Abstract
Global Navigation Satellite System positioning of gravity surveys permits geoid computation via Hotine’s integral. A suite of modifications is presented so that the user can tune the relative contributions of truncation and data errors in a combined solution for a regional geoid model from gravity disturbances.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Motivation
Most gravity surveys are now positioned using Global Navigation Satellite Systems (GNSSs), which deliver the 3D geodetic coordinates of each gravity observation after corrections for offsets between the GNSS antenna reference point and the gravity sensor. These can then be used to compute the gravity disturbance
where \(\gamma _S\) is normal gravity at the same 3D position as the gravity observation \(g_S\) that has been reduced to gravity datum and corrected for instrumental drift and tides. To a second-order approximation near the Earth’s surface, \(\gamma _S\) can be computed analytically from the GNSS-derived ellipsoidal height h and geodetic latitude \(\varphi \) of the observation point S using (e.g., Heiskanen and Moritz 1967, p 79)
with
where f is the geometrical flattening of the normal ellipsoid, m is the ratio of gravitational and centrifugal accelerations at the equator of the normal ellipsoid, a is the equatorial radius of the normal ellipsoid, k is the normal gravity constant, \(\gamma _{e}\) is normal gravity acceleration at the equator, and e is the first numerical eccentricity of the normal ellipsoid. Numerical values of these parameters for the GRS80 normal ellipsoid are given in Moritz (1980) and reprinted in the Geodesist’s Handbook.
While the downward continuation of gravity disturbances is beyond the scope of this article, it nevertheless requires that: (1) \(\delta g_S\) are downward-continued to the geoid to give \(\delta g\) before being convolved with Hotine’s kernel or some modification thereof; and (2) the gravity disturbances must correspond to a harmonic disturbing potential in the solution domain. The availability of these downward-continued gravity disturbances \(\delta g\) then allows for computation of the geoid N by Hotine’s integral (Hotine 1969, Chap 29), which is a solution to a fixed or the second boundary-value problem of potential theory in spherical approximation (e.g., Heiskanen and Moritz 1967, p 36).
One advantage of using gravity disturbances over gravity anomalies to compute the geoid is that they are not adversely affected by, e.g., uncertain or ambiguous realisations of vertical datums and their associated height systems. In Australia, for instance, the vertical datum contains a confirmed tilt with respect to the geoid (Featherstone and Filmer 2012), regional distortions [e.g., Featherstone et al. (2011) and the citations therein], and uses a height system that is not compatible with the geoid (Filmer et al. 2010). These will cause errors in the computed terrestrial gravity anomalies (cf. Heck 1990) that can propagate into the combined geoid solution unless modelled and/or filtered out (cf. Vaníček and Featherstone 1998; Wang et al. 2011; Featherstone et al. 2011).
Modern regional geoid models are often computed from an Earth gravitational model (EGM) augmented by one, more or all of land, airborne, ship-borne and altimeter-derived marine gravimetry, depending on data availability and the region of interest. GNSS is used to coordinate and navigate ship-borne gravity surveys, which results in better Eötvös corrections and horizontal locations so that crossover adjustments are more effective (cf. Wessel and Watts 1988). As the mean sea surface departs from the geoid by up to \({\sim }2\) m, account also has to be made for mean dynamic topography (MDT) in the derivation of gravity disturbances from satellite altimetry (Zhang 1998). Land and airborne gravity surveys are also coordinated with GNSS, thus facilitating the direct computation of gravity disturbances. A prior geoid model would be needed to determine gravity anomalies from GNSS-coordinated gravity surveys, resulting in a circular argument (cf. Vaníček et al. 1992), and reinforcing the benefit of directly using gravity disturbances in Hotine’s integral.
Hotine’s integral, or its inverse, has been used for: (1) geoid determination from GNNS-positioned airborne gravimetry (e.g., Schwarz and Li 1996; Kearsley et al. 1998; Forsberg et al. 2000; Novák and Heck 2002; Novák 2003; Novák et al. 2003; Alberts and Klees 2004; Serpas and Jekeli 2005; Sjöberg and Eshagh 2009) or land gravimetry (e.g., Kirby 2003), (2) marine gravity field and MDT estimation from satellite radar altimetry (e.g., Rapp 1980; Zhang and Blais 1993; Rapp and Wang 1994; Zhang and Sideris 1996; Zhang 1998), and (3) global Earth and planetary gravity field modelling (e.g., Sjöberg 1989; Barriot and Balmino 1992).
However, the spatial coverage of GNSS-coordinated gravity data is currently limited, so there is a need to modify Hotine’s integral to reduce the truncation error that results from the omission of gravity disturbances in the far zones beyond the area of interest. Admittedly, this truncation error can be reduced by the inclusion of a high-degree EGM (Sect. 2.2). However, the kernel modification and cap radius can be used additionally as a filter to tune the relative data contributions (cf. Vaníček and Featherstone 1998; Kern et al. 2003; Featherstone 2003a). It is this property that will be emphasised more than only reduction of the truncation error.
Early modifications to Stokes’s kernel (e.g., Molodensky et al. 1962) were formulated to reduce the truncation error alone because of limited spatial coverage of terrestrial gravity anomalies at that time. Many of the subsequent modifications also consider EGMs (Appendix A). The limited spatial coverage of gravity disturbances coordinated by GNSS is probably the same now as it was in the years soon after the portable gravimeter was developed, so the motivation for modifications to Hotine’s kernel is similar now to as it was then for Stokes’s kernel. However, the advent of new EGMs derived from dedicated satellite gravimetry (e.g., Pail et al. 2011) and EGM2008 (Pavlis et al. 2012) has changed the approaches to regional geoid computation for a couple of reasons: (1) the truncation and approximation errors are lessened considerably when a very high-degree EGM is used (Sect. 2.2), and (2) a satellite-derived EGM that is more reliable in the low and medium degrees allows for higher degrees of modification so as to place more reliance on the geoid derived from satellite gravimetry (e.g., Sects. 3.1 and 5).
Mostly in analogy to those already proposed and used for Stokes’s integral and gravity anomalies (Appendix A), this article will present: (1) deterministic modifications to Hotine’s integral, where the user can control the errors in some prescribed ways; (2) stochastic modifications to Hotine’s integral, where error spectra are embedded in an attempt to control the balance amongst data and/or truncation errors; and (3) their band-limited hybrid combinations, where they are combined according to the perceived relative benefits of each. For instance, the ‘user’ may have a good understanding of the data errors in parts of the geopotential spectrum so can use a stochastic modifier in those bands, but say use a deterministic modifier in other bands (cf. Sect. 5). The options are many, but the hybrid combinations do provide much more flexibility for the ‘user’.
