Variability in excitation–emission spectra among wastewater treatment units and applications of fluorescence-based water quality monitoring Open Access

TOC

Total organic carbon

DOC

Dissolved organic carbon

COD

Chemical oxygen demand

BOD

Biochemical oxygen demand

FWA

Fluorescent whitening agent

WWTP

Wastewater treatment plant

SBR

Sequencing batch reactor

E. coli

Escherichia coli

EEM

Excitation–emission matrix

FRI

Fluorescence regional integration

CFU

Colony forming units

DBP

Disinfection by-product

PVDF

Polyvinylidene fluoride

RFU

Relative fluorescence units

FI

Fluorescence index

HIX

Humification index

BIX

Biological index/freshness index

Lack of access to clean water, which is further harmed by the prevalence of waterborne pathogens, is a significant contributor to the burden of disease, morbidity, slowed economic growth, and poor public health in many developing nations (Balkhi et al. 2023). In order to address the water stress, strict implementation of water recycling, reuse, and zero discharge regulations through advanced monitoring systems are required. Dissolved organic matter (DOM) is a common pollutant found in water and wastewater, which primarily decides the quality of the water for the intended use. Organic contamination can be quantified either by cumulative metrics like chemical oxygen demand (COD), biochemical oxygen demand (BOD), total organic carbon (TOC), filtered dissolved organic carbon (DOC), or spectral absorption coefficient (Capodaglio et al. 2016). Aqueous DOM in wastewater consists of fulvic acids, humic acids, proteins, carbohydrates, and lipids. However, microbes utilize carbohydrates and amino acids rather than humic and fulvic acids, which are comparatively harder to degrade (Liu et al. 2019). The concentration of DOM is subject to changes undergone in the aquatic system both in terms of its source (e.g., terrestrial, anthropogenic, aquatic) and biogeochemical, physical, or chemical factors (Lee et al. 2019; Shi et al. 2021).

In wastewater treatment plants (WWTPs), the presence of DOM demands higher coagulant and disinfectant dosages, which can further lead to the formation of harmful carcinogenic disinfection by-products (DBPs) and facilitate microbial regrowth (Zhang et al. 2019). Therefore, it is crucial to identify different DOM elements and understand possible transformations in WWTPs to assess treated water quality (Shon et al. 2006). Organic matter estimation presents a common challenge for WWTPs in the global south. This is mainly because these facilities often utilize conventional laboratory methods like BOD and COD, which use chemicals, take a long time, and are prone to generating inaccurate results due to human error. Consequently, these methods are unreliable for on-site analysis. With the emergence of online analyzers for TOC data, laboratory measurements have been made relatively simpler.

The advantage of real-time online monitoring of WWTPs over traditional experimental methods is to rapidly identify the variations in wastewater characteristics, eliminating process disruptions and equipment breakdowns. Real-time monitoring can also help optimize treatment methods and save energy (Sorensen et al. 2018). A suitable predictive tool for real-time quality analysis is still a question that is being considered by WWTPs worldwide. One of the most widely used online analyzer concepts is photometric optical sensors using fluorescence. The fundamental idea of fluorescence revolves around the absorption of light, followed by excitation and emission. Fluorescence spectrometry is a rapid evaluation technique that generates contours from basic three-dimensional matrix data, also known as excitation–emission matrices (EEM). Technological improvements in light-source stability, data processing capacity, and scanning speed have made fluorescence spectroscopy an effective portable diagnostic tool with many applications. Fluorescence spectrometry is a highly selective and sensitive analytical method by which a detailed component analysis of complex organic matter pools (aqueous DOM-like) can be explored. Temperature, pH, the presence of fluorescence quenchers, and primary/secondary inner filtering effects are a few factors that can impact the fluorescence yield during analysis. However, the fluorescent matter in wastewater is influenced by significant changes in temperature and pH (Chen et al. 2003).

The current study aims to derive essential and reliable data from the fluorescence spectra of a field-scale wastewater treatment plant (WWTP) for real-time performance evaluation. The three-dimensional EEMs were utilized to characterize the DOM composition in wastewater. Different data processing methods were adopted to extract the relevant features of EEM. The study compared the variations in the fluorescence components across different treatment units in the plant and correlated the fluorescence indices (FI) with DOM. The study also investigated the changes in fluorescence components in biological and tertiary treatment units and identified the easily removable DOM in all the treatment units.

