Study of Water Level Variation in the Reservoir at the Ilha Solteira Dam (Brazil) Based on Geodesic Remote Sensing Estudo da Mudança de Carga Hídrica do Reservatório de Ilha Solteira com Base em Sensoriamento Remoto Geodésico

The construction of water reservoirs for the production of electricity or human consumption is very common in Brazil due to its high availability of water. However, in the last few years, the southeastern region of the country has been undergoing a period of drought that has greatly affected electricity generation, especially in the northwestern region of the state of São Paulo, an area that comprises one of the largest hydroelectric complexes in the world (Ilha Solteira and Jupiá). Ilha Solteira currently has a continuous-monitoring Global Navigation Satellite System (GNSS) station. These data were used in many geoscience studies investigating the Earth’s dynamics. This study evaluates the impact of the annual variation in the water level in the reservoir at geodetic altitude provided by the GNSS. Satellite images, GNSS solutions from the Nevada Geodetic Laboratory (NGL), referred to as IGS08, and daily quota and volume estimates from the Reservoir Monitoring System managed by the National Water Agency (ANA) were used. The results indicated a reduction in the reservoir water level and volume, which represented a variation of approximately 3.5 cm at the GNSS altitude.


Introduction
Brazil is one of the countries in the world with a great potential for energy production by hydroelectric plants due to the high availability of water resources. Hydroelectric plants represent almost 71% of the electricity generated and consumed nowadays in the country. Currently, there are over 640 reservoirs monitored by the National Water Agency, including Itaipu, Furnas, and Ilha Solteira (SAR, 2015).
The Paraná River, the second largest in South America, contains the Jupiá complex, the sixth largest complex of hydroelectric plants in the world. The Hydroelectric Plant of Ilha Solteira, which belongs to this complex, is located between the municipalities of Ilha Solteira (State of São Paulo, SP) and Selvíria (State of Mato Grosso do Sul, MS). With a 5,605 m-long dam and a 1195-km 2 reservoir, it is the largest plant of the Energy Company of São Paulo (CESP), with an installed capacity of 3,444 MW (CESP, 2015).
The metropolitan region of São Paulo is currently sufering the greatest water crisis since 1930, when the measurement of the levels in its reservoirs and rivers for water supply began. The Cantareira system, responsible for supplying almost half of the population of Great São Paulo (8.8 million people), operated with negative storage levels at the beginning of January in 2016, i.e., the system kept operating with the reservoir's technical level (dead volume). The water crisis also afected energy generation in the southeastern region, especially in the State of São Paulo. For instance, in December 2013, the Hydroelectric Plant of Ilha Solteira produced 2,026 MW (megawatts), and in the same month in 2014, it produced 1,117 MW, representing a 55% reduction (CESP, 2015). The impact of this reduction is observed in the values of the reservoir quota, which reached levels of 318 meters, far from the level of 323 meters, considered ideal by CESP. This reduction is relected in the water volume stored in the reservoir that, depending on the annual variation, could cause crustal movements. Events of this nature are observed on a global scale as well as a local scale and are associated with the elastic crustal deformation process. Elósegui et al. (2003) considered that surface loads, locally induced by lakes and glaciers, could be a signiicant source for the crustal deformation and could promote substantial crustal displacements. Davis et al. (2004) argued that the water redistribution at the Earth's surface also has a role in the load or weight changes in the Earth's crust. In addition, because it is supported by the sublithospheric mantle below, the crust will respond by deforming elastically in annual time scales.
Although the use of a Global Navigation Satellite System (GNSS) is widely employed for navigation, topography, and engineering, its use is not as pronounced in studies of the Earth's dynamics. Studies have demonstrated the potential of GNSS for measuring the Earth's elastic response due to surface loading (Blewitt, 2003;Bevis, 2005;Tregoning, 2009;Han et al., 2016).
By analyzing a geodetic station in Antuco (Chile), Bevis et al. (2003) veriied non-constant vertical movements of approximately 50 mm caused by volume variations in a lake located 20 km away from the station. The same authors, by analyzing the geodesic data from the station at Manaus, identiied a vertical displacement cycle with variations ranging from 50 to 75 mm caused by the annual variation of the Amazon River. Borsa et al. (2014) studied the west coast of the United States, which, since 2013, has undergone a severe drought. With the use of precipitation data and GNSS stations, it was possible to observe the average elevation change of 5 mm reaching 15 mm in the mountains of California. Prol et al. (2015) observed a strong anticorrelation between the increase in the water level of the Aquidauana River and the reduction of the geodesic altitude. The results showed that the GNSS could assist in the monitoring of the Pantanal looding level.
This study presents the efects of the annual variation in the water level of a reservoir using the geodesic altitude provided by the GNSS. The selected area was Ilha Solteira, which, in addition to representing the largest hydroelectric plant in the State of São Paulo and the third largest in the coun-

