Analysis of dominant land cover class based on land change cluster pattern in West Java Province

Study of land cover change is important in order to understand the global environmental change. One of the causes of land cover change is human activity in an effort to fulfil their needs, such as agriculture and settlement. In an administrative area, land cover change occurs in several locations and sometimes grouped in adjacent locations. Each location may have different type of land cover change. The understanding of land cover dynamics can be enhanced by identifying the dominant land cover change, which will lead to a better land cover management. By analysing each grouping area, it indicates the pattern and concentration of the type of land cover change, so it can appropriately support the regional management and development strategies. Hence, it is necessary to identify the characters of each grouping area so that the characteristics of the class of land cover change can be obtained. The dominant land cover change is identified by analysing the mode of land cover change in each area. The results show that the dominant land cover change in West Java in 2005-2010 are heterogeneous, with the top three broadest classes changes are rice field to field/moor (144,662.76 hectares), forest to fields/moor (36,339.84 hectares), and fields/moor to settlement (36,069.39 hectares). This study managed to identify 99.94% of the grouping areas are the dominant land cover change and 0.06% of the grouping areas are remained the same class. These results are expected to continue the development of land cover change model in West Java.


Introduction
There are three components that affect the global environmental change, i.e. the increasing concentration of carbon dioxide in the atmosphere, alteration of the global nitrogen cycle, and land use/ land cover change (1). Among these components, land use/ land cover change is the most important in understanding the global environmental change (1)(2)(3)(4). Therefore, research on land use/ land cover change is important in order to design the strategies to overcome sustainable challenges (2,4,5).
Land use/ land cover change is the result of human activities and impacting the environment, including climate change up to the Earth's ecosystem (6)(7)(8). Land changes that are often used to fulfil the human needs can reduce the portion available for other species and ecosystem functions (9). Their needs of shelters, for example, make forests turn into settlements (10).
Land cover changes are spreading anywhere in a particular area. So, it is necessary to analyse the patterns of the land cover changes. The information of land change patterns can be the basis for supporting the strategies in managing and planning the regional development appropriately (11). Land cover change patterns are not randomly or uniformly distributed, but clustered (9). It is because the land does not develop independently at each location, but affecting nearby and distant locations (12). For example, a research conducted in Bandung area, the result shows that land cover changes in Bandung area are clustered [15][AJK1]. Based on the former studies (13)(14)(15)(16)(17), cluster analysis is able to identify the pattern and concentration of data that being the object of the research, so the stakeholders can plan preventive actions appropriately. One aspect that needs to be observed from the cluster is the dominant land cover class change. Through this, it can identify the dominant land cover class change partially, not as a whole study area, so that in one study area can be found various dominant class changes occur in several locations. This identification can support smart and sustainable land use management practices (18). It is also expected to increase the understanding of the land cover dynamics, which can lead to more effective policy making, planning, and implementation by the government (19) [23][AJK2]. Therefore, a cluster-based analysis of dominant land cover class changes will be conducted in this study.

Data
The data used are land cover of West Java province in 2005 and 2010 in raster format with a 90 meters spatial resolution. These raster data have been classified into eight land cover class, i.e. forests, fields/ moors, plantations, settlements, rice fields, bushes/ shrubs, rivers/ lakes/ reservoirs/ situ, and fishponds/ ponds/ swamps. These data can be seen in figure 1 and figure 2 below.

Spatial Cluster
The data is processed by applying spatial cluster analysis to generate land cover change clusters in 2005-2010. The method used is Cluster and Outlier Analysis. This method identifies high clusters (High-High), low clusters (Low-Low), and spatial outliers using Anselin Local Moran's I statistics (20)(21)(22).
Spatial outliers represent spatial objects whose attribute values (non-spatial) are significantly different from the values in their spatial neighbourhoods (23,24). As an illustration, a low value point is surrounded by a neighbour with a high value (Low-High), vice versa (High-Low). In contrast to outliers in general statistics, which indicating extreme values in the data and must be removed from the analysis process, spatial outliers can provide additional information in various applications (23).
Based on Tobler's First Law of Geography (25), "everything is related to everything else, but near things are more related than distant things," adapted to the study of land cover change, the LH outlier has the possibility to change in the future. Because the central point that has not changed (L) has the potential to be affected by the neighbour that has changed (H). Based on this possibility, a land cover change cluster in this study consists of HH and LH. A total of 2590 cluster polygons were generated as can be seen in figure 3.

Identification of Dominant Land Cover Class Change
After obtaining the cluster, the dominant analysis is performed by calculating the mode of land cover change that occurs in each cluster. Mode is the most frequently occurring value in the data set (26). Mode is usually used to determine the level of occurrence of an event.
This stage is performed by overlaying land cover change data and cluster polygons. Then the highest frequency of land cover change is calculated in each cluster. By applying this, the dominant land cover class changes can be identified in each cluster.

Result and Discussion
The results of the analysis of dominant land cover class change are shown in table 1.   figure 4 and figure 5.  figure 4 and figure 5, it can be stated that the change of fields/ moors to settlements is the most because they are divided into small-sized clusters, while rice fields to fields/ moors clusters is divided into a bigger size.
Another thing to note is the change that dominant turn into settlements, since West Java's population growth is among the highest in Indonesia.

Conclusion
Based on the result of the analysis, it can be concluded that: 1. Overall, the dominant land cover change in each cluster in West Java is heterogeneous, with the three broadest class changes are rice fields turn into fields/ moors (144,662.76 ha), forests to fields/ moors (36,339.84 ha), and fields/ moors to settlements (36,069.30 ha). 2. Based on the total area of the clusters of land cover class changes in West Java in 2005-2010, this analysis managed to identify 99.94% of the dominant land cover class change and 0.06% are remained the same class.

Acknowledgments
The authors would like to thank the ITB Research and Innovation Program 2019.