SNTEMP (In)Frequently Asked Questions: Other Models

Q32. I have been a 12/90 document put together by Timber, Fish, and Wildlife and the Washington State Dept. of Natural Resources. It is entitled "Evaluation of prediction models and characterization of stream temperature regimes in Washington." (TFW-WQ3-90-006). This document compares different stream reach and basin models. The model USF&WS SSTEMP, version3.3, is a deal less accurate than a model known as TEMPEST (both stream reach models). Are you familiar with the model TEMPEST? Have there been any further studies looking at the accuracy of SSTEMP, version 9?

A32. - Yes, I am familiar with the TFW report, but not the TEMPEST model per se. As I recall, TEMPEST was more accurate, especially for maximum temperature prediction, than SSTEMP. The comparison of minimum temperatures is not truly relevant, as SSTEMP doesn’t really "predict" minimum temperatures; it just gives a heuristic guess. Both models work well on mean daily temperature, but TEMPEST seems to perform quite well, especially on small headwater streams, which was their objective. I do not know the general availability of TEMPEST, but recommend that you contact Kate Sullivan at Weyerhaeuser, sullivk@wdni.com, 206-924-6191, if you want to know more. [Added 12/2001]

Q45. You are probably not totally unbiased, but if you are trying to study the effects of reservoir release temps and flow regimes on downstream temperature, is there another model that comes close to doing the job? Or is SNTEMP the clear and obvious choice (assuming you have the budget to get the data for it)?

A45. There are numerous temperature models. Basically, they all work the same way. It’s more a matter of what’s in them and what isn’t. SNTEMP’s significant strengths are, in no particular order: 1) documentation, training, maintenance, and USE, 2) the shading algorithm, 30 the statistical tools for filling, smoothing, and goodness-of-fit calculations, and 40 requirement for mean daily values only. SNTEMP’s significant weaknesses are 1) the estimation of maximum daily temperatures and lack of minimum daily temperature calculation, and 2) the non-Windows user interface, by which I include the cumbersome nature of the data files, lack of integration of utilities, etc.

Qual-2e is a good model. Last I looked, it did not include shade (though you could cobble it in) and essentially required a finer time step on the meteorology, e.g., 3-hour. The dynamic models that are highly touted these days all require this finer meteorology. In addition, they require substantial computer time and much more detail on the hydraulic cross section. A head-up comparison between SNTEMP and CE-RIV-1 (I think it was) showed a slight improvement in modeling, but with a huge cost.

The other simpler models have merit too. TEMPEST and TEMP-86 may be better on small, forested headwater streams where a substantial portion of the flow is actually sub-surface. I’m not sure about this, however. The head-up comparison that was done between SNTEMP and TEMPEST showed a slight margin for TEMPEST, but I was never convinced that they ran SNTEMP as it should be run – always a problem.

My conclusion from watching developments is that little has changed in the last decade. If you know a program and its foibles, there is no compelling reason to switch. Someday, someone will have the $$ to develop a model that is easier to use. But until that time, they all perform about the same. [Added 12/2001]

Q70. A question on SNTEMP: I was conversing with a local temperature modeling expert, (Bob Beschta) who uses other models, and is mainly concerned with logging/grazing effects on shade, and the effects on stream temps. Folks from certain industries like to argue with him that shade is unimportant, because "the air temperature is high, and all that heat is transferred to the stream, so even if we let the trees grow up, the streams would still get too hot for salmon". I know that air temp is probably "numero uno" for SNTEMP sensitivity, but Bob’s opinion is that heat transfer from air to water is relatively unimportant, compared to direct solar energy hitting the stream. Any comments?

A70. On the solar issue, there are more and more people giving this line and they are probably right to some degree. Solar is relatively unimportant for mean daily temperature, but is definitely important for daily maximum temperatures, especially on small, shallow streams. I believe that SNTEMP underestimates the effect of solar on maximum temps given the empirical approach Theurer chose to implement, but I don’t know this for a fact. Try the sensitivity analysis portion of the new SSTEMP (once you get it) turning the solar radiation "on" and see what you think for maximum temperatures on a small shallow stream. I think you will find that solar is third behind air temp and humidity, but as usual, it depends. Also, don’t forget that solar is dependent on shade, i.e., you must in effect add the sensitivity of shade to that of solar, depending on how you look at it. I wish I could tell you more. The folks that are more discretely modeling the max daily temps are learning a lot. [Added 12/2001]

Q144. Other than SNTEMP, are you aware of any models that are well-geared toward the T vs. Q relationship? I have been hearing about a HeatSource model that DEQ runs here, but have no experience with it.

