Session Information
| Session | Poster Session | | Date | Monday (2008-04-07) | | Time | 5:00 PM - 7:00 PM | | Room | Grand Terrace |
Presentation Information
| Presenter | Jian Peng | | Title | The validity of landscape metrics in expressing spatial patterns | | Affiliation | Peking University | | Authors | Jian Peng, Yanglin Wang, Yuan Zhang, Jiansheng Wu | | Keywords | Landscape metrics, Multivariable regression, Sign effect, SIMMAP, Spatial pattern | | Presentation Type | Poster | Abstract:
Along with the widely used applications of remote sensing and geographical information system, spatial pattern analysis based on landscape metrics has become a fundamental component of research on landscape ecology. Behavior and validity of landscape metrics is always re-pondered by landscape ecologists. However, most of the existing works have been focused on the scale relations, the accuracy of source data, or the ecological implications of landscape metrics. Till now, few attentions have been paid to explore the essential characteristic of landscape metrics in expressing spatial patterns, although it is the primary step to evaluate the validity of metrics. In this paper, by setting 36 simulated landscapes generated by SIMMAP neutral landscape model as sample space and focusing on 23 widely used landscape metrics, we analyzed the validity of landscape metrics in expressing the complexity of such spatial pattern components as number of patch types, area ratio of each patch type and patch aggregation level, with the application of multivariate linear regression analysis method. The results showed that, all the metrics were effective in expressing spatial patterns despite remarkable differences in the object and extent of indication. According to the value of the determination coefficient, landscape metrics could be classified into two categories: (1) Type A included 14 landscape metrics with high correlation with independent variables; and (2) Type B included 9 landscape metrics with low correlation with independent variables. The categorization showed that: landscape metrics in type A mainly expressed the three components of spatial patterns in regression model; while besides the three, what the metrics in type B expressed also involved the other two components of spatial patterns, i.e. patch shape and contrast between neighboring patches. Furthermore, contrast analysis was also conducted to evaluate whether landscape metrics performed in simulated landscapes as they did in real landscapes. However, the results showed a distinct inconsistency between the performances of landscape metrics in simulated landscapes and real landscapes. Accurately, it is the difference of the correlation among spatial pattern components between simulated landscapes and real landscapes that led to this inconsistency. And when considering the very correlation, the changes of all the 23 landscape metrics against changing of number of patch types in simulated landscapes were consistent with those in real landscapes. Thus, the sign effect of spatial pattern components on landscape metrics was deducted, which could be illuminated as follows: when judging the monotonicity of the change of landscape metrics against changing of a certain component of spatial patterns, it was needed to consider not only the correlation between landscape metrics and the very component, but also the correlations between landscape metrics and the other components, and the correlations between the very component and the other components. Besides, the validity of landscape metrics in expressing spatial patterns should be set as the essential criterion for selecting good landscape metrics. According to the behaviors of landscape metrics in simulated landscapes, LPI and DIVISION were regarded as the best landscape metrics in expressing spatial patterns. One of the major advantages of this study is that it highlighted the necessity to analyze the validity of landscape metrics in expressing landscape patterns, as well as provided insights into the criterions for selecting good landscape metrics. |
|