Dissolved Organic Carbon Quality and Sorption of Organic Pollutants

Jan 12, 2015 - Regional climate change scenarios predict increased temperature and precipitation in the northern Baltic Sea, leading to a greater runo...
0 downloads 5 Views 2MB Size
Article pubs.acs.org/est

Dissolved Organic Carbon Quality and Sorption of Organic Pollutants in the Baltic Sea in Light of Future Climate Change Matyas Ripszam,*,† Joanna Paczkowska,‡ Joaõ Figueira,§ Cathrin Veenaas,† and Peter Haglund† †

Department of Chemistry, Umeå University, 901 87 Umeå, Sweden Department of Ecology and Environmental Sciences, Umeå University, 901 87 Umeå, Sweden § Department of Pharmacology and Clinical Neuroscience, Umeå University, 901 87 Umeå, Sweden ‡

S Supporting Information *

ABSTRACT: Regional climate change scenarios predict increased temperature and precipitation in the northern Baltic Sea, leading to a greater runoff of fresh water and terrestrial dissolved organic carbon (DOC) within the second part of the 21st century. As a result, the current north to south gradient in temperature and salinity is likely to be shifted further toward the south. To examine if such climate change effects would cause alterations in the environmental fate of organic pollutants, spatial variations of DOC quality and sorption behavior toward organic contaminants were examined using multiple analytical methods. The results showed declining contents of aromatic functional groups in DOC along a north to south gradient. Similarly, the sorption of a diverse set of organic contaminants to DOC also showed spatial differences. The sorption behavior of these contaminants was modeled using poly parameter linear energy relationships. The resulting molecular descriptors indicated clear differences in the sorption properties of DOC sampled in northern and southern parts of the Baltic Sea, which imply that more organic contaminants are sorbed to DOC in the northern part. The extent of this sorption process determines whether individual contaminants will partition to biota via direct uptake or through sorption to DOC, which serves as food source for bacteria-based food-webs.



carbon flow toward heterotrophic bacteria production.6 Future climate change scenarios suggest that the general north to south gradient will be shifted further south.1 Dissolved organic carbon is a diverse mixture of organic compounds that originate as degradation products, e.g., humic and fulvic acids, macromolecular carbohydrates, and amino acids.7,8 Sorption of organic contaminants to DOC is an important process affecting the environmental fate of these substances.9−11 Therefore, it is crucial to determine if the quantity, quality, and sorption characteristics of DOC differ throughout the aforementioned north to south gradient. Furthermore, models need to be applied that allow the prediction of the distribution of organic compounds between water and DOC in order to assess the effects of climate change on the environmental distribution of these pollutants in the future. Several methods are available for measuring the distribution of organic pollutants between DOC and water.12−14 The carbon-content normalized DOC-water distribution constant

INTRODUCTION The Baltic Sea is one of the largest brackish water bodies in the world, with a catchment area covering the majority of centralnorthern Europe. The Baltic Sea can be subdivided into three basins. The northernmost part is the Bothnian Bay. Further south, the Bothnian Sea is situated between the Bothnian Bay and the Stockholm archipelago, whereas the largest basin, the Baltic Proper, stretches from the Stockholm archipelago all the way to Sundet, the strait that separates Denmark from Sweden. Unique features of the Baltic Sea include its north to south temperature, salinity, and ecosystem productivity gradients that are controlled by factors such as air temperature, precipitation, or freshwater inflow from the north and mixing with the North Sea in the south.1 Climate change is likely to have significant effects on the Baltic Sea environment.2−5 Regional climate change models forecast a 2−4 °C increase in temperature, a 50−80% decreased ice coverage, and up to 22% increased precipitation in the northern Baltic Sea catchment area that will lead to a decrease in sea surface salinity. As a result of higher precipitation and decreased permafrost, increased washout of terrestrial dissolved organic carbon (DOC), primarily fulvic acids, to the Baltic Sea can be expected. This may lead to brownification of water and altered light conditions, thereby suppressing phytoplankton biomass production and shifting the © 2015 American Chemical Society

Received: Revised: Accepted: Published: 1445

September 10, 2014 January 7, 2015 January 12, 2015 January 12, 2015 DOI: 10.1021/es504437s Environ. Sci. Technol. 2015, 49, 1445−1452

Article

Environmental Science & Technology

Figure 1. Sampling locations throughout the Baltic Sea.

