Cargese Summer School on Flow and Transport in Porous and Fractured Media, Development, Protection, Management and Sequestration of Subsurface Fluids, July 20th - August 1st 2015
3. Residence time in the compartments of the water cycle
Aeschbach-Hertig, W., and T. Gleeson (2012), Regional strategies for the accelerating global problem
of groundwater depletion, Nature Geoscience, 5(12), 853-861.
5. Tracer Concentrations &
GW Ages
Hinsby K (2001): Freshwater – our most important resource. – In: Hinsby and Binzer “Freshwater our most
important resource – Geology and groundwater models”, special issue of Geologi – Nyt fra GEUS, nr.1 – 2001
6. Tracer Concentrations &
GW Ages
0
200
400
600
1940 1960 1980 2000
CFC-12(pptv)
c(tw) (mol/l) →water
c(tw) (pptv) →air
tr
Apparent age A
tw
)(1
winww tcCttA
Tracer concentration c
/R
l
tA w
Park, J., et al. (2002), Transport modeling applied to the interpretation of groundwater Cl-36 age, Water Resources Research, 38(5).
10. Exponential TTD for
aquifers at wells
/
1
R
H
etP
t
H: Mean aquifer depth
: Aquifer porosity
R: Aquifer recharge
Haitjema, H. M. (1995), On the residence time distribution in idealized
groundwatersheds, Journal of Hydrology, 172(1-4), 127-146.
11. GW Age, Transit Time Distribution, Mixing
No mixing (piston-flow model) Full Mixing (exponential model)
/R
l
/R
H
ttp
t
etp
1
Hl
12. Transit Time Distribution and Transport
Ginn, T. R. (1999), On the distribution of
multicomponent mixtures over generalized
exposure time in subsurface flow and reactive
transport…, Water Resources Research, 35(5),
1395-1407.
S
t
p
pp
t
p
u
Dv
Cornaton, F. J. (2012), Transient water age
distributions in environmental flow systems:
The time-marching Laplace transform solution
technique, Water Resources Research, 48.
13. Infering Transit Time Distribution from GW Age
▪ Apparent age A
▪ Direct problem
▪ Inverse problem
▪ Use of multiple tracers (multiple GW ages)
▪ Simplify the model of transit time distributions?
▪ Dirac, Exponential,…, Lumped Parameter Models
▪ Broad variety of natural distributions?
▪ Geological conditions, old versus young GW
▪ Sampling conditions
▪ Hydrological conditions
▪ Reduce the distribution to the mean, standard deviation, shape?
0
11
)()()( dttpttCCttcCttA wininwwinww
15. ▪ Fully-heterogeneous 3D models
Methodology
PhD S. Leray (2012), Caractérisation des aquifères de socle cristallin et de leur ressource en eau-
Apport des données d’ « âge » de l’eau, University of Rennes 1.
20. Transport model
▪ Advection, no diffusion
▪ Diffusion/dispersion vs pumping, heterogeneity
▪ Backward-time from the pumping well
ttdΓtp
Γ
Γ
x
w
sΓ
s
w
&),(),()(avec
sur0)).,(),((
sur0),(
0)0,(
0)()()),(
),(
.(
),(
*
"imposéCgrad"
*
imposée"C"
*
*
*
*
xxq
nxxq
x
x
x
xqx
34. San Joaquin Valley’s Aquifer
Transit Time Distributions
Lumped Parameter Models
Green, C. T., et al. (2014), Accuracy of travel time distribution (TTD) models as
affected by TTD complexity, observation errors, and model and tracer selection,
Water Resources Research(50), 6191 - 6213.
Predictions of
Nitrate
concentrations
35. Conclusions
▪ Large variety of Transit Time Distributions
▪ Sensitive to geological, hydrological, topographical constraints
▪ Limited number of Lumped Parameter Models
▪ Effective for bulk predictions on renewal times, nitrate concentrations
▪ Restrictions in the use of Lumped Parameters Models
▪ High influence of sampling (largely unknown)
▪ Tracer concentrations may be affected by reactivity, contamination,….
▪ Relating parameters to flow structures, hydraulic parameters
▪ Modification of boundary conditions, transient state
▪ Spatial variations in contaminant sources
▪ Combination of hydraulic and geochemical information
▪ Hydraulic Model give the shape of the distribution
▪ Tracers give the right order of magnitude
38. TTDs & flow patterns
Eberts, S. M., et al. (2012), Comparison of particle-tracking and lumped-parameter age-distribution models for
evaluating vulnerability of production wells to contamination, Hydrogeology Journal, 20(2), 263-282.
39. TTDs & flow patterns
Eberts, S. M., et al. (2012), Comparison of particle-tracking and lumped-parameter age-distribution models for
evaluating vulnerability of production wells to contamination, Hydrogeology Journal, 20(2), 263-282.
Hinweis der Redaktion
Purpose, objective
Orders of magnitude, Transit times in some aqufiers, mostly shallow aquifers
What is GW Age and how to measure it?
Is it worth to any predictions?
