Guide for Applying ICESat Inter-campaign Bias Corrections (ICBs)

Using ICESat GLAS altimetry data to assess temporal elevation changes requires consideration of long-period variations in range measurements that manifest themselves as constant elevation biases over each of the mission campaigns.  These "inter-campaign biases" (ICBs) have been evaluated by several research groups using a variety of approaches. In this document, NSIDC lists several available ICB corrections and provides self-reported information about how they were calculated.  No ICESat ICB correction is endorsed by the ICESat Science Team, NASA, or NSIDC. Users may decide whether or not to apply ICBs to their data and, if so, which ICB correction set is most appropriate for their particular application.

In general, all the ICB corrections have been determined by measuring apparent elevation change over a nearly-unchanging 'reference' surface spanning a wide area that includes a large number of GLAS laser shots for most or all of the ICESat campaigns. The reference surface is assessed with either an independent measurement of that surface's elevation through time or a well-founded assumption of near-zero change.

Two important corrections for ICESat data were recognized and assessed late in the mission that had a large effect on the apparent ICBs: 1) the range correction for GLAS detector saturation effects on the waveform (the 'saturation correction' – see NSIDC Saturation Correction Guidance; and 2) a data processing error arising from two different methods of determining the reference point in the outgoing and received GLAS waveform (geoid minus centroid, or 'G-C' correction; Borsa et al. 2014). Products through Release 633 had been incorrectly calculated from the centroid (amplitude-weighted center of leading and trailing edge thresholds) of the transmitted laser pulse to the center of the Gaussian fit of the return pulse (Zwally, 2013; Borsa et al., 2014). Applying the range correction for the G-C offset improved the range precision by 1.7 cm to <2 cm, and changed (but did not eliminate) ICBs.

Although the net effect of using ice-sheet data without the G-C corrections applied is very small if compatible bias corrections are applied (that is, if the ICBs were constructed so that the effect of G-C error is incorporated in them, unknowingly), the error in trends can be significant. Errors in an elevation trend without G-C correction have been assessed at –1.29 cm/yr for the L2a to L3j campaigns.

Users of ICESat data should note that in Release 634, the G-C correction is already applied on a shot-by-shot basis. [The revised saturation correction is available but not applied to the data].

Evaluation of ICBs

Following completion of the ICESat mission in 2009, ICB's were estimated by various investigators using differing reference surfaces, data releases, corrections, and other variations in methods. A summary of various ICBs is given in Scambos and Shuman (2016); however this did not include a full evaluation of the ICB methodologies. The purpose of this document is to discuss some evaluation criteria to five recent ICB studies, presented in spreadsheets in NSIDC Criteria for Evaluation of ICESat Inter-campaign Biases. These criteria provide a basis for evaluation of other ICBs previously published or not yet produced.

The six ICB studies we highlight are: 1) Borsa et al., 2017 (submitted): Stable terrestrial reference surface at the Salar de Uyuni, Bolivia with Rel 634; 2) Felikson et al., 2017 (in review; personal communication and Urban et al., 2015): Global ocean mean sea surface with Rel 634; 3) Zwally et al., 2015: Ocean level within leads in Arctic and Antarctic Sea Ice with Rel 634; 4) Richter et al., 2014: Ice surface above subglacial Lake Vostok, Antarctica with Rel 633 and G-C applied, 5) Hofton et al., 2013, Ice surface in East Antarctica (two regions) with Rel 633;  6) Schröder et al., Validation of satellite altimetry by kinematic GNSS in central East Antarctica.  Evaluation details for each of the six methods are given below.

ICB corrections for these studies are shown in Table 1.

