Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

The Egyptian Journal of Remote Sensing and Space Sciences 21 (2018) 265–271

Contents lists available at ScienceDirect

The Egyptian Journal of Remote Sensing and Space Sciences

journal homepage: www.sciencedirect.com

Polynomial interpolation methods in development of local geoid model


Rabindra Kumar Das ⇑, Sailesh Samanta, Sujoy Kumar Jana, Robert Rosa
Department of Surveying and Land Studies, Papua New Guinea University of Technology, Private Mail Bag, Lae 411, Morobe Province, Papua New Guinea

a r t i c l e i n f o a b s t r a c t

Article history: In geodesy three surfaces, the physical surface of the earth, the geoid and the reference ellipsoid are
Received 16 November 2016 encountered giving rise to orthometric height (h), the ellipsoidal height (H) and the geoidal separation
Revised 11 January 2017 (N). The orthometric height and the ellipsoidal height are with reference to the geoid and the reference
Accepted 26 March 2017
ellipsoid respectively. The vertical separation between the ellipsoid and the geoid is the geoidal separa-
Available online 10 April 2017
tion. A mathematical relation depicting the surface of the geoid with regard to the reference ellipsoid is
the geoid model. It relates the geoidal separation with the horizontal location.
Keywords:
The Global Navigational Satellite System provides precise location of points on the surface of the earth.
Geodesy
Orthometric height
The vertical location provided is the ellipsoidal height which needs conversion to a more usable format,
Ellipsoidal height the orthometric height. This is done by integrating ellipsoidal heights with a geoid model. The accuracy of
Geoidal separation conversion depends on the accuracy of geoid model. Therefore, development of geoid model has become
Geoid model a current area of research in geodesy.
Objective of this study is to develop a local geoid model by employing various polynomial models and
thereafter to analyse the accuracy of these models. The test area is in Papua New Guinea. The geometric
method is used for computation of the geoidal separation from ellipsoidal and orthometric heights and
thereafter the horizontal coordinates and the geoidal separation are used to develop the geoid surface
using second, third and fourth degree polynomials. The study shows that the third degree polynomial
provided an accuracy of ±20 cm.
Ó 2017 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-
nd/4.0/).

1. Introduction The advent of Global Navigational Satellite System (GNSS) such


as the Global Positioning System (GPS) has revolutionized the
In geodesy three basic surfaces are encountered: the physical precise positioning of points/features on the surface of the earth.
surface of the earth, the geoid and the reference ellipsoid. In rela- However the vertical positioning by this technique refers to the
tion to these three surfaces another three entities are defined: ellipsoidal height. Therefore professionals involved in the area of
the orthometric height (h), the ellipsoidal height (H) and the geoi- geomatics engineering including geodesists and surveyors convert
dal separation/undulation (N). The orthometric height is the height the ellipsoidal heights to a more usable simple format, the ortho-
with reference to the geoid whereas the corresponding ellipsoidal metric heights whose approximation is the Mean Sea Level height.
height is with regard to the reference ellipsoid. The vertical separa- This is done by integrating ellipsoidal heights with a global or
tion between the ellipsoid and the geoid is known as geoidal sep- regional or local geoid model. Accuracy of the geoid model reflects
aration. A mathematical relation depicting the surface of the geoid the accuracy of the converted orthometric heights. Therefore
with regard to the reference ellipsoid is known as geoid model. In a development of precise geoid models has become a current area
simpler form the model gives a relation between the geoidal sepa- of research in geodesy.
ration (N) and the horizontal location such as the Easting (x) and There are a number of mathematical models to relate the geoi-
Northing (y). dal separation N with corresponding locations in order to define
the geoid surface. Such models include simple polynomials
(Zhong, 1997), radial basis functions, multilayer perception neural
Peer review under responsibility of National Authority for Remote Sensing and network methods (Cakir and Yilmaz, 2014) etc. A back propagation
Space Sciences. artificial neural network was applied to develop a regional grid-
⇑ Corresponding author.
based geoid model using GPS derived ellipsoidal heights and the
E-mail addresses: rkdas2000@gmail.com (R.K. Das), rsgis.sailesh@gmail.com
(S. Samanta), sujoy2007@gmail.com (S.K. Jana), robert.rosa@pnguot.ac.pg (R. Rosa). orthometric heights (Lin, 2007). However, the polynomial models

http://dx.doi.org/10.1016/j.ejrs.2017.03.002
1110-9823/Ó 2017 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
266 R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271

are simple, easy to understand and adoption of these models for approximately bounded between 5° 120 to 5° 150 south latitude
development of geoid surface has been widely reported and and 145° 460 to 145° 490 east longitude. The terrain is nearly flat
therefore considered in the present study. The general form of a with undulations varying from 0.83 m to 14.50 m with regard to
polynomial model is: N (x, y) = a1xn + a2yn + . . .. . .. . .. . . Mean Sea Level. Thirty three (33) control points are available in
For example a second degree polynomial is: the study area. The location of the study area along with distribu-
tion of these points over corresponding Landsat 8 satellite image
Nðx; yÞ ¼ a1 x2 þ a2 y2 þ a3 xy þ a4 x þ a5 y þ a6 ð1Þ has been provided in Fig. 1. These points have their horizontal
Here a1 to a6 are the six coefficients of transformation. Similarly co-ordinates in terms of UTM easting and northing (UTM zone
the third and further higher degree polynomials are defined. No 55). Besides, these points have their orthometric heights fixed
A geoid model is developed in two stages. At the first stage the through differential spirit levelling and ellipsoidal heights through
geoidal separation (N) is determined either by Geometric Method GPS survey based on PNG 94, the National Geodetic Reference
or Gravimetric Method or Hybrid one. The geometric method uses System developed for Papua New Guinea during the year 2014.
the ellipsoidal height and the orthometric height whereas At the outset an exercise was carried out where all the thirty
gravimetric method employs the gravity anomaly to compute N. three control points are processed through the second degree poly-
Gravimetric method provides accurate geoid models only when nomial. The orthometric heights of the points are used along with
sufficient gravity data is available. In hybrid method both the geo- the corresponding ellipsoidal heights to compute the geoidal sepa-
metric method and gravimetric method are integrated to get the ration (N) using the relation N = H-h (in this case the geoid is above
optimum accuracy. The accuracy of a local geoid model may be the ellipsoid). Thereafter the horizontal coordinates Easting (x),
improved by integrating an existing gravimetric geoid model with Northing (y) and the geoidal separation (N) of these points are used
GPS derived ellipsoidal height and orthometric height (You, 2006). in computing the model coefficients of the second degree polyno-
The geometric method is simple, reasonably accurate and adopted mial (Eq. (1)). It may be noted that the second degree polynomial
in this study. In this method a number of control points uniformly has six unknown coefficients needing six control points. Therefore
distributed within the area of interest are selected. These points there are 27 redundant observations and the processing was car-
have their precise horizontal and vertical locations in terms lati- ried out with least square adjustment technique where the N of
tude and longitude (or easting- x and northing –y), the ellipsoidal these control points are again evaluated from the model by input-
heights (H) and also the orthometric heights (h). The geoidal sepa- ting their horizontal coordinates. The evaluated N from the model
ration (N) is computed using the simple mathematical relation is compared with the original N and from this a standard error of
H = h ± N, the positive or negative sign used according to the geoid ±0.49 m has been obtained. The other statistics extracted from
is above or below the reference ellipsoid (Dinter et al., 1997; the processed data include the range of errors and maximum error.
Bomford, 1971; Smith, 1997). Subsequently the same 33 points are processed through the third
In the second stage the geoidal separation N, and the horizontal degree and then the fourth degree polynomial and similar statistics
coordinates x and y of the control points are used in defining the has been extracted.
