A comparison of canopy cover derived from object-based crown extraction to pixel-based cover estimates

Canopy or vegetative crown cover is a popular structural characteristic that is described in both remote sensing and ecological research. There are, however, numerous methods and statistical measures described that can be used for the estimation and description of the proportion of canopy cover. Most methods of calculating vegetative cover from remotely sensed data are based on pixel-based measures in the form of ratios and indices. When calculating cover estimates from high spatial resolution (HSR) imagery the pixel-based methods also provide values of cover for non-canopy or between-canopy data especially in woodland regions such as tropical savanna. The non-canopy values are included in any calculations of cover. Therefore it can be argued that these measures are not truly a representation of canopy cover but of cover in general. This paper describes an object-based method for calculating canopy cover by extracting tree crowns from HSR imagery over tropical savanna woodland. Cover estimates from this method are then compared with those obtained from pixel-based statistical and ratio measures from medium resolution and HSR data. Relationships between the various measures are described.