# Methodology appendix

p. 189-192

## Plan détaillé

## Texte intégral

# The base maps of the provinces and the districts

1When studying the spatial organization of Thailand, the politico-administrative unit that we primarily used was the province. In 1992 the number of provinces increased from 73 to 76. With most provincial statistics we were able to include these three most recent creations and produce maps of 76 provinces for 1996 or 1995. The older 73-province mesh was adopted systematically in three instances: maps in the Population chapter, based on population and housing figures from before the 2000 census; most of the maps that show evolution between two dates, whatever the statistical sources used, and lastly maps showing ratios based on the number of inhabi-tants from population and housing census figures (population density, level of facility provision, income and consumption indicators *per capita*, ...). The dif-ference in the base maps (according to the statistics) is indicated by a different outline: in black, when the statistics specify the three provinces of Nong Bua Lam Phu and Amnat Charoen in the North-East, Sa Kaeo in the east; grey when these three provinces are statis-tically associated with their province of origin (Udon Thani, Ubon Ratchathani, Prachin Buri respectively).

2So as not to lose too much information by referring to even older provincial boundaries, three provinces created in the 1970s were the subject of specific treatment relating to population dynamics between 1970 and 1990: Phayao, Yasothon and Mukdahan created from Chiang Rai, Ubon Rat-chathani and Nakhon Phanom respectively. While the population of each, in absolute terms, is shown for 1990, the rate of population growth between 1970 and 1990, shown by shading in the circles, refers to the rate in their original province. The same system was used for Mukdahan on the maps of agricultural evolution which were based on 1976 data.

3At district level, within the same region, the territorial demarcation for statistics differed according to the different themes (population, agriculture, industry) due to the delay in producing data for the last districts created. We opted to produce a map that was adjusted to the statistical data, by grouping districts together, usually 2, to form a single spatial unit that was older but corresponded to the area covered by the statistics.

4Given the concentration of industrial activities in the central provinces, which are the smallest spatial units, a separate inset in chapter 5 on Industry presents 10 of these provinces in such a way that the symbols are legible.

# Processing and cartographic representation of the statistical information

5This atlas obeys the rules of cartography. Absolute values are represented by symbols and relative values by shading patterns, with graded series from pale to dark as the phenomenon intensifies. The size of the symbols is usually proportional to the measurement or count at a particular point or, less frequently, organized into classes: each size of symbol then corresponds to a defined range (e.g. Evolution of forest areas on plate 23 Deforestation).

6Concerning relative values, the series of spatial data were split into classes. Thus, different methods of discretization were used, according to the hypotheses that were put forward and the spatial structures that it seemed relevant to highlight. This is a complex and delicate operation, since the mapping has to be not only an exact restitution of the phenomena but also a legible representation of the reality being analyzed; to facilitate the process we used the statistical mapping software Cabral, developed by scientists at the IRD. Principal component analyses were carried out using SPSS software. Maps showing flow were produced with MapInfo.

7In order to inform the reader as to the choices that were made when dividing statistical data into classes, all the maps are accompanied by an explanatory figure, a histogram or a ternary diagram (see plate 71 Examples). The synthesis maps, based on a principal component analysis, are accompanied by a linkage tree and a legend table.

## 1 - Histograms

8These appear on the analytical maps and refer to the discretization of a variable. There is one column for each class created. The width of the column indicates the extent of the class and the surface is proportional to the number of spatial units that make up the class. The percentages show the proportion of units from each class in the total number of units. Three values are given: the class with the most spatial units, the class with the fewest spatial units and an intermediary class. The distribution given by the histogram is that of the complete statistical series even when the map has been cleaned up from symbols below a certain threshold which are too small to be easily legible.

## 2 - Ternary diagrams

9These appear on some structural maps to assess groups that are straightforward but significant. The diagram combines three variables with the total representing 100 % of the phenomenon under analysis. Each side of the equilateral triangle is allotted to one variable and is graduated from 0 to 100 % by lines parallel to the preceding side. Each spatial unit (the small circles) is pinpointed according to the coordinates of the three variables that define it. Classes are created according to the groups that are formed. In the legend, a table gives the ranges of percentages for each class: to facilitate reading, the values of the predominant variable or variables are shown in bold.

## 3 - Synthesis maps

10These combine a large number of variables, some that have been represented in the atlas (on analytical maps) and others that have not. One synthesis map can be found at the end of most chapters as a conclusion or for some specific themes (Main religions, Characteristics of tertiary product, etc.). They represent characteristic territorial structures and dynamics for every aspect of economic and social life in the country. Most of these analyses cover 73 provinces, rather than 76, as it is difficult to use missing values in this type of data processing.

11As a general rule, a principal component analysis (PCA) was carried out, followed by a hierarchical classification in order to define clusters of provinces. The PCA defines factors determined to best illustrate the selected variables (on average 75 to 85 % of the total variance). By synthesis analysis, the provinces are then defined by their coordinates in relation to these factors, according to the variables that they illustrate. In this way similarities and differences, associations and oppositions between variables can be determined. The hierarchical classification is based on these principal components.

