Georeferenced data as a tool for monitoring the concentration of population in Finland in 1970–1998

Georeferenced data as a tool for monitoring the concentration of population in Finland in 1970–1998 JARMO RUSANEN, TOIVO MUILU, ALFRED COLPAERT AND ARVO NAUKKARINEN Rusanen, Jarmo, Toivo Muilu, Alfred Colpaert & Arvo Naukkarinen (2003). Georeferenced data as a tool for monitoring the concentration of population in Finland in 1970–1998. Fennia 181: 2, pp. 129–144. Helsinki. ISSN 00150010. Changes in the distribution of population in Finland over the period 1970– 1998 are examined in terms of a co-ordinate system of 1 x 1 km grid cells. The results indicate that this system provides suitable areal units for a variety of statistical and GIS methods aimed at describing and explaining the spatial distribution of population. The system is capable of yielding more detailed information than heretofore on topics such as the concentration of population in the urban centres of Finland – a process that has been going on since the beginning of the last century, but has slowed down noticeably in the recent years. Jarmo Rusanen, Toivo Muilu, Alfred Colpaert & Arvo Naukkarinen, Department of Geography, PO BOX 3000, FIN 90014 University of Oulu, Finland. Email: [email protected]. MS received 07 January 2003. Introduction Demographic research has a long tradition within geography and represents a discipline that is eminently suited to the study of the spatial distribution of population and its changes. The results of these studies are required for planning purposes in many sectors of the society, e.g. determination of the capacity and location requirements for various public services at the national, regional or local level. Although it is primarily the authorities that have been interested in the direction of demographic trends, it is by no means a matter of indifference for individual citizens or families whether they live in an area of expanding or diminishing population. The general image of an area is to a considerable extent shaped by trends in population, as well as with structural changes of the area. There is thus an obvious need for statistical data that are both up to date and geographically more detailed than earlier, and for precise interpretations of such data. This poses new challenges not only for the producers and users of the statistics but also for those engaged in developing and applying analytical and interpretative methodologies. The geographical examination of population distributions and their temporal changes has typically relied on notions such as spatial concentration and deconcentration (Lipshitz 1996; Borgegård et al. 1995), or urbanization and counterurbanization (Håkansson 2000), while migration research has adopted perspectives such as that of the ‘turnaround’ phenomenon (Long & Nucci 1997; Lewis 2000). The actual terms in use vary not only according to the theoretical basis from which the research sets out but also in relation to the volume of migration concerned and its temporal duration. Our picture of the spatial concentration and agglomeration of population in Finland has largely been shaped by the publications of Hustich (1977) and Alestalo (1983). More recently, however, the present authors have discussed changes in the settlement structure and the concentration of population with the aid of georeferenced data and GIS methods (see Naukkarinen et al. 1993; Rusanen et al. 1997). As pointed out by e.g. Håkansson (2000), research into the distribution of population can be greatly affected by the employed geographical unit, so that models that apply at one spatial lev- 130 FENNIA 181: 2 (2003) Jarmo Rusanen, Toivo Muilu, Alfred Colpaert and Arvo Naukkarinen el do not necessarily hold good at another. One particularly difficult problem concerns boundary changes, which indicates that census statistics for different years do not necessarily apply to the same areal units and therefore, cannot be compared easily (Howenstine 1993; Openshaw 1995; Rees 1998). The requirements placed on population statistics and the usefulness of the data available can vary greatly from one group of users to another. Public authorities, for instance, are accustomed to using statistical means and definitions expressed in terms of administrative areas and deductions made from these when they require information on which to base their decisions. Researchers, in contrast, prefer to go back to the sources of the original raw data in order to test their hypotheses