Landscape Ecol Eng (2007) 3:21–31 DOI 10.1007/s11355-006-0008-6
ORIGINAL PAPER
Land use/cover change and its drivers: a case in the watershed of Lake Kasumigaura, Japan Takehiko Fukushima Æ Masae Takahashi Æ Bunkei Matsushita Æ Yoshinori Okanishi
Received: 21 October 2005 / Revised: 31 March 2006 / Accepted: 10 August 2006 / Published online: 15 September 2006 International Consortium of Landscape and Ecological Engineering and Springer 2006
Abstract Land use/land cover (LULC) changes in the watershed (2,157 km2) of Lake Kasumigaura during 1979–1996 (Period-1: 1979–1990, Period-2: 1990–1996) were analyzed, and their socio-economic and biophysical drivers were compared using time-series, highquality GIS datasets in order to examine the characteristics of a model forecasting the future LULC. The changes occurred over an area of more than 90 km2 during the overall period at changing rates of 0.22% year–1 in Period-1 and 0.25% year–1 in Period-2. Forestland decreased most in both periods at changing rates of 0.45% year–1 in Period-1 and 0.61% year–1 in Period-2. However, predominant changing patterns differed, i.e., from forest to golf course in Period-1 and from forest to artificial field in Period-2. Particularly in Period-2, a significant LULC change was observed in an area of high population increase on the edge of an already high-population area. Relationships examined among LULC change, population, and rate of population change suggested that the urbanized area was highly resistant to LULC change, and that such change was less frequent in areas of population decline. Statistical analyses indicated that the most influential
T. Fukushima (&) Æ B. Matsushita Graduate School of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba 305-8572, Japan e-mail:
[email protected] M. Takahashi Ministry of the Environment, Tokyo, Japan Y. Okanishi Master Program in Environmental Sciences, University of Tsukuba, Tsukuba, Japan
drivers for total LULC changes were population in Period-1 and distance from the Tokyo Station in Period-2. Since the change potentials differed between the periods, we could not assume a stationary process for the corresponding drivers. Somewhat low S values (indices for demonstrability) show that LULC changes in the watershed of Lake Kasumigaura occurred rather randomly, probably resulting in fragmentation of the landscape. Keywords LULC change Æ Lake watershed Æ Biophysical drivers Æ Socio-economic drivers Æ High-quality GIS dataset
Introduction Over the last decade, a variety of analyses and models on land use/land cover (LULC) change have been developed to address land management needs, and to better assess and project the future role of LULC in the functioning of the global/regional ecosystems. Crucial issues to be focused on and evaluated in the terms of LULC change usually include rural housing (Aspinall 2004; Robinson et al. 2005), forest fragmentation (McAlpine and Eyre 2002; Butler et al. 2004), agricultural expansion (Gobin et al. 2002; Soares-Filho et al. 2002; McConnell et al. 2004), suburbanization (Zipperer et al. 2000), and urban sprawl (Lopez et al. 2001; Walker 2001; Luck and Wu 2002; Wilson et al. 2003). A variety of models, e.g., regression, Markov Chain, cellular automata, have been applied to explain LULC changes (Baker 1989; Veldkamp and Lambin 2001). A watershed unit was sometimes employed for analysis since it not only indicates the natural (and
123
22
sometimes socio-economic) boundaries but has a close connection with water/mass cycles (Li et al. 2001; Irwin and Geoghegan 2001). Given that the landscape significantly influences downstream water chemistry (Johnson et al. 1997), the forecasting of future land use is a requisite factor in watershed assessment (Butcher 1999). In the present paper, we analyzed the watershed of Lake Kasumigaura which has suffered from population sprawl on land surrounding the Tokyo Metropolitan Area during the last decades, thereafter indicating fragmentation of the landscape (Matsushita et al. 2006). Furthermore, the watershed has generated nutrients and organic loads in Lake Kasumigaura, causing deterioration of the lake water, i.e., eutrophication. Thus, LULC changes have had a potentially profound impact on water/mass cycles, habitat distribution, lake water quality, etc., in this watershed. A variety of drivers, which include both biophysical (e.g., altitude, slope, soil type) and socio-economic (e.g., population, government policy, accessibility) attributes, have been related to LULC changes. However, only a few studies have analyzed the spatiotemporal patterns of LULC changes with sufficient accuracy and compared both types of attributes quantitatively and simultaneously (McConnell et al. 2004). In his case studies, Ichinose (2005) indicated that temporal stability in LULC changes should not be assumed. Thus, we used ground-based, 50-m-resolution GIS datasets that might improve LULC analysis, since they were often developed by combining intensive field investigation with interpretations of aerial photographs and satellite images; we compared the demonstrability of the drivers during two periods. In addition, we used the term ‘‘LULC’’ because the landscape types shown below were classified based on land use (LU) and land cover (LC) in a combined manner due to the development method of the datasets. Our objectives in this study are: (1) to characterize the LULC changes in the Lake Kasumigaura watershed with an area of more than 2,000 km2, based on the above time-series, high-quality GIS datasets; and (2) to compare the influence of both biophysical (altitude, topography, geology, and soil types) and socio-economic (population, accessibility to roads, stations, etc.) drivers on LULC changes during two periods in order to lay the ground work for a model forecasting future LULC. Materials and methods Study area The watershed of Lake Kasumigaura is located in the eastern part of Japan’s Kanto Plain, and covers an area
123
Landscape Ecol Eng (2007) 3:21–31
of 2,157 km2 (Fig. 1). With a surface area of 220 km2 and an average depth of 4 m, Lake Kasumigaura is known to be a eutrophicated lake. The climate of the area is similar to that of other regions on the nation’s Pacific side, with an annual average air temperature of about 14C and an annual precipitation of 1,250 mm. The predominant landscapes in the watershed are paddy fields (25.8%), forests (22.7%), plowed fields (21.2%), and water (10.6%). The area’s traditional industries are agriculture, livestock management, and fisheries, the products of which have contributed significantly to the development of the Tokyo area due to the fertility of the land and its proximity to the capital city. Moreover, many new industrial plants have been constructed, and industrial shipments from the area continue to rise year after year. Since the study area is only 60 km distant from Tokyo, there has been a rapid rise over the last 20 years in the number of residents, from around 810,000 in 1980, to 875,000 in 1985, 933,000 in 1990, 998,000 in 1995, and 1,026,000 in 2000 (calculated using population census data; Table 1), as well as recreational facilities and resort areas, all of which have contributed to the most rapid changes in the LU profile in decades. On the other hand, a decline in the population and its aging have begun in the northeastern part of the watershed. These trends highlight the practical need for an updated study of landscape changes to facilitate socioeconomic planning and analysis in this rapidly changing area. Production of time-series LULC map based on GIS datasets A vegetation GIS dataset (DS_GIS_NEJ) was compiled by The Natural Conservation Bureau of Japan’s Ministry of the Environment (Natural Conservation Bureau 1999). The dataset was developed based on the green census beginning in 1973 and has been updated about every 5 years. A vegetation survey was carried out for each period of the green census, and vegetation maps were produced for each survey. Field investigations were carried out at the Lake Kasumigaura watershed in 1979 (second survey), 1990 (fourth survey), and 1996 (fifth survey). Time-series LULC maps of these 3 years were then generated, based on the three corresponding vegetation maps (scale 1:50,000, UTM projection). We designated the period from 1979 to 1990 as Period-1 and that from 1990 to 1996 as Period-2. In order to analyze LULC changes with a variety of drivers usually expressed by geographic (lat/long) projection, the vector datasets were then converted to a raster format at resolutions of around 50 m (1.5 s) in latitude and 37.44 m (1.5 s) in
Landscape Ecol Eng (2007) 3:21–31
23
Fig. 1 Lake Kasumigaura and its watershed. LULC map in 1996
Table 1 Datasets used for analysis Name
Responsible organization
Information used
Year
Original format
Size in raster data
Vegetation GIS dataset Population census
Ministry of Environment Ministry of Internal Affairs and Communications Ministry of Land Infrastructure and Transport (MITI) MITI
LULC map Population
1979, 1990, 1996 1980, 1985, 1990, 1995, 2000
Vector Raster
50 · 37.4 m 1 · 1 km
Elevation
1975
Vector
50 · 50 m
Positions of national roads and free ways Positions of national roads and railway stations Positions of railway stations, interchanges of high ways and trunk roads Topographical, geological, and soil types
2002
Vector
50 · 37.4 m
1977, 1979, 1989, 1991 2002, 2003
Map
50 · 37.4 m
Raster
50 · 37.4 m
1975
Raster
1 · 1 km
Digital elevation map
Spatial map (1:25,000) Topological maps (1:25,000) Digital picture maps (1:25,000) National digital information
MITI MITI
MITI
Fig. 2 The regions where LULC has changed. a Period1 and b Period-2
123
24
longitude using a spatial analyst module of ArcGIS 8.2; the total number of grids was 1,205,629. Finally, we reduced the 50 landscape types in the DS_GIS_NEJ vegetation map to 14 new landscape types: water, artificial field, residential area, industrial area, urban area, park, paddy field, uncultivated paddy field, plowed field, golf course, orchard, artificial grassland, natural grassland, and forest. In this study, ‘‘artificial field’’ includes certain commercial areas separated from the continuous blocks of ‘‘urban areas’’, quarries in mountain areas, and bare soils converted from vegetated areas under pressure from recent development. Since these ‘‘artificial fields’’ are mainly located outside of urban areas, we put them in a separate category to differentiate them from others. Likewise, ‘‘residential areas’’ includes small villages and suburbs with greater vegetation coverage than that in ‘‘urban areas’’. Other watershed information The datasets used for analysis are listed in Table 1, including LULC maps. Population information for 1980, 1985, 1990, 1995, and 2000 was provided for corresponding meshes of 1 · 1 km from the regional mesh statistics based on the national census. When analyzing the relationship between LULC and population, we excluded: (1) 318 meshes, the area of which was not fully occupied by the Lake Kasumigaura watershed; (2) 138 meshes where more than 90% of the area was covered by water; (3) 13 meshes where rather steep population increases and decreases were observed at 5-year intervals, probably due to an error of population allocation in the meshes; and (4) 37 meshes which had both more than 1,000 persons and more than 0.2 km2 of artificial field. That last exclusion was because artificial field had various residential conditions. The number of meshes left for analysis was 1,664. Digital elevation maps with a 50 · 50-m mesh were used for elevation data. Positions of national roads and freeways were obtained from digital maps for spatial data on basic transportation infrastructure (1:25,000). Based on an investigation by Ibaraki Prefecture, information on those parts of them that had changed was also added to the database. Forty topographical maps (1:25,000) were used to extract the positions of trunk roads (national roads) and railway stations. Spatial variations in distances from railway stations in the watershed, Tokyo station, interchange of high ways and trunk roads were calculated using the digital picture maps (1:25,000). Classifications of topographical, geological- and soil types with a 1 · 1km mesh were derived from the National Digital
123
Landscape Ecol Eng (2007) 3:21–31
Information on natural topography (G01-56M). When analyzing the relationships between LULC change and such drivers, grids located in the lake area were excluded. Analytical methods The average sizes of the respective LULC patches were calculated using the landscape structure analysis program, FRAGSTATS (McGarigal and Marks 1995). We used the raster version (image map). The terms related to FRAGSTATS are explained in Matsushita et al. (2006). To analyze LULC changes, the following indices were calculated (see Appendix): LULC-change (i, j) indicates areas which have changed from LULC (i) to LULC (j) during Period-1 or Period-2; LULC (i) and LULC (j) are the i (i = 1,14) and j (j = 1,14) LULC, respectively. LULC-change% (i, j) expresses the ratio of LULC-change (i, j) to LULC (i) at the start of the period. In order to compare the levels of elucidating the LULC changes by different drivers, we first classified all the grids into ten groups based on the characteristic value of each driver, and designated an LULC-driver name (i, j, k) for LULC-change% (i, j) in case the grid was classified into the k group (k = 1,10). In particular, the change potential was calculated as follows: LULC driver name ðkÞ XX ¼ LULC driver name ði; j; kÞ: i