Material and Methods
The Municipality of Ilha Solteira is located in the northwestern region of the State of São Paulo and is part of the Mesoregion of Araçatuba ( Figure  1). With coordinates 20º 25' 58" S of latitude and 51º 20' 33" W of longitude, it has an average altitude of approximately 335 meters. Geomorphologically, Ilha Solteira is part of the Western Plains of São Paulo. The reservoir of the Hydroelectric Plant of Ilha Solteira, with a water surface of 1,195 km 2 , crosses four state borders, those of São Paulo, Minas Gerais, Goiás, and Mato Grosso do Sul (CESP, 2015).
To perform the study, Landsat 5 sensor TM, and Landsat 8 sensor OLI satellite images were used. The images were used as the basis for classiication of the area of the water surface in the reservoir at Ilha Solteira, where it was possible to verify the variation that occurred in the last several years.
The Landsat 5 scenes were obtained free of charge from the website of the National Institute of Space Studies. In turn, the Landsat 8 scenes were obtained from the website for Earth Explorer. To cover the entire reservoir, three images were used, with the following orbits/point: 222-073, 222-074 and 223-074. The years 2009, 2013, and 2014 were studied, requiring a total of nine scenes. All were from the month of September to reduce or eliminate cloud cover.
To exemplify the changes that took place in the area of the reservoir, a multi temporal composition was generated with the mid-infrared band, which corresponds to band number seven in Landsat 5 and 8.
The digital image processing (DIP) step was performed in diferent Geographic Information Systems platforms. For the georeferentiation, the PCI Geomatica software (PCI, 2003) was used. The Landsat 5 scenes were corrected based on the Landsat 8 scenes, which have high positional acuity (Storey et al., 2014;Roy et al., 2014).
The water surface area classiication was performed in the Ecognition software (Deiniens, 2013) which works with the object-oriented classiication mode. Only two classes were deined: water and anthropic. After initial tests, the values that adjusted best to the segmentation and subsequent classiication were: Scale parameter 50, Shape 0.1, and Figure 1 Location of the Municipality of Ilha Solteira, located in the northwest region of the State of São Paulo, with the altimetry map in the background.  (Sherman et al., 2015), where the values for the reservoir's surface area were obtained. This software also generated the multi temporal composition, as well as the maps presented in the study.

Study of Water Level Variation in the Reservoir at the Ilha Solteira Dam (Brazil) Based on Geodesic Remote Sensing
The data from the GNSS station of Ilha Solteira were obtained from the Nevada Geodetic Laboratory (NGL) website, in a text ile format (*txt) referred to as IGS08, derived from the latest version of the International Terrestrial Reference Frame (ITRF), ITRF2008.
Data from the Reservoir Monitoring System managed by the National Water Agency were also used, which provides daily readings of parameters, such as quota and efective volume as percentages.
The available data covered September 2010 to June 2015. The use of these data required their conversion to volume in cubic meters and then liters, where 1 L = 1 kg. This procedure was necessary to quantify the minimum weight exerted by the mass of water in the reservoir. This study did not consider the inluence of bottom sediments and underground water, it has considered only the water line as an adjustment factor for the local topography.
Finally, data smoothing was performed. This required the itting of the quota data with the altitudes provided by the GNSS (Up) by using a mobile mean in Microsoft Excel software. In addition, correlation graphs were generated and presented as the results of this study. By analyzing the multi temporal composition image (Figure 2), the water surface retreat is evident.

