You are on page 1of 38

IsoFit Documentation and User’s Guide Version 1.

2
L. Shawn Matott
State University of New York at Buffalo
Department of Civil, Structural and Environmental Engineering
Version 1.2

Contents
1 Introduction
1.1 BET Isotherm . . . . . . . . . . . . . . . .
1.2 Freundlich Isotherm . . . . . . . . . . . . .
1.3 Freundlich Isotherm w/ Linear Partitioning
1.4 Generalized Langmuir-Freundlich Isotherm
1.5 Langmuir Isotherm . . . . . . . . . . . . . .
1.6 Langmuir Isotherm with Linear Paritioning
1.7 Linear Isotherm . . . . . . . . . . . . . . . .
1.8 Polanyi Isotherm . . . . . . . . . . . . . . .
1.9 Polanyi Isotherm with Linear Partitioning .
1.10 Toth Isotherm . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

3
5
6
7
8
9
10
11
12
13
14

2 Input File
2.1 Entering Comments Into the Input File .
2.2 Specifying the Isotherm Type . . . . . . .
2.3 Specifying the Weighting Scheme . . . . .
2.4 Entering the Lab or Field Measured Data
2.5 Adjusting Isotherm Parameter Treatment
2.6 Adjusting the IsoFit Search Procedure . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

15
15
16
17
18
18
23

.
.
.
.
.
.

3 IsoFit Output Files

24

4 Running IsoFit

27

5 Example Problem

28

i

Glossary of Terms
Terms used in the various equations of this manual are defined as follows:
• m : the total number of laboratory measurements.
• Ci : the ith aqueous concentration [M/L3 ] (i = 1...m).
• qi : the ith estimated sorbed mass fraction, estimated via an isotherm
expression.
• X : A vector of isotherm parameters.
• f (X, Ci ) : an expression for qi that depends on the corresponding concentration (Ci ), the selected isotherm and it’s associated parameters
(X).
• Φ : the weighted sum of squared error (WSSE) objective function to
be minimized by IsoFit
• wi : the weight given to observation i
• qi,obs : the laboratory measured sorbed mass fraction
• p : the number of isotherm parameters, which depends on the type of
isotherm
• Xj : the value of the j th isotherm parameter (j = 1...p)
• XjL : the lower bound of Xj
• XjH : the upper bound of Xj
• Q0 : the maximum sorption density [M/M ]
• b : the affinity of adsorbent to the adsorbate [L3 /M ]

1

• Sw : the aqueous solubility of the sorptive material.e. • (1/nf ) : the Freundlich exponent [no units]. Freundlich w/ Partitioning. (1/n) • Kf : the Freundlich isotherm parameter [ (M/M )/(M/L3 ) ] • Kp : a linear partitioning parameter [L3 /M ] used by the linear isotherm and several dual-mode isotherms (i. Langmuir w/ Partitioning. • A : a lumped Polanyi isotherm parameter • B : a lumped Polanyi isotherm parameter • nt : the Toth exponent [no units] 2 . and Polanyi w/ Partitioning). • (1/ng ) : the Generalized Langmuir-Freundlich exponent [no units].

p Equation 1. (6) Langmuir with Linear Paritioning. which can be described as: Minimize Φ = W SSE = m X  2 wi (qi − qi. IsoFit implements the following isotherm expressions (1) Brunauer-EmmettTeller (BET).obs ) i=1 (1. (5) Langmuir. Ci ) .1) Where the form of f (X. The general form of the isotherm expression is qi = f (X.Chapter 1 Introduction IsoFit is a program that fits isotherm parameters to a given set of laboratory data by minimizing the weighted sum of squared error between laboratory measured data and corresponding isotherm computed values. or (10) Toth. (7) Linear. (3) Freundlich with Linear Partitioning. To solve the optimization problem.2) Subject to XjL ≤ Xj ≤ XjH . IsoFit utilizes a 3 . IsoFit treats the isotherm fitting process as a minimization optimization problem. (9) Polanyi with Linear Partitioning. (8) Polanyi. At present. j = 1. 1985) and FIT (Kinniburgh. Ci ) depends on the type of isotherm. i = 1. For isotherm expression(s) not currently supported by IsoFit. (4) Generalized Langmuir-Freundlich.2 calls for the adjustment of uncertain isotherm parameters so that the difference between model-predicted and field-observed sorbed mass fractions is minimized.. (2) Freundlich. 1999)..m (1. two software packages developed by the British Geological Survey may prove useful: ISOTHERM (Kinniburgh.

