Spatio-temporal evolution and prediction of carbon storage in Kunming based on PLUS and InVEST models

Study area

Kunming is directly northeast of central Yunnan Province (102°10’ ~ 103°40’ E, 24°23’ ~ 26°22’ N) in the central Yunnan-Guizhou Plateau (Fig. 1) (Fang et al., 2021), the land area of Kunming is about 21,011.41 km².

Kunming, the capital of Yunnan Province in China, houses most of the province’s population and GDP. It is the province’s most important economic development center and corridor for economic and cultural communications between China and Southeast Asian countries. Furthermore, Kunming has abundant forest resources and biodiversity, making it an important area in the Yangtze River Economic Belt and an eco-conservation shield upstream of the Yangtze River (Jia, Wang & Zhao, 2022). In October 2021, the 15th Conference of the Parties (COP15) of the UN Convention on Biological Diversity was hosted in Kunming. At the conference, new ideas were presented for conserving worldwide biodiversity. Ecological civilization refers to the sum of material and spiritual achievements made by human beings in accordance with the objective law of harmonious development of man, nature, and society. During the 14th five-year plan period of China’s national economic and social development, Yunnan Province pledged that it would strive “to become the vanguard of China’s ecological civilization construction” as a long-term goal and incorporate the aim to achieve “carbon peak and carbon neutrality” into the general arrangement of economic development and the establishment of an eco-civilization. To realize these goals, Yunnan should promote the establishment of an eco-civilization, which is not only a significant embodiment of Yunnan’s initiative to serve the national developmental strategy but also an important embodiment of integrating into the national development. Therefore, studying and forecasting the response of land utilization conversions to carbon storage in Kunming can help build a strong ecological security barrier in southwest China and provide theoretical support for reducing regional emissions.

Data sources and preprocessing

The main data used for this research mainly included land utilization data, carbon density, driving factors, and other data (Table 1). (1) Land utilization data: the land-use data (30 m × 30 m) was obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences. The data set was built by the Chinese Academy of Sciences through artificial visual interpretation on the basis of the national resources and environment database and the Landsat remote sensing image data as the main information source. The data were selected in time: 2000, 2010, and 2020, and the geographic coordinate system used was GCS_WGS_1984, which included 25 secondary land types and six primary land types (cultivated land, forest land, grassland, water area, construction land, and unused land). (2) Carbon density: the carbon density varies by land use type. The studies which had similar levels of natural resources (Ke & Tang, 2019) and climatic conditions (Yan et al., 2015; Tang, Xu & Ai, 2019) to that in the study area were selected as references. Then, the data were corrected according to the above-ground carbon density dataset of the terrestrial ecosystem of China in 2010 and the carbon density dataset of soil in 0 ~ 100 cm of the terrestrial ecosystem of China in 2010. Finally, the carbon density dataset for land utilization categories in Kunming was obtained (Table 2). (3) Driving factors: The driving factor data for future land prediction included physical factors and social factors. The physical factors involved the DEM, slope, annual average rainfall, soil type, annual average temperature, and distance to the water system. The social factors included the GDP, distance to the government office, distance to the highway, distance to the main road (first-class road and second-class road), population density, data source, and resolution, as shown in Table 1. Next, ArcGIS was used to unify the resolution to 30 m × 30 m and unified the geographic coordinate system to GCS_WGS_1984. (4) Other data: four Landsat 8 OLI and TIRS images in August 2020 were selected for radiometric calibration, atmospheric correction, splicing, and cropping to obtain remote sensing images of Kunming City to extract the impervious surface in 2020. NDVI was directly obtained from the vegetation index data MOD13A1.

Ecosystem carbon storage estimation

The Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model is an ecological model for evaluating and quantifying ecosystem services. In this study, the carbon storage and sequestration module of the InVEST model was used to calculate the carbon storage in Kunming. The basic assumption of this module is that the value of the carbon density of a specific land category is fixed, and the carbon storage of that land type can be acquired by multiplying the value of carbon density with that of the land area (Gong et al., 2022). The carbon storage module in the InVEST model divides the ecosystem carbon storage into four basic carbon pools (Lin et al., 2022), which include terrestrial biogenic carbon, subsurface biogenic carbon, soil carbon, and dead organic carbon. The sum of the carbon stock of the four carbon pools provides the value for the total carbon storage of the ecosystem in the area, which can be achieved using Eqs. (1) and (2) (Li et al., 2020b).

