Written by Brian Wahlin, Ph.D., P.E., D.WRE | WEST Consultants
Copyright © The RAS Solution 2016. All rights reserved.
In an earlier blog post, we discussed some details regarding modeling overflow gates in HEC-RAS.
In an earlier blog post, we discussed some details regarding modeling overflow gates in HEC-RAS.
These gates are typically used in irrigation canals. This blog post generated many interesting questions and comments regarding modeling
irrigation canals in general using HEC-RAS. Thus, we felt it might be
interesting to address some of these questions in a blog post devoted solely to
modeling irrigation canals and structures using HEC-RAS.
One of the questions from the overflow gates blog post was:
“I have a question about external
data interface for canal automation (EDI). I wanted to know if the project was
success full [sic] and could simulate
the real canal system.
I have seen in some articles that
using Saint Venant equations is too complicated for canal automation purposes
and needs lots of data. Instead they suggest to use industrial control system
approach. Furthermore, in a book which has been published by ASCE http://www.asce.org/templates/publications-book-detail.aspx?id=6965,
they suggest that to use similar approaches. I want to know your opinion.”
This question covers many topics from modeling irrigation
canals using HEC-RAS to designing automatic controllers for irrigation
canals. It also references a project we
completed several years ago where we linked an unsteady HEC-RAS model of the
Central Arizona Project (CAP) Canal to a Supervisory Control and Data
Acquisition (SCADA) program so that the operators of the CAP Canal could be
trained using a hydraulic model rather than by operating the real canal. CAP has an extensive SCADA system that allows
them to monitor and control the entire 336-mile canal from their central
control room. 
This is similar to the concept of how pilots are trained using a
flight simulator instead of flying a real plane. In a way, we created a “flight simulator” for
canal operators. To make this project
successful, we had utilized many of the HECRASController routines and had to
create a program that linked HEC-RAS to the SCADA system. We called this program the External Data
Interface (EDI). We are also pleased
that the question references the newly published ASCE Manual of Practice (MOP)131 Canal Automation for Irrigation Systems. This new publication is intended to be an update on recent advances in
canal automation. On a side note, I was
one of the editors for MOP 131, so thanks for the plug on the book!
In reading this question, we first need to clarify
something. As the reader points out, it is true that using the St. Venant
equations to design a controller is not a good idea because of the
non-linearity of those equations. Thus, controllers are typically designed using
some sort of linearized approximations of the system (i.e., industrial control
system approach).
However, when testing the performance of a designed
controller, it is no problem to test them via simulations using the full St.
Venant equations using unsteady HEC-RAS. In fact, you want to do this to make
sure that the designed controller (designed using simplified linear
approximations) will behave as expected in the real nonlinear world.
The EDI program that the reader refers to did work
successfully. We were able to model the unsteady hydraulics of the CAP Canal
successfully and had calibration data to verify the model performance. In this
particular project, we did not actually design feedback controllers or test the
controllers via hydraulic simulation as that was not the goal of the project.
Instead, we used the EDI as an interface between the HEC-RAS model of the canal
and the SCADA system used to manually monitor and control the canal
(manual-centralized control). As mentioned earlier, the EDI program was crucial
in making the canal operator “flight simulator” work. A schematic diagram of
how the “flight simulator” for irrigation canals is set up and how the EDI
program fits into this scheme is shown below. Note that the SimSuite program in
the figure is a program that emulated field hardware and maintained physical
information such as gate positions, water levels, and turnout flows. SimProctor is a program used to assist in the
training of the canal operators.
After posting the above response, the reader came back with
a series of questions. So, let’s go
through these follow-up questions:
1. “I was wondering if there is any publication available for the methodology which has been implemented in this project.”
Yes, the results of our “flight simulator” project for the CAP was published in the proceedings for a US Committee on Irrigation and Drainage conference a few years ago. We will also be speaking about this topic again at the upcoming ISHS conference in Portland in June 2016. A link to the ISHS conference paper is given here: https://drive.google.com/file/d/0B0bpiyLiUeRXRzBoUW9IRlpLQ28/view?usp=sharing
2. “How did you calibrate the model?”
CAP collects data for the canal on a regular basis. These data include water depths, gate openings, and flow rate estimates. We used this information to adjust gate discharge coefficients and Manning’s roughness coefficients to perform the calibration of our model of the canal. If you are doing an unsteady calibration, you will want to have the initial conditions start off at a steady state condition.
3. “Did you calibrate the pools one by one or calibrated [sic] couple of pools together?”
Adjusting the gate discharge coefficients and Manning’s roughness coefficients can affect the water levels upstream and downstream if the gates are submerged. Thus, we adjusted the gates one at a time, but always checked to make sure that the entire system calibration looked good as we were moving from one gate to the other.
4. “How did you model turnouts in HEC-RAS? Did you
have instability issues in the turnouts?”Yes, we modeled turn outs. The turnouts on the CAP Canal are locally automated to deliver a constant flow rate to the turnout regardless of the water level or flow rate in the canal. Thus, the turnouts were modeled as constant outflow hydrographs (unsteady). If you are doing a steady model, the turnouts would be modeled as a flow change location. If you want your turnout to have the flow vary as a function of head in the canal, then you can place a lateral structure where the turnout is located and then put a gate on that lateral structure. RAS will then calculate the flow through the gate based on the head from the water surface in the main canal. If you are not modeling the water after it leaves the turnout, just have RAS remove the water from the system. If you are modeling the water after it leaves the turnout, you will have to create another reach in RAS to model the lateral canal that the water from the turnout flows into. You should not see instability issues due to turnouts. The photo above shows a typical turnout structure on an irrigation canal.
5. “Can HEC-RAS model water levels with +/- 5
centimeter accuracy in irrigation canals?”