these studies compare numerous isotherms in the context of several environmental systems. and (iii) demonstrates IsoFit by applying it to an example problem. Kinniburgh (1986). (2004)]. (ii) describes the syntax for the IsoFit input file. Kennedy and Eberhart (1995)] with a nonlinear regression procedure [Levenberg-Marquardt. Katare et al. Allen-King et al. and Rabideau at al. Padmesh et al. (in review). BarteltHunt et al. (2005). 4 . The remainder of this manual (i) briefly describes each of the isotherm expressions supported by IsoFit. (2003). (2002).hybrid optimization search procedure that couples a global search heuristic [particle swarm optimization. Additional isotherm details may be found throughout the literature and include the studies of Allen et al.g. IsoFit first uses particle swarm optimization (PSO) to identify promising solutions while circumventing local minima. this is followed by a regression-based ”polishing” step that refines the PSO parameter estimates and generates regression statistics [e. Levenberg (1944) and Marquardt (1963)]. (2006).

Figure 1.1 contains a representative BET isotherm fit.1 BET Isotherm The BET isotherm expression (Brunauer et al. 1938) was originally developed to describe the adsorption of gases. Ci ) = Q bC  0 i  (Sw − Ci ) 1 + (b − 1)(Ci /Sw ) (1.. Figure 1. b]T . but has since been extended to aqueous-phase sorption.1: Example BET Isotherm Fit 5 .1. The BET isotherm expression is: f (X.3) Where the isotherm parameters are X = [Q0 .

Figure 1. 1924) is (1/nf ) f (X.2 illustrates the shape of a typical Freundlich isotherm fit. Ci ) = Kf Ci (1.2 Freundlich Isotherm The Freundlich isotherm expression (Freundlich.4) Where the isotherm parameters are X = [Kf . Figure 1. (1/nf )]T .1.2: Example Freundlich Isotherm Fit 6 .

Figure 1.3: Example Freundlich w/ Linear Partitioning Isotherm Fit 7 .3 Freundlich Isotherm w/ Linear Partitioning The Freundlich isotherm with linear partitioning (Weber at al.1. Figure 1.5) Where the isotherm parameters are X = [Kf . 1992) is expressed as a dual-mode formulation that incorporates both Freundlich and Linear isotherm behavior: (1/nf ) f (X. (1/nf ).3 illustrates the shape of a typical Freundlich w/ Linear Partitioning isotherm fit.. Ci ) = Kf Ci + Kp Ci (1. Kp ]T .

2000) is expressed as a combination of various components of the Langmuir and Freundlich isotherms: (1/ng ) Q0 (bCi ) f (X. Ci ) = (1. (1/ng )]T . Figure 1.4: Example Generalized Langmuir-Freundlich Isotherm Fit 8 .6) (1/n ) 1 + (bCi ) g Where the isotherm parameters are X = [Q0 . b.4 Generalized Langmuir-Freundlich Isotherm The Generalized Langmuir-Freundlich isotherm (Kano et al.1..4 illustrates the shape of a typical Generalized Langmuir-Freundlich isotherm fit. Figure 1.

1. 1916) is written as f (X.5: Example Langmuir Isotherm Fit 9 . Figure 1. Figure 1.7) Where the isotherm parameters are x = [Q0 .5 contains a representative Langmuir isotherm fit. Ci ) = Q0 bCi 1 + bCi (1. b]T .5 Langmuir Isotherm The Langmuir isotherm (Langmuir.

Ci ) = Q0 bCi + Kp Ci 1 + bCi (1..1. 1996) is expressed as a dual-mode formulation that incorporates both Langmuir and Linear isotherm behavior: f (X. Figure 1. Figure 1. b.8) Where the isotherm parameters are X = [Q0 .6 illustrates the shape of a typical Langmuir w/ Linear Partitioning isotherm fit. Kp ]T .6 Langmuir Isotherm with Linear Paritioning The Langmuir isotherm with linear partitioning (Xing et al.6: Example Langmuir w/ Linear Partitioning Isotherm Fit 10 .

Figure 1. Ci ) = Kp Ci (1.7: Example Linear Isotherm Fit 11 .7 Linear Isotherm The formula for the linear isotherm is f (X.7 contains an example linear isotherm fit.1.9) Where the isotherm parameter is X = [Kp ]. Figure 1.

may be expressed as:  −A log Sw /Ci B f (X.g.8 Polanyi Isotherm The Polanyi isotherm. Ci ) = Q0 10 (1.8 illustrates the shape of a typical Polanyi isotherm fit. A. Figure 1. B]T . Xia and Ball (1999 and 2000)]. Figure 1.8: Example Polanyi Isotherm Fit 12 .10) Where the isotherm parameters are X = [Q0 . as applied to solid-water interactions [e.1.

