How Old Is This Water?
An interactive primer on groundwater age — what the word can and cannot mean, how the radiometric clocks (¹⁴C, ³⁹Ar, ⁸¹Kr) work and where they fail, and why a residence time is a distribution rather than a number, from confined aquifers down to the water inside a tree. Builds on the decay law of Week 2; no dating background needed.
One water sample. Three careful methods. Three wildly different ages.
A well taps a shallow confined aquifer. Darcy's law, applied to the measured head gradient and a reasonable hydraulic conductivity, says the water took hundreds of thousands of years to arrive from the recharge area. Its radiocarbon activity is about a tenth of the pre-bomb recharge value — by the decay law you met in Week 2, less than 20,000 years. Its helium-4 content implies millions of years. (The situation is real; it is the worked example in Bethke & Johnson, 2008.)
Seen the new way, the sample is water that flowed along the aquifer and mixed with modern water recharging from the surface and with ancient water discharging from below. The Darcy estimate describes the flow path; the radiocarbon clock is dominated by the young fraction (a decaying tracer barely notices dead water, as Section 7 shows); the helium clock is dominated by the old fraction. Three honest instruments, three different questions answered. The disagreement is not an error to be explained away — it is information about the flow regime, and reactive transport modeling (Section 8) is how the field has learned to read it.
Cook & Böhlke call the enterprise "hydrochronology," by analogy with geochronology. The analogy is honest about the ambition — dating water as geologists date rock — and, as this week's papers show, honest ambition runs into an interesting problem: rocks mostly stay put, and water mixes.
"Age" is easy to define and easy to break
The definition sounds innocent: the age of groundwater at a point is the time that has elapsed since it entered the subsurface. Bethke & Johnson note, a little mischievously, that almost no textbook states even this — the texts jump straight to isotope formulas. Start slower, because the definition is where the trouble lives.
The packet picture (piston flow)
- A parcel of rain infiltrates, becomes a closed packet, and rides a flow line from recharge to well. Nothing enters or leaves it.
- Age then grows along the path at the pace of the flow: the age gradient is the reciprocal of velocity. Water moving 10 m yr⁻¹ is 100 years older every kilometre.
- Every classical dating formula in Section 3 quietly assumes this picture. One sample, one age, one equation.
The molecule picture (the "new thinking")
- Water molecules diffuse into and out of neighbouring aquitards, disperse along and across the flow, and leak between formations. There are no closed packets.
- So a sample is a crowd: each molecule has its own residence time. The sample's age is defined as the average over the crowd — and the crowd has a shape.
- Two samples with the same average can be utterly different waters. The shape, not the mean, decides what a well is telling you.
Same mean age, four different waters
Each panel below is a possible age distribution for a sample whose mean age is identical — the four sketches of Bethke & Johnson's Figure 2. Click through them. The x-axis is the age of individual water molecules; the curve is how much of the sample is that old.
Three ways a tracer can tell time
Water molecules carry no age information of their own. Every dating method therefore rides on a marker dissolved in, or part of, the water — and every marker keeps time in one of three ways (Cook & Böhlke §1.2; Bethke & Johnson's three classes). Click each clock.
Click a clock to see how it keeps time, what it needs to be trusted, and how it fails.
The ruler of time — and the holes in it
Each bar is the range of ages a tracer can usefully date; the axis is logarithmic, from one year to ten million. Hover or tap a bar for the tracer's story. Notice what Cook & Böhlke noticed: the widely available methods cluster in two windows — younger than ~50 years and roughly 2,000 to 40,000 years — leaving a gap between decades and millennia that only the hard-to-measure ³⁹Ar spans, and another beyond radiocarbon where ⁸¹Kr, ³⁶Cl and ⁴He take over.
The numbers behind the bars
| Tracer | Half-life | Clock type | Usable window (approx.) |
|---|---|---|---|
| ³H (with ³He) | 12.3 yr | draining / calendar | < ~50 yr |
| ⁸⁵Kr | 10.76 yr | calendar (rising) | < ~50 yr |
| CFCs, SF₆ | — | calendar (rising) | < ~50–70 yr |
| ³²Si | 152 yr | draining | ~30–1,000 yr; sorbs, unreliable to date |
| ³⁹Ar | 269 yr | draining | ~50–1,000 yr |
| ¹⁴C | 5,730 yr | draining | ~2,000–40,000 (to ~50,000) yr |
| ⁸¹Kr | 229,000 yr | draining | ~5×10⁴–1×10⁶ yr |
| ³⁶Cl | ~301,000 yr | draining / pulse | ~10⁵–10⁶⁺ yr |
| ⁴He | stable | filling | ~10⁴–10⁸ yr, semi-quantitative |
Half-lives as quoted in Bethke & Johnson (2008); windows as characterized in Cook & Böhlke (2000), Bethke & Johnson (2008), and Lehmann et al. (1993) — the ⁸¹Kr window is Lehmann et al.'s; the ³⁹Ar and ³²Si spans are the ranges conventional in the dating literature, implied by their half-lives. "Usable" assumes the initial condition and subsurface sources are under control, which is precisely what Sections 4 and 7 complicate.
¹⁴C, ³⁹Ar, ⁸¹Kr: dating old water
The week's title methods are the draining clocks for water older than living memory. Radiocarbon is the workhorse — cosmic rays make ¹⁴C in the atmosphere, it dissolves into recharge as CO₂, and its 5,730-year half-life spans the last ice age. The demo turns the crank on the decay law, then shows the two ways the answer lies to you.
The decay-clock calculator
Choose a nuclide, then set the measured concentration as a fraction of the initial (for ¹⁴C, think "percent modern carbon"). The curve is the decay law; the shaded band is the clock's comfortable range — roughly a tenth of a half-life to five half-lives. Outside it, the measurement drowns in precision limits (too young) or in background (too old).
Thief 1: dead carbon
The decay law needs C₀ — the ¹⁴C content the water started with. But groundwater dissolves calcite and oxidizes ancient organic matter along the way, and both add carbon with no ¹⁴C at all. The activity per gram of carbon drops, and the clock reads old without a single year passing. Drag the slider: q is the fraction of the dissolved carbon that actually came from recharge.
³⁹Ar — the gap-bridger
A noble-gas radionuclide (t½ = 269 yr) made by cosmic rays; chemically inert, so nothing sorbs it or dissolves extra into the water — theoretically the cleanest clock for the 50–1,000 year gap that ³H and ¹⁴C both miss. Two catches: measuring it has historically demanded heroic low-level counting on large samples (Loosli et al., 2000, in the Cook & Herczeg volume), and it has a subsurface source — neutrons from U and Th decay make ³⁹Ar from potassium via ³⁹K(n,p)³⁹Ar, so in U-rich rock the "atmospheric" signal needs auditing (Lehmann et al., 1993).
⁸¹Kr — the deep-time noble gas
Also inert, with t½ = 229,000 yr: the right decay constant for water tens of thousands to a million years old, where ¹⁴C is dead and ³⁶Cl is haunted by subsurface production and dead-chloride dilution. Lehmann et al. (1993) single out ⁸¹Kr (with ³H/³He at the young end) as the dating method with the least interpretive complication — its subsurface production is negligible — and its historical obstacle as purely analytical: counting a handful of atoms. Their conclusion aged well; atom-counting techniques have since turned both ⁸¹Kr and ³⁹Ar from theoretical into practical tools, which is why your reading list keeps them in the title of Week 3.
Where the ages live: groundwater stratigraphy
Before distributions, get the ideal geometry into your hands. In an unconfined aquifer of constant thickness receiving uniform recharge — Vogel's classic 1967 model, the backbone of Cook & Böhlke's §1.3 — age is stratified like sediment: youngest at the water table, oldest at the base, with the isochrons crowding together logarithmically at depth. Cook & Böhlke call reading this record "groundwater stratigraphy," and it is how single wells become recharge gauges.
