Math & Engineering

A stream tracer technique employing ionic tracers and specific conductance data applied to the Maimai catchment, New Zealand

Description
A stream tracer technique employing ionic tracers and specific conductance data applied to the Maimai catchment, New Zealand
Published
of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  HYDROLOGICAL PROCESSES  Hydrol. Process.  19 , 2491–2506 (2005)Published online 14 March 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hyp.5685 A stream tracer technique employing ionic tracers andspecific conductance data applied to the Maimaicatchment, New Zealand Michael N. Gooseff  1 * and Brian L. McGlynn 2 1  Department of Geology and Geologic Engineering, Colorado School of Mines, Golden, CO 80401, USA 2  Department of Land Resources & Environmental Sciences, Montana State University, Bozeman, MT 59717-3120, USA Abstract: The stream tracer technique and transient storage models (TSMs) have become common tools in stream solute andhyporheic exchange studies. The expense and logistics associated with water sample collection and analysis oftenresults in limited temporal resolution of stream tracer breakthrough curves (BTCs). Samples are often collectedwithout  a priori  or real-time knowledge of BTC information, which can result in poor sample coverage of the criticalshoulder (initial rise) and tail (post-steady state fall) of the BTC. We illustrate the use of specific conductance (SC)measurements as a surrogate for conservative dissolved tracer (Br  ) samples. The advantages of collecting SC data foruse in the TSM are (1) cost, (2) ease of data collection, and (3) well-defined breakthrough curves, which strengthenTSM parameter optimization. This method is based on developing an ion concentration (IC)–SC relationship fromlimited discrete tracer solute samples. SC data can be collected on a more frequent basis at no additional analysiscost. TSM simulations can then be run for the conservative tracer data derived from SC breakthrough curves andthe IC–SC relationship. This technique was tested in a 120 m reach of stream (2–60 m subreaches) in the MaimaiM15 catchment, New Zealand during baseflow recession. Dissolved LiBr was injected for 12 Ð 92 h, with Br  as theconservative ion of interest. Four TSM simulations using the OTIS model are optimized using UCODE to fit (1) Br  data derived from the Br  –SC relationship ( n D 1307 observations at each of two stream sampling sites), (2) allstream Br  data collected ( n D 58 in upper reach,  n D 60 in lower reach), (3) half of the stream Br  data collected,and (4) 20 stream Br  samples from each site. No two simulations resulted in the same optimal parameter values.Results suggest that the greater the frequency of observations, the greater the confidence in estimated parameter values.Br  –SC simulations resulted in the best overall model fits to the data, with the lowest calculated error variance of 6 Ð 37,narrowest 95% parameter estimate confidence intervals, and the highest correlation coefficient of 0 Ð 99942, among thefour simulations. This is largely due to the improved representation of the shoulder and tail of the BTC with thismethod. The IC–SC correlation method is robust in situations in which (1) changes in background SC data can beaccounted for, and (2) the data used to define the IC–SC relationship are representative of the range of data collected.This method provides more efficient sample analysis, improved data resolution, and improved model results comparedto the alternative stream tracer data gathering methods presented. Additionally, we describe a new parameterization of the cross-sectional area of the stream during flow recession, as a function of discharge, based on a stream hydraulicgeometry relationship. This variant of the OTIS model provides a more realistic representation of stream dynamicsduring unsteady discharge. Copyright  ©  2005 John Wiley & Sons, Ltd. KEY WORDS  transient storage; hyporheic zone; solute transport modelling; bromide; tracer test; specific conductance;OTIS INTRODUCTIONStream tracer experiments and subsequent transient storage modelling (TSM) are widely used tools to studyhyporheic exchange in a range of stream ecosystems (Bencala and Walters, 1983; Bencala  et al ., 1983; *Correspondence to: Michael N. Gooseff, Department of Geology and Geologic Engineering, Colorado School of Mines, 1516 IllinoisStreet, Golden, CO 80401, USA. E-mail: michael.gooseff@usu.edu  Received 2 September 2003 Copyright  ©  2005 John Wiley & Sons, Ltd.  