9 illustrates the shape of a typical Polanyi w/ Linear Partitioning isotherm fit. Ci ) = Q0 10 + Kp Ci (1. Figure 1.9 Polanyi Isotherm with Linear Partitioning The Polanyi isotherm with linear partitioning (Xia and Ball. 1999 and 2000) is a dual-mode expression that incorporates both Polanyi and Linear isotherm behavior:  B −A log Sw /Ci f (X. B.11) Where the isotherm parameters are X = [Q0 . Kp ]T . Figure 1.1. A.9: Example Polanyi w/ Linear Partitioning Isotherm Fit 13 .

12) Where the isotherm parameters are X = [Q0 . b. Ci ) =   n (1/nt ) 1 + (bCi ) t (1.1. Figure 1. 1971) is expressed as Q0 bCi f (X. nt ]T . Figure 1.10 illustrates the shape of a typical Toth isotherm fit.10 Toth Isotherm The Toth isotherm (Toth.10: Example Toth Isotherm Fit 14 .

• <variable> is a user-configurable IsoFit variable • [TAB] is the tab character • <value> is the user-specified value of the variable • [CR] is the carriage return-line feed character sequence Using this general syntax.1 Entering Comments Into the Input File Users may include helpful comments in the IsoFit input file by using the # character. in the following line the text ’this is an embedded comment’ would be ignored: IsothermType FreundlichIsotherm # this is an embedded comment 15 . as described in the following sections.Chapter 2 Input File The input file must be named IsoFitIn. the user specifies the isotherm type. and laboratory data. weighting scheme.txt and is formatted such that each line of user-supplied information uses syntax of the form: <variable>[TAB]<value>[CR] Where. designated as the comment character. any text that follows a comment character is ignored. 2. For example. Additional optional input file syntax can be used to fine-tune the behavior of the IsoFit search procedure and its treatment of isotherm parameters. When IsoFit parses the input file.

and Polanyi-Partition). the user must also specify the aqueous solubility of the sorptive material by including the following line in the input file: Solubility <val> 16 . Polanyi.Similarly. this makes it easy to determine which output files correspond to a given isotherm. BET. one at a time) • BET Isotherm • FreundlichIsotherm • Freundlich-PartitionIsotherm • LinearIsotherm • LangmuirIsotherm • Langmuir-FreundlichIsotherm (the generalized Langmuir-Freundlich isotherm) • Langmuir-PartitionIsotherm • PolanyiIsotherm • Polanyi-PartitionIsotherm • TothIsotherm NOTE : If the user selects the AllIsotherms options. If required by the chosen isotherm expression (i. IsoFit will loop through each isotherm. one at a time.2 Specifying the Isotherm Type The user selects the isotherm type by setting the IsothermType variable to one of the following self-explanatory values: • AllIsotherms (fits all isotherms. After completing each fit. the name of the corresponding isotherm will be appended to any output files that are generated. the entire line of text below would be ignored (since the comment character is the first character to appear in the line): # this entire line is treated as a comment 2.e.

2. wi = <val>. IsoFit calculates each corresponding weight as the inverse of the standard deviation (1/sdi ) using the following formulas: For the SorbedRelative scheme: wi = 1 1.96 = sdi <val> × Ci × <conv> 17 (2.3 Specifying the Weighting Scheme The user selects a weighting scheme by including a line that uses one of the following forms: • WeightingScheme[TAB]Uniform[TAB]<val>[CR] • WeightingScheme[TAB]SorbedRelative[TAB]<val>[CR] • WeightingScheme[TAB]AqueousRelative[TAB]<val>[TAB]<conv>[CR] • WeightingScheme[TAB]IndividualStdDevs Where the different weighting schemes are: • Uniform : uniform observation weights (i.Where <val> is the aqueous solubility. • AqueousRelative : the ith observation weight (wi ) is assigned relative to the corresponding aqueous concentration (Ci ).96 = sdi <val> × qi.. for i = 1.1) For the AqueousRelative scheme: wi = 1 1. By relating this relative error to a 95% confidence interval for each observation.e. <val> corresponds to the relative error associated with each observation.obs (2.m) • SorbedRelative : the ith observation weight (wi ) is assigned relative to the corresponding sorbed mass fraction (qi.obs ).2) . • IndividualStdDevs : observation weights are assigned using individual estimates of observation error (expressed in terms of standard deviations) For the relative weighting schemes.