The Vogel aquifer
Drag the sliders; the cross-section recolours by age and the profile replots. Hover either panel to probe a depth. The age at depth z is independent of where you drill horizontally — one of the model's quiet surprises.
Cross-section, coloured by age
Age–depth profile at a well
What people actually do with this
- Recharge from a dated profile. Fit the age–depth curve from multi-level samplers and R falls out. At Cape Cod, ³H and ³He peaks ~27 m below the water table gave a water-table velocity R/ε of 3.3 m yr⁻¹ (Solomon et al., 1995, in Cook & Böhlke §1.4.1); with ¹⁴C, recharge rates as small as 0.1–0.2 mm yr⁻¹ have been extracted (Murray Group, Leaney & Allison, 1986). Across methods, ³H/³He and CFCs have mapped recharge of 100–1,200 mm yr⁻¹; ¹⁴C covers 2–100 mm yr⁻¹.
- Velocity from ages along a confined aquifer. Carrizo Sand: ¹⁴C ages rise downgradient from <4,000 to >28,000 yr, giving 2.4 m yr⁻¹ at 15 km and 1.6 m yr⁻¹ at 50 km — agreeing with the hydraulics (Pearson & White, 1967). Age differences between wells dodge the initial-condition problem entirely (Phillips et al., 1989).
- The unsaturated zone is part of the story. The tracer's total residence time is root zone + unsaturated zone + groundwater (Cook & Böhlke Fig. 1.3). A ³H peak parked 4 m deep in the English Chalk in 1968 — rainfall's 1963–64 bomb peak, four to five years on — gave 0.9 m yr⁻¹ of downward crawl (Smith et al., 1970); at Cape Cod the unsaturated lag was 14 years. Forgetting this lag shifts every "recharge date" you infer.
- Groundwater as archive. Date the water, then read its passengers against the calendar: nitrate loading beneath Maryland farmland rising three- to six-fold from 1950 to 1990, tracking fertilizer use, with ~30% of applied N reaching the water table (Böhlke & Denver, 1995); sulphate recharge peaking with 1960s–70s acid deposition; a halocarbon history of the Danube reconstructed from a riverside aquifer (Böhlke et al., 1997). And with age in hand, slow chemistry gets rate constants no laboratory could measure: denitrification with a ~5,000-year half-life in the Kalahari (Vogel et al., 1981), dedolomitization at tenths of a µmol L⁻¹ yr⁻¹ over 30,000 years in the Madison aquifer (Plummer et al., 1990).
What a well actually samples
Wells are rarely points. A screen spanning the aquifer, a spring, a stream — each collects flow lines of every age at once, and the sample's tracer content is set by the whole age frequency distribution g(t), not by any single age. The classical "lumped-parameter models" are a small zoo of g(t) shapes, each the exact consequence of an idealized geometry you have already seen (Cook & Böhlke Table 1.1, Fig. 1.4; Maloszewski & Zuber's framework).
The distribution explorer
Pick a model and a mean transit time τ. The main panel is g(t) — the fraction of discharge in each age band; the dashed curve is the cumulative version. The readouts answer the question a distribution is for: how much of this water is young? Two axis views: in years, dragging τ stretches the whole family; in multiples of τ — the convention of Cook & Böhlke's Fig. 1.4 — every τ collapses onto one universal shape, so that view deliberately sits still as τ moves (and "older than 2τ", a property of shape alone, never moves in either view).
This is the machine that connects a distribution to a measurement: feed the tracer's input history through g(t), decay it, and out comes what the well should read today. Fitting measured concentrations of tracers with different input histories — ³H and CFC-12 and SF₆ together — can pin down both τ and the shape of g(t), because each tracer averages the distribution differently (Cook & Böhlke Fig. 1.5, computed for 1998 sampling; the same logic runs TracerLPM in Week 5).
