An Abrupt Decline in Global Terrestrial Water Storage and Its Relationship with Sea Level Change

Abstract

As observed by the Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow On (GRACE-FO) missions, global terrestrial water storage (TWS), excluding ice sheets and glaciers, declined rapidly between May 2014 and March 2016. By 2023, it had not yet recovered, with the upper end of its range remaining 1 cm equivalent height of water below the upper end of the earlier range. Beginning with a record-setting drought in northeastern South America, a series of droughts on five continents helped to prevent global TWS from rebounding. While back-to-back El Niño events are largely responsible for the South American drought and others in the 2014-2016 timeframe, the possibility exists that global warming has contributed to a net drying of the land since then, through enhanced evapotranspiration and increasing frequency and intensity of drought. Corollary to the decline in global TWS since 2015 has been a rise in barystatic sea level (i.e., global mean ocean mass). However, we find no evidence that it is anything other than a coincidence that, also in 2015, two estimates of barystatic sea level change, one from GRACE/FO and the other from a combination of satellite altimetry and Argo float ocean temperature measurements, began to diverge. Herein, we discuss both the mechanisms that account for the abrupt decline in terrestrial water storage and the possible explanations for the divergence of the barystatic sea level change estimates.

Supplementary information: The online version contains supplementary material available at 10.1007/s10712-024-09860-w.

Keywords: Climate change; GRACE; Sea level; Terrestrial water storage.

Fig. 1

Anomalies (relative to the 2003–2020 mean) of global mean terrestrial water storage (equivalent height of water over land) from GRACE and GRACE-FO after removing annual and semiannual components and the S2 and K2 tidal alias terms. Ice covered land regions excluded from the global mean are shown in Figure S1. The shading conveys the formal error estimates, which average about ± 0.9 mm

Fig. 2

Anomalies (relative to the 2004–2010 mean) of terrestrial water storage (GT) from GRACE (black) and GRACE-FO (blue) averaged over the region of South America delineated in Fig. 3. The time series was deseasonalized and smoothed (7-month moving window with seasonal and trend decomposition using Loess). The shading indicates the formal errors

Fig. 3

Map of mascons in which the GRACE/FO era minimum terrestrial water storage occurred after the start of 2015, color coded for the year of that minimum. The time series plotted in Fig. 2 is averaged over the South American region delineated in red

Fig. 4

Time series of three estimates of the contribution of TWS to BSL (as sea level change equivalent in mm): (1) GRACE/FO observations of BSL after subtracting the contributions of glacier and ice sheet melt water inputs (based on Jet Propulsion Laboratory GRACE/FO data) and of atmospheric water vapor content variations (using the ERA5 reanalysis); (2) satellite altimetry measurements of GMSL after subtracting the contributions of melt water inputs, atmospheric water vapor content, and thermal expansion (based on Argo float data); and (3) hydrological model output (sum of the climate-driven contribution from ISBA-CTRIP model and human-induced linear contribution based on WaterGAP Hydrological Model, WGHM). Annual and semiannual signals have been removed. A mask is applied to gravimetry, altimetry, and Argo data over the ocean, excluding latitudes beyond ± 60°, closed seas, and Indonesian seas and coastal areas up to 200 km from the coastline

Fig. 5

Deseasonalized time series of TWS anomalies (TWSA) from ERA5, WaterGAP, SLR, and GRACE/FO, excluding Greenland, Antarctica, and the gulf coast of Alaska region. Both unfiltered monthly SLR (thin gold line) and filtered (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) SLR data (thick gold line) are plotted. The SLR data represent the sum of mascons 1, 2, and 5–19 as shown in Figure S1. The model time series are plotted as TWS anomalies relative to a zero mean. The vertical offset of the observational time series is artificial and was inserted for visual clarity, as only the temporal variations are meaningful (the absolute anomaly values are not)

Fig. 6

TWS trends (mm/yr) over the period 2003–2022 in a GRACE/FO and b ERA5-Land; and c correlations of monthly TWS anomalies between GRACE/FO and ERA5-Land. Ice sheets and areas of significant glacial extent have been masked (see Figure S2)

Fig. 7

Deseasonalized time series of area-weighted TWS (mm, left Y-axis) over six near-continental regions based on GRACE/FO (black line) and ERA5-Land (blue line). The area-weighted scaling ensures that the sum of these contributions is the global TWS time series as shown in Fig. 5. Shading indicates global SST anomalies (°C, right Y-axis), relative to a 2003–2022 base period. See Figure S2 for definition of continental averaging domains

Fig. 8

Comparison of ERA5-Land, GRACE/FO, and complementary observations: a deseasonalized time series of global land average (60S-70N) TWS anomalies (mm) from GRACE/FO, ERA5-Land, and ERA5 vertically integrated moisture flux convergence, b ERA5-Land precipitation (P), evapotranspiration (ET), and runoff (RO), c same as b but for GPCP P, GLEAM ET and G-RUN RO. Units in b and c are mm d⁻¹, d 2003–2022 trend in ERA5-Land P-ET, e Same as d but for GPCP v3.2 P—GLEAM 3.6a ET. Units in d and e are mm d⁻¹/yr. All anomalies are relative to the 2003–2022 monthly resolved climatological mean

Fig. 9

a Time series of near-global (60° S–70° N) mean SST anomalies, Niño 3.4 (black), the Pacific Decadal Oscillation (PDO) index with 1-year smoothing (blue), and a North Atlantic tripole index (brown). Negative PDO corresponds to anomalous positive SSTs in the central N. Pacific. b Near-global ERA5 SST trend, 1980–2022, °C/yr. c Correlation between the Niño 3.4 index and GRACE/FO d(TWS)/dt, the latter lagged 2 months with 4-month smoothing applied. d Correlation between the PDO index and GRACE/FO TWS, 3-month smoothing applied