Impact of broad- and fine-scale patterns on regional landscape characterization using AVHRR-derived land cover data

As global environmental issues and the impact of global environmental changes attract increasing attention from the public and policymakers, the focus has been on developing indicators of ecosystem health at regional to global scales. This broadening of the scale of assessments has led to a heavy reliance on quantitative tools using an array of technological advances (e.g., GIS, remote sensing). The accuracy and quality of reports on ecosystem integrity or health obtained from broad-scale assessments depend to a large extent on understanding the limits of the data and the technologies used. As the public and policymakers increasingly require rapid, timely evaluation of ecosystem conditions, the use of data and technologies that allow real-time, relatively inexpensive, frequent assessments will increase accordingly. Although the limitations of these data and technologies are broadly understood, the practical implications for specific assessments are not always fully considered. Ar broad scales, key measures or indicators of ecosystem integrity are usually related to landscape patterns, such as patch abundance, size, and spatial distribution. Therefore, accurate and robust assessments of ecosystem health require a reliable methodology that captures relevant measures of ecosystem health. It is the purpose of this article to document one aspect of the development of reliable characterization and monitoring protocols, namely the appropriateness of the data used to characterize landscape patterns for assessing ecosystem integrity and ecological conditions. We assessed the use of coarse-resolution advanced very high resolution radiometer (AVHRR) derived data for characterizing broad-scale (ecoregions) and fine-scale landscape patterns over the 580,000 km(2) of the interior Columbia River basin. We analyzed patch and landscape indices, correlations, principal components, and a statistical comparison of an AVHRR-derived land-cover map with fine-resolution data. Differences in broad-scale landscape patterns among ecoregions were most successfully characterized using combinations of patch and landscape indices. Both classification levels differentiated ecoregion patterns, but the finest level provided increased discriminating ability. Ecoregion and classification levels interacted to produce results specific to particular cover types. AVHRR-derived data did not adequately characterize some important fine-resolution landscape features. We hypothesized that AVHRR-derived data would be in closest agreement with fine-resolution data when landscape texture was coarse. The results Indicated that landscape texture influenced the performance of AVHRR-derived data, but in an inconsistent manner, suggesting that coarse-resolution differences from fine-resolution data have limited predictability. Our results have important implications for characterizing and discriminating landscape patterns among and within ecoregions and along ecological gradients. Across areas with varying landscape heterogeneity, changes in relationships among ecoregion levels, classification levels, and data resolutions may increase opportunities for inaccurate representation of landscape patterns. The choice of a specific ecoregion level, classification level, and data resolution should be examined for patterns of inconsistent characterization before use in broad-scale ecological assessments.