BeginLabData <C1>[TAB]<q1>[TAB]<sd1> <C2>[TAB]<q2>[TAB]<sd2> ..4).<sdm> items and they may be safely omitted.EndLabData. 2. on the order of the errors in qi. NOTE : the current version of IsoFit treats measured aqueous concentrations. IsoFit won’t read the <sd1>..<Cm> are aqueous concentrations and <q1>... such as those discussed by Alper and Gelb (1990). and Rubichev (1993). Ci . If the IndividualStdDevs weighting scheme is not selected.4 Entering the Lab or Field Measured Data The user lists laboratory (or field) measured aqueous concentration and sorbed mass fraction values in a section wrapped by the following statements: BeginLabData. Barker and Diana (1974). The following aspects of a given parameter may be adjusted: 18 . as if they are error free.<qm> are corresponding sorbed mass fractions..<sdm> are estimated standard deviations of each sorbed mass fraction measurement (qi. If errors in Ci are significant (i. <Cm>[TAB]<qm>[TAB]<sdm> EndLabData Where. <C1>.. users may wish to adjust the manner in which parameters are treated by the hybrid search procedure. then one should consider using alternative regression techniques. .. .. If the IndividualStdDevs weighting scheme is selected. <sd1>.obs ). the standard deviation (sdi ) for each observation must be entered alongside each laboratory data point (see section 2.5 Adjusting Isotherm Parameter Treatment To enhance the performance of IsoFit.e. Carroll and Ruppert (2006). Orear (1982). NOTE : When the IndividualStdDevs weighting scheme is selected.where <conv> is a conversion factor that converts aqueous concentrations to an equivalent sorbed mass fraction. Kinniburgh (1986)... as shown below.. McCammon (1973). 2.obs ).

Ci ) − f Xj (1 + δXj ). log10 ) to a given parameter. <nameN> <lowN> EndRanges <high1> <high2> <trans1> <trans2> <step size1> <step size2> <highN> <transN> <step sizeN> 19 .. IsoFit will NOT employ parameter lumping. Langmuir. Langmuir-Partitioning. a characteristic that can interfere with the ”polishing” step of the IsoFit search procedure.EndRanges. To activate this feature. as shown below: BeginRanges <name1> <low1> <name2> <low2> . users have the option to lump the Q0 term with the b term.3 Ji. the user lists the new configuration of each parameter in a section that is wrapped by the following statements: BeginRanges. set to none) in preparation for the polishing step and subsequent post-regression statistics. all parameter transformations are removed (i..• Parameter Transformation : Users can direct the PSO search procedure to apply a transformation (e. . and Toth isotherms. which is computed using valuerelative finite-differences as shown in Equation 2. • Parameter Lumping: For the BET. Ci Xj × δXj  (2. users should include the following line in the input file: LumpedQ0*b yes To override the default parameter settings. forming a composite b ∗ Q0 parameter. By default. • Finite Difference (FD) Step Size : Users can adjust the step size (δXj ) used in the calculation of the Jacobian.g.3) • Lower and Upper Bounds : Users can specify lower (X L ) and upper (X H ) bounds for each parameter. .e.j = ∂qi ∆qi ≈ = ∂Xj ∆Xj f (Xj . This lumping may avoid excessive parameter correlation. Once the PSO search is complete.

