Testing mixing models of old and young groundwater in a tropical lowland rain forest with environmental tracers

Abstract

[1] We tested three models of mixing between old interbasin groundwater flow (IGF) and young, locally derived groundwater in a lowland rain forest in Costa Rica using a large suite of environmental tracers. We focus on the young fraction of water using the transient tracers SF 6 He. Because of their unique concentration histories in the atmosphere, combinations of transient tracers are sensitive not only to subsurface travel times but also to mixing between waters having different travel times. Samples fall into three distinct categories: (1) young waters that plot along a piston flow line, (2) old samples that have near-zero concentrations of the transient tracers, and (3) mixtures of 1 and 2. We have modeled the concentrations of the transient tracers using (1) a binary mixing model (BMM) of old and young water with the young fraction transported via piston flow, (2) an exponential mixing model (EMM) with a distribution of groundwater travel times characterized by a mean value, and (3) an exponential mixing model for the young fraction followed by binary mixing with an old fraction (EMM/BMM). In spite of the mathematical differences in the mixing models, they all lead to a similar conceptual model of young (0 to 10 year) groundwater that is locally derived mixing with old (>1000 years) groundwater that is recharged beyond the surface water boundary of the system.

The practicalities of using CFCs and SF6 for groundwater dating and tracing

“…Other studies which report and consider the presence of CFCs or SF 6 in waters with relatively low 14 C activity are those of Cook et al (2005) and Solomon et al (2010).…”

Residence time distributions for hydrologic systems: Mechanistic foundations and steady-state analytical solutions

“…This generality stems from the fact that, regardless of any specific system being considered (watershed, lake, ocean, etc…), water is moving and cycling into and out of neighboring systems, and the amount of time spent in any section of the connected network is an important consideration for many problems. Residence time has application to water quality, risk assessment, contaminant remediation, characterization, habitat restoration, toxicity, reaction rates, age dating, turnover times in lakes, and ocean circulation, amongst others (Cirpka and Kitanidis, 2001;Delhez et al, 1999;Maxwell et al, 2003;Neumann et al, 2008;Seeboonruang and Ginn, 2006;Solomon et al, 2010). Despite this wide range of applications, the principles of residence time are fundamentally the same in that they are all concerned with the amount of time water, or some element transported by it, has spent in the system.…”

“…Systems presented in section 3 -and more generally, any system whose flow paths are analytically described -can be plugged together to generate composite RTDs. Examples of linear combinations include the exponential and dispersion models (Stolp et al, 2010), exponential and shape-free models (Goderniaux et al, 2013), two piston flow models (Eberts et al, 2012), exponential and piston-flow models (Eberts et al, 2012;Solomon et al, 2010), two exponential-piston-flow models (Green et al, 2014), and multiple dispersion models (Engdahl and Maxwell, 2014;Long and Putnam, 2009;McCallum et al, 2014).…”

“…The hydrological literature abounds with studies based on analytical solutions for RTDs: (Bockgård et al, 2004;Chen et al, 2011;Knowles et al, 2010;Manning and Caine, 2007;Osenbrück et al, 2006;Solomon et al, 2010;Zouari et al, 2011) for a few recent examples. Since there is no practical way to empirically measure residence time distributions, a commonly used approach is parametric fitting of trial distributions to chemical concentrations (Turnadge and Smerdon, 2014 …”

Inference of long-term groundwater flow transience using environmental tracers: A theoretical approach

“…Environmental tracers-including anthropogenic or cosmogenic radioactively decaying tracers such as 3 H/ 3 He, 85 Kr,36 Cl, 39 Ar, and 14 C; tracers with variable atmospheric concentrations as a result of anthropogenic activities, such as CFCs, SF 6,3 H, and 85 Kr; and linearly accumulating ones, such as 4 He and 40 Ar-have been widely used for inference of the parameters of presumed forms of GADs [Long and Putnam, 2009;Massoudieh et al, 2012;Solomon et al, 2010] or for constraining them [Corcho Alvarado et al, 2007]. This is typically done by presuming a mathematical form for the age distributions (also known as lumpedparameter models, LPM) and then estimating their parameters using a maximum likelihood-based approach by finding the values of the parameters of the LPM that minimizes some measure of the difference between the modeled and measured tracer concentrations [Bauer et al, 2001;Long and Putnam, 2009;Solomon et al, 2010], probabilistic methods such as Bayesian inference [Massoudieh et al, 2012], or by using direct deconvolution techniques [Massoudieh and Ginn, 2011].…”