and construct their models. For this reason we will attempt to analyse and examine demographic data using various methods in terms of both administrative and non-administrative areal units, and to evaluate the subsequent results. Aims, data and methods We set out here to describe 1) the changes in population that have taken place within the spatial and settlement structure of Finland by means of 1 x 1 km grid cell data, 2) the methods available for the analysis of population data, and 3) the factors highlighted by the grid cell method that cannot be identified when using administrative areal units. Finally, the usefulness of the resulting information for the purposes of demographic research will be evaluated. Georeferenced data on Finland are available at many spatial levels. In practice, any individual citizen can be assigned a set of coordinates describing the geographical location of the house in which he or she lives. The resulting data can be aggregated to represent higher-level units, e.g. postal districts, parts of municipalities, whole municipalities, groups of municipalities or the NUTS hierarchy as used by the EU. It must be remembered, however, that the above-mentioned problem of comparability over time emerges in the case of these areal units whenever administrative areas are combined or boundaries are altered. On the other hand, the fact that the initial information is available at the individual level is regarded as a significant advantage of GIS methods (Martin & Higgs 1997; see also Openshaw & Turton 1996). Thus, the principal units employed here are the 1 x 1 km grid cells defined in the data produced by Statistics Finland for the years 1970, 1980 and 1990–1998 (see Rusanen et al. 1997). These grid cells are of fixed location and are independent of any changes in administrative boundaries. Statistical information on human activities in Finland has been available annually since 1987. Current methods enable Statistics Finland to revise their georeferenced population data very quickly, so that figures for the population on 31st December are available by the end of the following February, together with statistics for the individual municipalities. This annual information is based on registers, so that no separate population census is necessary as in the USA, for example (see Crews 2000). The second basic areal unit employed here is the municipality, or independent local government district. These districts have numerous rights and obligations that define them as administrative entities, including the right to levy taxes (Ministry of the Interior 2000). In accordance with the above principle, the populations of the municipalities are calculated here from the grid cell figures, and for this reason the figures differ slightly from those quoted in the official statistics, as for instance in 1998, 1.42% of the national population could not be assigned coordinates on the basis of their place of residence. However, this discrepancy can be assumed not to affect the results. It should be noted that we are not concerned here with studying the potential factors responsible for demographic trends in particular areas, e.g. birth-rates, mortality or migration – topics which have traditionally been to the fore in population geography (Ogden 1998; 1999). Instead, we focus on describing solely the concentration of population. This is defined here in two ways: as an increase in population density per square kilometre when using grid cell data, and as an increasingly larger proportion of the population living in a progressively smaller number of municipalities. Data on population changes in the 1 x 1 km grid cells and municipalities are analysed here by a number of methods, and assessments are made as to whether these methods yield similar or divergent results. The methods tested include the Gini Coefficient, which is more commonly used for analysing the spatial distribution of income FENNIA 181: 2 (2003) 131 Georeferenced data as a tool for monitoring the… rates (see Chakravorty 1994) rather than population (see Bradford & Kent 1986; Lovell-Smith 1993). In addition, the concentration of population is described in terms of mean values and the Moran and Geary Indices used in version 8.0 of the Arc/Info software. Changes in population are also described by reference to deciles and by classification of the ‘spatial demographic structure’. Long-term patterns are visualized by means of a rank size model and the kriging interpolation method contained in version 8.0 of Arc/Info. The Hoover Index that is frequently used in studies of population concentration (see Borgegård et al. 1995; Long & Nucci 1997) is not employed here, as the emphasis is on analysis at the grid cell level rather than at the municipality level. Decile analysis involves ranking the data units in size order from largest to smallest with respect to the studied variable, and dividing them into ten equal groups in terms of population described by the variable, regardless of whether the areal unit is a grid cell or a municipality. Population concentration as reflected in the Moran and Geary indices and the Gini coefficient Population density in Finland in 1998 relative to land area was very low by European standards, only 16.9 inhabitants per square kilometre. If, however, we use the grid cell material to calculate the mean density for inhabited cells only, we see that the density has increased steadily over the period examined here, and is now approaching 50 inhab./km2 (Table 1). The Moran and Geary methods allow a spatial autocorrelation index to be determined for the grid cell material as follows: when the Moran Index is positive and the Geary Index varies in the range 0–1 the distribution of population can be described as ‘similar’, ‘regionalized’, ‘smooth’ or ‘clustered’. If the Moran Index is negative and the Geary Index greater than 1, the corresponding terms are ‘dissimilar’, ‘contrasting’ and ‘checkerboard’ (for details, see Arc/Info User Manual 2000). In the present case, the gradual increase in the Moran Index and the simultaneous decrease in the Geary Index lend further support to the impression gained from the previous table of a process of consolidation of the inhabited area rather than a scattering or dispersal of population (Table 2). One typical technique employed to describe evenness or concentration in the distribution of population is the Lorenz diagram (Alestalo 1983), or its derivative, the Gini Coefficient (Fainstein 1996). Regardless of whether we use grid cells or municipalities as the areal units, the obtained Gini Coefficients indicate a concentration of population (Table 3), although the higher value of the grid cell material as compared to the municipality level also indicates the essentially local nature of this concentration. The fact that the majority of the population of Finland is located within a relatively small number of grid cells is indicative of the low proportion of built-up areas within the country’s total settled area. Measured in both of the above ways, the rate of change was greatest in the 1970’s, an observation that confirms the general impression regarding the concentration of population in Finland. It is consistent with the results of the examination by deciles published by Alestalo (1983). On the other hand, analysis of the trend over the last two decades in terms of the Gini Coefficient provides deviant results for the two sets of areal units. The rate of population concentration apparently has slowed down when assessed in terms of the grid Table 1. Number of inhabited grid cells and mean population density in Finland in 1970–1998 (Data: Statistics Finland). 1970 1980 1990 1992 1994 1996 1998 Number of inhabited grid cells 110477 104540 103242 103020 103036 103045 102873 Inhab./km241.3 44.1 47.8 48.4 48.9 49.1 49.4 Table 2. Moran (I) and Geary (C) Indices in 1970–1998 (Data: Statistics Finland). Year 1970 1980 1990 1992 1994 1996 1998 I 0.543 0.531 0.568 0.570 0.581 0.590 0.597 C 0.427 0.435 0.400 0.394 0.384 0.377 0.370 132 FENNIA 181: 2 (2003) Jarmo Rusanen, Toivo Muilu, Alfred Colpaert and Arvo Naukkarinen cell data, being only 1.1% in the 1990’s, whereas the figures for the municipalities indicate acceleration with a terminal rate of 3.5%. The result obtained from the grid cell data may indicate a decreasing trend in population concentration in the densest areas, the pattern being attributable to the exclusion of uninhabited grid cells and the concentration of population in a constantly decreasing number of cells and a more restricted geographical area. This trend, which is largely internal to individual municipalities, fails to be reflected in the analyses based on municipalities as the areal units. Decile analysis Decile analysis employs divisions of the total material into tenth parts. It can be regarded as a flexible, non-given means of classification, and since the boundaries of the deciles tend to vary from one year to the next, the method is well suited to studies of concentration or dispersal, e.g. in population or incomes. According to Alestalo (1983), the population of Finland was fairly evenly distributed over the deciles at the end of the 19th century, the agrarian society of the time showing little agglomeration of the population into urban centres. From that time onwards, however, the population gradually became concentrated in a smaller number of municipalities (Table 4), so that by the 1970’s, half of the country’s population lived in 52 municipalities, 10.1% of their total number. The trend continued so that in 1998, the corresponding figure was 33 municipalities, i.e. 7.3% of the total. Part of this effect may be attributed to the amalgamation of municipalities within the system of local government, but the principal factor has without doubt been the actual concentration of population within a progressively smaller number of towns and cities (Rusanen et al. 2000). Is this concentration visible in the grid cell data? For answering this, a decile analysis comparable to that in Table 4 is presented in Table 5. The deciles indeed indicate a continuation of the concentration process, so that half of the country’s population (the population of deciles 1–5) occupied a total of 1541 km2 in 1970, 1167 km2 in 1980, 1294 km2 in 1990 and 1284 km2 in 1998. The 1970’s were a period of heavy population concentration, whereas from the end of that decade onwards Finland was affected by a ‘turnaround’ phenomenon, as noted by several authors (Kauppinen 2000). This process was experienced in many western countries, entailing above all a Table 3. Concentration of population as shown by the Gini Coefficients for the years studied and percentage change over the period 1970–1998 (Data: Statistics Finland). Year Areal unit 1970 1980 1990 1998 Grid cell 0.78558 0.83450 0.85365 0.86324 Municipality 0.59228 0.62145 0.62962 0.65186 Change in Gini Coefficient (%) 1970–1980 1980–1990 1990–1998 Grid cell 6.2 2.3 1.1 Municipality 4.9 1.3 3.5 Table 4. Concentration of population by deciles of municipalities in 1970, 1980, 1990 and 1998 (Data: Statistics Finland). Number of municipalities Year Decile 1970 1980 1990 1998 1 1 1 2 1 2 4 4 3 3 3 6 5 5 4 4 151010 9 5 26201816 6 38292723 7 50424038 8 66585756 9 97888788 10 212 204 206 214 Total 515 461 455 452 FENNIA 181: 2 (2003) 133 Georeferenced data as a tool for monitoring the… decline in the popularity of urban areas as living environments for families with children. The inhabited area of Finland in 1998 amounted to ca. 30.4% of the total surface area of 338 145 km2, with a consistent decline of this proportion from one decade to the next since 1970. The most pronounced decline, almost 6000 km2, took place during the 1970’s, although it has also been claimed that the material for 1970 contained some errors in the coordinates determined for individual dwellings and that the actual net decrease was not necessarily as great as this. It should also be noted that the trend was a relatively steady over the period 1990–1998 relative to the 1980’s, although data on the final years of the decade are still lacking. Changes may also be observed in the population densities for the deciles (Table 6). The highest densities of all were recorded in deciles 1 and 2 in 1970, when Finland was experiencing a pronounced migration from the countryside into the towns and also abroad, primarily to Sweden (see Kauppinen 2000). At that time the mean population density in the top decile was around 9700 inhab./km2, whereas the corresponding figure in 1998 was about 6200. The maximum population density was reached in Helsinki, the capital. In the 1970 data the density peaked at 29 234 inhab./km2, whereas the figure for 1998 showed a reduction of more than a third, 19 172 inhab./ km2. A number of factors can be identified that contributed to the reduction of population density in Finland over the study period. The most significant factor was the transfer in the urban centres from dwellings to other uses, mainly offices and commercial premises. Other factors were changes in the structure of households, including reductions in the average family size and the number of families with children and an increase in the number of single-person households. It should be pointed out, however, that the mean population density of the two most densely inhabited deciles began to increase again in the 1990’s, partly on account of the efforts made to fill in the settlement pattern in built-up areas in accordance with the principles of sustainable development. Andersson (1988) recognizes four stages in the development of towns in Finland: urbanization, suburbanization, disurbanization and reurbanization. Urbanization stage refers to the growth of a central urban nucleus, while the suburbanization stage involves a slowing down and eventually cessation of this trend. At the disurbanization stage, Table 5. Numbers of inhabited grid cells by deciles in 1970, 1980, 1990 and 1998 (Data: Statistics Finland). 1970 1980 1990 1998 Inhab. km2 Decile N N N N in 1998 1 densest 47 66 84 82 4171–19172 2 settlement 116 133 159 158 2633–4170 3 207 200 234 234 1795–2632 4 384 297 336 334 1257–1794 5 787 471 481 476 881–1256 6 1 856 825 751 721 536–880 7 4 992 1 928 1 420 1 299 267–535 8 11 290 6 816 4 635 3 813 70–266 9 sparsest 22 331 19 590 16 579 14 980 20–36 10 settlement 68 462 74 209 78 561 80 770 1–19 Number of grid cells 110 472 104 535 103 240 102 873 Table 6. Population density by deciles in 1970, 1980, 1990 and 1998 (Data: Statistics Finland). Inhab. km2 Decile 1970 1980 1990 1998 1 9703 6980 5876 6202 2 3932 3464 3104 3219 3 2203 2303 2109 2173 4 1188 1551 1469 1523 5 579 978 1026 1068 6 246 558 657 705 7 91 239 348 392 8 40 68 106 133 9 20243034 10 7666 134 FENNIA 181: 2 (2003) Jarmo Rusanen, Toivo Muilu, Alfred Colpaert and Arvo Naukkarinen the trend is reversed, until the reurbanization stage marks new growth in the urban nucleus. It is interesting to consider how these stages might be reflected in an empirical decile analysis. The urbanization stage seems to have taken place in Finland before 1970, as the population of the most densely inhabited central areas (deciles 1 and 2) began to decline after that time. Hence, the 1970’s and 1980’s may be assigned to the suburbanization and disurbanization stages, as these figures in particular declined markedly at first and then levelled out somewhat in the 1980’s, when the changes were less pronounced in other respects, too. The 1990’s represent the reurbanization stage, in which the population of the most densely inhabited areas again started to increase. It is important to bear in mind when evaluating these findings that the stages are distinguished on the strength of only a single variable. A more precise consideration would call for an internal analysis of the structure of the ten largest cities, for example, and also other information relevant to the growth of urban areas, e.g. their development and planning policies. Population changes by deciles in 1990–1998 Dual trends in population density and in the number of inhabited grid cells are observable during the 1990’s, the cut-off point being reached in 1993. The minimum area occupied by half of the population of Finland, i.e. deciles 1–5, increased numerically over that time (Table 7a), implying a slight decline in population density in the urban nuclei and suburbs (Table 7b). From 1993 onwards, the trend towards denser communities can be observed, although not quite amounting to the figures recorded in 1970 and 1980. The most sparsely populated rural areas (decile 10) increased in number throughout the 1990’s, whereas their mean population density remained more or less stable. The increase may be attributed almost entirely to reductions in the population of some grid cells previously contained in decile 9, representing the rural areas proper, causing their transfer to decile 10. The reversal in the trend in 1993 is probably linked to the fact that this was the worst year of the economic recession in Finland, i.e. the periTable 7. Changes in the number of grid cells (A) and population density (B) in the 1990’s (Data: Statistics Finland). A. Number of grid cells Trend Decile 1990 1991 1992 1993 1994 1995 1996 1997 1998 1990– 1994– 1990– 1993 1998 1998 1 84858686858483 8282+– 2 159 162 164 164 163 161 160 160 158 + – – 3 234 238 241 241 240 239 236 235 234 + – 0 4 336 339 342 344 343 339 338 336 334 + – – 5 481 484 487 488 485 482 480 477 476 + – – 6 751 751 750 751 744 736 728 724 721 0 – – 7 1420 1406 1400 1390 1364 1345 1329 1316 1299 – – – 8 4635 4500 4409 4332 4214 4096 4012 3915 3813 – – – 9 16579 16305 16105 15930 15746 15548 15380 15204 14980 – – – 10 78561 78815 79038 79314 79652 80007 80299 80583 80770 + + + B. Population density inhab./km2 1 5876 5840 5802 5831 5923 6015 6100 6191 6202 – + + 2 3104 3064 3043 3058 3089 3138 3165 3173 3219 – + + 3 2109 2086 2071 2081 2098 2114 2145 2160 2173 – + + 4 1469 1464 1459 1458 1468 1491 1498 1511 1522 – + + 5 1026 1026 1025 1028 1038 1048 1055 1064 