Multi-temporal analysis of landslide susceptibility in the Greater Bogor Area and its relation to land use change and rainfall variation

Landslide susceptibility in the Greater Bogor Area is a significant concern due to its hilly topography, soil types, and high rainfall. This research examines the geographical and temporal variations in landslide susceptibility and the relationship between land use change, rainfall, and landslide occurrence. Landsat 7 ETM+, CHIRPS, and DEM satellite data were used to simulate landslide susceptibility. The study used 2000, 2010, and 2020 data to examine land use and precipitation changes. The weighted overlay method created a landslide susceptibility model that considered precipitation, land use, slope, rock, and slope. The association between land use change, rainfall, and landslide susceptibility was examined using correlation analysis. According to the findings, the Greater Bogor Area is highly prone to landslides, with a growing tendency in landslide events. 22.48% of the area has a very high susceptibility classification to landslide; meanwhile, 66.46% has a high susceptibility classification. Moreover, this study also found that the land use factor correlates more positively with landslide susceptibility than the rainfall factor. Overall, this study underlines the need to take land use dynamics and rainfall patterns when assessing landslide susceptibility and the need for effective mitigation measures in the area.


Introduction
Landslides have been one of the most frequent disasters occurring in Indonesia, with more than 1,000 occurrences from 1981 to 2007, or 49 events per year on average [1].Landslide occurrence is highly associated with the physical condition of a region, such as topography, type of soil, and runoff.The Greater Bogor Area (Bogor City and Bogor Regency) is an urban area with hills and mountainous morphological conditions.Almost half of the area, specifically 43.64%, has a sloping morphology, and 23.38% of the area has a morphology of 0-15% slope [2].Additionally, around 72.32% of the Greater Bogor Area contains high concentrations of clay that cause a high potential occurrence of landslides in the rainy season [3].
The historical data on the number of landslide events exposes the Greater Bogor Area's susceptibility to landslides.According to the Bogor Regional Disaster Management Agency (Badan Penanggulangan Bencana Daerah Kabupaten Bogor/BPBD Kabupaten Bogor), there were 1,154 landslide events recorded from 2019 to 2021.The data also indicates that the summary of landslide occurrence in Bogor Regency has an increasing trend from 212 landslides in 2019 to 514 landslides in 2021 or an increase of about 140%.West Java Province's open data portal revealed that from 2019 Factors that cause landslides can be separated into four main causal factors: ground conditions, geomorphological process, physical process, and human-made process [5].The changes related to those factors will also dynamically influence the land sturdiness and landslide occurrences.While some of the factors that may influence the dynamics of a landslide are naturally given, some rapid changes come from human activities.Notable human-made phenomena that drift the dynamics of landslide factors are land use and climate change caused by high socio-economic demand from urban areas.Land use change can trigger landslide events, altering the soil's ability to hold runoff [6].It also depleted green space, reducing the area's ability to hold water and increasing water runoff.Meanwhile, climate change refers explicitly to the change in precipitation rate related to hydrometeorological hazards.
Landslide susceptibility assessment is usually used to map the area's susceptibility to landslide occurrence.It is a quantitative method in which predicting where, when, and how often landslides will appear will give assessment and early warning for impacted groups and related stakeholders about landslide occurrence to prepare the mitigation and emergency action needed [7].There have been many studies on landslide susceptibility model's variation such as weighted, probabilistic, regression, geotechnical and factor of safety, likelihood, and weighted linear combination models [8].The weighted overlay method is considered one of the standard methods to estimate land susceptibility because it is relatively simple and easy to use [9].In Indonesia's context, this method has been used by authorised institutions and many referred by other local institutions to measure landslide susceptibility in Indonesia.Such as a publication from the Centre of Soil and Agroclimate Research (Pusat Penelitian Tanah & Agroklimat/Puslittanak) in 2014 [10].The critique of the method relies on the authorised analysis results and assumption on determining land susceptibility value.In contrast, the significance of each land susceptibility factor may differ in different contexts [9].This paper uses the geospatial approach to identify landslide susceptibility in the Greater Bogor Area context with the rapid growth of built-up areas and high precipitation.Among different factors that may trigger a landslide in a particular area, land use change and precipitation will be utilised to estimate landslide susceptibility in the Greater Bogor Area.The correlation between land use change and precipitation as factors with landslide empirical data is also conducted to measure both factorst on the landslide occurrence.Thus, the purpose of the research is as follows: 1. Assess the landslide susceptibility based on land use change, precipitation trend, and other factors.2. Identify land use change and precipitation correlation with landslide occurrence.