12* Linkage trees* were prepared using Ward’s method (ascending hierarchic classification). The aim is to identify classes that are relatively homogeneous based on characteristics (the variables) defined by an algorithm which begins with each observation (the provinces) in a separate class and groups the classes together until only one remains. Ward’s method is a method of hierarchical classification based on analysis of the variance to evaluate the distances between the groups formed, seeking to minimize the sum of the squares of the distances within each possible group. It is considered to be effective, tending as it does to create groups that are small in size, which can then be mapped directly. In the example on plate 71 (from plate 57 Tertiary sector facilities and activities), the classes that are closest together are 8 and 9: they would form a single class if we decided to present only 8 classes in all. The subsequent grouping concerns classes 5 and 6, then class 3 which attaches itself to this last group, then class 4 which joins the cluster (3, 5, 6), then 7, then 2. The two clusters (8, 9 and 2, 3, 4, 5, 6, 7) then join together. The last stage groups together class 1 (Bangkok) with all the other classes; this gives some indication of the very unusual features of this province.

13* The legend tables* were produced by analyzing class averages for the most significant variables. They present in schematic form the relative values of these averages, compared with the average for all the provinces. In the example, access to water supply via a public well is very much lower than the average for all the provinces in classes 1 and 2, a great deal lower in classes 3 and 5, and even in 6, equal to average in 4 and 7 and very much higher than average in 8 and 9. The “+” signs are not a mechanical indication of the development. Thus, the number of inhabitants per post office is very high for classes 8, 9 and even for 2, average for 1 and 4 and very low for classes 3, 5, 6 and 7, which are thus the best equipped in this matter.

## 4 - Graphic models

14The purpose of a graphic model is exactly the same as that of spatial analysis based on statistical data processing. It is produced to understand the organization of a given area and the rules that structure its spatial differentiation. The premise of the method is that the geographical space is produced, organised and structured by society (though without ignoring the role that physical conditions have to play). It is therefore possible to define elementary structures or choremes (Brunet, Roger. 1980. “La composition des modèles dans l’analyse spatiale”, *L’Espace Géo-graphique*, n° 4: 253-265) and examine how they combine together; in other words, find the structure of the structures.

15The choreme is a dual figure:

it is a fundamental arrangement of objects in a given space, an elementary structure like, for instance, the gravitation of secondary centers around a main center (conceptual dimension);

it is the representation of this concept by a hierarchy of points and an orbital circle (graphic dimension). The graphic representation of the choreme is the layout of the conceptual structure and not a simplified map.

16In graphic modeling, an experimental method governed by a series of hypotheses, we must decide in a rational manner how to represent as clearly as possible a spatial organization as a combination and composition of elementary structures or choremes. There are five stages:

look for the significant elements in the complexity of reality;

analyze them and see how they inter-relate;

control the representation techniques;

suggest a logical and coherent whole;

reach a degree of generalization so that a comparative analysis can be carried out.

# Calculations and conversion rates

17Population growth rates were calculated using a formula specific to this field:

18(((J21^ (1/20))* (I21^ (-1/20)))-1)*100.

19Here, for the growth rate between 1970 and 1990 and the province with the code 21, J21 represents the population in 1990 and I21, the population in 1970.

20Using population growth rates between the two censuses of 1990 and 2000, the 1996 population was extrapolated using this same formula: it is used for all calculations involving a ratio per inhabitant or *per capita*. The true growth rate was probably not regular between these two dates but this method appears to be more reliable and coherent than using figures from the Ministry of Interior based on estimates of registered provincial populations (voluntary formality by residents). These estimates are by no means perfect, and the same is true for the censuses, since for example, migrants working and living away from home in another province declare their residence to be in their village of origin.

21The multiplying coefficient of the number of industrial units between 1986 and 1996 was calculated as “the number of units in 1996 / the number of units in 1986”. This ratio seemed clearer than a proportion for expressing the spatial dynamics of industrialization.

22The conversion rates used were:

for areas, from

*rai*into ha: n rai x 0.16for money, the exchange rates from Baht into US$ established by the Bank of Thailand for the year(s) analyzed.

Le texte seul est utilisable sous licence Licence OpenEdition Books. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.

### Territories and urbanisation in South Africa

Atlas and geo-historical information system (DYSTURB)

Frédéric Giraut et Céline Vacchiani-Marcuzzo

2009

### Ouagadougou (1850-2004)

Une urbanisation différenciée

Florence Fournet, Aude Meunier-Nikiema et Gérard Salem (dir.)

2008

### Hà Nội, future métropole

Rupture dans l’intégration urbaine des villages

Sylvie Fanchette (dir.)

2015

### Atlas des pêcheries thonières de l’océan Atlantique / Atlas of Atlantic Ocean Tuna Fisheries / Atlas de pesquerías de túnidos del océano Atlántico

Alain Fonteneau

2009

### Atlas des pêcheries thonières de l’océan Indien / Atlas of Indian Ocean Tuna Fisheries

Alain Fonteneau

2010

### Atlas des pêcheries côtières de Vanuatu / Coastal Fisheries Atlas of Vanuatu

Un bilan décennal pour le développement / A 10-year Development Assessment

Espérance Cillaurren, Gilbert David et René Grandperrin

2002

### Hà Nội, a Metropolis in the Making

The Breakdown in Urban Integration of Villages

Sylvie Fanchette (dir.)

2016