For verification of this paper, please visit on http://www.socialresearchfoundation.com/researchtimes.php#8

Lakes are an important part of ecological system and important source of freshwater. Fresh water comprises of less than 2.5% of the total water reserve of the earth and thus it becomes important to keep a strong vigil on the day to day changes happening in these sensitive features which are home to several flora and fauna and basis of life for many communities around the globe. Advances in remote sensing and GIS have now made it possible to constantly monitor the changes occurring in the various types of surface features such as glaciers, lakes, urban area, forests etc. Thanks to the advances in remote sensing, it is now possible to evaluate the changes that have occurred in various features with the help of multi-temporal and multi-spectral satellite image data which is freely available for the general public and researchers both. Water surface change detection is one such area of application of remote sensing and GIS where satellite images aided with various tools can be used to detect the changes that have occurred in the spatial-temporal aspects of a particular water body. Usage of NDWI has been a common technique among geographers and remote sesning professionals for the water body extraction and change detection in the areal extent of the water body.This study too made use of NDWI for the study of Lake poopo , which surprisingly has not been studied in depth and no major work has been done to analyze the change in the lake shoreline even after being one of the largest lakes in South America. Thus this study is unique in terms of its area of study i.e Lake poopo, Bolivia.

There have been many studies involving the use of remote sensing and GIS for change detection of lake area. [12] Has used multiple indices such as NDWI, mNDWI, and SVI (Supervised classification) for the shoreline extraction and analysis of change of Lake Burdur, Turkey with the help of LANDSAT TM and ETM+ data. In a similar study [1,11] made use of NDWI, MNDWI, WRI and Principal component analysis for the change detection of Lake Urmia. In another study of Lake Chad, [6] made use of Supervised classification technique for the shoreline extraction and calculation of change in area of the lake. One thing common in all the studies mentioned above is, all the authors have recognised the superiority of NDWI for the accurate shoreline extraction and detection of the change in lake area over a period of time. According to [6] and [11], the results showed greater accuracy when using NDWI as compared to other methods such as NDMI and MNDWI. The results were more accurate in cases where there were no urban built up areas in the vicinity of the lake. As Lake Poopo had no built up areas in its proximity, thus NDWI was selected for the shoreline extraction and change detection in the area of the lake.

 Study Area and Sources of Data

Lake Poopó is a saline lake, situated in the Altiplano Mountains at an altitude of 3,700 meters in Bolivia. Lake Poopó used to have an approximate permanent water body area of nearly 3000 sq. km. But increased interference has changed the lakes morphology and water content which is clearly depicted in this study. The lake’s main tributary is the Desaguadero River, from which it used to get 92% of the water, the river flows in from Lake Titicaca, situated at the northern end of the Altiplano. Lake Poopó is a shallow lake with an average depth of 3 meters, making it more prone to evaporation and variations in surface area [2]. The lake is highly prone to drought which was the main factor behind the seasonal drying up of lake in the past, but increased human activities in the form of mining and diversion of water for irrigation purposes have changed the lake dynamics and affected it negatively. This has resulted in the semi- permanent disappearance of the lake and decreased water level as compared to past during rainy seasons.

Location of Lake Poopo

Fig.1. Source- NAIP hybrid imagery (ArcGIS Pro)

The study makes use of satellite imagery obtained from USGS Earth explorer from 6 August 1991, 23 August 2003, and 29 August 2011 for Landsat 4-5 TM (Path 233 and Row 73) and 29 August 2020 for Landsat 7 ETM+ (Path 233 and Row 73). To ensure clarity in the satellite image, cloud cover was kept less than 10%.. Scan line error of Landsat 7 were also corrected using Landsat toolbox of ArcGIS Pro for higher accuracy.

Methodology

Image Acquisition

For the purpose of analysis multi temporal Landsat 7 TM and Landsat 4/5 ETM+ imagery in Band 2 (Green) and Band 4 (NIR) were acquired through USGS Earth Explorer.

