SNTEMP (In)Frequently Asked Questions: Hydrologic Issues

Q15. Do you know of any SNTEMP studies done on rivers similar to the River D…, in Oregon? Here are the particulars that make it a little unusual. It is large, with a mean Q of about 5000 cfs. It is fast: though the gradient only about 0.002, the mean velocity is generally agreed to be about 1 m/sec. So the travel time through our 160-km reach below the dam is in the neighborhood of 2 days. It is mainly spring fed. The temperature at the top ranges from 7 to 12. At the bottom, the T range is 5 to 18C.

The annual Q range is small also. The range at the top is about 3500 to 7000 cfs; at the bottom, the range might be 5000 to 10000 cfs.

A15. No, the biggest river I can remember SNTEMP being applied to is the Colorado/Green. But it is warm and slow. The Trinity River in northern California is probably similar topographically, but flows are much smaller most of the time and the gradient is higher. [Added 12/2001]

Q16. Based on our preliminary results, it would appear that a Temp change at the top could persist for a long ways downstream, maybe the whole 160 km. Is this conceivable, given that the water moves fast and the difference between water and air temperatures is often small?

A16. Sure. In fact, my rule of thumb about meteorology dominating after about 30 km must be thrown out the window when dealing with such a large river. For example, a simple experiment with SSTEMP would show lots of what I call inertia – a bad name for the thermal mass of water. What I mean is that the heat in that mass cannot change very rapidly no matter what the ambient conditions. [Added 12/2001]

Q20. You often state that "this is not a hydrology model". I know that I must calculate the Q between each node pair. Are there any ready-made routines for calculating all the necessary Q’s? Or do I need to use a big spreadsheet or write my own program?

A20. You do not need to calculate lateral flow (accretions or losses). SNTEMP does that for you. But you do need the flow at each hydrology node, yes. Some program or spreadsheet is normally the way to go. Usually, one doesn’t know "internal" flows, so determining an average per km is about as good as you can do. [Added 12/2001]

Q71. I assume time-of-travel can be employed instead of Manning’s n? If so, what units are used - - presumably some rate (i.e., distance per time)?

A71. I believe this is documented in the ERRATA*.Txt file that should have come with SNTEMP. The units are seconds/kilometer such that any number in that field >1.0 will be treated as travel time. [Added 12/2001]

Q87. I am a recent graduate of …University's Program in Environmental Science, having received my M.S. in December 1998. My thesis research involved the relationships between stream flow and stream temperature on the lower River Y… in Washington State. The title of my thesis is "The Effects of Flow Variation on Stream Temperatures in the River Y…." I am interested in conducting some additional research on the effects of turbidity on stream temperatures. Because of extensive agricultural development in this region, the lower river has a significant problem with suspended sediment, which results in high turbidity during the summer months. Local agency personnel and other interested parties believe that the high turbidity and dark color of the water contribute to elevated stream temperatures. While this seems at least plausible, to my knowledge there has been no research conducted to substantiate this statement in the river. Are you aware of any studies or stream temperature models that address this phenomenon? To date, I have found very little information in the scientific literature, and I am wondering how thoroughly this relationship has been studied. I would appreciate any information you may have on this topic, or any suggestions for other contacts I might make.

A87. [From Fred Theurer] It is easy to believe that the solar (shortwave) radiation would be more readily absorbed in an opaque, rather then transparent, water column, especially if it is darkly colored. But I'm not certain that it is as important as it may initially appear. My recollection of the literature search at the time is that shortwave radiation is entirely absorbed within a transparent water column within a few meters, such as a lake. (I'm not certain, but I believe the number I read was 3 meters.) Also, the streambed will eventually intercept the shortwave radiation for transparent water and what is not absorb is reflected and has to pass back through the same water column. Remember that the solar radiation is almost always at an angle so that its path is lengthened. And finally, what is absorbed by the streambed is converted to longwave radiation (heat) which is subject to convective & conductive heat transfer at the streambed-water column interface back into the water. Obviously, there is some heat conduction down into the soil & water column below the streambed. Matthew Boyd may be able to offer some help with the streambed conduction.

[From Matt Boyd] This is a terrific discussion. Other resources would include Wunderlich (TVA) 1972, Beschta and Weatherred 1984, and Sellers 1965.

In terms of surface reflectance, we can easily calculate a coefficient of reflectance as a function of solar altitude. When the solar altitude is high (and short-wave radiant energy is greatest) reflectance is low. The effects of turbidity would probably not affect the mid-day reflectance experience in a stream. I know of no research that mathematically relates flowing water reflectance as a function of turbidity.

Water Column Absorption/Assimilation as a Function of Turbidity

We should assume that 50% of short-wave radiant energy is absorbed in the top 10 cm of the water column (Sellers 1965) and that nearly 100% of the short-wave radiant spectrum (0.36u to 0.76u) is absorbed in the upper 1 meter of the water column (Wunderlich, 1972). Further, radiant energy that traverses the water column will encounter the streambed, which has quartz like absorption properties and is very likely to attenuate remaining portions of the overall solar load. Beschta and Weatherred (1984) correctly assume that differential heating takes place in such a condition, but that eventually the heat energy absorbed in the streambed will be returned to the water column via conduction as a function of grain size. Any reflectance of short-wave radiant sources off the streambed will then traverse the water column, with same likelihood of attenuating in the water column that Sellers (1965) describes. Simple logic then would suggest that a stream with an average depth of 5 cm or greater would experience near 100% absorption of solar energy either in the water column or streambed... with all of the solar load that penetrates the water column either directly received by the water column or conducted to the water column from the streambed within 24 hours. With this said, we can put some perspective on the role that turbidity plays in the attenuation of radiant energy in the water column.

As John states below, in the diel cycle the water column will attenuate or receive via streambed conduction all of the solar load that penetrates the stream surface regardless of turbidity. Perhaps the only effect that turbidity may play is the increasing likelihood that radiant energy is attenuated before reaching the streambed.

Water is opaque to thermal (longwave) radiant energy. Regardless of turbidity a stream will attenuate 100% of the incoming thermal energy from the atmosphere and vegetative sources. However, the stream itself is a constant source thermal radiation (backradiation... termed by McCutcheon). It is possible that an increase in turbidity would affect the Emissivity of a stream and therefore alter the backradiation flux emitted from a stream. I would anticipate that such a function would be difficult to accurately quantify and prove negligible when compared to the overall heat energy balance that occurs in a natural stream environment.

[and from Bartholow] I wish I could help as I too have heard similar "hypotheses". When I was putting together IP#13, I looked for information on this and the related issues that (1) a dark stream may reflect more radiation than a clear stream, and (2) stream with dark cobble/boulders may experience greater daily maximum temperatures, but I found little credible information. The reservoir water temperature model I am using at the moment (CE-QUAL-W2) includes a formulation that adjusts the light penetration as a function of water clarity (algal biomass and particulate matter). This does make some small difference in the reservoir's thermodynamic stratification. They cite Bears Law (with no reference) as

This formulation is discussed along with other surface heat exchange material citing Edinger et al. (1974). Heat exchange and transport in there environment Research Project 49, rpt, 14 EPRI 74-049-00-3 Palo Alto CA

Experimental data has shown insensitivity of surface reflectance to the water's purity and turbidity (Viskanta and Toor 1972). Viskanta, R., and J.S. Toor. 1972. Radiant energy transfer in waters. Water Resour. Res. 8(3):595-608.

I must admit to a degree of skepticism about dark water in a free flowing environment. From my point of view, all (or almost all) of the radiation penetrating the water's surface is likely to be absorbed either by the water or by the substrate, and thus becomes a part of the stream's energy budget regardless of the fraction immediately absorbed by each. However, I must admit that I am used to thinking about this from a "mean daily" perspective, and not from a more dynamic point of view. I suppose the maximums could be somewhat perturbed. [Added 12/2001]

Q92. Can SSTEMP account for the effects of a pond and influencing downstream areas or does it just model effluent flows added to the stream as a pond that is not part of the stream? That is the situation we have here. These ponds will be instream and protecting the downstream areas from hollow fills [coal mining].