Previous authors who have investigated modifications to Hotine’s kernel comprise Jekeli (1979, 1980b); Guan and Li (1990); Sjöberg and Nord (1992); Vaníček et al. (1992); Zhang and Blais (1993); Zhang (1998); Novák (2003); Novák et al. (2003); Alberts and Klees (2004); and Sjöberg and Eshagh (2009), so their results will only be summarised as part of the review component of this paper. Importantly, most of these authors conclude that Hotine’s integral with gravity disturbances can be superior to Stokes’s integral with gravity anomalies. The other modifications to Hotine’s kernel presented herein will be based on adaptations of the principles previously applied to Stokes’s kernel, as well as some new formulations that can be applied back to Stokes’s or other kernels.
2 Basics
2.1 Spherical Hotine integral
In terms of spherical polar coordinates of spherical distance \(\psi \) and azimuth \(\alpha \) centred on each computation point, Hotine’s integral in spherical approximation reads
and the spherical Hotine kernel is
where \(P_n (\cos \psi )\) is the Legendre polynomial of degree n, and r is the radius to the surface of the normal ellipsoid, which can reduce the ellipsoidal correction for the spherical approximation to a manageably small value (cf. Claessens 2006, Chap 6), and \(\gamma \) is normal gravity on the surface of the normal ellipsoid (Eq. 3) as is demanded by Bruns’s formula (e.g., Heiskanen and Moritz 1967, p 85).
When the integration domain in Eq. (4) is truncated to a spherical cap of radius \(\psi _0\) centred on each computation point, this results in an approximation of the geoid height
where the corresponding truncation error (\(N=\widehat{N_1}+\Delta N\)) is
and \(\delta g_n\) is the \(n\)-th degree spherical harmonic of the gravity disturbance, and the integral in square parentheses yields the truncation coefficients. Recursion formulas for Eq. (7) are given in, e.g., the Appendices of Jekeli (1979). In the remainder of this article, only the integral forms will be presented instead of cluttering the presentation with [too many] recursions.
2.2 Inclusion of an EGM
One very simple way to reduce \(\Delta N\) in Eq. (7) is by subtracting the gravity disturbances computed from an EGM to spherical harmonic degree L (\(\delta g_L\)) from the observed and downward-continued \(\delta g\), and then add back the geoid contribution of the same EGM to the same degree. In physical geodesy, this is commonly referred to as the remove– compute–restore technique.
Albeit well known for Stokes’s integral (e.g., Vincent and Marsh 1974; Rapp and Rummel 1975), this strategy for Hotine’s integral seems first-attributable to Rapp (1980). This gives the residual gravity disturbance
Equation (6) then becomes
where \(N_L\) is the geoid undulation given by an EGM to degree L, and the truncation error becomes
If \(\left\| {\delta g^{L}} \right\| <\left\| {\delta g} \right\| \) over the far zones beyond the spherical cap (\(\psi _0 <\psi \le \pi \)), then \(\delta N^{L}\) should be \({<}\delta N\), indicating that this strategy can reduce the truncation error (cf. Vaníček and Sjöberg 1991). However, the practical implementation is not as simple as the theory may imply. Any EGM is imperfect, and if the terrestrial gravity data contain low-frequency errors, there will be leakage of errors during the evaluation of the last term in Eq. (9) (cf. Vaníček and Featherstone 1998). Therefore, it is important to consider the kernel modification not only as means to reduce the truncation error, but also as a filter to reduce leakage of any low-frequency errors from the terrestrial data (cf. Omang and Forsberg 2002; Featherstone et al. 2011; Wang et al. 2011).
Another benefit of including an EGM in is that the residual geoid height computed from the last term in Eq. (9)—as well as its variants presented in the remainder of this article—is smaller in magnitude (a metre or so vs. up to 100 m), so are less subject to approximation errors. Also, as EGMs become more homogeneously reliable—notably those derived from satellite gravimetry—the motivation for using the remove–compute–restore approach will become stronger.
In order to remain as general as possible, and in line with current widespread practice of utilising an EGM in regional geoid computations under the remove–compute–restore scheme, the following modifications will be applied only to Eqs. (5) and (9). The constants outside the integral terms will be abbreviated to \(\kappa =r{/}{(4\pi \gamma })\) and \(c=r{/}{(2\pi }).\)
3 Deterministic and hybrid modifications
3.1 Remove Legendre polynomials (modification D1)
A simple deterministic kernel modification is to subtract polynomial terms from Eq. (3). For Stokes’s kernel, this approach is generally attributed to Wong and Gore (1969), though de Witte (1967) alluded to it.
Vaníček et al. (1992) and Sjöberg and Nord (1992) have applied this approach to Hotine’s kernel, thus Eq. (9) becomes what herein is termed the spheroidal Hotine kernel
with the subscript D1 denoting this as the first deterministic modification and so on
and
This modification makes Eq. (11) less sensitive to low-frequency errors in the terrestrial gravity disturbances by partially filtering them out (cf. Omang and Forsberg 2002; Wang et al. 2011). However, the filtering is only ever partial because the truncated integration domain allows for spectral leakage (cf. Vaníček and Featherstone 1998).
In the context of band-limited kernel modifications, the summation in Eq. (12) does not necessarily have to start at degree \(n=0\) (cf. Featherstone 2003a). Such a band-limited modification to Stokes’s kernel was used by Li and Sideris (1994) and to Hotine’s kernel by Novák and Heck (2002) and Novák et al. (2003); also see Colombo (1977). Extending this yet further, and not violating the restriction \(M \le L\) (otherwise components of the combined geoid model will be omitted), the band-limited modification can be applied over multiple bands of the user’s choice. This modification is straightforward to implement in existing software as recursion routines for Legendre polynomials are widely available (e.g., Press et al. 2007, Chap 6.7).
Figure 1 shows an example of the D1-modified Hotine kernel (Eq. 12) in relation to the spherical Hotine kernel (Eq. 5). The degree of D1 modification has been chosen arbitrarily at \(M=50\). As the degree of modification increases, the D1-modified kernel oscillates more rapidly. Thus, high degrees of modification can slow the numerical integration because more nodes are needed to determine the integral mean value of the modified kernel over each element (cf. Hirt et al. 2011).
