Page 193 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

(b) (a)

(c)

Fig.2. Waveform re-tracking for pass 595, cycle 7 using threshold algorithm for different approach: a) mean waveform
re-tracking, b) first and original waveform re-tracking, c) Maximum correlation Approach re-tracking.

4. Sea Surface Topography Time Series and Evaluation with Tide Gauge Data

To create a time series (TS) SST and to compare it with the TG data following process have been done [25]:

 Instantaneous SST time series for each pass at each cycle using tracker and available re-tracker algorithm in Level-2
data were formed using Equation 3.
 Gross errors were detected and excluded from the Instantaneous SST time series data using 95% confidence level and
linear model (see figure-3).
 Merging median all single TS of each cycle to create a long time series of each pass.
Excluding outliers at each TS of each pass detected using a curve (Equation 17) which includes linear and quadratic
as well as trigonometric terms has been fitted to each TS at each pass and excluded using 3-sigma test [25].
2  2 

2
h(t )=a+b t +c t +d sin( t ) e cos( t ) (17)
i i i i i
T T
where h(ti) is the modeled SST at the time of ti, a, b, c, d, e are the unknown coefficients and were estimated using linear least
squares parametric adjustment method (LLSPA), T is the annual period (check figure-6).

 Since the SA and terrestrial datasets are referenced to different datum, the mean of the retrieved TS of each dataset
has been applied to SA data so all TG and SA datasets are referenced to the same datum.
 Evaluating each TS of each pass by TG data in terms of Root Mean Square Error (RMSE).
 SST time series have been selected based on the lowest RMSE at each pass and the selected TS merged together to
create the final TS in the study area. The process no. 4 repeated and compared with the TG data by RMSE.

All above mentioned process have been done for both Level-2 data (using available tracker and re-trackers) and Level-1
data (using original waveform, first meaningful sub-waveform, mean waveform per each cycle and developed re-tracking)
and compared with TG data.

6