Survey Review

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ysre20

Geoid determination through the combined least-

squares adjustment of GNSS/levelling/gravity
networks – a case study in Linyi, China

Dongmei Guo & Zhixin Xue

To cite this article: Dongmei Guo & Zhixin Xue (2020): Geoid determination through the combined
least-squares adjustment of GNSS/levelling/gravity networks – a case study in Linyi, China, Survey
Review, DOI: 10.1080/00396265.2020.1842642

To link to this article: https://doi.org/10.1080/00396265.2020.1842642

Published online: 16 Nov 2020.

Submit your article to this journal

Article views: 66

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ysre20
Geoid determination through the combined
least-squares adjustment of GNSS/levelling/
gravity networks – a case study in Linyi, China
Dongmei Guo∗ and Zhixin Xue
A detailed discussion of the adjustment problems used to combine GNSS/levelling/gravity network
data is provided in this paper. The two primary problems inherent to heterogeneous data networks,
namely, parametric models that describe the datums and systematic distortions among the
available data sets and stochastic models that describe the observational residuals, are
described. For parametric models, a relationship between the transformation parameters and the
effects of datums and systematic distortions inherent among different height data types is
established based on a least squares criterion. For stochastic models, the stochastic errors in
GNSS/levelling/gravity data are evaluated, and a Helmert variance component estimation
approach is introduced to refine weighting models. Finally, the proposed model is applied to
determine the hybrid geoid in Linyi, China. The numerical results validate the capability and
effectiveness of the proposed combined adjustment technique for hybrid geoid computations,
revealing an achievable external accuracy of ±1.22 cm compared with GNSS/levelling
measurements, which can be increased by 0.44 cm compared with classic adjustments of
GNSS/levelling/geoid height data.
Keywords: Geoid model, Combined adjustment, Deterministic model, Stochastic model, Accuracy analysis

1. Introduction where h denotes the ellipsoidal height, H represents the

With the wide use of GNSS (Global Navigation Satellite orthometric height and N designates the gravimetric
Systems), more and more attention has been focused on geoid height. The ellipsoidal and orthometric heights
high precise determination of geoid models with an aim are referred to the national reference frame and vertical
to replace the classical levelling-measured heights, via datum, respectively. Gravimetric geoid heights can be
GNSS measurements (Yanalak and Baykal 2001, Kotsa- adjusted with the aid of GNSS/levelling data for the pur-
kis et al. 2012, Sansò and Sideris 2013, Kaloop 2019). The pose of obtaining a ready transformation surface consist-
GNSS/levelling data is of high-precision but low-resol- ent with the national vertical datum (Kiliçoğ lu 2011,
ution while the gravity data is of low-precision but high- Yakubu et al. 2017). Existing methods for geoid determi-
resolution (Heiskanen and Moritz 1967, Sansò and Side- nation involve two steps (Sjöberg and Featherstone 2004):
ris 2013); therefore, the comprehensive utilisation of these (1) the construction of the gravimetric geoid model N
two kinds of data can make up for each other’s using geodetic boundary value problem (GBVP) and (2)
shortcomings. the combination of gravimetric geoid with GNSS/level-
Research has been conducted both in a practical and a ling data to obtain the hybrid geoid. Figure 1 shows the
theoretical point of view to establish a link among ellip- flowchart for the classical adjustment procedure.
soidal, orthometric and gravimetric geoid height data The method of Least Squares (LS) adjustment is
(Kotsakis and Sideris 1999, Kuroishi et al. 2002, Chen usually employed for the combination of heterogeneous
and Luo 2004, Fotopoulos 2005, Sansò and Sideris heights (Jiang and Duquenne 1996, Ahmad-Berger and
2013, Georgopoulos and Telioni 2015). The fundamental Kearsley 1997, Kotsakis and Sideris 1999). When com-
relationship between the triplet of height types can be bining the heterogeneous heights using equation (1), dis-
written as: crepancies may be caused by numerous factors such as
(1) datum inconsistencies intrinsic among the different
h−H −N =0 (1) height types, (2) possible systematic errors in the hetero-
geneous height data sets and (3) stochastic errors in
State Key Laboratory of Geodesy and Earth’s Dynamics, Institute of Geo-
GNSS/levelling and gravity data, which are always sub-
desy and Geophysics, The Chinese Academy of Science, Wuhan, People’s ject to unaccountable influences that cannot be comple-
Republic of China tely controlled (Fotopoulos 2005, Sansò and Sideris
∗
Corresponding author, email: [email protected] 2013). To date, the major component of these

© 2020 Survey Review Ltd

Received 6 August 2020; accepted 22 October 2020
DOI 10.1080/00396265.2020.1842642 Survey Review 2020 1
Guo and Xue. Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

1 Flowchart for the classical adjustment procedure

discrepancies in former studies was usually attributed to 2. Methodology of the combined

datum inconsistencies and systematic errors (Fotopou-
los 2005). Numerous models have employed the incor-
adjustment
poration of parametric models, such as polynomial Stokes’ formula is used to determine gravimetric geoid
fitting, multi-surface function and so on, into the above- heights from gravimetric anomalies. The spherical form
mentioned adjustment scheme to combine the height of Stokes’ integral can be written explicitly as (Heiskanen
data (Kavzoglu and Saka 2005, Tziavos 2012). However, and Moritz 1967)
we often accept rather coarse approximations for the


statistical properties of observations in LS adjustment R
N= DgS(c)d s (2)
(Henk and Kiers 1997, Hu et al. 1999). For example, 4pg
s
height measurements are usually considered indepen-
dent and often have equal precision (i.e. weight). Alter- where γ is the normal gravity, R is the mean radius of the
natively, large errors in gravimetric geoid heights have Earth, Δg denotes the gravity anomaly on geoid, S(ψ) is
been blamed on the inadequacies of gravimetric data- Stokes′ function, ψ is the geocentric angle between a data
bases with regard to their resolution and accuracy. The point and a computation point, and σ is the unit sphere.
statistical tool of variance component estimation Usually, the kernel function of equation (2) can be modi-
affords an efficient approach for calibrating geoid fied in combination with an existing global geopotential
error models (Kotsakis and Sideris 1999, Fotopoulos model (Novák et al. 2007) or a digital elevation model
et al. 2005). (DEM) to improve the accuracy and/or resolution of
In this paper, we propose a new method for combined the gravimetric geoid model (Featherstone 2013). In this
adjustment of GNSS/levelling/gravity data based on clas- study, the gravimetric geoid height was estimated using
sic adjustment method of GNSS/levelling/geoid heights. the remove-compute-restore technique (Heiskanen and
With this proposed method, both the datum transform- Moritz 1967):
ation parameters and the residuals of GNSS/levelling



data and gravity anomalies can be determined. In Section R
2, both a deterministic model used to describe the datum N= (Dg fa − DgEGM + tc)S(c)d s + NEGM
4pg
and systematic distortions among available data sets and s
a stochastic model used to describe observational + Nind (3)
residuals are determined. Here a classic multi-surface
function model is utilised to solve a mathematical adjust- where Δgfa is the free-air anomaly at the topographic sur-
ment problem and a Helmert variance component esti- face, tc is the terrain effect (Sjöberg 1993), Nind is the
mation approach is employed to estimate the weighting indirect effect on geoid (Sjöberg and Nahavandchi
models. In Section 3, the suitability of the proposed 1999). ΔgEGM and NEGM are the reference gravity
GNSS/levelling/gravity adjustment method is carefully anomaly and geoid height components synthesised from
evaluated through a practical application of the compu- global geopotential model, respectively.
tation of the hybrid geoid. Finally, some conclusions are Considering the observational residuals of the gravity
given in Section 4. measurements, the formula used to determine geoid

2 Survey Review 2020

 Guo and Xue Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

heights in equation (3) with gridded gravity anomalies can unknown parameters of the corrector surface through
be rewritten as the following discrete form: the following (Cui and Tao 1992)

n
x̂ = (BT (P−1 −1 T −1 T−1
−1 −1 T −1
Nir = (Dgj + yj − DgEGM + tc)K(cij )Dsj 1 + AP2 A ) B) B (P1 + AP2 A ) l
j=1 (12)
+ NEGM + Nind (4) Furthermore, we can also solve for the vectors of the
observational residuals for each of data types (Cui and
where n is the number of grids, Δgj is the mean gravity Tao 1992):
anomaly of the jth grid element, ʋj denotes the observa-
tional residuals of the gravity measurements, and Δsj is u = −P−1 −1 −1 T −1
1 (P1 + AP2 A ) (l − Bx̂) (13)
the area of each grid element as follows:
−1 T −1


