You are on page 1of 3

Workshop 9 Solution

Uncertainty Analysis using MCMC Optimization Results


Question 1: What was the peak 1-percent flow value computed in Task B? List
the major assumptions that are part of the analysis within Task B?

The peak 1-percent flow value computed in Task B was 57,100 cfs. The
following figure shows the peak flow plotted on the flow frequency curve
computed using the Bulletin 17C procedure in HEC-SSP.

The major assumptions within the deterministic analysis are:


1) The 1-percent precipitation results in the 1-percent flow.
2) There is no uncertainty in the initial conditions.
3) There is no uncertainty in the model parameters.
4) There is no uncertainty in the storm characteristics, temporal and spatial
patterns.

Question 2: List the major assumptions that are part of the Uncertainty
Analysis?

The major assumptions within the uncertainty analysis are:


1) The 1-percent precipitation results in the 1-percent flow.
2) Assuming storms occur in the Dec 1 – Feb 28 period.

1
3) Only 100 samples of model parameters is adequate to quantify model
uncertainty.
4) There is no uncertainty in the storm characteristics, temporal and spatial
patterns.

Question 3: What were the minimum and maximum peak flow values from the
100 sample uncertainty analysis?

The minimum and maximum 1-percent flow values were 49,400 cfs and 63,100
cfs, respectively. The following figure shows the minumum and maximum values
plotted on the flow frequency curve.

Question 4: Given the treatment of uncertainty in initial conditions and model


parameters, how confident are you that model results portray uncertainty in the
1-percent flood estimation?

Following a traditional approach, a best estimate, or sometimes a conservative


estimate, of initial conditions and model parameters are used to estimate
probabilistic floods (a single deterministic result). Incorporating uncertainty in
initial conditions and model parameters is a great opportunity for modelers to
better understand the watershed and its response. Understanding when major
floods occur and the typical condition of the watershed is important. Also,
incorporating the uncertainty in the storm characteristics is a major component of
probabilistic floods. The temporal pattern, storm movement and concentration of

2
intense precipitation, and area-reduction are needed to better quantify the total
uncertainty in probabilistic floods.

Another important topic are inherent assumptions within the modeling


software/algorithms. For example, the Clark Unit hydrograph method assumes a
linear precipitation-runoff response, as precipitation intensity increases, the runoff
response increases linearly. The HEC-HMS team is adding new modeling
options that provide additional options. HEC-HMS version 4.3 has the option to
vary the Clark unit hydrograph parameters based on precipitation intensity. The
modeler can create relationships so that the unit response happens more quickly,
with less attenuation, as precipitation intensity increases. There are also reach
routing methods that can model non-linear responses in the channel.

Since we have not explored uncertainty in the boundary conditions, and due to
the limited sample size in model parameters, the model results do not adequately
quantify the uncertainty in the 1-percent flow for the Cloverdale gage. The
following presentation about integration of HEC-HMS within a HEC-WAT model
shows an option for incorporating uncertainty in meteorologic boundary
conditions.

You might also like