3.2 Set kernel to zero at \(\psi _0\) (modification D2)
Another easy-to-implement deterministic modification is to set the kernel to zero at the truncation radius \(\psi _0\) by subtraction. There is some conjecture as to whether this type of modification, albeit for Stokes’s integral, should be first-attributed to Ostach (1970) or Meissl (1971), but the latter author has gained wider acceptance in the literature on modifications to Stokes’s kernel.
Rapp (1980) applied this strategy to Hotine’s integral, then attributing it to Jekeli (1980b), presumably because both papers were under consideration at around the same time; this is
with
and
Recursion formulas for the integral term in Eq. (16) are given in Guan and Li (1990) and Jekeli (1980b).
By forcing the kernel to zero at the truncation radius accelerates the convergence rate of the truncation error from \(\mathcal O \left( n^{{-}1}\right) \text{ to} \mathcal O \left( n^{{-}2}\right) \) (cf. Jekeli 1980a, 1981; Featherstone et al. 1998). However, faster convergence of a series does not necessarily guarantee smaller values of its coefficients. Since the truncation error is the sum of all terms (Eq. 16), the truncation error may even increase, as shown by Jekeli (1980a, 1981) for Stokes’s kernel. However, Guan and Li (1990) claim that this Ostach–Meissl-type modification to the Hotine kernel does decrease the truncation error.
3.3 Deterministic hybrid (modification D3)
The deterministic modifications in Eqs. (12) and (15) can be combined, as first proposed by Heck and Grüninger (1987) for Stokes’s integral, such that the Legendre polynomials are removed to some degree M such that (s.t.) the zero-crossing point of the D1-modified Hotine kernel in Eq. (12) coincides with the truncation radius \(\psi _0\); this is
with
and
This hybrid modification simultaneously exploits the partial high-pass filtering properties of Eq. (12) and the accelerated rate of convergence of the truncation error from Eq. (16). However, this hybrid is less flexible to implement because M and \(\psi _0\) are inextricably linked in this case. For instance, a small \(\psi _0\) will dictate that M has to be large (cf. Fig 1) and possibly too much filtering will occur, and vice versa.
3.4 Deterministic hybrid (modification D4)
To counteract the above restriction, an alternative combination of Eqs. (12) and (15) can be implemented, where the Legendre polynomials are removed to a user-chosen degree M, and then the kernel is set to zero at the truncation radius \(\psi _0\) by subsequent subtraction (cf. Heck and Grüninger 1987; Featherstone et al. 1998).
This type of hybrid modified Hotine kernel was first introduced by Alberts and Klees (2004), but they did not elaborate upon its properties. In the notation adopted herein, this is
with
and
This hybrid offers the same properties as Eqs. (15) and (18) in terms of the accelerated convergence rate of the truncation error, but does not suffer the same restrictions, thus giving the ‘user’ much more control over the degree of filtering and integration domain chosen. Alberts and Klees (2004) chose \(M=20\) and \(\psi _0 =5^{\circ }\) from simulations and experiments with airborne gravity in their case study area. Importantly, the parameter values chosen depend on the study area, data sets used, their resolution and spatial extent; indeed, this applies to all the modifications.
3.5 Molodensky-type approach (modification D5)
(Molodensky et al. (1962), Chap 7) presented an approach to reduce the \(\text{ L}_{2}\) norm of the truncation error for the spherical Stokes’s kernel. This was later adapted for a higher-than-second-degree reference spheroid by Vaníček and Kleusberg (1987) and Vaníček and Sjöberg (1991); also see Martinec and Vanicek (1997). The Molodensky et al. (1962) approach for the spherical [not spheroidal] Hotine integral (Eq. 5) was presented in Sjöberg and Eshagh (2009), and Novák (2003) presented a band-limited version of the same kernel. However, neither of these modifications allow for the inclusion of an EGM, whereas the presentation herein does.
In analogy, the Molodensky-type modification to the spheroidal Hotine kernel (Eq. 12) was presented by Zhang (1998) as
with
and
Once the values of \(\psi _0\) and M have been chosen by the ‘user’, the \(h_n (\psi _0)\) modification coefficients are determined from the solution of a set of linear equations. However, the formulation of (Zhang (1998), Sect. 2.2) is not particularly intuitive for practical application, and also contains two typographical errors. Thus, a clearer formulation is given below.
By inference from Vaníček and Sjöberg (1991, Eq.18), the \(\text{ L}_{2}\) norm of the Hotine truncation error (Eq. 25) is minimised when
Inserting Eqs. (24) and (12) in Eq. (26) gives
Using the abbreviation
leaves the desired system of (\(M+1\)) linear equations
A recursion formula for the second term in Eq. (29) (cf. Eq. 7) is derived in the Appendices of Jekeli (1979), and recursions for \(e_{nk} (\psi _0)\) (Eq. 28) are given in Paul (1973) or Hagiwara (1972, 1976). The right-hand-side of Eq. (29) is also the recursion used to compute the integral term in Eq. (13). For instance, the computer code from Featherstone (2003b) can be adapted to compute Eq. (29) and thence Eq. (24), as well as the other deterministic modifications presented herein.
3.6 Deterministic hybrid (modification D6)
The modified kernel in Eq. (24) can be forced to be zero at \(\psi _0\) by appropriate selections of \(h_n (\psi _0)\), e.g., by varying M and \(\psi _0\). This approach was first suggested for the Molodensky-modified spheroidal Stokes’s integral (cf. Sect. 3.5) by Featherstone et al. (1998), but it can also be applied to Eq. (24). In analogy to modification D3 (Sect. 3.3), \(h_n (\psi _0)\) are chosen such that the zero-crossing point of the modified Hotine kernel in Eq. (24) coincides with the truncation radius \(\psi _0\)
with
and
However, the practical evaluation of Eq. (31) can be quite cumbersome because trial and error has to be used to meet the condition that this modified Hotine kernel is zero at \(\psi _0\). For some starting values of \(\psi _0\) and M, the kernel in Eq. (24) has to be evaluated (involving the inversion of Eq. 29), then plotted to see if it is zero at \(\psi _0\). If not, then the values of \(\psi _0\) and/or M have to be adjusted until it is. Evidently, this may involve a lot of work and is not so attractive given the following option.
3.7 Deterministic hybrid (modification D7)
A far simpler way to set the D5-modified kernel (Eq. 24) to zero at \(\psi _0\) is by subtraction (cf. Sects. 3.2 and 3.4). This strategy was suggested by Featherstone et al. (1998) for Stokes’s kernel. It allows for more user control over the values chosen for \(\psi _0\) and/or M in terms of the filtering properties of the kernel.
with
and
Practical implementation just involves the evaluation of Eqs. (29) and (24) for the user-chosen values of \(\psi _0\) and M and subtraction of the value of \(H_\mathrm{D5} (M,\psi _0)\) at \(\psi _0\) for \(0< \psi \le \psi _0 \)
4 Stochastic and hybrid modifications
Stochastic kernel modifications can be more subjective than deterministic modifications because reliable error spectra of the data involved are not always available or reliable (e.g., Sjöberg and Hunegnaw 2000). The error spectra of EGMs can be unrealistic if they are derived only from the diagonal of their variance covariance (VCV) matrices and are global estimates, so do not necessarily reflect the errors in a particular region (cf. Sjöberg 2005). Attempts are sometimes made to ‘calibrate’ these error spectra; nevertheless, they still remain global estimates.
The error spectra of terrestrial gravity data are even more problematic to estimate (e.g., Kern et al. 2003) and can vary quite considerably from region to region. Most often, simple covariance models are used (e.g., Ellmann 2005a), which render the stochastic modifiers more akin to least squares collocation and thus subject to the same simplifying assumptions such as stationarity and isotropy. As such, the physical acceptability of the stochastic modifiers is arguably less than for the deterministic modifiers.
4.1 Wenzel-type approach (modification S1)
Wenzel (1981, 1982, 1983) implemented a Wiener-type filter in Stokes’s integral, which can also be applied to Hotine’s integral to give
with the subscript S1 denoting this as the first stochastic modification and so on
and
where \(\sigma _n^2 \{\delta g_\mathrm{EGM}\}\) is the error degree variance of the gravity disturbances from the EGM and \(\sigma _n^2 \{\delta g_\mathrm{T}\}\) is the error degree variance of the terrestrial gravity disturbances. This modification is—by necessity—restricted to the degree L of EGM used in the combined solution for the geoid, such that
leaving
to show that no specific attempt has been made to reduce the truncation error; it is just as large as for the truncated spherical Hotine integral (Eq. 7). Nevertheless, the truncation error is reduced already because of the use of the EGM to degree \(L\) (cf. Sect. 2.2).
4.2 Stochastic hybrid (modification S2)
Similar to the D1 modification (Sect. 3.1), the degree to which the Wiener-type filter is applied can be limited to any \(M \le L\) or any band(s) in that domain.
with, e.g.,
and \(\Delta N_\mathrm{S2} =\Delta N_\mathrm{S1} =\Delta N\), showing again that there is no reduction of the truncation error.
4.3 Stochastic hybrid (modification S3)
An Ostach–Meissl-type modification (cf. Sect. 3.2) can also be applied to Eq. (42), noting that if \(M=L\) it degenerates to Eq. (39) so can be implemented simply for both options.
with
and
This additional modification accelerates the rate of convergence of the truncation error.
4.4 Stochastic hybrid (modification S4)
This stochastic hybrid is an analogue of Heck and Grüninger (1987) and Featherstone et al. (1998) to achieve an accelerated rate of convergence of the truncation error without subtraction (Eq. 45), but by selecting the value of M for which the value of the kernel is zero at \(\psi _0\).
with
and
As for modifications D3 and D6 (Sects. 3.3 and 3.6), this requires cumbersome trial and error to determine the appropriate value of M, but is also complicated further by the choice of EGM used to provide \(\sigma _n^2 \{\delta g_\mathrm{EGM}\}\) and the model adopted for \(\sigma _n^2 \{\delta g_\mathrm{T}\}\) (i.e., via Eq. 38).
4.5 Sjöberg-type approach (modification S5)
Sjöberg (1980a, b, 1981, 1984a, b, 1986, 1991, 2003c) and Sjöberg and Hunegnaw (2000) have provided a series of incremental stochastic modifications to Stokes’s kernel, culminating in the variant in Sjöberg (2003b, Eq. 26). An attractive aspect of most of the Sjöberg-type modifications is that they attempt to simultaneously reduce the truncation error and errors originating from the EGM and terrestrial gravity data, or subsets thereof. The principal restriction is reliably estimating the error spectra of the terrestrial gravity disturbances \(\sigma _n^2 \{\delta g_T\}\), as well as the other caveats mentioned at the start of this Section.
Assuming that Sjöberg (2003b) gives the ‘final word’ on this class of modifications for Stokes’s integral, when applied to Hotine’s integral gives
with
and
Depending on the choices of M and L, the \(s_n\) modification coefficients are
where
and \(c_n \{\delta g\}\) is the degree variance of the gravity disturbances. For weighting the contribution of the EGM, the \(b_n\) coefficients are
Ellmann (2005a, 2012) provides computer code that can be adapted to compute the Sjöberg-type modifiers to Hotine’s kernel, albeit only with an isotropic and stationary covariance model for \(\sigma _n^2 \{\delta g_\mathrm{T}\}\).
4.6 Stochastic hybrid (modification S6)
An Ostach–Meissl-type modification (cf. Sects. 3.2 and 4.3) can also be applied to Eq. (49) to give
with
and
The case of varying the parameters in stochastic modification S5 to achieve a zero crossing of the kernel at the truncation radius \(\psi _0\) is not considered on the grounds of practicality; there are simply too many parameters to trial to make it feasible versus the more pragmatic approach proposed here.
5 Hybrid and band-limited modifications: some suggestions
This section is restricted to a brief discussion of only a few of the options possible, though there are many others; the final choices are left ultimately to the ‘user’. This style of presentation is deliberate to encourage the ‘user’ to experiment with various combinations and permutations so as to tune the data combination for their data sources and area(s) of interest. As alluded to earlier, there is no specific requirement to use any single modification in isolation or to any particular degree or truncation radius, especially when treating them as band-pass filters to reduce errors in the combined solution for the geoid. This applies to Hotine’s, Stokes’s and many other geodetic integrals.
Modern EGMs, particularly those derived from GRACE and/or GOCE satellite gravimetry, are far superior at modelling the low-frequency geoid than terrestrial data alone. As such, it is logical to apply an as-strong-as-possible filter to the terrestrial data, e.g., to the degree that the EGM is considered reliable, so as to rely more upon the low-frequency geoid provided by that EGM. The D1 modification (Sect. 3.1) is the most powerful filter because it removes the low-degree polynomial terms altogether, but the amount of filtering also depends on the truncation radius used (cf. Vaníček and Featherstone 1998). The S1 and S2 modifications (Sects. 4.1 and 4.2) are less effective high-pass filters because they depend on the estimates of \(\sigma _n^2 \{\delta g_\mathrm{T}\},\) noting that if \(\sigma _n^2 \{\delta g_\mathrm{T}\}=0,\) they degenerate to the D1 modification.