Transit Times:
unconfined, confined, deeper aquifers
Leakage: some relations between aquifers
Fundamentally a distribution
Residence Times, Renewal Times
Contaminant is still in the aquifer (has not been transfered)
Water has not been renewed, and in a broader sense, contaminant has not been flushed
Vulnerability: Effect of a change in nonpoint source contaminants
Groundwater Ages:
Some average over the aquifer: easy concept to assess vulnerability and sustainability, what is vulnerable is sustainable
Mean Residence Time
Observable: at the core of the concept, groundwater age can be approached, some groundwater age can be approached
Groundwater dating
How to measure
What to measure
Many ways to use and interprete tracers (Eberts, 2012: Introduction)
Fraction of young groundwater (<60 years)
Fraction of old groundwater (>1000 years)
Response to Non-point source contamination: dilution of this acontamination in the aquifer: distribution of groundwater age
Hydrogeology and Geochemistry
Interface between the 2 disciplines
Groundwater dating: Préciser le passage eau-air
Age is not a measure, it results from a model
Concentration (CFC, SF6, 3H, 4He, 14C, 39Ar) (IAEA, 2006)
Age is a transformation of concentration, observation c(tw)
Piston-flow models: no mixing, local-scale mixing to account for diffusion-dispersion processes
Point-like sampling, sampling at a given depth of the well
More in divergent zones, close to recharge
Characteristic time: distance over recharge
Exponential model: Global (Total) mixing (continuous stirred tank reactor)
Effect of the well screen: dispersion in the well
Obvious: broad distribution of times in a single aquifer, distribution of times is not necessarily an effect of tappling in multiple aquifers
Convergence of the flow lines towards the well (interaction of the well with the aquifer)
More generally in the convent zones, close to discharge
Characteristic time: characteristic thickness of the aquifer over recharge
Origin of simplified models from chemistry (Lumped parameter models)
For reactivity
Towards estimating optimal residence time: Damköhler number close to one
Large enough to be reacting
Small enough to optimize global reactivity
Piston-flow models: no mixing, local-scale mixing to account for diffusion-dispersion processes
Point-like sampling, sampling at a given depth of the well
More in divergent zones, close to recharge
Characteristic time: distance over recharge
Exponential model: Global (Total) mixing (continuous stirred tank reactor)
Effect of the well screen: dispersion in the well
Obvious: broad distribution of times in a single aquifer, distribution of times is not necessarily an effect of tappling in multiple aquifers
Convergence of the flow lines towards the well (interaction of the well with the aquifer)
More generally in the convent zones, close to discharge
Characteristic time: characteristic thickness of the aquifer over recharge
Transit time distribution is fundamentally the result of a model (LPM, mechanistic)
Complementarity between
geochemical perception: measured age (lag to an event-nuclear test, level of concentration)
Punctual but well calibrated
Hydraulic perception: flow, velocity, age gradient
Spatially exhaustive but uncalibrated distribtion of age
Surface libre inconnue (himp -> R trop imp// Qimp->h himalaya)
Détermination du champ de vitesse
Misinterpretation of data, bad reproduction of p(t)
Assessment of
(1) The ability of Lumped parameter models to match synthetic (simulated) Transit Time Distributions
(2) Their capacity to give sound predictions to change in nonpoint source contaminants (vulnerability)
2-parameter models good enough
Influence of parameters on predictions
1. Error on data, sampling, not analytics, what do data represent
No influence
Complexity of the TTD
Lumped Parameter Model
4 field sites: major US aquifers: well producing sites (quite permeable, with strong level of recharge)
Aquifers providing for 35% of the water public suppy of the US
Modesto (Central Valley Aquifer)
Semi-Arid conditions (300 mm/year)
Increase of pumping and irrigation, increase of recharge rates (up to 600 mm/year)
Unconfined to semi-confined (?) conditions
Shallow water table (10m to the surface)
Alluvial fan sediments with lenses of gravel, sand, silt and clay
Long-screen wells: downward flow migration and contamination in the absence of pumping
Tampa Florida
Subtropical humid climate (1140 mm/year)
Carbonate rocks, karst features: Underlying aqufiers (tapped)
Overlying confining unit of clays breached by sinkholes associated with the underlying carbonate rocks
Overlying unconfined aqufier (sand, clays, marls), Water Table, 3m from the surface
Woodbury (Connecticut)
Humid climate (1170 mm/year)
Unconfined Glacial aquifer system
Underlying crystalline basement (granite, schist, quartzite, gneiss)
Water Table : 3m from the surface
York, Nebraska
Humid, continental climate, 711 mm/year
Layered sedimentary basin settings
Uppermost confined aquifer (tapped)
Confining unit continuous (8m to 17m)
Overlying and underlying aquifers heavily used for irrigation
Observations of environmental tracers: SF6, CFCs, tritium, titriogenic (?) helium
2 faster aquifers: karst, unconfined
Karst: smaller volume, smaller variance
Glacial: larger volume, larger variance
2 slowest aquifers: deeper, confined
Disconnection to the surface: does not start at zero
Long well screen
Leakage: larger times, larger variance
LPMs and distributions
Close, built on the same assumptions
Differences between mean and median
Some details not represented (right? Importance?)
Parameters of the LPMs: mean transit time is not a good characteristic, the worst possible
Too large for the early times
Too small for the large times