Glassware including beakers, conical flasks, filtration units, and Petri dishes were procured from Borosil and MilliPore. Filter papers of pore size 0.45 μm and 47 mm diameter were procured from Merck Sigma Aldrich. The gridded membrane filters made of mixed cellulose esters of pore size 0.45 μm and 47 mm diameter for microbiological analysis were procured from Merck Sigma Aldrich. Chromocult coliform agar used was purchased from Himedia. The deionized water used for standard preparations and dilutions was from the ELGA LabWater system.

The WWTP located at the Indian Institute of Technology Delhi residential campus was selected for this study. This fully functional treatment plant reuses the wastewater collected from hostels and in-campus residential areas for gardening and horticulture purposes. The treatment capacity of WWTP was 500 cubic meters per day (m³/d). The preliminary treatment included screening, oil, and grease removal. A sequencing batch reactor (SBR) was used for secondary treatment, which was operated for cycles of 4 h duration. The 4-h cycle included 120 min for filling, 45 min for settling, and 75 min for decanting processes. After secondary treatment, the supernatant was sent to a chlorine contact tank, followed by dual media filters and a ultraviolet (UV)-disinfection unit collectively considered for tertiary treatment. The effluent was stored prior to further supply and distribution.

The DOC content of the filtered samples was measured using a Shimadzu TOC analyzer. Potassium hydrogen phthalate was used as the standard for calibration. The non-purgeable organic carbon measurements were done for better accuracy in correlating DOM component peaks with DOC (Saadi et al. 2006). Nutrient measurements such as nitrate-nitrogen (⁠-N) and ammonium-nitrogen (⁠-N) were carried out using a Metrohm IC analyzer as per the ion chromatographic technique (section 4110-APHA).

The membrane filtration method determined Escherichia coli and total coliform counts (section 9222-APHA) in the samples. Chromocult coliform agar was used as the media, which enabled the formation of colored colonies (as CFU/mL sample) of the bacterial strains, E. coli forming blue-colored colonies and total coliforms forming pink-colored colonies when incubated at 37 °C for 24 h. A sample volume of 100 mL was taken for bacterial analysis.

A Shimadzu RF6000 Spectrofluorophotometer was used for fluorescence measurements of the wastewater sample. LabSolutions RF software enabled 3D spectrum mode, which generated EEMs. The samples were filtered using 0.45 μm PVDF (polyvinylidene fluoride) filter paper before further analysis. EEMs generated contour plots representing fluorescence intensities in relative fluorescence units (RFU) across the specified excitation–emission wavelength ranges. All the measurements were taken at room temperature.

In order to reduce the inner filter effects that interfered with fluorescence readings, the UV absorbance values of the samples were kept below 0.15 through appropriate dilutions. To further validate the accuracy of our results, deionized water fluorescence measurements were incorporated between sample analyses. The experimental design included three repetitions to address the possible experimental errors. The excitation and emission wavelength ranges chosen for this study were 220–540 nm and 290–550 nm, respectively, with a data interval of 10 nm. The wavelength ranges ensured a reduced Raman scattering effect in the readings taken. A slit width of 10 nm and scanning speed of 6,000 nm/min were selected to show distinct fluorescence contour maps. The EEM data were then exported for further analysis and deduction. EEMs of deionized water were taken as blank EEMs, which were later subtracted from the measured readings for the Rayleigh scattering effect corrections (Rinnan & Andersen 2005).

Peak picking is the primary method used in the current study for EEM data analysis. Peak picking involves identifying and quantifying specific fluorescence peaks within EEMs to characterize different organic matter components. The technique was selected due to its simplicity in data interpretation, as this study primarily focused on fluorescence as a rapid monitoring tool for the performance evaluation of WWTPs. The excitation/emission wavelength ranges were selected based on maximum fluorescence intensities and reported in Table 1.

Fluorescence regional integration (FRI) is a technique used in fluorescence spectroscopy to analyze the EEMs of DOM in water samples. FRI involves dividing the EEMs into distinct regions based on fluorescence characteristics, such as protein-like, humic-like, and microbial-derived fluorescence. Each region represents specific components of the DOM pool, allowing researchers to identify and quantify different types of organic matter present in the sample. By integrating the fluorescence intensity within each region, FRI provides insights into the relative abundance and composition of different organic matter fractions. Five distinct zones, denoted as I, II, III, IV, and V, were identified in the EEM spectra of DOM (He et al. 2013). The regions are divided as shown in Table S1 in the supplementary information. The study evaluated WWTP performance and identified optimization opportunities by integrating fluorescence signals within predefined regions and comparing them with other chemical and microbiological analytical methods.

In the above equation, Ɵ_i represents the volume under intensity region at wavelengths ⁠, respectively. denote the wavelength increments for excitation and emission.

FIs such as the humification index (HIX) and freshness index/biological index (BIX) were utilized to elucidate the origin of organic matter in the samples. These indices provided valuable insights into the degree of humification and microbial activity, aiding in the differentiation between humic-like substances derived from terrestrial sources and autochthonous organic matter produced within the aquatic system. BIX is a fluorescence index calculated based on the excitation–emission spectra analysis. It is commonly used in environmental science to assess the biodegradability and microbial activity of organic matter in water samples. BIX was calculated using a standardized formula based on fluorescence peak intensities associated with protein-like fluorescence components, such as peaks T1 and B1. Specifically, BIX is determined by comparing the relative intensities of protein-like fluorescence peaks to humic-like fluorescence peaks, such as peak C, which represents humic substances derived from natural sources. FIs mentioned in Table S2 in the supplementary information were calculated to identify sources and other relevant information about DOM in the samples.

The peak intensities of X1 showed an average value of 52,741.71 ± 6,039.12 RFU for peak A and 6,685.26 ± 533.10 RFU for peak C, representing humic-like fluorescence. The X1 samples had the highest peak intensities of protein-like fluorescence, averaging 316,187.01 ± 86,170.68 RFU for peak T1 (tryptophan-like) and 392,777.81 ± 113,017.82 RFU for peak B1 (tyrosine-like). A 3D view of the EEM at X1 is shown in Figure 2(a).

Peaks T1, T2, B1, and B2, collectively called protein-like fluorescence, contributed the maximum to the X1 sample, as shown in Figure 2(b). Domestic wastewater fluorescence spectra are typically characterized by intense peaks in the < Em 380 nm region of protein-like fluorescence and significantly lower intensities for peaks A and C (Hur & Cho 2012; Carstea et al. 2016). Similar trends were also observed in studies conducted by Hudson et al. (2007), Carstea et al. (2016) and Derrien et al. (2019), where domestic sewage was the predominant source. Due to high tryptophan content, X1 had a greater concentration of anthropogenic organic matter than natural organic matter (Xiao et al. 2020). The possible interference from scattering made Peak B1 rarely examined in wastewater EEMs (Carstea et al. 2016), yet this fraction can demonstrate the freshness of contamination and the vicinity of the measurement location to the point of discharge.

Although detergents, one of the fluorescent whitening agents (FWAs) can often be found in domestic wastewater, the peak (Ex/Em: 370/410 nm) indicated the presence of FWAs was insignificant in the X1 samples. This can be due to the association of the fluorescence fraction of FWAs with particulate matter, preferably not seen in pretreated samples (Carstea et al. 2016). In addition to domestic wastewater, studies have discovered the presence of FWAs in landfill leachates and paper mill effluent (Graham et al. 2015).

FIs are other tools that help identify wastewater sources. The FI values in X1 samples ranged from 0.89 to 1.97. Domestic sewage typically has FI values of 2.1, which rarely change during treatment (Lin et al. 2023). Previous studies have highlighted the use of FI in identifying non-point source pollution (Jeon et al. 2022). However, FI was also found inappropriate for wastewater characterization compared to natural waters with many humic-like substances (Rodríguez-Vidal et al. 2020). Fluorescence analysis of DOM does not yield conclusive data regarding the biochemical composition or precise concentration of organic molecules (Fellman et al. 2010). Additionally, only a small portion of the DOM pool is likely responsible for its fluorescence. Temperature changes can cause modifications in fluorescence characteristics (Fellman et al. 2010). Additional challenges in utilizing and examining FRI include overlapping, multipeak response, and a non-linear correlation between the concentration of components and fluorescence intensity (Li et al. 2020). While FIs provide valuable insights into the sources and changes in organic matter across treatment units, their reliability and accuracy may vary depending on factors such as sample characteristics, environmental conditions, and treatment processes. Therefore, it is crucial to acknowledge the need for further research to validate the performance of FIs under various treatment scenarios and environmental conditions.

