Aspects influencing the selection of representative urbanization measures to quantify urban-rural gradients

The quantification of urban-rural gradients using urbanization measures has become standard practice in many urban ecological studies. Nonetheless, the choice of urbanization measures for a specific urban gradient still remains problematic. Increasing numbers of papers stress the importance of comparative urban ecological research, in an attempt to contribute to an understanding of the ecology 'of' cities. This implies that research in diverse urban areas globally should be comparable. This study follows an approach to quantify the urban-rural gradient in Klerksdorp previously followed in Melbourne, Australia with the goal to help elucidate the viability of creating a standard set of urbanization measures that is useful across continents. Satellite imagery and spatial analysis were used to calculate the values of 12 urbanization measures across a 900 km(2) landscape grid. Principal components analysis is commonly used to identify smaller subsets of measures to quantify urban-rural gradients. The results of this study indicate that factor analysis is more suitable than principal components analysis and ideal in identifying these independent measures of urbanization. The factor analysis revealed that landscape structure and demographic attributes are both essential characteristics of a city that needs to be accounted for in the choice of urbanization measures. Additionally, we identified seven aspects influencing the direct comparison of cities, namely: scale of analysis, spatial resolution, classification typology, accuracy of input data, specific measure equations, the type of statistical analysis and the habitat context. These aspects must be taken into consideration and resolved before effective comparative gradient research between cities can be achieved.