Multivariate correlations between imagery and field measurements across scales: comparing pixel aggregation and image segmentation

To successfully use remotely-sensed data in landscape-level management, questions as to the relevance of image data to landscape patterns and optimal scales of analysis must be addressed. Object-based image analysis, segmenting images into homogeneous regions called objects, has been suggested for increasing accuracy of remotely-sensed products, but little research has gone into determining image object size with regard to scaling of ecosystem properties. We looked at how segmentation of high-resolution Ikonos and medium-resolution Landsat images into successively coarser objects affected multivariate correlations between image data and eight percent-cover measurements of a sagebrush ecosystem. We also looked at changes in correlation as imagery was aggregated into larger square pixels. We found similar canonical correlations between field and image data at the finest scales, but higher for image segmentation than pixel aggregation for both images when scale increased. For image segmentation, correlations between the canonical variables and original field variables were invariant with respect to size of the image objects, suggesting linear scaling of vegetation cover in our study system. We detected a scaling threshold with the Ikonos segmentation and confirmed with a semi-variogram of the sample data. Below the threshold interpretation of the canonical variables was consistent: scale levels differed primarily in the amount of detail portrayed. Above the threshold, meaning of the canonical variables changed. This approach proved useful for evaluating overall utility of images to address an objective, and identified scaling limits for analysis. Selection of appropriate scale for analysis will ultimately depend on the objective being considered.