specific mathematical relation/model. For example the mathemat- On the basis of the statistics of the 33 points it was observed
ical model as in Eq. (1) may be considered. At least with six control that one point i.e. point No. 13 has shown significant higher error
points whose N, x, y are known, the coefficients from a1 to a6 may in all the three models with near identical magnitude. The stan-
be computed thus defining the exact mathematical relation. There- dard error obtained for each model was also near identical. The
after this model can be used to obtain the N value for any location range of errors encountered from second degree is from 1.79 m
by inputting the horizontal coordinates x and y (or latitude and to +0.94 m and the range of errors of other two models are also
longitude). near identical. It is worthwhile to mention that in redundant sur-
There are global geoid models such as EGM 1996 and EGM 2008 vey observations/measurements, it is essential to eliminate the
which supports the conversion of ellipsoidal heights to orthometric gross errors/mistakes before final processing. For this there are a
heights with accuracies varying between few centimetres to even a number of criterions in use. One of the criteria is to reject observa-
metre. Over areas covered with high quality gravity data, the dis- tions having error of 4.2 times the probable error (David, 2012).
crepancy between EGM 2008 geoid undulations and independent This is equivalent to 2.83 times of standard error. In Geodesy and
GPS/levelling data is of the order of ±5 to ±10 cm (Nikolaos et al., Photogrammetry one often sets 2 or 3 times of initial standard
2012). The accuracy of the conversion may be further improved error r as the accepted error of tolerance (Fan, 1997). In some
with adoption of regional or local geoid models. A regional geoid other citations even the E90 error (=±1.645 r) has been used as
model was developed for Jordan using gravity data and compared accepted error tolerance (Wijayratne, 2002). In the present study
with the GPS/levelling measurements yielding an accuracy of the criteria of ±2.83 r has been adopted to identify points possibly
about ±40 cm (Omar and Abdulla, 2007). Local geoid models are containing gross errors/mistakes whereas the criteria of ±1.645 r
developed for relatively small areas. However, a local geoid model has been used for analysing the points before accepting. Using
was developed for an area of 100 km  100 km in Botswana yield- the 2.83 r criteria, the point No. 13 was eliminated thus providing
ing an accuracy of about ±20 cm. A second degree polynomial 32 control points for further processing.
model was employed for the purpose (Manisa et al., 2016). A second exercise was taken up using the remaining 32 points
Under this back ground, this study is contrived with the objec- where the processing was carried out in a similar fashion like the
tive to develop a local geoid model by employing various polyno- first exercise through the second, third and fourth degree polyno-
mial models and thereafter to carry out comparative analysis of mials. The standard deviation/errors in this case are found to be
the efficacy of each model for defining the geoid surface which in ±0.34 m, ±0.31 m and ±0.31 m respectively for the corresponding
turn facilitates the conversion of ellipsoidal heights to orthometric models (Tables 1 and 2). So there is small improvement from sec-
ones. ond degree model to third degree model in terms of statistical
parameters including the standard error and range of errors but
there is no change in statistical parameters between the third
2. Study area, materials & methods and fourth degree polynomial. Therefore the third degree polyno-
mial is accepted for further processing and analysis.
The study area is the Madang town, the capital of Madang Another issue to be highlighted is the behaviour of four points 7,
Province of Papua New Guinea (PNG). It is about 25 sq. km 8, 10 and 12 when dealt with 2nd degree model using 32 points.
R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271 267

Fig. 1. Location of the study area and distribution of control point.

The errors for these points are in the range from ±1.645 r to ±2.83 further examined in terms their magnitude and spatial distribu-
r with exception of the point no. 10 showing an error of 0.96 m tion. For this, the 31 points are re-processed through the third
which is equivalent to the rejection limit of 2.83 r. Besides, this degree model alone where standard error is found to be ± 0.25 m
point has shown relatively higher error in 3rd degree model. with reduced range of errors. However the point No. 8 still contin-
Therefore this point is also assumed to have mistakes and accord- ued to show an error of 0.61, relatively at higher side. It is also
ingly rejected/eliminated. The magnitude of the errors for other 3 observed that the elimination of this point does not adversely
points is between 1.645 r and 2.83 r. However in this study, no affect the spatial distribution. Hence this point is eliminated,
point was eliminated on this basis but errors in these points are finally leaving 30 points for further processing. The 30 points are
268 R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271

Table 1
Computation of geoid surface using polynomial models.
Poi UTM UTM Ellips Orthom Geoid Residual with Residual with respect to third Residual with
nt Easting Northing oidal etric al respect to degree respect to
No (X) (Y) heigh height separ second degree fourth degree
t (H) (h) ation 33 32 33 32 31 30 33 32
in m from (N) = points points points points points points points points
levellin H-h
g in m
1 365442 9420672 83.28 14.50 68.78 0.63 0.32 0.63 0.20 0.18 0.18 0.61 0.20
2 365045 9423831 73.01 4.73 68.28 -0.13 -0.14 -0.05 0.01 -0.05 -0.07 -0.12 0.01
3 364165 9424082 81.78 13.32 68.46 0.17 0.16 0.00 -0.03 0.03 0.04 0.03 -0.03
4 368502 9423728 72.42 5.24 67.18 0.07 0.23 0.10 0.20 0.15 0.16 0.16 0.19
5 367637 9423003 73.11 5.91 67.20 -0.16 -0.18 -0.30 -0.10 -0.16 -0.23 -0.33 -0.10
6 367847 9422603 73.15 6.04 67.11 0.07 -0.03 0.11 0.16 -0.25 -0.21 0.10 0.16
7 367682 9422626 73.69 5.81 67.88 0.71 0.61 0.68 0.74 0.45 0.45 0.66 0.74
8 366243 9422505 71.66 4.37 67.29 -0.66 -0.70 -0.79 -0.76 -0.61 El -0.76 -0.76
9 366605 9422618 72.36 4.66 67.70 -0.12 -0.17 -0.28 -0.21 -0.09 -0.25 -0.25 -0.22
10 367798 9422292 69.37 3.19 66.18 -0.77 -0.96 -0.57 -0.77 El El -0.57 -0.77
11 365272 9421265 82.14 13.93 68.21 0.01 -0.05 0.01 -0.12 -0.08 -0.13 0.00 -0.12
12 366100 9421294 73.81 5.03 68.78 0.94 0.66 1.00 0.39 0.27 0.26 1.02 0.39
13 365920 9420920 70.57 4.45 66.12 -1.79 El -1.69 El El El -1.68 El
14 365593 9421257 73.22 4.90 68.32 0.24 0.08 0.24 -0.08 -0.05 -0.11 0.24 -0.08
15 368180 9423984 72.76 5.54 67.22 -0.27 -0.15 -0.23 -0.36 -0.32 -0.29 -0.25 -0.36
16 367209 9424119 70.82 2.56 68.26 0.17 0.15 0.24 0.10 0.08 0.12 0.25 0.10
17 367299 9422241 73.69 5.90 67.79 0.49 0.30 0.50 0.31 0.00 0.01 0.51 0.31
18 366311 9424358 68.98 0.83 68.15 -0.30 -0.44 -0.12 -0.24 -0.39 -0.32 -0.08 -0.24
19 365253 9420298 82.90 14.45 68.45 0.18 -0.20 0.11 -0.19 -0.18 -0.16 0.09 -0.19
20 365770 9423469 70.23 1.87 68.36 0.07 0.08 0.08 0.20 0.25 0.16 0.08 0.20
21 365725 9422967 71.68 3.31 68.37 0.17 0.21 0.14 0.24 0.35 0.22 0.14 0.24
22 366611 9423728 71.80 3.33 68.47 0.32 0.29 0.27 0.37 0.45 0.37 0.30 0.37
23 365177 9421839 70.60 2.46 68.14 -0.09 -0.01 -0.03 -0.06 -0.04 -0.10 -0.04 -0.06
24 365021 9421374 71.67 3.70 67.97 -0.31 -0.26 -0.25 -0.26 -0.26 -0.28 -0.26 -0.26
25 364825 9421084 74.33 5.79 68.54 0.19 0.24 0.23 0.36 0.32 0.35 0.25 0.36
26 364810 9422566 71.34 3.18 68.16 -0.14 0.07 0.07 0.02 -0.12 -0.09 0.06 0.02
27 364770 9420635 74.13 5.87 68.26 -0.16 -0.22 -0.23 0.05 0.06 0.09 -0.20 0.05
28 364968 9420656 79.67 11.55 68.12 -0.23 -0.36 -0.30 -0.22 -0.18 -0.18 -0.30 -0.22
29 365427 9420956 80.43 11.78 68.65 0.50 0.30 0.50 0.18 0.20 0.17 0.48 0.18
30 367844 9423920 70.32 2.47 67.85 -0.30 0.25 0.17 0.11 0.16 0.17 0.13 0.11
31 367364 9423761 69.97 2.06 67.91 0.24 0.07 -0.02 0.04 0.12 0.08 -0.04 0.04
32 367536 9423481 69.47 2.05 67.42 -0.43 -0.19 -0.36 -0.19 -0.10 -0.18 -0.39 -0.19
33 366678 9422069 73.04 5.21 67.83 0.20 0.02 0.13 -0.11 -0.17 -0.25 0.15 -0.11

Note: EL – Points eliminated in processing

again processed through the 3rd degree polynomial yielding stan- for developing the local geoid model whereas the 10 points are
dard error of ±0.22 m, maximum error of +0.45 m and range of kept aside as check points for assessing the accuracy of the model.
errors from 0.31 m to +0.45 m. The third degree model as well as the second degree model is used
A third exercise is carried out to examine the impact of using for the purpose. The points are selected on the basis of their spatial
reduced number of points. The third degree polynomial needs min- distribution as well as distribution in topographic relief range.
imum 10 control points. Therefore it was decided to examine 20 After processing the standard deviation in case of 20 points used
control points (two times the minimum required) and accordingly in 3rd degree model is found to be ±0.22 m. The 10 check points
the set of 30 points used above are segregated to two disjoint sets, provided standard error of ±0. 23 m. The range of errors for the
one consisting of 20 points and the other 10. The 20 points are used two cases is from 0.25 m to +0.44 m and 0.26 m to +0.34 m
R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271 269

Table 2
Abstract of statistics and error analysis of polynomial models

Degree of polynomial/—Statistical Modelling with 2nd degree Modelling with 3rd degree polynomial Modelling with 4th degree
parameters polynomial polynomial
33 points 32 points 33 points 32 points 31 points 30 points 33 points 32 points
Standard error (r) in m ±0.49 ±0.34 ±0.47 ±0.31 ±0.25 ±0.22 ±0.47 ±0.31
Range of errors 1.79 to 0.96 to 1.69 to 0.77 to 0.61 to 0.31 to 1.68 to 0.77 to
+0.94 +0.66 +1.00 +0.74 +0.45 +0.45 +1.02 +0.74
Maximum error 1.79 0.96 1.69 0.77 0. 61 +0.45 1.68 0.77
Point No.13 Pt. No 10 Pt No. 13 Pt No. 10 Pt 8: 0.61
2.83 r 1.39 0.96 1.33 0.88 0.71 0.62 1.33 0.88
Errors more than 2.83 r Pt. 13: Pt. 10: Pt. 13: Nil Nil Nil Pt. 13: Nil
1.79 0.96 1.69 1.68
Error in range 1.645 r 2.83 r & 1.39–0. 81 0.96–0.56 1.33–0.77 0.88–0.51 0.71–0.41 0.59–0.35 1.33–0.77 0.88–0.51
points in this range Pt. 12: +0.94 Pt. 07: +0.60 Pt. 12: +1.0 Pt. 7: 0.74 Pt. 8: 0.61 Pt 7: +0.45 Pt. 12: +1.02 Pt. 7: 0.74
Pt. 08: 0.69 Pt. 8: 0.75 Pt. 7: +0.45 Pt. 8: 0.75
Pt. 12: +0.66 Pt. 10: 0.77 Pt. 10: 0.77

Table 3
Geoid model– Third degree and second degree polynomial (Modelling with 20 points & Checking with 10).

Point UTM UTM Ellipsoidal Orthometric Geoidal Residuals using Residuals of 10 Residuals using Residuals of 10 Remarks
No Easting Northing height (H) height (h) from separation 20 points in check points 20 points in check points
(X) (Y) in m levelling in m (N) = H-h modelling (3rd deg) (3rd degree) modelling (2nd deg) (2nd degree)
1 365442 9420672 83.28 14.50 68.78 0.171 0.258
2 365045 9423831 73.01 4.73 68.28 0.017 0.125
3 364165 9424082 81.78 13.32 68.46 0.018 0.089
4 368502 9423728 72.42 5.24 67.18 0.194 0.168
5 367637 9423003 73.11 5.91 67.20 0.227 0.351
6 367847 9422603 73.15 6.04 67.11 0.221 0.325
7 367682 9422626 73.69 5.81 67.88 0.439 0.349
9 366605 9422618 72.36 4.66 67.70 0.215 0.291
11 365272 9421265 82.14 13.93 68.21 0.104 0.072
12 366100 9421294 73.81 5.03 68.78 0.251 0.480
14 365593 9421257 73.22 4.90 68.32 0.092 0.004
15 368180 9423984 72.76 5.54 67.22 0.292 0.110
16 367209 9424119 70.82 2.56 68.26 0.130 0.255
17 367299 9422241 73.69 5.90 67.79 0.014 0.025
18 366311 9424358 68.98 0.83 68.15 0.255 0.326
19 365253 9420298 82.90 14.45 68.45 0.185 0.224
20 365770 9423469 70.23 1.87 68.36 0.248 0.094
21 365725 9422967 71.68 3.31 68.37 0.302 0.199
22 366611 9423728 71.80 3.33 68.47 0.425 0.327
23 365177 9421839 70.60 2.46 68.14 0.052 0.011
24 365021 9421374 71.67 3.70 67.97 0.257 0.240
25 364825 9421084 74.33 5.79 68.54 0.346 0.297
26 364810 9422566 71.34 3.18 68.16 0.044 0.087
27 364770 9420635 74.13 5.87 68.26 0.075 0.130
28 364968 9420656 79.67 11.55 68.12 0.188 0.317
29 365427 9420956 80.43 11.78 68.65 0.175 0.247
30 367844 9423920 70.32 2.47 67.85 0.163 0.292
31 367364 9423761 69.97 2.06 67.91 0.085 0.096
32 367536 9423481 69.47 2.05 67.42 0.178 0.238
33 366678 9422069 73.04 5.21 67.83 0.249 0.184
Standard error (r): ±0.223 m ±0.225 m ±0.25 m ±0.23 m
Range of errors: 0.25 to 0.45 0.26 to +0.34 0.32 to +0.48 0.32 to +0.30

respectively. The standard error and other statistics of the third same. Using the rejection criteria of 2.83 times of standard error
degree model were found marginally better than the second degree one points was eliminated from the processing. After the elimina-
one (Table 3). The geoid model with 30 points and third degree tion, the 32 points provided the standard error with significant
polynomial is used to develop contours of the geoidal undulation improvement to ±0.34 m for 2nd degree model and to ±0.31 m
with 0.1 m interval, a visual display of which is provided in Fig. 2. for the third and fourth degree models.
Using the newly computed standard error at this stage a second
3. Results and discussion point was eliminated with the same rejection criteria. It is worth to
mention that the analysis of the ellipsoidal height of the two elim-
From the study, it is seen that when all the 33 points are used inated points shows an abrupt change with regard to surrounding
for developing the geoid model the standard deviation varies from points thus validating the wisdom of rejection. Besides the analysis
±0.47 m to ±0.49 m from second, third and fourth degree polyno- of standard errors and other statistics was carried out here from
mials. Also the range of errors encountered in each case is almost which it is observed that the third degree model is marginally
270 R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271

Fig. 2. A visual display of the geoid model using 30 points and 3rd degree polynomial.

better than the second one whereas the fourth degree is found to out for the same area (Rosa et al., 2016). However the second
be redundant. Therefore the final processing was done using 31 degree polynomial is marginally behind the third degree for the
points through the 3rd degree model to yield a further reduced purpose. It is worth to mention that this level of accuracy is reason-
standard error of ±0.25 m. Here one point has shown an error of ably fair considering the achievable accuracy of ±5 to ±10 cm in
0.61 m (more than 2r).This point was analysed in terms of its case of EGM 2008 (Nikolaos et al., 2012). From the experience
spatial distribution and from this it was found to be redundant obtained in this study it is recommended that number of control
hence was be eliminated. The final processing was carried out with points for third degree model may be not less than 30 (3 times
30 points and the third degree model, providing standard error of redundancy) which includes few points for checking. To achieve
±0.22 m (say ±20 cm). still higher accuracy, number of control points may be increased
The last exercise using 20 points for modelling with the third with increase in area so that it is with uniform distribution spa-
degree polynomial has provided standard error of ±0.22 whereas tially as well as topographically.
the standard error of 10 check points in this case was found to Assuming that the orthometric heights derived from ellipsoidal
be ±0.23 m That means the standard error in processing of 30 heights by using this level of geoid model will have an accuracy of
points, 20 points and 10 check points were very consistent ±20 cm, it is possible to use these heights for producing contours of
including the range of errors and maximum error. 0.70 m (say 1 m) interval as per prescribed accuracy standards. In
view of ever increasing demand of large scale maps with contour
4. Conclusion interval of 1 m, the space based technology of GNSS or GPS is going
to be a major contributor in providing precise locations including
Current study demonstrates that the geoid model developed heights.
using third degree polynomial can provide accuracy of the order At the beginning of processing there is a need to eliminate con-
of ±0.20 m, a significant improvement over an earlier study carried trol points having gross errors. For this preliminary processing of
R.K. Das et al. / Egypt. J. Remote Sensing Space Sci. 21 (2018) 265–271 271

all the control points may be used to assess the accuracy of the Fan, H., 1997. Theory of Errors and Least Squares Adjustment. Royal Institute of
Technology, Sweden, ISBN 9171702008.
model and then the points showing higher error may be eliminated
Lin, L., 2007. Application of a back-propagation artificial neural network to regional
on the basis of statistical parameters such as 2.83 r (say 3 r) or grid-based geoid model generation using GPS and leveling data. J. Survey Eng.
even 2 r. This will improve the accuracy of the geoid model and 133 (2), 81–89.
is in tune with well adopted adjustment procedures. Manisa, M., Das, R.K., Segobye, M., Maphale, L., 2016. Developing local geoid model
to assess the accuracy of orthometric heights from GPS-based ellipsoidal
heights in Botswana. Spatial Inform. Res. Korean Spatial Inform. Soc. 24 (5),
Conflict of interest 607–616. http://dx.doi.org/10.1007/s41324-016-0057-3.
Nikolaos, K.P., Simon, A.H., Steve, C.K., John, K.F., 2012. The development and
evaluation of the Earth Gravitational Model 2008 (EGM 2008). J. Geophys. Res.
There is no conflict of interest 117 (B4). http://dx.doi.org/10.1029/2011JB008916.
Omar, A., Abdulla, A., 2007. Preliminary study of the gravimetric local geoid model
in Jordan – case study. Ann. Geophys. 50 (3), 387–396.
Rosa, R., Jana, S.K., Das, R.K., Pal, D.K., 2016. Evaluation of orthometric heights from
References GPS survey using a geoid model – case study for Madang, Papua New Guinea.
Int. J. Advancement. Res. Technol. 5 (5), 9–16.
Bomford, G., 1971. Geodesy. Oxford University Press, Oxford, ISBN 9780198519195. Smith, J.R., 1997. Introduction to Geodesy – The History and Concepts of Modern
Cakir, L., Yilmaz, N., 2014. Polynomials, radial basis functions and multilayer Geodesy. John Wiley & Sons, Inc, ISBN 9780471166603.
perceptron neural network methods in local geoid determination with GPS/ Wijayratne, I.D., 2002. Survey Measurements and Adjustments, Course notes (SU
levelling. Measurement 57, 148–153. 3250), Michigan Technological University. Access on: June 2016, from www.
David, B., 2012. The Combination of Observations. Cambridge University Press, tech.mtu.edu/courses/su3250/survey%20Adjustments.pdf.
ISBN 9781107697614. You, R., 2006. Local geoid improvement using GPS and leveling data: case study. J.
Dinter, G., Illner, M., Jäger, R., 1997. A synergetic approach for the transformation of Survey Eng. 132 (3), 101–107.
ellipsoidal heights into a standard height reference system (HRS). In: Gubler E. Zhong, D., 1997. Robust estimation and optimal selection of polynomial parameters
and Hornik H. (Eds.), EUREF Publ. No. 6, Bayerische Kommission für die for the interpolation of GPS geoid heights. J. Geodesy 71 (9), 552–561.
Internationale Erdmessung, No. 58, München, Germany, ISBN 3769696212.

You might also like