Geoscience Reference
In-Depth Information
7.6 APPLICATIONS OF EAs IN GEOGRAPHY
Within the geographical literature, the use of EAs has primarily been limited to the application
of GAs across a range of different areas. The first main area is model calibration, which was also
one of the first applications of GAs in geographical research, that is, to determine the parameters
of a spatial interaction (SI) model (Diplock and Openshaw 1996; Openshaw 1998). More recently,
GAs have been used to successfully calibrate the parameters of cellular automata (CA) models
of urban growth, land use and the spread of forest fires (Li et al. 2007, 2013; Shan et al. 2008;
Al-Ahmadi et al. 2009; Vahidnia et al. 2013) where GAs have often been shown to be superior to
other search methods.
In the past, some limited work has been undertaken in the area of cartography, for example,
using GAs for automatic cartographic label placement, map generalisation, line simplification and
in the specification of optimal class intervals for choropleth maps (Dijk et al. 2002; Yamamoto et al.
2002; Armstrong et al. 2003).
Another main area where GAs have been applied is to multi-criteria evaluation, that is, to opti-
mise the weights of different factors and/or to choose optimal scenarios for land use planning, suit-
ability analyses and route planning. For example, Huang et al. (2004) used a GA to determine the
weights of different factors including safety, costs and security to calculate the suitability of routes
for transporting hazardous materials in Singapore, while Tsou (2010) used a GA for both the initial
population determination and route planning for safety. Other examples include shortest path opti-
misation (Mooney and Winstanley 2006), optimisation of a tourist sightseeing route (Huang et al.
2006) and rail line optimisation (Jha et al. 2007). GAs have also been used for allocating services
optimally, for example, the location and number of ambulance stations (Sasaki et al. 2010; Comber
et al. 2011) and medical centres (Aghamohammadi et al. 2013) as well as the recent work on land use
allocation framed as a multi-objective optimisation problem (Chandramouli et al. 2009; Neema and
Ohgai 2010, 2013; Cao et al. 2011, 2012; Haque and Asami 2011; Fotakis and Sidiropoulos 2012;
Gong et al. 2012).
Other miscellaneous areas of GA application include the design of electoral zones, land
parcels and other redistricting problems (Bacao et al. 2005; Datta et al. 2012; Joshi et al. 2012;
Demetriou et al. 2013), the creation of synthetic spatial data (Hermes and Poulsen 2012), the
identification of crime hot spots and other classification problems (Oyana and Dai 2010; Wu and
Grubesic 2010) and the estimation of missing urban areas in OpenStreetMap (Hagenauer and
Helbich 2012).
Although the majority of applications fall within the scope of human geography, there are also
examples within physical geography. These include the optimisation of a neural network model con-
figuration for prediction of suspended sediment (Heppenstall et al. 2007b), the estimation of land
surface biogeophysical variables (Liang 2007), the calibration of fuzzy membership functions for
an air pollution model of Tehran (Shad et al. 2009), the optimal design of a water quality network
(Park et al. 2006) and the assessment of groundwater pollution vulnerability in a Korean aquifer
(Ahn et al. 2011). There is also a very extensive literature on the use of EAs for the calibration of
hydrological models and in water resource planning and management (e.g. see Jain and Srinivasulu
2008 and Nicklow et al. 2010 for comprehensive reviews).
Other types of EAs have also been applied to geographical problems. Early work includes the use
of GP for breeding SI models (Diplock 1998) and the more recent application of GP in combination
with GAs to create gridded populations of urban and rural areas (Liao et al. 2010). As GEP is a more
recent development, applications have only more recently appeared in geography, for example, in
multisite land use allocation (Eldrandaly 2010) and in modelling evaporation and other hydrological
problems (Beriro et al. 2013, 2014).
In the next two sections, we present examples of GAs applied to optimising model parameters
and to breeding model equations.