(1) $$C_i=C_{i-above}+C_{i-below}+C_{i-soil}+C_{i-dead}$$

(2) $$C_{tot}={\sum }_{i=1}^n{C}_i\times S_i.$$

Here, $C_i$ indicates the carbon density of land utilization type i; $C_{i-above}$, $C_{i-below}$, $C_{i-soil}$, and $C_{i-dead}$ indicate the carbon density of terrestrial biogenic carbon, subsurface biogenic carbon, soil carbon, and dead organic carbon of land use type i, respectively; $C_{tot}$ indicates the total carbon storage in the region; $S_i$ indicates the area of land utilization pattern i; n indicates the total number of land use types.

Future land use prediction and scenario setting

Impervious surface extraction

The impervious surface area (ISA) is a typical land cover type and an important indicator for measuring and analyzing city development and ecology (Xu, 2009). The index method is generally adopted to extract the urban impervious surface. The representative spectral indices include the normalized building index (NDBI), the normalized difference impervious surface index (NDISI), and the enhanced normalized difference impervious surface index (ENDISI) (Duan, Zhang & Liu, 2022). ENDISI can better identify shadows of mountains and remains unaffected by terrain factors while identifying impervious surfaces. Kunming City is dominated by mountainous terrain. Selecting this index for extracting information on impervious surfaces can prevent mountain shadows from affecting the extraction accuracy of impervious surfaces. The ENDISI can be calculated using Eq. (3).

(3) $$ENDISI={\displaystyle \frac{\left(2Blue+MI{R}_2\right)÷2-\left(NIR+Red+MI{R}_1\right)}{\left(2Blue+MI{R}_2\right)÷2+\left(NIR+Red+MI{R}_1\right)}.}$$

Here, $\mathrm{B}\mathrm{l}\mathrm{u}\mathrm{e}$, $\mathrm{R}\mathrm{e}\mathrm{d}$, $\mathrm{N}\mathrm{I}\mathrm{R}$, $\mathrm{M}\mathrm{I}{\mathrm{R}}_1$, and $\mathrm{M}\mathrm{I}{\mathrm{R}}_2$ are the reflectance of blue, red, near-infrared, shortwave infrared 1, and shortwave infrared, two bands corresponding to the image.

Before extracting the information on the impervious surface, the normalized water index MNDW (Xu, 2008) needs to be used to mask the large area of water and snow over the study area. The MNDW index can be derived from Eq. (4).

(4) $$MNDWI={\displaystyle \frac{\left(Green-MI{R}_1\right)}{\left(Green+MI{R}_1\right)}.}$$

Here, $\mathrm{G}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}$, $\mathrm{M}\mathrm{I}{\mathrm{R}}_1$, and $\mathrm{M}\mathrm{I}{\mathrm{R}}_2$ are the reflectance of green, shortwave infrared 1, and shortwave infrared 2 bands corresponding to the image, respectively.

Using ArcGIS, 700 verification points were randomly generated in the working area after masking water and snow. With the data on the impervious surface extracted from the Landsat 8 image in 2020, the land use data and the remote sensing imagery of Kunming City in this period were selected, and the accuracy of the extracted impervious surface was determined by artificial visual interpretation. The Kappa coefficient was 0.7582, and the overall accuracy was 89.86%. It demonstrated that the accuracy of the impervious surface met the requirements of subsequent research.

Land-use change analysis

According to the overlay map of the area proportion of land utilization types (Fig. 3), the land in Kunming City between 2000 and 2020 was primarily forest land, accounting for approximately 45% of the total area studied. The total area of forest land, grassland, and cultivated land decreased between 2000 and 2020. From 2000 to 2020, the land use transfer matrix (Table 4) revealed that forest land decreased by 142.28 km², grassland decreased by 328.95 km², and cultivated land decreased by 278.07 km². The growth rate of construction land was high; construction land increased by 708.84 km², and the area proportion increased from 2.3% to 5.7%. The development process of S1 in 2030 was similar to the trend of 2000–2020. Woodland, grassland, and cultivated land in 2030 decreased by 123.86 km², 150.61 km², and 197.358 km², respectively. However, construction land increased by 451.136 km², and water area and unused land remained unchanged. In the S2 scenario, with ecological protection as the leading role, the area of forest land increased by 1.1%, and the growth rate of construction land slowed down. In the S3 scenario, where cultivated land conservation policy and economic development were considered, cultivated land and construction land increased by 55.82 km² and 209.28 km², respectively. In comparison, forest land and grassland decreased by 123.96 km² and 151.79 km², respectively.

