I'm doing some work on hydrograph separation using hydrochemical measures. Throughout the literature workers refer to the need to use so called 'conservative tracers' for hydrograph separation. I've always thought of conservative tracers as those chemical species (ions, molecules, isotopes) whose movement is not retarded by interaction with porous media and/or attenuation by biophysical processes (e.g. reduction). Using this criteria Cl, Br and the stable isotopes of water are often referred to as conservative (dissolved Si is also commonly used).
However, when looking to separate a hydrograph into component sources (aquifer, soil and surficial runoff) I see some workers note that the tracers most suited to hydrograph separation are those that are 'conservative with flow.' If a tracer is highly correlated with flow, -/+, then it is better suited to hydrograph separation than one that is poorly correlated with flow. Further, the same workers note that the suitability of a tracer for hydrograph separation varies with the physiographic setting of any given capture zone/catchment, noting it is unwise to assume that tracers suited to hydrograph separation at one site are suited at another.
Conservation of a tracer with flow suggests different concentrations of a given tracer within each key compartment supplying stream flow (e.g., aquifer, soil, surficial zone). In some of the sites I am investigating, Cl and Br are not conservative with flow (i.e., show no correlation) and exhibit little if any meaningful variation in concentration between surficial, soil and aquifer waters that I have sampled. In this setting Cl and Br have proven to be poor tracers of water source. Rather, Total Nitrogen exhibits the strongest correlation with flow (r >0.95), increasing with flow. The increase in TN is correlated with low TN in reducing aquifers, moderate TN in subsoils and highest TN concentrations in 'A' horizon and surficial waters of an intensively farmed catchment. In addition to TN there are other tracers, which are not classically defined as chemically conservative or 'ideal', which are also strongly correlated with flow. These flow conservative tracers also exhibit show strong variation in concentration between the compartments supplying stream flow.
Consequently, I'm tending towards the definition that the most suitable tracers for hydrograph separation are those that are strongly correlated with flow (r >0.9).
I was wondering if anyone else has addressed this issue as it appears there is conflicting definitions of what constitutes a conservative tracer when looking at stream hydrograph separation?
Kind regards,
Clint
You might try googling on end member analysis (Garrett is one of the co-authors I remember). In my opinion, you would want to identify components that are fairly unique to rainfall, interflow, surface runoff, groundwater, etc. For example, fine sediments are most likely in runoff, typically not rainfall, not groundwater. The problem in using just sediment, is the varied erosion sources as flow velocity, depth and available surface, bank and channel sources change, so it is not necessarily unique to surface runoff. If you separate mineral and organic sediment particles, or perhaps in some headwater systems, there may be some options for identifying uniqueness and lead to separation, but these headwater systems may have infrequent groundwater influence. I do not feel I should suggest what you use as conservative tracer, it may change with system.
Hi William, thanks for your reply. Helpful to hear your landscape based thinking. I've been looking at both EMMA (Hooper and others) and isotopic and geochemical mixing equation approaches (Pinder and Jones, 1969). Personally, I have found the work of Hendershot and O'Brien (1993) the most helpful in defining the suitability of tracers. As you also conclude, a number of other works have noted that the suitability of a tracer for hydrograph separation is often unique to the physiographic setting of a particular stream capture zone. I feel this is especially important for regions such as New Zealand, where spatial variability in geology, soil and climatic gradients are steep.
As you also suggest, O'Brien and Hendershot (1993) used linear correlation analysis to identify tracers that were defined as being reactive with flow. They suggested these tracers are good indicators of the hydrological flowpath. (For example, Si/K ratio appears to be a useful indicator of overland flow. It is weakly correlated across the entire flow range and yet strongly correlated with the flood flow component of the hydrograph). I'm defining these reactive tracers as those that are statistically significant with flow but have an r
I agree with your definition Clinton of conservative solute tracers as those that move through the system during the time frame of interest with little attenuation or transformation through biogeochemical processes. Exploring linearity of lack thereof among various solutes is an informative approach in this regard. I would also urge you to read the paper by Hooper, 2003, WATER RESOURCES RESEARCH, VOL. 39, NO. 3, 1055, doi:10.1029/2002WR001528 where he discusses various diagnostic approaches to explore when applying mixing models. Another issue to consider is whether there is a significant difference in concentrations of the solutes you are applying in the mixing model among the components that make-up streamflow. In your case, it may be baseflow/slow flow and rapid runoff/interflow, though some systems have a readily identifiable third source of runoff such as overland flow. Hooper mentions that there can be much more complex models as well with 4, 5, or more mixing components if justified. A final topic of discussion you raise is preference for selecting solutes that are correlated with flow. This is also an important consideration, and I would offer that obtaining satisfying model results is facilitated when the solute(s) of interest show predictable correlative behavior as streamflow rises and falls. However, it is not necessary to have this condition in order to have a successful mixing model solution, as long as there are significant differences in concentrations among the solutes for the various components of the hydrograph separation. Finally, examining the behavior of solutes that are not conservative relative to the behavior of those that are can be quite informative and may suggest the importance of biogeochemical processes within the context of your system.
Thanks Doug, for your answer, lots of key insights. I've accessed the Hooper 2003 paper and am looking forward to reading it. It looks like it addresses the linear mixing phenomenon which is great. Thanks again. Clint