Sources of intercampaign (IC) biases (cm)

Table 1. Published ICESat Intercampaign Bias Assessments

Hofton and others [2013]1

Richter and others [2014]1,2

Urban and others [2013]1,3

Zwally and others [2015]4

Schröder and others [2017]

Borsa and others (in prep) [2017]

ICESat Campaign

(86°S orbit ring)

(EA Divide)

(Vostok area)



(Polar Oceans)

(Vostok region)

(Salar de Uyuni)


-1.1 ±1.6

-5.1 ±5.5 

-2.8 ±1.4

-4.7 ±3.6



-3.2 ±1.0



+3.2 ±1.8

+0.4 ±5.5 

+2.1 ±1.2

+4.8 ±1.3



+3.6 ±0.9

+4.9 ±3.9


-1.0 ±4.3

-1.6 ±4.3

-0.7 ±1.1

+0.5 ±1.3



-0.1 ±0.9

+0.4 ±3.0


+7.1 ±5.5

+2.4 ±5.7

+6.3 ±1.0

+4.1 ±0.9



+5.2 ±0.8

+5.9 ±5.8


-2.7 ±4.0

-6.7 ±3.9

-3.2 ±0.9

+1.0 ±1.3



-3.1 ±0.9

+0.6 ±3.7


-2.5 ±3.1

-5.1 ±4.6

-4.0 ±0.9

+0.2 ±1.1



-1.7 ±0.8

+5.4 ±9.9


-3.6 ±3.9

-6.2 ±4.7

-2.9 ±0.8

+1.0 ±1.5



-2.1 ±0.7

+1.8 ±3.0


+1.6 ±2.5

-3.2 ±4.2

+0.4 ±0.8

+0.4 ±1.0



+1.9 ±0.7

-0.6 ±3.9


+1.8 ±1.6

-1.1 ±4.5

-1.1 ±0.8

+0.3 ±0.7



+0.2 ±0.7

-1.7 ±6.9


-2.2 ±2.3

-4.9 ±4.9

-1.7 ±0.8

+0.1 ±0.9



-1.0 ±0.6

-2.4 ±8.8


+3.2 ±0.8

-1.1 ±4.9

+2.4 ±0.8

+1.9 ±0.7



+2.4 ±0.6

-2.3 ±6.3


+1.2 ±1.8

+0.1 ±3.0

-1.0 ±0.8

+1.2 ±0.9



0.0 ±0.6

-0.4 ±9.5


  0.0 ±3.2

  0.0 ±3.3 

 0.0 ±0.9

  0.0 ±0.9



0.0 ±0.6

  0.0 ±3.4


+3.5 ±2.3

+1.3 ±5.2 

+3.1 ±1.0

-1.2 ±1.4



+2.5 ±0.6

-1.3 ±6.3


+6.2 ±2.6

+3.1 ±3.4  

+4.3 ±1.2

-0.7 ±2.0



+3.7 ±0.6

+2.8 ±3.6


+7.7 ±1.5

+7.3 ±4.5 

+5.0 ±1.2

+5.7 ±1.7



+5.9 ±0.6

+3.7 ±4.4


+14.7 ±3.0

+13.9 ±4.9 

+5.6 ±1.3

+11.2 ±7.3



+8.7 ±0.6

+6.6 ±9.9


+7.4 ±2.8

+4.2 ±4.4 

+6.0 ±1.3

+4.9 ±1.2



+6.1 ±0.6

+1.2 ±7.6

All published IC biases presented here have been adjusted to be relative to the Laser 3I campaign (zero). Uncertainties (±) are included if available.
1These data were based on Release 633 but include the Gaussian minus Centroid (G-C) correction. All other analyses used Release 634 data. See the original references for additional details.
2Ewart and others [2012] was updated for Helm and others [2014] and then Richter and others [2014] by including the G-C correction (see their Table 3, ΔHGC column). L2A value excludes the 8-day data early in that campaign.
3These bias numbers have been updated by Urban, pers. comm. [2015] as published in Scambos and Shuman [2016].
4From Zwally and others [2015], DSL equals the D value plus an Envisat-derived correction for sea level variations.

A critically important criteria is the correction for vertical movement of the reference surface, listed in line item 6 of the evaluation tables spreadsheet. This requires either an independent method of measurement or a well-justified assumption of near-zero vertical movement of the reference over the ICESat campaign period. We review this for the five highlighted ICB approaches here.

