Representing coastal land use in the island of Gran Canaria

Abstract

This map displays a geographic information system-based spatial analysis representing coastal land use in the island of Gran Canaria. It presents a method of summarizing coastal patterns of land use/cover into arc/sectors of a graph, setting up spatial units of analysis based on compass directions suitable to organize, analyse and depict spatial data. The method allows the easy detection of patterns and visualization of similarities between two or more sets of coastal land use/cover data. This paper outlines the methods used in designing the map.

Keywords:

• coastal areas
• land use
• spatial analysis
• geographic visualization

1. Introduction

In Europe, coastal management is a key topic in planning, with the European Commission operating a programme on ‘Integrated Coastal Zone Management’ (ICZM) from the mid-1990s. ICZM attempts to ‘balance the needs of development with protection of the very resources that sustain coastal economies’ (EEA, 2006, p. 7). The role and significance of coastal areas are well documented in several studies (Roth, Oke, & Emery, 1989; Small & Nicholls, 2003; Thom, Williams, & Diefenderfer, 2005), with the importance magnified in the case of small islands, such as Gran Canaria, where tourism development and associated commercial and residential growth have dramatically changed the coastal landscape.

Gran Canaria is the second most populous of the Canary Islands, with an area of 1560 km² and approximately 850,000 inhabitants. Some of Gran Canaria's coastal areas are highly impervious and urbanization rates along the coast are much higher than further inland. In recent decades, the coastal strip has experienced a rapid transformation, mainly due to tourist development. The impact of this transformation has been recorded in land-use/land-cover (LULC) data.

Although studies examining land change are diverse (Mas, 1999; Munsi, Malaviya, Oinam, & Joshi, 2010; Schulz, Cayuela, Echeverria, Salas, & Rey Benayas, 2010; Yuan, Sawaya, Loeffelholz, & Bauer, 2005), LULC representation and visualization methods can be greatly improved. As researchers have increasingly large amounts of LULC data, there is a continuous need for tools and methods to synthesize information.

The main goal of this Main Map is to propose a method to represent coastal LULC in islands, in a meaningful and concise manner. Using CORINE datasets as data sources, geographic information system-based spatial analysis was employed to represent the distribution of land use relative to a 5-km coastal buffer zone. The results of the analysis are visualized in the form of a diagram, summarizing coastal patterns of LULC into arc/sectors of a graph, setting up spatial units of analysis based on compass directions suitable to organize, analyse and depict spatial data. The map is capable of showing coastal land-use data, whilst depicting Gran Canaria's overall land-use pattern, thus allowing the easy detection of patterns over different years of observation.

2. Concepts and framework

Within the field of land change assessment, a key aspect of all studies is data presentation. In land change assessment, change matrices are a dominant tabular method to highlight LULC changes. These matrices keep track of LULC shifts from one category to other categories, and are widely applied in land change research (Xu, Liao, Shen, Zhang, & Mei, 2007). Nonetheless, for large areas or several study cases, they may turn out too extensive, thus becoming impractical and ineffective. Besides tabular form, LULC data may also be presented in graphical forms, such as maps, graphs and diagrams.

Once several LULC data sources become available for an area, LULC data might be treated as spatiotemporal data, which implies specific visualization methods. Several studies have covered visualization methods of spatiotemporal data (Andrienko, Dykes, Fabrikant, & Wachowicz, 2008; Claramunt, Jiang, & Bargiela, 2000; Monmonier, 1990; Peuquet, 1994). These graphical methods make it possible to easily draw visual impressions of spatiotemporal data, facilitate comparisons and present characteristics in a straightforward manner. Monmonier (1990) highlights several graphical methods to portray quantitative spatiotemporal data. Two of the main methods addressed are the ‘single-static-map’ (p. 30) and ‘multiple-static-maps’ (p. 30) strategies. ‘Single-static-map’ strategies incorporate the temporal dimension through techniques ranging from ‘complex point symbols, or temporal glyphs, to generalized trend-surface or flow-linkage maps’ (Monmonier, 1990, p. 30). The ‘multiple-static-maps’ strategy ‘juxtaposes two or more maps for a simultaneous visual comparison of time units’ (Monmonier, 1990, p. 30). The ‘multiple-static-maps’ strategy suits LULC change assessment particularly well, since each map presents a snapshot for a discrete period of time. Thus, if two or more maps are juxtaposed, the user can visually compare LULC patterns. This may be sufficient for most studies; nonetheless, issues arise once it becomes difficult to represent LULC data analysis, such as gains, losses and net change. The solution resides in resorting to other graphical representations. However, a major problem with graphical representations of geographic data lies in the need to display ‘both the attribute space familiar to the statistician and the geographic space that provides the necessary sense of place and relative location’ (Monmonier, 1990, p. 38).