One suggested strategy, but only in this author’s opinion (cf. Featherstone 2003a), is to use the D1 modification for the low degrees where the satellite-only EGM is superior to terrestrial data, then use other modifications in the bands where the satellite-only EGM starts to deteriorate, e.g., because of the attenuation of gravitation at satellite altitude. Assuming that GRACE and/or GOCE static gravity field models (e.g., Pail et al. 2011) are better than terrestrial gravity data below some degree L1, but start to deteriorate beyond this to degree \({L2}\,(\le L)\), e.g., the hybrid band-limited kernels from Eqs. (12), (39) and (41) can be combined to give
Likewise, hybrid band-limited versions of the more sophisticated modifiers can be implemented together with the D1 modifier, which can reduce the truncation and other errors. Just as two other examples, combining Eqs. (12), (24) and (50) gives
The above three examples can be extended or simplified depending upon one’s confidence in the terrestrial gravity error spectra, say where the S5 modification is applied in bands where the error spectra are known and the D5 modification applied where they are not. Naturally, there are many more and alternative options than suggested here.
Figure 2 shows two examples of the hybrid band-limited modified Hotine kernels. Equation (60) is calculated using L1 = 50, where polynomial terms are removed completely in the band \(0\le n\le 50\), and the Wenzel-type modifier (S1) is computed to L2 = 2,160 (i.e., \(51\le n\le 2,160\)). EGM2008 (Pavlis et al. 2012) was used to provide the error degree variances of the EGM and \(\sigma _n^2 \{\delta g_\mathrm{T}\}\) are very crudely assumed to be 0.01 mGal\(^{2}\) for all degrees in a band-limited implementation of Eq. (38). The example presented for Eq. (61) also removes polynomials to L1 = 50 and applies a band-limited Vaníček–Kleusberg-type modifier (D5) to L2 = 100 (i.e., \(51\le n\le 100\)) computed for a spherical cap radius of \(\psi _0 =10^{\circ }\).
As well as using band-limited modifiers, it is also possible to truncate the modified kernel to some degree L3 that is commensurate with the spatial resolution of the data (cf. Colombo 1977), and which can also avoid aliasing of high-frequency errors (cf. Kern et al. 2003; Novák et al. 2003). Thus, Eqs. (12), (60), (61) and (62) are adapted to omit the spherical Hotine kernel that contains infinite degrees (Eq. 5) and instead evaluate the kernel in bands that are driven by the perceived reliability of the EGM (L1), the types of modifications selected (L2) and the spatial resolution of the terrestrial gravity disturbances (L3). The overbar is used to distinguish these as the band-limited hybrid kernels.
Naturally, multiple bands and modifiers are possible by combining the above; again, the choices are left to the ‘user’ depending on their data source(s) and area(s) of interest.
The rate of convergence of the truncation error can be accelerated by setting the Hotine kernels to zero at the truncation radius, which removes the discontinuity. This zero can be achieved by simple subtraction or appropriate choices of L1, L2, L3, \(w_n, h_n\) and/or \(s_n\), but noting that some require iteration and lessen the amount of control that the ‘user’ has over the filtering properties of the modifications. Another way to remove the discontinuity is to taper the kernel (cf. Forsberg et al. 2003), though this was not presented as a means to accelerate the convergence, but instead to avoid spectral discontinuities, which can lead to Gibbs fringing when transforming the kernel from the spatial to the spectral domain in FFT geoid computations.
In the most generic form that can be applied to any of the modified Hotine kernels, this is
where \(H_*\) is the modified Hotine kernel, \(m_n\) are the modification coefficients (e.g., 1, \(w_n, h_n\) or \(s_n\)), and the tapering \(\alpha _n\) can be implemented by the following, but other methods of tapering can be used according to the user’s preference
where \(l_1\) is chosen by the ‘user’. Tapering could also be used if a combination of the modified kernels causes other discontinuities.
6 Closing remarks
As more and more terrestrial (land, marine and airborne) gravity observations are coordinated by GNSS, gravity disturbances are becoming available for regional geoid computation via Hotine’s integral. Until large surface areas are covered by these types of observations, the truncation error will be larger than desired. As such, this article has presented a suite of deterministic, stochastic, hybrid and band-limited modifications that can be trialled with gravity disturbances for regional geoid computation. Many are driven by modifications already well established for Stokes’s integral and gravity anomalies, some are new, and some can be applied back to Stokes’s and other geodetic integrals, particularly the band-limited and hybrid variants. However, the motivation for this work is not only the reduction of the truncation error, but also the optimal combination and filtering of the heterogeneous GNSS-based gravity data sources available for regional geoid computation.
These various modifications to Hotine’s kernel have been presented in a deliberately non-prescriptive manner so that the ‘user’ has total freedom to experiment with their combinations and permutations, provided that terms are not omitted, which will result in an incomplete spectral representation of the geoid model. It is suggested, but not necessarily recommended, that (1) D1 modifications (Sect. 3.1) are applied routinely because of the superior data now being provided by GRACE and/or GOCE EGMs, (2) the stochastic modifiers can be applied when the ‘user’ is confident with estimates of the error spectra of the data or wishes consider a stochastic interpretation of the gravity field, and (3) the deterministic modifiers can be applied when there is no reliable information of the error properties of the data or the ‘user’ does not wish to consider a stochastic interpretation of the gravity field.
