# PowerPoint Presentation

Document Sample

```					 Exercise 8.1e: Assess the likely importance of
potential new data to the predictions using dss,
css, and pcc (Book, p. 204-205)

Two potential
new
observations
to be collected
under
pumping
conditions

River gain or loss   Hydraulic head in layer 1,
along entire reach   row 9, column 18
Exercise 8.1e: Assess the likely importance of
potential new data to the predictions
Calculate dimensionless scaled sensitivities (dss) for the
two potential observations and calculate parameter
correlations (pcc) that include these observations.

Use the dss and pcc to evaluate potential improvement in
precision and uniqueness of the parameter estimates
under pumping conditions.

Do Exercise 8.1e (p. 204) and the Problem, including answering
Question 3: Is it worth waiting for new data under pumping
conditions?

Composite scaled sensitivity (css )

Absolute value of prediction
8.1e:
A100x      A100y      A100z

scaled sensitivity (pss )
css                                                     30

pss, dss,                                                            1                                                                     20

css                                                                                                                                     10

0                                                                     0
HK_1   K_RB     VK_CB    HK_2      RCH_1   RCH_2 POR_1&2 POR_CB
60                                          Parameter Name
css
dimensionless scaled sensitivity
Absolute value of composite or

dss, potential flow
40

30

20

10

0
HK_1   K_RB     VK_CB   HK_2     RCH_1     RCH_2
Parameter Name

Do the potential observations provide information about
parameters important to the predictions? By this analysis, no
Recall results of Exercise 8.1b
HK_1      K_RB       VK_CB     HK_2     RCH_1     RCH_2

This analysis        HK_1      1.00      -0.40      -0.90    -0.93     0.96      -0.90

showed that the      K_RB                 1.00      0.20      0.34     -0.32     0.32
predictions          VK_CB                          1.00      0.97     -0.97     0.97
depend on unique     HK_2             symmetric               1.00     -0.99    0.996
estimates of
RCH_1                                             1.00      -0.98
parameters that                     Observations only
are NOT uniquely     RCH_2                                                       1.00

estimated using              HK_1      K_RB       VK_CB     HK_2     RCH_1     RCH_2
the existing
observations.        HK_1      1.00      -0.16      0.078    -0.22     0.71      0.26

K_RB                 1.00      -0.61    0.013     0.25     -0.070

VK_CB                           1.00     0.33     -0.19     0.28
Tables 8.4 & 8.5,   HK_2             symmetric               1.00     -0.52     0.83
p. 199
RCH_1                                             1.00      -0.30
With predictions
RCH_2                                                       1.00
Exercise 8.1e:pcc
HK_1      K_RB       VK_CB     HK_2     RCH_1     RCH_2

Recall: Exercise          HK_1      1.00      -0.40      -0.90    -0.93      0.96     -0.90

8.1b showed that          K_RB                 1.00      0.20      0.34     -0.32      0.32
the predictions           VK_CB                          1.00      0.97     -0.97      0.97
required unique           HK_2             symmetric               1.00     -0.99     0.996
estimates of some
RCH_1                                              1.00     -0.98
parameters that                          Observations only
cannot be uniquely        RCH_2                                                        1.00

estimated with                    HK_1      K_RB       VK_CB     HK_2     RCH_1     RCH_2
only the existing
observations.             HK_1      1.00      -0.33      -0.43    -0.96     0.93      -0.89

K_RB                 1.00      -0.42     0.21    -0.060     0.10
Do the potential          VK_CB                          1.00      0.54     -0.50     0.53
observations help         HK_2             symmetric               1.00     -0.92     0.95
solve this
problem?                  RCH_1                                             1.00      -0.95
With potential observations
RCH_2                                                       1.00
By this analysis, yes!
opr statistic (percent decrease in

100
prediction standard deviation)

90
80                                                             Figure
70
60
8.12,
50                                                             p. 207
40
30
20
10
0
Prediction name

By the OPR analysis, are the potential observations important
to the predictions?

Are there other new head locations we should consider?
Results
Average OPR statistic at 100
years (averaged over x, y, z
transport directions):
Percent reduction in prediction
standard deviation caused by
observation at a model node in
layer 1.
Figure 8.12, p. 207
Where are the best locations for collecting additional
Is the new data worth waiting for?

Our analysis clearly shows that with the
new data we are likely to be able to
estimate more uniquely parameters
important to predictions.
The government officials want to see
the results of the uncertainty analysis
before they decide

```
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
 views: 0 posted: 2/12/2013 language: English pages: 8
How are you planning on using Docstoc?