Você está na página 1de 25


Ground-Water Management Problem Used for the Exercises

Ground-Water Flow System

The flow system is comprised of two confined aquifers separated by a confining unit. Inflow occurs primarily as areal recharge; there is also a very small amount of inflow across the boundary with the hillside. At steady-state, outflow occurs only as discharge to the river.

Management Issues

Ground-water management issues related to this flow system:

Pumping wells for water supply are being completed in aquifers 1 and 2. The effect of pumping on the river is of concern, because there is a minimum required discharge from the ground-water system to the river. A landfill is proposed in one corner of the study area. The landfill developers claim

(a) the landfill is outside the capture zone of the wells and (b) any leaking effluent will reach the river sufficiently diluted.

We are developing a flow model to help evaluate these claims.

Predicted transport from landfill

Will effluent go to well or river?


Ground-Water Flow Model

First, steady-state model without pumping will be developed and calibrated using available measurements of hydraulic-heads and discharge to the river, and will be used for preliminary evaluation of the issues.

Then, pumping wells will be installed, a long-term aquifer test will be conducted using the wells, and a transient model of the system will be recalibrated using the steady-state measurements as well as the additional drawdown and river discharge data collected during the test.

See flow model setup in Figure 2-1b (page 32) of Hill and Tiedeman. See model parameter definition and starting values in Table 3-1 (page 38). See true simulated conditions in Figures 2-1c and 2-1d (page 22-23).

MODFLOW Packages Used

Basic (BAS) and the Discretization input file Define model grid, including rows, columns, and layers Define types of model layer (confined; convertible) Layer Property Flow (LPF) Define hydraulic properties K, Ss, Sy, Anisotropy Recharge (RCH) Define distribution of areal recharge River (RIV) General-Head Boundary (GHB) Well (WEL) Define pumpage Used here for prediction runs only

Spatial Discretization -- Ground-Water Flow Model Grid

Figure 2-1b of Hill and Tiedeman (page 22)

Layer-Property Flow Package

Define hydraulic properties K, Ss, Sy, Anisotropy

Figure 2-1a of Hill and Tiedeman (page 34)

Model layer 1 is homogeneous Model layer 2 has K that increases linearly from beneath the river to the hillside

Recharge Package

Can apply recharge to model layer 1 or top active cell at each row, column location Input a flow rate (L/T). The program multiplies by area to get volume per time. Layer data. Parameters are used to define the recharge rate. For this problem, use 2 zones; no multiplication arrays.

EXPLANATION qn Positive qn indicates flow into the subsurface (A) GHB Slope = -Cn = -(KnAn)/Dn qn 3/T) the simulated flow rate at one cell (L (negative for flow out of the ground -water system) the hydraulic conductivity (L/T) of, for example, the riverbed or lakebed the thickness (L) of, for example, the riverbed or lakebed the area of the water body within the finite difference cell (L2) the conductance calculated using K n, Dn, and An. is the simulated hydraulic head in the ground water system adjacent to the head -dependent boundary (L); and is the water level in the water body or the (C) elevation of the drain (L) is the bottom of the streambed qn Positive qn indicates flow into the subsurface (C) RIV Slope = -Cn = -(KnAn)/Dn

Kn Dn

The River (RIV) and General-Head Boundary (GHB) Packages are used

qn = 0


An Cn

Negative q n indicates flow out of the subsurface


Hn Hn En (B) DRN hn Slope = -Cn = -(KnAn)/Dn

Both define headdependent boundaries: q=C(H-h)

qn qn = 0

qn = 0 Negative q n indicates flow out of the subsurface



Negative q n indicates flow out of the subsurface



Well Package

List data layer, row, column, pumpage rate Input volume of discharge per time. Outflow is NEGATIVE. For this problem, no pumpage parameters need to be defined, but you can use the parameter capability to define them if you like. [Why might you want to do that?]

Parameters and Starting Values

Flow-system property Horizontal hydraulic conductivity of layer 1, in m/s Vertical hydraulic conductivity of confining bed, in m/s Horizontal hydraulic conductivity of layer 2 in columns 1 and 2, in m/s Hydraulic conductivity of the riverbed, in m/s Recharge in recharge zone 1, in cm/yr Recharge in recharge zone 2, in cm/yr

Parameter name HK_1 VK_CB HK_2 K_RB RCH_1 RCH_2

Starting value 3.0 x 10-4 1.0 x 10-7 4.0 x 10-5 1.2 x 10-3 63.072 31.536

Table 3-1 of Hill and Tiedeman (page 38)

True Simulated Conditions

Figure 2-1 c and 2d of Hill and Tiedeman (page 22-23)


10 head observations, 5 in each model layer. Observed values shown in Table 3-2 (page 38) of Hill and Tiedeman. 1 steady-state flow observation: ground-water discharge to the river of 4.4 m3/sec.

Head Observations
Well identifier Observation name
Layer Row Col

Observed head (m)

Variance of well elevation measurement error (m2)

Variance of water-level measurement error (m2) 0.0025 0.0025 0.0025 0.0025 0.0025 0.0025 0.0025 0.0025 0.0025 0.0025

Variance of the observation (m2)

1 2 3 4 5 6 7 8 9 10

1.ss 2.ss 3.ss 4.ss 5.ss 6.ss 7.ss 8.ss 9.ss 10.ss

1 1 1 1 1 2 2 2 2 2

3 4 10 13 14 4 10 10 10 18

1 4 9 4 6 4 1 9 18 6

101.804 128.117 156.678 124.893 140.961 126.537 101.112 158.135 176.374 142.020

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

1.0025 1.0025 1.0025 1.0025 1.0025 1.0025 1.0025 1.0025 1.0025 1.0025

Table 3-2 of Hill and Tiedeman (page 38)

Weighting is accomplished using UCODE_2005

Here we break for a lecture introducing UCODE_2005

Weights on Observations

Weighted residuals are squared and summed in the objective function:

S (b) i (hi h'i (b))
i 1 nh 2

i (qi q'i (b))

i 1


i ( Pi P'i (b))
i 1





Need to weight the observations for two primary reasons:

So that residuals for observations with different units (e.g. heads and flows) can be summed in the objective function.

To account for different observations having different degrees of measurement error. Weighting is used to reduce the influence of observations that are less accurate, and to increase the influence of observations that are more accurate.

Weights on Observations

Defining the weights as being proportional to the inverse of the variance of measurement error meets both these needs:


The variance of measurement error is an estimate of the uncertainty of the observation. How do we quantify this? Need to quantify the uncertainty using: 1. A range that is symmetric about the measurement used for the observation. 2. A probability with which the true value is expected to occur within the range.

Weights on Observations: Example 1

A head observation is thought to be good to within 3 meters.

Quantify this as there is a 95-percent chance the true value falls within 3 meters of the measurement.

Use a normal probability table to determine that a 95-percent confidence interval is a value plus and minus 1.96 times the standard deviation, .

Thus, 1.96 x = 3 m, so = 1.53 m, and the variance 2 = 2.34 m2.

Weights on Observations: Example 2

A measurement of stream loss to a ground-water system is derived by subtracting two streamflow measurements, an upstream value of 3.0 m3/s and a downstream value of 2.5 m3/s .

Quantify uncertainty: The first measurement is considered to be slightly worse than the second. For the upstream measurement, the hydrologist believes that there is a 90% chance that the true value falls within 5% of the measured value; for the downstream measurement, there is a 95% chance that the true value falls within 5% of the measured value. Using values from a normal probability table, a 90% confidence interval is a value 1.65 times the standard deviation, . Upstream measurement: 1.65 x = 0.05 x 3.0 m3/, so = 0.091 m3/s. Downstream measurement: 1.96 x = 0.05 x 2.5 m3/, so = 0.064 m3/s. The loss of streamflow in the reach between these two measurements is 0.5 m3/s. How accurately is this loss known? Add variances!! The variance of the loss is 0.0912 + 0.0642 = 0.0124 (m3/)2.

Weights for the Steady-State Problem

Head Observations:

Elevation of each observation well has a variance of measurement error of 1.00 m2. Each water-level measurement has a variance of measurement error of 0.0025 m2. Thus, total variance measurement error = 1.0025 m2

Flow Observation:

Coefficient of variation of 10%; thus = 0.1 x 4.4 m3/s = 0.44 m3/s

DO EXERCISE 3.2d: Calculate weights on hydraulic-head and flow observations.

Model Fit Using Starting Parameter Values

DATA AT HEAD AND FLOW LOCATIONS OBSERVATION NAME hd01.ss hd02.ss hd03.ss hd04.ss hd05.ss hd06.ss hd07.ss hd08.ss hd09.ss hd10.ss flow01.ss MEASURED VALUE 101.804 128.117 156.678 124.893 140.961 126.537 101.112 158.135 176.374 142.020 -4.40000 SIMULATED VALUE 100.225 139.331 174.363 139.331 157.132 139.632 102.868 173.956 190.300 157.041 -4.86060 WEIGHTED RESIDUAL 1.577 -11.20 -17.66 -14.42 -16.15 -13.08 -1.754 -15.80 -13.91 -15.00 1.047

From UCODE-2005 output file. DO EXERCISE 3.3: Evaluate model fit using the starting parameter values.

RESIDUAL 1.579 -11.21 -17.68 -14.44 -16.17 -13.10 -1.756 -15.82 -13.93 -15.02 0.4606

WEIGHT**.5 0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999 2.27



Least squares objective function

Example Least-Squares Objective Function

S (b) i (hi h'i (b)) 2
i 1



i (qi q 'i (b)) 2

i 1



i 1

2 ( P P ' ( b )) i i i


System information related to model inputs, use in model development

Basic Steps for Model Calibration

Italicized steps are affected by using regression

Adjust parameter values and model construction


Related to model output. Use to calibrate model


Parameter estimation

Compare simulated & measured values using objective function

Alternative models

Evaluate model fit and estimated parameter values


Consider predictions

Evaluate predictions and prediction uncertainty

Societal decisions