Water Turbidity Mapping Using Sentinel-2A Imagery and Cloud Based Google Earth Engine in Saguling Reservoir

Turbidity represents the level of suspended sediments in water, that will contribute to a greater effect on the sedimentation process such as siltation in river and reservoir. Knowing the distribution of turbidity is expected to provide an overview of which parts of the reservoir area need special attention to reduce the sedimentation rate of the reservoir. The study location is the Saguling Reservoir. All stages in this study are processed in the Google Earth Engine, a cloud-based computing that only produce one output of water with turbidity index values only. The visualization shows that the normalized difference turbidity index (NDTI) varies with a minimum value index of -0.187228 and a maximum value index of 0.09871. The results of this study are sufficient to describe where the source of turbidity has the potential to become sediment in the reservoir which continues to settle and can gradually accelerate the lifetime of the dam. The map produced can provide an overview for stakeholders who have a task in managing water resources. The higher of turbidity show the worse condition of the catchment area or watershed area upstream. The conditions can be a consideration in planning engineering that might be done so as to reduce sedimentation that occurs in the reservoir.


Introduction
Sedimentation problem is a very important matter to urgently handled in a reservoir [1].Sedimentation can reduce the storage capacity of a reservoirs and affect conditions water quality.Sedimentation has relationship with turbidity level and water depth [2].Turbid water can be identified as cloudy water contain sedimentary material or materials pollutant, so that if the water becomes more turbid, then the water will contribute greater effect on the sedimentation process.Dealing with sedimentation problems is very important.Identification and monitoring process will be done effective and efficient if supported by technology [3].The process of sedimentation of waters in the reservoir is influenced by the level of turbidity [4].
Turbidity represents the level of suspended sediments in water also indicating water clarity or how clear is the water.It is mainly caused by the presence of silt, algae in a water body, or industrial waste disposed in the rivers by mining activity, industrial operations, logging, etc.According to the ASTM-International definition, turbidity is an expression of the optical properties of a liquid that causes light rays to be scattered and absorbed rather than transmitted in straight lines through a sample.
Traditionally, turbidity is analyzed by evaluating water samples taken during field measurements.Mapping of water quality of inland waters using remote sensing is being carried out since 1970s, with the launch of Landsat series of satellites.However, field studies are expensive, time and labor intensive, besides, during lockdown field surveys cannot be undertaken.Satellite remote sensing data is a good IOP Publishing doi:10.1088/1755-1315/1343/1/012027 2 alternative to field survey measurements which can capture both spatial and temporal variations in river turbidity levels [5][6][7] [8].Every feature on the surface of the earth behaves differently on interacting with electromagnetic radiation [9].Identification of the results of the sedimentation process on a water object can be done by utilizing the spectral response of the water object to certain wavelengths [10].Remote sensing is a technology that can be used for measurement and monitoring of sedimentation processes in areas that have wide coverage [11].
Sentinel-2 carries the Multispectral Imager (MSI).This sensor delivers 13 spectral bands ranging from 10 to 60-meter pixel size as shown in Table 1.Band combination is used to better understand the features in imagery.By using band combinations, the specific information can be extract from an imagery.There are band combinations that highlight geologic, agricultural or vegetation features in an image.The natural color band usually named visible band combination uses the red (B4), green (B3) and blue (B2) channels.The purpose is to display imagery the same way our eyes see the world.The bathymetric band combination is good for coastal studies, it uses the red (B4), green (B3) and coastal band (B1).As for Normalized Difference Turbidity Index (NDTI), the band combination is green (B3), red (B4) and VNIR (B8).
The Sentinel-2 instrument is made of 12 spectral bands with a 10 m resolution of visible bands (VI), 20 m resolution of vegetation red edge (VRE) bands, and short-wave infrared (SWIR) bands, in addition to three bands related to coastal aerosols and water vapor of 60 m resolution.Three different remotely sensed indices were obtained to represent three different water quality parameters, maximum chlorophyll index (MCI), green normalized difference vegetation index (GNDVI), and normalized difference turbidity index (NDTI) [12].
The purpose of this research is to find out how the condition of the turbidity of the reservoir has the potential to become sediment in the reservoir.Knowing the distribution of turbidity is expected to provide an overview of which parts of the reservoir area need special attention to reduce the sedimentation rate so as to extend the life span of the reservoir.The study location is the Saguling Reservoir.This reservoir is a reservoir that has the most upstream position in a series of Citarum cascade reservoirs that dam the Citarum River.The Saguling Reservoir also receives water supply from quite a number of large rivers from various directions.The large inundation area of the reservoir makes the IOP Publishing doi:10.1088/1755-1315/1343/1/0120273 method of utilizing satellite imagery and the cloud based google earth engine one of the easiest alternatives in identifying turbidity.

