search

OpenHydrology / StatisticalFloodEstimationTool

wxpython based tool for assessing flood flows using current UK FEH methods
GNU General Public License v3.0
2 stars 1 forks source link

improve qmed local adjustment #18

Closed neilnutt closed 9 years ago

neilnutt commented 10 years ago

Show hiflows suitability for qmed, urbext and record length. Automatically select those suitable for qmed. Show qmed adjusent factor as stations are selected

faph commented 10 years ago

I'm currently working on the floodestimation.collections module. This module allows straightforward retrieval of all catchments suitable as QMED donor like this:

collections = CatchmentCollections(db_session)
donor_catchments = collections.nearest_qmed_catchments(subject_catchment)
faph commented 10 years ago

See #31.

CatchmentsCollections object can be used. Call find_donor_catchments() explicitly (and modify results, if desired) or just call qmed(). By default, up to 20 nearby catchments within 500 km are now used. Adjustment factors are weighted by 1/distance**3 so you can use all nearby suitable donors without converging to an adjustment factor of 1.

The nearest_qmed_catchments() currently searches for:

Question:

Should this be further limited by catchments with URBEXT < 0.03 ?

If not, the adjustment factor for a given donor is based on its ratio between the "as rural" QMED from descriptors and the observed QMED. Given that the observed is the urban one, this factor is not quite right.

If yes, why bother calculating the donor adjustment factor from "as rural" given that the "as rural" ≈ "as urban"?

faph commented 9 years ago

Today I had a real-life site where the urbext is an issue with donors. The commercial package widely used was not offering to use the nearest gauging station presumably because its urban (station 84012). When I choose the nearest rural catchment as QMED donor, it suggested that further urban adjustment was not necessarily/possible because a donor was selected. Then it went on to use a rural QMED from descriptors times the donor adjustment.

Issues with comercial package:

Suggested approach for Open Hydrology floodestimation library:

This way all factors are consistent (except possibly that it doesn't consider that urbext may have increased in time, something the FEH suggest you should), and we don't double adjust for urbanisation (nor omit to adjust).

Agree?

mikerspencer commented 9 years ago

Been a long time, but I think I used to de-urbanise donor catchments and then apply the urban adjustment factor. I recall the feh says don't bother below a certain % urban. Given the adjustment scales based on urbanisation I used to always apply it knowing it would have little effect on rural locations, that I wouldn't forget to do it on urban catchments and there would be no grey area in the middle.

Sent from my phone, apologies if there are any typos. On 30 Oct 2014 18:01, "Florenz A. P. Hollebrandse" notifications@github.com wrote:

Today I had a real-life site http://wtp2.appspot.com/wheresthepath.htm?lat=55.824218&lon=-4.300070&lz=18&rz=15&lt=OS&rt=satellite&lov=None&rov=None&lgrat=None&rgrat=None where the urbext is an issue with donors. The commercial package widely used was not offering to use the nearest gauging station presumably because its urban (station 84012 http://www.ceh.ac.uk/data/nrfa/data/station.html?84012). When I choose the nearest rural catchment as QMED donor, it suggested that further urban adjustment was not necessarily/possible because a donor was selected. Then it went on to use a rural QMED from descriptors times the donor adjustment.

Issues with comercial package:

  • It doesn't use a good ("QMED suitable") nearby catchment.
  • It suggests that no urban adjustment is required.

Suggested approach for Open Hydrology floodestimation library:

  • Use all catchments marked as being suitable for QMED estimation, whether rural or urban
  • Calculate donor adjustment factor simply as QMED from AMAX records divided by QMED from descriptors including an urban adjustment factor .

This way all factors are consistent (except possibly that it doesn't consider that urbext may have increased in time, something the FEH suggest you should), and we don't double adjust for urbanisation (nor omit to adjust).

Agree?

— Reply to this email directly or view it on GitHub https://github.com/OpenHydrology/StatisticalFloodEstimationTool/issues/18#issuecomment-61139706 .

faph commented 9 years ago

Good. De-urbanising is what the FEH says you should do and as far as I can see that's the same as urbasing the QMED from descriptors (if you're only interested in the ratio).

I agree on the "apply urban always". "Rural" is simply "low urban".

faph commented 9 years ago

Now implemented. See #34.