Accepted 8 March 2004  2492  M. N. GOOSEFF AND B. L. MCGLYNN D’Angelo  et al ., 1993; Harvey and Bencala, 1993; Valett  et al ., 1996; Tate  et al ., 1995; Morrice  et al ., 1997;Mulholland  et al ., 1997; Runkel  et al ., 1998; Chapra and Runkel, 1999; Hinkle  et al ., 2001; Laenen and Ben-cala, 2001). The TSM provides a means of identifying stream solute transport processes including advection,dispersion, lateral in- or outflow, and transient storage due to in-channel dead zones and hyporheic exchange.Harvey  et al . (1996) and Wagner and Harvey (1997) discussed the reliability of the stream tracer experiment and subsequent use of the TSM, pointing out that the most information available within a solute breakthroughcurve (BTC) for parameterization of transient storage parameters is in the shoulder (initial rise in concentrationabove background) and tail (concentration decrease from plateau). Wagner and Harvey (1997) suggest thatone method for decreasing the uncertainty in the optimization of transient storage parameters is to increasethe sampling frequency during the time periods of the breakthrough curve at which the parameters are mostsensitive to the data collected.Several investigators have studied the suitability of various tracers for use in stream tracer experiments.Bencala  et al . (1983) found that rhodamine WT reacts with cationic co-tracers and/or sediment surfaces of streambeds. Ionic tracers such as dissolved Br  or Cl  might be more appropriate because they are morelikely to be conservative. Zellweger (1994) found that four dissolved salts, Na C , Br  , Li C and Cl  , werereliable dissolved tracers in an acidic mine drainage stream in Colorado. Beyond the general guidelines of the Stream Solute Workshop (1990), presentation of the stream tracer technique by Webster and Ehrman(1996), and the analysis of the TSM by Harvey  et al . (1996) and Wagner and Harvey (1997), there has been little specific information published on the use of dissolved tracers and execution or design of stream tracerexperiment techniques to refine our understanding of transient storage processes (e.g. hyporheic exchange),which increases hydrologic retention within the stream environment.In this study, we demonstrate an efficient method for implementing the stream tracer technique that reducesanalysis costs while increasing data resolution of the stream tracer BTC. This method utilizes a well-definedion concentration (IC)–specific conductance (SC) relationship. SC data are gathered frequently (secondsto minutes) and ionic tracer samples are collected less frequently (hours), throughout the entire range of BTC concentrations, providing thousands of data points to define the BTC. We simulate the results of aconservative stream tracer test in two subreaches of a 120 m stream reach during stream baseflow recession todemonstrate this method. We also use subsets of the larger data sets to analyse the effect of sampling frequencyon breakthrough curve analysis. The results are then analysed using model fit statistics to demonstrate theadvantages of this sampling approach, compared to simulating only the collected ion concentrations. Further,several solute transport metrics are computed to assess differences in model interpretation.SITE DESCRIPTIONWe performed a stream tracer experiment in a small first-order stream at the Maimai catchment (McGlynnand McDonnell, 2003a,b), an intensively studied long-term hydrological research site in New Zealand (seeMcGlynn  et al ., 2002 for details). The Maimai study area consists of multiple research catchments that formthe headwaters of the Grey River, located to the east of the Paparoa mountain range on the west coast of theSouth Island of New Zealand. Much of the hydrological research to date has been directed towards adjacent,similar catchments ( < 10 ha), sharing similar topographic, geologic and soil characteristics (Mosley, 1979;Pearce  et al ., 1986; Rowe  et al ., 1994; McGlynn  et al ., 2002). Slopes are short ( < 300 m) and steep (average34 ° ), with local relief of 100–150 m. The research described in this paper was conducted within the first-orderM15 catchment (2 Ð 6 ha). The M15 catchment is highly dissected with strongly convergent topography andnarrow riparian zones (combined channel and riparian widths range from  ¾ 3–7 m). The M15 catchment hasa mean slope of 38%, a stream channel slope of 16 Ð 7%, stream sinuosity of 1 Ð 06, and bank-full channelwidths ranging from 0 Ð 7 to 1 Ð 9 m. Wetted channel widths during the experiment were on the order of tensof centimetres, and stream depths were generally less than 10 cm. The catchment is underlain by moderatelyweathered, firmly compacted, early Pleistocene, Old Man Gravels that are effectively impermeable. Nearly Copyright  ©  2005 John Wiley & Sons, Ltd.  Hydrol. Process.  