The apparent age of a mixture
Now run the dating formulas of Section 3 on water that is honestly a mixture — and watch each class of clock fail in its own characteristic direction. This is Bethke & Johnson's Figure 3 made adjustable, and it is the single most useful diagram in the week.
Two waters, one sample
Blend a younger water A with an older water B and read the sample three ways. The true mean age is the flow-weighted average (the "age mass" rule of Section 8). Each method's apparent age is what its formula returns for the blended concentration.
Why decay clocks read young
- Concentration decays along a curve; mixtures sit on the straight chord between the end-members — and the chord lies on the young side of the curve, always.
- The bias is mild for similar ages (longitudinal dispersion) and extreme for young + very old: half modern water, half radiocarbon-dead water reads as one ¹⁴C half-life — 5,730 years — however ancient the old half is.
- The same geometry afflicts calendar clocks whenever part of the blend pre-dates the marker's history.
Why the filling clock doesn't
- ⁴He grows linearly with time, and a straight line is its own chord: mixing preserves the mean exactly (when the source rate is uniform).
- The catch is the premise. Aquitards often out-produce aquifers, the deep crust leaks ⁴He upward, and an unnoticed uranium deposit reads as stagnant water (Bethke & Johnson's error analysis).
- So the honest clock has the dirty source, and the clean-sourced clocks lie about mixtures. Every dating study lives somewhere on that trade.
Age mass: making "age" something that can be transported
If a sample's age is an average over molecules, age must be carried, mixed and diluted like a solute. Ages themselves don't add — mixing 20-year and 40-year water does not make 60-year water — so Goode (1996) found the quantity that does: age mass, the product of water mass and age. One kilogram of ten-year-old water holds ten kilogram-years. Age mass mixes linearly (the equal blend above is 30 years old), obeys a transport equation, and turns age from an isotope formula into a field you can model.
Read the terms: age disperses, age advects, and the "+1" says every kilogram of water everywhere grows one year older per year — a uniform, zero-order source that no real tracer quite matches. Age behaves like a reacting solute with the simplest source law in the world, which is exactly why real tracers (each with its own source law) are imperfect proxies for it.
Aquitards age aquifers simply by existing
Solve the steady transport equation across an aquitard sandwiched between two aquifers (Bethke & Johnson eqs. 19–24, their Figure 4). With no cross-flow the excess-age profile is a parabola peaking at b²/2D*. Axes are absolute — depth in metres inside a fixed ±100 m frame, excess age in years — so both sliders reshape the curve: b widens the aquitard and grows the peak as b²; D* stretches or shrinks it horizontally. Then note the readout the sliders cannot change.
Excess age across the aquitard
Reactive transport modeling: inverting the tracers together
The alternative to dating samples one formula at a time: build a 2-D flow model of the whole regime, transport each marker with its true source law (decay, in-situ production, basal flux), and adjust permeabilities until the simulated concentrations match every observation at once — heads, salinity and temperature included. The age field then falls out of the same model via the +1 equation. Three worked cases from Bethke & Johnson:
Age distributions in the soil–plant system
Everything so far scales down. Replace "aquifer" with a three-metre soil profile, "recharge" with sprinkler rainfall, "well" with a tree — and the age of water becomes a days-to-months question with the same mathematics. Block 3's paired papers do exactly this at Biosphere 2's enclosed rainforest, where the boundary conditions of a real catchment's problem are, for once, actually known.
The experiment (Evaristo et al., 2019)
The Biosphere 2 Tropical Rainforest: a 1,936 m² glass-enclosed mesocosm near Tucson, ~27 m tall, soils 1–3.5 m deep over a drained basement — so everything that leaves the bottom ("deep percolation," the stand-in for groundwater recharge) is collected and measured. The team switched the rain off for 68 days, then rewet the system with 66 mm of deuterium-labelled rainfall at +152‰ δ²H over four events — screamingly distinct from the −60‰ background — followed by 87 mm of ordinary rain over 13 events. For months afterward they tracked the label through bulk soil water at seven depths, the basement seepage, and weekly xylem water from five canopy species, sampled by clipping from a bosun's chair.
Read those three tiles together and the paper's two findings appear. Separation in time: water leaving as recharge moved 2–7× faster than water leaving through trees. Separation in space: the Bayesian mixing model says trees drew 89 ± 6% of their water from bulk soil-matrix storage rather than the freely draining "mobile" water — even though soil at 25 cm is older on average than the seepage collected at 3 m, which only makes sense if much of the seepage bypassed the soil matrix through preferential flowpaths. The old phrase "tightly bound water" earns a correction here: the matrix water trees used was not bound at all, merely slow to drain — the authors adopt Brantley et al.'s term "matrix water," and the two-water-worlds offset of Week 2 acquires a transport mechanism.
Gamma-shaped ages: the explorer
The 2019 analysis fits each breakthrough with a gamma transit-time distribution — the two-knob family (shape α, scale β; mean = αβ) that catchment hydrology reaches for first. α is the shape's whole personality: α < 1 gives an L-shaped, young-dominated distribution with a long tail; α ≈ 1 is exponential (the well-mixed reservoir of Section 6); α > 1 rises to a delayed hump. Load each fitted preset and read its story.
The breakthrough curve, parametric and phenomenological
The 2026 Comment reopens the same Biosphere 2 dataset with a teaching agenda: strip transit-time analysis down to two approaches simple enough for any lab group to run, apply both to every breakthrough curve (BTC), and be honest about what neither can promise. It is the shortest path in this course from "I have tracer data" to "I have a defensible timescale."
Parametric (mechanistic)
- Fit the BTC with a theoretical TTD and interpret the parameters. Two candidate shapes, two hypotheses about plumbing: gamma = serial (water queues through compartments in series — the sum of exponential steps), lognormal = parallel (many simultaneous pathways whose effects multiply).
- Selection by a six-criteria vote (AIC, BIC, R², RMSE, KS, Anderson–Darling): gamma won 7 of 8 soil curves; lognormal won 4 of 5 xylem curves. Serial soils, parallel trees — as a hypothesis, not a verdict; the paper is explicit that the fits are descriptive, and several are near-ties.
Phenomenological (data-based)
- No assumed shape. Read three numbers straight off the curve: tz, when the peak arrives; tμ, the concentration-weighted mean time; and Δt = tμ − tz, a tailing metric — how far dispersion and retention drag the mean past the peak.
- Plus a velocity: flowpath length over tμ — 2–14 cm d⁻¹ through the soils, 18 cm d⁻¹ to deep percolation, and 52–193 cm d⁻¹ up the trees.
- The cost of assumption-freedom: the numbers only integrate over the observed window, so where the tail is unfinished (the trees, above all) tμ is a lower bound.
Anatomy of a breakthrough curve
A synthetic BTC with two knobs: when the peak comes, and how heavy the tail is. Watch the three phenomenological metrics respond — then truncate the record and watch tμ quietly shrink, which is precisely why the Comment reports its empirical means as lower bounds.
What the reanalysis found — and proposed
- Preferential flow, quantified. In one soil pit the tracer peaked at 2 days at every depth from 15 to 130 cm — vertically coherent bypass channels — while the neighbouring pit's peaks walked down (2 → 20 days). Two pits a few metres apart, two different plumbing diagrams: a warning against single-profile generalization.
- "Xylem bias," a named hypothesis. Tracer velocities up the five trees spanned 0.52–1.93 m d⁻¹, and the fraction of labelled water at the xylem peak averaged only ~20% — roughly 80% of what trees transpired was older, stored water. The Comment proposes that stands route a disproportionate share of transpiration through hydraulically "advantaged" individuals — preferential flow's canopy-level analogue — offered explicitly as something to test, with sap-flux and storage measurements named as the test.
- Method matters at the tree scale. Parametric and data-based means nearly agree in soils (14 ± 2 vs 12 ± 2 days) but split two-fold in plants (59 ± 13 vs 29 ± 6 days) — the unfinished tails again. And a concentration BTC is not a TTD: without coupling concentrations to water fluxes, all such timescales are comparative metrics, not absolute means. The Comment says so of its own numbers, repeatedly — model that candour.
- The stakes. A table of ten climate and terrestrial-biosphere models (CLM, JULES, ORCHIDEE, …) shows none representing the weeks-to-months lag between infiltration and transpiration, or tree-internal storage, that both papers document. The "call for broader dialogue" of the title is aimed there — at ecophysiology, soil physics and catchment hydrology jointly instrumenting the next experiment.
2019 vs 2026: the same trees, different numbers — a discussion point, not an erratum
| Outflux | 2019 MTT (flow-weighted gamma) | 2026 MTT (parametric BTC fit) | 2026 tμ (data-based) |
|---|---|---|---|
| Seepage / recharge | ≈ 9 d | 22 d | 19 d |
| C. racemosa | ≈ 17 d | 53 d | — |
| H. elatus | ≈ 21 d | 45 d | — |
| H. crepitans | ≈ 62 d | 109 d | 47 d |
| P. indicus | excluded (no label uptake) | 18 d | 15 d |
Per-species tμ is quoted in the Comment's text only for the fastest and slowest trees; across species tμ spans 15–47 days. Same dataset, defensible choices at every step, materially different numbers. The 2019 study convolved the input in flow-weighted time and required tracer-mass accounting; the 2026 Comment fits concentration curves directly and says its timescales are experiment-specific comparisons. Nothing here is hidden — the Comment cites the 2019 uncertainty analysis as the fuller treatment. For the seminar: which choices moved the numbers, and which number would you put in a model? (Notice the pattern of Section 7 returning at days-scale: methods weight the same mixture differently.)
Four papers, three blocks, one argument
Cook & Böhlke supply the working machinery of groundwater dating; Bethke & Johnson dismantle its central assumption; the Block 3 pair rebuilds the same ideas at the scale of a soil profile and a tree. The guides below are maps, not substitutes: what each paper asks, where its load-bearing figures live, which equations to recognize, and where a careful reader might press. The notes and the critique are yours to write.
Cook, P. G., & Böhlke, J.-K. (2000). Determining timescales for groundwater flow and solute transport. Ch. 1 of Cook & Herczeg (eds.), Environmental Tracers in Subsurface Hydrology. Kluwer.
Thirty pages that set up the whole enterprise: tracer types, the transport models, and what age data are used for. The reading list cites the volume by its editors (Cook & Herczeg); the chapter itself is by Cook and Böhlke. doi:10.1007/978-1-4615-4557-6 (volume)
Bethke, C. M., & Johnson, T. M. (2008). Groundwater age and groundwater age dating. Annu. Rev. Earth Planet. Sci., 36, 121–152.
The conceptual reset: age as a property of molecules, samples as mixtures, and reactive transport modeling as the way forward. Thirty-two pages, generously illustrated, no field data of its own — every example is a model with a purpose. doi:10.1146/annurev.earth.36.031207.124210
Evaristo, J., Kim, M., van Haren, J., Pangle, L. A., Harman, C. J., Troch, P. A., & McDonnell, J. J. (2019). Characterizing the fluxes and age distribution of soil water, plant water, and deep percolation in a model tropical ecosystem. Water Resour. Res., 55(4), 3307–3327.