A144. The CE-Qual-W3 model is likely very good, but very data intensive. Qual-2E might work well if you have no or little shading. I have seen one or two iterations on HeatSource, but have no real experience with that model. It might be fun to try, especially since DEQ loves it. It is likely very good for maximum daily temperature, especially on such a short reach. Most of its development was on very short reaches. [Added 6/2002]

Q145. Do you have any suggestions for models to use for conditions near the freezing point?

A145. No. Chances are none of the models deal with the changing state of water near zero and I really don't know how serious this is. I do know that SNTEMP doesn't hesitate to calculate equilibrium temperatures below zero. This can screw you up if you are doing any filling or smoothing. Same problem in terms of trying standard regression equations. [Added 6/2002]

Q146. Do you have any other models that you recommend that can handle shading alternatives and will predict extremes and daily cycles?

A146. No, I do not. I believe that with additional calibration (the so-called a0, a1, a2, a3 parameters) that you would be able to calibrate the maximum temperature, but the rate of change of the diurnal cycle would not be possible to get out of SNTEMP directly. It does print out two values, the K1 values for mean and max temp, that give a hint about the rate of change of thermal loading, but I cannot do an adequate job of explaining them to you. You would need to examine IP#16.

The state of Oregon has a model that may be better for your purposes called HeatSource. Look at or around http://www.deq.state.or.us/wq/tmdls/wqanaltools.htm. I am not sure of the availability or support of this program, however. [Added 6/2002]

Q147. I am a graduate student and I am doing research on modeling water temperatures of streams for a project on the effects of global climate change. We hope to use the model SSTEMP and I would very much appreciate if you have time to answer a few questions about the model. Is there a version available that would allow the input of multiple data sets, such as a file containing a month's worth of data? I know that the SSTEMP version available for download from the web allows entry of only one data set at a time.

If there is no version of SSTEMP that allows for entering multiple data sets, do you know if SNTEMP have this capability? Might you be able to suggest an alternate model that might work well for modeling daily stream temperature given inputs of climatic, meteorologic, and physical stream variables?

A147. You are correct that SSTEMP does not allow multiple-day simulations. SNTEMP does, but is much more difficult to learn. The state of Oregon has a somewhat more detailed spreadsheet-based model like SSTEMP, called HeatSource that allows multi-day (in fact multi-hour) modeling. I am not sure of its availability, but if you poke around on the web, you will probably find it. [Added 6/2002]

Q158. Please tell me more about HeatSource.

A158. Here is a response I got from Oregon DEQ:

Heat Source is the primary temperature model that Oregon DEQ uses in TMDL development. Prior and current versions of Heat Source are been capable of modeling stream networks for one day at a high spatial and temporal resolution (100-foot distance step and 1 minute time step). Under a Federal Grant, Heat Source is being upgraded to enable simulation of various days and seasons, with dynamic inputs. In addition, this new version is planned to have a user manual associated with it (past versions have lacked).

For more detailed information on Heat Source methodology and use, it is helpful for people to review TMDLs that have been submitted to and/or approved by EPA. For example, the Tualatin River Subbasin TMDL, the Umatilla River TMDL, the Draft Upper Klamath Lake Drainage TMDL, and others are available for download at: http://www.deq.state.or.us/wq/TMDLs/TMDLs.htm

The following link contains an academic peer review of Heat Source:
http://www.deq.state.or.us/wq/HeatSource/HeatSource.htm

The following link is for the download site of GIS tools (TTools) that are used to assemble the input database for Heat Source:
http://www.deq.state.or.us/wq/TMDLs/WQAnalTools.htm

Hopefully this helps answer some of your questions.

Brian Kasper
Oregon DEQ - Watershed Management Section
811 SW Sixth Avenue
Portland, OR 97204
503-229-5074
kasper.brian@deq.state.or.us

Q159. Are you aware of any other dynamic temperature models worth considering?

