Using Very-Large-Scale Aerial Imagery for Rangeland Monitoring and Assessment: Some Statistical Considerations

Abstract

The availability of very-large-scale aerial (VLSA) imagery (typically less than 1 cm ground-sampling-distance spatial resolution) and techniques for processing those data into ecosystem indicators has opened the door for routinely using VLSA imagery in rangeland monitoring and assessment. However, for VLSA imagery to provide defensible information for managers, it is crucial to understand the statistical implications of designing and implementing VLSA image studies, including consideration of image scale, sample design limitations, and the need for validation of estimates. A significant advantage of VLSA imaging is that the researcher can specify the scale (i.e., spatial resolution and extent) of the images. VLSA image programs should plan for scales that match monitoring questions, size of landscape elements to be measured, and spatial heterogeneity of the environment. Failure to plan for scale may result in images that are not optimal for answering management questions. Probability-based sampling guards against bias and ensures that inferences can be made to the desired study area. Often collected along flight transects, VLSA imagery lends itself well to certain probability-based sample designs, such as systematic sampling, not often
used in field studies.With VLSA image programs, the sample unit can be an entire image or a portion of an image. It is critical to define the sampling unit and understand the relationship between measurements and estimates made from the imagery. Finally, it is important to statistically validate estimates produced from VLSA images at selected locations using quantitative data of the same scale and more precise and accurate than the VLSA image techniques. The extent to which VLSA imagery will be useful as a tool for understanding the status and trend of rangelands depends as much on the ability to build the imagery into robust
programs as it does on the ability to quickly and relatively easily collect VLSA images over large landscapes.