Long-term vertical land motion from double-differenced tide gauge and satellite altimetry data

We present a new approach to estimate precise long-term vertical land motion (VLM) based on double-differences of long tide gauge (TG) and short altimetry data. We identify and difference rates of pairs of highly correlated sea level records providing relative VLM estimates that are less dependent on record length and benefit from reduced uncertainty and mitigated biases (e.g. altimeter drift). This approach also overcomes the key limitation of previous techniques in that it is not geographically limited to semi-enclosed seas and can thus be applied to estimate VLM at TGs along any coast, provided data of sufficient quality are available. Using this approach, we have estimated VLM at a global set of 86 TGs with a median precision of 0.7 mm/year in a conventional reference frame. These estimates were compared to previous VLM estimates at TGs in the Baltic Sea and to estimates from co-located Global Positioning System (GPS) stations and Glacial Isostatic Adjustment (GIA) predictions. Differences with respect to the GPS and VLM estimates from previous studies resulted in a scatter of around 0.6 mm/year. Differences with respect to GIA predictions had a larger scatter in excess of 1 mm/year. Until satellite altimetry records reach enough length to estimate precise VLM at each TG, this new approach constitutes a substantial advance in the geodetic monitoring of TGs with major applications in long-term sea level change and climate change studies.     

Systematic errors in the Z-geocenter derived using satellite tracking data: a case study from SPOT-4 DORIS data in 1998

Within the scope of the Global Geodetic Observing System, Doppler Orbit Determination and Radiopositioning Integrated by Satellite – as a geodetic technique – can provide precise and continuous monitoring of the geocenter motion related to mass redistribution in the Earth, ocean and atmosphere system. We have reanalyzed 1998 DORIS/SPOT-4 (Satellite pour l’ Observation de la Terre) data that were previously generating inconsistent geocenter positions (?65 cm offset). We show here that this error is due to an incorrect phase center correction provided with the DORIS preprocessed data resulting from a +12 cm offset in the cross-track direction that has been confirmed since. We also conclude that a 1 mm error in the cross-track offset of non-yawing sun-synchronous SPOT satellites will generate a ?6.5 mm error in the derived Z-geocenter. Other non-yawing satellites would also be affected by a similar effect whose amplitude could be easily estimated from the orbit inclination     

Spatially variable surface elevation changes and estimated melt water contribution of Continental Glacier in the Wind River Range,Wyoming, USA: 1966–2011

Seasonal snow melt in the Wind River Range, Wyoming, has been ending earlier over the last several decades leaving the region to rely more on supplemental melt water from mountain glaciers. This leads to the necessity of understanding recent glacial changes. This study uses elevation data from 1966, 2006 and 2011 to calculate surface elevation and volume changes that have occurred on Continental Glacier. Results indicate a mean volume change of ?0.034 ± 0.02 km³ and surface elevation change of ?0.36 ± 0.19 m y^?1 between 1966 and 2006. Detailed spatial analysis shows that the glacier is divided into two sections which are thinning at different rates (lower section: ?0.06±0.19 m y^?1; upper section: ?0.51 ± 0.19 m y^?1). The upper section has experienced 97% of the thinning (or 742.5 × 10³ m³ of melt water equivalent per year) and increased its rate since 2006 by 27.5%.     

The gravitational effect of ocean tide loading at high latitude coastal stations in Norway

Gravity measurements close to the ocean are strongly affected by ocean tide loading (OTL). The gravitational OTL effect consists of three parts, i.e. a change in gravity caused by direct attraction from the variable water-masses, by displacement of the observing point due to the load, and by redistribution of masses due to crustal deformation. We compare the OTL gravitational effect of several global models to observed time-series of gravity to identify the best model for four arctic observation sites. We also investigate if the global models are sufficient for correcting gravity observations. The NAO99b model fits the observations best at three stations. At two stations (Tromsø and Bodø) the global models explain the variability in the observations well. At the other two (Honningsvåg and Andøya), a significant periodic signal remains after the OTL correction has been applied. We separate two of the gravitational effects, the direct attraction and the change in gravity due to displacement, to study the local effects. Simple geometric models of the water load and independent measurements from local tide-gauges are used to calculate these effects. This leads to improved correspondence with the OTL signal, hence demonstrating the importance of careful modelling of local effects for correction of gravity observations in coastal stations.     

