Open KHodgens opened 5 months ago
The consequences tables can be reorganized to be more intuitive based on the risk equation; grouped as exposure and consequences. Also it's unclear if present tables are reporting exposure for, e.g., pop >65, # structures, etc, or if that's a consequence estimate (expected annual lifeloss/structures lost/etc); need clear column labels Are the dollar values EAD? If so, need labels
Changing labels on the aggregation table should be a different issue than adding reporting by AEF and uncertainty ranges. Think of it this way - bug reports are changes to existing implementation, feature requests are adding functionality.
Ok do I need to separate and redo then?
For the reporting by AEF, please describe the "telling the story" process. What is the user functional role? (e.g. hydrologic engineer, economist), what information would be assessed from the system (e.g. geospatial distribution of flood, geospatial distribution of flood, extent of flood at the 100-year level), and how are they used?
Ok do I need to separate and redo then?
Since this thread is already established, it's for reference for future issues.
Functional roles: planning tech lead, economist, and engineer
Viewing tabular consequences by AEP assists with a 'tipping point analysis' to indicate when the system is overwhelmed. The tipping point may be best exploited by graphing EAD (or filtered component, e.g. commercial or residential infrastructure) vs. AEP.
The tipping points can be roughly tied to the HSI topographic profile versus SWL hazard by AEP comparison to roughly estimate the length over which flooding occurs, thus the magnitude of an intervention.
Can you describe how the "uncertainties" will be used. Specifically, what distribution and how are they used. Currently the system is modeling the value across the AEF distribution which is good for visualizing the AEF curve or aggregation at a selection of the cross-sections. This would be insufficient if we're looking to do statistical tests.
CHS outputs uncertainty bands by AEP for SWL and Hs. Consequence calcs can be performed with these data to provide a range of EAD for a given AEP. It is desirable to report ranges with the scoping tool to reflect the uncertainty in the decision making process.
Consequence calcs can be performed with these data to provide a range of EAD for a given AEP.
The current modeling supports only the expected scalar EAD for each AEP cross-section. The AEP distribution captures the variability across flood depth levels. Is the "range of EAD" indicative of another distributional dimension and what variability does that distribution captures?
Present formulation provides for the best estimate value. A damage distribution for each frequency can be generated using confidence limit values similar to the SWL hazard plot with uncertainty:
This could be the basis for FDA-like functionality as well... @HenryGeorgist @trietmnj @rnugent3 do you all agree?
@KHodgens can you provide a reference to the mentioned FDA functionality?
@HenryGeorgist can oakie model the stochastic bounds around the expectation at each AEF cross section? Looks symmetrical, is it just a norm dist?
@trietmnj , no I can't, it's an anecdotal statement from what I've heard in our discussions which is why I asked you fine folks if you agree.
@trietmnj , no I can't, it's an anecdotal statement from what I've heard in our discussions which is why I asked you fine folks if you agree.
@HenryGeorgist @rnugent3
@KHodgens it sounds to me that you're looking for EAD with uncertainty, which is quite different and more complex (Monte Carlo) than running three hazard scenarios for a lower bound, mean, and upper bound.
I think that you could absolutely calculate an EAD estimate for the lower bound, mean, and upper bound of the hazard, but I'd be very careful to display that you're showing EAD for different scenarios, not EAD with uncertainty.
@rnugent3 Are those hazard curves scenario-based? The bounds on the graph look like there is a distributional assumption extending out from the expectation.
the bounds of the hazard values are assumed normal. For SCT, I do intend on running stratified monte carlo (stratified on the hazard, but naive on the consequences) to construct an ead with uncertainty for each structure that should be better than scenario analysis like @rnugent3 hints. I would bin the error distribution on hazard into a set of uniformly derived probability slices (on the uncertainty) and sample consequences for each bin on the error in hazard, for each frequency and combine the result.
I worry about calling it EAD with uncertainty while being naive in the economic uncertainty.
Uff, more normal distributions...
Contact Details
No response
Feature request item
...
Is the feature related to problem
The consequences tables can be reorganized to be more intuitive based on the risk equation; grouped as exposure and consequences. Also it's unclear if present tables are reporting exposure for, e.g., pop >65, # structures, etc, or if that's a consequence estimate (expected annual lifeloss/structures lost/etc); need clear column labels Are the dollar values EAD? If so, need labels Consider also reporting by AEP and giving the user the options to filter data so they can tell the story Include uncertainty so ranges are displayed in tables rather than best estimate values.
Priority of requested feature.
Medium
Feature alternatives
none
Additional context
none