Where <name> is the name of an isotherm parameter. <trans> is the transformation type (either ’log10’ or ’none’)..EndLangmuir-FreundlichRanges • For Langmuir : BeginLangmuirRanges. Valid choices for <name> depend on the isotherm specified in IsothermType and whether or not parameter lumping is employed: BET Isotherm ’b*Q0’ (if lumping).. ’b’..EndLangmuirRanges 20 .. <low> is the lower bound. ’Q0’ (if not lumping).EndBET Ranges • For Freundlich : BeginFreundlichRanges. ’(1/n)’... and/or ’b’ lumping). and/or ’(1/n)’ Langmuir-PartitionIsotherm ’b*Q0’ (if lumping). ’b’. and step size is the value-relative FD step size (δX). <high> is the upper bound.. ’b’.. and/or ’Kp’ TothIsotherm ’b*Q0’ (if lumping). ’Q0’ (if not PolanyiIsotherm ’Q0’.. ’b’. These isotherm-specific wrappers replace the normal BeginRanges.. ’a’.. ’Q0’ (if not FreundlichIsotherm ’Kf’ and/or ’(1/n)’ Freundlich-PartitionIsotherm ’Kf’. and use the following syntax: • For BET : BeginBET Ranges. and/or ’b’ Polanyi-PartitionIsotherm ’Q0’.EndRanges wrappers.. then a specific parameter-range wrapper for each isotherm must be included to override the default parameter settings. and/or ’b’ lumping). ’Q0’ (if not Langmuir-FreundlichIsotherm ’Q0’.EndFreundlichRanges • For F-P : BeginFreundlich-PartitionRanges. and/or ’n’ NOTE : if the user selects AllIsotherms for the Isotherm type. and/or ’Kp’ LinearIsotherm ’Kp’ LangmuirIsotherm ’b*Q0’ (if lumping). and/or ’Kp’ lumping).EndFreundlich-PartitionRanges • For GLF : BeginLangmuir-FreundlichRanges. ’a’.

. 21 .. L-P = Langmuir w/ Partitioning.. IsoFit will use the default isotherm parameter settings given in Table 2.EndPolanyi-PartitionRanges • For Toth : BeginTothRanges. GLF = Generalized LangmuirFreundlich..• For L-P : BeginLangmuir-PartitionRanges.EndLinearRanges • For Polanyi : BeginPolanyiRanges. and P-P = Polanyi w/ Partitioning..EndTothRanges Where F-P = Freundlich w/ Partitioning.EndLangmuir-PartitionRanges • For Linear : BeginLinearRanges..1 (located on the following page) unless the user overrides them in the parameter ranges section.EndPolanyiRanges • For P-P : BeginPolanyi-PartitionRanges.....

001 log10 none 0.00 1 × 106 1.00 0.001 1 × 10−6 log10 Lower 1 × 10−6 1. b ∗ Q0 b Kp Q0 .001 1 × 10−6 1 × 106 1 × 106 1.00 log10 0. b ∗ Q0 bVal Kp Q0 A B Kp Q0 A B Q0 .1: Default Isotherm Parameter Configurations 22 1.001 1 × 10−6 0.001 1 × 10−6 none 1 × 106 1. b ∗ Q0 b Kf (1/n) Kf Kp (1/n) Q0 b (1/n) Q0 .00 1 × 106 1.001 1 × 10−6 0.00 none log10 Upper 1 × 106 1 × 106 1.Isotherm BET Freundlich F-P GLF Langmuir L-P Linear Polanyi P-P Toth Parameter Q0 .001 1 × 10−6 none 1 × 106 log10 none log10 none Table 2.00 0. b ∗ Q0 b n Transformation FD Step Size log10 0.001 1× 10−6 0.001 1 × 10−6 1 × 106 log10 log10 0.00 10.00 10.00 .001 1 × 10−6 1 × 106 log10 0.

parameter sets) to evaluate during each generation. 23 .e. <pop> is the number of particles (i. and np is the number of isotherm parameters.2.6 Adjusting the IsoFit Search Procedure The user can adjust two parameters of the IsoFit search procedure by including the following syntax in the input file: MaxGens PopSize <gens> (default value is 20 × np) <pop> (default value is 20 × np) Where <gens> is the maximum number of generations to use in the particle swarm optimization procedure.

providing a rough indication of the goodness-of-fit of the chosen isotherm type. including: • Error variance and standard error of the regression. (iii) a run record detailing each iteration of the optimization algorithm. • 95% linear confidence intervals for each isotherm parameter. 24 . • Normality of residuals analysis. • Parameter variance-covariance matrix.2) and. measures of the sensitivity of simulated observations (qi ) to changes in isotherm parameter values. as a result. identifying observations having the greatest impact on the isotherm fit. • Observation residuals and correlation between measured and fitted observations (Ry ).Chapter 3 IsoFit Output Files IsoFit uses the OSTRICH calibration/optimization program to perform the minimization of equation (1. a test of the assumption that residuals are normally distributed. • Cook’s D and DFBETAS influential observation measures. (iv) the resulting optimal isotherm parameters and objective function value. (ii) a summary of the basic configuration variables.txt. the output is stored in a file named OstOutput0. parameter standard error and parameter correlation coefficients. • Beale and Linssen measures of model non-linearity. • Parameter sensitivities. This output file will contain the following elements (i) a GNU Public License disclaimer. and (v) a comprehensive list of regression statistics.