19 , 2491–2506 (2005)  A STREAM TRACER TECHNIQUE EMPLOYING IONIC TRACERS  249330 years of hydrological research at the Maimai catchments has shown the catchments to be highly responsiveto rainfall by shallow throughflow systems in  ¾ 1 m deep hillslope and riparian zone soils, with no evidenceof deep groundwater contributions (McGlynn  et al ., 2002).METHODS  IC–SC relationship method  The typical objective of stream tracer experiments is to generate a solute BTC of a conservative dissolvedtracer. Dissolved ions that have low or no background in most natural waters, such as Li C , Cl  , Br  andNa C , are commonly used as tracers. Ideally, the perturbation of stream solute loads is clearly traceabledownstream. The timing, magnitude and shape of the BTC provide data for transient storage simulations,in which transient storage parameters such as exchange rate and size of storage zone are calculated. Streamtracer solute concentrations should reach plateau, or steady state, at downstream sampling locations in orderto accurately characterize solute transport characteristics (Stream Solute Workshop, 1990; Scott  et al ., 2003a).One effective method of defining downstream BTCs is to gather stream grab samples and analyse them forconservative tracer concentrations (e.g. Bencala and Walters, 1983; Tate  et al ., 1995; Runkel  et al ., 1998).Chemical analysis of these samples is time-consuming and costly, thus many stream BTCs are more sparselydefined than scientists may desire.The IC–SC method is defined by collecting specific conductance and ion concentration data at streamsampling locations during an ionic stream tracer experiment. Data logging specific conductance probes collecthigh-frequency data from the centre of the stream, and stream grab samples from the same location are takenless frequently. Grab samples are later analysed for the ionic tracer of interest and a well-defined IC–SCrelationship is developed. The entire SC data set collected at each location can then be converted into ICdata (background IC and SC values are subtracted from BTC data so that only the SC change caused by thestream tracer injection is measured) and simulated using a stream solute transport model.The greatest advantage of the IC–SC method is the ability to maximize data resolution and minimizecostly stream sample chemical analyses when gathering BTC data during stream tracer experiments. Further,as the shoulder (initial rise), plateau (steady-state condition), and tail (concentration fall) of the BTC passesa sampling site, field personnel can make immediate informed decisions when to sample stream water forion analysis, being sure to capture a complete range of IC and SC values. Alternatively, the BTC can beoversampled in the field and the SC BTC data can be used to select key IC samples for analysis. This method,or a similar method, has been used by several investigators in previous studies (Butturini and Sabater, 1999;Fellows  et al ., 2001; Mulholland  et al ., 2001). However, none have outlined the method or quantified theenhanced reliability of higher resolution BTC data for TSM modelling.This method is purposely based on a developed IC–SC relationship. One tempting alternative would beto use SC alone, but such an approach does not account for temporal or spatial changes in SC from lateralinflows or upstream boundary conditions. Alternatively, the collection of real-time dissolved ion data with anion-specific probe (e.g. Hall  et al ., 2002; Paul and Hall, 2002) could provide the same result, a high-frequencyIC BTC, provided that probe sensitivity was comparable to laboratory ion analysis methods. Stream tracer experiment  A stream tracer experiment was performed in two 60 m stream reaches in catchment M15. Discharge datawas collected at a 90 °  V-notch weir located at the end of the experimental reach, 120 m downstream of the injection site (Figure 1). Discharge at 0 m and 60 m was regressed from the weir hydrograph based oncomputed travel times during the stream tracer experiment, and a proportional multiplication of the weirhydrograph such that the total injected mass of tracer was observed at the downstream sites (60 m and120 m). Specific conductance data was collected on a 2-min interval over a period of 4 days at 60 m and Copyright  ©  2005 John Wiley & Sons, Ltd.  Hydrol. Process.  19 , 2491–2506 (2005)  2494  M. N. GOOSEFF AND B. L. MCGLYNNFigure 1. (A) Hydrograph and (B) specific conductance data for M15 stream from 11 April 2000 to 06 May 2000, data collected at a weirlocated at the end of the experimental reach. Discharge data for 0 m and 60 m hydrographs regressed from weir data, transport characteristicsand synoptic data (see text). Curves in panel B represent best fit second-order polynomials 6 days at 120 m downstream of the injection point (Figure 1B) with Campbell Scientific model CS-247specific conductivity and temperature probes (Campbell Scientific, Inc., Logan, UT). From this long-termdata, a second-order polynomial model was fit to the SC data before and after the stream tracer experiment sothat the IC–SC relationship could be normalized to the SC induced by the addition of the ionic tracer alone.All SC data during the stream tracer experiment was normalized to account for changes in background SC,prior to developing the IC–SC relationship.A LiBr stream