Plain Language Summary
Filling a reservoir prevents that water from flowing back to the ocean,
thus causing sea level to fall. But sea level does not change by the
same amount everywhere: the mass of the water in the reservoir causes
sea level to rise in locations near the reservoir, but fall in locations
that are farther away. We use databases that include the locations and
capacities of reservoirs to estimate how constructing reservoirs has
changed sea level, and explore how these changes have varied in space
and in time. We find that while constructing reservoirs since 1900 has
caused sea level to fall on average, there are some locations that
experience a dramatic sea level rise, as much as 40 mm, over a short
period of time, usually only a few years. As more reservoirs are built,
we expect this trend to continue. In fact, reservoirs that are currently
being planned—if completed—will generate a sea level rise of a few
millimeters in some low-lying coastal areas. This rise would be in
addition to the rise in sea level from many other known factors.
1 Introduction
Global and regional changes in sea level are driven by a wide range of
processes, including the redistribution of mass from melting glaciers
and ice sheets, the continued adjustment of the solid Earth to ice mass
changes during the last glacial period, steric expansion of ocean water,
ocean circulation changes, and both natural and artificial changes in
terrestrial water storage (Cazenave et al ., 2018; Kopp et
al ., 2015). Proper characterization of each of these different
components of the sea level budget is necessary for accurate projections
of future changes in sea level. In this context, the role of terrestrial
water storage remains underexplored and serves as the focus of the
present study. We provide the most complete picture to date of the
impact that artificially impounded water has had on sea level over the
last century, resolved both spatially and temporally, and an estimate of
future sea level changes due to projected reservoir construction.
Chao et al. (2008) use the International Commission on Large Dams
World Register of Dams (WRD) (www.icold-cigb.org) database to construct
an estimate of the global mean sea level (GMSL) change due to the
construction of 29,484 reservoirs worldwide. They estimate the total
volume of impounded water to be 10,800 km3,
corresponding to a GMSL fall of ~0.55 mm
yr-1 in the half century prior to their publication.
Sea level changes associated with the impoundment of water on land will
be geographically variable (Fiedler and Conrad, 2010) and each reservoir
will have a unique sea level “fingerprint,” or gravitational,
rotational, and deformational (GRD) response to mass redistribution
(Gregory et al ., 2019). The redistribution of water from the
ocean to the reservoir will (1) increase the gravitational attraction of
the reservoir on the surrounding water and thus raise the local sea
surface height, (2) induce a change in Earth’s moment of inertia, and
(3) drive local crustal subsidence. Indeed, relative sea level (RSL)
will rise within ~2000 km of a reservoir being filled,
despite a drop in GMSL, and it will fall by increasing amounts at larger
distances from the source of impoundment. The local signal, which can
have a peak value an order of magnitude larger than the GMSL change
associated with the impoundment, is primarily a result of processes (1)
and (3). Calculating the global sea level pattern associated with water
impoundment requires knowledge of both the size and location of a
reservoir.
The WRD database adopted by Chao et al. (2008) includes the
largest global tabulation of reservoirs and the most complete estimate
of the total volume stored in those reservoirs. However, it does not
provide locations for these reservoirs and thus cannot be used to
generate maps of the associated sea level change. To compute such a map,
Fiedler and Conrad (2010) adopt a dataset (Vörösmarty et al .,
1997) that includes the locations of 674 reservoirs currently built and
scale their result upwards to match the GMSL cited by Chao et al .
(2008). We extend their analysis in three ways. First, we make use of a
much larger database of reservoirs. Second, we explore both the spatial
and temporal patterns of RSL change associated with water impoundment.
Finally, we project the signals into the future using a database of
planned dam construction (Zarfl et al ., 2015).
2 Reservoir Databases
In the present study, we use two different databases of reservoir
construction. For the time period 1900 – 2011, we use the Global
Reservoir and Dam (GRanD) database (Lehner et al. , 2011), which
aims to geospatially reference all reservoirs with a capacity of more
than 0.1 km3. It contains 6,329 reservoirs that were
completed after 1900 and reports their location, year of construction,
and capacity (Fig. 1a). The total volume of these reservoirs is 5,979
km3, which is less than 8,300 km3,
as reported in the WRD (Chao et al ., 2008). Time series of
integrated water volume impoundment (and equivalent GMSL fall) for the
GRanD database is shown in Fig. 2 (blue histogram). Although the GRanD
database contains only ~72% of the volume reported in
the WRD, we do not scale our total impounded water volume to match the
WRD values because the primary purpose of this study is to estimate the
spatial variability of sea level, and water impoundment not included in
the GRanD database has an unknown geographical distribution.