Figure 3.1 Time series of Lake Powell elevation variations based on observations collected by multiple radar altimeters.
(Source: USDA, G‐REALM, Time series of Lake Powell elevation, U.S. Department of Agriculture)
Satellite radar altimeters have been used for monitoring the elevation variations of large lakes and reservoirs since the 1990s (Birkett, 1994). The underlying principle used for radar altimetry is to infer the distance between the nadir pointing satellite and the water surface by measuring the travel time of a radar signal emitted and then reflected back to the sensor. This technology has been applied primarily to the remote sensing of ocean topography (Fu & Smith, 1996). Its usage in monitoring inland surface water levels, however, has been increasingly recognized as a practical option (Crétaux et al., 2011). To date, a suite of satellite altimeters has collected elevation data of several hundred lakes and reservoirs. Primarily in the frequencies from 5 GHz to 37 GHz, the repeat cycles of these sensors range from 10 days to 35 days. Despite the low spatial resolution, the narrow swath, and the large footprint size associated with radar altimeters, they have made great contributions in quantifying large inland water bodies globally (Gao et al., 2012). The U. S. Department of Agriculture’s Global Reservoirs/Lakes (G‐REALM; https://ipad.fas.usda.gov/cropexplorer/global_reservoir) and French Space Agency Centre National d’Etudes Spatiales’ (CNES) Hydroweb (http://hydroweb.theia‐land.fr/) databases have served as the representative data portals for historical and near‐real‐time observations. Without counting their overlaps, G‐REALM and Hydroweb have reported elevations of 379 and 268 large lakes and reservoirs, respectively.
Nonetheless, there are several limitations with regard to developing a generalized drought index directly from surface water elevations observed by radar altimeters. First, there is a dearth of data continuity due to the limited lifespan and limited spatial coverage of the sensors. Using Lake Powell as an example, even though it is the second largest reservoir in the United States (in terms of storage capacity), there was no altimeter overpassing it from 2002 to 2008 (Figure 3.1). Second, both the data quality and repeat period vary significantly among sensors. Third, although the reservoir elevation is uniquely related to the storage, the elevation–storage relationship varies drastically among reservoirs due to bathymetric differences. For instance, Gao et al. (2012) showed that the slope coefficients of the elevation–area relationships for the 34 global reservoirs range from 0.15 (Toktogul) to 210.28 (Aydarkul). This suggests that for an elevation increment of 1 m, the area of Lake Toktogul will increase 0.15 km2, while the area of Lake Aydarkul will increase 210.28 km2. Furthermore, the differences between the storage change values will be even much larger. As a result, it is difficult to compare drought severity among different reservoirs by comparing their elevation anomalies.
Reservoir/lake elevations also can be measured by the Geoscience Laser Altimeter System (GLAS) onboard the Ice Cloud and Land Elevation Satellite (ICESat). With a footprint of 70 m, the data collected by ICESat/GLAS has a much higher spatial resolution than those measured by radar altimeters. As such, ICESat/GLAS can observe more lakes/reservoirs of smaller sizes (Phan et al., 2012; Zhang et al., 2011), which offers a unique advantage over radar altimeters. However, ICESat has a long repeat period (91 days), and the satellite took measurements only from January 2003 to August 2010. The short lifetime and low temporal resolution have limited the applications of using ICESat/GLAS to monitor reservoir elevations operationally at near real time. Similar to the radar altimeters, ICESat/GLAS has a large spacing between tracks, resulting in sparse spatial coverage.
3.2.2. Reservoir Storage
Storage is the reservoir variable that is the most direct indicator of drought severity. For instance, the Texas Water Development Board (TWDB) uses the Reservoir Storage Index (RSI) as a reference to support water management decisions. The RSI is defined as the percent of storage capacity that the reservoir conservation pool is filled to at a given time (TWDB, 2017). In Melo et al. (2016), the percentage of total reservoir storage (relative to the system’s maximum capacity) was used to quantify the drought severity in the Parana River Basin. The results were further compared with the SPI to identify the linkage between meteorological and hydrological drought. In Figure 3.2, the time series of observed total reservoir storage over the Brazos River Basin (Texas) are compared with the precipitation data and the SPI. While the onset of hydrological drought is usually slightly lagged behind that of meteorological drought (Van Loon, 2015), the recovery of the former can be much slower than that of the latter when conditions are extreme (e.g., the 2011 record drought). On one hand, a reservoir‐storage‐based drought index can offer valuable information to help mitigate socioeconomic drought at both local and regional scales. On the other hand, such efforts have not been pursued fully, largely because the observed reservoir storage data are typically not shared.
With remotely sensed elevation and/or surface area data, reservoir storage can be estimated using equation 3.1:
where V, h, and A represent storage, surface elevation, and area, while the subscript c stands for capacity. In some studies, elevations collected by radar altimeters and water area from image classifications were combined in order to calculate the storage. For instance, Birkett (2000) was the first to combine altimetry data obtained from the TOPEX/Poseidon satellite with area images from the Advanced Very High Resolution Radiometer (AVHRR) for estimating Lake Chad storage variations. More recently, Duan & Bastiaanssen (2013) calculated the water volume variations of Lake Mead, Lake Tana, and Lake Ijssel from Landsat imagery data and elevations collected from four operational satellite altimetry databases: G‐REALM, River Lake Hydrology (RLH), Hydroweb, and ICESat‐GLAS level 2 data. The drawback of this approach is that satellite elevation data and area imagery need to be available simultaneously during the study period. Despite their high accuracy, results based on ICESat are so sparse that they are suitable only for trend analysis and not for monitoring purposes. To overcome this constraint, alternative approaches were developed by first estimating the elevation–area relationships such that the storage can be calculated from elevation or area directly using equation (3.1) (Gao, 2015). Crétaux et al. (2011) showcased the capability of monitoring reservoir storage variations at a global scale by applying radar altimetry data to prior developed elevation–area relationships. Gao et al. (2012) leveraged the availability of both radar altimetry data and Moderate Resolution Imaging Spectroradiometer (MODIS) area data to maximize the temporal coverage. Zhang et al. (2014) established elevation–area relationships based on ICESat elevations and enhanced MODIS water area estimates, which allowed for monitoring relatively small reservoirs that do not have overpassing radar altimeter tracks. By combining the Joint Research Centre (JRC) Global Surface Water (GSW) data set and the Database for Hydrological Time Series of Inland Waters (DAHITI), time series of storage variations for 135 global lakes and reservoirs between 1984 and 2015 were generated by Busker et al. (2018). These studies serve as