Borsa et al., 2017 (dry lake bed) use GPS measurements on a dry salt lake to establish that the surface exhibits little seasonal deformation and that the mean interannual deformation is near zero. However, the total number of shots over the lake bed is far fewer than the other assessments, raising concerns that the few tracks involved may not represent other regions well. Urban et al. 2015 (mean ocean surface) apply a constant 3 mm/yr correction for global sea level rise, but do not correct for concurrent variations in ocean dynamic topography, which can be significant (e.g., Rio and Hernandez, 2004). These variations may be on the order of 10 cm/yr in some ocean regions; however, the study assumes that these dynamic effects have a near-zero mean over the global ocean. Wave effects are similarly not corrected, but are assumed to have near-zero mean impact because of the very large averaging area and the smaller footprint and waveform approach to elevation determination for ICESat (relative to radar altimetry). This method includes only unsaturated laser returns over the ocean, so any saturation correction effects are minimal.

In Zwally et al., 2015 (polar ocean leads), corrections are made for ocean level changes by using concurrent Envisat radar altimetry of the same surface in leads and polynyas within the Arctic and Antarctic sea ice pack. A similar approach is routinely used for detection of leads and measurement of sea ice freeboard (e.g., Kwok et al., 2007), but concerns have been raised about its accuracy for measurement of an absolute sea level surface (Kern and Spreen, 2015) owing to the sensitivity to parameters used to subset the leads-only surface. These concerns and others were discussed in Scambos and Shuman, 2016 and Zwally et al., 2016. The measured reflectivity over leads and polynyas in Zwally et al. is 0.42, and 0.53 on adjacent sea ice. Effects from saturation should be small. This approach should also account for seasonal and inter-annual temporal variations and for spatial variations in ocean dynamic typography, which can be significant (e.g., Giles et al., 2012).  Longterm changes in global sea level due to changes in the gravitational field rise also vary regionally, particularly near polar ice sheets, and the method makes an adjustment intended to correct for that as well. Further, leads and small polynyas have generally low wave heights.

Richter et al. 2014 (Antarctic Lake Vostok) rely on a repeat measurements of a GPS stake network, and kinematic profile resurveys, to conclude near-zero surface elevation change and hydrostatic equilibrium (Ewert et al., 2012) for the snow surface above Sub-Glacial Lake Vostok and its surroundings in East Antarctica. However, the accuracy of those measurements and conclusions have been disputed (Zwally et al., 2015; Zwally et al., 2016a; Zwally et al., 2016b). Richter et al. (2016) discussed the disputed aspects, and further work with continued GPS measurements still indicates near-zero vertical motion for the region (Schroeder et al., 2016). Some (Zwally et al., 2015) but not all (Helm et al., 2014; McMillan et al., 2014) radar altimetry analysis of the region shows elevation increase in the area of >1 cm/yr. While the saturation correction was applied in the Richter et al. analysis (and related earlier work by Ewert et al., Gunther et al.), there are remaining concerns that residual effects of saturation even after correction could remain. Borsa et al., 2016, have concluded that these are negligible.

Hofton et al. 2013 (stability of East Antarctic plateau) calculate ICBs beginning with the assumption that two broad regions in East Antarctica have near-zero elevation change over the ICESat campaign periods, and apply a saturation correction to Rel. 633 with G-C applied, and use cross-over analysis to extract ICB numbers for the different campaigns. They reference the elevations collected. While this assumption is supported by other studies, Zwally et al. 2015 report a significant elevation increase for the region.

For further details on each of these methods, see the NSIDC Criteria for Evaluation of ICESat Inter-campaign Biases assessment tables.

Brief Review of All Past Published ICB assessments

Below we list a literature review of all efforts to assess absolute or relative campaign biases for the period 2005-2017, in reverse time order. Note that any table or figure number referenced below refers to entries in the original publication. This listing compiles the known ICB assessments for the user to refer to and investigate, but makes no evaluation of accuracy, quality, or appropriateness for a particular application. See these links for a discussion of ICESat Data Releases:

Borsa et al., 2017 (submitted)
Intercampaign biases for saturation-corrected Release 634 ICESat data are determined by differencing ICESat elevations from a GPS-derived digital elevation model (DEM) over the Salar de Uyuni, Bolivia.  The survey area covered by the DEM is 45 x 54 km, has no visible relief, and is stable at the cm level on annual timescales. Aside from transformations into the GPS frame of reference, no adjustments are made to the ICESat data.