References
Alberts B, Klees R (2004) A comparison of methods for the inversion of airborne gravity data. J Geod 78(1–2):55–65. doi:10.1007/s00190-003-0366-x
Barriot JP, Balmino G (1992) Estimation of local planetary gravity fields using line of sight gravity data and an integral operator. Icarus 99(1):202–224. doi:10.1016/0019-1035(92)90183-8
Claessens SJ (2006) Solutions to the ellipsoidal boundary value problems for gravity field modelling. PhD thesis, Curtin University of Technology, Perth, Australia
Claessens SJ, Hirt C, Amos MJ, Featherstone WE, Kirby JF (2011) The NZGEOID09 model of New Zealand. Surv Rev 43(319):2–15. doi:10.1179/003962610X12747001420780
Colombo OL (1977) Optimal kernels for band-limited data. UNISURV Report G27, The University of New South Wales, Sydney, Australia pp 77–87
de Witte L (1967) Truncation errors in the Stokes and Vening Meinesz formulae for different order spherical harmonic gravity terms. Geophys J R Astron Soc 12(5):449–464. doi:10.1111/j.1365-246X.1967.tb03125.x
Denker H, Barriot J-P, Barzaghi R, Fairhead D, Forsberg R, Ihde J, Kenyeres A, Marti U, Sarrailh M, Tziavos IN (2009) The development of the European gravimetric geoid model EGG07. In: Sideris MG (ed) Observing our changing earth. Springer, Heidelberg, pp 177–187. doi:10.1007/978-3-540-85426-5
Ellmann A (2005a) Computation of three stochastic modifications of Stokes’s formula for regional geoid determinations. Comput Geosci 31(6):742–755. doi:10.1016/j.cageo.2005.01.008
Ellmann A (2005b) Two deterministic and three stochastic modifications of Stokes’s formula: a case study for the Baltic countries. J Geod 79(1–3):11–23. doi:10.1007/s00190-005-0438-1
Ellmann A (2012) Using high-resolution spectral models of gravity anomaly for computing stochastic modifications of Stokes’s formula. Comput Geosci 39(2):188–190. doi:10.1016/j.cageo.2011.07.00
Evans JD, Featherstone WE (2000) Improved convergence rates for the truncation error in gravimetric geoid determination. J Geod 74(2):239–248. doi:10.1007/s001900050282
Featherstone WE, Olliver JG (1994) A new gravimetric determination of the geoid of the British Isles. Surv Rev 32(254):464–478
Featherstone WE, Evans JD, Olliver JG (1998) A Meissl-modified Vaníêk and Kleusberg kernel to reduce the truncation error in gravimetric geoid computations. J Geod 72(3):154–160. doi:10.1007/s001900050157
Featherstone WE (2003a) Band-limited kernel modifications for regional geoid determination based on dedicated satellite gravity field missions. In: Tziavos IN (ed) Gravity and geoid 2002. Ziti Editions, Thessaloniki, pp 341–346
Featherstone WE (2003b) Software for computing five existing types of deterministically modified integration kernel for gravimetric geoid determination. Comput Geosci 29(2):183–193. doi:10.1016/S0098-3004(02)00074-2
Featherstone WE, Holmes SA, Kirby JF, Kuhn M (2004) Comparison of remove–compute–restore and University of New Brunswick techniques to geoid determination over Australia, and inclusion of Wiener-type filters in reference field contribution. J Surv Eng 130(1):40–47. doi:10.1061/(ASCE)0733-9453(2004)130:1(40)
Featherstone WE, Kirby JF, Kearsley AHW, Gilliland JR, Johnston GM, Steed J, Forsberg R, Sideris MG (2001) The AUSGeoid98 geoid model of Australia: data treatment, computations and comparisons with GPS-levelling data. J Geod 75(5–6):313–330. doi:10.1007/s001900100177
Featherstone WE, Kirby JF, Hirt C, Filmer MS, Claessens SJ, Brown NJ, Hu G, Johnston GM (2011) The AUSGeoid09 model of the Australian height datum. J Geod 85(3):133–150. doi:10.1007/s00190-010-0422-2
Featherstone WE, Filmer MS (2012) The north–south tilt in the Australian height datum is explained by the ocean’s mean dynamic topography. J Geophys Res Oceans 117:C08035. doi:10.1029/2012JC007974
Filmer MS, Featherstone WE, Kuhn M (2010) The effect of EGM2008-based normal, normal-orthometric and Helmert orthometric height systems on the Australian levelling network. J Geod 84(8):501–513. doi:10.1007/s00190-010-0388-0
Forsberg R, Olesen A, Bastos L, Gidskehaug A, Meyer U, Timmen L (2000) Airborne geoid determination. Earth Planets Space 52(10):863–866
Forsberg R, Strykowski G, Iliffe JC, Ziebart M, Cross PA, Tscherning CC, Cruddace P, Stewart K, Bray C, Finch O (2003) OSGM02: a new geoid model of the British Isles. In: Tziavos IN (ed) Gravity and geoid 2002. Ziti Editions, Thessaloniki, pp 132–137
Guan Z, Li Y (1990) The determination of oceanic geoid using modified Hotine integral. In: Rapp RH, Sanso F (eds) Determination of the geoid: present and future. Springer, Heidelberg, pp 86–94
Hagiwara Y (1972) Truncation error formulas for the geoidal height and deflection of the vertical. Bull Géod 46(4):453–466. doi:10.1007/BF02522052
Hagiwara Y (1976) A new formula for evaluating the truncation error coefficient. Bull Géod 50(2):131–135. doi:10.1007/BF02522312
Heck B, Grüninger W (1987) Modification of Stokes’s integral formula by combining two classical approaches. In: Proceedings of the XIX IUGG General Assembly, vol 2. Vancouver, Canada, pp 319–337
Heck B (1990) An evaluation of some systematic error sources affecting terrestrial gravity anomalies. Bull Géod 64(1):88–108. doi:10.1007/BF02530617
Heiskanen WA, Moritz H (1967) Physical Geodesy. Freeman, San Francisco
Hirt C, Featherstone WE, Claessens SJ (2011) On the accurate numerical evaluation of geodetic convolution integrals. J Geod 85(8):519–538. doi:10.1007/s00190-011-0451-5
Hotine M (1969) Mathematical geodesy, ESSA Monograph 2. Department of Commerce, Washington
Jekeli C (1979) Global accuracy estimates of point and mean undulation differences obtained from gravity disturbances, gravity anomalies, and potential coefficients, Report 288. Ohio State University, Columbus, Department of Geodetic Science
Jekeli C (1980a) Reducing the error in geoid undulation computations by modifying Stokes’s function, Report 301. Ohio State University, Columbus, Department of Geodetic Science and Surveying
Jekeli C (1980b) Comparison of undulation difference accuracies using gravity anomalies and gravity disturbances. Bull Géod 54(2):137–147. doi:10.1007/BF02521243
