Reservoirs play a fundamental role in the hydrological planning of the central valley of Chile as they provide water for human and animal consumption, energy generation, and crop irrigation, especially during the summer season. In agriculture, reservoirs represent a significant source to keep the food security standard for more than half of the population of the country. The water management plans need complete records of their volume to calculate rules of operation or future scenarios; however, currently, these time series include gaps that do not allow better analysis, which increases uncertainty. To address this, the authors test a methodology to assess Sentinel 2 imagery through normalized difference water index (NDWI). The results correctly represent the temporality and seasonality of reservoir dynamics; however, the magnitude of the changes is not well represented when the reservoir is delivering water. This research allows more data-based planning of water resources in the central zone, contributing to better decision-making and more efficient water management.
TopIntroduction
Reservoirs play a fundamental role in watershed management since they allow water to be delivered during dry periods, both for human consumption and for agricultural use, increasing water security, allowing the generation of hydroelectric power, and the possibility of mitigating the effects of floods. (Deng et al., 2020; Scola et al., 2014). They also alter the dynamics of river sedimentation, the concentration of certain pollutants, and can alter the regime of the basins (Kondolf et al., 2014). In Chile, there are three categories of reservoirs according to their volume and the height of their wall. Small reservoirs (category A) with a wall between 5 and 15 m and a capacity between 5 x 104 m3 and 1.5 x 106 m3, medium-sized reservoirs (category B) with a wall between 15 and 30 m and a capacity between 1.5 x 106 m3 and 6.0 x107 m3 and large reservoirs (category C) with a wall greater than 30m, and a capacity equal to or greater than 6.0 x107 m3 (Ministry of Public Works - Chile, 2013).
The 26 large reservoirs currently built have a combined volume of 12,961 million cubic meters, including reservoirs of 22 million cubic meters such as Conchi or Tutuvén, to larger ones such as Colbún (1,544 mm3), and Laja (5,582 mm3).) (General Water Directorate, 2021a). These have played a fundamental role in preventing major consequences of the drought in Chile (Garreaud et al., 2017), maintaining food security, and irrigation security in the main valleys of the country (Cordova, 2017; DGA, 2013). The records of these reservoirs include both the outgoing flow from the wall, and the registered volume based on the elevation, and allow knowing the resources delivered and their timing; however, these time series are not complete due to multiple problems: (a) measurement instrument failures, (b) problems associated with floods or other risks, (c) measurement planning or administration problems (Tencaliec et al. al., 2015).
An alternative to reconstructing these time series is through satellite images (Klein et a., 2017; Yao et al., 2019). This technique has several advantages, given its low cost, its global coverage, its availability since the 1980s, its independence from other hydrological stations, and the multiple associated validation methodologies (Arvor et al., 2016).
The use of satellite images in this field embodies several challenges. In the first place, the need for a complete database of reservoirs that includes basic data such as coordinates or geographic information, date of construction, and maximum volume. Globally, there is the GOODD project with more than 38,000 reservoirs (Mulligan et al., 2020), GRAND that includes more than 7,320 with an area greater than 0.1 km3 (Lehner et al., 2011), Global WaterPack (Klein et al. ., 2017) or the AQUASTAT database (FAO, 2021). In Chile, the official database is developed by different offices of the Ministry of Public Works, publishing monthly the volume of the main bodies of water (Direccion General de Aguas, 2021a).
Secondly, the selection of images, since there are daily information sources such as MODIS but with a large cell size of 250 to 500m (Friedl et al., 2015) has been used as support in the generation of time series of reservoir volume (Deng et al., 2020; Zhang et al., 2014), while others with bi-weekly review with better spatial resolution such as Landsat (30m) or Sentinel (10 to 60m). These have been used individually by Van Den Hoek et al., (2019) or in combination with Schwatke et al. (2019) or Militino et al., (2020). Additionally, it should be noted that not all images can be used due to sensor problems such as Landsat 7 (Hossain et al., 2015; Jabar et al., 2014) or cloud cover (Foga et al., 2017). ; Zhao & Gao, 2018)