General purpose Freundlich isotherms for cadmium, copper and zinc in soils

Abstract

Summary Assessing the accumulation and transport of trace metals in soils and the associated toxicological risks on a national scale requires generally applicable sorption equations. Therefore Freundlich equations were derived for Cd, Zn and Cu using multiple linear regression on batch sorption data from the literature with a wide variety of soil and experimental characteristics, and metal concentrations ranging over five orders of magnitude. Equations were derived based on both total dissolved metal concentrations and free metal activities in solution. Free metal activities were calculated from total metal concentrations taking into account ionic activity, and inorganic (all metals) and organic complexation (Cu only). Cadmium and Zn were present in solution predominantly as free ions, while Cu was present as organic complexes. Since actual dissolved organic carbon (DOC) concentrations were not available they were estimated using an empirical field relation between DOC and organic matter content. The logarithmic transformation of the Freundlich constant for Cd was regressed on the logarithmic transformations of cation exchange capacity (CEC) (H+) and dissolved Ca, and for Zn with CEC and (H+). For Cu the log–log regression model of the Freundlich constant included the solid:solution ratio of the batch to account for dilution of DOC in the batch as compared with the field. The explained variance for the fitted Freundlich equations was 79% for Cd, 65% for Cu and 83% for Zn, using log‐transformed adsorbed concentrations and soil solution activities. The Freundlich adsorption models underestimated metal contents determined from 1 m HNO3 digestion on field samples, up to a factor of 6 (Cd and Cu) or 10 (Zn).

Application of neural network model for the prediction of chromium concentration in phytoremediated contaminated soils

“…This indicates that the results of ANN model are better than that of MLR models for all amendment cases. Our finding agrees well with those of van der Zee and van Riemsdijk (1987), Andersen and Christensen (1988), Streck and Richter (1997), Römkens and Salomons (1998), Tiktak et al (1998), Schaap and Lei (1998), Elzinga et al (1999), Sarmadian and Taghizadeh Mehrjardi (2008). …”

“…Generally two common methods are used to develop prediction models, regression methods and artificial neural networks (ANN). Several multiple linear regression (MLR) models have been developed over the past 20 years to predict the sorption of trace metals in soils (van der Zee and van Riemsdijk, 1987;Andersen and Christensen, 1988;Reinds et al, 1995;Streck and Richter, 1997;Römkens and Salomons, 1998;Tiktak et al, 1998;Elzinga et al, 1999;Schug et al, 2000). With MLR methods, the relationships between soil inputs (properties) and soil output characteristics have to be stated a priori in the regression models.…”

“…It has been reported that ANNs provide superior predictive performance compared to conventional mathematical methods including MLR models (van der Zee and van Riemsdijk, 1987;Andersen and Christensen, 1988;Streck and Richter, 1997;Römkens and Salomons, 1998;Tiktak et al, 1998;Schaap and Lei, 1998;Elzinga et al, 1999;Sarmadian and Taghizadeh Mehrjardi, 2008). In regression models in many soil engineering situations, the input-output relationships are highly complex and are not well understood.…”

Impact of cadmium levels in fertilisers on cadmium accumulation in soil and uptake by food crops

“…To calculate the dissolved Cd concentration in solution a non-linear model was used similar to that developed by a.o. Tiktak et al (1998), Elzinga et al (1999), and . Solution concentrations in a wide range of soils were obtained by a 1:10 (w:v) extraction using 0.002 and 0.01 M CaCl 2 solutions ) which appear to mimic in situ soil solution concentrations of Cd (de Greyse et al, 2003).…”

Application of an inverse neural network model for the identification of optimal amendment to reduce copper toxicity in phytoremediated contaminated soils

“…Care must therefore be taken concerning the selection of soil parameters. Although parameter correlations are observed and may be strong in some cases, existing soil analysis methods typically adopt the assumption of independent parameters (van der Zee and van Riemsdijk, 1987;Andersen and Christensen, 1988;Streck and Richter, 1997;Römkens and Salomons, 1998;Tiktak et al, 1998;Elzinga et al, 1999;Annadurai and Lee, 2007;Sarmadian and Taghizadeh Mehrjardi, 2008;Minasny et al, 2004;McBratney et al, 2003;Behrens et al, 2005;Buszewski and Kowalkowski, 2006;Anagu et al, 2009 ;Gandhimathi and Meenambal, 2012).…”