Study Area
The study area, Reservoir of the Saguling Dam, is an embankment dam on the headwater of Citarum River in West Java, Indonesia.It is located about 40 km from Bandung city, or about 12 km to the south of the main road from Bandung to Jakarta, and about 90 km at the upstream of the Jatiluhur Lake as shown in Figure 1.The lake was built on the Citarum River, the largest river in West Java Province, in February 1985.
The primary purpose of the dam is hydroelectric power generation but it also provides for water supply and aquaculture when passing through a series of the Citarum cascade with the Cirata and Djuanda (Jatiluhur) reservoirs which are downstream.The dam is rock-fill embankment-type with watertight core that withholds a reservoir with a capacity of 982.000.000m 3 at maximum elevation 643 m.Jatiluhur reservoir is located at an altitude of 111 m above sea level with a surface area of 53 km 2 and maximum depth of 92 m.The source of water comes from Citarum river with basin area of 660,000ha with length of 270km which is flowing to Saguling, Cirata and Jatiluhur reservoirs.

Google Earth Engine
The Google Earth Engine (GEE) is a cloud computing platform designed to store and process huge data sets (at petabyte-scale) for analysis and ultimate decision making.
The data catalog houses a large repository of publicly available geospatial datasets, including observations from a variety of satellite and aerial imaging systems in both optical and non-optical wavelengths, environmental variables, weather and climate forecasts and hindcasts, land cover, topographic and socio-economic datasets [7].All of this data is preprocessed to a ready-to-use but information-preserving form that allows efficient access and removes many barriers associated with data management.
Since its inception in 2010, GEE usage was investigated using articles drawn from a total of 158 journals.The study showed a skewed usage towards developed countries as compared to developing regions such as Africa, with Landsat being the most widely used data set.It categorized the papers into five main themes, such as vegetation mapping and monitoring, landcover mapping, agricultural applications, and disaster management and earth sciences.It has demonstrated the power of GEE platform in handling huge data sets at various scales and building automated programs that can be used at an operational level.[13]

Calculation of Normalize Difference Wetness Index (NDWI)
Study the water quality of the river, in terms of change in turbidity, the optical remote sensing data from Sentinel-2 data were analyzed for change in reflectance and hence change in turbidity.The water pixels were estimated using the normalized difference water index given by McFeeters as provided in Equation (1).The NDWI makes use of reflected near-infrared radiation and visible green light to enhance the presence of such features while eliminating the presence of soil and terrestrial vegetation features.
It is suggested that the NDWI may also provide researchers with turbidity estimations of water bodies using remotely-sensed digital data [14] Where Rg and Rnir are reflectance in green and near infrared (NIR) bands.These wavelengths were selected to maximize the typical reflectance of water features by using green light wavelength, to minimize the low reflectance of NIR by water features and to take advantage of the high reflectance of NIR by terrestrial vegetation and soil features.
NDWI used to differentiate water from dry land or more suitable for mapping water bodies.The body of water has low radiation and strong absorption in the visible infrared wavelength range.NDWI uses the near infrared and green bands of remote sensing imagery based on their occurrence.It can improve water information efficiently in most of the cases.It is smooth on built-up ground and often ends up in overly high bodies of water.

