The spectra of GRACE (2013-14)
Hughes and Williams 2010 (PDF) treated sea level spectra as light spectra. I and my co-authors applied their methodology to GRACE, revealing the first “full color” GRACE maps and comparing them to the Noah/GLDAS water storage model.
Some maps of GRACE data:
RMS= All freqs, like B/W pic
Annual= Only 1 freq, like a blue filter
Full color maps Tues. at AGU poster G23A-0778.— Bryan Killett (@BryanKillett) December 5, 2013
Map GRACE power spectra onto visual light spectrum to see GRACE data in full color. #AGU13http://t.co/X2WN8b8m0L pic.twitter.com/id6xyLLbZD
— Bryan Killett (@BryanKillett) December 9, 2013
Compare prev. colorful GRACE spectra to these Noah/GLDAS water storage model spectra. #AGU13http://t.co/afKr1sMKzx pic.twitter.com/zKUNuVF2eV
— Bryan Killett (@BryanKillett) December 10, 2013
Annual dominates GRACE spectra. Except in Paraguay, which has a ~1.5 year cycle. Why? #AGU13http://t.co/afKr1sMKzx pic.twitter.com/sRzSWXUdjH
— Bryan Killett (@BryanKillett) December 11, 2013
Paraguay is red in Noah/GLDAS, so even slower than GRACE's ~1.5 year cycle. Itaipu dam? #AGU13http://t.co/afKr1sMKzx pic.twitter.com/7gpTzQR247
— Bryan Killett (@BryanKillett) December 12, 2013
Update: Here are new images (68 MB) and output data (421 MB).
- Newer GRACE spectra (“SS76y” on bottom row) are less colorful than the preliminary GRACE data used in the 2013 AGU poster (“SS23x” on top row). This indicates more wide spectrum power is present in the new GRACE spectra, which makes them look more similar to GLDAS spectra. Also, a new triple colorscale (modeled after the colorscales in Hughes and Williams 2010 figure 1) shows the spectra of light curves with three different widths; wider light curves make the colorscale more faded and “pastel”. The triple colorscale is unfinished; still need to decide how to automatically and objectively normalize three different colorscales- right now the colorscale on the left (with the most narrow light curve) is too dark.
- This experiment tests another way to detect tides: map the acceleration spectrum (without inverting for surface mass) immediately around a tidal frequency (O1 here) onto the visible light spectrum. The resulting map of O1 has a spatial pattern that roughly corresponds to that of the O1 tidal acceleration map obtained by using least squares to find the best-fit O1 acceleration amplitudes.
- GRACE spectra based on the older “SS23x” GRACE solutions (top row) are compared to GRACE spectra based on CSR Tellus, JPL Tellus and GAIA’s prototype monthly GRACE maps.
In the Hughes and Williams 2010 appendix, matrices 4 and 5 should be swapped, as well as the two parts of equation 7.