ð1Þ
j
In this case, sizes of the respective groups were set to be nearly equal. In addition, value of the LULCchange rate (i) was expressed by LULC change rate ðiÞ X ¼ LULC change% ði; jÞ=LULC ðiÞ:
ð2Þ
j
Secondly, we calculated v2 values (see Appendix) for LULC-driver name (i, j, k) (k = 1,10) or LULC-driver name (k) for testing whether or not the distributions were uniform and then normalized them with v2s , which was the average of v2 values when all values of LULC-driver name (i, j, k) except the one for k were zero. We consider that the larger S (v2/v2s ) indicates a higher demonstrability of the driver for LULC change. The v2 test was also applied for evaluating the significance of the difference in LULC–driver name (i, j, k) between the periods. In addition, the coefficient of variation (Cv) in LULC-driver name (i, j, k) (k = 1,10)
586.12 0.12 0.28 0.03 0.00 0.00 0.00 0.02 0.00 0.22 0.00 0.00 0.00 0.00 586.80
Paddy field
1990
0.04 533.78 0.08 0.00 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 533.92
Forest 0.50 2.06 494.86 0.01 0.02 0.00 0.00 0.15 0.00 0.06 0.00 0.00 0.00 0.00 497.64
Plowed field 0.00 0.00 0.01 241.65 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 241.66
Water
Cell values in km2
Paddy field Forest Plowed field Water Residential area Artificial field Urban area Orchard Golf course Natural grass Industrial area Artificial grass Park Uncultivated paddy 1996 Sum
1990
582.31 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 582.31
Paddy field
1996
0.00 514.24 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 514.24
Forest 0.04 1.18 490.86 0.04 0.00 0.01 0.00 0.00 0.00 0.05 0.00 0.00 0.00 0.00 492.18
Plowed field 0.00 0.00 0.00 241.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 241.20
Water
During Period-2 (1990–1996) (in order of area in 1990)
Paddy field Forest Plowed field Water Residential area Artificial field Urban area Orchard Golf course Natural grass Industrial area Artificial grass Park Uncultivated paddy 1990 Sum
1979
During Period-1 (1979–1990) (in order of area in 1990)
Table 2 LULC change matrix
0.01 0.04 0.00 0.00 162.68 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 162.73
Residential area
0.49 2.58 1.75 0.01 156.84 1.08 0.05 0.17 0.00 0.15 0.00 0.00 0.00 0.03 163.14
Residential area
3.77 11.73 5.88 0.40 0.39 66.83 0.49 0.08 0.23 1.24 0.30 0.06 0.02 0.06 91.50
Artificial field
3.47 6.61 4.77 0.15 0.34 50.32 0.03 0.53 0.29 0.55 0.02 0.00 0.00 0.00 67.08
Artificial field
0.00 0.00 0.00 0.00 0.00 0.00 58.62 0.00 0.00 0.00 0.00 0.00 0.00 0.00 58.62
Urban area
1.00 1.88 1.44 0.01 0.06 0.46 53.91 0.07 0.00 0.28 0.00 0.01 0.00 0.00 59.11
Urban area
0.00 0.00 0.00 0.00 0.00 0.00 0.00 30.31 0.00 0.00 0.00 0.00 0.00 0.00 30.31
Orchard
0.12 0.21 0.31 0.00 0.02 0.00 0.00 30.04 0.00 0.00 0.00 0.00 0.00 0.00 30.69
Orchard
0.66 5.91 0.90 0.02 0.05 0.23 0.00 0.29 29.61 0.02 0.00 0.00 0.00 0.00 37.68
Golf course
1.68 11.88 2.45 0.03 0.10 0.02 0.00 0.04 12.66 0.67 0.00 0.28 0.02 0.00 29.84
Golf course
0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 24.03 0.00 0.00 0.00 0.00 24.04
Natural grass
0.24 0.21 0.07 0.41 0.00 0.00 0.00 0.00 0.00 24.42 0.00 0.00 0.00 0.00 25.35
Natural grass
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 15.13 0.00 0.00 0.00 15.13
Industrial area
0.08 1.82 0.86 0.00 0.00 0.63 0.00 0.02 0.00 0.08 11.92 0.00 0.00 0.00 15.42
Industrial area
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.83 0.00 0.00 3.83
Artificial grass
0.07 0.33 0.28 0.00 0.00 0.03 0.00 0.03 0.00 0.00 0.00 3.16 0.00 0.00 3.90
Artificial grass
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.54 0.00 1.54
Park
0.07 0.40 0.48 0.00 0.00 0.02 0.00 0.01 0.00 0.01 0.00 0.00 0.55 0.00 1.55
Park
0.00 0.81 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.78 1.62
Uncultivated
0.01 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.70 0.84
Uncultivated
586.80 533.92 497.64 241.66 163.14 67.08 59.11 30.69 29.84 25.34 15.42 3.90 1.56 0.84 2,256.94
1990 Sum
593.89 562.01 507.63 242.29 157.40 52.56 53.99 31.07 12.96 26.45 11.95 3.45 0.57 0.73 2,256.94
1979 Sum
Landscape Ecol Eng (2007) 3:21–31 25
123
26
was by the maximum value pffiffiffiffifficalculated and pffiffiffiffiffinormalized 10; T ¼ Cv= 10 to compare the demonstrability of the drivers in the same manner as S.