In order to represent attribute and geographic space, the current map proposes an adaptation of rose diagrams to represent LULC data. Rose diagrams are circular histograms in which the frequency of vector data in predefined azimuthal classes is plotted as sectors of circles with a common origin (Baas, 2000). Early applications of rose diagrams can be found in Curray (1956) where the rose diagram is used to show direction as well as magnitude. Over the years, the method has become widely used in the Earth sciences (Baas, 2000). Rose diagrams are common ways of visualizing geographic data, and extensively used in several studies. Examples include the analysis of spatial patterns of vegetation fire (Brivio, Grégoire, Koffi, & Ober, 1997), terrain pattern recognition (Miliaresis, 2008) and spatial orientation of urban expansion (Xu et al., 2007). Since LULC data are not directional per se, it needs to be analysed through appropriate methods in order to be represented with rose diagrams. In order to do this, Xu et al. (2007) employed ‘concentric circle and sector analysis methods’ (p. 20). According to the authors, the ‘concentric circle method’ is ‘effective for analysing the quantity and distribution of different categories of land use with respect to distance from a pre-determined urban center’ (Xu et al., 2007, p. 20). On the other hand, ‘sector analysis’ can transpose the cardinal directions to a circular graph. With these methods, Xu et al. (2007) utilized graphs to illustrate the spatial orientation of urban expansion. The authors named these graphs ‘rose diagrams of urban expansion’ (p. 23). So far, however, there has been little discussion about alternative methods to represent LULC data. This Main Map proposes an enhancement of existing methods in order to present a diagram for evaluations of coastal LULC gains and losses.

3. Methods

Gran Canaria's CORINE 1990 and 2006 datasets provided the map's data sources. CORINE datasets (http://www.eea.europa.eu/data-and-maps) are part of the programme started in 1985 by the European Community to generate digital land-cover maps covering Europe. The availability of comparable datasets using similar source data and having the same technical characteristics (1:100,000 scale and 25 ha minimum mapping unit) allows a quantitative characterization and assessment of land change, over a period of two decades. The first iteration of the CORINE data covered the reference year of 1990 with subsequent releases covering the years 2000 and 2006. The latest 2012 update is still under production.

The map consists of four figures. The two top figures depict Gran Canaria's overall land use in the two years under analysis, whilst the bottom two figures shows a coastal land-use diagram for each year. Empirical knowledge of the island's landscape has established the importance of working with compass directions, since the trade winds carry moisture to the northeast of the island making this area cooler, wetter and more favourable to agriculture, which has been the island's economic driving force up to the mid-twentieth century. As such, the North-northeast (NNE) and East-northeast (ENE) sectors have become more populated and with stronger urban dynamics, and this has been recorded in the LULC data.

In order to process the data, and with the origin in the island's centroid, the island was divided into eight regions. These eight regions are meant to represent the cardinal and intermediate directions, thus dividing the island into eight geographic sectors. The coastline identified in the CORINE dataset was then extracted and buffered in 1 km increments, up to 5 km, thus creating five buffers. These buffers were then intersected with the CORINE data set polygons for 1990 and 2006. Finally, these five buffers, filled with land-use data, were intersected with the eight regions representing the island's cardinal and intermediate directions. The final result of this process is, for each of the eight regions, five coastal buffers filled with land-use data for 1990 and 2006. The objective of this procedure was to compute the percentages of land-use classes in each of the intersected buffers within the eight regions.

