Estimating the road edge effect on adjacent Eucalyptus grandis forests in KwaZulu-Natal, South Africa, using texture measures and an artificial neural network

Abstract

SPOT-5 multispectral and panchromatic image data were used to compute texture measures to estimate the road edge effect on adjacent Eucalyptus grandis forests. Employing a stepwise selection algorithm enabled the selection of optimal texture measures that were input into a backpropagation artificial neural network. The R² of best models ranged from 0.67 to 0.89 for DBH, TH, BA, Volume and LAI on an independent test data set, with a root mean square error (RMSE) range of 0.01–5.36% for the respective variables. The result is critical for understanding and spatially predicting the road edge effect on adjacent vegetation.

Keywords:

• SPOT-5
• texture measures
• road edge effect
• artificial neural networks
• Eucalyptus grandis

Acknowledgements

The authors would like to acknowledge the support of the National Research Foundation, Mondi-SA and the CSIR, which enabled the successful completion of this paper.