Suspended Particulate Matter Analysis of Pre and During Covid Lockdown Using Google Earth Engine Cloud Computing: A Case Study of Ukai Reservoir

Abstract

Presence of suspended particulate matter (SPM) in a waterbody or a river can be caused by multiple parameters such as other pollutants by the discharge of poorly maintained sewage, siltation, sedimentation, flood and even bacteria. In this study, remote sensing techniques were used to understand the effects of pandemic-induced lockdown on the SPM concentration in the lower Tapi reservoir or Ukai reservoir. The estimation was done using Landsat-8 OLI (Operational Land Imager) having radiometric resolution (12-bit) and a spatial resolution of 30 m. The Google Earth Engine (GEE) cloud computing platform was used in this study to generate the products. The GEE is a semi-automated workflow system using a robust approach designed for scientific analysis and visualization of geospatial datasets. An algorithm was deployed, and a time-series (2013–2020) analysis was done for the study area. It was found that the average mean value of SPM in Tapi River during 2020 is lowest than the last seven years at the same time.

Introduction

Rivers are the dominant source of fresh water, even though water covers 70% of the earth’s surface. The river waters are subjected to a high rate of pollution owing to the disposal of domestic sewage and industrial effluents and so on. Furthermore, climate change, increasing population and other anthropogenic factors have escalated the freshwater crisis. Water Resources Management is the immediate need to optimize the available natural water flows. Artificial reservoirs are constructed for various purposes like irrigation, power generation, flood control, etc. Rivers carry pollutants and sediments and eventually deposit them in reservoirs. Sedimentation is an essential factor that threatens the sustainability and longevity of reservoirs. It reduces the storage capacity of the reservoir. Besides, water quality is an important parameter that needs to be evaluated from time to time for sustainability for aquatic organisms and human beings. Various parameters determine the water quality, such as DO, BOD, COD, pH, turbidity, ammonia, etc. DO or dissolved oxygen is a very important measure of water quality as it impacts the phytoplankton health, algal growth sustainabilty (Gorde and Jadhav 2013). Biochemical oxygen demand or BOD is the measure of the amount of biodegradable organic material present in the water. COD or chemical oxygen demand is the amount of oxygen required to oxidise all the substances (even biologically decomposed) in is the water (Chandra, 2012) and a very useful tool for contamination monitoring. pH is a very well known parameter for water quality. Any water sample below or above pH 7 is a indication of water contamination.

Turbidity measures the amount of total suspended organic and inorganic materials in water (Giardino et al., 2017). It is an optical property and has a relationship with the reflectance in water bodies (Sravanthi et al., 2013). Several studies (Garg et al., 2017, Doxaran et al., 2002) have reported an increase in reflectance in the visible region (red region) with an increase in turbidity. Apart from anthropogenic processes, several natural processes also like floods and erosion, influence the amount of suspended sediments in the water column. The suspended sediments reduce the penetration of sunlight which lowers the primary productivity. Toxic metals and organic matter can get associated with suspended sediments causing eutrophication and water pollution (Point et al., 2007; Stoichev et al., 2004). Thus, turbidity can be used as an indicator of water quality. Suspended particulate matter (SPM) is one of the key water quality parameters for assessing the pollution of a waterbody. SPM causes a range of aquatic discomforts and plethora of environmental damage (Dersch, 1986) including problems such as benthic concealing, irritation of fish gills and hindrance of light for photosynthesis etc. (Davies-Colley, 2002). The presence of SPM in a waterbody or a river can be caused by multiple reasons such as other pollutants by the discharge of poorly maintained sewage, siltation, sedimentation, flood and even bacteria.

