Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming...

18
Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi Symposium Rome 6-10 July 2009

Transcript of Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming...

Page 1: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Numerical aspects of the omission errors due to limited grid size in

geoid computations

Yan Ming WangNational Geodetic Survey, USA

VII Hotine-Marussi Symposium Rome 6-10 July 2009

Page 2: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

• Objective: to determine the maximum grid size at which the omission error is less than 1-cm

• Omission errors are estimated by using the 3” SRTM elevation data, focus is on the spectral band corresponding to grid size 5’ to 3”

• Results are compared with those based on Kaula’s rule.

• Conclusions and discussions

Overview

Page 3: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

1. Spherical harmonic series up to ultra high degree and order Numerical difficulty to expand the potential of the topography into degree and order of 216,000 (3” resolution); Kaula’s rule can be used

2. Newtonian integrationNumerically doable. Computation only need to be extended in 1ºx1º integration area around computation point for both potential and gravity.

Methods of estimation of the omission errors

Page 4: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

• Newtonian integral

• Assumption 1: All short wavelengths of the gravity field (< 10km resolution) are due to the residual topography (isostay uncompensated)

• Assumption 2: Constant density of the residual topography

• Assumption 3: Contribution of the wavelengths shorter than 3” (<=90m) are negligible

Method used in this work

Page 5: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

• Newtonian integral

where and are the radial distances to the reference surface and the Earth’s surface, respectively.

Potential of the residual topography

drrkddrrl

GV refS

r

r

S

ref

),(2

Srrefr

S

ref

r

rrefS drrl

Grrk 2),(

Page 6: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

• SRTM3” elevation data for North and central America and the Caribbean (10°≤ ϕ ≤ 60° ; 190° ≤ λ ≤ 308°).

• It contains 13,992,031 gaps assigned an elevation value of –1.

• SRTM30/GTOPO30 30”x30” global DEM is used to fill-in the gaps.

Data Used

Page 7: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

1. Compute 5’, 2’ and 1’ bloc-mean values using the SRTM 3” elevation data

2. Using the block-mean values as reference surfaces, compute the potentials of the residual topography

3. Convert these potentials to geoid heights by using Bruns formula

4. Using the block-mean values as reference surfaces, compute the gravity of the residual topography

Computation procedure

Page 8: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Elevation along latitude band 41º24’15”. It peaks at 4157 m in the Rocky Mountains. The smaller peak is in

the Appalachian.

Page 9: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Heights after removing 1’, 2’ and 5’ block-mean values

Page 10: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Statistics of residual heights (meters)

Grid size Mean RMS value Min/Max

1’ -0.6 32.4 -342/531

2’ -0.3 44.8 -488/749

5’ 2.2 79.3 -696/1498

Page 11: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Geoid omission errors at grid size 1’, 2’ and 5’, in cm

Page 12: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Statistics of the geoid omission errors (cm)

Grid size Mean RMS value Min/Max

1’ (~1.9 km) -0.00 0.09 -0.8/0.9

2’(~3.7 km) -0.06 0.54 -1.2/8.6

5’(~9.3km) 0.07 1.08 -4.3/11.4

Page 13: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

. Results (RMS values) based on Kaula’s rule (Jekeli 2008)

This Work Kaula’s Rule. For 1’ grid size:0.1 cm 0.5 cm (Kaula)

. For 5’ grid size: 1.1 cm 3.0 cm (Kaula)

. The geoid omission errors by Kaula’s rule are several times larger

. Question: does Kaula’s rule overestimate the power of the gravity field at ultra high frequencies?

. This study provides ranges of geoid omission errors:

for 1’: -0.8/0.9 cmfor 5’: -4.3/11.4 cm

Comparisons with Jekeli’s results

Page 14: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

What are the contributions of the residual terrain to gravity at 5’, 2’ and 1’ grid size?

Page 15: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

-20

0

20

40

60

80

100

240 250 260 270 280

Longitude

-20

0

20

40

60

80

100

240 250 260 270 280

Longitude

-20

0

20

40

60

80

100

240 250 260 270 280

Longitude

Gravity of residual terrain at grid size 1’, 2’ and 5’ (mGal)

Page 16: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

Statistics of gravity of residual terrain (mGal)

Grid size Mean RMS value Min/Max

1’ 0.4 1.4 -05.4/23.6

2’ 0.7 2.4 -11.1/40.8

5’ 1.6 5.5 -18.2/114.5

Page 17: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

. The geoid omission error at 5’ grid size reaches dm in high rough mountains. It is not suitable for cm-geoid.

. The geoid omission error at 1’ grid size is less than 1 cm

everywhere in CONUS, even in very rough areas. It seems 1’ grid size is sufficient for cm-geoid.

. The omission errors are smaller than those obtained by Kaula’s rule. Overestimation by Kaula’s rule at ultra high frequency band?

. Is the assumption that the gravity field in the frequency band

5’ – 3” is purely due to the residual terrain reasonable and accurate?

Conclusions and discussions

Page 18: Numerical aspects of the omission errors due to limited grid size in geoid computations Yan Ming Wang National Geodetic Survey, USA VII Hotine-Marussi.

. The RMS values of the gravity omission errors are 1.4 and 5.5 mGal for 1’ and 5’ grid size, respectively.

. The extreme values are 23.6 and 114.5 mGal for 1’ and 5’ grid

size, respectively.

. The removal of omission errors from the gravity observations should help improving the quality of block-mean computation and data gridding.

. Constant density assumption should not change the conclusions.

Conclusions and discussions (cont.)