Volume 49, Issue 19 e2022GL100551
Research Letter
Free Access

Greenland Interannual Ice Mass Variations Detected by GRACE Time-Variable Gravity

Zhen Li

Zhen Li

State Key Laboratory of Geodesy and Earth's Dynamics, Innovation Academy of Precision Measurement of Science and Technology, Chinese Academy of Sciences, Wuhan, China

University of Chinese Academy of Sciences, Beijing, China

Search for more papers by this author
Benjamin Fong Chao

Benjamin Fong Chao

Institute of Earth Sciences, Academia Sinica, Taipei, Taiwan

Search for more papers by this author
Zizhan Zhang

Zizhan Zhang

State Key Laboratory of Geodesy and Earth's Dynamics, Innovation Academy of Precision Measurement of Science and Technology, Chinese Academy of Sciences, Wuhan, China

Hubei Luojia Laboratory, Wuhan, China

Search for more papers by this author
Liming Jiang

Liming Jiang

State Key Laboratory of Geodesy and Earth's Dynamics, Innovation Academy of Precision Measurement of Science and Technology, Chinese Academy of Sciences, Wuhan, China

Search for more papers by this author
Hansheng Wang

Corresponding Author

Hansheng Wang

State Key Laboratory of Geodesy and Earth's Dynamics, Innovation Academy of Precision Measurement of Science and Technology, Chinese Academy of Sciences, Wuhan, China

Correspondence to:

H. Wang,

[email protected]

Search for more papers by this author
First published: 29 September 2022
Citations: 2


To better understand the impact of climate forcing on Greenland ice sheet (GrIS), we study the GrIS mass variations on the interannual timescale as observed by the satellite mission of Gravity Recovery and Climate Experiment (GRACE). By employing the method of empirical orthogonal function) analysis, we find: (a) That the GrIS interannual mass variations are significantly correlated with the Pacific Decadal Oscillation, suggesting a connection to the changes of the Icelandic Low (a permanent low-pressure system) related to PDO; (b) An East-West Costal Dipole related to precipitation anomaly subjects to the North Atlantic Oscillation; (c) Certain contribution of the Atlantic Multidecadal Oscillation to GrIS mass variations in the form of temperature and runoff anomalies.

Key Points

  • Greenland interannual mass variations are correlated with the Pacific Decadal Oscillation

  • We report an East-West Coastal Dipole related to the North Atlantic Oscillation

  • The contribution of the Atlantic Multidecadal Oscillation is also captured

Plain Language Summary

For the past few decades, Greenland ice sheet (GrIS) has undergone accelerated melting under impacts of the global climate changes, leading to significant change in the sea level. GrIS has been well monitored by the Gravity Recovery and Climate Experiment (GRACE) satellite mission that observes Earth's gravity change due to mass transports since its launch in 2002. Here we use GRACE data to understand the responses of GrIS to climatic forcing, which can reveal certain phenomena that are hitherto unreported: (a) We show that the Pacific Decadal Oscillation plays a significant role on GrIS variations; (b) We also report a “seesaw” pattern of an East-West Costal Dipole related to the North Atlantic Oscillation; (c) The Atlantic Multidecadal Oscillation also contributes to GrIS mass variations.

1 Introduction

As the largest body of fresh water on Earth with full amount equivalent to a global sea level rise by ∼70 m (Bamber et al., 2001; Cazenave et al., 2009), the Antarctica and Greenland ice sheet (GrIS) have experienced a most pronounced melting (van den Broeke et al., 2009; Harig & Simons, 2012), and become major contributors to sea-level rise (Rignot et al., 2011). GrIS is highly sensitive to climatic changes, as Greenland has experienced rapid warming episodes over the past few decades (Hall et al., 2013). Climatic parameters, such as temperature and precipitation have shown intensified relationship with GrIS mass balance. For example, using GRACE and regional climate model, Seo et al. (2015) found the most recent period acceleration of GrIS can be explained by runoff and ice discharge increase and precipitation decrease. Wu et al. (2022) further found mass variations of southern Greenland subregions mainly affected by pronounced precipitation. As temperature and precipitation patterns generally reflect large-scale atmospheric circulations, B. Zhang et al. (2019) showed that the atmospheric circulations anomalies as measured by the Greenland Blocking Index (GBI) are significantly related to GrIS transient mass changes—high GBI is usually associated with abnormally mass loss while low GBI can lead to mass gain. In addition, the North Atlantic Oscillation (NAO) also plays an important role on Greenland's precipitation with a distinct east–west asymmetry (Bjørk et al., 2018). However, the climate driving mechanism for GrIS mass balance is still yet to be understood.

Since its launch in 2002, GRACE satellite mission has documented GrIS mass variations on a monthly basis with a spatial resolution of ∼300–400 km (Tapley et al., 2004), providing reliable observation of understanding toward the climatic changes over Greenland (e.g., Velicogna et al., 2020; B. Zhang et al., 2019). While most of GRACE works have studied the secular trend (e.g., Jacob et al., 2012; Sasgen et al., 2012; Velicogna et al., 2020), here we focus on the interannual timescale of the GrIS variations, a critical aspect to understand GrIS’ response to climatic changes.

2 Data and Preparation

2.1 GRACE Data

We used the GRACE RL06 gridded mascon solutions from Center for Space Research, University of Texas, for the period from 2003 to 2015. The missing data during the study period is filled by linear interpolation. No additional requirements for post-processing (such as spatial smoothing and de-striping filtering) are needed for the mascon solution, that could avoid potential errors caused by different post-processing steps.

The mascon solution inverses surface mass changes by two-steps regularization matrices (Save et al., 2016). The first regularization matrix aims to suppress the signal leakage of the land signals into the ocean, and generates an intermediate solution having little or no leakage. The intermediate solution then is further optimized by the second regularization matrix, meanwhile the stripe errors in GRACE data are greatly reduced. The conducted data pre-processing, such as replacement of low degree coefficients (Sun et al., 2016; Swenson et al., 2008), glacial isostatic adjustment correction (Peltier et al., 2018) and atmosphere and ocean de-aliasing (Dobslaw et al., 2017), has been made ensuing that mascon solution can be used directly. The mascon data is provided on 0.25° latitude-longitude global grids in terms of equivalent water height (EWH) in centimeter.

Aiming at the interannual ice-mass redistribution, the (monthly) mascon time series urn:x-wiley:00948276:media:grl64914:grl64914-math-0001 at each grid are first decomposed into a quadratic trend, two seasonal signals (annual and semi-annual terms) plus the residual signal, in the form of:
with angular frequency urn:x-wiley:00948276:media:grl64914:grl64914-math-0003. All coefficients are then estimated via linear least-squares regression by minimizing the variance of Residual(t). Residual(t), obtained by subtracting out the fitted signals as above, constitutes the target signal of interannual mass variations to be input into the analysis below.

2.2 Climatic Index Series

To investigate the climatic connections with GrIS variation phenomena, climatological cycles of the Pacific Decadal Oscillation (PDO), the NAO, and the Atlantic Multidecadal Oscillation (AMO) are considered in terms of their indices.

The PDO Index is defined as the pattern and time series of the first empirical principal component of the sea surface temperature (SST) over the mid-latitude North Pacific north of 20°N (Mantua et al., 1997). A PDO-positive phase favors a cold SST anomaly in western North Pacific, while the opposite occurs in PDO-negative phase. NAO consists of a north-south dipole of anomalies, with one center located over Greenland and the other of opposite sign spanning the central latitudes of the North Atlantic (Hurrell, 1995). And AMO is an ongoing coherent mode of natural variability occurring in the North Atlantic Ocean, based on the average anomaly of SST, typically over 0–80°N (Kaplan et al., 1998).

2.3 Climatic Parameters

To investigate the intrinsic drivers of GrIS mass change, a set of climate parameters are implemented. The 500-hPa geopotential height data is adopted to understand the changes of atmospheric circulation. Positive geopotential height anomaly usually indicates high atmospheric pressure that always links to warmer air, while negative anomaly represents low atmospheric pressure related to colder air (Rajewicz & Marshall, 2014). Besides, precipitation, runoff, and surface temperature data are also investigated. Those parameters are available as monthly means on a 1/10° latitude-longitude global grid provided by the fifth generation of European Re-Analysis (ERA5) data set. For each parameter, we remove the secular linear trend and seasonal terms by least squares regression, and obtain its anomaly variations.

2.4 Surface Mass Balance (SMB) From Regional Climate Model

SMB is the mass change due to precipitation, runoff, sublimation processes, and dominates the interannual mass changes of GrIS (e.g., B. Zhang et al., 2019). We adopt SMB data modeled by MAR (Modèle Atmosphérique Régional) version 3.12 (Fettweis et al., 2017) to verify the outputs of GRACE. MAR is a coupled regional climate model which is forced by reanalysis meteorological data (Fettweis et al., 2013). MAR utilizes a multilayered energy balance model (Brun et al., 1992) that determines the energy exchanges between the ice sheet, atmosphere and oceans, and generates different constituents of ice-mass balance. The monthly MAR data set adopted here is provided at 1 urn:x-wiley:00948276:media:grl64914:grl64914-math-0005 1 km grids over the entire Greenland with the time span from 1980 to 2020.