(KDOC, L kg−1) is commonly used to quantify this partitioning process. Solid-phase microextraction (SPME) coupled to gas chromatography−mass spectrometry (GC−MS) is one of the more powerful techniques commonly used in the field for this purpose. Generally, SPME is used to representatively measure the aqueous concentrations of organic compounds in a multiphase system. Several studies have employed SPME to determine KDOC coefficients of different organic compounds.15−18 Most of these studies considered only isolated humic or fulvic acids to model the sorption process. Only a handful of reports have described in situ measurement of KDOC values in natural systems, mostly using lake water or sediment pore water of generally high DOC concentrations (>10 mg/ L).19,20 This paper presents the results of the first in situ organic pollutant−DOC sorption experiments conducted using natural (brackish) pelagic seawater. The partitioning of organic compounds between DOC and water has been modeled generally by using the octanol−water (KOW) distribution coefficient as a measure of hydrophobicity. This relationship has been proven to be applicable only within specific compound classes. In the past few years, more advanced generic approaches have been developed that use linear solvation energy relationships (LSERs) to predict sorption of organic contaminants to DOC.17,21 The general form of the Abraham’s general solubility LSER model is given in eq 1: Log KDOC = c + eE + sS + aA + bB + vV

The uppercase letters in eq 1 represent the molecular descriptors of the organic contaminants. A and B parametrizes the hydrogen bond acidity and basicity. The parameters E, S, and V stand for the polarizability, dipolarity, and molecular size, respectively. Naturally occurring DOC can be perceived as submicrometer droplets of organic solvent in water as a separate phase. Partitioning of small organic compounds between water and DOC thus can be regarded as a distribution process between water and an organic solvent. Accordingly, the lowercase solvent coefficients in eq 1 account for differences between water and DOC with respect to molecular interactions with the organic contaminants. The chemical interpretation of the individual parameters has been discussed elsewhere in detail.21,22 Molecular characterization of DOC generates useful information that can be correlated with sorption properties of different organic compounds. For more than ten years, it has been possible to detect molecular constituents of both marine and terrestrial DOC through the use of ultrahigh resolution methods, such as Fourier transform ion cyclotron resonance (FT−ICR) MS.23,24 For bulk characterization of carbon quality, techniques such as in-solution proton nuclear magnetic resonance (1H NMR) and solid state 13C NMR,25−27 infrared or excitation−emission 3D fluorescence spectroscopy28,29 and specific UV absorbance (SUVA) measurements have been used.30,31 In the present study, we examined the effects of aromaticity on the sorption of organic contaminants by using 1 H NMR and the SUVA-parameter.

(1) 1446

DOI: 10.1021/es504437s Environ. Sci. Technol. 2015, 49, 1445−1452

Article

Environmental Science & Technology

(Table 1). The concentrations of the added compound were at least 100-fold higher than the expected natural background.

The goal of this study was to investigate the spatial differences in the quality of DOC and differences in sorption behavior (KDOC) throughout the open Baltic Sea along a north to south gradient. The focus was on semipolar organic compounds (log KOW = 2.5 to 6.4), for which partitioning to DOC is of high importance. The KDOC values were determined in situ using water samples collected in field to avoid losses or alterations of DOC that frequently occur when applying isolation or concentration techniques.

Table 1. List of Investigated Chemicals regulated chemicalsa POP convention pentachlorobenzene HCB α-HCHb β-HCHb γ-HCHb δ-HCHb endosulfan Ib endosulfan IIb mitotane water framework directivea biphenyl naphtalene fluoreneb phenanthrene anthracene hexachlorobutadiene trans-chlorfenvinfos alachlor atrazine trifluralin octylphenol