tracer injection solution with a Br  concentration of 9 Ð 6 ð 10 4 µ eq l  1 was injected into thecentreline of stream flow at a rate of 0 Ð 037 l min  1 for 12 Ð 92 h, beginning at 0120, 12 April 2000. Injectionrate was controlled using a Campbell Scientific CR-10x data logger (Campbell Scientific, Inc., Logan, UT) andFluid Metering pump model #PM6013 (Fluid Metering, Inc., Syosset, NY). We performed calibration checkson the injection rate throughout the tracer injection and found the injection rate to be constant throughout thetracer test. Stream water was collected from the thalweg for Br  analyses manually and with Isco automatedsamplers (Isco, Inc., Lincoln, NE) to establish a relationship between SC and Br  . Copyright  ©  2005 John Wiley & Sons, Ltd.  Hydrol. Process.  19 , 2491–2506 (2005)  A STREAM TRACER TECHNIQUE EMPLOYING IONIC TRACERS  2495A synoptic survey was performed, collecting samples at 5 to 10 m intervals from the weir to the injectionpoint over a period of 0 Ð 6 h after reaching steady-state conditions in the stream. Stream grab samples werefiltered prior to analysis through 0 Ð 45  µ m glass fibre syringe filters and then analysed for Cl  and Br  bystandard ion chromatography (Dionex model DX-120) methods at the Boulder, CO USGS laboratory. TSM simulations The OTIS model (Runkel, 1998) was used to simulate the conservative Br  breakthrough data for each of the reaches: ⊲ water column ⊳∂C∂t D Q A∂C∂x  C 1  A∂∂x    AD∂C∂x   C ˛⊲C S  C⊳ C q  L ⊲C  L  C⊳ ⊲ 1 ⊳⊲ storage zone ⊳dC S dt D ˛ A A S ⊲C  C S ⊳ ⊲ 2 ⊳ where  A  is the cross-sectional area of stream (m 2 ),  A S  is the cross-sectional area of storage zone (m 2 ),  ˛  isthe storage zone exchange coefficient (s  1 ),  C  is the main channel Br  concentration ( µ eq l  1 ),  C S  is thestorage zone Br  concentration ( µ eq l  1 ),  q  L  is the lateral inflow rate (m 3 s  1 m  1 of stream length),  C  L is the lateral inflow Br  concentration ( µ eq l  1 ),  D  is the dispersion coefficient (m 2 s  1 ),  x   is the distancedownstream (m),  t  is time (s), and  Q  is stream flow rate (m 3 s  1 ). Optimization of the TSM parameters  D ,  A ,  A S ,  q  L  and  ˛  was accomplished using UCODE (Poeter and Hill, 1998), as prescribed by Scott  et al . (2003a).We ran four simulations to test the IC–SC approach. In Simulation 1, Br  observations were derivedfrom a linear Br  –SC relationship developed at each sampling site. Stream samples were taken over thefull range of Br  concentrations and SC levels, to establish a robust relationship. This method produced ahigh temporal resolution data set, as SC was recorded at a 2-min interval in stream water at both 60 m and120 m downstream of the injection point, resulting in 1307 BTC data points at each site through the tracerexperiment. In Simulation 2, all of the stream Br  samples were used to simulate Br  transport downstream.This data set was more sparse, yet still intensive relative to standard TSM sampling schemes, composed of 55 observations at 60 m, and 60 observations at 120 m. In Simulation 3, half of the Br  observations ateach site were used in the TSM parameterization, on a nearly 2-h interval (28 observations at 60 m, and 30observations at 120 m). In Simulation 4, 20 Br  observations from each sampling site were used to optimizea TSM simulation.Solute transport simulations in unsteady discharge conditions are possible with OTIS, although one mustprovide changing discharge values and cross-sectional area (  A ) data as model input parameters. Previousstudies during unsteady discharge have utilized stream flow routing models to provide such data for solutetransport modelling (Runkel  et al ., 1998; Scott  et al ., 2003b). The consequence of this approach is that thestream cross-sectional area (  A ) cannot be optimized within the solute transport model. To enable optimizationof   A  during changing discharge and work within a single model structure, we used a simple hydraulic geometricrelationship to simulate stream cross-sectional area as a function of stream flow:  A D cQ 0 Ð 6 ⊲ 3 ⊳ where  c  is a coefficient. The OTIS code was modified to compute  A  as prescribed by Equation (3), and  c was optimized for each simulation method, for each reach. Simulations 1–4 incorporate the formulation of Equation (3) to calculate  A  as a function of   Q . We made one additional simulation (Simulation 5) applied to theBr  –SC data (similar to Simulation 1), using the standard OTIS model, without the Equation (3) formulationto test its utility.  Model performance assessment  In order to assess parameter optimization performance for the Br  –SC and the Br  observation simulations,we present several model fit and parameter estimation statistics generated during UCODE regression runs Copyright  ©  2005 John Wiley & Sons, Ltd.  Hydrol. Process.  19 , 2491–2506 (2005)
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks