Investigating the potential of near-infrared spectroscopy for the detection of low-concentration CECs in water

Date
2023-03
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: The presence of so-called contaminants of emerging concern (CECs) in water is a significant issue in many countries. Their low concentrations, complex matrices, and wide range of compositions (with varying physical-chemical properties) pose many challenges to their rapid identification and quantification. Consequently, there is a need for the development of reliable, fast, and low-cost monitoring techniques that warn of any irregular changes in CEC-related water quality, which can then prompt detailed analysis by standard methods. Near-infrared (NIR) spectroscopy, in conjunction with chemometrics, is one such technique that could potentially be utilised for near-real-time monitoring and detection of selected CECs. Therefore, this study endeavoured to develop a better understanding of the potential of NIR spectroscopy for detection of changes in the water spectrum that may indicate a CEC-related deterioration in water quality. The applicability of the method was considered by performing multivariate analysis on the NIR spectral data of deionised water spiked with surrogate CECs typically associated with anthropogenic pollution. Standard deviations of less than 0.16% and less than 0.71%, during the repeatability and reproducibility tests, respectively, suggested a high precision of the NIR method. The ratio of the standard deviation over the mean were less than 0.2% for the repeatability tests, and less than 1.16% for the reproducibility tests. This allowed for the development of a CEC classification model. A pre-feasibility study was firstly conducted, considering the following five CECs at concentrations varying between 0.001 mg/L and 1000 mg/L: carbamazepine, caffeine, efavirenz, sulfamethoxazole and trimethoprim. Positive pre-feasibility results – indicating that NIR spectroscopy indeed has potential as a rapid screening method – prompted further detailed experiments with three typical surrogate CECs, viz acetaminophen, benzotriazole and caffeine. Exploratory data analysis using principal component analysis (PCA) revealed that two spectral regions, 1300–1600 nm and 1600–2200 nm, where the appropriate wavelength regions in which to perform the investigation. The PCA score plots of the surrogate CECs showed good separation between the NIR spectra of the test chemicals and the deionised water. However, for acetaminophen and benzotriazole good separation was only possible at high concentration ranges (10–1000 mg/L) – notably higher than typical concentrations at which they occur in polluted surface water. On the other hand, caffeine samples showed separation at all the tested concentrations (0.00001–1000 mg/L), and thereby suggesting the potential application of NIR spectroscopy for the detection of caffeine at typical concentrations found in the environment. Partial least squares discriminant analysis (PLS-DA) showed that it was possible to differentiate between acetaminophen and the deionised water at a minimum concentration of 10 mg/L using the PLS-DA model with a 92.6% classification accuracy. For caffeine, differentiation was possible at a minimum concentration of 0.01 mg/L, with model classification accuracies of 87%. These results corroborated the results from the exploratory analysis with PCA. Differentiation between benzotriazole and the deionised water using PLSDA was successful to a minimum concentration of 0.1 mg/L, coinciding with findings from previous studies. The results of some preliminary evaluation tests performed on river water samples were unsatisfactory. As expected, it showed that NIR measurements in water containing salts and organics (potentially also several CECs) would be influenced adversely by such dissolved components, thereby complicating potential realtime application of these methods. Further in-depth research is therefore required, where a wider variety of matrices (such as surface water, groundwater, and wastewater), are included in the experimental design. NIR spectroscopy combined with multivariate data analysis has shown promising potential for detecting contaminants of emerging concern at mg/L and even high µg/L levels. However, implementation of the method as a (near)-real time early warning system for detecting CECs at environmental concentrations (at ng/L levels) is not yet a practical possibility.
