accuracy of 82.15% and KHAT of 0.65. These results are promising for the future management of pitch canker disease at a landscape scale.