**7. Conclusions**

The surface charge amphoteric characteristics will confer to VADS physical/ chemical properties absolutely different to constant-charge soils, where soil

*Advanced Sorption Process Applications*

prediction power [49–53]. PP-LFERs are multiple linear regression (MLR) models that employ several solute- or sorbate-specific descriptors as independent variables and their fitting coefficients are denoted as they describe system-specific, soluteindependent properties. In this sense, descriptors and their coefficients quantitatively describe the energetic contribution of different types of sorption coefficients [53, 54]. The major advantages of the PP-LFER approach are its solid mechanistic

In the last decade, different PP-LFER models for organic contaminants sorption on soil estimation have been proposed. Endo et al. [52] proposed two PP-LFER models at environmentally relevant concentrations. However, these models lack reliable PP-LFER descriptors for environmentally relevant chemicals (e.g. pesticides, pharmaceuticals and highly polar compounds, acids, bases, and ionic compounds). This deficiency also has been identified for PP-LFER models developed for high sorbate concentrations previously reported [55]. The PP-LFER models' reviews up to now mainly have been calibrated estimating log*Koc* data of classical pollutants such as PCBs and PAHs and also of organic compounds that have chemical structure comparatively simple than chemicals of current environmental concern. These are often multifunctional or complex organic chemicals like pesticides and pharmaceuticals. The first reliable PP-LFER model for soil-water partitioning was calibrated with data from 79 polar and non-polar compounds that cover a more diverse and wider range of chemical classes than other PP-LFERs published. The model of Bronner and Goss [49] was validated using the experimental data for about 50 pesticides and pharmaceuticals not involved in the calibration set. This has potential to correctly estimate the *Koc* data for multifunctional or complex organic chemicals like pesticides and pharmaceuticals. However, Sabljic and Nakagawa [53] suggest still important drawbacks to the general applicability of the developed model. In view of the scope of this section, we recommend the review made by Sabljic and Nakagawa [53] around this topic.

On the other hand, little attention has been paid to the general applicability of the calibrated PP-LFERs for predicting sorption to soils considering the diversity of soil mineralogy, variable surface charge, OC structures and their interactions [51]. The evaluation of possible applications of PP-LFERs in the study of partitioning of

**6.4 Mechanistic approach: QSAR models for sorption of ionisable pesticides**

In the last decade, different authors developed equations to predict the sorption of ionisable and non-ionisable compounds in soils or sediments [25, 58–61]. Several models have expanded their applicability domain including soil properties and ionisation effects [48, 58, 59]. Franco et al. carried out a surface acidity correction, because the two units proposed by Bintein and Devillers [59] are dependent on soil properties, related to the surface potential of the colloid [25]. These researches suggest a general non-linear equation based on *LogKow* for neutral and ionic species (a fragmentation of *LogD*) and the speciation of monovalent acids, monovalent bases and amphoteric species. Franco et al. aimed to predict pH-dependent *Kd* values of organic acids, considering speciation as a function of soil pH and species-specific partition equilibrium [60]. This modification of their previous models by replacing their constant terms pHopt by a varying pH range allowed that the modified model performs significantly better than the original model for organic acids [25]. The two molecular descriptors, pKa and *logPn*, and the two soil descriptors, OC and pH, used in the model have a major impact on the sorption of ionisable chemicals. Nevertheless, it was not successful to develop the analogous modified model for bases due to the contradictory

ionic organic chemicals is a subject of ongoing research [56, 57].

grounds and the use of uniformly measured calibration data.

**118**

effect of pH on the total sorption.

composition (i.e. SOM), mineralogy and variable charge are key components of most VADS, controlling soil sorption of INIH, representing an environmental substrate that may become polluted over time due to intensive agronomic uses. The *pseudo-second-order* model and TSNE have been the models that best describe the kinetics parameter and solute sorption mechanism, respectively, of INIH on VADS. These models are also necessary in order to develop and validate QSAR models to predict pesticide sorption on VADS to prevent potential contamination of water resources and predict environmental risks. In this regard, the last section of this chapter illustrates briefly some of the advances of QSAR models established for predicting the soils' sorption of pesticides with a focus on the mechanistic interpretation. In the generation of QSAR models, the statistical approach is the most used with a posteriori mechanistic interpretation, possibly due to complex sorption mechanisms of pesticides on soils. In the mechanistic approach (a priori mechanistic interpretation), few studies have paid attention to the diversity of soil mineralogy, texture, variable surface charge, OC structures and their implication on sorption of ionisable pesticides. Finally, the use of solute sorption mechanism models and QSAR models for pesticide sorption in soils will contribute to a better understanding of behaviour of pesticides on VADS.
