Combining Landsat, VIIRS Night-time Light, and Sentinel-1 SAR for Spatial Flood Risk Assessment in Coastal Area: an Earth Engine Cloud Computing Process

The intensity of development in coastal areas stimulates various potential issues such as flood disasters. This study aims to demonstrate the importance of latest methods and geospatial data as inputs for coastal spatial planning policies in efforts to reduce flood disaster risks. Leveraging spatial analysis with cloud computing through Google Earth Engine (GEE), this research assesses flood risk components—hazards, vulnerability, and capacity. The method involves processing SAR Sentinel-1 data to map flood inundation as a representation of hazards, analyzing Landsat and WorldPop data to evaluate vulnerability, and assessing capacity by utilizing VIIRS nighttime light level imagery to determine economic activities. The chosen research study location is the coastal area of Pekalongan due to frequent flood disasters throughout the year. The results demonstrate that cloud computing is capable of assessing flood risks. The flood inundation model using SAR data covers an area of 2,780 hectares with an accuracy of 96.75%. The analysis also reveals the highest vulnerability level, reaching 15.7% (946.32 hectares) of the total area. The capacity analysis indicates a medium to high level of 15% (913.6 hectares). The assessment of flood risks in the coastal area is dominated by the medium to very high-risk class, covering 43% (2,631.84 hectares) of the area. In conclusion, integrating cloud-based flood risk modeling into spatial planning is crucial, considering disaster resilience for sustainable human habitats.


Introduction
Global discussions on coastal disasters often focus on the increasing occurrence of damaging floods in terms of financial and environmental impacts [1][2][3][4].The rapid pace of development, coupled with limited land availability for the growing population, has resulted in problems when not properly managed [2,3].By 2100, almost 15-23% of the world's population is predicted to reside in coastal areas [7].This situation threatens the sustainability of coastal living [8,9].
Extreme climate change significantly contributes to rising sea levels, eventually leading to floods [6].The IPCC predicts a rise of up to 48cm by 2050 and an ongoing increase of 16mm per year until 2100 [10].Strategies and adaptive mitigation measures need to be developed to reduce the risk of disasters in coastal areas, particularly floods [5].This problem becomes "homework" for all stakeholders, particularly the government, considering that coastal areas undergo continuous dynamics 3 Pekalongan coast.The research area includes 21 villages in the regency and 13 villages in the city, with a total area of 6,035 hectares (Figure 1).The coastal area of Pekalongan is highly vulnerable to tidal flood threats.In June 2022, this disaster caused physical and financial losses to the surrounding community.Based on a primary survey (Figure 2), there are areas affected by tidal floods, such as along Tegaldowo Highway and several roads leading to Api-Api Village, which resulted in disrupted community connectivity.The damage caused by tidal floods also contributes to the issue of land subsidence, which cannot be avoided and ultimately affects the capacity of the community [23].This is due to the proximity of residential areas to rivers and the Java Sea coastline [8].1).The flood model results need validation to proceed to the next stage.Therefore, the researchers conducted a primary survey to obtain validation points.We collected the data from publicly accessible sources such as GEE, Restservices, and OpenStreetMap.The data was analyzed using geocloud-computing techniques, and the results are available via the provided link (code.earthengine.google.com/f5c2f984c053c8ea574bfcd4040d084e).

Methodology
This research uses a quantitative technique to collect data from secondary and primary surveys.Primary data was obtained by conducting fieldwork as a form of validation for the developed model.In contrast, secondary data is acquired through literature review, downloading open-source imagery facilitated by a Cloud platform, and accessing relevant institutional websites.The methodology is depicted in the research framework (Figure 3), from goal formulation to the final results.

Hazard
We utilized the Change Detection And Threshold (CDAT) method to model flood inundation using SAR Sentinel-1 data [11,18].The script used for this method was provided by UN-SPIDER (https://www.un-spider.org/advisory-support/recommended-practices/recommended-practice-googleearth-engine-flood-mapping/step-by-step).Then the results of this method were adjusted with DEM data and interpolated validation points to generate a depth model.
In order to classify floods, it is necessary to compare two temporal SAR images: one taken before the flood and one taken during the flood.This change detection technique is called the Ratio Image [24].This technique utilizes the calculation of the ratio of the backscatter coefficient values in the two images [25].When an image is brighter, the ratio value increases, and when it is darker, the ratio value decreases.For mathematical expression (Eq.1), the RI formula can be employed.