The construction land in the four periods was concentrated in the main urban area of Kunming (Wuhua District, Panlong District, Xishan District, Guandu District, and Chenggong District) and the surrounding areas (eastern Anning City, northern Songming County, and northern Jinning District) (Fig. 4). According to the statistical yearbook of Yunnan Province in 2020, the main urban area of Kunming comprised 63.17% of the population and generated 73.76% of the GDP of the city. A large population and capital flow led to increased construction, and ecological lands, such as forest land and grassland, decreased considerably. The land utilization types in Luquan County and Xundian County are mainly woodland and grassland, which act as important ecological barriers in Kunming City. The Jiaozi Snow Mountain Nature Reserve is in Luquan County and is an important water conservation ecological reserve in Kunming City. The overall territorial spatial layout of Kunming is: the southern Dianchi Lake basin is the core of the economy and population, and the northern mountainous area is the ecological security area.

Analysis of space and time variation of ecosystem carbon storage

The carbon storage in Kunming in 2000, 2010, and 2020 was 1.146 × 108 t, 1.139 × 108 t, and 1.120 × 108 t, respectively, indicating a continuous decrease, with a total decrease of 2.619 × 106 t in 20 years. The reduction of carbon storage in 2010–2020 was the highest, with a decrease of 1.917 × 106 t. During this period, the economic development and urbanization of Kunming were rapid, and the demand for land use was relatively high.

In 2030, carbon storage is predicted to be 1.102 × 108 t in the S1, and the loss of ecosystem carbon stock is greater. The ecosystem carbon storage might decrease by 1.818 × 106 t compared to that in 2020. In 2030, carbon storage is predicted to be 1.136 × 108 t in the S2 scenario, which is 1.601 × 106 t higher than that in 2020, indicating that ecological protection can restore ecosystem carbon storage in Kunming. Finally, carbon storage in the S3 is predicted to be 1.105 × 108 t in 2030, and the ecosystem carbon storage might decrease by 1.479 × 106 t compared to that in 2020, which is less than the carbon loss in the S1. That is, the prediction results of ecosystem carbon storage under different scenarios are S2 > S3 > S1, which is consistent with the prediction results of previous studies (He et al., 2022; Wei et al., 2023). As shown in Fig. 5, the terrestrial biogenic carbon, subsurface biogenic carbon, dead organic carbon, and soil carbon in the S2 scenario in 2030 are predicted to be higher than those in the S1 and S3 scenarios. The soil carbon gap was found to be the largest, which was 1.539 × 106 t and 1.250 × 106 t higher than the soil carbon in the S1 and S3 scenarios, respectively. The terrestrial biogenic carbon storage, belowground carbon storage, and dead organic carbon storage in the S1 and S3 scenarios were similar. In contrast, the soil carbon storage in S3 was higher than that in S1, which was consistent with the changes in total carbon storage.

The overall content of ecosystem carbon stock in Kunming is high, showing a ‘north high, south low’ distribution (Figs. 6 and 7). Carbon storage is primarily concentrated in Luquan County, Xundian County, and Yiliang County to the east and west of the research area, respectively, whereas the main urban area of Kunming City to the south of the study area, is highly urbanized and has less carbon stored. From 2000 to 2030, the main urban area of Kunming experienced the greatest change in the spatial layout of carbon storage. According to our findings using the S1 scenario, construction land in the main urban area expanded dramatically from 2000 to 2030, resulting in a loss of regional ecosystem carbon storage every year due to rapid urbanization. Under the S2 scenario’s ecologically protective development, the carbon stock in Kunming’s main urban area increased in 2030 compared to 2020. When considering the cultivated land conservation policy, eco-friendly policy, and economic development, carbon storage in the main urban area of Kunming was lower than that in 2020, but the reduction range was narrower than that in the S1 scenario. The results showed that carbon stock in the main urban area of Kunming could strongly affect the change in the ecosystem carbon storage in the whole city.