y = P−1 −1
2 A (P1 + AP2 A ) (l − Bx̂)
T
(14)
Dsi = d s (5) From equations (12)–(14), we can see that the para-
sj metric model Bx for describing the systematic distortions
and the weighting model of observation residuals are
The kernel function K(ψij) can be expressed as
important in the combined height adjustment. Here a
R classic multi-surface function model is utilised to solve a
K(cij ) = S(cij ) (6) mathematical adjustment problem from a geometrical
4pg
perspective considering that this function approach is an
Equation (4) can be expressed in a matrix form as effective interpolation method for solving mathematical
follows: surface adjustment problems from a geometrical perspec-
y = y c
+ A y (7) tive (Chen and Luo 2004, Yu 2013). A Helmert variance
m×1 m×1 m×n n×1 component estimation approach (Zhang et al. 2019) is
employed to estimate the weighting models, since this
where
 m is the number T of GNSS/levelling points, method is highlighted as the best for selecting the regular-
y = N1r N2r · · · Nmr , each component Nir in the
isation parameter compared with other methods (Fan
vector
 y is  given in equation (4),
T 2017). Figure 2 shows the flowchart for the proposed
y = y1 y2 · · · yn , the component in row i and col-
adjustment procedure.
umn  k of the matrix T A is aij = K(cij )Dsj ,
yc = N1c N2c · · · Nmc , and each component in
the vector yc can be written as
3. Case study

n
Nic = (Dgj − DgEGM + tc)K(cij )Dsj + NEGM 3.1. Available data
j=1 A high resolution digital terrain model was obtained from
+ Nind (8) the Shuttle Radar Topographic Mission (SRTM) with a
resolution of 90m (Denker 2005, Farr 2007). Here
Then, a combined adjustment can be conducted using SRTM 90m dataset was used to provide the high-fre-
the theoretical relation between GNSS/levelling heights quency information of the gravity field. Figure 3 gives
with geoid undulations. Considering the Gauss-Markov the SRTM 90m topographic data.
model, the system of observation equation can be The terrestrial gravity data including 995 gravity
expressed as, measurements are referenced to National Gravity Funda-
z + u − y = B × x (9) mental Network 2000 (Chen 2007). These gravity
m×1 m×1 m×1 m×t t×1 measurements are of station spacing about 4 km. Figure
4 shows the distribution of gravity stations. Terrestrial
where z is the vector comprising the ellipsoidal heights at
gravity data obtained were checked with 18 absolute grav-
co-located GNSS/levelling sites, u is the vector of observa-
ity values which are shown in Figure 4 denoted by aster-
tional residuals for the GNSS/levelling heights, the (m×t)
isk, and a standard deviation value (STD) of ±23 μGal
design matrix B depends on the parametric model (m is
can be achieved. Furthermore, since the approach pro-
the number of observational equations and t is the number
posed in this study requires regularly gridded gravity
of unknown parameters), and x is a vector containing the
anomalies, the discrete gravity values were interpolated
unknown parameters.
into a gridded form using Shepard’ surface fitting method
Substituting the solution of equation (7) into equation
(Park et al. 2004).
(9) and letting l = z–yc, we can obtain the following obser-
Global gravitational model (GGM) was employed to
vation equation for the combined adjustment:
provide a reference gravity anomaly (Pavlis 2008, Roy
Bx + Ay − u − l = 0 (10) et al. 2017, Pail 2018). In this study, a criterion for
selecting the optimal GGM was established with mini-
For m > t, the solution for the unknown parameters mal STD of the difference between terrestrial gravity
(Wolberg 2006) can be obtained by applying the LS mini- data and the reference gravity data computed from
misation principle of: high-order GGMs such as EGM2008 (Earth Gravita-
uT P1 u + yT P2 y = minimum (11) tional Model 2008), XGM2016 (Experimental Geopo-
tential Model 2016) and EIGEN-6C4 (European
where P1 is the weighting matrix of GNSS/levelling Improved Gravity model of the Earth by New tech-
heights, P2 is the weighting matrix of gravity anomalies. niques-6C4). The statistical results are listed in Table 1.
According to equations (10) and (11), we can obtain the Here EIGEN-6C4 up to degree 2190 was used to compute