LULC data are not directional; nonetheless, the proposed diagram with eight sectors and five concentric rings transposes LULC data to geographic units that can serve as basis for further analysis, since it sets up spatial units of analysis based on compass directions suitable to organize, analyse and depict spatial data. The relationship between land use/cover and prevailing winds (trade winds) is clear in many tropical and subtropical islands (e.g. in the Caribbean islands with landscape difference between Windward and Leeward, and in the Atlantic subtropical islands with a North/South difference) and explains why it is important to use rose diagram-like displays to analyse spatial data in these islands.

3.1. Map design

The coastal land-use diagram was built by bisecting the angles of a circle by 45°. The eight radial dividers provide the orientation of the cardinal directions and the intermediate directions. In this map, five 1-km buffers have been created. Therefore, the circle was further divided into six concentric rings. The sixth and last interior ring is void of data. The result of this process is a circle, divided by six concentric dividers and eight radial dividers. Thus, the concentric dividers represent the buffers, whilst the radial dividers represent the cardinal and intermediate directions, and divide the circle into the same eight regions (geographic units) that the island had been divided. One of the method's drawbacks is that in the case of an elongate island, oriented in a cardinal direction, some sectors would contain much more area than others, thus the importance of normalizing LULC data as a percentage.

Taking into account the percentage that each class occupies on each buffer in each sector, the concentric and radial dividers can be used to represent the computed data. As with a histogram, the diagram areas should be proportional to the frequency of the data. Since a concentric divider (buffer) in each radial divider takes 45°, 45° represents 100% of the occupation. Taking this into account, we can compute the degrees for each land-use class. For example,

This calculates how many degrees a LULC class needs to bisect the radial divider. Now, in order to represent the data, we need to take into account the angular difference in each radial divider. Starting from 0°, to represent data in the first radial divider (NNE) ending at 45°, 45° would show that a single class occupies 100% of the land use in that buffer and in the NNE sector. Since we want to calculate the degrees that a single class requires to be represented in this first sector, we subtract the value for (1) from the upper boundary value. This is performed iteratively for each class. Therefore, the proposed diagram allows representation of LULC data in both attribute and geographic space. Land-use percentages make the attribute space, whereas the diagram's eight sectors and five concentric rings represent the island's geographic space.

The map design follows the ‘multiple-static-maps strategy’ (Monmonier, 1990, p. 30), which juxtaposes graphics for a simultaneous visual comparison of time units. In this map, graphics are juxtaposed for 1990 and 2006. As can be seen in the map, this method allows the reader to infer trends over time, including coastal land-use gains and losses. Since the map was intended to be viewed only in printed form, it has an ISO standard A4 page format. Visual relationships were designed to achieve appropriate visual hierarchy and optimize visual contrast. Given that the map depicts only three classes, a decision was made to avoid colour use and design the map in greyscale to increase legibility whilst improving the portrayal of variation in the data. Geographic labelling was hand-made to allow final manual tidying of the text. For the typography, the map uses the Calibre font in sizes ranging from 5 to 13 pt.

4. Conclusions

Graphical representation of LULC provides valuable information for planners and land resource managers. Overall, the map's method can be used to (1) easily draw visual impressions of coastal LULC data; (2) facilitate comparisons among study areas and (3) uncover underlying trends of land change. By presenting a form of summarizing coastal patterns of LULC into arc/sectors of a graph, this method can simplify complex spatial data in a single graphical presentation within a geographic context, which otherwise would occupy extensive tables of data. By setting up geographic units based on compass directions, the method uses spatial units of analysis without resorting to administrative units. And since this method may be easily customized to fit other study areas elsewhere, it can incorporate its graphical dimension into broader approaches contributing to the systematic and representative analysis of LULC.

Software

Spatial analysis and data manipulation were accomplished using Esri ArcGIS 10, with map layouts exported to the Illustrator file format. Land-use statistical analysis was performed using Microsoft Excel 2007. Finally, coastal land-use diagrams, map composition and labelling were made using Adobe Illustrator CS6.

Acknowledgements

Financial support for this research from the Portuguese Foundation for Science and Technology is greatly acknowledged (Grant: SFRH/BD/69396/2010). The author would also like to express his sincere gratitude to Professors Javier Gutiérrez Puebla, Patrick J. Kennelly and Félix Angel González Peñaloza, for their valuable suggestions on the previous drafts of this map.