Effects of land-use conversion on carbon storage

Land utilization/land cover significantly affects vegetation cover and biomass. It is also the primary purpose for distributing and changing carbon stock in regional terrestrial ecosystems (Zhang et al., 2022b). The dynamic changes in carbon stock induced by the changes in main land use types in Kunming from 2000 to 2030 are shown in Fig. 8. From 2000 to 2030, according to the S1 scenario, carbon storage decreased mainly because of the shift from forest land and grassland to other land use types. Although the policy of returning farmland to the forest has partly promoted the transfer of cultivated land to forest land, the expansion of construction land encroaches on forest land, leading to a sharp decrease in the ecosystem carbon reserve. Under the S2 scenario, in 2030, the increase in ecological lands, such as forest land and grassland, was predicted to be the key reason for the increase in carbon stock. Under the S3 comprehensive development scenario, in 2030, the cultivated land conservation policy promoted increased cultivated land carbon storage, while the forest land and grassland carbon stock decreased. The carbon density of forest land and grassland was higher than that of cultivated land, resulting in a decrease in the overall ecosystem carbon storage. The outcomes suggested that the conversion in land utilization type is consistent with the changes in the ecosystem carbon stock, and land use/land cover directly affects ecosystem carbon stock.

Spatial correlation analysis of ecosystem carbon storage

Spatial correlation is segmented into spatial global autocorrelation and spatial local autocorrelation (Luo, Ai & Jia, 2022). In this study, Moran’s I index was employed to represent the spatial global autocorrelation of ecosystem carbon storage in Kunming (Xiong et al., 2021) and Getis-Ord Gi * was employed to measure the spatial local autocorrelation (Zhang et al., 2020a). This study divided Kunming into a 2 km × 2 km grid, and the carbon storage data in the three development scenarios for 2000–2020 and 2030 were linked to the grid. Based on this scale, the spatial correlation of carbon stock in Kunming was analyzed.

In terms of global autocorrelation, the spatial Moran’s I values of carbon storage in Kunming were 0.5583 in 2000, 0.5546 in 2010, 0.5635 in 2020, 0.5900 in 2030 S1, 0.5672 in 2030 S2, and 0.5763 in 2030 S3, indicating a significant spatial global autocorrelation in carbon storage in Kunming (all values were greater than 0). The results of the carbon storage hotspot analysis under the three development scenarios in Kunming from 2000 to 2030 are shown in Fig. 9 for local autocorrelation. From 2000 to 2020, the area designated as the hotspot of carbon stock in Kunming reduced, while the area considered to be a coldspot increased. Except for the northeast of Dongchuan District, the coldspot area in other areas, especially in the main urban area of Kunming, increased. In the last 20 years, the hotspots of carbon stock in the research area were scattered in the northwestern, western, central, and eastern regions, including the east and west sides of Luquan County, Xishan District, the western part of Songming County, the northern part of Yiliang County, and the eastern part of Guandu District. Carbon storage was not only higher in these areas, but high-value areas of carbon stock were also gathered in these areas, primarily situated in zones with less construction land, excellent plant coverage, and more ecological land. The coldspot area was mostly distributed in the northeastern, central, and southern locations of Kunming, i.e., the northeastern part of Dongchuan District, the eastern part of Songming County, the eastern part of Yiliang County, the northern and western parts of Shilin County, and the central part of the main urban area of Kunming City. These areas underwent rapid land development, had complex topography and geomorphology, and fragmented distribution of ecological land, forming an area with low carbon storage. In the S1 scenario, in 2030, the coldspot area in the main urban areas of Anning City and Kunming City increased. In the main urban area, the coldspot area in the S2 scenario was slightly smaller than in 2020. In the S3 scenario, the coldspot area in 2030 was similar to that in 2020, indicating that eco-friendly and cultivated land protection policies were beneficial in reducing the loss of ecosystem carbon stock in Kunming City.

