Effects of patch characteristics and within patch heterogeneity on the accuracy of urban land cover estimates from visua
- PDF / 772,597 Bytes
- 15 Pages / 547.087 x 737.008 pts Page_size
- 25 Downloads / 187 Views
RESEARCH ARTICLE
Effects of patch characteristics and within patch heterogeneity on the accuracy of urban land cover estimates from visual interpretation Weiqi Zhou • M. L. Cadenasso
Received: 5 January 2012 / Accepted: 6 July 2012 Ó Springer Science+Business Media B.V. 2012
Abstract Landscape ecology links landscape pattern to ecological function. Achieving this goal hinges on accurate depiction and quantification of pattern, which is frequently done by visually interpreting remotely sensed imagery. Therefore, understanding both the accuracy of that interpretation and what influences its accuracy is crucial. In addition, imagery is pixel-based but landscape pattern exists, more realistically, as irregularly shaped patches. Patches may contain only one feature type such as trees, but, in some landscapes, patches may contain several different types of features such as trees and buildings. Using a patch-based approach, this paper investigates two types of variables—whole-patch and within-patch— that are hypothesized to influence the accuracy of visually estimating the cover of features within patches. A highly accurate reference map, obtained from object-based classification, was used to evaluate the accuracy of visual estimates of cover within patches. The effects of the variables on the accuracy of
these estimates were tested using logistic regressions and multimodel inferential procedures. Though all variables significantly affected the accuracy, the within-patch configuration of features is the most significant factor. In general, errors of cover estimates are more likely to occur when patches are smaller or have more complex shapes, and features within a patch are (1) more diverse; (2) more fragmented; (3) more complex in shape; and (4) physically less connected. These results provide an important first step towards a quantitative, spatially explicit model for predicting error of cover estimates and determining under what circumstances estimation error is most likely to occur.
W. Zhou (&) State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Shuangqinglu 18, Beijing 100085, China e-mail: [email protected]
Introduction
M. L. Cadenasso Department of Plant Sciences, University of California, Mail Stop 1, 1210 PES, One Shields Avenue, Davis, CA 95616, USA
Keywords Visual interpretation Object-based classification HERCULES Accuracy Land cover composition and configuration Pattern analysis Spatial heterogeneity Urban systems
Landscape ecology focuses on understanding the reciprocal link between pattern and process (Turner et al. 2001; Wu and Hobbs 2002). Building this understanding requires accurate quantification of landscape pattern at the grain and extent appropriate for a specific research question (Gustafson 1998; Turner 2005; Shao and Wu 2008). Quantifying
123
Landscape Ecol
landscape pattern primarily relies on thematic maps, and these maps are frequently derived from visual interpretation of remotely sensed im
Data Loading...