Before the advancement of remote sensing, early literatures used secchi disks to measure the water transparency. It was very time and cost-consuming as well as the extension of coverage was very restricted. Remote sensing methods made it easier for the researchers to overcome all these hurdles by establishing the relationship between water-leaving reflectance and turbidity (SPM) (Curran et al., 1987). It has been well-established that with increasing turbidity, reflectance is also increasing. Doxaran et al. (Doxaran et al., 2002) showed that the reflectance from 400 to 1000 nm increases with turbidity. Wei et al. (Wei et al., 2021)the inferred in SPM study that the ascending order of the reflectance is proportional with the increasing turbidity. This kind of empirical algorithms are used in multiple successful studies using linear, log-linear and exponential relationship between the satellite images and reflectance (Yunus et al., 2020; Nechad et al., 2010; Doxaran et al., 2002; Tassan, 1994). A wide range of satellite sensor data has been used to estimate aquatic turbidity or Total Suspended Matter (TSM) globally such as SeaWiFS,Resources ats, VIIRS, Landsat, Resourcesat-2, and recently launched Sentinel − 3 satellites (Wei et al., 2021; Chander et al., 2018; Doxaran et al., 2002; Tassan, 1994). This study has used the Landsat 8 data to analyse and assess the SPM algorithm.

The new decadeof 2020 saw the outbreak of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), or COVID-19, although it was initially reported in December 2019 in the Wuhan province of China. The first case in India was reported on 30th January 2020. It started spreading at a rapid rate, being highly transmissible. In March 2020, the World Health Organisation (WHO) declared COVID-19 a pandemic (Gupta et al., 2021; Sood et al., 2020b). In order to combat the contagion of the deadly virus, unprecedented measures were taken (Sood et al., 2020a). A nationwide 21-day lockdown was initially imposed from 25th March 2020, which got extended until the end of May, and subsequently unlock started from June. All industrial, transportation, and anthropogenic activities were restricted to some extent. For the past few decades, anthropogenic activities have been one of the primary sources of pollution in various compartments of the Environment (Barwal et al., 2020; Gao et al., 2018). A positive effect of the lockdown on the Environment in terms of improvement in air quality (Gautam et al., 2020; Lokhandwala and Gautam, 2020, Chauhan and Singh, 2020; Collivignarelli et al., 2020; Dantas et al., 2020; ) and water quality (Braga et al., 2020, Garg et al., 2020, Yunus et al., 2020, Lotliker et al., 2020, Mishra et al., 2020, Chakraborty et al., 2021) was reported. According to Sun et al. 2021, high-frequency remote sensing assessment of Wuhan lake, China has shown a steep 16% decline in mean turbidity. Tripathi et al., 2021 have shown that near Patna, Bihar where the discharge of industries and sewage is highest has a witnessed rapid decrease of sedimant concentration in the lockdown period. An extensive study (Kumar et al., 2021) has been done on the Ganga river with Sentinel-2 satellite data. Multiple points over the entire course of the river for Chl-a, CDOM, and TSM is taken. The study revealed that the Covid-19 lockdown has a very vivid effect on the water quality and hence quite assuredly inoftion over the different water bodies may have impactful results overall.

During the lockdown period, field data could not be collected, so remote sensing techniques were used to determine the changes in the Environment. The parameters which represent water quality are chromophoric dissolved organic matter (CDOM), suspended particulate matter (SPM), and Chlorophyll-a (Chl-a). In this study, remote sensing techniques were used to understand the effects of pandemic-induced lockdown on the suspended particulate matter (SPM) concentration in the lower Tapi reservoir or Ukai reservoir. The estimation was done using Landsat-8 OLI (Operational Land Imager) having radiometric resolution (12-bit) and a spatial resolution of 30 m. The Google Earth Engine (GEE) cloud computing platform was used in this study to generate the products. The GEE (cloud computing platform) is a semi-automated workflow system using a robust approach designed for scientific analysis and visualization of geospatial datasets. There is a massive data catalog made up of Earth-observing remote sensing and imagery many other environmental, geophysical, and socio-economic datasets (Gorelicks et al., 2017). An algorithm was deployed, and a time-series (2013–2020) analysis was done for the study area. GEE has proven to be an important tool for policymakers and researchers through plethora of standpoints. Enormous data storage with versatile data types (social, demographic, DEM, climate data) (Mutanga 2019). The applications of the GEE cloud computing platform cover the gargantuan spectrum of scientific and sociological arrays such as vegetation mapping (Poortinga et al. 2018), landcover mapping (Lee et al., 2018), agricultural applications (Aguilar et al., 2018), earth and meteorological observation (Sproles et al., 2018), disaster management (Lobo et al. 2018). In summary, the possibility of handling, monitoring, mapping huge data, and automating programs for operational level becomes easier with GEE and this has turned out to be a huge upgrade for combatting environmental problem. The list of studies have been launched globally to compare the status of air, ,water, soil and other sociological, economical and industrial standpoint before and after Covid-19 (Shami et al. 2021). Very few, among the aforesaid literatures rarely targeted the sediment problems and status of any river choked by the pollutants in the previous years (Sun et al. 2021) in India (Kumar et al., 2021; Tripathi 2021) that too with a different approach such as cloud platform workframe.

In this study, the main focus is finding the intensity of sediment deterioration of an extremely polluted river of India during the shut down phase of the nearby industrial plants during the 1st phase of Covid-19 lockdown. Our another aim is to establish the GEE as an impactful tool for the assessment of pre and during covid situation for the environmental managers, scholars and researchers for there respective studies.

Study Area and Data

Location Description: Ukai Reservoir

Tapi is a major river in Western India, and Ukai is the largest reservoir in Gujarat. It was constructed in 1972 for the purpose of irrigation, flood protection, power generation, and fisheries development. Ujjania et al., 2015, reported the higher suspended particulate matter as an indicator of pollution in the Ukai reservoir. The major source of suspended particulate matter is organic material from anthropogenic activities like fishing, domestic sewage, agricultural practices. The high amount of suspended load is most likely to cause flooding. During the lockdown period, as there was a reduction in anthropogenic activities, a reduction in suspended particulate matter is hypothesized.

The study area (Fig. 1) is Ukai reservoir in Southern Gujarat, India. The reservoir is situated on the Tapi River and is around 90 km from the Surat city in Gujarat. It is located between longitudes 73°32’25”-78°36’30"E and latitudes 20°5’0”-22°52’30"N and has a catchment area of 62,225 km². The Tapi river originates from Multai (Madhya Pradesh), then flows through Maharashtra and then into Gujarat and joins the Arabian Sea in the Gulf of Cambay. It can be divided into three zones viz. Upper Tapi basin, Middle Tapi Basin, and Lower Tapi Basin (LTB). The portion between Ukai Dam to the Arabian Sea is considered as LTB and it is estimated to be 122 km.

Fig. 1.

Study area location map

Data and Tools Used

The data used here is Landsat 8 OLI images with 30 m resolution and they are exported from the GEE platform after applying median composite and SPM algorithm over the images. Descriptions are given in the Table 1. Total 77 Landsat 8 (OLI) images were acquired (Path/Row: 147/45, 147/46, 148/45) (Fig. 2). All the images are level 2, collection 2 images from the GEE catalogue, meaning the images are radiometrically calibrated and atmospherically corrected surface reflectance derived data.

Table 1.

Dataset used for the study

Dataset	Description	Properties	Source
USGS earth explorer Landsat 8	Path/Row: 147/45, 147/46, 148/45	30 m	Google Earth Engine (GEE)catalogue
Random GPS points	12 points (6 each on northern and southern sides)	Lat and Long	Google Earth

Fig. 2.

Number of images of 2013 to 2020 divided into 3 row and path

The tool mainly used is GEE for fundamental image processing for the SPM concentration mapping and for map-making purposes ArcMap is used. Descriptions of the tools are given in the Table 2. The success of this study lies in the approach of GEE for retrieving the SPM images which is a sustainable method for the researchers and policy-makers without using any software or downloading unnecessary large data.

Table 2.

