Exploring Spatial Scale in Geography
Rating: 
The opening sentence says it all: “spatial scale is central to geography and to all disciplines concerned with the spatial arrangements of properties”. I think it fair to say that you can substitute a number of other words: ecology and conservation seem to fit but I’m sure there are others. The preface continues by arguing that the extent of a phenomenon might show different properties dependant on the resolution of the view to which some give the title of ‘paradox of scale’. The text continues to explore these ideas in the opening chapter. It’s made very clear that scale is a key topic but it might be better to say an understanding of scale and the implications thereof is key. It’s also made clear that there are numerous interpretations of the scale question and so devoting the second chapter to such issues is vital.
The chapter starts with that school hazard – the map scale. Large-scale maps cover small areas and vice versa. Simple when to get the idea of fraction, but not otherwise! It soon becomes clear that even trying to define scale is fraught with difficulty. Are we talking about measurement or variation? Perhaps we were thinking of the resolution over which we take our measurements. The exercise of species numbers and quadrat size should be enough to convince people about that! Interestingly, the work turns to consider scale dependence; the notion that objects close together are more likely to be similar. Of course this, in turn, depends on the scale of measurement! Other ideas are introduced but the basic notion of complexity remains. The text moves on to consider some of the issues raised in more detail, the first of which is the problem of the modifiable area. Let’s take an example. Map the distribution of trees on a hillside. These points might form a pattern, but the analysis of this pattern depends not on the trees but on the scale of the measurements. The zones you create to describe patterns are arbitrary; the case of the raster file is apposite! This moves on to look at the ecological fallacy – the idea of making inferences about individuals from aggregate data. From here, we move to spatial autocorrelation. The central concept is that items close together are likely to be similar. The problem is that this violates basic statistical understanding which suggests that points are independent of one another. Whereas this can be overcome in sampling if we are looking at a field issue, when it’s mapped it becomes a spatial issue. This in turn is linked to the size of the units used to determine the map. From this, the reader is taken into technical details of the various measures that can be used to best determine the scale of analysis. Not all point data have to be independent. It might well be that, for the variable under study, there are linkages. For example, voters might be independent but they might vote as a neighbourhood (not least because of the socio-economic impacts of a neighbourhood. The question then arises as to the best scale to show this. There are tests – a range of regression methods are described using the book’s main example – a useful device to highlight different issues with the same data. Of course, this uses real data. There are arguments that fractal analysis can create understanding by taking a more theoretical approach. It’s use might be limited but it serves to highlight the impact of scale-based decision-making. So far, the issue has been one of deciding on the scale and seeing how appropriate it is to find an explanation. The next chapters consider how changing the scale of observations might alter the output. A final chapter summarises findings whilst a companion website, provides
This is an intriguing book. The notion of scale is one that is under-appreciated. It is too easy to just take whatever is given of find something that might fit the data without really considering the implications of what one is doing. Patterns might appear that in real life are not there and vice versa. For this alone, the text is worth consulting. However, this is not the introductory text one might expect. It does require considerable understanding of both statistical analysis and basic GIS concepts. Given this, the text feels drawn more in one direction than another. It’s focus on maths means that the arguments about basic concept and procedure are less accessible. In trying to cater for both the tyro and the more advanced, the book runs the risk of solving the issues of neither.
