LETTERS PUBLISHED ONLINE: 2 DECEMBER 2012 | DOI: 10.1038/NGEO1652 Increased water storage in North America and Scandinavia from GRACE gravity data Hansheng Wang 1 * , Lulu Jia 1,2,3 , Holger Steffen 4,5 , Patrick Wu 4 , Liming Jiang 1 , Houtse Hsu 1 , Longwei Xiang 1,2 , Zhiyong Wang 6 and Bo Hu 1,2 Space-borne gravity data from the Gravity Recovery and Climate Experiment (GRACE) have revealed trends in present- day continental water storage in many parts of the world. In North America and northern Europe, it has been difficult to provide reliable estimates because of the strong background signals of glacial isostatic adjustment 1 . Attempts to separate the hydrologic signal from the background with numerical models 2 are affected by uncertainties in our understanding of the precise glacial history and mantle viscosity 3,4 . Here we use a combination of GRACE data and measurements from the global positioning system to separate the hydrological signals without any model assumptions. According to our estimates, water storage in central North America increased by 43.0 ± 5.0 Gt yr -1 over the past decade. We attribute this increase to a recovery in terrestrial water storage after the extreme Canadian Prairies drought between 1999 and 2005. We find a smaller rise in water storage in southern Scandinavia, by 2.3 ± 0.8 Gt yr -1 . In both North America and Scandinavia, our computed increases in water storage are consistent with long-term observations of terrestrial water level. We suggest that the detected mass gains in terrestrial water storage need to be taken into account in studies on global sea-level rise. Natural and anthropogenic stresses such as climate change, drought and deluge, increasing water use, land use and agricultural practices affect surface- and groundwater resources globally. A small change in the hydrological cycle may have a significant socio-economic impact, which, for example, has recently been observed in the form of groundwater depletion in northwest India probably leading to reduction of agricultural output and shortages of potable water 5 . It is thus of importance to determine the spatial and temporal variability in continental water storage 6 . The GRACE satellite mission has proved to be an invaluable tool in monitoring such hydrological changes with global coverage and sufficient spatial and temporal resolution 5,7 . However, such studies have been mainly limited to areas where the GRACE signal is assumed to be solely related to hydrology or where other overlapping processes can be removed as accurately as possible. This limitation is due to the fact that GRACE detects combined mass changes that can arise from hydrology, solid earth, cryosphere, oceans, atmosphere and tides 4 . Although the last two can be adequately removed during the data processing, one has to make sure that signals from other sources do not overlap. For example, the ongoing ice melt in Greenland and Antarctica is accompanied by glacial isostatic adjustment 8 (GIA), and the dominating GIA signal in North America and northern Europe is altered by water storage changes 2,4 . 1 State Key Laboratory of Geodesy and Earth’s Dynamics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China, 2 University of Chinese Academy of Sciences, Beijing 100049, China, 3 National Earthquake Infrastructure Service, Beijing 100036, China, 4 Department of Geoscience, University of Calgary, Calgary T2N 1N4, Canada, 5 Lantmäteriet, 80182 Gävle, Sweden, 6 Department of Surveying Engineering, Shandong University of Technology, Zibo 255049, China. *e-mail:whs@asch.whigg.ac.cn. In both cases, one has to hope that other data sets, observed or modelled, exist that quantitatively describe the unwanted signal so that it can be removed. For North America and northern Europe the issue becomes complicated if the GIA signal is desired. This is because until now there is no complete network of surface hydrological observations in those areas that can deliver an almost perfect hydrological signal. Imperfections in the hydrology model will contaminate the deduced GIA signal. In spite of this, the general practice is to use the hydrology model from the Global Land Data Assimilation System 9 (GLDAS/NOAH) or the WaterGAP Global Hydrology Model 10 (WGHM) to remove the hydrology signal 4 . Nonetheless, the models—and especially GLDAS/NOAH—are still used to reveal the GIA signal from GRACE (refs 2,4,11). In turn, if one is interested in the hydrological signal in the two regions, one has to find a method to remove the GIA contribution. One common approach is to use GIA models 1 based on ICE-5G (ref. 12) or Australian National University ice model (ANU-ICE; ref. 13) ice histories and selected mantle structures. Owing to large uncertainties in the deglacial histories and mantle rheology 3,4 , these models are incapable of precisely estimating the water changes from GRACE data in these previously glaciated areas. For example, it has been shown with the help of independent geodetic observations that the often used ICE-5G model contains too much ice west of Hudson Bay 14–17 . An alternative approach is to use observed data. The authors of ref. 18 proposed a relation between GIA-induced gravity and radial displacement signals that can be used to remove the effects of GIA from gravity and global positioning system (GPS) observations, and therefore helps separate the signals of present mass transport. In ref. 19, ICESat and GRACE data were combined to separate the Antarctic present-day mass balance and GIA signals. Recently, the authors of ref. 20 separated the present global water transport and GIA by using GRACE, GPS data and an ocean bottom pressure model. However, that study focused only on ice-mass loss and GIA in Antarctica and Greenland, and ice loss in Alaska. The authors of ref. 21 presented a new empirical relationship between GIA-induced gravity and radial displacement signals providing the means to separate simultaneously present-day mass loss and ongoing GIA in polar areas. In this study we propose a separation approach based on a modified extension of the method in ref. 18, which allows the self- determination of the hydrological contribution from GRACE and GPS data. Hence, the separation is not GIA model dependent. As an example we show the hydrological trend in North America and 38 NATURE GEOSCIENCE | VOL 6 | JANUARY 2013 | www.nature.com/naturegeoscience © 2013 Macmillan Publishers Limited. All rights reserved.