Processing RADS Data 

LAXMIKANT DHAGE

COAS OREGON STATE UNIVERSITY

 

 

 

1

Declaration I hereby declare that the work incorporated in this project is original and carried out at the College of Oceanic and Atmospheric Sciences, Oregon State University.

LAXMIKANT DHAGE B Tech 6th Semester Department of Mechanical Engineering Indian Institute of Technology, Guwahati.

 

2

Certificate

This is to certify that this project is an authentic record of the research carried out by Mr. Laxmikant Dhage under my supervision and guidance at COAS, OSU.

COAS, OSU. 22 July 2010

 

Dr. Ted Strub Professor 104 COAS Admin Bldg College of Oceanic & Atmos. Sciences Oregon State University

3

Table of Contents     

Declaration Certificate Acknowledgements Summary of work 1.

Introduction

2. Methods 3. Results and Figures 4. References

 

4

Acknowledgements

It is my pleasure and privilege to express my deep gratitude to Prof. P. Ted Strub for giving me the opportunity to carry out the project under his guidance. I owe much to Dr. Corinne James, Research Associate for helping me throughout the period of this project. Lastly, I thank my parents, and family members for their moral support and constant encouragement, without which the work carried, would not have been possible.

LAXMIKANT DHAGE B Tech 6th semester Department of Mechanical Engineering IIT Guwahati.

 

5

Summary of Work Our main emphasis was on creating a precise processing strategy on RADS database and obtain smooth Sea Level Anomaly field. Methodology involves editing of each and every corrective field with some specific sequential steps. We looked carefully into each and every data field involved in the construction of Sea Level Anomaly. These steps involve applying respective limits, removing outliers and smoothing with weighted moving average. Finally smooth SLA is interpolated into standard grid points. We removed the spatial and time series average from the gridded SLA data so that deviation from mean can be observed clearly. We started our work with pass 28 off Oregon coast and then implemented similar processing steps for other passes. Processed Sea Level Anomaly data is used to study seasonal oceanic variations. We plotted monthly mean of 7 years Jason 1 satellite data for all of those passes.

 

6

Introduction

Altimetry is a technique for measuring height. Satellite altimetry measures the time taken by a radar pulse to travel from the satellite antenna to the surface and back to the satellite receiver. Combined with precise satellite location data, altimetry measurements yield sea-surface heights. Altimetry satellites basically determine the distance or range from the satellite to a target surface by measuring the satellite-to-surface round-trip time of a radar pulse. The return yields information on the global distribution and variability of sea surface height, ocean swell amplitude and scalar wind speed. The shape of the return yields significant wave height and the magnitude yields the scalar wind speed. If the satellite orbit is precisely determined and the range is corrected for a variety of atmospheric, ocean surface and solid earth factors, then these measurements allow determination of changes in sea surface height (SSH) due to tides, geostrophic currents and other oceanographic phenomena to an accuracy of 2-3 cm. Several different frequencies are used for radar altimeters. Each frequency band has its advantages and disadvantages: sensitivity to atmospheric  

7

perturbations for Ku-band, better observation of ice, rain, coastal zones, land masses ... for Ka-band. The principal objective is to measure the range R from satellite to the sea surface. The pulse interacts with the rough sea surface and part of the incident radiation reflects back to the altimeter. The range R from satellite to mean sea level is estimated from round -trip travel time by

( Ŕ - ∑ ∆Rj ), where Ŕ= c×t/2, range computed neglecting refraction based upon the free

space speed of light c ∆Rj, j=1,.... are corrections for various components of atmospheric refraction and for biases. Now range is transformed to a fixed coordinate system by precision orbit determination of height H of the satellite relative to a specified reference ellipsoid approximation of the geoid. Then it is converted to the height h of the sea surface relative to the reference ellipsoid by h= H - R = H - ( Ŕ - ∑ ∆Rj )

Accurate estimation of R and H are not sufficient for oceanographic applications of altimeter range measurements. The above sea surface height h is the superposition of a number of geophysical effects. In addition to the dynamic effect of oceanographic currents, h is affected by undulation of geoid hg, tidal variations hT and the ocean surface response ha to atmospheric pressure loading. The dynamic sea surface height is thus estimated as hd = h - hg - hT - ha = H - ( Ŕ - ∑ ∆Rj ) - hg - hT - ha

 

8

Altimeter Noise: The presence of water vapor, dry gases and free electrons in the atmosphere reduces the propagation speed of radar pulse. Additional biases in the range estimates are introduced by non Gaussian distribution of the wave field. Failure to account for the effects of the atmospheric refraction and sea-state bias introduces errors in the range measurements that are more than an order of magnitude larger than 1 cm. Therefore these corrections must be applied to transform the round-trip travel time into a range estimate with the high degree of accuracy required for oceanographic applications of altimeter data.