A92. The model really knows nothing about ponds or their effect on hydrology or release temperatures. You give the model those values, and it transports heat and flow downstream. If the ponds affect hydrology or release temps, that is something to be reckoned with outside the model.

Q122. What happens if inflows are more than outflows? This would be a reach that is either losing to groundwater or where there is a diversion in the reach. I guess if there are diversions, it could be better to break the reach into pieces. Can the model handle a losing reach, or a reach where the location and magnitude of individual diversions is not readily available?

A122. In answer to your questions, if inflows are greater than outflows, the stream is simply a losing stream. Water, and heat, leave the stream uniformly throughout its length. I have occasionally noticed some computational problems in losing reaches, but luckily they are not subtle -- they stick out like a sore thumb. And they seem to occur when the stream is majorly losing, not just a small amount. So, in general, the model handles a losing reach just fine. [Added 6/2002]

Q123. Now that I have the model calibrated. I need to run some scenarios at different flows and at different years. I do not have temp and flow data for that year at the headwater for the model. I am trying to set up a zero discharge to predict the temps for that year's flow data.

For the zero discharge input files, is the flow data all zero for all days? Is the segment characteristics for the zero discharge the same as the headwaters or is it an average of the headwater and zero?

A123. If you mean that you wish to use a zero flow headwater, that might be a bit unusual. Zero flow headwaters are useful if you need to simulate an unmonitored tributary that has some, but not a lot, of influence on the stream. It could be that switching to this mode in your case could be argued that you are using the model out of its calibrated domain. I'm not really sure what alternative you have though except to make assumptions, i.e., if the headwater flow was X then this is the response you get from the system. [Added 6/2002]

Q124. I've been working on an application of SNTEMP in the M. River Basin and have gotten the model to run. The model runs with no warnings or errors, however the initial temperatures at the zero flow headwater nodes are 0.00 in all instances. I initially left the temperature field blank for these nodes in the hydrology data file so that the model would default to the annual average air temperature. When that didn't work I filled the initial temperature field with the annual average air temps, but it is still showing 0.00 in the output files (KVRTRNS, KVMETR, KVRHYDR).

Is this a problem that you've seen before? Any insight that you can provide will be greatly appreciated.

A124. It has been a long time since I have looked at a zero flow headwater and do not clearly remember what I have seen it show for temperatures, but the fact is that it doesn't matter. If there is zero flow, we don't care whether the temperature is zero or 100 because there is no heat if there is no volume. What happens then is that the stream gains lateral accretions (typically at groundwater temperature) as it goes downstream. Given enough distance, the headwater stream will be near its "equilibrium" by the time it gets to the next downstream node. Just make sure that SNTEMP is calculating lateral flow temperatures correctly by looking at Table 6 (I think) and also look to see that the water temperatures predicted by the model look pretty good as they accrue downstream. The model will never be as good as if you had measurements, but as you know, we don't always have them. [Added 6/2002]

Q213. Is the model capable of mixing groundwater input (based on groundwater distributed flow into the stream) with surface water flow. I think this would have influence on stream temperature. Please let me know if the model can do this.
A213. Yes, the models incorporate mixing of lateral accretions on gaining streams between points of known discharge. However, effects due to hyporheic flows are not directly modeled.

Q214. Lateral flow is computed based upon the difference between flows at hydrology nodes. I have a losing stream (lots of diversions and water rights claims) so I would think that lateral flow would be computed as zero or negative. But I have a positive inflow. Yet on another stream, I have zero or negative inflow for a decreasing flow but on this stream I haven't added in any D nodes.
Does the program see the D nodes as taking water out and therefore compensate by allowing lateral flow in? If I had more water flowing out through a D node than could be accounted for by the flow downstream (e.g. flow starts off at 10cfs, I take out 8 cfs, then have a downstream node with 4 cfs - 2 more cfs than should be possible), does the program account for this by adding lateral flow?

