Close-range vertical photography for measuring cover changes in perennial grasslands

Abstract

We describe a method of close-range vertical photography and digital image analysis for measuring changes in total projective cover in perennial tussock grasslands of semi-arid Australia. Repeated photographs of permanent plots (1 m2) were classified using supervised image analysis, providing a clear and objective record of the effects of single-burns on grass cover relative to controls. Computer simulations of the photographic set-up indicated that errors due to camera perspective were consistently less than 4% across a range of cover classes. Other errors, including misclassification error, were not quantified because simplified laboratory tasks indicated that conventional field methods, such as point sampling and visual estimation, provided unreliable estimates of grass cover and were therefore not suitable benchmarks for assessing the photographic method. However, the presented data indicate that the photographic method was sufficiently accurate and precise to measure treatment effects over time and to elucidate relationships between independent growth parameters across a range of cover conditions. In addition, the photographic method was inexpensive, involved minimal field time, and utilised commercial software to classify images. It has the clear advantage over more traditional methods of providing outputs that are readily archived for retrospective studies.