MATERIALS AND METHODS Sampling and Total DOC Determinations. Samples were collected throughout the Baltic Sea along a north−south gradient from the 20th to the 23rd of September, 2013. The water samples were obtained from a sampling/online measurement ferry box station set up by the Swedish Meteorological and Hydrological Institute (SMHI) housed on a cargo ship. The sampling depth was approximately 2 m, and the samples were collected in 2 L Pyrex glass bottles (VWR International, Stockholm, Sweden). Figure 1 illustrates the sampling locations. The dead volume of the sampling equipment was approximately 20 L. Hence, sampling was carried out after discharging the first 30−35 L of water. The water was transferred into a 20 L metal canister (Millipore) that was rinsed three times with approximately 1−2 L of the sample before filling. The canister was connected to compressed air and the water was filtered through 150 mm diameter Whatman 0.7 μm GF/F filters (VWR International). Pyrex borosilicate bottles (2 L) were rinsed three times with filtered seawater before being filled with the sample, after which the bottles were wrapped in aluminum foil and stored at 4 °C. The filters and glassware that came in contact with seawater were precombusted at 400 °C for 12 h. Upon arrival to port, the samples were rapidly transported to the laboratory (approximately 45 min), where they were conserved using a 0.05% NaN3 solution to avoid bacterial transformation of the dissolved organic carbon. The samples were stored for a maximum of 2 weeks at 4 °C before further processing. Collaborators at Pelagia AB (Umeå, Sweden) performed determinations of DOC concentration. The samples (100 mL) were analyzed on a TOC 5000 system (Shimadzu TOCVCPH). Before analysis, all samples were filtered through 0.45 μm filters (Minisert, Hydrophilic) and thereafter acidified by adding 1 mL of 2 M HCL. The detection limit was 0.2 mg L−1 and the relative standard deviation was 4.5%. One field blank and one laboratory blank were processed in parallel to the samples. Determination of KDOC Values. For the KDOC determinations, 16 mL of the water samples were first filtered into 20 mL SPME vials through 0.45 μm membrane disc filters (Millipore) using a glass syringe (precombusted glassware). The samples were subsequently spiked with a set of test compounds (Dr. Ehrensdorfer, Teddington, U.K.) at concentration levels ranging from 1.6−41 μg L−1 and then left to equilibrate for 12 h in a water bath (set to the surface water temperature measured at each respective sampling point). A basic set of test compounds was selected from the E.U. water framework directive and the Stockholm POPs convention lists.32 In addition to those legacy pollutants a number of emerging contaminants (current use pesticides, organophosphorous triesters, etc.) were added in order to expand the coverage of functional groups and physicochemical properties

other water contaminants current use pesticides chlorothalonil chlorothal dimethyl chlorpyrifos methyl pendimethalin picoxystrobin diflufenican diazinon

miscellaneous tributyl phosphate tris(1,3-diisopropyl)phosphate triphenyl phosphate dibutyl-phthalate 4-bromophenol 4-bromoaniline 2,4-dibromoaniline 2,4,6-tribromoaniline

a

Water framework directive chemicals that are also on the POP convention list excluded. bLow sorption to DOC. Compound outside the SPME methods applicability domain.

Pollutant analyses were performed using online liquid immersion solid-phase microextraction - gas chromatography−mass spectrometry (SPME−GC−MS). SPME was performed in the kinetic (nonequilibrium) mode. A detailed description of both log KDOC analytical methods and calculations can be found elsewhere.33 A brief description follows below. An MPS-XL autosampler purchased from Gerstel (Gerstel GmbH & Co.KG, Mülheim an der Ruhr) was used for SPME, utilizing 20 mL vials and 7 μm thick fused silicapolydimethylsiloxane (PDMS) and 65 μm thick polystyrenedivinylbenzene-PDMS SPME fibers (Supelco, Sigma-Aldrich, Stockholm, Sweden). GC−MS analysis was carried out on an Agilent (Agilent Technologies Sweden AB, Kista) 7890N GC system coupled to a LECO HRT time-of-flight MS. A list about the monitored ions and some molecular properties of the organic compounds is provided in Supporting Information (SI) Table S1. For the determination of total aqueous concentrations, reference samples not containing any DOC were prepared using artificial seawater34 matching the salinity, NaN3 content, and temperature of the seawater samples. Calibration samples were prepared using artificial seawater in the same manner as the reference samples for 4 different spiking concentrations of contaminants. After equilibration, the free concentrations of organic pollutants were determined in the calibration, reference and seawater samples using SPME−GC−MS. Using these SPME measurements, the difference between the aqueous concentrations of the reference and the seawater samples were assessed and the partitioning constants of the 1447