AFRIKAANS OPSOMMING: Die teenwoordigheid van sogenaamde besoedelingstowwe van opkomende kommer (CECs) in water is ʼn beduidende kwessie in verskeie lande. Hul lae konsentrasies, gekompliseerde matrikse en wye reeks samestellings (met variërende fisiese-chemiese eienskappe), bied baie uitdagings tot hul spoedige identifikasie en kwantifisering. Gevolglik bestaan daar ʼn behoefte aan die ontwikkeling van betroubare, vinnige en lae koste-moniteringstegnieke, wat waarskuwing kan bied teen enige onreëlmatige veranderinge in CEC-verwante watergehalte. Hierdeur word die gebruik van omvattende ontleding deur die standaardmetode aangespoor. Naby-infrarooi (NIR) spektroskopie, in samewerking met chemometrie, is een tegniek wat moontlik gebruik kan word vir intydse monitering en bespeuring van geselekteerde CECs. Hierdie studie het dus gepoog om ʼn beter begrip te ontwikkel van die potensiaal van NIR spektroskopie om veranderinge in die waterspektrum op te spoor wat ʼn CEC-verwante agteruitgang in watergehalte kan aandui. Die toepaslikheid van die metode is ondersoek deur meerveranderlike analise op die NIR spektrale data van gedeïoniseerde water uit te voer wat met surrogaat CECs toegedien was, wat tipies met antropogeniese besoedeling verbind word. Die standaardafwyking vir die herhaalbaarheid- en herproduseerbaarheidstoetse, was minder as 0.16% en 0.71%, onderskeidelik, en het dus voorgestel dat die NIR-metode hoë akkuraatheid toon. Die verhouding van die standaardafwyking teenoor die gemiddeld was minder as 0.2% vir die herhaalbaarheidtoets, en minder as 1.16% vir die herproduseerbaarheidstoetse. Dit het die ontwikkeling van 'n CEC-klassifikasiemodel moontlik gemaak. Daar was eerstens ʼn voor-uitvoerbaarheidstudie uitgevoer, wat die volgende vyf CECs in ag geneem het by konsentrasies wat gewissel het tussen 0.001 mg/L en 1000 mg/L: carbamazepine, kaffeïen, efavirenz, sulfamethoxazol en trimetoprim. Positiewe resultate van die voor-uitvoerbaarheidstudie – wat daarop gedui het dat NIR spektroskopie wel potensiaal het as ʼn vinnige siftingsmetode – het gelei tot verdere omvattende eksperimente met drie tipiese surrogaatCECs, naamlik: acetaminophen, benzotriazole en kafeïen. Verkennende data-analise, deur van hoofkomponentanalise (PCA) gebruik te maak, het aan die lig gebring dat twee spektrale streke, 1300-1600 nm en 1600-2200 nm, die toepaslike golflengtestreke is waarin die ondersoek uitgevoer moet word. Die PCA telling-grafieke van die surrogaat CECs het goeie skeiding tussen die NIR spektra van die toetschemikalieë en die gedeïoniseerde water getoon. Goeie skeiding vir acetaminophen en benzotriazole was egter slegs moontlik in die hoë-konsentrasiegebiede (10-1000 mg/L), wat noemenswaardig hoër is as die tipiese konsentrasies waarby hierdie chemikalieë in besoedelde oppervlakwater voorkom. Aan die ander kant het kafeïenmonsters skeiding by al die getoetste konsentrasies (0.00001-1000 mg/L) getoon, en dus dui op die potensiële toepassing van NIRspektroskopie vir die bespeuring van kafeïen by tipiese konsentrasies soos dit in die omgewing voorkom. Parsiële kleinste kwadratediskriminantanalise (PLS-DA) het getoon dat dit moontlik is om tussen acetaminophen en die gedeïoniseerde water te onderskei tot by 'n minimum konsentrasie van 10 mg/L deur gebruik te maak die PLS-DA model met ʼn klassifikasie akkuraatheid van 92.6%. Vir kafeïen was onderskeiding moontlik tot by 'n minimum konsentrasie van 0.01 mg/L, met model klassifikasie akkuraatheid van 87%. Hierdie resultate het die resultate van die verkennende analise met PCA bevestig. Onderskeiding tussen benzotriazole en die gedeïoniseerde water was suksesvol tot 'n minimum konsentrasie van 0.1 mg/L met behulp van PLS-DA, wat ooreenstem met bevindinge van vorige studies. Die resultate van ʼn paar voorlopige evalueringstoetse wat op rivierwatermonsters uitgevoer is, was onbevredigend. Soos verwag, het dit getoon dat NIR-metings in water wat soute en organiese stowwe bevat (moontlik ook verskeie CECs), nadelig beïnvloed sou word deur sulke opgeloste komponente, en sodoende potensiële intydse toepassing van hierdie metodes bemoeilik. Daarom word verdere in-diepte navorsing vereis, waar 'n groter verskeidenheid matrikse (soos oppervlakwater, grondwater en afvalwater), by die eksperimentele ontwerp ingesluit kan word. NIR-spektroskopie, gekombineer met meerveranderlike data-analise, het belowende potensiaal getoon vir die bespeuring van besoedelingstowwe van opkomende kommer by mg/L en selfs hoë µg/L-vlakke. Implementering van die metode as 'n (byna)-intydse vroeë waarskuwingstelsel vir die bespeuring
Description
Thesis (MEng)--Stellenbosch University, 2023.
Keywords
Citation