Tools used for the study

S.No.	Software	Version	Description
1	GEE	-	Cloud-based platform for remote sensing application enabled with JavaScript API
2	ArcMap	10.1	Used for Map generation

Methodology

Retrieving the Median Images of Each Year

The purpose of current study is to develop an automated framework using Google Earth Engine to evaluate the time series analysis of SPM in the Ukai dam, India from 2013 to 2020, specifically on the month of April 1st to June 7th of each year (lockdown period). Total 77 Landsat 8 (OLI) images were acquired (Path/Row: 147/45, 147/46, 148/45) (Fig: 1). All the images are level 2, collection 2 images from the GEE catalogue, meaning the images are radiometrically calibrated and atmospherically corrected surface reflectance derived data. The single median image of each year has been taken and SPM algorithm equation is applied on all of them. None of the images of the 3 path and rows were fully covering the reservoir so median composite approach is taken. It is to be noted that median is taken instead of mean to reduce the impact of outliers.

The first approach is to filter out the Landsat-8 data according to geometry, cloud cover year and months. Utilized functions are filterBounds()- to filter out the images lying over the study area. filterMetadata(name, operator, value)- to filter out the images that has more than 5% of cloud cover. filter(ee.Filter.calendarRange(startyear, endyear, ‘year’)- to fetch the data from 2013 to 2020. filter(ee.Filter.calendarRange(startmonth, endmonth, ‘month’)- to pick the data only for the desired month. These commads over the Landsat – 8 data brought total 77 images of 2013 to 2020 of the months April to June specifically. Clip() is used to extract the desired reservoir from the whole imageries. Select (band name)- we only need the red band so this function we automatically take the red band of the Landsat-8.

The next part creates a loop that will stack the the images according to the year as a median composite. Function that are used here, map(funtion{})- to operate desired function over the maps. Median()- processing the images a single median images without outliers and artifacts. This will return the 8 median images from the collection. After that SPM algorithm is applied in the median images which is explained in the Sect. 3.3.

Detection of Red Band Reflection Change

It has been accounted for in the writing and demonstrated that, because of the adjustment of suspended sediments of the water, the variation in the spectral reflectance in visible region of the spectrum are critical (Brezonik et al., 2005; Liedeke et al., 1995; Ritchie et al., 1976). Writing proposes that even a single band, whenever picked appropriately, can give a robust estimate of suspended sediments (Gholizadeh et al., 2016; Nechad et al., 2010; Pavelsky and Smith, 2009). It was proposed that a single red band can be utilized to gauge the suspended sediments in water (Shi and Wang, 2009; Hellweger et al., 2007; Miller and McKee, 2004). The single band concept was used where the reflectance increases with the increase in suspended sediment concentrations. In the present study red band is utilized for analyzing the change in spectral response due to varying suspended sediment concentration across the reservoir. The reflectance on each water pixel of each annual median images (2013–2020) was classified from high to low. It was viewed as the sediment concentration increases, the reflectance in the red band also increases and vice versa. In this manner, the pixels with high reflectance in red band are viewed as high suspended sediment concentration, while low reflectance as low suspended sediment concentration.

For measuring the reflectance values of red band over the northern and southern portion of the reservoir, ESRI’s Arc Map tool, Extract Multi Points Values to Points, is used. With the help of this tool the cell values that coincide spatially to a specified point feature class from one or more raster can be easily extracted and the value can be recorded as an attribute in the point feature class. In Table 3, the Global Positioning Syatem (GPS) points are listed, which were used to extract the reflectance values across the northern and southern regions of the reservoir.

Table 3.

GPS Points over the northern and southern part of the Waterbody

S.No.	Northern Region		Southern Region
	Longitude	Latitude	Longitude	Latitude
1	73.841	21.421	73.641	21.235
2	73.862	21.437	73.627	21.258
3	73.82	21.415	73.697	21.217
4	73.889	21.445	73.697	21.253
5	73.889	21.461	73.703	21.201
6	73.824	21.388	73.673	21.23

SPM Algorithm

Previous study of Nandita et al., 2015 showed that the in-situ measurement of average value of SPM in Ukai dam ranges between 70 and 300 mg/L around all the seasons in 2014 (Nandita et al., 2015). Nechad et al., 2010 suggested a single band empirical model for the assessment of SPM in any waterbody. With the support of previous study it is safe to assume that the applying Nechad et al., 2010 empirical method will retrieve the SPM concentration data efficiently (RMS value < 10 mg/L).