Results and discussion Characteristics of LULC changes in Lake Kasumigaura watershed The regions in which LULC has changed are shown in Fig. 2. Those changed areas were 55.99 km2 (2.48% of the watershed) and 34.96 km2 (1.55%), indicating changing rates of 5.09 km2 year–1 (0.22% year–1) and 5.83 km2 year–1 (0.25% year–1) for Period-1 and Period-2, respectively. When excluding the specified meshes shown in the previous section, the changing rates were calculated to be 0.24 and 0.29% year–1 for Period-1 and Period-2, respectively. The changing rate was somewhat higher for Period-2 compared with Period-1. This is probably because, during Period-2, the so-called ‘‘Bubble’’ (expansion of monetary economy not accompanied with economical fundamentals during 1985–1990 and its breakdown after 1990) ceased, and emigration from the Tokyo Metropolitan Area was observed. In contrast, the ‘‘Bubble’’ started, and a high rate of immigration into the Metropolitan area occurred during Period-1 (Tampo 2002). Forest decreased most significantly (50.4 and 56.3% of total decrease for Period-1 and Period-2, respectively), followed by plowed field (22.8 and 19.4%) and paddy field (13.9%, 12.8%). The decreasing rates of forest were 0.45 and 0.61% year–1 for Period-1 and Period-2, respectively, which were slightly lower than the historical rates experienced in Rondonia, Brazil (1.47% year–1 during 1978–1987), Costa Rica (1.73% year–1 during 1940–1983), and Peninsular Malaysia
Fig. 3 LULC change rate (in order of area in 1990)
123
Landscape Ecol Eng (2007) 3:21–31
(2.47% year–1 during 1972–1982) (Dale et al. 2000). However, the rates observed in the Lake Kasumigaura watershed were especially high in Japan because the forested areas in Japan had changed only negligibly in the previous few decades (http://env.go.jp/policy/ hakusyo). On the other hand, golf course (30.7 and 23.1%) and artificial field (29.9 and 70.6%) increased most, followed by residential area (11.3 and 0.1%). A large increase in golf course and artificial field would bring about a rise in surface runoff, effluent nutrients and toxic chemicals, producing further deterioration of the original ecosystem (Butcher 1999; Matsushita et al. 2006). LULC changes from forest to golf course (21%) were highest in Period-1, followed by those from forest to artificial field (12%) and from plowed field to artificial land (9%). On the other hand, changes from forest to artificial field (34%), forest to golf course (17%) and plowed field to artificial land (17%) were the dominant LULC changes in Period-2 (Table 2). In addition, values of the LULC-change rate (i) expressed by Eq. 2 are higher in forest, grassland and uncultivated paddy field (Fig. 3). During Period-1, the LULC-altered regions, particularly golf course, were distributed rather uniformly throughout the watershed. In contrast, regions converted to artificial field were concentrated in the southwestern part of the watershed near the cities of Tsukuba, Tsuchiura, Ushiku, and Ryugasaki during Period-2 (Fig. 2). In both periods, the mean patch sizes converted into golf course were the largest among all LULC changes; those at Period-1 (33.6 ± 40.4 ha) were much higher than those at Period-2 (9.8 ± 20.2 ha). These sizes seem rather small compared with the mean golf course area of around 70 ha (http://www.recpas.or.jp). This is because golf courses are usually fragmented by roads and/or because this classification is sometimes applied
Landscape Ecol Eng (2007) 3:21–31
27
is lax. Thus, many small-scale LULC changes are made without difficulty. Socio-economic and natural drivers for LULC change Population
Fig. 4 Influence of population on LULC change. a Population density versus LULC change, b population density versus forest change and c population change rate versus LULC change
to small practice facilities such as driving ranges. During Period-1, the sizes of converted patches were 4.1 ± 11.6, 6.8 ± 11.0, and 4.4 ± 11.2 ha for artificial field, urban area, and residential area, respectively and those were 3.5 ± 9.1 ha for artificial field during Period-2. The guideline regulating land development are only applied to area larger than 5 ha, and enforcement
In both periods, relatively higher values of LULCpopulation (k) were observed in areas of low-population-density (LPD) (
c.300 km–2) (Fig. 4a). Since population density showed a good correlation with the combined total of residential and urban areas (r = 0.81, 0.82, and 0.82, P < 0.01, for population 1980 versus area 1979, population 1990 versus area 1990, and population 1995 versus area 1996, respectively), the above result suggests that urbanized (urban and residential) areas were highly resistant to LULC change. This tendency was also observed in LULC-population (i, j, k) in case i = forest and j = golf course and artificial field. In addition, the conversion of land to golf course compared with that to artificial field occurred more often in less populated area (Fig. 4b). An association can be observed between an increase in LULC change with population change, i.e., lower LULC change occurred in areas of decreasing population while higher LULC change was seen in areas of increasing population during Period-2, whereas during Period-1 areas with only a minor