Atmospheric sources of error Dry troposphere: The dry tropospheric range delay varies with the amount of atmospheric mass between sensor and the surface or equivalently with sea level pressure. Correction for this delay uses the surface pressure fields produced by ECMWF. For JASON1 based on the ECMWF rms pressure accuracy of 3 mbar, the associated range error is 7 mm. Wet troposphere: The wet tropospheric range correction is designed to correct the observed altimetric sea-surface elevations for the propagation delay caused by atmospheric water vapor and cloud liquid water. Although it is smaller than the corresponding `dry tropospheric range correction,' the wet correction is more complex, with possibly rapid variations in both time and space. The correction can vary from just a few millimeters in dry, cold air to more than 40 cm in hot, wet air. Ionospheric Free Electrons: Ionospheric refraction of altimetric radar signals is determined by the dielectric properties of the upper atmosphere associated with presence of free electrons. The ionospheric range derived from dual-frequency altimeters has an error about 5 mm.

Sea-state bias: It is generated by ocean swell and mainly consists of electromagnetic (EM) bias and tracker or skewness bias. The sum of these two is called as sea-state bias. Skewness bias can be reduced by post processing, however EM bias cannot be further reduced.

 

9

Electromagnetic bias: It refers to the apparent depression of the mean sea level generated by interaction of the radar pulses with the physical wave properties. In an ocean swell field, EM bias occurs because the troughs of even pure sinusoidal waves are better reflectors than the crests, so that the mean reflecting surface is depressed below mean sea level. Tracker or skewness bias: In the calculation the algorithm assumes that the wave amplitude has a Gaussian distribution. Because the actual waveform is non-Gaussian or skewed, the tracker generates an additional negative offset. This is an instrument error and can be reduced by post processing.

Error in orbit determination: At short time scales, uncertainties in the satellite orbital position are the largest source of range error. Orbit errors divide into single pass errors, which are associated with a single range estimate, and the error associated with monthly or greater time scale averages over spatial scales of few hundred kilometers.

Environmental Sources of uncertainty In addition to the height changes generated by geostrophic flows, sea surface height is also physically altered by ocean tides and the inverse barometer effect. Tides are generated by the relative motion of Earth, moon and sun; the inverse barometer effect is the surface response to spatially variable changes in sea level pressure. Tides: Ocean tides occur at specific discrete frequencies. Tides produce elevation changes of 1-3m, and except for very large ocean waves are the largest contributor to ocean surface variability. Given tidal models, most of the tidal signals can be removed from the altimeter range retrieval, which greatly improves the accuracy of geostrophic height retrieval. Inverse barometer: Atmospheric pressure exerts a downward force on the sea surface. Spatial and temporal variations of this force are compensated at least partially by variations of the sea surface elevation. These variations of the sea surface elevation are unrelated to sea surface topographic features associated with geostrophic currents and therefore must be removed to obtain accurate estimates of the dynamic sea surface height hd. The in 

10

verse barometer correction works well over the open ocean, but breaks down in small marginal seas and in the western boundary currents. Although the inverse barometer and dry troposphere corrections are both functions of sea level pressure, they differ fundamentally. The dry troposphere correction yields an electromagnetic range delay independent of surface displacement; inverse barometer effect is a physical surface displacement. Similar to the dry troposphere correction, the inverse barometer effect is removed using ECMWF surface pressure fields. The error in this correction is about 3 cm.

RADS DATA: The RADS data base consists of one metafile and one or more data files per satellite pass. Ascending passes have odd numbers; descending passes have even numbers. The data files contain the binary data. The meta files describe the contents, for example, data type, units, creation history etc. of the data in one data file. There is only one metafile for each pass. The data files are grouped in one directory for each cycle. These cycle directories are then grouped into one directory for each mission phase, which are finally part of one directory per satellite. To read and manipulate the data, the standard netCDF tools, like ncdump (that comes with netCDF package), GMT(Generic Mapping Tools),nco(NetCDF Operators) can be used.

