salinity interpolation in corpus christi bay

Salinity Interpolation in Corpus Christi Bay

Presented by Ernest To

May 1, 2007.

Corpus Christi Bay Testbed

Montagna stations

SERF stations

TCOON stations

USGS gages

TCEQ stations

Hypoxic Regions

NCDC station

National Datasets (National HIS) Regional Datasets (Workgroup HIS)

USGS NCDC TCOON Dr. Paul Montagna TCEQ SERF

ET 20061116

Datacube diagrams

08/02/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

39

34

309

24

8

21

Salinity in Corpus Christi Bay

08/16/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

39

34

309

24

8

21

Salinity in Corpus Christi Bay

08/23/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

39

34

309

24

8

21

Salinity in Corpus Christi Bay

08/30/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

39

34

309

24

8

21

Salinity in Corpus Christi Bay

Kriging

• Kriging is BLUE (best linear unbiased estimator).• Kriging is a function that predicts value and error

estimates by using data values and their spatial configuration as inputs.

• Requires:– Stationarity, i.e. mean and variance are invariant with

translation• to support covariance modeling

– Normality• to support linear estimation

• Advanced Kriging methods can deal with non-normality and non-stationarity.

Test for Normality

Tests also performed on subsets of the data.

Examples of Variograms(Variograms for salinity data collected on 8/2/2005)

Major azimuth = N60EMajor range = 15,000 mSill = 6

Minor azimuth = N150EMinor range = 1800 m

Dip angle = 0Vertical range = 1 m

Process flowchart

Performnormality tests

(IDL)

Plot Variograms(IDL)

Perform 3Dkriging (IDL)

Visualize resultsusing voxels (IDL)

Database

Gamv.exe(GSLIB)

KT3D.exe(GSLIB)

Variogram parameters, e.g. range, sillanisotropy, azimuth, dip, variogram model, etc.

transformationparameters, λ

Probability plots

Variograms

Voxels

• Voxels = volume pixels or 3D pixels• A voxel volume is formed by superpositioning four 3D

arrays:– Red array + Green array + Blue array +Opacity array

• Manipulation of the opacity array can make inner voxels visible

Plotted with data from head.dat from IDL 6.3 examples

Kriging Results for Aug 2, 2005.

08/02/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

3934

309

24

8

21

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

3934

309

24

8

21

Kriging Results for Aug 16, 2005.

Kriging Results for Aug 23, 2005.

08/23/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

3934

309

24

8

21

Kriging Results for Aug 30, 2005.

08/30/2005

Ingleside

Port Aransas

PackeryChannel

Laguna Madre

OsoBay

1

2

12

11

10/D

14

18

3934

309

24

8

21

Space-Time Integration

35

40

45

50

8/1/2005 8/6/2005 8/11/2005 8/16/2005 8/21/2005 8/26/2005 8/31/2005

Salinity (psu) Salinity vs t

timeline

What happened in between the observations?

? ? ?

Conclusion

• Created set of tools to investigate data, perform 3D interpolation and visualize results.

• Analysis framework can be expanded to 4D and modified to incorporate data from deterministic models.

• One step towards understanding space-time integration.

Next steps

• Understand factors causing salinity patterns observed in August 2005.

• Expand on framework to perform space-time kriging

• Incorporate results from salinity models into kriging model using kriging with external drift.

Questions?

Backup slides

35

40

45

50

8/1/2005 8/6/2005 8/11/2005 8/16/2005 8/21/2005 8/26/2005 8/31/2005

Salinity (psu) Salinity vs t

Variance