population change showed somewhat higher rates of LULC change (Fig. 4c). Throughout both periods, LULC change was less frequent in areas of decreasing population, due partly to an increasing trend in the per-capita average of residential areas (25 m2 in 1980 to 33 m2 in 1993; http://wp.cao.go.jp). As to the change from forest to golf course, the areas of only minor population change (–3 to +6 km–2 year–1) showed higher LULC change rates compared to those with sharply increasing or decreasing rates. In Period-2, a dramatic LULC change from forest to artificial field was found in areas with high rates of population growth (>27.6 km–2 year–1), most of which were located around HPD areas (Figs. 1, 2). Natural drivers In this watershed, the proportions of topographical, geological and soil types in each mesh were principally dependent on its elevation (Fig. 5), i.e., delta in topology, mud and silt in geology, and gley lowland soils are predominant at altitudes below 10 m, while
123
28
Landscape Ecol Eng (2007) 3:21–31
Fig. 5 Proportions of topographical, geological, and soil types versus elevation
Fig. 6 Influence of elevation on LULC change. a Elevation versus LULC change and b elevation versus forest change
loam plateau in topology, loam in geology, and andosols dominate from 11 to 40 m. Therefore, we analyzed only elevation with respect to LULC change rates because of the similarity in their distributions. In both periods, the LULC change rates at altitudes from 17 to 26 m were higher than those at other
123
Fig. 7 Distance from facilities on LULC change. a Distance from interchanges of highways versus LULC change, b distance from Tokyo Station versus LULC change and c distance from Tokyo Station versus forest change
altitudes, indicating the general characteristics of the LULC change potential (Fig. 6a). However, those characteristics varied depending on the type of LULC change. The LULC elevation (k) from forest to golf course was high at a relatively high altitude (>71 m in
Landscape Ecol Eng (2007) 3:21–31
29
Comparison of demonstrability for LULC change
Period-1 and 36–70 m in Period-2), suggesting the advantages in cost and recreational location (Fig. 6b). In contrast, conversions from forest to artificial field (as well as those from forest to residential and urban area in Period-1) showed higher values at altitudes from 7 to 30 m, due probably to their inhabitability.
Respective S values corresponding to the above-mentioned drivers were compared with respect to the total LULC changes, including those from forest to artificial land and from forest to golf course (Table 3). A significant difference from uniform distribution was confirmed (P < 0.001; v2 test for goodness of fit) for all factors except distance from a railway station to explain the change from forest to artificial land in Period1. In addition, a difference in the LULC–driver name (i, j, k) between the two periods was also confirmed (P < 0.001; v2 test for goodness of fit), suggesting the existence of a non-stationary process for the corresponding drivers in LULC change. Regarding LULC changes as a whole, population was the most influential, followed by the distance from trunk roads and the elevation in Period-1; on the other hand, the distance from Tokyo station and the elevation overwhelmed other drivers during Period-2. On the whole, S values were higher in Period-2 than in Period-1. As to the change from forest to golf course, the distances from trunk roads in Period-1 and from the railway station in Period-2 showed the highest S values, whereas the distance from Tokyo Station overwhelmed other drivers in both periods with regard to the change from forest to artificial field. The shift of dominant drivers and the temporal change in demonstrability most probably resulted from a socio-economic trends, i.e., ‘‘Bubble’’, ‘‘Sprawl of the Tokyo Metropolitan Area’’. However, the highest S in
Proximity to facilities These found weak or negligible influences of the distances from trunk roads, railway stations and highways on LULC changes (data not shown except for highways in Fig. 7a; see Comparison of demonstrability for LULC change). In a similar manner, the distance from Tokyo ranged from 52.6 to 92.0 km indicating the proximity to the Tokyo Metropolitan Area did not clearly influence the LULC change rate during Period1. In contrast, a remarkable decrease in that rate with the distance was found in Period-2 (Fig. 7b). In particular, the change from forest to artificial field showed the latter tendency in Period-2, whereas there was no clear tendency in LULC change from forest to golf course with distance in either period (Fig. 7c). The LULC change from forest to artificial field largely occurred in the southwestern part of the watershed which proved advantageous for commuting and attending school due to the relatively short distance from Tokyo station and to the fact that the area had not been previously developed, indicating the pressure created by the sprawl of the Tokyo Metropolitan Area. Table 3 S (c2/c2s) (%) (see text)
All changes From forest to golf course From forest to artificial field
1979–1990 1990–1996 1979–1990 1990–1996 1979–1990 1990–1996
Population
Elevation
Distance from trunk road
Distance from railway st.
Distance from interchange
1.5 1.8 4.1 3.9 0.9 1.0
1.1 2.6 0.3 2.8 0.5 3.8
1.1 1.6 4.1 5.2 1.2 2.7
0.2 1.6 1.2 10.8 0.4 1.4
1.7
0.7 2.7 2.2 1.3 1.9 14.6
Distance from interchange
Distance from Tokyo
1.6 6.6
Distance from Tokyo
Table 4 Coefficient of variation (Cv) normalized by Cvmax (see text or Appendix)
All changes From forest to golf course From forest to artificial field
1979–1990 1990–1996 1979–1990 1990–1996 1979–1990 1990–1996
Population
Elevation
Distance from trunk road
Distance from railway st.
0.12 0.13 0.24 0.25 0.09 0.09
0.10 0.16 0.16 0.21 0.28 0.19
0.11 0.12 0.19 0.24 0.10 0.15
0.04 0.13 0.12 0.34 0.08 0.12
0.13 0.25 0.14
0.08 0.16 0.17 0.13 0.14 0.34
123
30
Landscape Ecol Eng (2007) 3:21–31
Table 3 (14.6%) indicates similar values (11.1%) when changes in the ten groups occurred at the rates of 0, 0, 0, 0, 0, 20, 20, 20, 20, and 20% (information = ln 2 = 1), respectively. Such relatively low S values indicate that LULC changes in the watershed of Lake Kasumigaura occurred rather randomly, since the influence from the drivers was not critical. This indicates the possibility of LULC changes everywhere in the watershed, which combined with the modest mean LULC change size mentioned above would probably bring about fragmentation of the landscape. Our analysis using a coefficient of variation suggested results similar to those obtained with pffiffiffi S2 (Table 4) where T was nearly proportional to S (r = 0.68, P < 0.01).
prerequisite for forecasting future LULC changes. Further studies will be necessary on the relationship between such trends and LULC changes over the coming decades, since it is one of the key issues determining the structure of an LULC model. The development induced in the watershed of Lake Kasumigaura by the new railway lines should also be focused on (e.g., the Tsukuba Express opening in 2005). In addition, the relatively low S values (indices of demonstrability) show that LULC changes in the watershed occurred rather randomly, since the influence of drivers was not crucial. This tendency should be investigated in future with regard to fragmentation of the landscape, i.e., the influence of LULC changes on landscape metrics.
Conclusions
Appendix: Indices for analyzing LULC change
We investigated LULC changes in the watershed of Lake Kasumigaura during 1979–1996 and compared their socio-economic and biophysical drivers using time-series GIS datasets with a 50-m-resolution. Those changes occurred over an area of more than 90 km2 during the overall period, at changing rates of 0.22% year–1 in Period-1 and 0.25% year–1 in Period-2. Forest decreased most in both periods at changing rates of 0.45% year–1 in Period-1 and 0.61% year–1 in Period-2, showing varying predominant patterns, e.g., changes from forest to golf course in Period-1 and from forest to artificial field in Period-2. Significant LULC changes were observed, particularly in Period-2, in areas of major population increases surrounding urbanized areas. Statistical analyses indicated that the population in Period-1 and the distance from Tokyo Station in Period-2 were the most influential drivers of overall LULC changes. Various populations classified by age, income, occupations, etc., should be considered as the drivers in future studies. The relationships among LULC change, population, and the rate of population change suggested that the urbanized areas were highly resistant to LULC changes and that such changes were less frequent in areas of decreasing population. These tendencies indicate the irreversibility of LULC change, which will significantly affect the future LULC. With this in mind, we need to monitor those places that have been converted into artificial field but have not yet been used. Since change potentials differed between the periods, we could not assume the existence of a stationary process for the corresponding drivers in LULC changes. This is probably because socio-economic trends might affect the dominant drivers as well as the relationships between LULC changes and drivers. Therefore, a projection of socio-economic trends is a
LULC-change (i, j): Areas which have changed from LULC (i) to LULC (j). LULC (i) and LULC (j) are the i (i=1,14) and j (j=1,14) LULC, respectively. LULC-change% (i, j): Ratio of LULC-change (i, j) to LULC (i) at the start of the period. LULC-driver name (i, j, k): LULC-change% (i, j) in case the grid was classified into the k group of the driver. PP LULC driver name ðkÞ ¼ LULC driver
