Figure 1. Map of reservoirs and their capacities. Global maps of reservoir locations, with the size of symbols corresponding to (log) capacity, and color corresponding to year of completion. Only larger reservoirs are plotted. (a) shows the GRanD database (Lehneret al., 2011); (b) shows the database of Zarfl et al. (2015).
In the GRanD database, standard filling rates suggest that we can assume that the smaller reservoirs filled over the course of a year. However, the same is not true for the largest reservoirs in the database. In the case of the largest nine reservoirs, we use various records of the filling duration or average discharge rate (e.g., Loo, 2007) and assume a steady filling rate. For these nine largest reservoirs, the average volume added per year is approximately 23 km3. The 49 largest reservoirs in GRanD database have a capacity larger than 23 km3, and for all these reservoirs we adopt a filling rate of 23 km3 yr-1. Each reservoir smaller than this is assumed to have filled the year the dam was completed. These filling rates are accounted for in the time series of Fig. 2.
For the future projections, we use a database compiled by Zarfl et al . (2015), which includes dams that will be built for hydropower and whose power generation capacity exceeds 1 MW (Fig. 1b). The dataset contains no direct information on the expected impounded volume of each reservoir; however, it lists the planned hydroelectric capacity. We use this value to estimate the impounded volume, following Grill et al . (2015):
V = (3.19 x 106 m3MW-1) P (1)
where V is the reservoir volume and P is the power generation capacity. Furthermore, because these dams were not complete at the time of the publication of Zarfl et al . (2015), many do not have an estimated year of completion. Instead, 3,565 reservoirs are listed as being in one of two broad categories, either in the construction phase (n = 501, or 15% of the total) or the planning phase (n = 2894, or 85%). Using these categories, and the equation above, we estimate that the reservoirs under construction will impound 663 km3 and that those being planned will contribute an additional 1,500 km3. Adding these values to results for the GRanD database yields the time series of integrated water volume impoundment shown in the inset to Fig. 2, assuming, following Zarflet al. (2015), that all dams will be completed by the year 2040 (blue histogram).
3 Results
3.1 Temporal Variation in GMSL
Removing 5,979 km3 of water from 1900 to 2011 according to the GRanD database corresponds to a GMSL fall of 16.6 mm, or an average 0.15 mm yr-1. As noted above, these values are 72% of the 23 mm and 0.21 mm yr-1 reported in Chao et al. (2008). Impounding the 663 km3(under construction) and 1,500 km3 (planned) from the Zarfl et al. (2015) database will cause a further globally averaged drop in sea level of 1.8 mm and 4.2 mm, respectively, for a total of 6 mm. Thus, over the period 2020 – 2040, the mean GMSL rate associated with water entering reservoirs will likely be –0.3 mm yr-1, which is larger than the average rate that occurred over the past century.
The above numbers will increase when accounting for natural seepage of water into the land surrounding the reservoir. After a reservoir is built, water will slowly seep into the neighboring land. As this occurs, the flow of water into the reservoir continues to recharge it, but the water that seeps from the reservoir remains relatively localized and does not generally reach the ocean, adding to the total impounded water. While the rate of seepage depends on numerous local factors (e.g ., Harr, 1962), for simplicity we follow Chao et al.(2008) and use a single equation to estimate the seepage rate at each reservoir:
V (t ) = 0.05 V 0t 1/2 (2)
where V 0 is the reported volume of the reservoir,t is the time in years since the water was impounded, andV (t ) is the total amount of water that seeps by timet . According to this equation, seepage contributed an estimated 1,878 km3 of water by 2011, for a total impounded volume of 7,857 km3. This corresponds to a total GMSL drop of 21.8 mm from 1900 – 2011 (Fig. 2, red histogram in main frame and inset). By 2040, we estimate that this value will increase to 30.1 mm, assuming for the seepage term that every “under construction” reservoir is completed by 2025, and every “planned” reservoir is completed by 2035. (We make the assumption that, for example, every “planned” reservoir is constructed in the decade 2030 – 2039; assigning every “planned” reservoir to the year 2039 will result in a seepage rate that is too low. We choose a date in the middle of the decade to give the best sense of the GRD fingerprint including seepage.)
Analysis of the GRanD time series indicates that the GMSL time series is characterized by markedly different rates in three distinct time windows: from 1900 to 1949 the volume of impounded water rose gradually; from 1950 to 1979 the rate of water impoundment increased dramatically, and from 1980 to 2011, new construction slowed but seepage became increasingly significant. Predictions from the Zarfl et al.(2015) database show a further dramatic increase in water impoundment in the next two decades. Many of these reservoirs will be near the coast (Fig. 1b), so a complete characterization of the effect this impounded water will have on sea level will help refine coastal hazard assessment.