Methodology
This study attempts to temporally assess the spatial variation of landslide susceptibility and determine the relationship between land use and rainfall on the susceptibility condition in the Greater Bogor Area.This study used several satellite data such as Landsat 7 ETM+, Climate Hazard InfraRed Precipitation with Station (CHIRPS), and Digital Elevation Model (DEM) to perform landslide susceptibility modelling.

Acquisition Data
The data used are satellite and long-term meteorological data taken for ten years starting from 2000, 2010, and 2020.Landsat 7 ETM+ images are used for land use modelling.CHIRPS data is used to obtain rain parameter data.Supporting parameter data, such as soil and geological vectors, are taken from existing data.

Rainfall analysis
The rainfall data per day collected from the Indonesian Meteorological, Climatological, and Geophysical Agency (Badan Meteorologi, Klimatologi, dan Geofisika/BMKG) and CHIRPS data were validated to the closest BMKG rainfall and weather station, which is the Citeko Meteorological Station.The analysis of regional rainfall is continued by finding the value of the biggest daily rainfall data for each year from 1991 to 2020 and frequency analysis involving an 80% probability of all daily rainfall data for each CHIRPS point is conducted.To validate this analysis, the Pearson correlation is employed, as depicted in the following formula (1): C i = CHIRPS daily rainfall (mm) O i = BMKG daily rainfall (mm) Furthermore, the rainfall trend analysis can then be concluded using the slope value.The regional rainfall in Bogor Regency will be analysed spatially using the IDW interpolation in ArcGIS software by making the isohyet map.The isohyet employed an 80% probability and encompassed four distribution approaches: Normal distribution, Log-Normal distribution, Gumbel distribution, and Log Pearson III distribution.The process of determining rainfall proceeds by choosing the most suitable distribution to minimise the statistical approximation, with the estimation of the 80% probability frequency [11].

Land Cover Change Modelling
The land use (LU) data were derived from Landsat 7 ETM+ Level 2 Collection 2 Tier 1 using supervised classification methods.Supervised classifications were done using the Random Forest (RF) machine learning algorithm, an algorithm proven to have high accuracy for land use classification.This accuracy comparison with other algorithms, such as Support Vector Machine (SVM) and Classification and Regression Trees (CART), was verified by utilising a confusion matrix to calculate the resulting land use classification's overall accuracy and kappa index [12].Further spatial and temporal filtering was done to produce the processed surface reflectance at the Greater Bogor Area in 2000, 2010, and 2020.For accurate measurements, the training sample areas were divided into 80% sampling for classification and 20% testing [13].
Land cover change in the Greater Bogor Area was measured by comparing 2000, 2010, and 2020 land cover maps.The MOLUSCE (Modules for Land-Use Change Simulation) plugin in QGIS software was used to examine land cover change quickly.MOLUSCE uses several criterias, including DEM (digital elevation model), slope, and distance from the road and from built-up area.This will act as spatial variable maps in the MLP-ANN learning process to predict the effect of percentage land-use change on landslide susceptibility in the Greater Bogor Area.MOLUSCE operates using an equation to calculate the difference in the land use percentage, which can be seen in ( 2): LCC = Land cover change percentage (%) LC i = Land cover area for the recent year (km 2 ) LC 0 = Land cover area for the previous year (km 2 ) A = Area of Bogor Regency (km 2 )

Landslide susceptibility modelling
The weighted overlay method was chosen to build the landslide susceptibility model.This method is considered semi-quantitative and is mostly used for landslide susceptibility evaluation and hazard zonation studies [9].Data processing and collection were done using Google Earth Engine (GEE) and ArcGIS software, along with data from the Indonesian Government database and Climate Hazards Group InfraRed Precipitation with CHIRPS database.The weighted overlay method produces landslide susceptibility distribution in the Greater Bogor Area and is classified in Table 2, from which the intervals were derived (3).Very Low 1 -1.9 The final data processing step is to complete the weighted overlay method.Every output of data processing was made sure to be in a raster data format.Data harmonisation was done by matching each reclassified factor layer's spatial resolution and coordinate reference system (CRS) using the nearest neighbour resampling method.After that, every factor layer was summed up according to its respective weights, as mentioned in (4).

Correlation of landslide susceptibility with variables of land use change and rainfall
Multiple correlations were done to measure the linear association of more than two variables.In this study, rainfall (x) and land use change (y) were seen as independent variables, meanwhile landslide susceptibility (z) was the dependent variable.The formula of statistical analysis used for validation is as seen in ( 5).

Location of the study
Greater Bogor Area is an administrative area located inside the West Java Province, Java Island of Indonesia.The area comprises two parts: an administrative city, Bogor City, and a regency, Bogor Regency.This area has an area of 2,986.2km 2 .According to its morphology, the area consists of lowland areas in the northern area and mountainous areas in the southern area.The administrative map of the Greater Bogor Area as the study area can be seen in the following figure.

Historical landslide data in Bogor
The Greater Bogor Area is prone to landslides due to its physical characteristics.According to the Bogor Regional Disaster Management Agency (BPBD Kabupaten Bogor), there were 1,154 landslide events recorded from 2019 to 2021.The data also indicates that the number of landslide events in the area is increasing, as shown in Figure 2. Based on the figure above, landslides in Bogor usually increase from October until February.Furthermore, those months are also in line with the rainfall pattern in Indonesia.This fact can also be a noteworthy factor, especially in areas where an increased number of landslides are expected during the rainy season, given that Bogor is a particularly susceptible region [14].Furthermore, the monthly rainfall data processing from CHIRPS data also showed the same pattern as the historical landslide data in Bogor, as shown in Figure 3 below.The same pattern of rainfall and landslide events in Bogor might need significant consideration as Bogor becomes more vulnerable to landslides when the rainfall season occurs.Based on the validation test results in Table 4., the average correlation value from two locations is 0.136.Such correlation indicates that the two types of rain data have different daily rainfall patterns.This fact is caused by differences in data collection methods at the BMKG Stations and CHIRPS, where the station data uses observational data of rain falling to the earth's surface.In contrast, CHIRPS data uses data from cloud thickness, supported by the large Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values showing that the CHIRPS daily rainfall data is overestimated compared to the BMKG Stations daily rainfall data.The daily rain data used to calculate the landslide susceptibility is the most complete daily rainfall assessment point, CHIRPS daily rainfall data.
The data from CHIRPS daily measurements were gathered from various points in 30 years.These data were taken into account when determining the minimum data length necessary to establish a standard for the typical precipitation levels in a particular region [15].Then, to conduct the frequency analysis with an 80% probability, the data is transformed into the immense daily rainfall value for the same 30-year duration.The frequency analysis process involves utilising four distributions, including the Normal, Log-Normal, Gumbel, and Log Pearson III.The aim is to identify the most suitable distribution that minimises statistical estimation error and accurately determines the frequency of rainfall at an 80% probability [16].The result of the rainfall map from the frequency analysis is presented in Figure 4.  shows the spatial distribution of maximum rainfall in the Bogor area, processed from 80% probability from 1991 to 2020.More intense daily rainfall was in Bogor City, from 100 to 150 mm, while other areas experienced less extreme rainfall.The more extreme rainfall in urban areas of Bogor occurred due to the Urban Heat Island, as recent studies highlight the trend of this phenomenon impacting more concentration of the area with precipitation increase [17].Although rainfall in other areas experienced less than in urban areas, rainfall was also categorised as extreme, according to BMKG.Based on those facts, it is also necessary to consider the rainfall trend in Bogor within 30 years.The daily rainfall trend of Bogor itself can be seen in the figure below.very high with the value of over 500 mm/month, according to BMKG.The high rainfall impacts the number of water resources in Bogor and the susceptibility to hydrometeorological hazards such as landslides, especially in the slopes and mountainous areas.Frequent landslides in Bogor mainly occurred in the discharge zone of the groundwater basin, and it was commonly caused by disrupted slope stability from intensive human activities [18].

Land cover change
Land use is one of the parameters studied in this research to see the influence of land dynamics on landslide susceptibility assessment.This study demonstrates the use of Landsat 7 satellite data to model land use change during 2000, 2010, and 2020.An accuracy assessment was conducted to evaluate the quality of the classification results and assess the effectiveness of sampling pixels into different land use categories.The overall accuracy of the classified imagery was compared to the actual land cover conditions obtained from the corresponding ground truth data with overall accuracy values of 97.62%, 99.52%, and 87.19%.The dominant land use type classified in Bogor is a low-zoned area, which covers about 70% of the total study area, followed by agricultural area, forest, and built-up area.Land use change significantly influences the distribution shaping the landscape of the region's landslide susceptibility in the Greater Bogor Area.Over the decades, the occurrence of land use change has led to the transformation of land use patterns that can increase the susceptibility of certain areas to landslides.Land change refers to alterations in land cover and land-use patterns, including deforestation, urbanisation, agricultural activities, and infrastructure development.These changes can profoundly impact slope stability and increase certain areas' landslide susceptibility.Clearing forests for agriculture, settlements, or industrial purposes exposes the slopes to increased water runoff, leading to a higher likelihood of slope failure and landslides.The loss of vegetation reduces the roots' binding effect, which helps stabilise slopes and prevent soil erosion.Without this natural protection, the soil becomes more prone to erosion and landslides, especially during intense rainfall.The cumulative effect of land changes on the distribution of landslide susceptibility in Bogor is evident in the spatial patterns of landslide-prone areas.Vulnerable zones coincide with areas undergoing extensive land-use changes, such as deforestation and urbanisation.

Landslide susceptibility
Based on a review of topographic characteristics, soil types, and high rainfall intensity, landslide susceptibility appears to be a significant problem in the Greater Bogor Area.According to the spatial data analysis of landslide susceptibility in the area, Southern Bogor appears to have a higher landslide susceptibility than the northern side.Further explanation about the spatial distribution of landslide  As the landslide susceptibility estimation was done from 2000 until 2020, comparing the landslide susceptibility change is possible.Spatial analysis of landslide susceptibility estimation exposes an increase in landslide-prone areas.The most noticeable change can be seen in a few hilly areas of Cigudeg, Nangguang, Pamijahan, Klapanunggal, or Citeureup Sub-district.As for the central northern area of Bogor and its urban area, there are few changes regarding landslide susceptibility.The scoring method used to estimate susceptibility resulted in the classification of landslide-prone areas, divided into four categories: very low, low, high, and very high.7 indicates the amount of the area of each landslide susceptibility class, which is mostly dominated by high levels of landslide susceptibility.The result indicates that the area with a very low classification of landslide susceptibility is rare and almost nothing compared to the whole area of the Greater Bogor Area.Meanwhile, the area with a low classification of landslide susceptibility is around 11-12% of the total area located in the central-north side of the Greater Bogor Area.Referring to the visual interpretation of the landslide susceptibility map and its factors, the agglomeration of this low landslide susceptibility area is associated with the urban area of Bogor, which features a gentle slope with a value of below 15%.
According to the visual interpretation of the landslide susceptibility map between 2000-2020, the Greater Bogor Area is always dominated by a high level of landslide susceptibility.It is confirmed by the data shown in Table 5 that high landslide susceptibility areas dominate 66-69% of the area, located in hilly areas mostly occupied with bushes and shrubs.The slope level in this area has reached above 15%, which theoretically can exacerbate the potential of landslide occurrence.Moreover, the land use type dominated by shrubs and bushes could not hold the amount of runoff produced in this high precipitation area.Furthermore, very high classification of landslide susceptibility dominated almost 20% of the total area, mostly appearing in the same area with high landslide susceptibility, specifically in the southern side where precipitation is between 3,800 -4,600 mm/yr.However, even though several areas with very high classification landslide susceptibility have low rainfall, the result is still very high because of the scoring on land use and slope factors.Most of these landslide hazard classifications have slopes above 45%.
Landslide susceptibility rose in heavily vegetated forest land converted to agricultural land from 2000 to 2020, suggesting the impact of land use and land cover change on landslide susceptibility.According to the study, slopes with volcanic lithology are more prone to landslide susceptibility in the Greater Bogor Area, and the increased susceptibility to landslides in the southern highlands can be attributed to the steeper slopes and the prevalence of older deposits, primarily characterised by volcanic lithology or volcanic sediments, and the intensity of rainfall is high, resulting in the magnitude of erosion forces on steep slopes.
As mentioned before that to analyse the landslide susceptibility, slope, land use, rainfall, and soil type data are used.However, between 2000-2020, only land use and rainfall data changed significantly each year.The information about slope data and soil type remained the same.However, landslide susceptibility changes still occurred, which strongly correlated with the rapid alteration of land use due to urbanisation and anthropogenic activities, as well as climate change-induced rainfall change.Increased rainfall intensity and soil type variables can influence landslide susceptibility through their capacity to trigger erosion, especially in Bogor Regency.Landslide susceptibility change is hypothetically affected by the change of several precipitation class areas.As can be witnessed in the rainfall map of 2000-2020, a change in areas classified as 'very wet' occurred in the same area of the increased landslide susceptibility.

