Summer school 2016: Problem with results in Uncertainty analysis DSS file

Summer School Activity Forum

Problem with results in Uncertainty analysis DSS file

Dear all,

Till now I could not find a solution for accessing all outflow results (from all parameter realisations) in the dss file generated from an Uncertainty Analysis in HEC-HMS. Only the characteristic results (mean, mean+ or - std. dev., etc) are available from the dss file. The data for all realisations appear in the HEC-DSSVue as available, but in fact they are not.

This persistence of this problem means that what is explained as a possible approach in section 6.1.1 of the guidance document (Approach 1) will not be possible. Please use the alternative approach (Approach 2 - explained under section 6.1.2) for the parameter sensitivity. Many of you have already realised this, I think, and have started developing the demo application with Approach 2.

Apologies for this inconvenience.

All the best,

Andreja

Dear Andreja,

I have some question related somehow with your post and related also with the assignment we have to do. In the document called "Development of a sensitivity analysis application for an event based HEC-HMS model" at page 12, paragraph 6.1.2 where is given Approach 2 it is said " The sampling of parameters may be random from a probability distribution, or ordered from the specified range". What I have understood from the UI design and from the rest of the doc is that the user introduces one value for each parameter and we run one simulation, obtain one graphic which we display. So I do not understand what is about this sampling of parameters which is again mentioned at page 13, paragraph 6.2, in the given steps to follow, at step 2

"Generate sufficiently large number of precipitation coefficients, random or ordered, within the specified range. (With these coefficients the base precipitation input will be multiplied)"

So, what precipitation coefficients should be sampled? Is not just the quantity of the rainfall which the user adjust, so finally there is only one distribution of the rainfall over hours, and for that one distribution we run the simulation?

Secondly, what means that the base precipitation must be multiplied by these coefficients?

I present below our user interface we have so far, to make more clear the approach we took in solving the assignment.

Link below shows the image

https://i.imgsafe.org/53b2525486.jpg

Any clarifications on the above questions is highly appreciated.

Thank you

Sorin

Show parent

Dear Sorin,

There is a small misunderstanding here, I think. Judging from your figure with the User Interface it seems that you are developing the application in such a way that for each parameter (or precipitation input) value the user should run the model and see the results. That is the purpose of the 'Run' button at the bottom of the application, I suppose.

However, our idea of the application was to have all the runs of the model with all possible parameter (or precipitation inputs) values pre-run and stored (e.g. in files) and the demo application would just be reading the stored results as the user manipulates the sliders to change the parameter values or precipitation inputs. This serves better the purpose of analyisng the sensitivity, because the user can immediately notice the influence - without running the model.

With this, I hope that it becomes clearer what is meant with "The sampling of parameters may be random from a probability distribution, or ordered from the specified range". In order to have the pre-runs of the model you need to select a sufficiently large number of such values of parameters or precipitation inputs. You can do that in you code by random sampling (e.g. from uniform probability distribution), or by using ordered steps of parameter values within the giving range (e.g if the parameter has range between 0 and 1, you can decide to have 200 model runs by varying the parameter in steps of 0.005: 0; 0.005, 0.01,0.015, etc till 1)

We thought that this approach would also be easier for the short period available for this summer school. However, adding the functionality of actually running the model from the user interface with a given value is also nice and can be added.

In any case - you can proceed as you find more useful, but I hope that this clarifies the misunderstanding.

All the best,

Andreja

Show parent

Thank you very much for reply, that clarifies the requirements.

Best regards,

Sorin

Show parent