Detection of Changes in Glacier Mass Balance Using Satellite and Meteorological Data in Tirungkhad Basin Located in Western Himalaya

Himalayan glaciers and their mass balance are poorly sampled through direct mass balance measurements. Thus, based on Landsat datasets of ETM⁺ (2000), ETM⁺ (2006) and TM (2011), mass balance studies of 32 glaciers was carried out using accumulation area ratio (AAR) method in the Tirungkhad river basin, a tributary of Satluj River, located in western Himalayan region. The overall specific mass balance was negative varying from ?27 cm (2000) to ?41 cm (2011). Out of 32 glaciers, 27 glaciers (81.2 %) showed negative mean mass balance and 5 glaciers (18.7 %) showed positive mean mass balance. Mean of specific mass balance for the year 2000, 2006 and 2011 was found to be ?48 cm, ?55 cm and ?0.61 cm respectively, in case of glaciers with negative mass balance while in case of glaciers with positive mass balance, it was 0.67 cm (2000), 0.56 cm (2006) and 0.47 cm (2011). The investigations suggested a loss of ?0.034 km³ of glacial ice for 2000, 0.036 km³ for 2006 and 0.038 km³ for 2011 respectively. The negative mass balance of the glaciers since 2000 correlates well with the increasing trend of annual mean temperature and decreasing trend of precipitation (snow water equivalent (SWE) and rainfall). Based on Mann Kendall test the temperature and SWE trends were significant at 95 % confidence level, however, the rainfall trend was insignificant.     

Secular changes in Earth’s shape and surface mass loading derived from combinations of reprocessed global GPS networks

The changing distribution of surface mass (oceans, atmospheric pressure, continental water storage, groundwater, lakes, snow and ice) causes detectable changes in the shape of the solid Earth, on time scales ranging from hours to millennia. Transient changes in the Earth’s shape can, regardless of cause, be readily separated from steady secular variation in surface mass loading, but other secular changes due to plate tectonics and glacial isostatic adjustment (GIA) cannot. We estimate secular station velocities from almost 11 years of high quality combined GPS position solutions (GPS weeks 1,000–1,570) submitted as part of the first international global navigation satellite system service reprocessing campaign. Individual station velocities are estimated as a linear fit, paying careful attention to outliers and offsets. We remove a suite of a priori GIA models, each with an associated set of plate tectonic Euler vectors estimated by us; the latter are shown to be insensitive to the a priori GIA model. From the coordinate time series residuals after removing the GIA models and corresponding plate tectonic velocities, we use mass-conserving continental basis functions to estimate surface mass loading including the secular term. The different GIA models lead to significant differences in the estimates of loading in selected regions. Although our loading estimates are broadly comparable with independent estimates from other satellite missions, their range highlights the need for better, more robust GIA models that incorporate 3D Earth structure and accurately represent 3D surface displacements.     

利用GRACE、GPS和绝对重力数据监测斯堪的纳维亚陆地水储量变化

重力卫星GRACE(gravity recovery and climate experiment)监测斯堪的纳维亚半岛陆地水储量变化会受到冰川均衡调整(GIA)信号的严重影响。首先根据该地区绝对重力和GPS并址观测数据计算了GIA重力和垂直位移的实测线性比值,利用该比值和GPS网观测的垂直位移速度场得到了GIA重力。然后,对GRACE观测的重力变化速率进行GIA重力改正,进而可分离陆地水储量变化趋势,避免了使用GIA模型所带来的巨大不确定性,并根据观测数据完整估计了所得结果的不确定性。最后与水文模型作对比分析。结果表明,实测的GIA重力-垂直位移线性比值为0.148±0.020μGal/mm(1Gal=10-2 m/s2),该结果检验了Wahr的理论近似值且与北美实测的结果非常接近。2003年1月至2011年3月期间,斯堪的纳维亚半岛陆地水储量存在明显的增加趋势,信号的主体位于半岛南端的维纳恩湖附近,总的水量增加速率为4.6±2.1km3/a,数据观测期间的累积增加水量为38±17km3。研究结果与WGHM水文模型的结果有较好的一致性,相关系数达到0.69,而与GLDAS水文模型的相关性略小。     

A fixed full-matrix method for determining ice sheet height change from satellite altimeter: an ENVISAT case study in East Antarctica with backscatter analysis