𝑅𝐼 = (𝜎°𝑣𝑣(𝑎𝑓𝑡𝑒𝑟 − 𝑓𝑙𝑜𝑜𝑑))/(𝜎°𝑣𝑣 (𝑝𝑟𝑒 − 𝑓𝑙𝑜𝑜𝑑))
(Eq. 1) Where: σ°vv(after-flood) = The backscattering coefficient of the image during a flood of VV polarization σ°vv(pre-flood) = The backscattering coefficient of the image before it occurs Threshold calculation is one of the stages in flood modeling that identifies flood areas [18,20].This technique utilizes the results from the RI calculation while still creating sample data based on land use classification.In order to identify areas affected by floods, we categorize them into two types: water and urban areas.Eq.2 we will use is called σwater.
(Eq. 2) Where: Water depth is determined through the process of flood modeling.According to the analysis results, we require specific location data showing the depth of flooding obtained from a primary survey.The DEM data indicates the slope of the coastal area, and the data taken is sloped greater than 3%, which is the threshold for flood-prone areas [24].This modeling results in the flood inundation depth selected based on slopes greater than 3%, thus resulting in several classes that predict flood inundation in Coastal of Pekalongan.

Vulnerability
The vulnerability model requires three indicators: population density, land use, road network, and healthcare facility points [15].The modeling process (Table 2) is conducted by weighting the factors based on the modified approach [15].The vulnerability model (Eq.3) calculation involves assigning weights while considering the extent of potential flood inundation.

Capacity
Capacity modeling combines physical and economic aspects in Coastal of Pekalongan.The model aligns with the Sendai Framework for Disaster Risk Reduction 2015-2030, focusing on two priorities: understanding disaster risk and investing in disaster resilience (Table 3).These priorities are further divided into three indicators to assess the level of risk reduction achieved.

Table 3. Capacity Modeling Index Capacity Priority
Indicator Index Total

Understanding of capacities for disaster risk reduction
The high level of the community's economy 0-1 1

Invest in disaster risk resilience
Ensuring the existence of health facilities in order to strengthen public resilience 0-1 Ensuring the effectiveness of the community in reaching safe points 0-1 Physical capacity modeling involves healthcare facilities and road networks, reflecting that more facilities and infrastructure correspond to a greater capacity [27,28].In Coastal of Pekalongan, the level of physical capacity is determined using the Kernel Density method.The results are scaled from 0 to 1, with a higher value indicating a higher capacity.
Economic capacity modeling utilizes night-time satellite imagery.This imagery can reveal human activities and economic conditions in a given area, moreover, this imagery can estimate Gross Regional Domestic Product (GRDP) down to the village level [10,22].For this study, we have adopted and modified the approach to suit the conditions in Coastal of Pekalongan.Night-time satellite imagery is accessed through GEE, allowing for data retrieval and analysis.
The light levels in the imagery are adjusted using the 'avg_rad' function in GEE, measured in nW/m 2 .sr.In order to match the current conditions, we use the Lpkl (Eq.4) formula to convert night-time data into W/m 2 .sr,which is the standard unit for measuring economic activity.The values obtained are categorized into five groups based on activity level, ranging from high to low.
Then the reclassified data is combined with the administrative boundaries of Coastal of Pekalongan, incorporating Gross Regional Domestic Product (GRDP) at constant prices in millions per city/regency using Zonal Statistics.We have developed a new method to calculate the PDRBkel/des (Eq.5) at the village level.This method involves using a balanced approach where we divide the PDRB of the village by its population.

Risk
We combine the results of hazard, vulnerability, and capacity models to map the risk of tidal floods using weighted proportions.To determine the weight of each component, we use the fuzzy membership method and then apply the raster calculator to each one.The final output of the model yields the flood risk index in Coastal of Pekalongan.The quantification formula (Eq.6) for the tidal flood risk in Coastal of Pekalongan is Rpkl21.