Calculation of Normalize Difference Turbidity Index (NDTI)
Where Rr is reflectance in the red band.The reflectance of pure water is more in green than the red wavelength region.However, the water pixels were identified as mentioned above, the Equation (2) was applied using these two bands to map NDTI.The higher value of turbidity yields a high value of NDTI and vice versa.
The use of green and red band reflectance to estimate variable suspended sediment concentration in the stratified water column.Therefore, later the results were validated through mapping of NDTI, which is a normalized ratio of red and green band reflectance [15] It was noticed that red and NIR bands are more sensitive towards turbidity estimation [3].These bands are very useful to quantitatively estimate the turbidity in the absence of field observed data for optically deep water.Based on the finding, it can be deduced that in the absence of ground observed data, remote sensing approach can be used for preliminary estimates on water quality.
Elhag [12] was successfully utilizing remote sensing data acquired from Sentinel-2 to estimate chlorophyll, nitrogen content and water turbidity.The mean values of raster data showed a high correlation with the actual data from the conducted laboratory examination.

Data Processing
Data processing to investigate reservoir turbidity begins with determining the coordinates of the polygon area study.The coordinates of the study area are as in Table 2.A computational function is derived from the polygon to facilitate clipping to the extent shapefile.Subsequently, the acquired Sentinel-2A image is imported.A consolidated image is then generated from this collection using the mosaic function, incorporating bands 3 (Green), 4 (Red), and 8 (VNIR).The Normalized Difference Water Index (NDWI) is computed, resulting in a binary raster.The Normalized Difference Turbidity Index (NDTI) is calculated, and the outcome is constrained by the water mask.Lastly, the map layer is incorporated into the analysis.This research was conducted by using a stack of several indices.This index stack not only offers a quick overview of some landscape features but also shows the arithmetic basis for quantitative analysis and classification of images using index values in this case used both NDWI and NDTI.Given the variety of indices available, the potential for modifying and extending this type of analysis is virtually unlimited.

Result and Discussion
All stages in this study are processed in the GEE, a cloud-based computing so that they only produce one output.The only one output is an image that only displays bodies of water that have turbidity index values.In conventional satellite image processing, this data processing produces several images, including the initial image clip from the Sentinel-2 image which is the study area for processing, sometime its need more than one images for one study area related the satellite scene recorded, the second output image is the result of NDWI extraction which is the output of the separation process between water bodies and non-water bodies, the third image is a water body with a turbidity index value and no turbidity index value, the fourth image is the final output or a water body image that only has a turbidity index value.This is an advantage of cloud computing, especially GEE, where data processing is carried out very efficiently without burdening the memory of computer devices used for processing.As is the case in this case, for the entire study area which is the all area of the Saguling Reservoir, it only takes less than three minutes plus three minutes for the export process for image storage.The total time required for this processing is only six minutes.It is possible that if more hardware specifications and faster internet speed are used, the process from start to storage only takes less than five minutes or even three minutes.

Figure 3. Output of NDTI calculation results from GEE
The output resulted from image processing in GEE is an NDTI image with a pixel value between minus one to positive one.If displayed using standard software it only produces a black and white image.The output results cannot present the variation in the index value of the NDTI.Image can only be interpreted in a simple way.If the image is displayed on a white background (Figure 3), the display is slightly clearer with several black areas on the right, bottom and several scattered black dots which are areas of water bodies that have an NDTI value.This shows that the value of the NDTI result is only read as a binary pixel value in the standard image opener software.This visual makes the image output unable to give a visual of the turbidity conditions in the Saguling Reservoir water body area.In order to present visuals that can be analyzed visually, visualization is needed using special geographic information system (GIS) software, in the case of this study using the open source QGIS software as software to visualize the NDTI index value of the output.Based on the visualization results using the QGIS software, the range of index values can be presented in a gradual color display as shown in Figure 4.The visualization shows that the NDTI index varies with a minimum value index of -0.187228 and a maximum value index of 0.09871.The visualization shows the range of NDTI value in the water bodies of the Saguling Reservoir.The minimum value indicates a low turbidity while the more positive it indicates a higher turbidity.This visualization also shows how the results of the NDTI clip process can be seen from the results of the NDTI overlay layer which are presented in a gradual red color overlayed with the base map using Google satellite imagery in QGIS.Not all water bodies on Google satellite imagery are covered by a gradual red color which indicates the NDTI value.In addition, the output display before visualization does not clearly describe several scattered dots as well as the visualization results with QGIS.The output of the results of the process using GEE has been georeferenced so that when the overlay process is automatically in accordance with the coordinates of the base map used.
The NDTI map of the Saguling Reservoir shows that the distribution of turbidity is only around the water bodies at the inlet of the reservoir from the river, whereas in the water bodies that are further downstream near of the main dam there is no turbidity is detected.this can be explained logically that turbidity is carried by water discharge, the farther from the inlet and the deeper the water depth so that the turbidity is not detected.The distribution of the highest turbidity is shown in a gradual, darker red color and is located at the reservoir inlet of the main dammed river named Citarum River, which can be seen from the map on the lower right.Furthermore, turbidity with a medium level based on classification originates from the south which is the reservoir inlet of the Ciminyak and Cibitung Rivers.Followed by turbidity at the inlet of the reservoir from the Cijambu River.The contribution of turbidity from other rivers such as Cilalanang, Cijeruk and Cipatik is not visible in the analysis results which are only presented at the red dots around the water body.
The results of this study are sufficient to describe where the source of turbidity has the potential to become sediment in the reservoir which continues to settle and can gradually accelerate the lifetime of the dam.Dam lifetime is calculated based on how long a body of water in a reservoir can hold water.When the sediment is high, the water cannot be controlled by the dam so that the dam can be said to have failure because it can use for flood control, raw water sources, or for electricity generation.In addition, the map produced can provide an overview for stakeholders who have a task in managing water resources.The higher of turbidity show the worse condition of the catchment area or watershed area upstream.The conditions can be a consideration in planning engineering that might be done so as to reduce sedimentation that occurs in the reservoir.For areas with medium levels of turbidity, conservation can be carried out in the upstream areas of the catchment area, so that the level of erosivity is reduced which directly impacts on reduced material that the river carries to the reservoir.
Overall, the results of this study can be said that NDTI analysis utilizing satellite imagery and GEE is very helpful for detecting turbidity in large areas of water.The limitations of software capabilities and hardware limitations to process high-resolution satellite images such as Sentinel are not felt during processing.The results also show a logical level, although it would be better if field tests were carried out by taking water samples and checking turbidity levels in the laboratory.

Conclusion
The NDTI map of the Saguling Reservoir shows that the distribution of turbidity is only around the water bodies at the inlet of the reservoir from the river, whereas in the water bodies that are further downstream near of the main dam there is no turbidity is detected.The distribution of the highest turbidity is shown at the reservoir inlet of the main dammed river named Citarum River.Furthermore, turbidity with a medium level based on classification is the reservoir inlet of the Ciminyak and Cibitung Rivers.Followed by turbidity at the inlet of the reservoir from the Cijambu River.
Knowing the distribution of turbidity is expected to provide an overview of which parts of the reservoir area need special attention to reduce the sedimentation rate so as to extend the life span of the reservoir.NDTI analysis utilizing satellite imagery and GEE is very helpful for detecting turbidity in large areas of water.

Research Limitation and Future Study
This research is limited to the use of satellite imagery in this case specific used Sentinel 2a image to provide an overview of the turbidity index value in the Saguling reservoir.This research does not include accuracy tests in the field or accuracy tests by taking and turbidity values in the laboratory.Further research can be carried out by completing the accuracy test process and comparing it with data from the turbidity test results in the laboratory.In addition, further research can also be carried out by comparing time series to determine changes in turbidity index due to changes in land use and infrastructure upstream of water bodies.

Figure 1
Figure 1 Location of Saguling Reservoir3.MethodologyThe methodology employed in this study aims to unravel the dynamics of turbidity and sedimentation processes in the Saguling Reservoir, leveraging the robust capabilities of Google Earth Engine for comprehensive analysis.Turbidity, indicative of suspended sediment levels in water, plays a pivotal role

Table 2 .
Coordinate of Study Area