1068 + + + 6 657 661 665 668 677 687 695 701 705 + + + 7 348 353 356 361 369 376 381 386 391 + + + 8 106 110 113 116 119 123 126 130 133 + + + 9 30303131323233 3334+++ 10 6,3 6,3 6,3 6,3 6,3 6,3 6,3 6,3 6,3 0 0 0 FENNIA 181: 2 (2003) 135 Georeferenced data as a tool for monitoring the… od in which the unemployment peaked. Similarly, housing production declined steadily, to reach its lowest ebb in 1994. This was followed by a period of economic recovery (SVT 1999), with a new stimulation of building activity, especially in the metropolitan area of Helsinki and other large municipalities. This trend was accompanied by a rise in population densities, as the housing capacity of the built-up areas increased while the land areas concerned remained more or less constant. The increased housing density was hence reflected in the population density. The above observation demonstrates the usability of the grid cell data as an aid in monitoring changes in spatial structure, and potentially achieving detailed explanation. This method allows, for instance, observing small changes in population within a municipality very easily. The spatial demographic structure in 1970–1998 The repeatability of the decile method allows it to be applied to any country or any areal unit, and any researcher can arrive at the same results. It should be remembered, however, that such accurate geofererenced data are available in selected countries only. We will turn our attention now to structural features of the spatial distribution of population and the changes detected in the distribution in the cross-sectional data for 1970, 1980, 1990 and 1998, with particular reference to developments during the 1990’s. The classification employed may be referred to as the ‘spatial demographic structure’, as it represents an attempt to describe the relation between population density and settlement structure. Hence, it does not take into account the functional elements normally implied in the term regional structure, e.g. dwellings, jobs and the aspects of infrastructure that support these. The concept has been used earlier for classification purposes by Räisänen et al. (1996) and Rusanen et al. (1997). We will first consider the situation over the whole of Finland and subsequently concentrate the analysis on one specific region, Kainuu. The interpretation provided here is based on the assumption that the character of a grid cell can be deduced from the size of its population. The classification concerned is not necessarily applicable to all areas or to all countries, and it has been constructed knowing well that the distinction between rural and urban is by no means unambiguous (see Malinen et al. 1994; Berry et al. 2000). It is impossible to build a model that would apply equally well to all countries and under all conditions. The following classification of spatial demographic structure is used: Inhab./km2Element of spatial demographic structure 1–5 Scattered settlement 6–20 Rural areas proper 21–100 Rural areas with built-up features 101–1000 Build-up areas and suburbs with mostly private housing More than High-rise centres and suburbs of 1000 major cities Examination of the situation in the years mentioned above indicates that the most densely populated areas, with over 1000 inhab./km2, grew most rapidly in the 1970’s, the rate of growth diminishing in the 1980’s and reaching its slowest in the 1990’s (Fig. 1). The areas of suburban private housing departed from this pattern somewhat, however, as these underwent their greatest population growth in the 1980’s. By contrast, the transitional category between rural and urban conditions, that with densities of 21–100 inhab./ km2, declined in population throughout the period studied here, although most markedly in the 1970’s and least so in the 1990’s. The rural areas proper decreased substantially in extent, while the areas of scattered settlement increased somewhat in both total population and extent, largely as a result of contractions in the population of grid squares previously included in the category of rural areas proper. The trend in population density over the whole country in the 1990’s was polarized. The population of the densely inhabited areas increased steadily throughout the decade, amounting to a total rise of 4.6%, or 110 000 persons, between 1990 and 1998 (Fig. 2). The suburban private housing population grew in a similar manner, with the exception of a small decline in 1997. The population of the urban-rural transition zone with