airborne gravity data quality assessment · 2017. 5. 16. · airborne gravity data quality...
TRANSCRIPT
Airborne Gravity Airborne Gravity Data Quality AssessmentData Quality Assessment
Sandra PreauxMay 2016
7701 Greenbelt Rd., Suite 400Greenbelt, MD [email protected]
“Every careful measurement in science is always given with the probable error ... every observer admits that he is likely wrong, and knows about how much wrong he is likely to be.”— Bertrand RussellIn The Scientific Outlook (1931, 2009), 42. (Part 2 of a longer quote.)
The ConundrumHow do we estimate errors in airborne gravity data?How do we estimate errors in airborne gravity data?
• We don’t know what is correct– EGM 2008 and other models give a good estimate but if they were perfect we
would not need to measure
• You cannot actually make the same measurement twice– All repeat measurements are only approximate due to differing environmental
conditions and limitations on the ability to repeat a flight path
• Comparing to other data amounts to comparing tangerines, oranges and grapefruit?– Spectral content and areal coverage differences make comparisons with
ground and satellite data difficult– Continuation (upward or downward) are problematic and exacerbate errors
• Precision– Repeat Tracks– Adjacent Tracks– Crossovers– Grids
• Accuracy–Models– Satellite Data– Ground Data
“Although this may seem a paradox, all exact science is dominated by the idea of approximation. When a man tells you that he knows the exact truth about anything, you are safe in inferring that he is an inexact man. ”— Bertrand RussellIn The Scientific Outlook (1931, 2009), 42.(Part 1 of the longer quote.)
Precision: Repeat TracksThings to considerThings to consider• Good weather conditions for both passes• Same direction vs opposite direction– Autopilot behavior– Biases: Eötvös, scaling factors– Prevailing winds
• Position Differences– Altitude constraints– Ground track variations– Differing sample spacing
Precision: Repeat TracksSuggested MethodSuggested Method• Ensure both lines are well trimmed• Compute gravity disturbance or use 2nd order
free air adjustment to nominal altitude• Interpolate to nominal horizontal positions• Statistics of interest–Mean, RMS*, STD between tracks– Correlation coefficient* RMS is overly influence by outliers, so rarely provides a meaningful estimate of uncertainty
Precision: Adjacent Tracks• Ensure all tracks are well trimmed• Compute gravity disturbance or use 2nd order free air
adjustment to nominal altitude
Examine correlation between adjacent tracks
Treat adjacent tracks similarly to widely spaced repeat tracks. Interpolate to common points.
Interpolate to points on center data track from adjacent tracks
• A correlation length of ~15Km is a good estimate for the gravity field• Depending on line spacing you may compare several adjacent lines. • RMS, STD should increase and correlation decrease with distance
Precision: Crossovers
• Mean, RMS, STD• Identify Biases or Noisy Data• Tracks, Blocks or Between Blocks• Since we filter, crossovers need to be viewed
with some skepticism especially in regions with highly variable fields.
Precision: CrossoversMap and histogram of crossovers from GRAV-D Block TS01 over Puerto Rico, collected in 2009.
Precision: Crossovers
Precision: Grids
• Grid data and use statistics of points within each grid cell to estimate that cell’s error
• Compare Grids made from subsets of data – Oversampled along track 10x or more
(With newer 20Hz systems it will be closer to 200x)
– Assign points by track• Direction flown• Alternate tracks
Precision: Grids
• Full Field Gravity– Large values– Overall agreement easy to see– Includes altitude and normal gravity which may dominate
other features
• Gravity Disturbance– Smaller values– Larger context of gravity field retained normal gravity and
altitude effects are removed
• FFGMeasured – FFGModeled– Small deviations much more apparent– Context of larger field absent
Model Comparison: By Track
Model Comparison
Model Comparison
Model Comparison
Accuracy: Ground & Satellite
• Difficulties arise from the large differences in altitude, aerial extent and sampling geometry
• Upward continuation and filtering can provide useful comparison but must be undertaken with care.
• Comparison of models with and without each data component may offer the best evaluation–More on this will be presented on Wednesday.
δg: EGM2008 vs GOCO05S in Alaska
Conclusions• Precision– Repeat tracks, crossovers and grid comparisons all
provide useful information– No single method provides a complete picture
• Accuracy– Track by track comparisons with models (FFG, GD,
or FFG-model) provide useful information– Direct comparison with ground and satellite
measurements is difficult
Much more work is needed in this area!