Flood Hazard at Coastal
The spatial model of tidal flood involves analyzing flood potential and depth.Using the Google Earth Engine (GEE) platform and customized code from UN-SPIDER, a flood potential model was developed to effectively extract flood extent.This analysis utilized the Sentinel-1 dataset, an active sensor imagery capturing data every six days [23,24].This efficiency supports the use of SAR data for identifying tidal flood potential in Coastal of Pekalongan.Historical records before and after flood events were chosen as reference periods: pre-flood from March 25, 2021, to April 2, 2021, and postflood from December 25, 2022, to January 6, 2023.VV polarization was selected for flood identification due to its suitability [32].Setting a threshold value of '1.35' based on references improved accuracy in flood area determination for the specific location [33].
The result of the process yielded a potential flood extent of 2,780 hectares (Table 4).To refine the model results, the researchers conducted field and online news surveys to validate the model.Out of 123 sampled points, the flood extent model did not capture eight (Figure 4).The tidal flood predominantly inundated densely populated residential areas and infrastructure such as road networks, markets, educational facilities, bridges, and fish ponds.The flooding of infrastructure, especially road networks, will undoubtedly hinder community activities.The tidal flood affected major roads, specifically Jeruksari Street and Tegaldowo Street, which serve as alternative routes to Pekalongan Regency.The flood extent model provided an extraction; therefore, the DEM data plays a crucial role in determining the depth of the flood.After analyzing the depth model, a Fuzzy Membership analysis was conducted to produce the flood hazard index in Coastal of Pekalongan (Figure 5).The higher the value (1) of the index, the greater the vulnerability to flood hazards.

Coastal Flood from Vulnerability Perspective
The development of the vulnerability model in Coastal of Pekalongan needs to consider physical aspects such as road networks and the number of healthcare facilities, environmental aspects by examining land use, and social aspects by looking at population density in an area.

Psychical Vulnerability
The analysis process of physical vulnerability was divided into two components, namely road networks and healthcare facilities.The development of this model uses the kernel density method, where the more dense an area is with road networks or facilities, the lower the vulnerability in that area [34].Model results reveal a notable vulnerability in the coastal of Pekalongan (Figure 6).This vulnerability is particularly influenced by numerous ponds in Coastal of Pekalongan that contribute to the susceptibility of the region's road networks.Upon scrutinizing the road network, it becomes evident that the eastern coastal part of the city experiences fewer damaged vulnerable zones.Analyzing the physical vulnerability of healthcare facilities in the area demonstrates that those concentrated in urban areas exhibit lower vulnerability.Conversely, higher vulnerability levels are scattered across various locations, often aligning with the distribution of built-up areas.Notably, the presence of facilities in coastal areas significantly impacts the created vulnerability model.

Environment Vulnerability
The researcher developed an environmental vulnerability model by utilizing land use distribution.Land uses involving human and production activities are highly vulnerable [35].The author created four land use classifications: ponds and water bodies, dry fields, plantations, settlements, and cropland (Figure 7).This vulnerability model was constructed using Landsat-8 data.Cropland and settlements have the highest vulnerability values due to the presence of human activities and their production values [14].The classification results showed four classes, each with its vulnerability value.Water bodies had the lowest value of 0, dry fields and plantations had a relatively low value of 1, cropland had a moderate value of 3, and settlements had the highest vulnerability value of 5. Regarding spatial distribution, Coastal of Pekalongan has low environmental vulnerability mainly because water bodies cover a larger area than other classes.They account for 39% of the total area (Figure 8).

Social Vulnerability
In this research, spatial modeling of social vulnerability refers to population density using the dasymetric method.Dasymetric results better represent the distribution of the population exposed to disasters due to the consideration of delimitation boundaries in the form of settlements [36].The data used for analysis is WorldPop, which was accessed through GEE.This data provides the distribution of population density up to the most recent year.The data needs to be validated, so the author compares it with data per administrative area (village) in Coastal of Pekalongan.The model of population density distribution shows that each pixel in a raster represents approximately two individuals residing in that area.Based on the obtained model results, it was observed that population density is distributed more towards Pekalongan City (Figure 9).In contrast, the District areas have lower population density distribution.This indicates that population growth is linear with settlement patterns, where the more significant the extent of settlement development, the higher the population density.From a vulnerability perspective, areas with higher population density also represent higher vulnerability.Tirto Village is the area with the highest vulnerability class, covering an area of 117.32 hectares.Looking at the overall picture, Depok Village has the largest total area of 413 hectares with a moderate vulnerability class.

Spatial Model of Coastal Flood Vulnerability
The quantification of the vulnerability index in the spatial model was carried out using the fuzzy membership method, where the higher the predetermined value, the higher the vulnerability level in an area.The calculation considers the weighting prepared by the researcher with the formulation VPKL21.Based on the vulnerability index analysis, it was found that flood vulnerability has a high index in areas that fall within settlements (Figure 10).This model is consistent with the high population density in urban areas due to the dasymetric method [36].The areas with the highest vulnerability were also concentrated in Pekalongan City, which aligns with their physical vulnerability level.The low vulnerability class dominates Coastal of Pekalongan, covering an area of 3,675 hectares (Table 5).The presence of water bodies, the dominant land use classification, is the basis for the large extent of the low vulnerability class.On the other hand, the vulnerable class has a relatively insignificant area of only 495 hectares.This assessment indicates that Coastal of Pekalongan falls into the category of moderate vulnerability to coastal flood disasters on a large scale.However, special attention must be given to the highly vulnerable class, as it has a high population density.

Psychical Capacity
Based on the developed model results, Coastal of Pekalongan has a moderate road density capacity when viewed spatially.Considering the pattern of road networks in Coastal of Pekalongan, road density tends to be concentrated in urban areas, and the denser the roads, the higher the capacity.The highest road density was in the eastern coastal area, predominantly in Pekalongan Utara Sub-district, Pekalongan City.Based on field surveys, alternative roads connecting Pekalongan Regency are affected by flood inundation.This problem has implications for the movement of the community and disrupts various activities, such as economic activities.The physical capacity model of healthcare facilities was developed using the kernel density method, where the denser the distribution of healthcare facilities, the higher the population's vulnerability to accessing those facilities.Healthcare facilities can reflect resilience in facing flood disasters [37].The more healthcare facilities available, the higher the capacity level.The researcher used healthcare facilities such as hospitals and health centers as healthcare infrastructure that can cope with disasters.Based on the model results, the density of healthcare facilities tends to be concentrated in Pekalongan City (Figure 11).The availability and completeness of facilities in an area affect the capacity model [27,37].

Capacity Seen from the Level of the Economy
The economy is a crucial component for measuring the capacity of communities to deal with issues such as coastal floods.There is a correlation between human activities and the economy, where higher activity levels in an area indicate a growing economy [16,22,31].To measure this relationship, the researcher utilized spatial data that captures community activities, namely night-time satellite imagery.The researcher developed an economic capacity model using VIIRS/Night-time V/1 imagery, which can be accessed openly through the GEE platform.The higher the light level in an area at night, the higher the correlation with economic activity [39].night-time satellite imagery also can quantify Gross Regional Domestic Product (GRDP) at a smaller scale using proportion-based methods [21].The data were classified into four classes based on the light level (Figure 12).The basis for this classification is that the brighter the night-time light, the higher the night-time activity in Coastal of Pekalongan.Night-time activity indicates that such areas also have high urban activity.The level of economic activity forms the basis for the resilience and readiness of communities to strengthen their physical and non-physical foundations.Areas with no night-time light are considered to have nondominant economic activity.Therefore, the researcher used the total area (AdTotal) as the divisor to quantify GRDP using the PDRBkel formulation.The proportional results were then interpolated using GEE to generate continuous distribution.The analysis results show that economic activity tends to dominate coastal areas (Figure 13).The presence of highways connecting regencies and cities in Java Island leads to high economic activity in Coastal of Pekalongan.In contrast, the southern areas of Pekalongan Regency have lower dominant activity due to their physical characteristics, such as being mountainous.Economic capacity in a region can be seen through its Gross Regional Domestic Product (GRDP) level [21].Based on the developed model results, it was found that the economic capacity in Coastal of Pekalongan is concentrated in Pekalongan City.The total proportion of GRDP per kelurahan in Coastal of Pekalongan is Rp 3,754,110.67.When the value of GRDP is broken down per administrative area (kabupaten/kota) in Coastal of Pekalongan, Pekalongan City has a significantly higher value of Rp 3,299,986.In comparison, Pekalongan Regency has a much lower value of Rp 356,125.43 (Table 6).The significant difference in these values indicates differences in the economic level in Coastal of Pekalongan.The disparity in proportion values is primarily due to the distribution of high night-time light imagery in Pekalongan City.This result aligns with the higher capacity levels predominantly found in Pekalongan City.