Research on driving factors of ecosystem carbon storage

Bivariate spatial autocorrelation analysis of carbon storage and the influencing factors

Depending on the results of Pearson’s correlation analysis, two indicators that had the strongest relevance with the carbon storage of the ecosystem in Kunming were selected, i.e., impervious surface coverage and NDVI, and bivariate spatial autocorrelation analyses were performed between the two indicators, and carbon storage data. The bivariate ‘Moran’s I index represents bivariate global spatial autocorrelation. A total of −0.379 was the bivariate ‘Moran’s I index of impervious surface coverage and ecosystem carbon storage. The ‘Moran’s I index’s negative value indicated a global negative correlation between impervious surface coverage and ecosystem carbon storage. The bivariate NDVI and ecosystem carbon storage index, Moran’s I, was 0.450. The presence of a positive ‘Moran’s I index value indicated a global positive correlation between NDVI and ecosystem carbon storage.

Figure 10a depicts the bivariate LISA cluster map of impervious surface coverage and carbon storage in the Kunming ecosystem, which reflects the two’s local agglomeration characteristics in space. The impervious surface coverage and carbon storage of the ecosystem showed opposite values of the two-pole agglomeration characteristics in space, which mainly were high-low agglomeration and low-high agglomeration, i.e., carbon stock in the area with low impervious surface coverage was higher, and carbon stock in the area with high impervious surface coverage was lower. The bivariate LISA clustering results of NDVI and ecosystem carbon storage in Kunming (Fig. 10b) showed that NDVI and ecosystem carbon storage had a similar value for the polar agglomeration characteristics in space, mainly high-high agglomeration and low-low agglomeration. Thus, the regional ecosystem carbon storage with higher vegetation coverage was higher, and the regional ecosystem carbon storage with lower value coverage was lower. These results suggested a local negative correlation between impervious surface coverage and ecosystem carbon storage and a local positive correlation between NDVI and ecosystem carbon storage.

Contribution of land use driving factors

Social and natural factors mainly drive land use change (Gong et al., 2022). Social factors include population density, GDP, distance to road, distance to the government office, etc. Natural factors include terrain factors, such as slope and elevation, and climatic factors, such as mean annual rainfall and mean annual temperature. In this study, 11 driving factors of social economy and climate were selected to forecast land utilization in 2030, and the contribution of each driving factor was evaluated (Table 6). A higher contribution of a factor indicated a greater impact of the driving factor on local land use evolution.

The driving factors with the highest contribution to cultivated land were GDP (0.166), DEM (0.111), and population density (0.104). Cultivated land expansion is inextricably linked to population and economic development because rapid population and economic growth will lead to greater food demand (Liao et al., 2021). Climate conditions differed depending on altitude. Agricultural farming was closely linked to climatic conditions, so cultivated lands were generally distributed in low-altitude farming areas. DEM (0.183), population density (0.125), and slope were the factors that had the greatest impact on forest land (0.117). The slope and altitude were important topographic factors influencing vegetation growth. In general, areas with a small slope and a low altitude are better suited for vegetation growth. Higher population density areas have less forest land area, and population density limits forest land expansion. GDP (0.133), population density (0.117), and distance to the water system (0.104) were the driving factors with the greatest contribution to grassland. The GDP and population density restricted grassland expansion, while the water system promoted grassland expansion. The area closer to the water system was more prone to grassland expansion. The driving factors with the highest contribution to construction land were population density (0.154), GDP (0.143), and the mean annual temperature (0.122). A greater population density and a higher GDP were associated with a greater demand for and more expansion of construction land. The mean annual temperature affected the development of construction land by changing the population density. The driving factors with the highest contribution to the unused land were DEM (0.306), the mean annual temperature (0.199), and the mean annual rainfall (0.156). The climate in high-altitude areas is harsh, and the perennial temperature is low, which is unfavorable for vegetation growth and human habitation.

Suggestions on carbon sink function restoration and governance

The period from 2000 to 2020 was an important stage of economic development in Kunming. In the last 20 years, the gross domestic product increased from 6,262.853 to 67,337.909 million yuan, the population increased from 4.8094 to 8.463 million, and the construction land expanded from 484.21 to 1,193.05 km². In the last 20 years, the area of forest land and grassland has decreased significantly, resulting in a decrease in ecosystem carbon storage in Kunming. Taking timely measures to protect the ecosystem is necessary to reduce the loss of carbon reserves, according to the predictions of the trend continuation scene, ecological protection scene, and comprehensive development scenario. According to the “14^th Five-Year Plan for National Economic and Social Development of Yunnan Province and the Outline of the Visionary Goals for the Year 2035”, Yunnan Province focused on “making new progress in the building of eco-civilization by 2025, and building the vanguard of China’s ecological civilization by 2035” (People’s Government of Yunnan Province, 2021). Responding to climate change and achieving carbon peak and carbon neutrality are also necessary for future development. Kunming needs to incorporate the dual-carbon goal into the overall future development and foster a comprehensive greener shift in environmental development.