Jekeli C (1981) Modifying Stokes’s function to reduce the error of geoid undulation computations. J Geophys Res Solid Earth 86(B8):6985–6990. doi:10.1029/JB086iB08p06985
Kadir MA, Fashir HH, Omar K (1999) A regional gravimetric co-geoid over South East Asia. Geomat Res Australas 71:37–56
Kearsley AHW, Forsberg R, Olesen A, Bastos L, Hehl K, Meyer U, Gidskehaug A (1998) Airborne gravimetry used in precise geoid computations by ring integration. J Geod 72(10):600–605. doi:10.1007/s001900050198
Kern M, Schwarz K-P, Sneeuw N (2003) A study on the combination of satellite, airborne, and terrestrial gravity data. J Geod 77(3–4):217–225. doi:10.1007/s00190-003-0313-x
Kirby JF (2003) On the combination of gravity anomalies and gravity disturbances for geoid determination in Western Australia. J Geod 77(7–8):433–439. doi:10.1007/s00190-003-0334-5
Li X, Wang Y-M (2011) Comparisons of geoid models over Alaska computed with different Stokes’s kernel modifications. J Geod Sci 1(2):136–142. doi:10.2478/v10156-010-0016-1
Li Y, Sideris MG (1994) Minimization and estimation of geoid undulation errors. Bull Géod 68(4):201–219. doi:10.1007/BF00808101
Martinec Z, Vanicek P (1997) Formulation of the boundary-value problem for geoid determination with a higher-degree reference field. Geophys J Int 126(1):219–228. doi:10.1111/j.1365-246X.1996.tb05280.x
Meissl P (1971) Preparations for the numerical evaluation of second-order Molodensky-type formulas, Report 163. Ohio State University, Columbus, Department of Geodetic Science and Surveying
Molodensky MS, Eremeev VF, Yurkina MI (1962) Methods for Study of the external gravitational field and figure of the earth. Translated from the (1960) original by the Israeli programme for the translation of scientific publications. Jerusalem, Israel
Moritz H (1980) Geodetic reference system 1980. Bull Géod 54(3):395–405. doi:10.1007/BF02521480
Nahavandchi H, Sjöberg LE (2001) Precise geoid determination over Sweden using the Stokes–Helmert method and improved topographic corrections. J Geod 75(2–3):74–88. doi:10.1007/s001900000154
Neyman YM, Li J, Liu Q (1996) Modification of Stokes and Vening-Meinesz formulas for the inner zone of arbitrary shape by minimisation of upper bound truncation errors. J Geod 70(7):410–418. doi:10.1007/BF01090816
Novák P, Heck B (2002) Downward continuation and geoid determination based on band-limited airborne gravity data. J Geod 76(5):269–278. doi:10.1007/s00190-002-0252-y
Novák P (2003) Optimal model for geoid determination from airborne gravity. Studia Geophysica et Geodaetica 47(1):1–36. doi:10.1023/A:1022274821011
Novák P, Kern M, Schwarz KP, Sideris MG, Heck B, Ferguson S, Hammada Y, Wei M (2003) On geoid determination from airborne gravity. J Geod 76(9–10):510–522. doi:10.1007/s00190-002-0284-3
Omang OCD, Forsberg R (2002) The northern European geoid: a case study on long-wavelength geoid errors. J Geod 76(6–7):369–380. doi:10.1007/s00190-002-0261-x
Ostach OM (1970) On the procedure of astro-gravimetric levelling. Studia Geophysica et Geodetica 14(2):222–225. doi:10.1007/BF02585629
Pail R, Bruinsma S, Migliaccio F, Förste C, Goiginger H, Schuh WD, Höck E, Reguzzoni M, Brockmann JM, Abrikosov O, Veicherts M, Fecher T, Mayrhofer R, Krasbutter I, Sansó F, Tscherning CC (2011) First GOCE gravity field models derived by three different approaches. J Geod 85(11):819–843. doi:10.1007/s00190-011-0467-x
Paul MK (1973) A method of evaluating the truncation error coefficients for geoidal height. Bull Géod 47(4):413–425. doi:10.1007/BF02521951
Paul MK (1991) On some possible local relationships between geoidal height and gravity anomaly. Manuscr Geod 16(2):177–190
Pavlis NK, Holmes SA, Kenyon SC, Factor JK (2012) The development and evaluation of the Earth Gravitational Model 2008 (EGM2008). J Geophys Res Solid Earth 117(B4):B04406. doi:10.1029/2011JB008916
Petrovskaya MS (1988) Simplified formulas for geoid height evaluation. Bull Géod 62(2):161–170. doi:10.1007/BF02519223
Petrovskaya MS, Pishchukhina KV (1990) Methods for compact approximation of geoid height. Manuscr Geod 15:253–260
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (2007) Numerical recipes: the art of scientific computing, 3rd edn. Cambridge University Press, New York
Rapp RH, Rummel R (1975) Methods for the computation of detailed geoids and their accuracy, Report 233. The Ohio State University, Columbus, USA, Department of Geodetic Science and Surveying
Rapp RH (1980) A comparison of altimeter and gravimetric geoid in the Tonga Trench and Indian Ocean areas. Bull Géod 54(2):149–163. doi:10.1007/BF02521244
Rapp RH, Wang Y-M (1994) Dynamic topography estimates using GEOSAT data and a gravimetric geoid in the Gulf Stream region. Geophys J Int 117(2):511–528. doi:10.1111/j.1365-246X.1994.tb03948.x
Schwarz K-P, Li YC (1996) What can airborne gravimetry contribute to geoid determination? J Geophys Res Solid Earth 101(B8):17873–17881. doi:10.1029/96JB00819
Serpas JG, Jekeli C (2005) Local geoid determination from airborne vector gravimetry. J Geod 78(10):577–587. doi:10.1007/s00190-004-0416-z
Sjöberg LE (1980a) Least squares combination of satellite and terrestrial data in physical geodesy. Annales de Geophysique 37(1):25–30
Sjöberg LE (1980b) Least squares combination of satellite harmonics and integral formulas in physical geodesy. Gerlands Beiträge zur Geophysik 89(5):371–377
Sjöberg LE (1981) Least squares combination of satellite and terrestrial data in physical geodesy. Ann Geophys 37(1):25–30
Sjöberg LE (1984a) Least squares modification of Stokes’s and Vening Meinesz’s formulas by accounting for the truncation and potential coefficient errors. Manuscr Geod 9:209–229
Sjöberg LE (1984b) Least squares modification of Stokes’s and Vening Meinesz’s formulas by accounting for errors of truncation, potential coefficients and gravity data, Report 27. University of Uppsala, Sweden, Department of Geodesy
Sjöberg LE (1986) Comparison of some methods of modifying Stokes formula. Bollettino di Geodesia e Scienze Affini 45(3):229–248
Sjöberg LE (1989) Integral formulas for geopotential coefficient determination from gravity anomalies versus gravity disturbances. Bull Géod 63(2):213–221. doi:10.1007/BF02519152
