Long-term Change of Water Transparency in Lake Singkarak Observed from Remote Sensing Data

The water quality data for supporting lake management in Indonesia are very limited due to financial constraints. Satellite data has a big potential as a source for providing data to retrieve water quality information. In this study, we used Landsat TM/ETM+ remote sensing data and applied an empirical model for estimating the water transparency (Zsd) to represent water quality in Lake Singkarak. We used 230 scenes of pre-processed Landsat TM/ETM+ images to generate a long-term SD database for Lake Singkarak from 1987 to 2020. The visual and statistical analysis shows the change of Zsd in Lake Singkarak. In the period before 2000, the Zsd were generally lower than 2 meters (m). From 2000 to 2005 the Zsd continuously increased from 1.5m to 4m. Lastly, from 2005 until 2020 Zsd were steady: 2m > Zsd < 4m. The satellite-based Zsd estimations captured the three regimes of water quality in Lake Singkarak. These estimations will thus provide useful data for lake managers and policymakers.


Introduction
Water quality data are crucial for supporting lake management.Conducting a regular assessment of water quality measurement on a lake remains a challenge due to labor, time, and budgetary constraints.As a result, there are extremely few water quality records for the majority of Indonesian lakes.
Remote sensing data has been employed as a supplementary data source in the extraction of data related to the water quality of aquatic environments (e.g., [1][2][3][4][5][6]).Therefore, it can offer possibilities to create a database on water quality for lakes and reservoirs in Indonesia.Considering the extensive historical data records of Landsat 5-TM and 7-ETM+, these remote sensing datasets were chosen to create a water quality database over the long term.[7][8][9].An optimum combination of a 30 m spatial resolution and 16 days temporal resolution was considered sufficiently effective in capturing changes in water quality within Lake Singkarak.
One of the important water quality parameters is the water transparency (Zsd) measured with a Secchi disk.Zsd (unit: meter) is a measurement value of water clarity for assessing the water quality conditions.[10].Zsd is noticed as the most frequently utilized and straightforward parameter in limnological measurements due to its easily interpretable values [11].Xu et al [12] documented the 1313 (2024) 012016 IOP Publishing doi:10.1088/1755-1315/1313/1/012016 2 utilization of Landsat TM was for obtaining extended water clarity data, and findings suggested that anthropogenic activities might contribute to the prolonged decline in water clarity trends.
Lake Singkarak is a large and deep lake in the west part of Sumatera Island.Lake Singkarak has various water uses including fisheries, hydro-power, irrigation, domestic, and tourism [13].Lake Singkarak is one of 15 national priority lakes; with sedimentation, water quality deterioration, and depletion of an endemic fish catch (ikan bilih) problems [13].Numerous investigations regarding water quality have been conducted in Lake Singkarak [14:17].However, a long-term analysis of water quality change has not been done yet.Accordingly, the main objectives of this study are to (1) generate longterm water clarity for Lake Singkarak from historical Landsat TM and ETM+ images from 1987 to 2020 and (2) detect temporal water quality patterns.

Study area
Lake Singkarak is the second-largest lake on Sumatera Island after Lake Toba.It is located in the Province of West Sumatra.Geographically located between 100.48°E to 100.60° E and -0.53° S to 0.7° S; at 360m above sea level (Figure 1).The water surface is about 108 km 2 ; with the catchment area is about 1,034 km 2 .The ratio of the lake's watershed area and the water surface is 9.57.Lake Singkarak is a deep lake with a maximum depth of 268m, with a water retention time of 20.4 years (Lake volume/outflow rate).

Pre-processing of Satellite Data
A sum of 769 Landsat TM/ETM+ satellite images were acquired from the USGS (United States Geological Survey) website [18].However, due to the image quality and the cloud coverage, only 203 scenes remained useful for further processing.To calculate Zsd from satellite data, a set of 'Preprocessing data' is required to standardize long-term time series data [19].The 'Pre-processing data' includes the extraction of water pixels [20:21], filtering [22], Digital Number (DN) to radiance conversion [23], and performing atmospheric correction [24].

Statistical Analysis
An R package called "ggplot2" [26] was utilized to visually display the average of Zsd calculated from each Landsat image.A "Locally wEighted Scatterplot Smoothing" (LOESS) statistic method was implemented to derive the trends of long-term Zsd.LOESS applies the Savitzky-Golay filter to draw a trend line from scattered points by local polynomial regression and has found extensive application in the analysis of time-series data [27][28][29].A simple linear regression, R 2, and p-value were used to describe the fluctuation of Zsd and the significant values.

Field Observation data
In situ water transparency (Zsd) data was gathered from twelve field works, with 3~10 sites for each survey.The locations were recorded using a GPS receiver.A white and black 20-cm-diameter standard Secchi disk was used to measure the water clarity values.The average Zsd values are given in Table 1.
Table 1 2 is the data visualization of water transparency in Lake Singkarak.The Black point is the average Zsd calculated from one satellite image (estimated); the outlier was removed by excluding the 2.5 % lowest and 2.5% highest data.The red point is the average Zsd measured in the field (observed).The Black curvy line is the Locally wEighted Scatterplot Smoothing" (LOESS).This line is helpful to connect the Time series data and detect the pattern.The Black straight line is the simple linear trend using all-year data.The Blue straight line is the simple linear trend using data from 2000 to 2020.The Red straight line is the simple linear trend using five yearly data (i.e., up to 1990, 1990-1995, 1995-2000, 2000-2005, 2005-2010, 2010-2015, 2015-2020).The number of total satellite data is 769 images, however, after the non-water pixel was removed, and images with less than 50% pixels were excluded, only 203 images were used for further analysis.An increased Zsd was observed using all data (from 1987 to 2020), R

The performance of the selected Zsd estimation in capturing the fluctuation of water transparency in Lake Singkarak
Even though Lake Singkarak is categorized as one of the Indonesian priority lakes [13], the availability of measured water quality data was sparse and patchy.In this study, our in situ Zsd data were collected from our field survey, previous research, and publication.However, we only have twelve in situ Zsd data sets, as can be seen in Table 1.This condition increases our motivation to utilize the remote sensing data archived to extract and build a water quality database with the Zsd parameter to trace back the lake condition from the very first time Landsat data was available until recent days.
Comparing the estimated with the corresponding in situ Zsd, we found one Zsd value is underestimated, four Zsd values are over-estimated, and four Zsd values were relatively close to in situ Zsd.Nevertheless, it is important to emphasize that the trend of long-term changes in Zsd derived from historical Landsat images is more reliable compared to a single Zsd estimation [19] The Zsd model showed robustness in estimating Zsd using hundreds of images in Lake Singkarak.Our research successfully processed all available images (769 images), and finally generated 203 dates Zsd data from Landsat-TM/eTM+.The minimum, maximum, average, and St.d (standard deviation values) of estimated Zsd are 0.4m, 6.02m, 2.80m, and 1.31m respectively.This extensive Zsd estimated range shows that our Zsd estimation model works well for turbid, medium up to clear water environment conditions.Thus, our satellite derived Zsd estimations effectively captured the fluctuations in water quality conditions in Lake Singkarak (please see Figures 2 and 3).

Water transparency shifting in Lake Singkarak
The visual and statistical analysis shows the change and tendency of Zsd in Lake Singkarak.The satellite-based Zsd estimations captured the three regimes of water quality in Lake Singkarak.In the period before 2000, the Zsd are generally lower than two meters (m).From 2000 to 2005, the Zsd continuously increased from 1.5m to 4m.Lastly, from 2001 until 2020, Zsd was steady, 2m > Zsd < 4m.
Geographically, Lake Singkarak is situated near Lake Maninjau.In the early 1990s, fish aquaculture was introduced in Lake Singkarak, mirroring practices in Lake Maninjau.However, the profitability of Lake Singkarak did not match that of Lake Maninjau.Risdawati (2011) [30] indicated that a fish parasite (Cirolana.sp) likely contributed to the low fish productivity in Lake Singkarak.Consequently, fish cage culture is not widely adopted in the lake.Accordingly, we assumed that there is no relationship between fish farming activities and the water quality change in Lake Singkarak.

Conclusion
We successfully generated long-term water transparency (Zsd) using a data archive of Landsat TM and ETM+ images spanning from 1987 to 2020.During the study period, the Zsd in Lake Singkarak has three regimes of water quality in Lake Singkarak.In the periods before 2000, the Zsd are generally lower than two meters (m).From 2000 to 2005, the Zsd continuously increased from 1.5 m to 4 m.Lastly, from 2005 until 2020, Zsd were steady, 2 m > Zsd < 4 m.Remote sensing data archived has demonstrated as valuable sources to address the data gap, trace-back and detection temporal water quality changes.This water quality change information will be valuable for lake managers and policymakers.A Further study to identify the driving factors that cause the water transparency shifting in Lake Singkarak should be performed.An in-depth exploration of these driving forces will contribute to an enhanced understanding of the lake's ecosystem dynamics, facilitating more targeted and effective conservation and management efforts.Such investigations will enhance our ability to implement informed and sustainable measures to preserve the ecological aspect of Lake Singkarak and contribute to broader environmental conservation initiatives.

Figure 2 .
Figure 2. The water transparency fluctuation in Lake Singkarak from 1987 to 2020.Blackpoint: average Zsd estimated from Landsat.Redpoint: observed Zsd.Black line: the trend analysis using LOESS.Blue line: the all-year simple linear trend.Red line: specific five yearly simple linear trends.