Table 1. Details of the Landsat Images acquired through USGS Earth Explorer

Fig.2. Images acquired from Landsat 4/5 and Landsat 7 showing subsequent decrease in area.

Usage of Water Indices

Water indices are an efficient way to extract water pixels [4]. There are many water indices which have been used in the past few years. [3] Made use of NDVI (Normalized difference vegetation index) to detect water surface area. [7] Made use of the NDWI for the water extraction. Mcfeeters NDWI is regarded as the first generation of water indices [4] and is the most widely used index.

NDWI is seldom confused with NDMI as both make use of moisture index but there is a sufficient difference between the two in terms of their method of calculation and level of accuracy. The NDMI makes use of NIR-SWIR bands to detect moisture contents in the chlorophyll of the leaves, whereas NDWI uses Green and NIR band for the monitoring of water content in various water bodies such as lake, enclosed seas. In the NDWI -1 to 0 represents no vegetation or no water content, whereas +1 represents water content.

 NDWI ( Normalized Difference Water index)

The use NDWI was suggested by Mcfeeters to delineate surface water features and enhance their presence in satellite. The NDWI works with the reflected NIR radiation and visible green light to boost the appearance of water features and removes the pixel values associated with non-water body features. The use of NDWI is not limited to the extraction of water body and has been conveniently used for estimation of water turbidity [9].

NDWI = (Green band – NIR band) / (Green band + NIR band) [7]

NDWI Applications

Whenever we require to enhance a water body from its surroundings and perform an analysis on the body, the NDWI is used as it water reflects more in Green band than NIR band, the difference between them is always positive. NDWI uses this property of spectral reflectance of water and separates non water pixels from water pixels which makes it easier to differentiate water bodies from non-water bodies [9].

NDWI Advantages

NDWI method has following advantages over manual digitization of water extent:

a)     It is an objective approach

b)     It gives consistent reproducible results

c)     It has faster processing since all water pixels in an entire image scene can be mapped in one processing  

Water Body Extraction Using Spectral Water Indexes

NDWI is primarily used for the differentiation of water and non-water body areas. It uses the spectral reflectance property of water and extracts the water pixels from the raster dataset. As water reflects more in visible green band, NDWI uses this property to enhance the water bodies by separating non water pixels from water pixels and illuminating the water bodies for better visibility which helps greatly while performing change detection analysis .Generally there is threshold value of ‘greater than 0’ is assigned to the water pixels which makes it easier to separate water pixels from non-water pixels. 

                                                                    

Fig3. ^[5] Spectral profile of water

As it is evident in the spectral profile of water given above indicated by Blue line, water reflects great in green band but reflects little in NIR band. This difference between the reflectance in Green band and NIR band pixel value can be used to map water pixels.

By dividing the difference between green and near infrared band with their sum, we scale NDWI value from -1 to +1

Fig. 4. NDWI layers for year 1991, 2003, 2011 and 2020

Identification of bad pixels using ‘Condition’ tool

Condition tools allow map the desired water pixels from the raster layer. It performs a conditional if/else evaluation on each of input cells of input rasters.

In our case all the values greater than 0 were regarded as water pixels and values below 0 were excluded. All he pixels that are not water is mapped as No data. No data is a concept in GIS to identify pixels we do not require.

                                                                          Fig 5a. CON layer for year 1991 and 2003

Extraction and Merging of Lake Shorelines         

To extract lake shorelines as polygon feature class, the raster images were converted to polygon using ‘Raster to Polygon’ tool. These layers were then merged using Merge tool which combines multiple input data sets into single new output dataset. All input classes must have same geometry type. Example- several point feature classes can be merged but a line feature cannot be merged with a polygon feature class [8, 10].

Fig.7a. Merged shorelines for visualization of change in Lake Shoreline

Fig.7b. Bar chart depicting year wise change in water area of Lake Poopo

Upon the analysis of the data derived through the processing of images taken over a span of 29 years, it is observed that area of Lake Poopo has decreased by 92.48 percent since 1991. Largest lake area was observed in the year 1991 which was 2410 square kilometers .Since then the lake has been continuously declining. From the Table 4 it is clear that the intensity of lake decline has been continuously increasing since 1991. The largest decline was observed between 2011 and 2010, where in just a span of 10 years the lake declined by 80.17%.

 Table 2.

Causes of decline of the lake

Following reasons are regarded as the reasons for reduction in area of Lake Poopo-

1. Melting of Andes glacier, which result in decreased water level river entering the lake and thus less water intake causes reduction in water level
2. Change caused by frequent droughts.
3. Diversion of water from River Desaguadero for irrigation and mining purposes.
4. Change in water level of Lake Titicaca, which causes reduced volume of water in river and thus the loss of water due to evaporation cannot be compensated.
5. Heavy mining activities in the region surrounding Lake Poopo cause a change in the dynamics of lake.

1. Acharya, T. D., Yang, I. T., Subedi, A., & Lee, D. H. (2016). Change Detection of Lakes in Pokhara, Nepal Using Landsat Data. Proceedings of the 3rd International Electronic Conference on Sensors and Applications, 15–30 November 2016; Available Online: Https://Sciforum.Net/Conference/Ecsa-3. https://doi.org/10.3390/ecsa-3-e005 2. Blair, Laurence (4 January 2018). "The ecological catastrophe that turned a vast Bolivian lake into a salt desert". The Guardian. ISSN 0261-3077. 3. Domenikiotis, C., Loukas, A., & Dalezios, N. R. (2003). The use of NOAA/AVHRR satellite data for monitoring and assessment of forestfires and flood. Natural Hazards and Earth System Sciences, 3, 115–128 4. Huang, C., Chen, Y., Zhang, S., & Wu, J. (2018). Detecting, Extracting, and Monitoring Surface Water From Space Using Optical Sensors: A Review. Reviews of Geophysics, 56(2), 333–360. https://doi.org/10.1029/2018rg000598 5. Lenhardt, J. (2021, January 5). Spectral Profiles: Improve Classification Before You Click Run. ArcGIS Blog. https://www.esri.com/arcgis-blog/products/arcgis-pro/imagery/spectral-profiles-classification/ 6. Mahamat, A. A. A., Al-Hurban, A., & Saied, N. (2021). Change Detection of Lake Chad Water Surface Area Using Remote Sensing and Satellite Imagery. Journal of Geographic Information System, 13(05), 561–577. https://doi.org/10.4236/jgis.2021.135031 7. McFeeters, S. K. (1996). The use of the normalized difference water index (NDWI) in the delineation of open water features. International Journal of Remote Sensing, 17(7), 1425–1432 8. Merge (Data Management)—ArcGIS Pro | Documentation. (n.d.). Pro.Arcgis.Com. https://pro.arcgis.com/en/pro-app/latest/tool-reference/data-management/merge.htm 9. Normalized Difference Water Index: NDWI Formula And Calculations. (2022, February 7). EARTH OBSERVING SYSTEM. https://eos.com/make-an-analysis/ndwi/ 10. Raster to Polygon (Conversion)—ArcGIS Pro | Documentation. (n.d.). Pro.Arcgis.Com. https://pro.arcgis.com/en/pro-app/latest/tool-reference/conversion/raster-to-polygon.htm 11. Rokni, K., Ahmad, A., Selamat, A., & Hazini, S. (2014). Water Feature Extraction and Change Detection Using Multitemporal Landsat Imagery. Remote Sensing, 6(5), 4173–4189. https://doi.org/10.3390/rs6054173 12. Sarp, G., & Ozcelik, M. (2017b). Water body extraction and change detection using time series: A case study of Lake Burdur, Turkey. Journal of Taibah University for Science, 11(3), 381–391. https://doi.org/10.1016/j.jtusci.2016.04.005