Sjöberg LE (1991) Refined least-squares modification of Stokes formula. Manuscr Geod 16:367–375
Sjöberg LE, Nord T (1992) Geoidal undulation computation by modifying Stokes’s kernel versus Hotine’s kernel from gravity anomalies. Manuscr Geod 17:135–140
Sjöberg LE, Hunegnaw A (2000) Some modifications of Stokes’s formula that account for truncation and potential coefficient errors. J Geod 74(2):232–238. doi:10.1007/s001900050281
Sjöberg LE (2003a) Comparison of four deterministic modifications of Stokes’s formula. Geomat Res Australas 78:85–106
Sjöberg LE (2003b) Improving modified Stokes’ formula by GOCE data. Bolletino de Geodesia e Scienze Affini 61:215–225
Sjöberg LE (2003c) A general model of modifying Stokes’ formula and its least-squares solution. J Geod 77(7–8):459–464. doi:10.1007/s00190-003-0346-1
Sjöberg LE (2005) A local least squares modification to Stokes’s formula. Studia Geophysica et Geodaetica 49(1):23–30. doi:10.1007/s11200-005-1623-7
Sjöberg LE, Eshagh M (2009) A geoid solution for airborne gravity data. Studia Geophysica et Geodaetica 53(3):359–374. doi:10.1007/s11200-009-0025-7
Smeets I (1994) An error analysis of the height anomaly determined by a combination of mean terrestrial gravity anomalies and a geopotential model. Bolletino di Geodesia e Scienze Affini 53(1):57–96
Šprlák (2010) Generalized geoidal estimators for deterministic modifications of spherical Stokes function. Contrib Geod Cartogr 40(1):45–64
Stokes GG (1849) On the variation of gravity on the surface of the Earth. Trans Cambridge Philos Soc 8:672–695
Vaníček P, Kleusberg A, Chang RG, Fashir H, Christou N, Hofman M, Kling T, Arsenault T (1987) The Canadian geoid, Report 129. University of New Brunswick, Fredericton, Canada, Department of Geodesy and Surveying Engineering
Vaníček P, Kleusberg A (1987) The Canadian geoid—Stokesian approach. Manuscr Geod 12:86–98
Vaníček P, Zhang C, Ong P (1990) Computation of a file of geoidal heights using Molodensky’s truncation method, Report 147. University of New Brunswick, Fredericton, Canada, Department of Geodesy and Surveying Engineering
Vaníček P, Sjöberg LE (1991) Reformulation of Stokes’s theory for higher than second-degree reference field and modification of integration kernels. J Geophys Res Solid Earth 96(B4):6529–6539. doi:10.1029/90JB02782
Vaníček P, Zhang C, Sjöberg LE (1992) Comparison of Stokes’s and Hotine’s approaches to geoid computation. Manuscr Geod 17(1):29–35
Vaníček P, Featherstone WE (1998) Performance of three types of Stokes’s kernel in the combined solution for the geoid. J Geod 72(12):684–697. doi:10.1007/s001900050209
Vincent S, Marsh J (1974) Gravimetric global geoid. In: Veis G (ed) Proceedings of the International Symposium on the use of Artificial Satellites for Geodesy and Geodynamics. National Technical University, Athens, pp 825–855
Wang Y-M (1993) On the optimal combination of potential coefficient model with terrestrial gravity data for FFT geoid computations. Manuscr Geod 18(6):406–416
Wang Y-M, Saleh J, Li X, Roman DR (2011) The US Gravimetric Geoid of 2009 (USGG2009): model development and evaluation. J Geod 86(3):165–180. doi:10.1007/s00190-011-0506-7
Wenzel H-G (1981) Zur Geoidbestimmung durch Kombination von Schwereanomalien und einem Kugelfunktionsmodell mit Hilfe von Integralformeln. Zeitschrift für Vermessungswesen 106:102–111
Wenzel H-G (1982) Least squares spectral combination by optimal integral kernels. In: Brenneke J, Lelgemann D, Torge W, Wenzel HG (eds) Validation of SEASAT-1 altimetry using ground truth in the North Sea region, Report 263. Deutsche Geodaetische Kommission, Frankfurt
Wenzel H-G (1983) Geoid computation by least squares spectral combination using integral kernels. In: Proceedings of the International Association of Geodesy General Meeting, Tokyo, Japan, pp 438–453
Wessel P, Watts AB (1988) On the accuracy of marine gravity measurements. J Geophys Res Solid Earth 93(B1):393–413. doi:10.1029/JB093iB01p00393
Wichiencharoen C (1984) A comparison of gravimetric geoid undulations computed by the modified Molodensky truncation method and the method of least squares spectral combination by optimal integration kernels. Bull Géod 58(4):494–509. doi:10.1007/BF02523696
Wong L, Gore R (1969) Accuracy of geoid heights from modified Stokes kernels. Geophys J R Astron Soc 18(1):81–91. doi:10.1111/j.1365-246X.1969.tb00264.x
Zelin G, Zuofa L (1992) Modified Stokes’s integral formulas using FFT. Manuscr Geod 17:227–232
Zhang C, Blais JAR (1993) Recovery of gravity disturbances from satellite altimetry by FFT techniques: a comparative study. J Geod 70(1):51–64. doi:10.1007/BF00815486
Zhang C, Sideris MG (1996) Ocean gravity by analytical inversion of Hotine’s formula. Marine Geod 19(2):115–136. doi:10.1080/01490419609388075
Zhang C (1998) Estimation of dynamic ocean topography in the Gulf Stream area using the Hotine formula and altimetry data. J Geod 72(9):499–510. doi:10.1007/s001900050189
Acknowledgments
I would like to thank the three anonymous reviewers, and the handling editor (Prof. Chris Jekeli) for encouraging me to add Appendix A. Disclaimer: The review in Appendix A is incomplete as not all the literature is available to me, such as defence reports or in languages other than English.
Author information
Authors and Affiliations
Corresponding author
Appendix A: Review and classification of modifications to Stokes’s kernel
Appendix A: Review and classification of modifications to Stokes’s kernel
Partial reviews of the modifications to Stokes’s kernel are given in Featherstone (2003b) and Ellmann (2005b). Featherstone (2003b) covers all the deterministic modifications with the exceptions of Sjöberg (2003a) and and Evans and Featherstone (2000). Ellmann (2005a) covers three stochastic modifiers by Sjöberg (1984b; 1991; 2003c). However, the “examples of use” column in Table 1 also includes some review materials.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
About this article
Cite this article
Featherstone, W.E. Deterministic, stochastic, hybrid and band-limited modifications of Hotine’s integral. J Geod 87, 487–500 (2013). https://doi.org/10.1007/s00190-013-0612-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00190-013-0612-9