3 Method

We shall analyze the interannual ice mass redistribution by the EOF method, a form of the general principal component analysis (e.g., Preisendorfer, 1988), which has been widely applied to investigate ice-mass variability (e.g., Li et al., 2022; Mémin et al., 2015). We construct the space-time data matrix D(x,y;t), where x and y are the geographic grid points and t is the (monthly) time tag. The EOF method then decomposes the space-time data set into orthogonal standing-wave oscillations or modes delimited with the target region and time span, each of which consists of a spatial pattern multiplied by the corresponding temporal variability. Thus, the eigen-solution of the covariance matrix urn:x-wiley:00948276:media:grl64914:grl64914-math-0006 are obtained using the singular value decomposition, forming a superposition of orthogonal standing oscillations, or eigen-modes, with an estimated variance percentage, expressed in the form of:
urn:x-wiley:00948276:media:grl64914:grl64914-math-0008 is the eigenvector solution or the spatial pattern of the kth EOF mode, and urn:x-wiley:00948276:media:grl64914:grl64914-math-0009 is the corresponding temporal function or the projection of D onto urn:x-wiley:00948276:media:grl64914:grl64914-math-0010 obtained by the inner product:
we normalize the spatial pattern urn:x-wiley:00948276:media:grl64914:grl64914-math-0012 with respect to its standard deviation, so that urn:x-wiley:00948276:media:grl64914:grl64914-math-0013 manifests the actual physical amplitude of the target quantity, in our case the EWH in unit of cm (for detail see, Chang & Chao, 2014; Chao & Liau, 2019).

4 Results

4.1 EOF Mode 1: Interannual Mass Variations Correlated With PDO

Figure 1 shows the first three EOF modes, which collectively account for ∼80% of the total variance. They effectively retrieve the sought interannual signals. The leading EOF Mode 1 (Figure 1a) with 35.6% of the total variance is mainly concentrated along southern coastal Greenland. The corresponding time function shows strong interannual undulations, particularly during the latter half of the studied period 2009–2015. Duplicated in Figure 2a, it is found to be significantly correlated with the PDO Index, the correlation coefficient is 0.53 at near-zero time shift (Figure 2b). To let such correlation coefficient be over, say, 99% statistical confidence level only requires the effective degree of freedom (DOF) to be over ∼20 (see Chao & Chung, 2019), which is easily met by our present broadband data that has the DOF estimated to be ∼32 (for detail see, Text S1 in Supporting Information S1).

Details are in the caption following the image

Empirical orthogonal function (EOF) mode solutions of the interannual mass variations over Greenland (a–c) Modes 1–3 respectively. Upper panels are the spatial patterns (normalized with respect to standard deviation), and lower panels are the time series exhibiting equivalent water height (EWH) in cm.

Details are in the caption following the image

Empirical orthogonal function (EOF) Mode 1 correlates with Pacific Decadal Oscillation (PDO). (a) Time series of EOF Mode 1 (duplicated from Figure 1a and normalized) and PDO Index, overlaid with the pink and blue shaded region indicating respectively the strong positive and negative PDO phases (the yellow shaded years indicates the stable PDO phase). (b) cross-correlation coefficients between the two time series of (a) as a function of relative time shift, where negative value means PDO leading.

As a most important ocean-atmosphere coupling process over the Northern Hemisphere oceans (Svendsen et al., 2018), PDO has significant impact on Arctic climatic changes; particularly, it has been shown that Arctic sea ice loss is regulated by PDO (e.g., Notz, 2009; S. Zhang et al., 2020). Figure 3 presents the behavior of the 500-hPa geopotential height anomalies under different PDO conditions, as well as the corresponding temperature variations. From 2003 to 2006, the PDO Index oscillated mainly within ±0.5, and we shall refer to it as a PDO-stable phase (the yellow shaded region in Figure 2a). The geopotential height in the PDO-stable phase shows a slight positive increase (Figure 3a), and the surface temperature anomaly in Figure 3b is close to zero. Therefore, the ice mass has been relatively constant. However, during the strong PDO-negative phase, for example, 2011 to 2013 as shown in Figure 2a (the pink shaded regions), the geopotential height over western side of Greenland increased significantly (Figure 3c), corresponding to an exceedingly warmer surface temperature over almost the entire ice sheet (Figure 3d). Such extraordinary warming could lead to significant ice melting, coinciding with the abnormal downswing of the ice mass in Figure 2a. The opposite trend occurs during a strong PDO-positive phase (the blue shaded region in Figure 2a), for example, 2009 to 2010, and 2014 to 2015, negative geopotential height anomaly is evident over South Greenland (Figure 3e). That is consistent with negative surface temperature anomaly along the southwest coast (Figure 3f) which reduces glacial melting and favors the upswing mass accumulation.