The controlled experiment: a drought, a deuterium-labelled rewetting, and transit-time distributions for seepage, soil and transpiration in one closed system. Open access (CC BY-NC-ND). doi:10.1029/2018WR023265
Evaristo, J., Wright, C., Bauser, H. H., Knighton, J., Johnson, D. M., & Kim, M. (2026). Tracer labelling and transit time modelling in soil–plant systems: perspectives and a call for broader dialogue in ecohydrology. Ecohydrology, 19(1), e70182.
A Commentary that reuses the 2019 dataset as a teaching testbed: two accessible modelling routes, a named hypothesis (xylem bias), and an argument about what land-surface models are missing. Open access, with code and data links. doi:10.1002/eco.70182
Four situations to reason through
Each scenario gives the kind of evidence a Week 3 reader should now be able to interrogate. Choose the reading you would defend first, then compare with the debrief.
Ten questions before the meeting
Immediate feedback, no grade, no record. If you miss one, the linked section is the fix.
Glossary
The Week 3 working vocabulary, plus a few terms you will meet again in Weeks 5 and 11. Search or browse.
The Week 3 readings — and two optional companions
Papers are not posted for this course: retrieving them from the citation and DOI — via the UGA Libraries, GALILEO, or the publisher — is part of the training. Two of this week's four are open access.
Cook, P. G., & Herczeg, A. L. (eds.) (2000). Environmental Tracers in Subsurface Hydrology, Ch. 1 — Determining timescales for groundwater flow and solute transport (Cook & Böhlke).
Tracer types and time windows; Vogel-type age models and the lumped-parameter zoo; recharge, velocity, archive and reaction-rate applications. The week's machinery. doi:10.1007/978-1-4615-4557-6
Bethke, C. M., & Johnson, T. M. (2008). Groundwater age and groundwater age dating. Annu. Rev. Earth Planet. Sci., 36, 121–152.
Age as a mixture property; the three clock classes and their mixing biases; age mass, aquitards, and reactive transport inversion. The week's argument. doi:10.1146/annurev.earth.36.031207.124210
Evaristo, J., et al. (2019). Characterizing the fluxes and age distribution of soil water, plant water, and deep percolation in a model tropical ecosystem. Water Resour. Res., 55(4), 3307–3327.
A drought–rewetting D₂O experiment at Biosphere 2: transit times for seepage (~9 d) versus transpiration (17–62 d), and matrix-water use by trees. Open access. doi:10.1029/2018WR023265
Evaristo, J., et al. (2026). Tracer labelling and transit time modelling in soil–plant systems. Ecohydrology, 19(1), e70182.
The 2019 dataset reread two ways — parametric (gamma/lognormal) and phenomenological (tz, tμ, Δt) — plus the xylem-bias hypothesis and the case to the land-surface-modelling community. Open access. doi:10.1002/eco.70182
Suckow, A. (2014). The age of groundwater — definitions, models and why we do not need this term. Appl. Geochem., 50, 222–230. · Lehmann, B. E., Davis, S. N., & Fabryka-Martin, J. T. (1993). Atmospheric and subsurface sources of stable and radioactive nuclides used for groundwater dating. Water Resour. Res., 29(7), 2027–2040.
Suckow pushes this week's logic to its provocative conclusion — retire the word "age" altogether in favour of tracer-specific statements. Lehmann and colleagues audit every nuclide's atmospheric and in-situ sources, and conclude ³H/³He and ⁸¹Kr carry the least interpretive baggage. Not discussed in a block.
Coming after the holiday — Groundwater dating II: young water
The under-50-year window gets its own meeting: ³H/³He as a self-starting clock, CFCs and SF₆ as calendar clocks with atmospheric complications, and the lumped-parameter models of Section 6 turned into a working tool (the TracerLPM workbook). Everything this week said about mixtures applies with the signs filled in.
Carry-forward question, for your proposal topic: for the water your question cares about, what age range do you expect — and does a tracer exist whose window, source terms, and mixing behaviour could actually resolve it? If the answer is "not cleanly," that is not a dead end; it is a proposal rationale.