DOI: 10.1021/es504437s Environ. Sci. Technol. 2015, 49, 1445−1452

Article

Environmental Science & Technology

are chemically stable. The samples for the SUVA measurement were filtered through Whatman 0.2 μm polycarbonate membrane filters (GE Healthcare, Uppsala, Sweden) aboard the ship and subsequently stored in amber glass bottles in the dark at 4 °C until analysis. Ultraviolet−visible spectra of DOC were recorded using an Evolution 600 spectrophotometer (Thermo Fischer Scientific, Hägersten, Sweden) from 250 to 800 nm with 2 nm step resolution. Milli-Q water was used as reference. Absorbance values measured at 254 nm were normalized to the DOC concentration in the sample. Calculation of Abrahams General Solubility Model Parameters. Abraham solubility parameters for the spiked organic contaminants were obtained using the Absolv function within the ACDLabs PhysChem software package (Advanced Chemistry Development Inc., Toronto, Canada). Experimental data for many spiked compounds was present in the Absolv database. For compounds not contained in the database, the parameters were calculated from the model embedded in Absolv (SI Table S1). The acquired parameters were imported to SIMCA software (Umetrics; Umeå, Sweden) alongside log KDOC coefficients for each sampling point and were used to create multiparameter linear regression models.

pollutants between DOC and water calculated as follows (eq 2): KDOC =

C DOC ‐ bound Cwater

(2)

In this equation, KDOC stands for the DOC−water distribution coefficient (L kg−1), CDOC‑bound and Cwater (ng μg−1 and ng mL−1) are the sorbed and freely dissolved aqueous concentrations of organic contaminants, respectively that were calculated using calibration curves. The limit of detection of KDOC determination was calculated using the cumulative standard deviation (CSD) of the aqueous concentrations. CSD levels were calculated by error propagation from replicates (n = 4) of both the reference and seawater samples. If the concentration drop was less than two times the CSD, then the measured log KDOC was declared as insignificant. Characterization of Dissolved Organic Matter (DOM). DOM for 1H NMR spectroscopy was isolated from seawater that had been sequentially filtered through 0.7 and 0.45 μm filters. After filtration, the samples were acidified to pH = 2 with 2 M HCl and concentrated using BondElut PPL (1g and 6 mL) SPE cartridges (Aglient, Santa Clara, CA, U.S.A.). The samples were acidified in order to improve extraction of phenolic compounds and organic acids.35 Precipitation of humic substances can occur at low pH, but we did not observe any precipitate in the filtrates; which is in accordance with observations made by Dittmar et al.35 This may be explained by the low DOC concentration in the Baltic Sea (ca. 5 mg/L). The SPE columns were preconditioned with 5 mL of methanol and then equilibrated with 5 mL acidified Milli-Q water. Subsequently, 2 L of seawater sample was gravity fed through the SPE cartridges. The columns were afterward rinsed with 5 mL of acidified (pH = 2) Milli-Q water, dried under a nitrogen stream at room temperature and eluted using two times 5 mL methanol into precombusted amber vials. The samples were evaporated to dryness and redissolved in 500 μL D2O with pH set to 12 using sodium hydroxide. This procedure was repeated three times in order to remove residual water. After the third cycle, the samples were transferred to 5 mm inner diameter NMR tubes (178 cm from Emperor), spiked with 20 μL of a 100 mM TMSP solution for spectrum referencing and then kept in darkness at 4 °C until analysis. All the proton NMR spectra were collected with a Bruker Avance III NMR spectrometer operating at 400.13 MHz (B0 = 9.4 T) and coupled with the Topsin 3.2 software package. The NMR instrument was run at 297.9 K using a 5 mm z-gradient PABBO BB/19F probe. Spectra were referenced to deuterated trimethylsilyl propanoic acid (TMSP) at 0 ppm (Cambridge Isotope Laboratories, Inc.; Tewksbury, MA, U.S.A.). Solvent suppression was performed via excitation sculpting (zgesgp) with a 4.1 s acquisition time, 1 s relaxation delay (d1) and 64 transients collected for all samples. SUVA measurements were performed using previously collected samples (23rd to 25th of August, 2011) due to insufficient sample amounts. No significant bias in the data is expected because the samples were collected following the same procedure at the same sampling locations. Furthermore, bias is not expected because of the large input of terrestrial riverine DOC (10 times the atmospheric and point source inputs, respectively), and the residence time of the Baltic Sea DOC is long (3−5 years)36 and the aromatic moieties of DOC



