Beginning with the Mossy Creek and Long Glade Run TMDLs, completed back in...eesh, 2004!...my colleagues and I at Virginia Tech started considering the effects of low flows on physical and behavioral changes in the stream. We noticed at that time that very low flows (Long Glade Run in particular was observed to go dry on a regular basis) cause HSPF to simulate hyper-concentrated bacteria - I typically explain this to my students as HSPF simulating flow happening down to the last little molecule of water hopping down the stream, and trying to cram millions of bacteria into that molecule. Of course in this case the concerns are with direct discharges to the stream from animals standing in the stream and illegal discharges from residences, as overland flow contributions won't be an issue at low flows (they only occur during the high flows associated with storm events) and permitted point source discharges come in with a significant volume of water that tends to prevent low flows.
The reason this happens is that HSPF simulates the stream using a function table (FTABLE) to represent the hydraulic properties of a reach. This FTABLE includes columns for depth, surface area, volume, and discharge, and HSPF enters the table for a given volume to interpolate the other three properties. As a result, the entries in the FTABLE create a series of smooth-sided stacked trapezoids. The lowest entry in the FTABLE is required to be zero volume, so HSPF will continue to interpolate flows all the way down to zero volume in the stream.
Clearly this is problematic. The first, physical, issue is that the stream bottom is not smooth - it is rough - and there is a period of time when there is still water (volume) in the reach but there is no flow, when the water is stored in a series of disconnected pools. This means from a modeling point of view that there is water, and direct discharges into that water can occur, but they will not flow downstream. Imagine a cow standing in the puddles, defecating - the cow pie will certainly contribute a large volume of bacteria to the puddle it hits, but because the puddle doesn't connect to any other puddles, the bacteria do not have the opportunity to move downstream. Additionally, the bacteria will have time to die off before the stream flow returns to a normal level. To address this physical situation, we began adding what we called a 'flow stagnation volume' to the reach - an entry in the FTABLE, immediately after the required zero flow entry, that has a small volume but no discharge. This allows the model to appropriately simulate cessation of flow when the volume falls to a level when all water is actually in a series of disconnected pools. The water and bacteria are not lost, simply held until flow increases. This will appropriately affect ALL sources of bacteria.
The second issue is a little harder to describe mathematically. This issue is a behavioral one and is twofold. First, if the water in the stream is running in a narrow rivulet, where before cows might have stood in the stream for relief from heat, insects, etc. and had their hind ends over the water, now the stream will provide little relief from heat and pests and will be a much smaller target for the defecated material to hit. This means that at some point the cows may still drink from the stream, but it is much less likely that their manure will actually be deposited in the stream. The same could be said of wildlife, though perhaps the restrictive depth would be lower because the animals are smaller. The second issue has to do with the water availability. If there is indeed an intermittent stream in a farmer's field, a logical assumption is that he must provide an alternative water source to the livestock when the stream goes dry or nearly dry. There is a lot of anecdotal evidence and some research literature to suggest that cows with alternative water sources will spend as much as 90% less time in and around the stream. Thus, if the water is low and we assume at those times the farmer must provide another water source, we can assume that the cows will be physically removed from the stream. In a similar manner, if the stream is getting low, it stands to reason that highly mobile wildlife (e.g., waterfowl, which also happen to be the worst offenders in terms of defecating in the stream) will fly away to wetter areas, again physically removing themselves from the stream.
To address this second complex behavioral issue, we institute a "stage cutoff" on animal contributions to the stream. This means in a practical sense that we export the depth of the water from the stream as an hourly timeseries, then create a multiplier from that timeseries where depths above the critical level are given a multiplier of 1 and depths below the critical level are given a value of 0. The new multiplier timeseries is then multiplied by the input direct deposit timeseries to create a filtered input timeseries for HSPF. Thus, we can represent the animals being physically removed from the stream when the flow drops.
The flow stagnation volume was first implemented in the Beaver Creek TMDL in 2005 and has been used in all subsequent TMDLs developed at or in conjunction with Biological Systems Engineering at Virginia Tech.
The stage cutoff method was first used in the Mossy Creek & Long Glade Run TMDL, and subsequently used in the Beaver Creek, Lick Creek, and Old Womans Creek TMDLs completed under my direction. It's also been used in some TMDLs developed by my peers at Virginia Tech. Unlike the flow stagnation, which is a physical representation I feel confident is applicable everywhere, I evaluate the need for the behavioral representation (i.e., the cutoff) on a case-by-case basis during water quality calibration.
I've now set the backdrop for the current conundrum I'm facing with low flows in my current project...but I think this post is long enough for now, so I'll tell you all about it next week!
No comments:
Post a Comment