That really depends on how accurate your data are. If you have poor information about your canal (e.g., you don’t have as-built plans or there has been settlement in the lining or there is lots of seepage), then you may not be able to get 5 cm accuracy. Probably the largest uncertainty in the canals will be the flow rates. You most likely will not know the flows in the canal to with +/- 5% unless you have very accurate flow measurement devices installed. One suggestion would be to run a sensitivity test on the canal by varying the flows to get an idea regarding how much the water surfaces will fluctuate due to the uncertainty in your flow rates.
Typically in an irrigation system, the water level upstream of a check gate is the “most important” depth. The predicted water level at this location heavily depends on the gate discharge coefficient and the Manning’s roughness coefficient. So if you do not know the gate discharge coefficient (and don’t have any calibration data to determine what it should be), then the water levels upstream of the check can be off by quite a bit. The plot below shows the calibration results for an irrigation canal, where the diamond shapes indicate observed water levels. You can see from this plot that our calibrated model agrees well with some of the observed data points. For other data points, there is some difference between the computed results in HEC-RAS and the observed data. Most likely, the areas where our calibration did not agree well were a result of the uncertainty in the observed data or flow rates.
That really depends on how accurate your data are. If you have poor information about your canal (e.g., you don’t have as-built plans or there has been settlement in the lining or there is lots of seepage), then you may not be able to get 5 cm accuracy. Probably the largest uncertainty in the canals will be the flow rates. You most likely will not know the flows in the canal to with +/- 5% unless you have very accurate flow measurement devices installed. One suggestion would be to run a sensitivity test on the canal by varying the flows to get an idea regarding how much the water surfaces will fluctuate due to the uncertainty in your flow rates.
Typically in an irrigation system, the water level upstream of a check gate is the “most important” depth. The predicted water level at this location heavily depends on the gate discharge coefficient and the Manning’s roughness coefficient. So if you do not know the gate discharge coefficient (and don’t have any calibration data to determine what it should be), then the water levels upstream of the check can be off by quite a bit. The plot below shows the calibration results for an irrigation canal, where the diamond shapes indicate observed water levels. You can see from this plot that our calibrated model agrees well with some of the observed data points. For other data points, there is some difference between the computed results in HEC-RAS and the observed data. Most likely, the areas where our calibration did not agree well were a result of the uncertainty in the observed data or flow rates.
6. “What parameters did you put from [sic] designed controller in HEC-RAS?”
Feedback controller design for irrigation canals in a complex topic. We like to design the controllers for an irrigation canal using a linear approximation of the real canal using something called the integrator-delay model. Using this model, the only two parameters needed to design the controller would be the delay time and the backwater surface area behind the pool. Both of these parameters can easily be obtained using your HEC-RAS model. These parameters are needed for each pool in your system. Of course, these parameters can change as the flow rate changes, so if you design your controller for one flow rate and operate it at a completely different flow rate, the controller performance will not be as expected. For canal pools that are completely under backwater, controller design is more complicated because of resonance that results from reflection waves. Details on modeling canals for controller design can be found in the ASCE MOP 131. To actually design the controller, you can use techniques like Linear Quadratic Regulator or Model Predictive Control. All of these topics are described in the ASCE MOP 131.
7. “Is it possible to use such approach for earthen
canals? Do I need to re-calibrate the model after couple of years due to
sediment and weeds?” Yes, you can use HEC-RAS to model an earthen canal. Sediment and weeds can influence the Manning’s roughness coefficients. Thus, the canal may respond differently over time due to weed growth and sediment, so this is something that you should consider. The figure below shows weeds growing along the banks of an earthen irrigation canal as well as along the bottom of the canal. These weeds were seasonal and only appeared in the summer months. Another issue would be canal seepage in earthen canals. If you performed a seepage test on a canal pool, you could get an estimate of the rate of water loss from the pool. You could then model this seepage from the canal using unsteady outflow hydrographs or possibly the groundwater interflow internal boundary condition. Without performing a seepage test, you would really need very good (and detailed) calibration data to determine the seepage losses.
8. “Do you have any instability issues due to
sudden opening/closing of the gates?”
You can have instabilities any time you make a sudden change in RAS while using unsteady flow. However, we have typically been able to get around these problems by adjusting the time step, the cross-sectional spacing, and the HTABs. The CAP Canal model we mentioned had an inverted siphon immediately downstream of the check gates. When the gates were slammed shut, the siphon would dry up and crash the model. In this unique situation, pilot channels helped eliminate the instability.
9. “Have you built any physical model in a lab for
irrigation canals and made a HEC-RAS model of the same canal and then compared
the results?”
No, we have not done that and do not know of any situations where this has been done.
10. “Is it possible to use HEC-RAS to model sediment
transport in irrigation canals? Is this model reliable?”
Yes, it is possible to use HEC-RAS for sediment transport in irrigation canals. A sediment transport model for an irrigation canal would be just as reliable as any other sediment transport model, which means it really depends on how the model is set up, what data are available, etc. Care needs to be taken to use the correct sediment transport function for the canal. Alternatively, there is a model developed specifically for sediment transport in irrigation canals called SETRIC, which was developed by IHE in the Netherlands. If you are concerned about modeling the sediment transport in an irrigation canal, SETRIC may be a good option for you.
Yes, it is possible to use HEC-RAS for sediment transport in irrigation canals. A sediment transport model for an irrigation canal would be just as reliable as any other sediment transport model, which means it really depends on how the model is set up, what data are available, etc. Care needs to be taken to use the correct sediment transport function for the canal. Alternatively, there is a model developed specifically for sediment transport in irrigation canals called SETRIC, which was developed by IHE in the Netherlands. If you are concerned about modeling the sediment transport in an irrigation canal, SETRIC may be a good option for you.
