Hi
I am working on a project about estimating streamflow in ungauged basins.
We have many methods to modelling river streamflow and one of them is Regionalization.
Regionalization is a process to transferring information from a gauged basin to ungauged or poorly gauged basins.
We have many schemes for using Regionalization but HEC-HMS modelling was not among them.
I have two question:
1. Can we use HEC-HMS for modelling river streamflow?
2.Can we use HEC-HMS to use Regionalization method?
Thanks
Amirhossein Haghighat,
1) Yes and No! Hec-HMS is not a river flow modeling software. It is a rainfall- runoff software, it can do limited flood routing with simple methods (hydrological modeling NOT Hydraulic Modeling) with MK or Kinematic Wave in prismatic channel methods which ignores momentum equation (or simplified version of that). For that purpose you have to use HEC-RAS.
2) You can use that but still have to do some heavy lifting manually. But you can write shell to automat that pre/post steps.
Hope it helps,
Kaveh
1. You can use HEC-HMS to model streamflow, assuming that you are just interested in modelling the hydrograph at a particularly location, and not wanting to model more complex issues like flood inundation (see Kaveh's comment).
2. This question is back to front. You would be using a regionalisation method to infer the parameters of the HEC-HMS model at ungauged sites. There are a number of regionalisation methods including similiarity (transfer the parameters for a similar catchment), regression (relate parameters to catchment attributes), using response signatures (relate things like runoff coefficient, BFI to catchment attributes and use this to constrain model parameters). The regression approach is problematic due to the noise involved and the non-linearity in the model. The similiarity approach is the simplest. Personally, I prefer the response signatures approach (see Yadav et al. 2007 Article Regionalization of constraints on expected watershed respons...
andConference Paper Regionalisation of Rainfall-Runoff Models
)
The issue here is the number of parameters you are using, and the uncertainty in the resulting prediction. You need to minimise the number of parameters that you are estimating through regionalisation (even when using the response signature approach). Otherwise, there will be a high uncertainty in the predicted values. Do you have enough gauged sites to keep some in reserve for testing the regionalisation?
Thanks Barry Croke for your perfect answer.
Actually I have read some articles about Rainfall-Runoff modelling by Regionalization, but I didn't find the application of this method on estimating streamflow in natural channel (Rivers) stations. Some people said that we can not use this method for river discharge estimating.
So it would my pleasure to know your idea about using Regionalization for river discharge estimating in ungauged stations.
On the other hand, in most studies, people used this method by considering many catchments ( 20 - 100 catchments) so the question is here that how many donor stations is necessary for having satisfactory results in target stations?
Thanks
The aim of regionalisation methods is to estimate the parameter values at ungauged locations, so that a model can be used to simulate the river discharge at that location. You could argue that regionalisation doesn't work well, and there are signficant uncertainties in the model predictions, but saying you cannot use this method for river discharge estimation is a bit too strong in my view. You can do it, the question is how good are the estimates.
As I said the simplest approach is the similarity-based approach. The complication is regarding how we measure similarity. One approach is to use several different similarity measures as see what the impact is on the model prediction. Ideally, you need a reasonable number of gauges to select from, but you can do this with a small number of gauges - the issue is that you would not be able to reliably estimate the uncertainty in the predictions.
The response signatures approach can potentially use few gauges, though again, the issue is how well you can estimate the uncertainties. The Yadav et al paper adopted an approach of estimating the likely range of values for each response signature - the tighter you can make these the lower the uncertainty in the model predictions. Having more gauges would enable you to give better estimates of the probability distribution for each response signature for the selected ungauged site. This means that the more gauges you have with good quality observations, the smaller the uncertainty in the model predictions.
The answer is that you can use fewer gauges, but the quality of the predictions will suffer. How many gauges you will need depends on what uncertainty in the model predictions is acceptable. The other factors that can influence the uncertainty in the model predictions are: uncertainty in the input climate data used by the model, the observed streamflow data, the data used to estimate the response characteristics (topography, soil, veg, geology, ...), the number of parameters being estimated, and how well the model structure performs at predicting flows in your study area. This makes it impossible to say how many gauges you would need, other than to say use as many as you can. It is also a good idea to have some gauged sites available for testing the regionalisation - these are gauges which are not used in setting up the regionalisation of the parameters, they are treated only as ungauged sites, and the results from the regionalisation compared to the observed values.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication