The Road to RL05 Srinivas Bettadpur for the UTCSR GRACE Team
Framework for Product Improvement What are limiting models? What can KBR data tell us? Methodology will be refined No changes planned to flight segment Obs a priori models - PredictProcess Estimate x + Interpret Product improvements are planned Changes made during the re- processing will affect interpretation
Summary Points Changes to parametrization will continue to reduce Level-2 product artifacts. We are exploring methods for improving the Level-1B data products. What does the data itself tell us about sub-monthly variability?
Methodology: For Example
RL km smoothing
Going in right direction… 300 km smoothing
Level-1 Product Improvement Previous experience suggests candidates for further improvement. KBR1B: –Removing high-frequency noise ACC1B: –Thermal corrections –Modeling of heater spikes and other artifacts? SCA1B: –In review… Other options remain to be explored.
What can the KBR data tell us? Analyze spatio-temporal distribution of KBR data to investigate reasons for residual sub-monthly variability of the data. Typical approaches: –Remove “best-fit” orbit from KBR data. –Some kind of low-pass filtering is necessary for noise reduction. –Frequently, numerical differentiation is used to enable assessments of range-acceleration data residuals. This Study: Looks at “pre-fit” residuals only –400 days studied from May 2006 through May 2007 –Two modeling standards used: RL04 RL04 + mean-field update + annual model of variability –POD with gps, acc, att and kbr data - in daily orbital arcs
KBR Fit Statistics Improved models reduce variance dominantly in the mid-frequency range. (Not shown): The daily raw KBR1B statistics are dominated by noise at 0.26 micron/sec levels in the high- frequency bands. –This must be removed for effective “regional” study of the data. Why is there such a large day-to-day variability in the RMS contributions from each frequency band?
A Note on the Mechanics “Glitches” will typically show up as large North-South patterns in the maps drawn from this data. –This study did not use the same level of rigorous editing used in making monthly solutions. User Resources: The project provides an SOE with complete details necessary to spot and avoid these problems.
Month-to-Month Variations Each panel shows the standard deviation of “pre-fit” accelerations within each 3° bin. “Blobs” are approx ±6 nanometer/sec^2 (full scale is 9 nm/s^2). “Blobs” in the same location wax and wane through the year. What is all this?
Close-up : July 2006
Attempted Fit to Tides (N2, M2 & S2) Tides appear to account for no more than ≈2 nm/s^2 signal rms over small regions. An observation from images not shown: Fitting monthly N2 & M2 harmonics does much better at variance reduction, and across more of the blobs.
The M2 Solution
The N2 Solution
The S2 Solution For the purpose of de-aliasing, it may be sufficient to devise a tidal correction model in the data-space (range-rate) as a short-term “fix”, but…
So What is All the Rest ? Some combination of: –True sub-monthly variability –Residual mean-field errors (particularly from very small but high-amplitude anomalies) –Data “glitches” Is the energy in these features reasonable? “Blob-by-Blob” investigation of the structure is next… Reminder - None of these signals survive the monthly gravity adjustment process. –EXCEPTING the data glitches that are edited in Level-2 Processing.