Details are in the caption following the image

The 500-hPa geopotential height and temperature anomalies over Greenland for three different conditions. (a–b) Geopotential height anomaly for a Pacific Decadal Oscillation (PDO)-stable condition and the corresponding temperature anomaly of (b); (c–d) similar to that of (a–b), but for strong PDO negative phase; (e–f) similar to that of (a–b), but for strong PDO-positive phase.

The response of Greenland's temperature to atmospheric circulation is complex, as B. Zhang et al. (2019) have shown the GBI also accounts for Greenland's temperature anomaly, such as high GBI being related to extreme high temperature in 2012. The 500-hPa geopotential height anomalies shown in Figure 3 may represent a joint effect of PDO and GBI. Figure S1 in Supporting Information S1 shows the time series of Greenland's temperature anomaly, the GBI and PDO Index, as well as their low frequency variations, for the period of 2003 through 2019. Both PDO and GBI evidently contribute to temperature anomalies over Greenland, for example, the positive temperature anomaly during 2010 is relate to high positive GBI, while the extreme high-temperature during 2012 is a joint effect of GBI and PDO. In addition, temperature anomalies in 2015 (negative) and 2019 (positive) are related to continuous high PDO Index. Besides GBI, we found that PDO also has significant impact on the Greenland's temperature, whereby a negative PDO warms Greenland and a positive PDO does the opposite.

To investigate the possible physical mechanism thereby PDO drives the surface temperature of Greenland, we extend the study area to the middle latitudes and focus on the 500-hPa geopotential height in summer and winter, seasons that associate with mass melting and accumulation. Compared to the said PDO-stable condition (Figure S2a in Supporting Information S1), larger geopotential height is evident at high latitudes during the PDO-negative phase (Figure S2b in Supporting Information S1), particularly in regions under the control of the Icelandic Low (IL), a permanent low-pressure system near Iceland. The winter geopotential height variation related to a positive PDO is opposite to that related to a negative PDO. High latitudes are characterized by smaller geopotential heights (Figure S2d in Supporting Information S1) than the winter geopotential heights averaged during PDO-stable condition (Figure S2c in Supporting Information S1), especially over regions near the IL. It suggests that changes in geopotential height at high latitudes show more significant variations than those at mid-latitudes during high PDO Index, especially for regions around the IL.

Previous studies have reported similar changes in the Aleutian Low with PDO, a permanent low-pressure system in the North Pacific, as atmospheric system over oceans is closely related to SST (Dong et al., 2013). We also find similar changes in the IL (Figure S3a in Supporting Information S1). A negative PDO warms the seawater over southern Greenland (Figure S3b in Supporting Information S1) and strengthens the upper air pressure, weakening the IL. In contrast, the IL is strengthened during a positive PDO caused by the colder seawater (Figure S3c in Supporting Information S1).

The impaired IL during the PDO-negative phase results in a weaker pressure gradient over southern Greenland, which could decrease zonal wind and slow the progression of upper-level flow (Francis & Vavrus, 2012). Weaker zonal mean wind is generally associated with increasing atmospheric blocking events (Barriopedro & Garcia-Herrera, 2006) and causes more persistent extreme weather conditions, such as droughts and heat waves. Meanwhile, a weaker pressure gradient causes airflow to be easily disturbed, resulting in a more meandering flow and the advection of warm air from middle latitudes toward the polar region. Both these effects can increase temperatures in Greenland, corresponding to the expected geopotential ridge elongation shown in Figure S2b in Supporting Information S1, and increases surface meltwater runoff that drops mass balance. In contrast, a positive PDO reinforces the air pressure of the IL, leading to the geopotential ridge moving southward, with the strengthened pressure gradient promoting the movement of air flow and preventing warm flow from middle latitudes. Therefore, a positive PDO favors Greenland cooling. A detail flow chart of the possible mechanism is shown in Figure S4 in Supporting Information S1.

4.2 EOF Mode 2: East-West Coastal Dipole (EWCD) Caused by Precipitation Anomaly Relative to NAO

Proceeding, EOF Mode 2 (Figure 1b) accounts for 24.8% of the total variance and shows a clear dipole or a seesaw pattern along the coast, mostly between East and West Greenland. We shall refer to this mode of ice-mass variations as the Greenland East-West Coastal Dipole (EWCD). With strong interannual undulation, EWCD's time history undergoes two pulses during the studied period, one is during 2003–2006 peaking in 2004, and another during 2007–2015 peaking in 2011.

Since EOF Mode 1 has been identified as representing mass variations induced by PDO, we now remove it from the interannual GRACE data. Then for a more refined perception of the second mode, we conduct two separate EOF analyses on the residual data, with each restricted to one subregion of the two strongest oscillations, namely coastal East Greenland and coastal West Greenland (Figure S5 in Supporting Information S1). The leading modes of the two separate EOF analyses now show a clear negative correlation in their time series (as expected), with a peak correlation coefficient value reaching −0.71. For the periods 2003–2006 and 2012–2015, West Greenland shows a gradual mass increase, with a significant mass decrease between the two periods. East Greenland exhibits the opposite behavior, for example, decreasing mass for 2003–2006 and 2012–2015, and increasing mass for 2007–2011.

A similar counteracting pattern of precipitation anomaly over Greenland had been reported by Seo et al. (2015) and Bjørk et al. (2018) on a decadal timescale, which may provide a scope for the origin of EWCD. The seesaw spatial patterns can be seen in precipitation anomalies (Figure 4) for the above three most pronounced phases: 2003–2006, 2007–2011, and 2012–2015. During the two periods 2003–2006 and 2012–2015, precipitation exhibited positive anomalies from West Greenland, and negative anomalies were evident over Southeast Greenland (Figures 4a and 4c). While opposite precipitation anomalies evidently exhibited in 2007–2011 (Figure 4b) coinciding with the opposite ice mass variations according to the EOF-based time series for the two subregions. Variabilities in precipitation coincide well with the ice mass anomaly, implying that EWCD originates from the precipitation anomaly.

Details are in the caption following the image

Precipitation anomaly over Greenland. (a) 2003–2006; (b) 2007–2011; (c) 2012–2015.

Precipitation over Greenland, particularly seasonal variations, is closely related to large-scale Arctic circulation (Koyama & Stroeve, 2019) of NAO. We further assess the apparent connection between seasonal mode of NAO and the interannual ice mass balance. Figure S6 in Supporting Information S1 presents the climatological seasonal mean of the interannual ice mass balance for both West and East Greenland based on the (monthly) time series in Figure S5 in Supporting Information S1, as along with the seasonal NAO Index. We statistically analyze the correlation between the glacier mass change and NAO (Table S1 in Supporting Information S1). Ice mass variations in spring and fall do not consistently correlate with NAO; however, a significant correlation is evident in summer and winter. The correlation of summer and winter mass variations of East Greenland with the seasonal NAO Index is negative, their correlation coefficients are −0.64 and −0.74, respectively. Whereas a positive correlation for West Greenland is evident, with the correlation coefficients in summer and winter both are 0.76.

In summary, EOF Mode 2 shows an EWCD that is caused by precipitation anomaly in relation to NAO. As most studies mainly focused on the temporal response of the precipitation to NAO (e.g., Peng et al., 2021; Seo et al., 2015), particularly seasonal precipitation, we present the spatial coherence between GrIS in view of glacier mass variations and NAO. The dipole pattern revealed by EOF Mode 2 implies an opposite impact of NAO on East and West Greenland both for precipitation and glacier mass variations.

4.3 EOF Mode 3: Contribution of AMO to Glacier Mass Variations

EOF Mode 3 (Figure 1c), which accounts for 18.1% of the total variance, shows largest amplitude mostly concentrated along North and South Greenland. When a 12-month time shift is allowed (see Figures S7a and S7b in Supporting Information S1), the correlation coefficient of the corresponding time series with AMO is ∼0.43 (the DOF is ∼25), relatively weak meeting only 95% confidence level, may subject to the limited GRACE data length inadequate to capture their correlation on longer timescales (Chylek et al., 2009). We present below certain pertinent, albeit peripheral, climate parameters that may link AMO to GrIS mass variations.

We first employ EOF on Greenland's surface temperature anomaly and attempt to identify the contribution of AMO to temperature variability (Figures S7c and S7d in Supporting Information S1). The time series of the leading mode of the temperature anomaly, which account for 64.1% of the total variance, show significant correlation with AMO, their correlation coefficient is 0.64 (the DOF is ∼23), which well meets 99% confidence level. Generally, runoff is closely related to changes of surface temperature (Francis & Vavrus, 2012), we then employ EOF on Greenland's runoff anomaly (Figures S7e and S7f in Supporting Information S1). Accounting for 44.6% of the total variance, the spatial pattern of the leading mode is much pronounced over southern Greenland (Figure S7e in Supporting Information S1), and the corresponding time history is also found to be correlated with AMO (Figure S7f in Supporting Information S1) with the correlation coefficient is 0.50 (the DOF is ∼15). Above analysis suggests that AMO has import contribution to Greenland's temperature and runoff anomalies.

5 Discussion

We also do EOF on MAR data set to verify the outputs of GRACE. Figure S8 in Supporting Information S1 shows the first three EOF modes of MAR data, which collectively account for ∼90% of the total variance. They show a basically consistence with that of GRACE, both for spatial and temporal modes, confirming the reliability of GRACE outputs. Accounting for 74.0% of the total variance, the time series of MAR EOF Mode 1 also correlate with PDO, with the correlation coefficient is 0.42 (the DOF is ∼50), meeting 99% confidence level. The consistency is particularly well during continuous high PDO Index, for example, the strong PDO-positive phases around 1983, 1993, 1997, and 2015 lead to pronounced ice mass accumulation, while strong PDO-negative phases around 1985, 1990, 1995, 2012, and 2020 coincide with the significant ice mass losing. Mode 2 of MAR (with 9.9% of the total variance) reveals the relationship of GrIS mass variation with AMO, their correlation coefficient is 0.40 (the DOF is ∼40), which also meets the 99% confidence level. Finally, the spatial part of MAR EOF Mode 3 (with 5.7% of the total variance) shows a similar diploe pattern with GRACE data, further confirms the heterogeneity in Greenland ice mass. However, the time series of EOF Mode 2 don't show significant agreement with NAO, not surprisingly, because the opposite behavior of East Greenland and West Greenland toward NAO could weaken the correlation.

In addition, the wavelet spectra (Figure S9 in Supporting Information S1) evidently present certain evidences linking GrIS mass variations to aforementioned climatic factors. In particular, the ∼3-6-year periodicity of MAR EOF Mode 1 seen in wavelet spectra is also found in PDO Index, whereas EOF Mode 2 and AMO Index both show ∼6–10-year oscillation. Besides, NAO may account the ∼3-year periodicity in EOF Mode 3. The wavelet spectra reveal the correlation between GrIS mass variations and those climatic factors mainly concentrates in low frequency band.

6 Conclusions

Focusing on the impact of climatic forcing on GrIS, we employ EOF method on interannual ice-mass redistribution as observed by GRACE mascon data over Greenland for the period from 2003 to 2015. Several facts found in the scenarios of the three phenomena during the studied period are worth noting: (a) We show that the interannual mass variations of GrIS are significantly correlated with PDO. PDO affects Greenland's temperature by the connection to changes of the Icelandic Low, positive PDO cools Greenland, while negative PDO warms Greenland. (b) The EWCD suggests the east–west asymmetry spatial coherence between GrIS in view of mass balance and NAO, subjecting to the formerly reported NAO-related precipitation anomaly. (c) AMO contributes to GrIS interannual mass variations in the form of temperature and runoff anomalies. All findings are matched by outputs of regional climate model, confirming its authenticity. Our new founding will further our understanding of GrIS's dynamic variations for future climate changes.


We would thank Dr. Shujing Shen for her helpful discussion. This work was supported by the National Key R&D Program of China (2017YFA0603103), the National Natural Science Foundation of China (41974009, 42074094), the Key Research Program of Frontier Sciences, the Chinese Academy of Sciences (QYZDB-SSW-DQC042, QYZDB-SSW-DQC027), the Open Fund of Hubei Luojia Laboratory (220100044) and the Taiwan Ministry of Science and Technology (111-2116-M-001-024).

    Data Availability Statement

    All Data is openly available in public repository. The GRACE data is downloaded from http://www2.csr.utexas.edu/grace/, and the ERA5 reanalysis data set is available at https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5. The Climatic Index series are obtained from https://www.psl.noaa.gov/data/climateindices/. MAR data is available at https://doi.org/10.5281/zenodo.6939913.