RESULTS AND DISCUSSION DOC Concentrations and Characteristics. The DOC analysis of the samples yielded similar results throughout the Baltic Sea, with levels ranging from 4.3 to 6.6 mg L−1 (SI Table S2). A two tailed t test was performed for the mean values in each basin assuming unequal variances. The test yielded pvalues above 0.25 in all comparisons (n = 3, 4, and 5 depending on the number of sampling points). This indicates no significant differences in the concentrations of DOC (CDOC) in the open sea among the main basins (two-tailed t test, p < 0.05). A higher CDOC value (6.6 mg L−1) was measured at the sampling point nearby the mouth of river Kemijoki (Sample 1), the northernmost sampling site, even though flocculation of DOC has been shown to happen at very low salinities.37−39 In contrast, lower DOC concentration (3.9 mg L−1) was measured in the Kattegat area, outside Gothenburg, which can be explained mainly by dilution with Atlantic water. The measured concentration levels are summarized in SI Table S2, along with other parameters acquired from the online monitoring system of the SMHI sampling station aboard the ship. The measured 1H NMR spectra are presented in SI Figure S1. A continuous decrease of the signal in the aromatic region (6.5−9 ppm) was apparent when comparing samples collected in a north to south gradient. To further investigate the increased percentage of aromatic groups in DOC, the ratio of aromatic carbon (6.5−9 ppm) to aliphatic carbon (0.5−4.5 ppm) peak areas was examined along this gradient using the 1D NMR processor function of ACDLabs. The peak areas in the aliphatic region were normalized to 100 and then the peak areas of the aromatic region were calculated in relation to the aliphatic peak areas. The obtained area ratios were further normalized to the measured DOC concentrations. Figure 2 illustrates the normalized DOC concentration values plotted against the sampling locations from north to south. Higher values in the north indicate more aromatic characteristics of the Baltic Sea DOC than in the south. This reflects the greater contributions of allochthonous carbon to DOC in the north and more of autochthonous carbon in the south.36 The lower ratio (0.95) measured at the river mouth (sample 1) could be because of the presence of a more bioavailable, or less 1448

DOI: 10.1021/es504437s Environ. Sci. Technol. 2015, 49, 1445−1452

Article

Environmental Science & Technology

There were significant spatial differences in the KDOC values. These are presented in Figure 4 and will be discussed in the

Figure 2. Ratio of aromatic to aliphatic carbon integrated peak areas obtained from 1H NMR signals normalized to the DOC concentration and plotted against the sampling location.

Figure 4. Spatial differences in log KDOC coefficients of tributyl phosphate (⧫), hexachlorobenzene (−), phenanthrene (■), and 2,4dibromoaniline (▲).

water-soluble, aliphatic fraction of DOC of terrestrial origin. Thus, the average half-life of this DOC fraction is likely to be shorter, due to biodegradation or precipitation, as compared to DOC in the open sea. The differences in aromatic character among the various samples were further investigated using SUVA. The absorbance at 254 nm was normalized to the DOC-concentration and plotted against the sampling points from north to south (Figure 3). The absorbance data correlated well with the 1H NMR data

order from north to south. In general, a declining trend was observed for chlorinated aromatic compounds, such as hexachlorobenzene but also chlorthal dimethyl and mitotane (2,4′-DDD). For polycyclic aromatic hydrocarbons (PAHs) including phenanthrene, a basin dependency was observed. The partitioning constants showed an initial increase toward the south and then a sharp decrease upon reaching the Baltic Proper. Most alkylated organophosphates, phosphoester- and phosphothioester-like compounds, some of which are used as pesticides, showed no clear spatial differences (e.g., tributyl phosphate, Figure 4), with the exception of triphenyl phosphate, which behaved in a similar manner to the aromatic pesticides, presumably because of its planar phenyl substituents. Bromoanilines were used as markers for brominated aromatic compounds. 2,4- dibromoaniline showed the same declining trend as for the other halogenated aromatic compounds, but 2,4,6-tribromoaniline showed no such tendency. This could be because of its bigger molecular size and the hindering effect of the relatively large bromine atoms on the aromatic ring. Proton transfer reactions of the bromoanilines are not expected to take place at the pH range of the Baltic Sea samples (7.8−8.1) based on the calculated pKa values. The difference in carbon quality, i.e., in aromaticity (Figures 2 and 3), may explain the decline in log KDOC values observed for planar chlorinated pesticides such as penta- and hexachlorobenzene or chlorthal dimethyl along the same gradient. Interestingly, the magnitude of log KDOC of PAHs increased slightly from the Bothnian Bay to the Bothnian Sea before dramatically decreasing in the Baltic Proper. This effect is difficult to explain as it does not seem to affect the hydrophobic planar chlorinated pesticides. It would therefore be interesting to study other PAHs to better understand this phenomenon, e.g., the 16 PAHs prioritized by the US-EPA, but also to use linear solvation energy relationships (LSERs) to further characterize the intermolecular interactions. Linear Solvation Energy Relationships. The Abraham general solubility parameters of the contaminants were used to create LSER models and describe the sorption properties of DOC from different parts of the Baltic Sea. After the experimental determination of log KDOC values, the solvent properties were calculated as molecular descriptors of the bulk DOC at each location. On the basis of the results, three models were developed, one for each basin of the Baltic Sea. Good models were obtained for the northernmost and southernmost