A159. I was recently contacted by Dr. Richard Polehn at the University of Washington about some temperature modeling software he had developed. RiverTemp can be used to estimate water temperatures and, to a lesser extent, dilute soluble concentrations along a branched, flowing river network. The program models each river reach in one-dimension and allows for both steady state and transient computations. The model does not allow time varying volume flow rates, but does allow variable flow by location. The temperature computation allows for convective heat exchange, evaporative energy exchange, and radiative long wave heat exchange with the atmosphere. In addition, solar heating of the water is calculated based on longitude, latitude, day of the year, and time of day. Note that these energy sources can be highly variable in time. The program allows time- and location-variable atmospheric conditions, inlet and lateral flow temperatures. More detail can be added to a computation by subdividing an existing reach or by subdividing the time intervals.

The program is designed to be highly visual. It emphasizes graphical construction of a river network and visual interpretation of the computational results. The results are fascinating to watch, with the figure associated not doing justice to the dynamic nature presented.

Dr. Polehn maintains a Web site for his RiverTemp software and provides an email contact. As usual, USGS cannot endorse this software, but we wanted you to be aware of it.

Q247. Topic #9: How Time Flies Pg. 67, paragraph 4 You mention that a model other than SNTEMP must be used to analyze hydropeaking. Do you have any suggestions for an alternative?

A247. It depends on your objective and the accuracy you require. If you need excellent accuracy, CE-QUAL-W2 is likely the best model. I'd talk to Tom Cole with the Corps Of Engineers about your needs. If all you really need to do is 'bracket' the possibilities, you can run SNTEMP for the high flow and low flow conditions and the 'answer' will be in-between.

Q248. I represent X. PUD, which owns the Lake X. Hydroelectric Project. X. PUD has applied for a new license from FERC and in the process of that proceeding is re-creating a river that's been dry for 76 years (the X. River), and that used SNTEMP as part of its process. The X Department of Ecology approved of the work done by X. PUD, but subsequently some have questioned the continued viability of SNTEMP as a model. They claim it is outdated and no longer a prevalent model in the United States, and that instead, models such as HeatSource are used much more often.

My question is whether you know of any data I can rely on to support our position that SNTEMP is still a widely used and appropriate model? If so, please let me know where I can find it. Thank you very much for your time.

A248. Thanks for your inquiry concerning SNTEMP. In fairness, I think there is some truth to what the some are saying, though perhaps for somewhat different reasons than they might have in mind. SNTEMP is "old" in that it is principally a mean daily model and extra work may need to be done to achieve acceptable maximum daily temperature estimates. Daily maximum temperatures are, for better or worse, getting to be the more commonly used regulatory standard. I say it this way because I am not yet convinced that daily maxima are the best indicator of biological consequences, but I seem to be in a minority. SNTEMP might also be considered "old" for a variety of other reasons, depending on the circumstances, including inability to handle fluctuating flows, inability to simulate reservoir thermodynamics, and implementation in the DOS computing environment. This last point, as you know, has been updated by the work Tom Payne and his staff has done, but remains true for many people. Personally, I believe that SNTEMP has many productive years remaining in the natural evolution of models. It is highly cost-effective to collect data for and run, unlike many other models, especially the ones which do handle fluctuating flows and reservoir dynamics, but also compared to ones that do a better job of predicting daily maxima.

As for applications of SNTEMP, I can assure you there are many. In fact, there are many more than I know about because people do not necessarily tell me about them all. I find references to several applications a year where I have never heard from those applying the model. I feel gratified that there are so many applications, though I wish I had more of a library of references. In any event, I have appended a list of references, not just applications, I know about below.

Today, I see more SNTEMP applications in the TMDL arena and other land use evaluations (shade restoration, storm water, grazing, and timber management) than FERC relicensing. In looking through my recent correspondence log, I see references to dealing with urban Best Management Practices, Forest Service timber management plans, state forest service fire recovery efforts, university use, and international applications in Laos and Canada. Probably the biggest group of applications seems to be in the state of Washington where, I suppose, their DEQ and Public Works departments consider it the standard. You might want to check with them for more on this.

