Land Use Modeling Scenarios using Spatial Dynamic Model in Badung Regency

Management of land use is necessary to maintain the balance between human needs and environmental aspects. Known as the highest-income region in Indonesia, Badung Regency has substantial demand for land utilization for tourism activities. The conversion of productive agricultural land to tourism areas would threaten the sustainability of natural ecosystems in Badung. This study aims to predict the future land use in Badung Regency in 2033 through two development scenarios, Bussiness As Usual (BAU) and Rapid Economic Growth (REG). The simulation was carried out using spatial dynamic models of cellular automata. Several satellite imageries, municipal data, field surveys, and influencing factors of land transition have been utilized as input for the model. The result indicates that the residential area dominated the development in the BAU scenario with the direction of residential development spreading to the south, and new tourism development portrayed in the north close to the agricultural area. While in the REG scenario, tourism land use had significant growth that does not only approach agricultural land but also protected forest areas in the north. The model accuracy showed moderate agreement according to the Kappa index both in BAU and REG scenarios. The finding of this study could support the decision maker to create a better policy considering the potential impact of land use growth direction.


Introduction
Physical development continues to occur because of the increasing population that requires land and space to live with all its derivatives.This phenomenon indirectly causes pressure on various protected land functions to preserve the ecosystem [1].Spatial planning regulations have been made, enforced, and implemented since, but are frequently oriented toward economic considerations.Therefore, many problems occur, such as floods, droughts, forest fires, etc. Management of land use planning is necessary to maintain the balance between environmental aspects and human needs [2].
The use of land use and land cover maps are well-known as effective instruments for facilitating planning processes, assisting with decision-making, tackling environmental problems, and managing natural resources [3,4].The Regional Spatial Plan (RTRW) in Indonesia as the formal regulation in 1318 (2024) 012004 IOP Publishing doi:10.1088/1755-1315/1318/1/012004 2 urban development has utilized a land use map in its plan construction.However, the regional plan has an issue with the aspect of time and duration of its implementation.The law and institutional barriers make it impractical for the land use map to be modified immediately.In contrast to land use maps, land use modeling can be produced quickly and repeatedly.
Land use modeling displays the process of land use change that has occurred or potentially occurred, in a spatial manner and time-based, with influential factors [5].The Cellular Automata (CA) model is one of the dynamic models that are raster-based and have a neighborhood effect [2,6].Physical and socio-economic factors could be integrated into the simulation process to define future land use.With this model, a spatial planning assessment might be performed.Any spatial change that happened and affected spatial plans, such as natural disasters, could be handled by running the simulation.In general, the model will support reliable data and information for decision-makers unceasingly.The land use change simulation would give an insight into the optimal utilization of land use [2,4].
Badung Regency is known as one of the highest-income tourist-based regions in Indonesia.The accessibility and tourist attractions along the coast have an impact on how the southern part of Bali develops along the coast from the western to the eastern side [7].It could be challenging to control the land use change, such as the conversion of productive agricultural land.Since it was believed to threaten the sustainability of coastal ecosystems.Hence, it also will affect social shifting such as changes in community social institutions [7].However, tourism is a vital sector that promotes other economic activities in Badung.For this reason, efficiency in land management is fundamental to the stability of economic and environmental matters.
This study aims to build land use modeling scenarios in Badung Regency using spatial dynamic models under two scenarios: Bussiness As Usual scenario and Rapid Economic Growth scenario.It will provide information on the direction of development for Badung Regency in 2033, as well as the dominant land use and the model performance for each scenario.Hopefully, this study will establish valuable knowledge for land use management in Badung Regency.

Study Area
The study was located in Badung Regency, Bali, with a 418,52 km 2 area.Covering the entire southern coast of Bali as a low elevation area with 27 meters above sea level and creating a thin form to the north toward Beratan lake, Badung is divided into 6 districts: Kuta Selatan, Kuta, Kuta Utara, Mengwi, Abiansemal, and Petang (fig 1).The population has reached 548.191 people based on census data in 2020 [8].In 2033, the residential area is projected to occupy 10.333 hectares of land [9].
The main activity in Badung is derived from tourism and is followed by agriculture [8].The majority of tourism activity is natural scenery, such as sand beaches, cliff beaches, lakes, and mountains.Badung also has a rich culture embedded in people's daily activities that attract plenteous visitors to come.On the other side, farming used to be a primary sector in Badung and indigenous society has rules to protect their agricultural land.However, with the rising demand for the tourism sector, farming land gets suppressed.If beforehand tourist activity focused on coastal areas, in recent times the trend showed that the location of tourism business spread into agricultural land in the last decade [7].

Data
Land-use data was obtained from various sources such as municipal public data, field surveys, and satellite imagery.The official data was derived from the Regional Spatial Plan (RTRW) document for Badung Regency.The data includes the existing land use of Badung in 2018 as the input for the spatial dynamic model and the planning zone for 15 years for the rate of land use change.The land use types from RTRW were first identified and then organized into 10 classes: forest, grassland, agriculture, bare land, water body, residential, commercial, tourism, road, and utility.The  In addition to land use data, driving factors is a necessary input for the model.Numerous factors influence land use change.Previous studies indicated that there are two main categories: biophysical factors and social-economic factors.Each of these categories could be elaborated into many factors.Topography, bedrock and soil type, water resources, weather, accessibility, and existing land use are examples of factors within the biophysical category.The socioeconomic category contains population, income, technology, culture, and institutional factors [3,5].Nevertheless, in this study, the factors were limited to several factors as shown in Table 1.Driving factors were listed from the literature review and matched with the available data, then the value of each factor was issued with the Analytical Hierarchy Process (AHP) [10].

Methods
This study applied the cellular automata concept, a method to produce land use prediction through modeling and spatial simulation [6].Future land use change was predicted using a series of modeling processes that include factors that affect the demand for particular land use.This research mainly utilized GIS software for spatial data processing and LanduseSIM software, which helps in land use modeling [6].
The land use simulation in this study was developed in two scenarios: Business As Usual (BAU) and Rapid Economic Growth (REG).The difference between both scenarios lies in the modeling rules and land demand.The BAU scenario is a typical scenario that uses land demand according to the trend rate and has tourism as the highest hierarchy.The total area growth from this scenario was coming from the area development of existing land use 2018 data and machine learning model in 2021 [11].As for the REG scenario, land demand was obtained from the Badung Regency Spatial Plan (RTRW), and residential ranked as the first order.Therefore, the expected growth value will be different from the BAU scenario.Both scenarios have protected areas such as forests as the constraint.
The simulation comprises 4 stages: data preparation, data simulation, visualization, and accuracy assessment.In the preparation stage, initial land use maps, driving factor maps, and constraint maps were prepared, including the total amount of expected growth in the predicted year.The initial land use was derived from the 2018 land use map from municipal/government data.Initial land use is then reclassified according to the land use class required in the modeling.The driving factors map was obtained from basic data from the government and the weight came from AHP.The constraint map displays a protected area that cannot be converted into any type of land use.The expected growth in the prediction year, which is the next 15 years in this study, was obtained from the calculation of land use trends from 2018 to 2021 in the BAU scenario or the total area in the RTRW of Badung Regency in the REG scenario.
These four datasets were processed using a GIS application and then came into the simulation stage to create transition rules.Land use change modeling in Badung Regency for the next 15 years was carried out with output from 2018 to 2033 in a raster-based form with a 20x20 m cell size.The final stage was visualization and accuracy assessment.The accuracy step is addressed to assess the validity of the model built.The accuracy tests used in this model are overall accuracy and the kappa index.In this case, overall accuracy generally uses a confusion matrix that displays a comparison between the real value (real/truth) and the predicted value in pixels.The overall accuracy value shows the percentage of pixels that are classified correctly.Overall accuracy is calculated by adding up the correct values (corresponding between real and predicted) divided by the total number.The kappa index measures the suitability between classifications and real values [12].The accuracy test was carried out using GIS software and Microsoft Excel.The data required for accuracy testing is ground truth points from the Badung Regency Survey in 2021 and maps from the simulation results of land use models in Badung Regency for 2021.

Business As Usual Scenario
Land use of Badung regency in 2018 was dominated by agriculture and residential.The simulation result in the BAU scenario for 2033 captured the agricultural land as the largest, followed by residential areas.However, the total area of farming has decreased, while the settlement has increased rapidly.The model outcome displayed the development of residential areas in the south (Kuta Selatan district) and intensified tourism in the densely populated area of the central part (Kuta district).Moreover, several new tourism areas have appeared next to the agricultural land on the north side (Mengwi district).However, the growth of commercial land is hard to be detected.Figure 3 below indicates a few locations that encounter notable changes in the Business As Usual scenario.Table 3 illustrates the total area of land use transition from 2018 to 2033.Land use classes that experienced changes occurred in grassland, agriculture, bare land, residential, and commercial.Whereas forests, water bodies, tourism, roads, and utility have been restricted to convert into other land uses because of the land constraints set before the simulation.The area of grassland, agriculture, and bare land had declined and shifted to settlement, commercial, and tourism.
BAU scenario regulates the highest hierarchy for development was tourism, followed by commercial and residential.However, the result shows residential land has substantial growth, from 8,913 hectares in the initial year to 10,319 hectares in the next 15 years.The tourism area only expanded from 233 hectares to 289 hectares, and the commercial has slightly increased to 10 hectares area.The kappa index for the land use model in this scenario was found to be 0.4111 and categorized as moderate agreement.The number explained the consistency between prediction land use from the model in 2021 and the actual land use from ground truth is fair enough.
Land demand for the BAU scenario was derived from the conversion rate of each land use in the last three years.Existing land use in 2018 to land use model in 2021 with 93% accuracy was compared to create the conversion rate.From this process, residential land had increased certainly because the population growth, whereas tourism areas had obtained limited areas for growth as well as commercial area.On detailed investigation, settlement and tourism have a high possibility to be mingled, such as home buildings used as guest houses.Therefore, specified data about the guest house, hostel, and/or lodging could escalate the model's accuracy.