Survey Review 2020 3

Guo and Xue. Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

2 Flowchart for the proposed adjustment procedure

3 SRTM 90 m topographic data (unit: m)

4 Survey Review 2020

 Guo and Xue Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

4 The distributions of the gravity stations

the reference gravity anomalies. The STD of differences of 142 well-distributed GNSS benchmarks (shown as dots
between the reference gravity data and terrestrial gravity in Figure 5) are treated as control points. The precisions
data is ±5.24mGal. of the ellipsoidal heights at the control points are better
165 GNSS/levelling stations evenly distributed were than 1 cm. The GNSS/levelling observations at these con-
used. Their ellipsoidal heights are referenced to 2000 trol points were used to perform a LS adjustment in com-
national geodetic coordinate system (Yang 2009) and bination with the gravimetric geoid heights, while the
the orthometric heights are referenced to National Verti- remaining GNSS benchmarks were treated as check
cal Datum 1985. Figure 5 shows their distribution. A total points to test the external accuracy of the final hybrid
geoid model.
The statistics regarding the DTM, gravity data and
Table 1 Statistics of the GGMs precision compared with GNSS/levelling data used in this study are shown in
terrestrial gravity data (unit: mGal)
Table 2. The DTM data range from 1.15 to 1125 m,
Standard thus it can be seen that Linyi is a hilly territory with
Minimum Maximum Mean deviation great relief. The maximum and minimum value of gridded
gravity data in this area is 50.7 and −6.6 mGal, respect-
EGM2008 −24.17 27.65 −2.95 ±5.29 ively, and the average value is 9.3 mGal. And the geo-
XGM2016 −24.72 28.91 −4.34 ±7.58 metric geoid heights obtained from GNSS/levelling
EIGEN-6C4 −24.29 27.51 −3.45 ±5.24
measurements have an average value of −0.8 m.

Survey Review 2020 5

Guo and Xue. Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

5 GNSS benchmarks with known orthometric heights (dot symbols denote the control points, asterisk symbols denote the
check points)

3.2. Construction and accuracy analysis of as a parametric model which solves a surface adjustment
local geoid model problem by the data of points from a geometrical view.
In the adjustment method elaborated upon in Section 2, The differences between geometric geoid heights and
the classic method is used with a multi-surface function gravimetric geoid are expressed as


m
z(x, y) = aj u(x, y; xj , yj ) (15)
Table 2 Statistics of the DTM, gravity data and GNSS/ j=1
levelling data

Standard where aj are coefficients to be determined. They are calcu-

Minimum Maximum Mean deviation lated using the LS method. θ(xi,yi;xj,yj) is a kernel func-
tion such that
DTM (unit:m) 1.1 1125.0 155.7 ±112.6
Gridded free-air −6.6 50.7 9.3 ±7.1 1
gravity
anomalies (unit: u(x, y; xj , yj ) = [(x − xj ) + (y − yj ) + d ]2
2 2 2
(16)
mGal)
GNSS/levelling −3.9 2.1 −0.8 ±1.4 where δ is the smoothing factor which dominates the
geoid heights shape of reciprocal hyperboloid. Here, δ is established as
(unit:m)
2 km through empirical testing.

6 Survey Review 2020

 Guo and Xue Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

An iterative Helmert variance component estimation The iterative method is based on the following pro-
approach was employed to estimate the weighting cedure. First, suppose that we have two estimates ŝj01
model in consideration of the numerous sources of data and ŝj02 obtained at the jth step by using a conventional
that are involved. The initial weight matrix P associated Helmert variance component estimation approach.
with the observations take the forms, Then, in the next iteration, a new set of variance com-
   −1  ponents can be computed by (Gao 2008):
P1 0 C1 0 ⎧
P= = (17)
0 P2 0 C2−1 ⎪ 2 j+1
⎨ {ŝ01 } = {ŝ201 }j
j j (19)
⎪ {ŝ2 } + {ŝ202 }
where C1 is the covariance matrix for GNSS/levelling ⎩ {ŝ202 } j+1 = 01
heights which come directly from errors of measurement 2
(Cui and Tao 1992, Fotopoulos 2005). C2 is the covari- And the weighting model can be renewed for the next
ance matrixes for gravity anomalies which are obtained iteration:
from according to the Hirvonen model of a planar isotro-
pic covariance function (Meier 1981, Kotsakis 2007). ⎧ j+1
⎨ P1 = I
⎪
j
C0 {ŝ201 } (20)
C2 (P, Q) = (18) ⎪ j+1
⎩ P2 = 2 j j
1 + (rPQ /a)2 {ŝ02 } /P2
where the variance C0 is the value of the covariance func- We can then update the covariance matrixes of the
tion for the argument rPQ = 0 (Schwarz and Lachapelle observables with the estimated values of the variance
1980), rPQ is the planar distance between points P and components. The process will terminate after all of unit
Q, α is the correlation length of the gravity anomaly weight variances satisfy a convergence criterion, e.g.
field (Moritz 1977), and α = 45 km. [ŝ201 ŝ202 ]T  [1 1]T .

6 The hybrid Geoid model LinyiGEOID (unit: m)

Survey Review 2020 7

Guo and Xue. Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

The method of geoid refining based on the LS adjust- stochastic errors among each data type is most important
ment model described in Section 2 comprises three factor to take into account. Compared with classic adjust-
steps: first, gravimetric geoid heights yc in equation (7) ments of GNSS/levelling/geoid height data, our revised
were computed at the GNSS/levelling points from the formulation can take full advantage of heterogeneous ter-
remove-compute-restore process described in equation restrial data and accurately evaluate stochastic errors in
(8), and the gravimetric geoid model in the gridded format the source data.
can also be computed from this process with gravity
anomalies; second, unknown parameters of corrector sur-
face and observational residuals for GNSS/levelling 4. Conclusions
heights and gravity data were determined with equations In this study, a combined adjustment method for GNSS/
(12), (13) and (14) over co-located GNSS/levelling bench- levelling/gravity network data is given to determine the
marks (e.g. control points); finally, the corrector surface in local geoid model in an attempt to find optimal trans-
a gridded format can be obtained using these parameters formation models aimed at replacing traditional levelling
and subsequently used to correct the gravimetric geoid surveys with GNSS measurements. Compared with clas-
model. Figure 6 shows results of the hybrid geoid model sic adjustments of GNSS/levelling/geoid height data,
LinyiGEOID, which ranges from −4.10 to 2.47 m. our revised formulation can take full advantage of hetero-
To estimate the accuracy of LinyiGEOID, we analyse geneous terrestrial data and accurately evaluate the sto-
its internal and external accuracies. First the hybrid chastic errors in the data sources.
geoid model has been interpolated to the location of the Two problems, namely, the use of deterministic models
GNSS/levelling points using the method of multi-surface and stochastic models, can be resolved simultaneously
function given in equation (15). Then, the internal accu- through this hybrid adjustment method. Deterministic
racy was conducted from statistical analysis of the differ- models can reduce systematic distortions originating
ences between the hybrid geoid and GNSS/levelling from datum inconsistencies and distortions in height
observations at the 142 control stations. The minimum, data. Stochastic models allow us to individually validate
maximum and STD values of the internal accuracy are various observational errors in data sources. In particular,
−1.15, 1.47 and ±0.57 cm, respectively. A more reliable Helmert variance component estimation approach has
check of the external accuracy is conducted by examining been employed as a useful tool for refining weighting
the differences between the hybrid geoid and the GNSS/ models.
levelling observations at the 23 check points denoted as Two models, namely, the classic GNSS/levelling/geoid
asterisks in Figure 3. The STD value in these differences model and the proposed GNSS/levelling/gravity model
is ±1.22 cm. The statistical results of the internal accuracy are presented to illustrate the theory. Terrestrial gravity
and external accuracy are listed in Table 3. data, DTM, global geopotential model, and GNSS/level-
We further assess the superiority in GNSS/levelling/ ling heights are optimally used to establish the high-pre-
gravity method relative to the classical GNSS/levelling/ cise geoid model over Linyi, China in this study.
geoid method. The GNSS/levelling/geoid method Numerical results indicate that the procedure proposed
employs the classic method represented by equation (1) in this study can apparently improve the accuracy of the
to solve the general adjustment model. This method local geoid model compared with the widely used
assumes that the systematic distortions among these GNSS/levelling/geoid method and an external accuracy
three height types have been sufficiently removed through of ±1.22 cm can be achievable compared with GNSS/
a reliable parametric model. Here the multi-surface func- levelling measurements.
tion is also employed as the parametric model. Both
internal and external accuracy values are listed in Table
3. Furthermore we assess the differences between the Acknowledgements
geoid heights computed using EIGEN-6C4 and the Dongmei Guo carried out the deterministic model
GNSS/levelling observations at the 23 check points, and research. Xue Zhixin conducted the stochastic model
the statistical results are listed in Table 3 as well. The research. All of the authors read and approved the final
STD value in these differences is ±4.76 cm. From the manuscript.
results, we can see that the method proposed in this
paper can significantly improve the accuracy of the hybrid
geoid. Therefore, we can conclude that estimation of the Funding
This work is supported by the Basic Frontier Science Research Program
of Chinese Academy of Sciences [grant number ZDBS-LY-DQC028] and
Table 3 Accuracy analysis of geoid models compared with Independent project of State Key Laboratory of Geodesy and Earth’s
GNSS/levelling measurements (unit: cm) Dynamics [grant number E025011010].

Standard
Minimum Maximum deviation
References
GNSS/ Internal −1.15 1.47 ±0.57 Ahmad-Berger, Z. and Kearsley, A.H.W., 1997. Geoid height optimis-
levelling/ accuracy ation via sequential least squares method. In: J. Segawa, H.
gravity External −1.62 2.82 ±1.22 Fujimoto, and S. Okubo, eds. Gravity, geoid and marine geodesy.
method accuracy International association of geodesy symposia, Vol. 117. Berlin:
GNSS/ Internal −1.97 1.90 ±0.63 Springer.
levelling/ accuracy Chen, J.Y., et al., 2007. Establishment of 2000 national geodetic control
geoid External −2.82 3.35 ±1.66 network of China and it’s technological progress. Acta geodaetica
et cartographica sinica, 1, 1–8 (in Chinese).
method accuracy
Chen, Y. and Luo, Z., 2004. A hybrid method to determine a local geoid
EIGEN-6C4 model −16.8 1.90 ±4.76
model-case study. Earth planets space, 56, 419–427.

8 Survey Review 2020

 Guo and Xue Geoid determination through the combined least-squares adjustment of GNSS/levelling/gravity networks

Cui, X. and Tao, B., 1992. Generalized surveying adjustment. Beijing: Kotsakis, C., and Sideris, M.G., 1999. On the adjustment of combined
Surveying and Mapping Press (in Chinese). GPS/levelling/geoid networks. Journal of geodesy, 73, 412–421.
Denker, H., 2005. Evaluation of SRTM3 and GTOPO30 Terrain Data in Kuroishi, Y., Ando, H., and Fukuda, Y., 2002. A new hybrid geoid model
Germany. In: Jekeli C., Bastos L., Fernandes J. (eds) Gravity, Geoid for Japan, GSIGEO2000. Journal of Geodesy, 76, 428–436.
and Space Missions. International Association of Geodesy Meier, S., 1981. Planar geodetic covariance functions. Reviews of geophy-
Symposia, vol 129. Springer, Berlin, Heidelberg. sics, 19 (4), 673–686.
Fan, Q., et al., 2017. Implication of adaptive smoothness constraint and Moritz, H., 1977. On the computation of a global covariance function.
Helmert variance component estimation in seismic slip inversion. Report No. 255 of the Department of Geodetic Science.
Journal of geodesy, 91, 1163–1177. Columbus, OH: The Ohio State University.
Farr, T.G., et al., 2007. The shuttle radar topography mission. Reviews of Novák, P., Kostelecký, J., and Klavklokoč ník, J., 2007. Testing global
geophysics, 45, RG2004. doi:10.1029/2005RG000183. geopotential models through comparison of a local quasi-geoid
Featherstone, W.E., 2013. Deterministic, stochastic, hybrid and band- model with GPS/leveling data. Studia geophysica et geodaetica,
limited modifications of Hotine’s integral. Journal of geodesy, 87, 53, 39–60.
487–500. Pail, R., et al., 2018. Short note: the experimental geopotential model
Fotopoulos, G., 2005. Calibration of geoid error models via a combined XGM2016. Journal of geodesy, 92, 443–451.
adjustment of ellipsoidal, orthometric and gravimetric geoid height Park, H., Jung, H.B., and Kim, K., 2004. A new approach for lofted B-
data. Journal of geodesy, 79, 111–123. pline surface interpolation to serial contours. International journal
Fotopoulos, G., Kotsakis, C., and Sideris, M., 2005. Estimation of var- of advanced manufacturing technology, 23, 889–895.
iance components through a combined adjustment of GPS, geoid Pavlis, N.K., et al. 2008. An earth gravitational model degree 2160:
and levelling data. In: F. Sansò, ed. A window on the future of geo- EGM2008. Presented at the 2008 general assembly of the
desy. International association of geodesy symposia, Vol. 128. Berlin: European Geosciences Union, April 13–18, Vienna.
Springer. Roy, P., Sai Krishnaveni, A., and Vinod Kumar, K., 2017. Geological
Gao, B., et al., 2008. Statistical analysis of negative variance components evaluation of EIGEN-6C4 and GOCE derived gravity models in