Results and Discussion
It is possible to observe some islands (highlighted in red) that formed in the reservoir. For spectral targets, such as the water, the mid-infrared tends to absorb the electromagnetic radiation to a greater extent. In turn, for vegetation areas and exposed soil, the mid--infrared is mostly relected. The analysis of the daily readings of the Reservoir Monitoring System showed the changes in the classiication process of the reservoir area and the multi temporal composition. One of the analyzed parameters was the variation in volume. Initially, a correlation test was performed between the quota data and the volume converted to cubic meters, obtaining a correlation of 99 %. According to the oicial data of CESP, the total volume of the Ilha Solteira reservoir is 21,060 million m³. Figure 3 shows the annual variation in the volume of the reservoir. This occurs due to the loss of water by annual evaporation, which is approximately 10% of the storage volume (Hernandes, 2005).
In terms of volume in cubic meters, the levels remained similar between the years 2011 and 2013 because the Ilha Solteira reservoir has a system of low control that, depending on the time of the year, could be reduced to maintain its operational levels.
With the annual variation of the volume and quota automatic control, there is an up and down movement in the water level when the data of the  Figure  4 presents the graphs with the quota and Up data. In Figure 4A, the values do not present the mobile mean and have a correlation of 66%. After the curve itting process ( Figure 4B), the observed correlation was 88%, which, in addition to improving the visual aspect of the curves, facilitates the interpretation of the data.
It is evident from the graph that, when the reservoir quota levels are high, the water volume in the system is consequently higher. This causes the altitudes observed by the GNSS to have an inverse behavior (anticorrelation). The process of adding load to the surface causes the crust to be pushed down, while an upward movement is observed when there is a decrease in the load.
During the study years, there is an inversion in the curves of the graph between the end of 2013 and the beginning of 2014. Table 1 shows the annual mean (September 2010 to June 2015) of the water volume in the reservoir in cubic meters, of the volume in liters, and of the weight (gigatonnene) of the mass of water, in order to deine the minimum necessary weight to cause variations in the water levels of the Ilha Solteira reservoir observed by the GNSS.

Year
Volume ( crease reached approximately 3.5 cm. When the quota value is considered, the reduction reaches approximately 4 m. In comparison to similar studies performed on the west coast of the United States (Borsa et al., 2014) estimated a deicit of approximately 240 gigatonnes. However, according to Prol et al. (2015), the altitude variations caused by the decrease in the load did not occur instantaneously, and they associated this with an event called "Temporal Translation."

Conclusions
The results obtained in this study indicate the potential of the GNSS application, when combined with other data sources, to study reservoir level variation and hydrology.
Additionally, with the reduction in the water level and, consequently, the volume of the reservoir, especially between 2013 and 2015, the Up response was an increase of approximately 3.5 cm in the GNSS altitude, in response to the decrease in the reservoir level. This suggests that the local topography can undergo vertical variations within a year.
These variations are most likely caused by crustal elastic deformation. In some cases, the overload produced by the reservoir could alter the local stresses and cause induced seismic activity, as reported in other locations of Brazil, such as near the Hydroelectric Plant of Capivari Cachoeira, NE of Curitiba-PR. Special attention is drawn to the increase in the water pressure in the pores and fractures of the rock massif contributing to a decrease in the rock and soil resistance to shear strain. Therefore, the presence of GNSS stations near reservoirs of hydroelectric plants or large natural water storage systems, such as the Pantanal, would help in understanding the Earth's dynamic compression.