Based on the distribution of carbon stock in Kunming from 2000 to 2030 (Fig. 7), Luquan County, Xundian County, and Dongchuan District in the north and Yiliang County and Shilin County in the southeast were found to be the essential sources of carbon storage in Kunming. Luquan County is a high-altitude mountain area; the mountain area takes up 98.4% of the whole area of the county, the mountain slope is large, and the forest coverage rate is around 55.4%. Considering that Luquan County is dominated by agricultural development, based on these advantages, forestry, agricultural development, and ecological green development can be organically combined to carry out characteristic economic forests, fruit trees, and other afforestation. Focusing on the positioning of the ecological containment function area of Kunming, Luquan County, Xundian County, and Dongchuan District needs to strengthen ecological conservation construction and improve the protection system of nature reserves and forest parks. Strive to build the Jiaozi Snow Mountain Nature Reserve as an important water source conservation area and ecological protection guard not only in Kunming but also in the middle and upper reaches of the Yangtze River. These areas should build a strong ecological security defense line in the north of Kunming. As an ‘ecological civilization county of Yunnan Province’, Shilin County should lead the construction of the national ecological civilization and help Yunnan in achieving its target of ‘double carbon’. Considering that the tourism industry is important in Shilin County, eco-tourism should be promoted to construct an ecological civilization. Shilin County, as a typical karst area, should actively solve the problem of restoration of ecological vegetation and strengthen the implementation of comprehensive control of returning farmland to forest and rocky desertification. Yiliang County has an abundance of forest resources and species diversity. The forest coverage rate of the county is 46.2%. Yiliang County can utilize its resource advantages, strengthen the protection of forest resources based on the current condition of forestry resources. Through the construction of an economic forest and flower planting base, simultaneously develop an ecological civilization and forestry economy and build an ecological security barrier with Shilin County in the southeast of Kunming. During the period of 2000 to 2030, the carbon storage in the main urban area of Kunming City, especially in the Dianchi Lake Basin, was found to decrease every year. Promoting the ecological management of the Dianchi Lake Basin is imperative for the development of Kunming City. The main urban area of Kunming City is located in the red line of water conservation and eco-protection of plateau lakes and the upper reaches of the Niulan River. Thus, its resource advantages need to be fully utilized to drive the holistic and sustainable development of the plateau lake basin. It is also necessary to improve the ‘three lines and three zones’ delineation standards, strictly control the Dianchi Lake Basin’s wetland parks and basic farmlands, conduct rational land and space planning, and improve the efficiency of land resource utilization. Ways to meet the needs of urban expansion while also protecting the environment must be determined urgently for the development of Kunming City. Solving this problem is also the key to achieving the dual carbon goals and the establishment of eco-civilization in Yunnan Province.

Research prospect

With the PLUS and InVEST models, the space-time changes in carbon storage in Kunming City were predicted and evaluated. The influence of land utilization conversion on carbon stock was analyzed. The correlation between impervious surface coverage and vegetation index, and regional ecosystem carbon storage was analyzed. The findings provided new ideas for sustainable development in the future. This study, however, had some limitations. First, the InVEST model’s carbon storage calculation module had some flaws. Many factors influence the ecosystem’s carbon stock. The InVEST model ignored the effects of climate, topography, hydrology, and other conditions in favor of focusing solely on land use change. Furthermore, the model ignored the effect of interannual changes in carbon density. Second, when using the PLUS model for future land cover, only 11 driving factors were considered; however, the actual land utilization evolution is influenced by a large number of physical and human factors. In future studies, the measured data might be used to determine the dynamic carbon density. Besides land use, other influencing factors might be considered to comprehensively evaluate the regional ecosystem carbon storage. While performing land-use simulations, we should select as many driving factors as possible to predict the future requirement of land utilization and improve the prediction accuracy of future land utilization patterns.