HOW.TO-urtap-with-Wiener

How to URTAP with Wiener filters

Gravity, Atmacs, Sea level

These notes are an excerpt from C.\Users\HGs\_Present_\NKG\GA2014\tap-results.txt

Wf-predicted series using Atmacs and the g-residual: hampered by low coherence at high frequency,
which could be a consequence of earthquakes/record gaps.
(There is little to be expected from a partial approach in the
three-set of RIOS BARL and residual. But we should give it a try.)

First, set environment parameters

cd ~/Ttide/SCGsource urtap-openend-iwf.ins

Make a backup copy:

cp o/g100202-OPNEND-1h.ra.ts tmp/g100202-OPNEND-1h.ra.ts

Two runs:     sasm06 @ sasm06-rios-atmacs.ins :BARI
     sasm06 @ sasm06-rios-atmacs.ins :RIBA
with the input swapped
and two runs
   sasm06 @ sasm06-atmacs-ra+wwfs.ins :RIOS     sasm06 @ sasm06-atmacs-ra+wwfs.ins :ATMACS
     predict-with-wwf[-l 24,24,24,24]
One

urtapt @ urtap-openend-iwf.ins :U > ! urtap.log

with dff -0.95 to produce a new PEF,
then one more with diff=0.9 applying the new PEF:

cp o/gr-1h-new-150409.pef tmp/wwf.ra-1h.pef
make-wwf-links $WDATE RCNT urtapt @ urtap-openend-wwf.ins :U > ! urtap.log

Experience from late summer 2014:

Still, we don't get the residual RMS below 8:
tslq o/g100202-OPNEND-1h-wwf.ra.ts RMS-dev=   8.0138D+00and POLI remains low (regular value not inside error bounds)

RIBA 65 <o/RIOS-from-Barlap.wwf.ts>                       -0.11182 +- 0.00719 1)          5) BARI 66 <o/Barlap-from-RIOS.wwf.ts>                        0.01367 +- 0.00463     2) RIOS 67 <d/bp-rin-oso.ra.ts>                               0.06670 +- 0.00142 1) BAWF 68 <d/atmacs-ReLo-1h.wwf.ts>                          0.28042 +- 0.01346        3) BAGL 69 <d/bag100202-OPNEND-1h.ts>                        -0.12471 +- 0.01805           4) BARL 70 <d/barlap100202-OPNEND-1h.ts>                     -0.81094 +- 0.00269        3) POLX 71 <d/PMG100202-OPNEND-1h.ts -LP|X>                  -0.00010 +- 0.00020 POLI 72 <d/PMG100202-OPNEND-1h.ts -LP|I>                   0.80790 +- 0.01719

1) complement each other
3) complement each other
4) now with the sign right, still rather small
5) SURPRISINGLY BIG, delayed response of water level (perhaps distance Onsala Ringhals?)
   While rel.accy BARL = 0.33%, accy. RIBA 6.4%
2) contribution is small, the others are equal rank:

        TsStatus Band x      SolP         RMS      SolP*RMS   Fraction ---------------------------------------------------------------------------- 65 33 RAW_SGNL RIBA . -1.1182E-01 1.4433E+01 1.6139E+00 0.003495     :: <o/RIOS-from-Barlap.wwf.ts> 66 34 RAW_SGNL BARI .   1.3672E-02 7.3570E+00 1.0059E-01 0.000218     :: <o/Barlap-from-RIOS.wwf.ts> 67 35 RAW_SGNL RIOS .   6.6703E-02 2.2300E+01 1.4875E+00 0.003221     :: <d/bp-rin-oso.ra.ts> 68 36 RAW_SGNL BAWF .   2.8042E-01 4.2039E+00 1.1789E+00 0.002553     :: <d/atmacs-ReLo-1h.wwf.ts> 69 37 RAW_SGNL BAGL . -1.2471E-01 7.4172E+00 9.2499E-01 0.002003     :: <d/bag100202-OPNEND-1h.ts> 70 38 RAW_SGNL BARL . -8.1094E-01 4.1555E+01 3.3699E+01 0.072968     :: <d/barlap100202-OPNEND-1h.ts> 71 39 RAW_SGNL POLX . -1.0313E-04 1.1979E+01 1.2354E-03      2.67E-06 :: <d/PMG100202-OPNEND-1h.ts -LP|X> 72 40 RAW_SGNL POLI .   8.0790E-01 1.1592E+01 9.3648E+00 0.020278     :: <d/PMG100202-OPNEND-1h.ts -LP|I>__________________________________________________________________________________________________________________
The last run (2014-06-30) of urtap-openend-wwf.ins was made with dff=(1,-0.8)
Saved in o-Atm-new-wwf-1h/   keyword "-new-wwf"
Predictions using Wiener filters: source predict-with-wwf
also:
      tmp-Barlap-RIOS-1h.wwf (input: d/bp-rin-oso.ra.ts )
      tmp-RIOS-Barlap-1h.wwf (input: d/barlap100202-OPNEND-1h.ts )

RESULTS: RIBA 65 <o/RIOS-from-Barlap.wwf.ts>                       -0.15135 +- 0.00601 BARI 66 <o/Barlap-from-RIOS.wwf.ts>                        0.05906 +- 0.00491 RIWF 67 <d/rios-1h.new-wwf.ts>                             1.45573 +- 0.01587 BAWF 68 <d/atmacs-ReLo-1h.new-wwf.ts>                      1.80533 +- 0.04812 RIOS 69 <d/bp-rin-oso.ra.ts>                               0.08764 +- 0.00157 BAGL 70 <d/bag100202-OPNEND-1h.ts>                        -0.21237 +- 0.01389 BARL 71 <d/barlap100202-OPNEND-1h.ts>                     -0.84472 +- 0.00251 POLX 72 <d/PMG100202-OPNEND-1h.ts -LP|X>                  -0.00034 +- 0.00045 POLI 73 <d/PMG100202-OPNEND-1h.ts -LP|I>                   0.87296 +- 0.01011
tslq o/g100202-OPNEND-1h-new-wwf.ra.ts
    RMS-dev=   5.4817D+00

The coefficients moved closer to expected values
The uncertainties became smaller at long-periods (POLI e.g. down from 0.01719) (owing to dff)
The RMS of the ra-residual became smaller (down from 8.0138D+00 !!!)

Plots:
psp-plot -> 4me/wwf_psp.png wwf_psp_lin.png wwf_pdg.png
shows major residual signal ~ M6

plot-vcv -O ~/www/4me/ -o vcv-new.ps 1,999 tt/g100202-OPNEND-1h-new-wwf.vcv
-> http://holt.oso.chalmers.se/hgs/4me/vcv-new.png
shows major parameter correlations for M4 & Sqa   POLI & Sa,I
The residual shows bad disturbances in June 2014 ( _Present_\NKG\GA2014\g100202-OPNEND-1h-new-wwf.ra.png )
which the tide-press-plot does not reveal. Mended by more careful closure of data gaps
(basis for the wwf-predicted series). There are problems with the Bubbler end of June, 2014.

There is a jump Oct. 23, 2013 ( _Present_\NKG\GA2014\g100202-OPNEND-1h-new-wwf.ra-jump.png )
We changed the coldhead on Oct. 25 after failure Oct. 24, see log files in ~/TD/RAW_o054/L11310*.054
With a BoxCar (coeff = 17.54 +- 0.22) we arrive at
    RMS-dev= 4.6570D+00

FINALLY FINALLY!
2014-07-01 Results saved in o-Atm-new-wwf-1h/
Better selection of OSO-bubbler time period and repairing the series;
reviewing the tide gauge analysis and make it run through to the open-end;
recalculating the Wiener filters
=> the lowest and smoothest residual ever: 2.28 nm/s^2
=> the expected POLI amplitude: 1.163 +- 0.007
Now we need a hell of a lot of plots.

.bye