 

11

Methods

We started processing by selection of valid ocean data. STEPS 1] Initially we sort out the appropriate data by applying maximum and minimum limits on specific data fields. For the construction of sea level anomaly all fields with a positive option number described in RADS manual are checked against respective limits. A data field with a factor of 0 means that correction is neither added nor subtracted from sea surface height, but may be considered as an edit criterion in the construction of sea level anomaly. However a negative option means that the data field is not used for creation of the sea level anomaly, nor for its editing. Following are the limits of all the data fields having positive option number mentioned in RADS manual. (all values are in m) Altitude: >=1300000 & <=1400000 Range: >=1300000 & <=1400000 Wet_tropo: >=•.6 & <=0 Ionosphere: >=•.40 & <=.40 Ocean Tide: >=•5 & <=5 Dry_tropo: >=•2.4 & <=•2.1 Inv_baro_mog2d: >=•1.0 & <=1.0 Solid_tide: >=•1.0 & <=1.0 Ocean_Tide_got47: >=•5 & <=5 Load_Tide_got47: >=•5 & <=5 Pole_Tide: >=•0.1 & <=0.1 Sea_State_Bias_cls: >=•1.0 & <=1.0 Mss_dnsc08: >=•200 & <=200 Refframe: >=•1.0 & <=1.0 Sig_Wave_Height: >=0 & <=8 Back_Scatter_Coeff: >=6 & <=27 Wind_speed: >=0 & <=30 Sigma_Range: >=0 & <=.4 Sigma_SWH: >=0 & <=2.1 No_avg_measurements: >=15.5 & <=20.5  

12

2] Appropriate flags are turned on so that bad data can be eliminated. To select the appropriate flags we looked into each and every flag. There are total 16 flags. 1) Hardware Status ⇒ 0=OK,1=bad 2) Satellite on track ⇒ 0=OK, 1=suspect 3) Continental ice ⇒ 0=no, 1=yes 4) Quality of dual•frequency ionosphere ⇒ 0=OK, 1=bad 5) Altimeter land flag ⇒ 0=water, 1=land 6) Altimeter Ocean/Non•Ocean Flag ⇒ 0=ocean, 1=non-ocean 7) Radiometer land Flag ⇒ 1=land, 0=water 8) Corruption of Altimeter measurements ⇒ 0=OK, 1=rain or ice 9) Corruption of Radiometer measurements ⇒ 0=OK, 1=rain or ice 10) Quality of the 23.8/22/18.7 Ghz Brightness temperature ⇒ 0=OK, 1=bad 11) Quality of the 36.5/37/34.0 Ghz Brightness temperature ⇒ 0=OK, 1=bad 12) Quality of range estimate: ⇒ 0=OK, 1=suspect 13) Quality of the wave height measurements ⇒ 0=OK, 1=suspect 14) Quality of backscatter coefficient estimate/wind speed measurements ⇒ 0=OK, 1=suspect 15) Tracking mode ⇒ 0=OK, 1=coarse,acquisition 16) Quality of orbit ⇒ 0=OK, 1=degraded We turned on the following flags: 4) Quality of dual•frequency ionosphere 7) Radiometer land Flag  

13

8) Corruption of Altimeter measurements 9) Corruption of Radiometer measurements 10) Quality of the 23.8/22/18.7 Ghz Brightness temperature 11) Quality of the 36.5/37/34.0 Ghz Brightness temperature 12) Quality of range estimate 13) Quality of the wave height measurements 14) Quality of backscatter coefficient estimate/wind speed measurements 15) Tracking mode 16) Quality of orbit Then a precise Editing strategy is used. First we did impose an editing criteria both on altimeter measurements and on the correction terms.