A new method, called the fixed full-matrix method (FFM), is used to compute height changes at crossovers of satellite altimeter ground tracks. Using the ENVISAT data in East Antarctica, FFM results in crossovers of altimeter heights that are 1.9 and 79 times more than those from the fixed half method (FHM) and the one-row method (ORM). The mean standard error of height changes is about 14 cm from ORM, which is reduced to 7 cm by FHM and to 3 cm by FFM. Unlike FHM, FFM leads to uniform errors in the first-half and second-half height-change time series. FFM has the advantage in improving the accuracy of the change of height and backscattered power over ORM and FHM. Assisted by the ICESat-derived height changes, we determine the optimal threshold correlation coefficient (TCC) for a best correction for the backscatter effect on ENVISAT height changes. The TCC value of 0.92 yields an optimal result for FFM. With this value, FFM yields ENVISAT-derived height change rates in East Antarctica mostly falling between \(-3\) and 3 cm/year, and matching the ICESat result to 0.94 cm/year. The ENVISAT result will provide a constraint on the current mass balance result along the Chinese expedition route CHINARE.     

The determination of a one year mean sea surface height track from Geosat altimeter data and ocean variability implications

One year (November 1986 to October 1987) of Geosat altimeter data with improved orbits produced at The Ohio State University has been used to define sea surface heights for 22 ERM and one year averaged Geosat track. All sea surface heights were referenced to the single reference track through the application of geoid gradient corrections. The root mean square (RMS) gradient correction was on the order of ±1 cm although it could reach 20 cm with data points in trench areas. 10 values used to form the mean were considered. Although this study was initially driven by a need for a good reference sea surface for geodetic applications the formation of the reference track yields information on the variability of the ocean surface in the first year of the Geosat ERM. The RMS point variability was ± 12.6 cm with only a very small number of values exceeding 50 cm when a depth editing criteria was used. Global plots of the sea surface variability clearly reveal the major ocean currents and their variations in position in the year. Examination of the 1° × 1° averaged sea surface height variations show average and maximum variability values as follows: Gulf Stream (29 and 50 cm); Kurshio Current (24 and 49 cm); Agulhas Current (24 and 52 cm) and the Gulf of Mexico (18 and 31 cm). These magnitudes may be dependent on the radial orbit correction procedure. To investigate this effect sea slope variations were also computed. These results also showed clear current structures but also high frequency gravity field information despite efforts to average out such information. The data described in the paper is available from the authors for numerous other studies, some of which are suggested in the paper.     

GPS measurements of ocean loading and its impact on zenith tropospheric delay estimates: a case study in Brittany, France

 The results from a global positioning system (GPS) experiment carried out in Brittany, France, in October 1999, aimed at measuring crustal displacements caused by ocean loading and quantifying their effects on GPS-derived tropospheric delay estimates, are presented. The loading effect in the vertical and horizontal position time series is identified, however with significant disagreement in amplitude compared to ocean loading model predictions. It is shown that these amplitude misfits result from spatial tropospheric heterogeneities not accounted for in the data processing. The effect of ocean loading on GPS-derived zenith total delay (ZTD) estimates is investigated and a scaling factor of 4.4 between ZTD and station height for a 10° elevation cut-off angle is found (i.e. a 4.4-cm station height error would map into a 1-cm ZTD error). Consequently, unmodeled ocean loading effects map into significant errors in ZTD estimates and ocean loading modeling must be properly implemented when estimating ZTD parameters from GPS data for meteorological applications. Ocean loading effects must be known with an accuracy of better than 3 cm in order to meet the accuracy requirements of meteorological and climatological applications of GPS-derived precipitable water vapor. Received: 16 July 2001 / Accepted: 25 April 2002 Acknowledgments. The authors are grateful to H.G. Scherneck for fruitful discussions and for his help with the ocean loading calculations. They thank H. Vedel for making the HIRLAM data available; D. Jerett for helpful discussions; and the city of Rostrenen, the Laboratoire d'Océanographie of Concarneau, and the Institut de Protection et de S?reté Nucléaire (BERSSIN) for their support during the GPS measurement campaign. Reviews by C.K. Shum and two anonymous referees significantly improved this paper. This work was carried out in the framework of the MAGIC project (http://www.acri.fr/magic), funded by the European Commission, Environment and Climate Program (EC Contract ENV4-CT98–0745). Correspondence to: E. Calais, Department of Earth and Atmospheric Sciences, Purdue University, West Lafayette, IN 47907-1397, USA. e-mail: ecalais@purdue.edu Tel. : +1-765-496-2915; Fax:+1-765-496-1210     

An Effective Model to Retrieve Soil Moisture from L- and C-Band SAR Data

This study investigated an appropriate method for soil moisture retrieval from radar images and coincident ground measurements acquired over bare soil and sparsely vegetated regions. The adopted approach based on a single scattering integral equation method (IEM) was developed to establish the relationship between backscatter coefficient and surface soil parameters including volumetric soil moisture content and surface roughness. The performance of IEM in 0–7.6 cm is better than that in 0–20 cm. Moreover, IEM can simulate correctly the backscatter coefficients only for the root mean square (RMS) height s < 1.5 cm at C-band and s < 2.5 cm at L-band by using an exponential correlation function and for s > 1.5 cm at C-band and s > 2.5 cm at L-band by using Gaussian function. However, due to the difficulties involved in the parameterization of soil surface roughness, the estimated accuracy is not satisfactory for the inversion of IEM. This paper used a combined roughness parameter and Fresnel reflection coefficient to develop an empirical model. Simulations were performed to support experimental results and to highlight soil moisture content and surface roughness effects in different polarizations. Results showed that a good agreement was found between the IEM simulations and the SAR measurements over a wide range of soil moisture and surface roughness characteristics. The model had a significant operational advantage in soil moisture retrieval. The correlation coefficients were 77.03 % at L-band and 81.45 % at C-band with the RMSEs of 0.515 and 0.4996 dB, respectively. Additionally, this work offered insight into the required application accuracy of soil moisture retrieval at a large area of arid regions.     

Thermosteric Effects on Interannual and Long-term Global Mean Sea Level Changes

We investigate global mean sea level (MSL) changes and different geophysical contributions at interannual and long-term (decadal) time-scales. Thermosteric effects of global MSL changes are estimated from ocean temperature anomaly data for the period 1955–2003 from the World Ocean Database 2001 (WOD01), plus additional data processed through June 2004. Estimates based on WOD01 show significant differences to previously published results based on similar temperature anomaly data from the World Ocean Database 1998 (WOD98), especially during the period overlapping with the TOPEX/Poseidon satellite altimeter mission. During this period (1993–2004), the WOD01-estimated thermosteric contribution of global MSL change is less than half of the estimate from WOD98 (1.3 ± 0.1 vs. 3.0 ± 0.6 mm/year), as compared to the rate of 2.6 ± 0.06 mm/year observed by satellite altimeters. The larger uncertainty in ocean temperature profiles and incomplete data collection in WOD98, especially in the later years (1997 and 1998) appear to be the major error sources to the overestimated steric effects by WOD98. During the entire 50-year period, the steric effect on global MSL change amounts to about 0.34–0.39 (±0.05) mm/year. Strong interannual and decadal variability exists in estimated thermosteric contributions to the global MSL change, and (surprisingly) the thermosteric effect does not show any pronounced contribution to the strong interannual variability during the 1997/1998 El Niño/La Niña event. Our analysis based on the National Centers for Environmental Prediction reanalysis atmospheric model and the National Oceanic and Atmospheric Administration Climate Prediction Center global land data assimilation system indicates that atmospheric water vapor and terrestrial water storage changes show strong interannual variability that is well correlated with observed global MSL change, and could have significant effects on interannual global MSL changes.     

The M 2 ocean tide loading wave in Alaska: vertical and horizontal displacements, modelled and observed

Crustal deformations caused by surface load due to ocean tides are strongly dependent on the surface load closest to the observation site. In order to correctly model this ocean loading effect near irregular coastal areas, a high-resolution coastline is required. A test is carried out using two GPS sites located in Alaska, where the ocean tide loading effect is large and consequently observed easily by relative positioning with GPS. The selected sites are Fair (Fairbanks) and Chi3 (located on an island that separates Prince William Sound from the Gulf of Alaska). Processing of hourly baseline solutions between Fair and Chi3 over a period of 49 days yields a significant ocean tide loading effect. The data are processed using different strategies for the tropospheric delay correction. However, the best results are obtained when 1-h ZTD (Zenith Tropospheric Delay) parameters for hourly solutions are used. In this case ocean tide loading is not absorbed into the ZTD parameters. Hence, ocean tide loading can be well resolved in the GPS data analysis. In addition, the M ₂ ocean tide wave in the Gulf of Alaska has a very large amplitude. Although the horizontal M ₂ ocean tide loading amplitude in general is only about 1/4 of the vertical M ₂ ocean tide loading amplitude, the differential horizontal M ₂ ocean tide loading displacements are nevertheless measurable using differential GPS (DGPS). When using the GOT99.2 ocean tide model and taking the coastal structure into account, the predicted differential vertical M ₂ amplitude and Greenwich phase lag due to ocean tide loading are 19.3 mm and 110.2 degrees respectively, while GPS measurements yield 21.3 ± 1.0 mm and 99.7±2.8 degrees. Similarly, the predicted differential horizontal M ₂ amplitude and Greenwich phase lag (in the north–south direction) are 4.5 mm and –77.0 degrees, while GPS yields 5.4 ± 0.3 mm and –106.3±3.3 degrees. Only the north-south component of the differential horizontal M ₂ ocean tide loading wave is considered, because the east–west component is too small for the processed baseline and not detectable using DGPS.     