Spatial Model of Capacity from Flood
When correlating with the Sendai Framework, Coastal of Pekalongan falls under Priorities I and III.The presence of models such as the economy falls under the priority of disaster risk reduction efforts with community self-reliance capacity.The economic level was measured through the Gross Regional Domestic Product (GRDP) in Coastal of Pekalongan, which tends to be concentrated in urban areas, particularly areas with high night-time light activity.The dense road network and healthcare facilities also indicate higher capacity in dealing with flood events in Coastal of Pekalongan (Figure 14).Based on the calculations, it was known that the lowest capacity class dominates Coastal of Pekalongan, covering an area of 3,736 hectares (Table 7).Factors such as accessibility, supporting facilities, and a high level of economy influence the obtained model results.These findings could be evidenced by Pasirkratonkramat Village, which has the most significant area.Pasirkratonkramat Village has high road density, a high economic level, and easy access to healthcare facilities.

Coastal Flood Risk Assessment
The flood risk model was developed using the Rpkl21 formula, which calculates the three main risk components: Hpkl, Vpkl21, dan Cpkl..It was found that low-risk levels dominate the risk index in Coastal of Pekalongan.This low-risk level dominates Coastal of Pekalongan, especially in the southern coastal areas.The western coastal areas have a relatively moderate risk level, dominated by plantation (terraced fields) and cropland.The hazardous level has a similar distribution to the vulnerability analysis previously conducted, which was scattered in areas dominated by settlements.Based on the developed vulnerability model, the population density overlay with settlements shows a high population density (Figure 15).The quantification of flood risk in coastal of Pekalongan indicates that the moderately risky to high-risk classes follow the flood hazard from the GEE model.This assessment indicates that the obtained realtime rob inundation data can identify areas at moderate risk.
However, when considering the capacity components in Coastal of Pekalongan, it is observed that even in areas with high-risk levels, the community can still withstand the impact.This model aligns with the previously developed economic capacity level, where the eastern part of Pekalongan City has a high Gross Regional Domestic Product (GRDP).The high GRDP in an area indicates the community's ability to sustain itself amidst various challenges, including tidal floods.Based on the calculations, it can be observed that Coastal Pekalongan was predominantly characterized by a very low-risk level, covering an area of 3,231 hectares (Table 8).This classification is justified by the dominant land use class of water bodies, accounting for 39% of the total 4 classes.The highest level of very high risk, amounting to 6.75 hectares, is found in Pekalongan Utara District.The high population density is the primary factor influencing the elevated risk in the Pekalongan Coast, particularly in urban areas.
The model results have successfully assessed the risk of utilizing cloud computing platform data.The hazard model, utilizing SAR data, effectively extracted inundation information in real-time.This study confirms the effectiveness of the model, where SAR-based flood mapping represents the current form of risk assessment [14].The vulnerability model developed in this study was a crucial consideration in the risk assessment process.This result aligns with the findings some researcher which emphasizing the significance of road networks and the number of healthcare facilities as rescue points in reducing risk [10,27].The land-use-based vulnerability model also played a vital role in risk assessment.The Landsat-8 imagery used in this study represents the current land use.Risk assessment can be derived from land-use classifications of affected areas [9,33], thus corroborating the findings of this research .The capacity component considers data derived from GEE, namely Nighttime Imagery.This imagery has successfully mapped the distribution of community economic capabilities based on their Gross Domestic Product (GDP).This research correlates with the findings which calculated GDP based on nighttime imagery using a proportional method [16,22].In contrast to the study which treated nighttime imagery as vulnerability [15], this research considers nighttime imagery as capacity.This result introduces a gap in risk assessment.However, it can also be an advantage as risk assessment can be dynamic, meaning it can utilize various data as long as they represent each component involved.
The differences observed in the results of this study compared to previous research provide a novel contribution to the field of disaster management, particularly in disaster mitigation planning.Previous studies only considered two components: hazard and vulnerability [14,[41][42][43].This study attempted to incorporate the capacity component in flood risk calculations, and the obtained results demonstrate its relevance to the existing conditions.For example, considering the previously established economic indicators, it was found that the eastern part of Pekalongan City has a high Gross Regional Domestic Product (GDP).The higher GDP in a region indicates the community's resilience in coping with various challenges, including flood risks (Figure 16).The findings of this research support the use of cloud computing optimization for disaster risk assessment, aligning with previous studies.Disasters are dynamic and holistic [35]; therefore, using upto-date and comprehensive data and considering multiple components is crucial for accurate risk assessment.Flood risk mapping is a non-structural disaster mitigation approach that can inform policymaking.This research also addresses the limitation of previous study that highlighted regarding the need to simultaneously implement structural and non-structural disaster mitigation measures [37].