3] After applying the maximum and minimum limits and appropriate flags turned on, the behavior of all the corrective terms are analyzed along track. Abrupt changes are assumed to be associated with erroneous data. Outliers are removed by means of 3 sigma standard deviation filter (sigma is the standard deviation of along track data). Either we can remove only outlier points or both outliers and their adjoining neighbors as sometimes adjoining neighbors may contain erroneous data. This strategy rejects much more data than the classical one, even if the altimeter measurement is useful. 4] In the next step, all the corrective terms are recomputed using a weighted average filter (weight = 1 – q6, q = d/f where d is the distance from the point of interest in km and f is 21 km). While performing this step we made sure that if there is no data continuously for two points on either side or both sides then we should neglect that outlier instead of regaining the data point using the filter. The above step insures that we don’t extrapolate the data anywhere. 5] Once all the corrective fields are edited, we formed the SLA . There is one multiplication factor for the correction, corresponding to each data field. When its factor is +1 a data field should be added to the sea surface height, when -1 it will be subtracted. When any of the fields fails due to limits or flags or 3 sigma deviation filter, no sea level anomaly is created at that point.  

14

SLA= Altitude• range • Wet_tropo • Dry_tropo • Ionosphere • Solid Tide • Ocean Tide • Load Tide • Pole Tide • Sea State Bias • Mean Sea Surface. In the end, the created sea level anomaly is checked against specific limits(-5m to 5m). Now we can either apply the same steps to the formed SLA or directly project SLA into standard CTOH grids.

Application of steps 3 and 4 to the SLA field:. I] Outliers are removed by means of 3 sigma standard deviation filter (sigma is the standard deviation of along track data). II] Removed outliers are regained using a weight average filter.

Projection of SLA field on standard CTOH grids: There were 2 ways : 1] Simply linear interpolating into CTOH grids without using the weighted average filter. We projected SLA field on standard CTOH grid points using linear interpolation. Here we made sure that we don’t interpolate the data points where there is no data on both sides for approximately 14 km from standard CTOH grid point. 2] Doing it in a single step (Projecting while smoothing) using the weighted average filter described in step 4. At each of the standard grid points smoothed data is computed by weighted moving average, instead of just linear interpolation.

Removing spatial and time series averages: To eliminate the bias, a spatial average of Sea Level Anomaly more than 2 degees offshore is removed from SLA field. Also after computing SLA for all the cycles on the standard grid, a time series average corresponding to each grid point is removed .  

15

Monthly mean of Sea Level Anomaly: Now finally, this processed and gridded sea level anomaly field is used to study the seasonal variation. We plotted monthly mean average of approximately 7 years of gridded SLA field data for different passes. While plotting we made sure that if there is no data for greater than 30%, the point is left blank instead of taking mean.        

Results & Conclusions We observed some large-scale dynamics for the passes of east coast of India compared to those for the west coast. For example pass no. 192,116 and 40 showed consistent high dynamics. For the months Jan to Apr we observed northward flow for approximately 50 km off shore and then southward flow from approximately 50 km off shore to 200 300 km off shore .There was a sudden change in the direction of flow when we looked at mean SLA for the month of May, i.e. near shore flow turned from northwards to southwards and flow from approximately 50 km offshore to 200 - 300 km offshore also changed its direction from southward to northward. This change in flow direction was consistent in almost all the passes of east coast of India. During the post Monsoon months i.e. from Oct. to Dec, instead of two regions, i.e. approximately 50 km off shore and from 50 km off shore to 200 to 300 km offshore, having different flow directions, we observed southward flow for both of the regions. Although the west coast didn’t show much dynamics, there  

16

was some pattern in flow directions for pre Monsoon (Feb-May) Monsoon (Jun-Sep) and post Monsoon (Oct- Jan) seasons. For the pre Monsoon season, we didn’t observe significant dynamics but we observed gradually northward flow when we enter Monsoon season from May. During Monsoon season flow was Northward and it starts gradually changing its direction towards southward as we enter into post Monsoon season. We observed strong northward flow during months of July, Aug and Sep for almost all the west coast passes. For the post Mpnsoon Season we observed southward flow for all the west coast passes.

 

17

Figures

 

18

 

 

19

Above figure involves Smoothing SLA with weighted average filter along  with each corrections whereas following figure involves only smoothing of 

 

20

every correction not the final SLA

 

21

Following are some plots of monthly mean SLA data fields.

 

22

 

23

 

24

 

25

References Chelton, D. B., J. C. Ries, B. J. Haines, L. Fu, and P. S. Callahan: Satellite altimetry, in Satellite Altimetry and Earth Sciences, L. Fu and A. Cazenave, eds. San Diego: Academic Press, 2001. Seelye Martin: An Introduction to Ocean Remote Sensing: Cambridge University Press 2004. RADS version 3.1 User Manual and Format Specification: Remko Scharoo, May 2010

 

26