All this said, I'd like to be clear that my preference is not for SNTEMP per se, but rather for the application of cost-effective modeling and solution techniques that apply the best data and information to solve (or at least address) the problems at hand. Too often we apply a model because we know it, not because it is the necessarily the best for the job.
Then I appended the list of references found at http://www.fort.usgs.gov/Products/Software/SNTEMP/SNTEMP_refs_DEV.asp

Q249. I'm going to be using SNTEMP as part of my master's research work. I'll be using it to look at in-stream temperatures under different management practices within the City of G., including riparian management, low flow channels, etc. I've been talking to a few people about what I'm going to be doing and that I will most likely be using SNTEMP. A common question I get is why SNTEMP has minimum time steps of a day even when most of the input can be obtained on shorter time steps (flows, air temperatures...). This question generally comes from people interested in the thermal impacts on fish. I usually answer that the model uses daily averages of all inputs and that I don't know how much benefit there is of having hourly temperatures over daily averages and maximums. I haven't looked at many papers regarding instream temperatures and fish health, but the ones I have seem to deal with maximum or daily average temperatures. Although my research is not looking directly at fish habitat, I think I need to be able to answer this question. I know that SNTEMP is meant to be used with PHABSIM and therefore must have some biological relevance. How do you usually answer this question? Is it that SNTEMP would have to be completely rewritten to provide hourly temperatures or is it that there is no real justification to have temperature readings shorter than a day?

Follow-up - I see myself getting talked into providing hourly readings for temperature. If this is the case, would it be possible to alter the SNTEMP code to make it work with smaller time steps? I personally don't feel confident doing this, but my advisor has a programmer that should be able to.

A249. This is a good question. I will be forwarding a recent exchange to you that address part of your question. Basically, I see a transition to finer time steps, though I personally do not necessarily believe that the extra resolution is warranted. But then fish biologists are often convinced that they have to have it, especially if they know that they can.

Follow-up - Unfortunately, the foundation for SNTEMP is steady state, not dynamic. Modifying the software would not be appropriate. It would be better to use a different model. HeatSource or CE-QUAL-W2 come to mind. Both, especially W2, are quite demanding.

Q250. I was looking over the FAQ on the SNTEMP website and noticed a section on other temperature models. There were a couple out there that seemed worthwhile investigating their merits. In particular, HeatSource and RiverTemp. I was able to track down HeatSource but a Google search on RiverTemp proved fruitless. Do you have any leads on how to get a copy of that model?

A250. I don't know why I didn't list the contact as Dr. Richard Polehn at Washington State University (rpolehn@beta.tricity.wsu.edu). RiverTmp is a very interesting model, and certainly visually fascinating. I am not at all clear whether Dr. Polehn really intends to make his model generally available or not. Please contact him for more information.

A251. Yes there is, but I am not familiar with the details. Occasionally I see announcements for a program called the CORMIX Mixing Zone model. Please see http://www.mixzon.com for more details or email information@mixzon.com

Q252. I have been using SNTEMP for some time and have found it to be a great model. Unfortunately I now need a model that will run at an hourly time step. Do you know of a public domain model that would have this capability and work as well as SNTEMP?

A252. The dynamic stream temperature models I hear about that are used most often are two Corps of Engineer's models, CE-QUAL-W2 (originally for reservoirs but now handles rivers), and HEC-5Q (or its new generation). There is also a great deal of use of RMA-11 and Heat-Source. The Corps models are supported to some degree I believe, but I have never been quite sure about the other two. I'm sure you can track down the information on each on the web. It is my belief that all these models will require significantly more investment in time and data than you have been used to with SNTEMP.

Q253. I was applying your water temperature model SSTEMP and other water temperature model that I created. The results of the simulations are very good. But your model SSTEMP and mine fit very well with the observed temperature when I introduce the same stream flow as input and as output. Why do you believe that this one happens?

A253. Without more detail, I really cannot carefully answer your question. When the inflow and outflow are equal, there are no accretions to the stream. If your reach in fact does have sizable accretions, and including accretions worsens the prediction, then my first supposition is that the estimate for groundwater temperature is in error. This may be especially true if you are using the assumption that ground temperature is the same as mean annual air temperature. Though this is a good rule of thumb (as an annual estimate), it may or may not be very good on a seasonal basis.

Q254. The SNTEMP and stream segment models seems to be able to do what I require and I have been unable to find any home grown models with a similar functionality from the United Kingdom. Is SNTEMP still widely used by government agencies in the USA? I would like to be able to justify my choice of selecting SNTEMP to use by making such a claim!

A254. Yes, SNTEMP and SSTEMP are still widely used. I would estimate that there is more use by non-government agencies than within the government, but most of the work is being done by consultants in support of government-related projects. In fairness, there is also growing use of several hydrodynamic models that at least claim to be more accurate than SNTEMP, especially in some situations. Though I do believe that these models can do a better job predicting maximum daily water temperature, it is very unclear to me whether they are better enough to justify the considerable added expense of using them.