Generation and Validation of two Day Composite Wind Fields from Oceansat-2 Scatterometer

The Oceansat-2 scatterometer (OSCAT) of the Indian Space Research Organization (ISRO), provides surface wind speed and direction with a spatial resolution of 50 km × 50 km. With a revisit time of 2 days it had provided ocean surface wind vectors over the global oceans. In the present work, an attempt has been made to generate two day composite of OSCAT wind vectors using Data-Interpolating Variational Analysis (DIVA) and compare them with daily composite winds to check how better is the two day composites in comparison to daily composites. The daily and two days composite wind vectors of zonal (U) and meridional (V) components have been validated with wind measurements from in situ buoys and Advanced Scatterometer (ASCAT) for the year 2012 over the tropical Indian Ocean region. The statistical comparison with the in situ measurements and ASCAT has shown that the two-day OSCAT wind composites are slightly better than the daily composite winds. The improvement in the statistics can be attributed to the use of ascending and descending passes pertaining to two days which results in fewer gaps between passes, thereby reducing the interpolation errors.     

Analysis of elevation changes in relation to surface characteristics for six glaciers in Northern Labrador,Canada using advanced space-borne thermal emission and reflection radiometer imagery

Advanced space-borne thermal emission and reflection radiometer imagery and Digital Elevation Models were used to analyse surface elevation changes of six glaciers in Northern Labrador. Results indicate an average surface thinning of0.94 ± 0.49 m y^?1 (water equivalent) between 2000 and 2009. Three glaciers had an average elevation change of ?1.16 ± 0.55 m y^?1 (water equivalent) whichis three times the thinning rate found in a study from 1981 to 1983 ?0.36 ± 0.10 m y^?1 water equivalent). Analysis of surface characteristics in relation to elevation changes shows expected results of rapid thinning in bare ice areas and near zero change in accumulation areas. Debris covered areas of three glaciers show expected results of moderate thinning, but three other glaciers indicate high rates of thinning. Variability in thinning rates suggests possible influences in the type ofdebris and/or variations in climate such as increased rainfall.     

Short note: the experimental geopotential model XGM2016

As a precursor study for the upcoming combined Earth Gravitational Model 2020 (EGM2020), the Experimental Gravity Field Model XGM2016, parameterized as a spherical harmonic series up to degree and order 719, is computed. XGM2016 shares the same combination methodology as its predecessor model GOCO05c (Fecher et al. in Surv Geophys 38(3): 571–590, 2017. doi: 10.1007/s10712-016-9406-y). The main difference between these models is that XGM2016 is supported by an improved terrestrial data set of \(15^\prime \times 15^\prime \) gravity anomaly area-means provided by the United States National Geospatial-Intelligence Agency (NGA), resulting in significant upgrades compared to existing combined gravity field models, especially in continental areas such as South America, Africa, parts of Asia, and Antarctica. A combination strategy of relative regional weighting provides for improved performance in near-coastal ocean regions, including regions where the altimetric data are mostly unchanged from previous models. Comparing cumulative height anomalies, from both EGM2008 and XGM2016 at degree/order 719, yields differences of 26 cm in Africa and 40 cm in South America. These differences result from including additional information of satellite data, as well as from the improved ground data in these regions. XGM2016 also yields a smoother Mean Dynamic Topography with significantly reduced artifacts, which indicates an improved modeling of the ocean areas.     

Initial results of precise orbit and clock determination for COMPASS navigation satellite system

The development of the COMPASS satellite system is introduced, and the regional tracking network and data availability are described. The precise orbit determination strategy of COMPASS satellites is presented. Data of June 2012 are processed. The obtained orbits are evaluated by analysis of post-fit residuals, orbit overlap comparison and SLR (satellite laser ranging) validation. The RMS (root mean square) values of post-fit residuals for one month’s data are smaller than 2.0 cm for ionosphere-free phase measurements and 2.6 m for ionosphere-free code observations. The 48-h orbit overlap comparison shows that the RMS values of differences in the radial component are much smaller than 10 cm and those of the cross-track component are smaller than 20 cm. The SLR validation shows that the overall RMS of observed minus computed residuals is 68.5 cm for G01 and 10.8 cm for I03. The static and kinematic PPP solutions are produced to further evaluate the accuracy of COMPASS orbit and clock products. The static daily COMPASS PPP solutions achieve an accuracy of better than 1 cm in horizontal and 3 cm in vertical. The accuracy of the COMPASS kinematic PPP solutions is within 1–2 cm in the horizontal and 4–7 cm in the vertical. In addition, we find that the COMPASS kinematic solutions are generally better than the GPS ones for the selected location. Furthermore, the COMPASS/GPS combinations significantly improve the accuracy of GPS only PPP solutions. The RMS values are basically smaller than 1 cm in the horizontal components and 3–4 cm in the vertical component.     

Applying GNSS and DTM Technologies to Monitor the Ice Balance of the Horcones Inferior Glacier, Aconcagua Region, Argentina

This work provides results and analysis of the behavior in a reconstituted debris covered glacier, Horcones Inferior (HIG). This glacier is located at 32° 41′ S and 69° 57′ W in the Aconcagua region, Mendoza, Argentina. The HIG has experienced surges events in 1984 and 2003. GNSS techniques and Digital Terrain Models (DTMs) derived from ASTER optical imagery, were used to detect movements and altimetric changes of the HIG in the 2001–2012 period. Based on a GNSS semi - continuous station, N and E mean velocities of 0.9 cm/d and 1.1 cm/d and in U direction were 1.1 cm/d and 1.5 cm/d were obtained respectively in 2009–2010 period. Additionally, GNSS profiles showed a velocity range between 0.5 cm/d and 2.8 cm/d. Finally, through a DTM comparison the altimetric change before and after the last surge in 2003 was obtained. The applied techniques allowed accurate and reliable change detection of the HIG.     

利用卫星重力探测南极冰盖质量变化

针对不同GIA模型在南极地区的改正差异较大,分析了采用不同GIA模型分析得到的南极冰盖融化误差值,研究了南极冰盖质量变化加速度。研究结果表明,扣除IJ05模型影响后,整个南极冰盖质量以(-81.5±4.2)Gt/a的速度和(-12.3±6.5)Gt/a2的加速度融化,对海平面上升贡献分别为(0.23±0.01)mm/a和(0.03±0.02)mm/a2。西南极处于加速融化状态,速度由2003—2008年间的-37.7Gt/a增加至2009—2013年的-156.0Gt/a。东南极冰盖在2009年前处于稳定,2009年后逐渐积累,且积累速度不断增加。2003-2013期间内东南极,西南极冰盖质量变化速度及加速度分别为(28±13.4)Gt/a、(9.8±2.8)Gt/a2、-(108.1±3.5)Gt/a和-(21.1±2.9)Gt/a2,南极冰盖融化主要来自西南极。南极冰盖周年变化影响较强,每年10月左右其质量变化振幅达到最大,半周年变化影响最大在西南极,S2潮波振幅在Ronne冰架附近较大,振幅最大值达到25mm。     

Bias in GRACE estimates of ice mass change due to accompanying sea-level change

Observations of spatio-temporal variations in the geopotential using the GRACE satellites have been used to estimate recent mass fluxes from polar ice sheets and glaciers. However, these estimates have not considered the potential bias associated with the migration of water that accompanies the ice melt. This migration is driven by the diminished gravitational attraction of the melting ice reservoir, and this migration, as well as the crustal loading it induces, will contribute to the observed geopotential anomaly. The extent to which this contribution contaminates the ice mass flux estimates depends on how far the smoothing filters applied to the GRACE data extend beyond the ice margins into the ocean. Using the Antarctic Peninsula as a case study, we estimate the magnitude of this bias for a range of melt areas and Gaussian smoothing filter radii. We conclude that GRACE estimates of ice mass loss over the Antarctic Peninsula are systematically overestimating the loss by up to 10  $\%$ for filter radii of less than 500 km.