in the estimation of variance components. In: P. Xu, J. Liu, and A and around Karakoram shear zone, Leh, India. Journal of the geo-
Dermanis, ed. VI Hotine-Marussi symposium on theoretical and logical society of India, 90, 51–61.
computational geodesy. International association of geodesy sympo- Sansò, F. and Sideris, M.G., 2013. Geoid determination: theory and methods.
sia, Vol. 132. Berlin: Springer. Lecture notes in earth system sciences. Berlin: Springer-Verlag.
Georgopoulos, G.D. and Telioni, E.C., 2015. Determination of local Schwarz, K.P., and Lachapelle, G., 1980. Local characteristics of the grav-
geoid model in Attica Basin Greece. Survey review, 42, 109–114. ity anomaly covariance function. Bulletin geodesique, 54, 21–36.
Heiskanen, W. and Moritz, H., 1967. Physical geodesy. San Francisco: Sjöberg, L.E., 1993. Terrain effects in the atmospheric gravity and geoid
WH Freeman. corrections. Bulletin geodesique, 67, 178–184.
Henk, A. and Kiers, L., 1997. Weighted least squares fitting using ordin- Sjöberg, L.E. and Featherstone, W.E., 2004. Two-step procedures for
ary least squares algorithms. Psychometrika, 62 (2), 251–266. hybrid geoid modelling. Journal of geodesy, 78, 66–75.
Hu, X., Liao, X., and Huang, C., 1999. Data weighting and solution Sjöberg, L.E. and Nahavandchi, H., 1999. On the indirect effect in the
assessment in combination. Journal of geodesy, 73, 391–397. Stokes–Helmert method of geoid determination. Journal of geo-
Jiang, Z. and Duquenne, H., 1996. On the combined adjustment of a desy, 73, 87–93.
gravimetrically determined geoid and GPS levelling stations. Tziavos, I.N., et al., 2012. Adjustment of collocated GPS, geoid and
Journal of geodesy, 70, 505–514. orthometric height observations in Greece geoid or orthometric
Kaloop, M.R., et al., 2019. Optimizing local geoid Undulation model height Improvement. In: Steve Kenyon, Maria Christina Pacino,
using GPS/levelling measurements and heuristic regression and Urs Marti, eds. Geodesy for planet earth. International associ-
approaches. Survey review. doi:10.1080/ 00396265. 2019.1665615. ation of geodesy symposia, Vol. 136. Berlin: Springer.
Kavzoglu, T. and Saka, M.H., 2005. Modelling local GPS/levelling geoid Wolberg, J., 2006. Data analysis using the method of least squares. Berlin:
undulations using artificial neural networks. Journal of geodesy, 78, Springer-Verlag.
520–527. Yanalak, M. and Baykal, O., 2001. Transformation of ellipsoid heights to
Kiliçoğ lu, A., et al., 2011. Regional gravimetric quasi-geoid model and local leveling heights. Journal of surveying engineering, 127 (3), 90–103.
transformation surface to national height system for Turkey Yang, Y.X., 2009. Chinese geodetic coordinate system 2000. Chinese
(THG-09). Studia geophysica et geodaetica, 55, 557–578. science bulletin, 54 (15), 2714–2721.
Kotsakis, C., 2007. Least-squares collocation with covariance-matching Yu, B., et al., 2013. The application of an improved multi-surface func-
constraints. Journal of geodesy, 81, 661–677. tion based on earth gravity field model in GPS leveling fitting.
Kotsakis, C., Katsambalos, K., and Ampatzidis, D., 2012. Estimation of Geo-informatics in resource management and sustainable ecosystem,
the zero-height geopotential level WoLVD in a local vertical 399, 141–148.
datum from inversion of co-located GPS, leveling and geoid Zhang, B., et al., 2019. Precipitable water vapor fusion: an approach
heights: a case study in the Hellenic islands. Journal of geodesy, based on spherical cap harmonic analysis and Helmert variance
86, 423–439. component estimation. Journal of Geodesy, 93, 2605–2620.

Survey Review 2020 9