• A Runs Test for determining whether residuals are serially autocorrelated. On completion of IsoFit. isotherms with low MMR measures are preferred. Detailed descriptions of the elements of the OstOutput0. 1978) – Hannan and Quinn’s Criterion (HQ) (Hannan and Quinn.buffalo.txt file. As such. normal.g.groundwater.1 contains an example IsothermOut. As with the Runs Test. 1974) – Corrected Akaike Information Criterion (AICc) (Hurvich and Tsai.• Regression matrices.txt output file. users may be interested in viewing the IsothermOut.edu/software/Ostrich/OstrichMain. including the Jacobian.html In addition to the OstOutput0. 25 . and inverse-normal matrices.txt output file are provided in the OSTRICH manual. • The Durbin-Watson test for serial autocorrelation of residuals. 1992) Poeter and Anderson (2005) reviewed these measures in the context of groundwater flow modeling and preferred the AICc measure. IsoFit reports the following MMR measures: – Akaike Information Criterion (AIC) (Akaike. computed following the guidelines of White (1992) and Durbin and Watson (1971). Each MMR measure may be viewed as an estimate of the amount of information that is lost when a model (i. • Multi-Model Ranking (MMR) Statistics that compute alternative measures of model fit. MMR measures are designed to help with the identification of the ”best” isotherm for a given data set. 1994) – Bayesian Information Criterion (BIC) (Schwarz. the Durbin-Watson test is only meaningful if the lab data is entered in an appropriate order. Figure 3. this file will contain the optimal parameter values of the user-specified isotherm. For the Runs Test to be meaningful. which is available online at www. the lab data should be entered in an appropriate order (e. and are based on statistics derived from information theory. ordered by increasing value of observation) within the LabData section of the input file. along with the laboratory data that was used by the fitting procedure.txt file. an isotherm) is used to approximate reality. computed following the guidelines of Sweed and Eisenhart (1943). 1979) – Kashyap’s Information Criterion (KIC) (Kashyap.e.

120000E-001 5.Isotherm Output File IsothermType Langmuir b*Q0 2.011496E+000 8 3.898533E+000 NumObs 11 i Concentration q 0 1.096651E+000 Q0 9.821797E+000 10 5.txt Output File 26 .075377E+001 b 2.124671E+000 Figure 3.963232E+000 4 1.614204E-307 3.660000E-001 2.000000E-002 5.700000E-001 3.180000E-001 1.578017E+000 7 3.440287E+000 9 4.095293E+000 6 2.857685E-001 2 6.555648E+000 5 2.350082E-306 1 3.530000E-001 4.900000E-002 1.880000E-001 4.1: Example IsothermOut.250000E-001 4.170000E-001 3.251026E+000 3 1.

exe executable. 27 . (ii) generate output files. the progress of the search procedure is displayed in a DOS console window. this can be accomplished in a graceful fashion by creating an empty text file named OstQuit.txt. it will (i) finish up the current step of the search procedure. As IsoFit runs.exe at a command line prompt. where <path> is the path to the location of the IsoFit. IsoFit is run by typing <path>/IsoFit.Chapter 4 Running IsoFit After creating the input file. when OSTRICH detects the existence of such a file. If the user wishes to abort the IsoFit program before the search procedure completes. and (iii) quit.

01 48.78 314. wherein various materials were considered as possible amendments to conventional landfill liners.19 437.00 43.63 197. Ci (mg/L) 0.583.1 contains the lab 28 . a series of column experiments were performed to determine the sorption of trichloroethylene (TCE) onto a compacted clay liner amended with granular activated carbon (GAC).47 217.85 406.928.55 252.71 Table 5.758.80 97.34 370.Chapter 5 Example Problem The following example is derived from a research project conducted by Dr. and it was hypothesized that amendment of traditional clay liners with such materials could substantially retard diffusive flux of organic contaminants from the landfill into underlying groundwater.obs (mg/kg) 0.1: TCE sorption onto liner amended with GAC In this example. Table 5.327. The IsoFit input file used for this example is provided in the Demo directory of the IsoFit download.00 15.71 32.57 qi. The materials considered were known to have high sorptive capacities for organic compounds commonly found in landfill leachate.76 508.02 161.850.03 68. Bartelt-Hunt.194.402. Shannon L.013.35 137.