The following equation of SPM concentration evaluation from the water-leaving reflectance value of the red-band is.

SPM =

ρ is the water-leaving reflectance of red band of Landsat 8 satellite, Aρ and Cρ are the wavelength-dependent empirical co-efficients of the equation; Aρ = 289.29, Cρ = 0.1686. Parameters Cρ was standardised using standard inherent optical properties (IOPs). Aρ is calibrated using least-square regression model of in-situ data of turbidity and reflectance. After retrieving the median images of each year the above-mentioned SPM is applied to the median images of the corresponding year. This process salvaged the single band empirical model of retrieving surface water SPM for this paper.

In GEE, the whole process of applying the SPM equation starts with assigning the constatns to variables to simplify the visualisation. ee.Image.constant()- to assign variables of each component (co-efficients) of the equation. Image.expression()- assign the SPM equation or expression to each the median image.

Last thing was to downloading the images using Expression.image.toDrive() and save them in the drive. The entire methodlogy is explained in Fig. 3.

Fig. 3.

Schematic diagram of the methodology used

Results

Yearly Trend of SPM in Ukai Dam

In this section the results and maps of the paper is discussed. Figure 4 shows the mean SPM value.

Fig. 4.

Yearly graph of mean SPM in Ukai reservoir

of each year, 2020 has the lowest value of mean SPM. Image of 2013 and 2018 shows huge amount of SPM concentration. On contrary, 2014, 2016 and 2019 shows moderate account of SPM. The SPM concentration of 2020 is 7% less than the average of SPM concentration of last.

seven years (2013–2019) which very significant for a river with this much pollution. The mean rainfall data from the CHIRPS (Climate Hazards Group InfraRed Precipitation With Station Data) daily of the same exact period over the study area reveals that there no such linear relationship between rainfall and the turbidity. The Fig. 5 shows that in rainfall (mm) is pretty inconsistent with results availed from the SPM algorithm but it also requires further studies with the greater number of data points to find significant relationship between rainfall and turbidity.

Fig. 5.

Mean rainfall from 1st April to 7th June of each year

The average SPM of the water body is generally between 66 and 70 mg/L in the month of March-June. SPM of the year 2018 shows the highest mean concentration of the SPM (70 mg/L). Figure 6 states that the changes in overall mean percentage of SPM in the reservoir than the previous year. The concentration increased from 2013 to 2015 and dropped in the year 2016. It again started to increase till 2018 and steeply dropped in 2020 mainly because of lockdown period all over India. All this graphs and maps suggest the decrease in SPM in 2020. Negative values shows the decrease in percentage of SPM concentration.

Fig. 6.

Deviation of SPM concentration of each year than the previous one

Reflectance Value of Median Images Each

It has been observed that there was a significant variation in the reflectance of the red band. Higher the turbidity, higher will be red band reflectance and vice-versa (Tripathi 2021; Garg et al., 2020). Since the red band have generally less obstruction from the base and return backscattered energy from suspended particles for the most part. The reflectance over the northern portion of the reservoir was high than that of over the southern portion of the reservoir (Fig. 7a h). This is because of the high sediment concentration over the northern part that southern part. In 2020 (Fig. 7 h), it can be observed that change in reflectance of the red band across the reservoir is quite low than any other image (2013–2019) i.e. Figure 7a g, it might be possible because of the Covid-19 lockdown which results in shut down of industry and various anthropogenic activities resulting in lesser concentration of suspended sediments in the reservoir.

Fig. 7.

