Tools › Environmental › Groundwater › Groundwater Plume Transport › Method
Groundwater Plume Transport: Method
Open the Groundwater Plume Transport tool
The Domenico Solution, Its BIOSCREEN and BIOCHLOR Forms, What It Assumes and When It Misleads
Scope
The page estimates concentrations in a dissolved plume downgradient of a continuing source, with the analytical model that the U.S. Environmental Protection Agency distributed in two spreadsheet tools: BIOSCREEN (petroleum hydrocarbons; User's Manual EPA/600/R-96/087 and the Version 1.4 revisions) and BIOCHLOR (chlorinated solvents; EPA/600/R-00/008). Both rest on the solution of Domenico (1987). The page computes three models side by side, as BIOSCREEN does: transport with no decay, with first-order decay, and with BIOSCREEN's instantaneous reaction. It is a screening calculation: a first estimate of how far and how fast a plume may travel, of whether a receptor may be reached, and of how much decay matters.
Not covered: BIOCHLOR's sequential decay of parent and daughter compounds and its two reaction zones, BIOSCREEN's plume mass balance and mass flux, pumping wells, layered or non-uniform flow, and any transport through the unsaturated zone.
The model
Groundwater flows uniformly along x at the seepage velocity vs = K i / ne (hydraulic conductivity, hydraulic gradient, effective porosity). Linear, reversible sorption slows the contaminant to v = vs/R, with the retardation factor R = 1 + ρb Koc foc / ne (bulk density, organic-carbon partition coefficient, fraction of organic carbon). The source is a vertical rectangle across the flow at x = 0, of width Y and of thickness Z below the water table, held at the concentration C0. With dispersivities αx, αy, αz and a first-order decay coefficient λ the manuals write
C(x, y, z, t) = (C0/8) fx fy fz
fx = exp[ x (1 − s) / (2 αx) ] erfc[ (x − v t s) / (2 √(αx v t)) ], s = √(1 + 4 λ αx / v)
fy = erf[ (y + Y/2) / (2 √(αy x)) ] − erf[ (y − Y/2) / (2 √(αy x)) ]
fz = erf[ (z + Z) / (2 √(αz x)) ] − erf[ (z − Z) / (2 √(αz x)) ]
with z measured down from the water table, so that the plume spreads downward only (the BIOSCREEN 1.4 and BIOCHLOR form). With αz = 0, the default of both programs, fz = 2 inside the source thickness and the plume does not thicken. At long times the erfc term tends to 2 and the profile becomes steady: on the centerline the concentration falls as exp[x (1 − s)/(2 αx)] times the transverse factors.
Two choices the two programs make differently are offered as dropdowns. (1) BIOSCREEN uses the single term of fx above. BIOCHLOR adds the second term of the exact one-dimensional solution, exp[x (1 + s)/(2 αx)] erfc[(x + v t s)/(2 √(αx v t))] (after Martin-Hayden and Robbins 1997); with it, a very wide and thick source reproduces the one-dimensional solution of Ogata and Banks (1961) exactly. The second term matters only near the source or where x is a small multiple of αx. (2) BIOSCREEN applies λ with the retarded velocity, which amounts to decay at the same rate in the dissolved and sorbed phases; BIOCHLOR assumes decay in the dissolved phase only, which replaces λ by λ/R. The half-life is converted with ln 2 (BIOSCREEN rounds it to 0.693).
Source zones and a declining source
A source that is strongest in the middle is built, as in both programs, by superposing up to three Domenico sources of increasing width: the outer source carries the outer concentration over the whole width, the next one adds the step up to the middle concentration, and the center one the last step. The page asks for the width of the center zone and the width of zones 2 and 3 on each side.
A source may decline as C0 exp(−ks t). The half-life is entered, or follows from the soluble mass M0 in the source zone by BIOSCREEN's relation ks = Q C0 / M0 (manual Appendix A.3), where Q = vs ne × (source width × thickness) is the groundwater flow through the source and C0 the width-averaged concentration. The manual calls this an experimental relationship that probably underpredicts the removal rate. As in BIOSCREEN, the concentration at x uses the source strength at the time t − x/v, when that water left the source; where the front has not yet arrived the page takes that time as zero (the page's own rule; the manual does not say).
Instantaneous reaction
BIOSCREEN's third model assumes that biodegradation is limited by the supply of electron acceptors, not by a rate. Its biodegradation capacity is BC = Σ (concentration of electron acceptor consumed, or of by-product produced) / (utilization factor), with the factors the manual gives for BTEX: oxygen 3.14, nitrate 4.9, ferrous iron 21.8, sulfate 4.7, methane 0.78 mg per mg. The no-decay solution is run with every source concentration raised by BC, and BC is then subtracted from the result everywhere; negative values are set to zero. The manual advises caution for retardation factors above 6. Use the factors only for the compounds they were derived for; otherwise enter BC directly.