A214. SNTEMP calculates the lateral flow between true hydrology nodes where instream flow is given, not at diversions where you specify the diversion flow. Then it computes the lateral accretions per km between the nodes. If there are intervening diversion nodes, it certainly has all the information it needs to compute the flow at and prior to the actual diversion, so it probably does this, though I am not exactly sure. What would happen if you specified more diversion that would have been in the channel at that point, I don't know, but the software would likely either quit with an error message or would give obviously wrong answers below the diversion. I do not believe the program would add any additional accretion to make up the difference -- "SNTEMP is not a hydrology model."

Q215. Thanks for getting back to so quick about my last question; your answer was very helpful. However, I have another question today this time about Q nodes and the hydrology data file. I have a number of validation nodes in my model where I collected temperature data over the summer but don't have measured flows at all the same points, although I can generate flows at those points using a hydrologic model I already have of the area. In the hydrology data file, do the validation nodes require both temperature and discharge data? If so, then does the discharge data at the validation nodes act as discharge data at Q nodes? I understand that Q nodes are used to change the discharge in the river and are useful when groundwater discharge rates change. I just wanted to know if I need to have Q nodes that correspond to my V nodes or if V nodes only require temperature data and so reserve Q nodes for when I have changes in groundwater or lateral flow.

Thanks again for all your help - you have definitely earned a spot in the acknowledgments section of my thesis!

A215. Interesting that you should bring this up. I am just now in the process of struggling to remember all this myself on the X in NY. The basic answer is that both V and Q nodes are hydrology nodes (along with K, D, P, R and all skeleton nodes). You must supply discharge at all hydrology nodes but Q nodes do not require temperatures. There is generally no reason to co-locate a Q and a V node. In other words, if you have a V node, you don't need the Q node. The V will act as the Q, even if there is no change in the basic accretion rate.
It is often the case that you do not have flows at V nodes and must estimate them one way or the other. If you have a hydrology model, that's great.

Q216. I also have a question about the data provided in the KVRMETR file, in particular the lateral inflow data. I believe this is calculated as the difference in flow between two hydrology nodes and is assumed to be groundwater. But when I check the difference between two nodes it does not agree with what is stated in the KVRMETR file - it remains at 0 even though there is a difference. However, I did adjust flow at some points to make sure there was groundwater inflow in some areas and this was effective at calibrating the model (but the lateral flow in KVRMETR stayed at zero). Any idea why there might be a discrepancy here? I can easily enough find out what the groundwater contributions are on my own, but I just want to make sure I understand what is going on "inside" the model.

A216. Lateral inflows are indeed computed as simple mass balance differences between points of "known" flow, i.e., all hydrology nodes. This may or may not be assumed to be "groundwater" in the sense that if you leave out the accretion temperatures from the hydrology file, SNTEMP will add the accretions at "ground" temperature (where ground temperature is computed as the mean annual air temperature adjusted for that reach's elevation). If, however, you do supply the accretion temperature, then that is the temperature that is used.

Now, if I understand you, you say that SNTEMP prints out zero accretion when there should have been some. I have never seen this before and find that hard to believe. If you can send me an example, I'd like to see it. Are you looking at Table 6? Zero lateral flow should only occur when you have two adjacent hydrology nodes with the exact same discharge for a given time period.
Later - Perhaps if I had paid closer attention to your earlier message I would have realized what was going on. The lateral flow printed in the KVRMETR file (Table 6) is the flow that accretes below the node listed on each line, not above, and ditto for all other reach-specific variables. That could have made for a confusing time during calibration, no?

