Efficiency of landscape metrics characterizing urban land surface temperature

Abstract

Identification of the relationship between landscape metrics and urban land surface temperature (LST) provides a basic understanding of the interaction of landscape pattern and ecological process. However, the evidenced relationships between landscape metrics and LST are still uncertain, and cannot provide fundamental support to landscape management. Other than a test of statistical significance, four judgments (median correlation value; temporal variation and statistical scale; threshold; and effects of key variables) are considered “efficient” criteria in this study. Partial correlation and piecewise linear regression are used to focus on the indicators of land cover proportion, biophysical proportion, and mainly area- or shape-related landscape configuration indicators. The results show: (1) land cover proportion can almost substitute for area-related landscape configuration indicators; (2) landscape composition is more efficient than configuration as an indicator because of its relatively stable temporal correlation values at different statistical scales; (3) the evident landscape composition threshold of vegetated land surfaces (50–70% for land cover proportion and 0.2–0.3 for biophysical proportion) is more indicative in application than the linear relationship for unvegetated land surfaces; and (4) landscape metrics are better correlated with LST in high temperature than in low temperature, and urban area weakly influences this correlation (R2 < 0.2). Additionally, the configuration metrics at a landscape level are not recommended for characterizing LST. Depended on the efficient indicators, such as the application of thresholds, landscape planning can be linked to the quantitative observations and statistical evidence in landscape metrics.