Rapid Economic Growth Scenario
Differing from the BAU scenario, the REG scenario generated the development of tourism more significantly than residential.Tourism land use was expanded in the southern (Kuta Selatan district) and northern regions of the district (Mengwi and Abiansemal).Several new tourism areas have emerged in the northernmost region, close to the forest land (Petang district).In addition, the commercial land appeared remarkable in this scenario as shown in Figure 4.
Kuta Selatan has built and intensified several new tourist sites in the last decade such as Pandawa Beach in 2010 and Garuda Wisnu Kencana in 2018.The government also has a policy to support Kuta Selatan as a tourism center with consideration of the limited fertility of land and landscape potential.On the other hand, Kuta had been known as the major tourist destination in Badung.The largest part of travel accommodation located in Kuta since then.This occasion could explain the massive growth of tourism in the REG scenario.Based on Table 4, the changes occurred in grassland, agriculture, bare land, commercial, and tourism.The factor rules in the REG scenario are slightly different from the BAU scenario.The land constraint was defined only for forests, water bodies, roads, and utility.Furthermore in this scenario, residential has a higher hierarchy than tourism, followed by commercial.The results showed that the tourism area has increased extensively from 180 hectares to 3819 hectares in 2033.While the residential area was built up similar to the BAU scenario, commercial land has grown to around 1519 hectares.The accuracy assessment showed that overall accuracy for the REG scenario was 53% with a Kappa Index of 0.439, so the validity of the model can be categorized as moderate agreement.

Conclusion
This study conducted land use modeling scenarios using the spatial dynamic model that adopts the concept of cellular automata in Badung Regency from 2018 to 2033.The model was developed under two scenarios: Business As Usual (BAU) and Rapid Economic Growth (REG).Future land use change was predicted using a series of modeling processes that include factors that affect the demand for particular land use and transition rules.According to the simulation results, for the BAU scenario, residential development is more prevalent than the development of other land uses.The direction of residential development spreads to the south, and there is new tourism development in the north around the agricultural area.The southern and northern parts of the BAU scenario develop into residential.As for the REG scenario, the tourism area is growing significantly.It is noticeable that southern and northern parts develop into tourism in the REG scenario.Commercial land is also seen developing in this scenario.Meanwhile, residential land becomes denser in the central part of the district.The model accuracy for the BAU scenario was obtained at 52%.Meanwhile, in the REG scenario, it was slightly higher at 53%.
Based on the simulation result, the land use manager could use either of both scenarios with each scenario followed by some considerations.The BAU scenario provides a reasonable prediction because it was derived from the actual growth of the Badung region.In this scenario, the progress of tourism land portrays a modest rate while residential land develops significantly.One of the advantages of this scenario is that most of the protected areas such as forests and agriculture are preserved.However, it would probably not stimulate revenue for the region.Therefore, the homebased industry that supports tourism and tourism activities or experiences without specific land required should be reinforced in this scenario.
On the other hand, the REG scenario shows an ambitious forecast for the development of tourism activity in Badung.Tourism does not only approach agricultural areas but also protected forest areas in the north.This scenario would result in better revenue for the region because the plenteous area is addressed for tourism, but it probably would damage the ecosystem.For that reason, utilizing this scenario requires regulation to maintain the ecosystem balance.For example, create zoning regulations for accommodation in the mountain forest area.
This study mainly analyzes the physical aspect of urban development, which is just one of the influential factors of urban dynamics.Furthermore, it still excludes the customs rules that could be various among regions.Therefore, further research could utilize more non-physical factors and specific constraints such as custom rules for each region.

3
classification was created according to model requirements.All the source data were reclassified into these classes and then analyzed in the raster form.

Table 3 .
Land use area development REG Scenario (unit: hectares).