Pearson's r value of 0.898 and significance value p < 0.001 were observed for the selected variable. Regression statistics of the peak T1-DOC model are presented in Table S4 in the supplementary information. BOD had a measured value that was five times more than DOC content. According to studies, Peaks T1 and C have also been abundantly present in municipal wastewater and functioned as a tracer for contaminants in sewage-influenced natural water sources (Hudson et al. 2008; Bridgeman et al. 2015).

Raman and Rayleigh scattering effects interfered with fluorescence peaks. The second-order Rayleigh scattering interfered with the region representing fulvic acid, while second-order Raman scattering notably hindered the reading resembling soluble microbial products. After the scattered peak elimination by self-quenching effects, the inner filter effect modification adjusted the lowered fluorescence intensity to its actual values, especially at the low excitation–emission region (Kumar Panigrahi & Kumar Mishra 2019).

Peaks A and C in X3 showed 0.57 and 4.55% increase in intensity, respectively. This rise in intensity values, primarily seen in peak C of lower emission wavelength regions (humic-like fluorescence), can be explained by the breakdown of the aromatic structures of humic molecules into their respective smaller counterparts (Sgroi et al. 2017). Peak T1 fluorescence intensity decreased by 52.98%, whilst peak B1 fluorescence intensity decreased by 66.84% as shown in Figure 3(b). Various studies have demonstrated that treatment techniques cause the fluorescence intensity of peak T1 to noticeably reduce from influent to effluent (Reynolds & Ahmad 1997; Ahmad & Reynolds 1999). Protein-like fluorescence is directly related to bacterial communities' growth stage (Cammack et al. 2004; Elliott et al. 2006). Therefore, peak T1 is the labile proportion of DOM that is preferentially destroyed during the treatment process due to its biodegradability. The intensity variation corresponding to peak T2 showed both an increase and decrease in values from X1. This is attributed to the dynamic nature of peak T2 components consisting predominantly of soluble microbial products and biopolymers. The ambiguity in the intensity variation cannot be used to measure the efficacy of the treatment plant.

Among the indices shown in Table S6 in the supplementary information, the HIX value increased from X1 to X3, indicating a reduced carbon-to-nitrogen ratio after tertiary treatment. The freshness index, or BIX, was consistent across all samples, indicating that terrestrial species had no impact on the treatment units' operations or distribution networks. This suggested that microbially derived products majorly contributed to the high HIX value. The peak T1/ peak C ratio showed an average decrease of 54.23% after the treatment. This index depicted the reduction of the relative content of BOD to DOC in X3.

The degree of DOM removal is crucial in ensuring safe effluent discharge. DOM removal includes the removal of aromatic, hydrophobic fractions with the highest DBP formation potential (mainly trihalomethane), which interferes with disinfection reactions (Bieroza et al. 2010). Its structural and chemical properties change simultaneously during the reaction. In this way, predicting influent and effluent DOM content via fluorescence reading makes it a potentially useful tool to assess the potential for DBP formation in real time.

Regions I, II, and IV (Figure 4(a)–4(c)), which represented protein-like species, constituted 66% of the X1 fluorescence. This agreed that protein-like species were predominant in domestic wastewater influent. The protein-like fluorescence regions showed an overall reduction from X1 to X3, as shown in Figure 4(d). Due to the arbitrary area division, a significant contribution by fulvic and humic-like peaks can be noticed from regions III and V. However, humic and fulvic-like intensities were less in the samples. As a result, it is difficult to track the changes that are taking place in humic-like molecules. Further research is needed to explore the application of FRI in WWTP monitoring.

Peak A intensity increased by 3.68% in X2 and further by 0.56% in X3. A possible explanation for the rise in fluorescence during secondary treatment can be due to the polycondensation and humification taking place in the SBR because of microbial interaction (Andrade-Eiroa et al. 2013; Gabor et al. 2014). Due to the chlorine reactivity of humic compounds in the contact tank, the subsequent percentage rise of peak A in X3 is comparatively less. During chlorine disinfection, the DOM reaction prediction becomes more ambiguous. This depends on the amount and type of intermediate chlorination species formed due to chemical reactions (oxidation or substitution). Therefore, the rise in X3 intensity is not directly linked to microbial activity but rather caused by photodegradation interactions that may take place during UV disinfection (Baker et al. 2015). The addition of humic-like substances to the water from the conveyance system connecting treatment units might be another likely cause of this rise.