Conclusion
This research successfully modeled the coastal flood risk using Cloud Computing.The findings indicate that the potential inundation area almost dominates Coastal of Pekalongan, covering approximately 2,870 hectares.The inundation model was developed using a modified code provided by UN-SPIDER.VV polarization was considered more effective in identifying coastal floods in areas with low slopes, specifically the coastal region.
The constructed models serve as the basis for risk assessment in the coastal area.For instance, the capacity model was computed using night-time satellite imagery and the proportion method.This approach can calculate a more detailed Gross Regional Domestic Product (GDP).In Pekalongan Coast, the GDP is higher in urban areas.The high night-time lights signify increased economic activity, aligning with the model's results (higher night-time lights in urban areas along the coast).
The risk assessment was based on the formula R = HV*(1-C), where a more significant capacity leads to lower risk.The model results show that the risk follows vulnerability, as the highest-risk areas were in urban areas.There is a correlation between high risk and capacity, as these areas have numerous affected buildings, some of which may be damaged and uninhabitable.Community resilience and financial issues play a crucial role in measuring risk, as indicated by the model of economic capacity.
Optimizing data from cloud computing proves to be beneficial for disaster mapping.The accessibility of cloud computing is a significant advantage, and the research results demonstrate high accuracy by field conditions.Eight of the 123 validation points were deemed invalid due to their location in high-elevation areas.
The research findings have several limitations and strengths.The model's results still require accuracy testing for each component, particularly those that involve predictive modeling.This limitation can be addressed by validating the model using the kappa index based on previous years' data.Although a wealth of open-access data is available in the Google Earth Engine (GEE), researchers still need to cross-check the data resolution.Ensuring consistency in data resolution for each component will facilitate the overall process.While data can be accessed openly, it is recommended for researchers to validate the model directly in the field to ensure its alignment with existing conditions.

Figure 2 .
Figure 2. Condition of Infrastructure and Buildings Affected by Flood (Primary Survey, 2023)

N=
Number of samples of water bodies μwater = The average of the pixel values of a body of water σwater = Standard deviation of the pixel value of a body of water

Figure 6 .
Figure 6.Map of Physical Vulnerability to Coastal Flood (Analysis, 2023)

Figure 7 .
Figure 7. Map of Environment Vulnerability to Coastal Flood (Analysis, 2023)

Figure 8 .
Figure 8. Distribution of Land Use Area on Coastal of Pekalongan (Analysis, 2023)

Figure 9 .
Figure 9. Map of Social Vulnerability to Coastal Flood (Analysis, 2023)

Figure 11 .
Figure 11.Map of Physical Capacity to Coastal Flood (Analysis, 2023)

19 Figure 16 .
Figure 16.Coastal Residents' Buildings that Survive and are not Against Flooding (Survey, 2023) This research requires Sentinel-1 VV polarization imagery, Rescue Point data, Road Network data, Landsat-8 imagery, World Population data, and VIIRS Night-time Day/Night Band Composites Version 1 (Table

Table 1 .
Data Research

Table 2 .
Vulnerability Model Aspect Weighting

Table 4 .
Percentage and Area of Flood Hazard

Table 5 .
Percentage and Area of Flood Vulnerability

Table 6 .
Rate GRDP (IDR) and Area of Economic Capacity Figure 13.Map of Economic Capacity to Coastal Flood (Analysis, 2023)

Table 7 .
Percentage and Area of Capacity from Coastal Flood

Table 8 .
Percentage and Area of Flood Risk