densities of 21–100 inhab./km2, declined from 1993 onwards, however, and simultaneously the area contracted slightly. The category of rural areas proper declined in both population and total area throughout the studied period, whereas scattered settlement category increased slightly in both total population and area. 136 FENNIA 181: 2 (2003) Jarmo Rusanen, Toivo Muilu, Alfred Colpaert and Arvo Naukkarinen Fig. 1. Population changes in different parts of the spatial demographic structure over the whole of Finland in 1970–1998 (Data: Statistics Finland). Fig. 2. Population changes in different parts of the spatial demographic structure over the whole of Finland in 1991–1998 (Data: Statistics Finland). The Kainuu region in Northern Finland, selected here for more detailed examination, is characterized by economic and demographic recession. It belongs to the Objective 1 EU support areas on account of its low income levels and sparse settlement. Its population, which has been on the decline since the 1960, was 93 218 persons in 1998, and the region has consistently been one with pivotal unemployment in the whole country. The demographic trend in Kainuu during the 1990’s departed markedly from that obtained for the whole country (Fig. 3). The most densely inhabited areas experienced a population decline from 1992 onwards, and the same trend affected the private housing areas from 1996 onwards. Correspondingly, the transition zone and the rural areas proper lost population throughout the decade, as in the whole country, and the population increase in the areas of scattered settlement FENNIA 181: 2 (2003) 137 Georeferenced data as a tool for monitoring the… appears to have come to an end in 1998. This was probably the first occasion during the century when the population of Kainuu decreased in every single category of its spatial structure. It is probable that demographic trends in the sparse population density classes in Fig. 9 were the same as in Kainuu. Elsewhere they have resembled those of the whole country. Finland became a member of the EU in 1995. Kainuu was an Objective 6 area in 1995–1999 and is currently an Objective 1 area for the period 2000–2006. Based on the above results one can argue that the development measures implemented in the region under the EU programmes have failed to improve the negative demographic trend in all elements of the spatial system. Detailed analyses of population trends are possible only with the aid of georeferenced data that allow precise location of the population units. A ‘sliding scale’ evaluation of population density based on local level data helps avoiding the fallacies that arise when one examines the whole country or large aggregate areas (Martin 1991). The unravelled negative demographic trend in Kainuu serves as an example of the use of grid cell data for monitoring regional development. The rank size model Rank size models represent a well-established method of geographical research that has been applied to the study of hierarchical structures in various areal units and changes taking place in these structures. The units employed for this purpose are usually towns or other administrative or functional entities (see, e.g. Bradford & Kent 1986; Das & Dutt 1993). The method involves arranging the data units in size order and presenting the results in diagrammatic form. We intend to apply the rank size approach to the 1 x 1 km2 grid cells, and use the density classes for describing the types of settlement and any changes in population. For comparison purposes, the approach was done at two levels: the whole country and the Kainuu region. The 452 municipalities of Finland varied in population from a mere 100 up to 500 000 inhabitants in 1998 (Fig. 4). With the municipality level as the areal unit of interest, the rank size model indicates that about one hundred largest municipalities showed an increase in population over the period 1970–1998, while the others experienced population decline. At the grid cell level, the rank size model fails to provide detailed information of spatial demographic structure as efficiently the classification (Figs. 5 and 6). It does, however, serve well in highlighting the decline in the total number of inhabited grid cells over the whole country, i.e. the contraction in the area of human settlement. Furthermore, it emphasizes certain major turning points such as the beginning of the decline in population in the most densely inhabited areas Fig. 3. Population changes in different parts of the spatial demographic structure in Kainuu in 1991–1998 (Data: Statistics Finland). 