Figure 3. Specific UV-absorption values plotted against the sampling location.

(SI Figure S2, R2 > 0.9), although there were smaller differences in the UV absorption than in the NMR aromatic:aliphatic ratio between the Bothnian Bay and Bothnian Sea. In summary, there appeared to be a clear decline in the aromatic nature of dissolved organic carbon from north to south. DOC-Water Distribution Constants (KDOC) of Organic Pollutants. The KDOC values were determined for 36 organic model contaminants, including both legacy and emerging contaminants, representing a wide range of functional groups and properties. The log KDOC values for all these compounds are presented in the SI Table S3. The sorption of HCHs, endosulfans, and fluorene to seawater DOC was too low to produce accurate data,33 and therefore these compounds were excluded from the analysis. The data from the west-coast of Sweden (Sample 15) was also omitted as it was not possible to judge its representativity (only one sample collected in the area). Error values based on the reproducibility of the method are presented elsewhere.33 1449

DOI: 10.1021/es504437s Environ. Sci. Technol. 2015, 49, 1445−1452

Article

Environmental Science & Technology

Figure 5. Linear regression (A and B) and solvent parameters determined from LSER models (C and D) for the Bothnian Bay (A and C) and Baltic Proper (B and D).

presence of a large number hydroxyl groups capable of both donating and accepting hydrogen bonds in Baltic Proper DOC may explain these trends. This may indicate a higher DOC residence time, and thus higher susceptibility to photooxidation, in the Baltic Proper. The coefficient associated with dipole−dipole or dipole− induced dipole (Keesom) interactions (s) decreases from north to south. This decline may be associated with the previously discussed increase in hydrogen bond donating ability (b). There seemed to be a greater number of hydroxyl groups in the DOC from southern waters, which could promote the formation of hydrogen bonds. The excess molar refraction (e) of different DOC phases increased from north to south, and had a greater weight in the Bothnian Bay model as compared to the Baltic Proper model (Figure 5C,D). This parameter describes the polarizability contributions from n- and π-electrons and generally correlates with the molecular size of the solute.22 However, the aromaticity is probably more important for explaining the large differences between basins. This property correlates with the π-electron polarizability and was shown by NMR (Figure 2) and UV spectroscopy (Figure 3) to differ between the north and south. Thus, DOC in the north is probably more likely to participate in charge transfer interactions and form electrondonor−acceptor complexes. This may explain the increased sorption of PAHs and chlorobenzenes (Figure 4) by DOC from the northernmost basin, i.e., the Bothnian Bay. The difference in the free energy of cavity formation (v) was not statistically significant, which means that the energy required to form a cavity of the same size in the DOC from the north and south was the same. This parameter indicates that the strength of the intermolecular interactions among the units making up the DOC phase are very similar in the Bothnian Bay and the Baltic Proper.

basin, i.e., for the Bothnian Bay and Baltic Proper, which showed good fit with the empirical data (R2Y values of 0.76 and 0.82), good internal cross-validation predictability (Q2 values of 0.71 and 0.72),40 and root-mean-square values of 0.26 and 0.25, respectively. In the cross-validation, the data are divided into 7 parts and 1/7th of the data is removed. A model is built on the remaining 6/7th of data and the data left out are predicted from the new model. The process is repeated 6 times until all data have been evaluated. R2Y = 1 indicates perfect description of the data by the model, whereas Q2 = 1 indicates perfect predictability. Models with Q2 values