Authors: MAHYAR MASOUDI*, National University of Singapore & Campus for Research Excellence and Technological Enterprise (CREATE), Puay Yok Tan, National University of Singapore, Daniel Richards, Singapore-ETH Centre, ETH Zurich & Campus for Research Excellence and Technological Enterprise (CREATE)
Topics: Spatial Analysis & Modeling, Land Use and Land Cover Change, Quantitative Methods
Keywords: spatial pattern, landscape metrics, spatial scale, resolution, analytical unit, modifiable area unit problem
Session Type: Virtual Paper
Start / End Time: 8:00 AM / 9:15 AM
Room: Virtual 9
Presentation File: No File Uploaded
Comparison of results from spatial pattern analysis using landscape metrics is halted by varying combinations of metrics used in different studies and the uncertainty around metrics’ behaviour across analytical units. We used Singapore and ran 11 metrics on vegetation and impervious land covers derived from three imageries of Landsat 8, Sentinel-2 and Worlview-3 to examine if generalizations could be made regarding the behaviour of metrics across analytical units and the possibility to identify a smaller subset of common metrics. While resolution and extent of analysis were kept constant at each run, the analytical unit was progressively increased from 60 m x 60 m to 1200 m x 1200 m and the response of each metric was plotted, followed by fitting an equation to the response curve. While we demonstrated that different metrics behaved differently with increasing analytical unit, we found that each metric’s behaviour was consistent across resolutions and land cover, and that all metrics generally behaved predictably. We fitted three types of function to explain the metrics’ behaviour: 1) linear, 2) power, and 3) logarithmic. Further, 4 distinct patterns of 1) area, 2) aggregation and fragmentation, 3) shape complexity and 4) connectivity were commonly identified in the pattern of vegetation and impervious covers across analytical units and resolutions; the specific metric representing each component differed though depending on the specific combination of land cover, analytical unit and resolution of the base map. Our findings will facilitate spatial pattern analysis using landscape metrics and comparison of results across studies.