144 FENNIA 181: 2 (2003) Jarmo Rusanen, Toivo Muilu, Alfred Colpaert and Arvo Naukkarinen Haartsen T, P Groote & PPP Huigen (2003). Measuring age differentials in representations of rurality in The Netherlands. Journal of Rural Studies 19, 245–252. Håkansson J (2000). Changing population distribution in Sweden. GERUM – Kulturgeografi 2000: 1. Howenstine E (1993). Measuring demographic change: the split tract problem. Professional Geographer 45: 4, 425–430. Hustich I (1977). Suomen “kutistumisen” vauhti hiljenee (The speed of the “shrinkage” of Finland is slackening). Terra 89: 4, 214–221. Kauppinen J (2000). Muuttoliike Suomessa vuosina 1989–1994 koordinaattipohjaisten paikkatietojen perusteella (English summary: Migration of population in Finland in 1989–1994, examined from coordinate-based GIS data). 155 p. Siirtolaisuusinstituutin tutkimuksia A 22. Langford M & DJ Unwin (1994). Generating and mapping population density surfaces within a geographical information system. The Cartographic Journal 31, June, 21–26. Lewis G (2000). Changing places in a rural world. The population turnaround in perspective. Geography 85: 2, 157–165. Lipshitz G (1996). Spatial concentration and deconcentration of population: Israel as a case study. Geoforum 27: 1, 87–96. Long L & A Nucci (1997). The Hoover Index of population concentration: A correction and update. Professional Geographer 49: 4, 431–440. Lowell-Smith EG (1993). Regional and intrametropolitan differences in the location of freestanding ambulatory surgery centers. Professional Geographer 45: 4, 398–407. Malinen P, R Keränen & H Keränen (1994). Rural area typology in Finland – a tool for rural policy. 78 p. University of Oulu, Research Institute of Northern Finland. Research reports 123. Martin D (1991). Geographic information systems and their socioeconomic applications. 182 p. Routledge, New York and London. Martin D & G Higgs (1997). Population georeferencing in England and Wales: basic spatial units reconsidered. Environment and Planning A 29: 2, 333–347. Martin R (2000). In memory of maps. Transactions of the British Geographers, New Series 25: 1, 3–5. Ministry of the Environment (2000). <http://www. vyh.fi/ajankoht/tiedote/ym/tied99/ ym12299.htm>. 15.12.2000 Ministry of the Interior (2000). <http://www: intermin.fi/eng/municipal/municipalities.html>. 15.12.2000 Muilu T & J Rusanen (2003). Rural definitions and short-term dynamics in rural areas of Finland in 1989–1997. Submitted manuscript. Naukkarinen A, J Rusanen & A Colpaert (1993). The future of rural settlement structure in Finland. Geography Research Forum 13, 61–70. Ogden PE (1998). Population geography. Progress in Human Geography 22: 1, 105–114. Ogden PE (1999). Population geography. Progress in Human Geography 23: 4, 617–631. Openshaw S (1995). The future of the census. In Openshaw S (ed). Census User’s Handbook , 389– 411. GeoInformation International, Glasgow. Openshaw S & I Turton (1996). New opportunities for geographical census analysis using individual level data. Area 28: 2, 167–176. Räisänen S, J Rusanen & A Naukkarinen (1996). Socio-economic grid data and GIS for analyzing the changes in the Finnish countryside. Second Joint European Conference & Exhibition on Geographical Information, March 27–29, 1996. Barcelona, Spain. Geographical Information. From Research to Application through Cooperation Vol 1, 651– 660. Rees PH (1998). What do you want from the 2001 Census? Results of an ESRC/JISC survey of users views. Environment and Planning A 30, 1775– 1796. Rusanen J, A Naukkarinen, A Colpert & T Muilu (1997). Differences in the spatial structure of the population between Finland and Sweden in 1995 – a GIS viewpoint. 46 p. Statistics Finland Research Reports 221. Rusanen J, T Muilu, A Colpaert & A Naukkarinen (2000). Asutun Suomen väestömäärä ylittää 50 asukasta neliökilometrillä. Tietoaika 5/2000, 14– 17. Silvasti T (2002). Maaseudun muuttuvat merkitykset – tuotannon tilasta kulutuksen kohteeksi. (Changing meanings of the countryside – from a production space to an object of consumption.) Yhdyskuntasuunnittelu (Community Planning) 40, 11–25. Wheeler J O (1998). Mapphobia in Geography? 1980–1996. Urban Geography 19: 1, 1–5. SVT (Suomen virallinen tilasto) (OSF = Official Statistics of Finland) (1999). Rakentaminen ja asuminen. Vuosikirja 1999. (Construction and Housing. Yearbook 1999). Rakentaminen 1999: 20. 226 p. Tilastokeskus. Statistics Finland.