Declining source: the page does not compute the instantaneous-reaction model together with a declining source. The equation printed in the Version 1.4 revisions and the text of the manual describe that combination differently, and neither reading reproduces the instantaneous-reaction row of the manual's own Keesler AFB example (the closest is within about 0.13 mg/L of the printed values).
Dispersivity
Dispersivity is seldom measured and is the main calibration parameter in both programs. The relations offered are those printed in the BIOSCREEN manual: αx = 0.83 (log10 Lp)2.414 with the plume length Lp in meters (Xu and Eckstein 1995; with lengths in feet the page converts with 3.28 ft per meter, as the manual prints the relation, so that the manual's examples reproduce), αy = 0.10 αx and αz = 0 (the BIOSCREEN defaults); or αx = 0.1 Lp (Pickens and Grisak 1981) with αy = 0.33 αx and αz = 0.05 αx (the ASTM values quoted in the manual). The manual stresses, after Gelhar and others (1992), that reported dispersivities at a given scale range over two to three orders of magnitude.
Results
The centerline table gives the concentration at eleven distances at the entered time, at y = 0 and z = 0, for each model; the first row is the source concentration itself. The receptor table gives the concentration at the point (x, y, z) now and at steady state, the first time at which the receptor reaches the entered standard (found by scanning time and then bisecting), and the plume length: the farthest centerline distance at or above the standard, now and at steady state. The breakthrough chart shows the concentration at the receptor against time.
Units
Lengths go through the site's unit dictionary (feet or meters). Velocity, conductivity, time and concentration are plain numbers in the units chosen in their dropdowns, and a change of the unit system does not convert them. A year is 365 days, as in BIOSCREEN. The concentration dropdown changes labels only, except that the biodegradation capacity and the soluble mass are converted to the unit chosen.
What it assumes
Uniform, steady, one-directional flow in a homogeneous, isotropic aquifer; a plane source of fixed position, constant or exponentially declining; dispersion proportional to velocity with constant dispersivities and no molecular diffusion (so not for clays or very slow flow); linear, instantaneous, reversible sorption; first-order decay at one rate everywhere downgradient of the source and none inside it; no pumping, no recharge, no vertical gradients, no change of conditions along the path. The manuals list the same limits.
When it misleads
The Domenico solution is not exact. It multiplies a one-dimensional solution by transverse spreading terms in which the travel time is replaced by x/v. Srinivasan, Clement and Lee (2007) show that it is a true solution only when the longitudinal dispersivity is zero and carries a finite error otherwise. West, Kueper and Ungs (2007) compare it with an exact solution and report that its accuracy is highly variable, that for a medium sand it underpredicts centerline concentrations by as much as 80 percent in some cases, and that the error grows with dispersivity, time and the number of dimensions. Guyonnet and Neville (2004) find the differences from the rigorous solution small on the centerline when advection dominates, and large away from the centerline, especially with decay. (These statements are taken from the papers' abstracts; the papers themselves were not read for this page.) Read the off-centerline and near-source values, and any case with a large longitudinal dispersivity, with that in mind.
Parameters dominate the answer. A decay half-life fitted to one plume, a dispersivity from a scale relation and a conductivity from a few slug tests can each be wrong by a factor of several, and the plume length responds strongly to all three. In the Hill AFB example of the BIOSCREEN manual the first-order half-life had to be lowered from 2 years to 0.1 year to match the observed plume length, and the match in the middle of the plume remained poor.
Other traps. The first-order model ignores the supply of electron acceptors and can overstate decay; the no-decay model ignores decay altogether; real sources are not planes and real aquifers are layered, so the true plume may dive, split or follow a channel. The computed time to reach a standard is the arrival of a dispersed front in an idealized aquifer, not a prediction for a real well. Results are not a substitute for monitoring data, a site-specific model or a regulatory determination.
Verification
The library reproduces the printed centerline tables of the manuals' examples: Hill AFB (BIOSCREEN manual, Figure 7: no decay, first-order decay and instantaneous reaction, all thirty values to the printed digit), Keesler AFB with a source declining from 2,000 kg (Figure 3: no decay and first-order decay, within 0.001 mg/L; source concentration 13.544 mg/L) and the no-degradation row for TCE at Cape Canaveral (BIOCHLOR manual, Figure A.8, within 0.004 mg/L). The printed values are matched only with the retardation factor computed from the bulk density, Koc and foc of the input screens, not with the rounded value those screens display. An independent computation with a scientific library agrees to ten figures on sixty seeded cases, and the two-term form reproduces the one-dimensional Ogata-Banks solution.