Change in the reflectance values (especially of 5 h) all over the reservoir suggests that the change of red band reflectance is correlated with variation of SPM: (a) 2013, (b) 2014, (c) 2015, (d) 2016, (e) 2017, (f) 2018, (g) 2019, and (h) 2020

To verify that several points over north and south part of the reservoir were taken spatially and then reflectance value of each point were extracted. After that mean percentage reflectance were obtained for each image for the year 2013 to 2020. The percentage reflectance graph in the Fig. 8 showed that the change in reflectance of the red band in northern portion is significantly higher than that of southern portion of the reservoir. Thus, the concentration of suspended sediments is always high in the northern part because of the contractive shape of the catchment reservoir. In the year 2020 (Fig. 7 h), the percentage reflectance is significantly low in comparison to any other year which tells that the concentration of suspended sediment is less because of the lockdown effect across the country.

Fig. 8.

Variation in the surface reflectance over thereservoir

SPM Concentration Maps

Figure 9 Suspended particulate matter (SPM) concentrations estimated for the Ukai Reservoir: (a) 2013, (b) 2014, (c) 2015, (d) 2016, (e) 2017, (f) 2018, (g) 2019, and (h) 2020. Last image of 2020 shows clear reduction of SPM.

Fig. 9.

Suspended particulate matter (SPM) concentrations estimated for the Ukai Reservoir: (a) 2013, (b) 2014, (c) 2015, (d) 2016, (e) 2017, (f) 2018, (g) 2019, and (h) 2020. Last image of 2020 shows clear reduction of SPM.

Figure 9 (a-h) shows the images of SPM concentration (mg/L) over Ukai reservoir from 2013 to 2020. It is visually very prominent that the SPM on year 2020 has reduced drastically compared to the previous years. The Northern part of the reservoir has shown greater decrease of SPM as the northern inlet of the river carries water from the industrial area and due to complete closure of the factories on the banks of Tapi river, the water carried less pollutants which effectively lowered the SPM on the surface water. The SPM concentration of the southern part of the reservoir remained more or less unchanged. The explanation can be the width of the reservoir. The waterbody is distributed over a larger area so the SPM maintained uniformity and change is not vividly noticeable but the overall change in SPM is observed. The similar trend has been observed in Yunus et al., 2020. It took Vembanad lake of Kerala for the assessment of the SPM concentration using Landsat 8 images and found the same pattern of the variation in SPM all over the lake. The lake witnessed decreased SPM during the Covid-19 lockdown period compared to other years. Aswathy et al., 2021 applied same method over Astamudi Lake and saw almost 40% reduction of SPM compared to the average SPM of last 5 years. These studies establishes the findings of this paper as a concrete evidence of SPM variation with the lockdown period.

Conclusion

This paper has dealt with the effect of Covid-19 lockdown in Ukai reservoir of Tapi river using remotely sensed data and cloud-based platform Google Earth Engine. This study is effective in such a way that it did not need to download any of the 77 Landsat 8 images and it also did not required any software for any kind of image processing, editing or manipulating. A few lines of code created a flow of semi-automated data processing which delivered desirable result.

The research work has considered the one of the most polluted rivers of Gujarat to evaluate the influence of the global pandemic situation on SPM concentration. The study followed the algorithm of Nechad et al., 2010 but in an automated workflow which has provided satisfactorily good results. The satellite-based data of Tapi river have shown that the average mean value of SPM in 2020 is the lowest at the same period of time than the last seven years. This is mainly happened because of the closure of the industrial plants in the banks of the river and tourism industry. These results also suggest that the primary reason for the high SPM in the Tapi river are above mentioned pollution sources. The satellite data are very useful for change detection or trend analysis let it be water quality analysis or Land Use/ Land Cover classes. Here, this paper has used the Landsat 8 OLI data for the water quality investigation with the images taken by the satellite for the last 8 years (2013-2020).

Further study is required to be done using finer resolution (Sentinel- 2). There are many other pollutants effecting the water quality of the reservoir and algorithms and empirical models available (Goddijn-Murphy & Williamson, 2019; Choe et al., 2008) which can be helpful for understanding the detrimental situation of water quality of the study area. Overall, the study served the purpose of detecting the changes in SPM of the study area with the help of Google Earth Engine platform.