Peak C values showed a constant rise in SBR and the tertiary treatment unit. An intensity increase of 2.39%, followed by a further increase of 2.96%, was observed in the X2 and X3 samples, respectively. Studies by Yu et al. (2015) observed that this increase might be a result of the formation of soluble microbial products due to increased retention time in SBR. The fraction associated with peak C may also contain a higher proportion of recalcitrant humic moieties and humic-proteinaceous substituent links that are robust enough to prevent direct interaction with chlorine in the contact tank (Maqbool et al. 2020).

From Figure 3(b), a similarity in the trend of intensity variation can be seen for protein-like fluorescence T1, B1, and B2 and humic-like fluorescence A and C. The peaks B1 and T1 had previously been related to the biodegradable portion of DOM. Peak B1 levels fell by 71.70% in X2 and then rose by 15.02% in X3. Peak T1 levels decreased by 62.17% in X2 and increased by 24.29% in X3, respectively. Peak B1 showed a more significant effective reduction than peak T1 in comparison. The microbial decomposition causes peaks B1 and T1 intensities to decline in X2. The reduction in protein-like fluorescence peaks T1 and B1, during secondary treatment signifies the degradation of biodegradable organic matter, indicating the effectiveness of the treatment process in removing pollutants. This observation suggests that secondary treatment units play a crucial role in reducing the organic load of wastewater, thereby improving the overall quality of treated effluent. The reduction of peaks further reveals that this DOM fraction was assimilated into the biomass during the biological treatment.

Additionally, complex biodegradable organic components are broken down by aerobic microorganisms into their simpler substitutes. DOC and COD readings decreased from X2 to X3, unlike peaks T1 and B1. Therefore, the increase of peaks T1 and B1 in X3 (Figure 3(b)) can be due to the presence of other unknown fluorophores formed during photodegradation reactions during UV disinfection. The breakdown of active aromatic molecules into minor compounds has increased peaks T1, B1, and C in the lower emission wavelengths (Świetlik & Sikorska 2006; Korshin et al. 2018). Peak T2 showed marginal differences as these included tryptophan-like components generated from microbial by-products, which are highly labile and are influenced by external factors such as temperature, time, pH, etc. Therefore, a detailed study on peak T2 composition is required.

Several studies have explored fluorescence data to determine changes undergone by DOM across seasons. In the current investigation, samples taken in 2022 showed higher peak intensity values for all peaks than those collected in 2021. This can be due to the increased loading of wastewater from fully occupied and functional hostels compared to the previous pandemic year. Long-term data from the WWTPs can be used to derive a correlation among organic matter (OM) and specific peaks by statistical analyses, which can further be integrated into fluorescence-based online and thus help in real-time continuous monitoring and optimization of the treatment plants.

This study investigated the applicability of fluorescence spectroscopy for performance monitoring of WWTPs. Based on peak picking and FRI techniques for 3D EEM contours, the variations in wastewater characteristics, specifically DOM, are investigated. Characterization revealed substantial reductions in pollutants were observed after the complete treatment process: TSS was reduced by 94.38%, DOC by 83.27%, and COD by 85.38%. Microbial analysis identified the presence of fecal and total coliforms in the effluent, indicating the need for ongoing monitoring and treatment. Moreover, fluorescence spectroscopy highlighted variations in DOM composition, with protein-like fluorescence dominating the influent and undergoing degradation during treatment. Notably, peaks associated with biodegradable organic components showed significant reductions during treatment; the protein-like peaks T1 and B1 intensities showed a reduction of 62.17 and 71.7%, respectively, during secondary treatment and further decreasing by 52.98 and 66.84%, respectively, by tertiary treatment. However, peaks like A and C are increased after tertiary treatment due to the formation of refractory humic and protein-bound components and photodegradation in the UV-disinfection unit. Additionally, the HIX increased from influent to effluent, indicating a reduced carbon-to-nitrogen ratio after tertiary treatment, with an observed increase of 0.14–0.515. Meanwhile, the BIX remained consistent across all samples, suggesting microbial activity as a major contributor to DOM. These findings highlight the potential of fluorescence spectroscopy as a rapid and informative monitoring tool for DOM in WWTPs, offering valuable insights into pollutant composition and treatment efficacy. Future research efforts are required to incorporate quantitative measurements and long-term monitoring to further validate the utility of fluorescence spectroscopy for real-time assessment and optimization of wastewater treatment processes.

The authors would like to acknowledge the support received from IIT Delhi. No funding was received to conduct this study.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.