19 508.00e06 none 0. This procedure helps prevent the Levenberg-Marquardt algorithm from falling into a local minima.35 32.63 48.00 15.2 shows the OSTRICH setup summary.76 406402.02 314850. given the non-linear form of many isotherm expressions. 29 . The first run uses a Particle Swarm Optimization algorithm to place the starting point for the subsequent Levenberg-Marquardt algorithm very close to the global minimum.80 252327. Although Figure 5.03 197583.01 137758.1 contains the corresponding IsoFitIn.0 1.2 suggests that the OSTRICH program ran only two iterations of the Levenberg-Marquardt algorithm.71 43928.001 (1/n) 0.001 EndRanges Figure 5. in actuality IsoFit runs OSTRICH twice.00 BeginLabData 0.85 217.0 1.00 0.txt input file for the Freundlich isotherm type. a distinct possibility. optimal parameter set and some of the basic statistical output from the IsoFit output file (OstOutput0.78 97.1: GAC-TCE Freundlich Isotherm Input File Figure 5.47 370194.57 437013.00 none 0. run record.55 68.measurements from the column experiments and Figure 5.71 EndLabData MaxGens 50 PopSize 50 Solubility 1000 BeginRanges Kf 0. IsothermType FreundlichIsotherm WeightingScheme Uniform 1.txt).34 161.

2: Partial List of IsoFit Output 30 .Figure 5.

The AICc values indicate that the Polanyi isotherm expression provides the best fit of the sorption data. Figure 5. Figure 5.txt file.3 summarizes the resulting parameter estimates for each isotherm. along with linear confidence intervals. Note that the parameter lumping option was enabled for the BET.3: Summary of Parameter Estimates for the Example Problem By adjusting the IsothermType parameter in the IsoFitIn.4 summarizes some of the diagnostic statistics computed by IsoFit and reported in the OstOutput0.Figure 5.txt output file. 31 . IsoFit runs were performed for all ten isotherms. L-P and Toth isotherms. Langmuir.

4: Summary of Selected Diagnostics for the Example Problem 32 .Figure 5.

pg. Ruppert DR. 2002. Gelb RI. 25. 848-856. 2003. 9851016. pg. Brunauer SP. Standard Errors and Confidence Intervals in Nonlinear Regression: Comparison of Monte Carlo and Parametric Statistics. Allen SJ. vol. vol. Allen-King RM.IEEE Transactions on Automatic Control. Teller E. no. 33 . New modeling paradigms for the sorption of hydrophobic organic chemicals to heterogeneous carbonaceous matter in soils. Bartelt-Hunt SL. American Journal of Physics. 1938. 224-227. shale and two organoclays for use as sorptive amendments in clay landfill liners. Comparison of Optimized Isotherm Models for Basic Dye Adsorption bye Kudzu. Emmett PH. vol. A new look at the statistical model identification. Journal of the American Chemical Society. Burns SE. pg. Advances in Water Resources. Ball WP. Evaluation of granular activated carbon. Diana LM. Grathwohl P. 19. Measurement Error in Nonlinear Models: a modern perspective. 7. Johnson PA. pg.Bibliography Akaike H. 1974. 309-319. 1990. Adsorption of Gases in Multimolecular Layers. Gan Q. 60. Barker DR. 2005. 42. pg. Chapman and Hall: London. 94. 2006. vol. 1974. 716723. and rocks. 131. Matthews R. Smith JA. pg. vol. Journal of Physical Chemistry. sediments. Rabideau AR. Alper Js. 4747-4151. Carroll RJ. DOI:10. Journal of Geotechnical and Geoenvironmental Engineering. Simple Method for Fitting Data when Both Variables Have Uncertainties. vol.1061/(ASCE)1090-0241(2005)131:7(848). vol. pg. Bioresource Technology. 143-152. UK. 88.