Felikson et al., 2017 (doi: 10.1109/TGRS.2017.2709303) (in press)
Intercampaign biases are estimated using release 634 GLAS data over the global oceans (Urban and Schutz, 2005). A bias is estimated for each campaign as the mean difference between ICESat elevation measurements and a reference mean sea surface, including a constant sea level rise rate of 3 mm/year, over the global oceans.

Urban et al. 2015 (doi: 10.1017/jog.2016.59)
ICESat biases are determined using Release 634 data, following the methodology of Urban and Schutz (2005), i.e., by taking the difference between ICESat elevations and mean global sea surfaces. A constant mean sea level rise rate of 0.3 cm/year is also included in the biases. ICESat bias values are reported in Scambos and Shuman (2016), Table 1. The values are updated from previous work by Urban et al. (2012).

Zwally et al. 2015 (doi: 10.3189/2015JoG15J071)
ICESat biases were determined by comparing sea surface heights over open water and thin ice in leads and polynyas in the Arctic and Antarctic sea-ice packs with ICESat data from Release 634.  Additionally, adjustments were made for sea surface height variations measured by Envisat radar altimetry, and gathered concurrently with ICESat.  Values are found in Table 6, for both the ICESat-measured and Envisat adjusted biases.

Helm et al. 2014 (doi: 10.5194/tc-8-1539-2014)
ICESat biases are determined following the work of Ewert et al. (2012), using Release 633 data (with G-C?). The values can be found in Table 1.

Gunter et al. 2014 (doi: 10.5194/tc-8-743-2014)
ICESat biases were determined following the methodology of Gunter et al. (2009) and Riva et al. (2009), by choosing a low-precipitation region in East Antarctica. This was defined as all high-elevation regions with less than 21.9 mm Equivalent Water Height/year snow accumulation. The value was set by selecting a continuous low-precipitation zone isolated from areas of steep topography. Repeat-track ICESat footprints from release 633 having at least 80% laser spot overlap in the low-precipitation zone were used to generate biases. Additionally, an firn density model, or FDM, (based on RACMO2; Lenaerts et al., 2012) was used to adjust the median elevation change in the low-precipitation zone, accounting for firn height changes. The results are found in Table 1.

Richter et al. 2014 (doi: 10.1002/2014JF003228)
Relative ICESat biases were determined by a regional crossover adjustment (7559 crossovers) over the ice surface within the area of subglacial Lake Vostok, East Antarctica. Repeated GNSS observations at 56 sites distributed all over the lake area (2001-2013) as well as repeated kinematic GNSS profiling around Vostok station (2001-2013) have shown independently and consistently that the ice surface elevation above the lake has not changed significantly throughout the ICESat mission. Moreover, the ice above the lake has been shown to be in hydrostatic equilibrium (Ewert et al. 2012) and this allows them to extrapolate the stability of the ice surface from the GNSS observation sites over the entire lake area. Data used were from Release 633, with and without G-C correction applied. Results are shown in Table 3.

Hofton et al. 2013 (doi: 10.1002/2013GL057652)
Two sets of ICESat biases were independently determined from comparison of inter-mission (LVIS to ICESat) and intra-mission (ICESat to ICESat) crossovers in Antarctica. For both estimates, ICB's were estimated using only data from areas where spatial and temporal variations due to climate related surface processes were minimal. For ICB's estimated from inter-mission crossovers this was the area along 86S between 110E and 155E. For ICB's estimated from intra-mission crossovers, this was an area of 393,765km2 of East Antarctica.  For both comparisons, the elevation differences between near-coincident footprints were obtained and ICB's calculated from the average value in a cell. The ICB values are determined from Release 633 data, corrected for G-C and saturation effects, and are adjusted for the effects of climate related processes, and can be found in Table 2 for each study region. The comparison involving LVIS had long wavelength topography removed from the elevations prior to extracting differences between overlapping footprints. The author's preferred solution for the ICESat ICB's are those from the 86S study region using intersensor differences from which long-wavelength topography is removed and to which corrections for ICESat elevation errors and elevation changes due to climate-related surface processes.

For citations from earlier work, see: Inter-campaign Bias Corrections Citations and References

See Also: NSIDC Criteria for Evaluation of ICESat Inter-campaign Biases