123
i
j
ði; j; kÞ : LULC change potential. P LULC change rate ðiÞ ¼ LULC change% j
ði; jÞ=LULCðiÞ : LULC change rate. P 2 v ¼ ðfk fcÞðfk fcÞ=fc; in which fk is the LULCk
driver name (i, j, k) (k = 1,10) or LULC-driver name (k), fc the LULC-driver name (i, j, k) (k = 1,10) or LULC-driver name (k) that are uniformly distributed. v2s is the average of v2 values when all values of LULCdriver name (i, j, k) except the one for k were zero. S ¼ v2 =v2s : Cv is the coefficient of variation (standard deviation divided by average) in LULC-driver name (i, j, k) (k = 1,10). pffiffiffiffiffi Cvmax is the maximum value of Cv ( 10 in this case). T¼ Cv=Cvmax : References Aspinall R (2004) Modelling land use change with generalized linear models—a multi-model analysis of change between 1860 and 2000 in Gallatin Valley, Montana. J Environ Manage 72:91–103
Landscape Ecol Eng (2007) 3:21–31 Baker WL (1989) A review of models of landscape change. Landsc Ecol 2:111–133 Butcher JB (1999) Forecasting future land use for watershed assessment. J Am Water Resour Assoc 35:555–565 Butler BJ, Swenson JJ, Alig RJ (2004) Forest fragmentation in the Pacific Northwest: quantification and correlations. Forest Ecol Manage 189:363–373 Dale VH, Brown S, Haeuber RA, Hobbs NT, Huntly N, Naiman RJ, Riebsame WE, Turner MG, Valone TJ (2000) Ecological principles and guidelines for managing the use of land. Ecol Appl 10:639–670 Gobin A, Campling P, Feyen J (2002) Logistic modelling to derive agricultural land use determinants: a case study from southeastern Nigeria. Agri Ecosys Environ 89:213–228 Ichinose T (2005) Modeling of land use/cover change (in Japanese with English abstract). Environ Sci 18:403–410 Irwin EG, Geoghegan J (2001) Theory, data, methods: developing spatially explicit economic models of land use change. Agri Ecosys Environ 85:7–24 Johnson LB, Richards C, Host GE, Arthur JW (1997) Landscape influences on water chemistry in Midwestern stream ecosystems. Freshw Biol 37:193–208 Li X, Peterson J, Liu G, Qian L (2001) Assessing regional sustainability: the case of land use and land cover change in the middle Yiluo catchment of the Yellow River basin, China. Appl Geogr 21:87–106 Lopez E, Bocco G, Mendoza M, Duhau E (2001) Predicting land-cover and land-use change in the urban fringe: a case in Morelia city, Mexico. Landsc Urban Plan 55:271–285 Luck M, Wu J (2002) A gradient analysis of urban landscape pattern: a case study from the Phoenix metropolitan region, Arizona, USA. Landsc Ecol 17:327–339 Matsushita B, Xu M, Fukushima T (2006) Characterizing the change in landscape structure in the Lake Kasumigaura Basin, Japan using a high-quality GIS dataset. Landsc Urban Plan (in press)
31 McAlpine CA, Eyre TJ (2002) Testing landscape metrics as indicators of habitat loss and fragmentation in continuous eucalypt forests (Queensland, Australia). Landsc Ecol 17:711–728 McConnell WJ, Sweeney SP, Mulley B (2004) Physical and social access to land: spatio-temporal patterns of agricultural expansion in Madagascar. Agri Ecosys Environ 101:171–184 McGarigal K, Marks BJ (1995) FRAGSTATS: spatial pattern analysis program for quantifying landscape structure. Forest Science Department, Oregon State University, Corvallis Natural Conservation Bureau (1999) The dataset for GIS on the Natural Environment of Japan, Version 2. Japan Environment Agency, Tokyo, CDROM Robinson L, Newell JP, Marzluff JM (2005) Twenty-five years of sprawl in the Seattle region: growth management responses and implications for conservation. Landsc Urban Plan 71:51–72 Soares-Filho BS, Cerqueira GC, Pennachin CL (2002) DYNAMICA—stochastic cellular automata model designed to simulate the landscape dynamics in an Amazonian colonization frontier. Ecol Modell 154:217–235 Tampo N (2002) Development of social capital in population decreasing phase. Japan Society of Civil Engineers, Tokyo (in Japanese) Veldkamp A, Lambin EF (2001) Predicting land-use change. Agri Ecosys Environ 85:1–6 Walker R (2001) Urban sprawl and natural areas encroachment: linking land cover change and economic development in the Florida Everglades. Ecol Econ 37:357–369 Wilson EH, Hurd JD, Civco DL, Prisloe MP, Arnold C (2003) Development of a geospatial model to quantify, describe and map urban growth. Remote Sens Environ 86:275–285 Zipperer WC, Wu J, Pouyat RV, Pickett STA (2000) The application of ecological principles to urban and urbanizing landscapes. Ecol Appl 10:685–688
123
