The impact of spatial scale: exploring urban butterfly abundance and richness patterns using multi-criteria decision analysis and principal component analysis
Modelling changes in biodiversity has become a necessary component of smart urban planning practices. However, concepts such as biodiversity are often evaluated using area-based composite indices, the results of which are heavily reliant on specific parameters chosen. This paper explores the design and implementation of a butterfly biodiversity index by comparing two widely accepted modelling techniques: principal component analysis and spatial multi-criteria decision analysis (MCDA). A high degree of scale dependency has been demonstrated in previous studies exploring the use of area-based composite measures. To evaluate the impact of scale, each model was assessed at two different spatial resolutions. The outcomes were analyzed, mapped and compared using ordinary least squares, geographically weighted regression and global Moran’s I to evaluate relative biodiversity patterns across the City of Toronto, Canada. The Urban Butterfly Index - City of Toronto includes a geodatabase that consists of shapefiles of various geographic information used to model urban biodiversity and the butterfly observation points provided by the Ontario Butterfly Atlas. The Ontario Butterfly Atlas (OBA) is a program created and administered by the Toronto Entomologists’ Association (TEA), a non-profit organization that aims to educate and inform the public about local insect populations. Volunteers submit species observations throughout the year including important attribute data such as location (i.e., GPS coordinates or detailed location description to be verified), species name, observation date, adult and immature counts (TEA 2019). The OBA dataset was used as the benchmark variable in each regression analysis performed (i.e., species abundance or richness per area). Each geographic dataset was aggregated and summed by areal unit (i.e., Census Tracts or Dissemination Areas) and combined into a set of composite index scores using either principal component analysis (PCA) or spatial multi-criteria decision analysis (i.e., Weighted Linear Combination). Tabular files used in the Ordinary Least Squares (OLS) and Geographic Weighted Regression (GWR) are also provided and include the benchmark variable of butterfly observations per area (i.e. "SA07_Ha" and "SR07_Ha"), as well as the final PCA (i.e., "PCAF") and WLC scores (i.e., "WLCF"). Findings indicate that the impact of spatial scale was significant, whereby the coarser resolution models (i.e., Census Tracts) were found to be more highly correlated with biodiversity, compared to the finer resolution models (i.e., Dissemination Areas). The results of this study contribute to a growing body of literature that explores key conceptual questions regarding the robustness of GIS-based MCDA, the impact of scale in urban ecology studies, and the use of composite indices to manage spatial ecological data.