Correlation of landslide with land use change and rainfall
Based on the correlation coefficient, which quantifies the relationship between rainfall and land use change concerning landslide susceptibility, both variables exhibit a negligible correlation, as indicated by coefficients close to zero shown in Table 8.The analysis reveals a positive correlation between land use change and areas prone to landslides in the Greater Bogor Area.Spatially, the distribution of highly susceptible landslide areas varies with changes in land use over time.The transition from densely vegetated areas to open areas and agriculture contributes to an increased susceptibility to landslides.However, the analysis suggests that rainfall has a limited impact on changes in the area susceptible to landslides, with a weak negative correlation.
Nevertheless, thick volcanic soil in areas susceptible to landslides poses a significant susceptibility assessment factor when subjected to frequent heavy rainfall [19].The Greater Bogor Area experiences a high rainfall rate, typically 300-500 mm/month.Intense rainfall leads to increased water retention, resulting in higher discharge and volume, exerting pressure on the soil particles and triggering movement in the passive clay soil on the slopes.In summary, the scientific analysis suggests that while there is a weak relationship between rainfall and land use change with landslide susceptibility in the area, converting land from high-density vegetated areas to open areas and agriculture increases landslide susceptibility.The thick volcanic soil in landslide-prone areas can pose a danger during periods of high-frequency rainfall.

Conclusion
The Greater Bogor Area is highly prone to landslides due to high rainfall and land use change.Land use change positively correlates to the distribution and susceptibility of landslide-prone areas.Deforestation, urbanisation, and agricultural activities have altered land cover patterns, making particular places more vulnerable to landslides.Vegetation loss affects slope stability and adds to soil erosion, making the soil more prone to landslides, particularly after heavy rain events.The spatial analysis of landslide susceptibility shows that the southern portion of Bogor is more susceptible to landslides than the northern area, caused by intensive land use change in high-slope areas.The area is experiencing increased landslide events, notably from October to February, corresponding to the rainfall pattern.
On the other hand, the correlation between rainfall and landslide susceptibility is minimal, showing that rainfall alone does not substantially influence changes in the area susceptible to landslides.However, the study emphasises the importance of considering the long-term rainfall trend in the area.Although annual rainfall has fallen over the last three decades, monthly rainfall in the area was still classified as high and very high, according to BMKG.The area is vulnerable to landslides due to the high land use change, rainfall severity, and steep slopes, and volcanic soil.The findings, in summary, highlight the intricate interplay between land use change, rainfall patterns, and geological factors in determining landslide susceptibility in the Greater Bogor Area.These factors become crucial for more effective landslide management, for example, enhancing land use planning and regulation, monitoring and early warning systems, and vegetation and erosion control.Effective landslide assessment should consider these elements to limit the potential effects of landslides and safeguard the local community's safety.
rxz = correlation between rainfall and landslide susceptibility ryz = correlation between land use change and landslide susceptibility rxy = correlation between rainfall and land use change rz,xy = multiple correlation

Figure 1 .
Figure 1.Map of Greater Bogor Area

Figure 4 .
Figure 4.shows the spatial distribution of maximum rainfall in the Bogor area, processed from 80% probability from 1991 to 2020.More intense daily rainfall was in Bogor City, from 100 to 150 mm, while other areas experienced less extreme rainfall.The more extreme rainfall in urban areas of Bogor occurred due to the Urban Heat Island, as recent studies highlight the trend of this phenomenon impacting more concentration of the area with precipitation increase[17].Although rainfall in other areas experienced less than in urban areas, rainfall was also categorised as extreme, according to BMKG.Based on those facts, it is also necessary to consider the rainfall trend in Bogor within 30 years.The daily rainfall trend of Bogor itself can be seen in the figure below.

Figure 5 .
Figure 5. Daily rainfall trend in Bogor for three decades Based on the figure above, the rainfall trend in the first and second decades did not significantly increase, although the trendline showed positive R 2 values.Compared to that, the increasing daily rainfall trend became more apparent in the last decade, where the R 2 was 0.0206.Bogor has a monsoonal type of rainfall with a drier season in June-July and more rainfall in October-November.Bogor experienced a decreasing trend of rainfall within 30 years.The average annual rainfall in 1991-2000, 2001-2010, and 2011-2020 were 5510, 5063, and 4767 mm, respectively.However, the monthly rainfall in Bogor is still classified as high at a value of 300-500 mm/month and .1088/1755-1315/1313/1/012025 11 susceptibility in the area shows in the following maps, where the area that is coloured red symbolises the area with the classification of very high susceptibility of landslide occurrence.

Table 1 .
Research data

Table 4 .
Result of Daily Rainfall Validation

Table 5 .
Accuracy Assessment of Land Use Supervised Classification

Table 6 .
Land Use Change in Greater Bogor Area in 2000, 2010, 2020

Table 8 .
Correlation Land Susceptibility with Land Use Change and Rainfall