Q217. I've picked up my SNTEMP model again after putting it on the shelf for several months. We are about to have a planning meeting with the stakeholders in the watershed. I'm sure they will be interested in what the capabilities of the SNTEMP model are. After reading through the documentation in more detail, is it correct that in the hydrology data file you must supply discharges for each time period at every node defined in the hydrology node file? In other words, if my hydrology node file looks like this:

must I supply daily mean Q's for each of those nodes or only the nodes I have data for? The bigger issue is how can anyone run this model for basins with only one or two stream gages? I have attached my hydrology node file for the watershed being modeled. I do have some historical data at a couple locations but certainly not the coverage I assume is required for the model to converge.

Follow-up - Thanks for the quick response. My hydrology node file looks OK because that is how it theoretically should look to provide an accurate conceptual model. However, I only have historical discharge at the Q nodes and a couple of H nodes. Otherwise I don't have any data to attach to the many T, J, and B nodes in the hydrology data file.

On a different note, if I read the manual correctly, I only need observed temperatures at H nodes with a nonzero discharge. Therefore, assuming I could estimate discharge at each node in my hydrology node file I only need to have water temperatures for the H nodes that have some flow, right?

A217. In general the answer to your question is yes. The mantra is, "This is not a hydrology model." Boundary conditions at the periphery of the system must be known or assumed. Flows at internal nodes are usually calculated (by the modeler) as simple mass balance, linearly interpolated by distance, or (if you have the data) by regression or drainage area, or some other fancier technique.

There is a small exception, which I hate to mention but for the sake of completeness. The model does permit a small number of missing flows. In this case, it fills in the missing values with the average of all "measured" flows for that location at that time period across all years of data. This is not a very good solution and I don’t generally recommend using it.

Follow-up - Yes. Temperatures are only required at H nodes (with the exception you stated), or other peripheral nodes (e.g., P, R, S), except that they must be supplied (when possible) for V nodes.

Q218. It looks like flow must be provided at all H, B, S, T, Q, K, V, and J nodes in the network. I assume this means that the model does not calculate or assign flow to downstream nodes based on assigned flow at a starting point (H nodes) and subsequent diversions and additions of flow. For example, if a segment of the network has one each of H, D, and T nodes, then SNTEMP will not calculate a flow at the T node based on flow supplied at the H node and flow diverted at the D node. Is this correct? If this is correct, then, for this example, does the assigned flow at the T node have to equal flow at the H node minus flow at the D node? In other words does Flow(H) – Flow(D) = Flow(T) ? It seems this would have to be the case if no other flow was added to or diverted from the stream in this example segment. Alternatively, can the assigned flow value at the T node be greater (or less) than Flow(H) – Flow(D), which would “build in” an assumption of some non-point gain or loss of flow throughout the segment, or is this what Q nodes are for? Thanks for you assistance with this, I swear I’m not a dummy….

A218. The mantra is, "This is not a hydrology model." Flow must be specified for all hydrology nodes except in very special cases. In effect, you are responsible for calculating the mass balance, including all estimates of gains and losses -- and groundwater (lateral flow) accretions or percolations. In turn, SNTEMP takes what you specify and (re)computes any groundwater gains or losses as a rate per kilometer. You can see these computed lateral flows in one of the output tables.

Yes, I know this is a pain, but at least it makes you specify what your assumptions are.

Q219. Just so that I am clear on the hydrology/flow aspect of SNTEMP... it sounds like if I designate different flows between a upstream node and the next downstream node (i.e. a H node and a T node), SNTEMP will calculate the gains (or losses) as a rate per kilometer. For example, if an H node has a flow of 10 cms, and 2 kilometers downstream a T node has flow of 8 cms, then SNTEMP will account for the difference by "adding" 2 cms to the segment between these two nodes at a rate of 1 cms per kilometer. I just want to be sure about this before running the model.

A219. Yes, that is exactly how things work. As you might imagine, there are some special cases. The easiest to understand is if there are intervening D, R. or P nodes. In these cases, the model is smart enough to calculate the average lateral flow (positive or negative) but does not know how to allocate that to individual sub-reaches (e.g., H-to-D and D-to-T). For this reason, the computed lateral flow is assumed to be constant throughout the entire hydrology reach (H to T).