Biometrika. 309-315. Eberhart RC. Kinniburgh DG. Testing for Serial Correlation in Least Squares Regression III. 58-61. Freundlich H. Kalos A. Autoregressive model selection in small samples using a bias-corrected version of AIC. 1995. vol. Series B. Kennedy J. 20. Tsai C-L. Watson GS. Journal of the Royal Statistical Society. Kluwer Academic Publishers: Dordrecht. vol. Hannan E. no. 2. IEEE Transactions on Pattern Analysis and Machine Intelligence. NJ. NY. Optimal choice of AR and MA parts in autoregressive moving average models. pg. 137-157. In: Bozdogan H. Kinniburgh DG. 131-138. editor. West D./Japan Conference on the Frontiers of Statistical Modeling: An Informational Approach. 4. 1971. The determination of the order of an autoregression. Kano F. pg. no.Durbin J.S. 1986. Report WD/ST/85/02 . Katare S. 1999. British Geological Survey: Wallingford. vol. 1985. ISOTHERM: A Computer Program for Analysing Adsorption Data. Netherlands. Abe I. pg. 1994. Piscataway. Surface Science. A hybrid swarm analyzer for efficient parameter estimation. Report WD/93/23 . 895904. 1942-1948. no. vol. vol.Environmental Science and Technology. UK. 244. Dutton and Company Publishers: New York. 2000. no. In Proceedings of the IEEE Congress on Evolutionary Computation. pg. Ueda I. 99-104. Quinn B. Particle swarm optimization. 1-19. Kamaya H. vol. 58. 41. Kashyap R. 1992. 1. UK. pp. General Purpose Adsorption Isotherms. 190-195. 1924. pg. 467. 34 . FIT User Guide. Proceedings of IEEE International Conference on Neural Networks. The Elements of Colloidal Chemistry. British Geological Survey: Keyworth. Kinniburgh DG. 2004. pg. 1979. pg. Hurvich CM. Proceedings of the First U. 1. translated by George Barger. pg. 1. Fractal Model for Adsorption on Activated Carbon Surfaces: Langmuir and Freundlich Adsorption.

69. 1916. A distributed reactivity model for sorption by soils and sediments. 37-44. 11. Hung. 1971. Journal of the American Chemical Society. Application of Two. Poeter E. 1311-1315. no. vol. pg. vol. Schwarz G. Allen-King R. Advances in Water Resources. no. Anderson D. 164-168. vol. no. A Method for the Solution of Certain Problems in Least Squares. pg. Levenberg K. American Journal of Physics. 2006. 2. in review. 2005. 597-605. Least-squares fitting with errors of measurement in both variables. Acad.and Three-Parameter Isotherm Models: Biosorption of Acid Red 88 onto Azolla microphylla. pg. no. The Annals of Mathematical Statistics. SIAM Journal on Applied Mathematics. Orear J. Multimodel ranking and inference in groundwater modeling. Weber WJ. vol. Measurement Techniques. 14. 2. 5. pg. Katz L. pg. 2221-2295. pg. vol. vol. Solids. McCammon RB. vol.. Sekaran G. no. 1993. 365-375. Quarterly of Applied Mathematics. vol. Toth J. 1. 1944. Nonlinear Regression for Dependent Variables. Tables for testing randomness of grouping in a sequence of alternatives. Rubichev NA. Vijayaraghavan K. vol. Marquardt D. Mathematical Geology. 461-464. Velan M. no.Langmuir I. Part I. 4. 6. 66-87. Rabideau AJ. 38. 912-916. 10. 11. vol. 1. 4. 431-441. Annals of Statistics. Eisenhart C. 43. pg. 1973. no. 1-2. Padmesh TVN. 1982. Acta Chem. Sweed FS. Least squares when both variables have uncertainties. An Algorithm for Least-Squares Estimation of Nonlinear Parameters. vol. McGinley PM. pg. pg. 1943. 1992. Bioremediation Journal. 50. Ground Water. Estimating the dimension of a model. pg. Matott LS. 311-317. no. 12. State Equations of the Solid-Gas Interface Layer. The Constitution and Fundamental Properties of Solids and Liquids. Modeling groundwater transport with nonlinear isotherms. 36. pg. Jiang Z. 1963. 2006. Conceptual basis and equilibrium 35 . 10.

pg. 36 . pg. 1246-1253. no.assessments. White KJ. 26. 1999. pg. 10. Environmental Science & Technology. Competitiive sorption between atrazine and other organic compounds in soils and model sorbents. The Durbin-Watson Test for Autocorrelation in Nonlinear Models. 370373. 24322440. Environmental Science & Technology. Xia G. vol. Ball WP. 1996. 2000. pg. The Review of Economics and Statistics. no. Polanyi-Based Models for the Competetive Sorption of Low-Polarity Organic Contaminants on a Natural Sorbent. 30. 34. Environmental Science & Technology. Environmental Science & Technology. vol. vol. 2. 74. pg. Xia G. Gigliotti B. Xing B. 33. 1992. 19551962. 262-269. 8. vol. Adsorption-Partitioning Uptake of Nine Low-Polarity Organic Chemicals on a Natural Sediment. Ball WP. vol. no. Pignatello JJ.