https://orcid.org/0009-0009-2223-4899
Figures
Abstract
Low-cost, scalable tests hold great potential to increase the availability of water safety data globally, especially in resource-limited settings. We evaluated the Lishtot TestDrop Pro, a commercially available, low-cost triboelectric-based sensor marketed for real-time determination of water contamination, including microbiological safety. We prepared a total of 199 water samples by serial dilution of raw wastewater influent into deionized water to produce a range of E. coli counts from <1 to too numerous to count (TNTC, > 300 per dilution plate) CFU/100 mL to evaluate the sensor. We tested these waters using the manufacturer’s instructions. We collected sensor readings using three settings on each of five devices, generating nearly 8,955 individual measurements to compare against E. coli measured via standard membrane filtration assays (EPA Method 1604), a common metric of drinking water safety. We found no statistical correlation between the sensor score and E. coli presence in water: 43% of sensor readings indicated “safe” when wastewater and E. coli was present, and 56% of samples lacking E. coli were deemed “unsafe” by the sensor when wastewater was absent. The Lishtot TestDrop Pro is not an accurate method for measuring microbiological water safety.
Citation: Purvis T, Nguyen T, McHugh C, Yeolekar S, Ding E, Wang J, et al. (2026) The Lishtot TestDrop Pro device cannot reliably indicate microbial water safety. PLOS Water 5(6): e0000442. https://doi.org/10.1371/journal.pwat.0000442
Editor: Sara Marks, Eawag, SWITZERLAND
Received: September 16, 2025; Accepted: May 26, 2026; Published: June 18, 2026
Copyright: © 2026 Purvis et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data have been made available for review in the supplementary information.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Water Quality Monitoring (WQM) is an important tool for maintaining or improving public health. Drinking water is an important exposure route for pathogens [1–3], toxic metals [4], and other contaminants [5–7]. While communities and governments can take preventative measures to reduce public health risk, WQM is often necessary to validate control measures [8]. WQM can also enable communities and public health practitioners to identify consistent risks and trends in drinking water quality [3]. Existing methods for WQM may often be cost-prohibitive or time-intensive. For measuring pathogens in drinking water, standard methods often require either a delay in results to culture a target organism or expensive equipment to identify and analyze genetic materials [9–11]. Dependence on capital-intensive standard methods may make WQM unfeasible in rural and remote regions, or in resource-limited areas where standard methods and laboratories are not financially or logistically feasible [12,13]. These barriers may lead to scarcity in health-relevant data generation for WQM.
Seeing the importance of WQM and the difficulties of generating such data, the water and health sector prepared guidelines for generating new acceptable technologies and methodologies. The WQM sector, through UNICEF, developed criteria for a low-cost (<$1000 USD), novel methodology, with additional performance requirements [14]. The criteria note that for rapidly evaluating health risks posed by pathogens in drinking water, methods should be capable of differentiating risk levels, as discussed below, in a timely manner (less than 2 hours) [14]. One proposed technology, the Lishtot TestDrop Pro, is a commercially available real-time water quality sensor, marketed for consumer use. Marketing material for the sensor suggests that it is capable of measuring Escherichia coli (E. coli) by measuring the proteins it releases, and potentially other contaminants like heavy metals, PFOS, and pesticides [15]. Despite known limitations, E. coli is widely used to indicate fecal contamination of water, and therefore the potential presence of enteric pathogens [16]. According to the manufacturer, the technology works by measuring triboelectric effects in a water sample. Triboelectric effects, or surface electrification, can occur when objects move across one another and create a buildup of static electricity on one surface [17]. The Lishtot sensor marketing appears to claim that swirling a water sample in a plastic cup generates enough distinguishable triboelectric effect above the water level to measure water quality, including E. coli contamination.
No prior peer-reviewed studies have evaluated the performance of this sensor in indicating water safety. One unpublished report presents ambiguous findings [18], and initial testing by our team revealed poor performance in indicating microbial contamination [19]. Given the lack of peer-reviewed evidence to date, we sought to evaluate the sensor’s ability to predict microbiological water safety via a systematic study of the sensor output across a range of water samples spiked with wastewater. We compared sensor results to a standard metric for water safety: E. coli counts in water as measured by membrane filtration.
Methods
Methods: Measurements
We prepared samples representing contaminated drinking water supplies (n = 199) by spiking serial dilutions of wastewater influent from the Orange Water and Sewer Authority (OWASA) into de-ionized (DI) water, representing fecal contamination of drinking water. We collected wastewater influent in a clean 1-Liter HDPE bottle at the OWASA site and transported it on ice to the laboratory for processing. We prepared dilutions of influent ranging from 10-9 to 10-3 to produce E. coli ranging from non-detect (<1 CFU/100-mL, Low Risk), 1–10 CFU/100-mL (Intermediate Risk), 10–100 CFU/100-mL (High Risk), and >100 CFU/100-mL (Very High Risk) as described in Table 1 below. We did not measure total dissolved solids or organic matter in this study. We used the 1000-mL diluted wastewater samples to take measurements with the Lishtot TestDrop Pro sensor.
We took sensor measurements by pouring 100 mL of well-mixed samples into a thin-walled polystyrene cup and gently swirling the 100 mL sample as specified by the manufacturer [15]. We poured a new 100-mL sample into a new cup from the original 1000-mL wastewater dilution for each sensor tested (n = 5). Each sensor contains three buttons (Tap, Bottle, or Environment). The test buttons of Tap and Bottle are said to test if the parameters are above the EPA guidelines, though the manufacturer does not mention the difference between these buttons or the use case of the Environment button [15]. Using the same sample cup for all measurements on the same device, we took three measurements using each button for a total of nine measurements per test. We took three measurements with the Tap button, then three measurements with the Bottle button, then three measurements with the Environment button. We briefly swirled the cup before each measurement. We took measurements by pressing and holding the relevant test button two to three inches from the wall of the cup, and in one motion, bringing the sensor to the surface of the cup, about half an inch above the water level, then releasing the button, as recommended by the manufacturer [15]. All measurements produced a water quality score from 0-99 (0 representing the worst water quality), which were stored on the technology’s mobile app, connected via Bluetooth. Lishtot reports representative score ranges of 99–90 as “very high quality water”; 81–90 as “no significant contamination detected”; 65–80 as “low level of contamination detected”; 50–65 as “intermediate level of contamination”; 20–50 as “high level of contamination”; and 0–20 as “heavily contaminated water” [15]. We tested each button three times on each of the five separate Lishtot TestDrop Pro sensors across 154 samples. We also tested known negative (n = 22) and known positive (n = 23) samples by measuring laboratory blanks of DI water and raw mixed influent, respectively.
We compared the sensor scores against the standard method of Membrane Filtration (MF) on BD Difco™ Chromogenic Dehydrated Culture Media: MI Agar (Maryland, USA). For each MF sample, we took a 100-mL sample from diluted wastewater by pipette and passed it through a 0.45-µm pore filter paper, and then transferred to MI agar before incubation at 35°C for 24 hours while inverted, per the EPA method 1604 [10]. E. coli colonies on MI agar present as blue dots under ambient light, which two authors counted by eye. We completed MF measurements in triplicate for each wastewater solution evaluated. While we conducted the pretest of the MI agar and the filter control, we did not conduct the phosphate-buffered dilution water control test [10]. We departed from method 1604 in counting by including readings of an individual plate up to 300 CFU/100 mL to include as much data as possible for analysis.
Methods: Analysis
We evaluated the sensor for its ability to accurately predict the WHO E. coli risk category (Low risk as <1 CFU/100 mL, Intermediate risk as 1- < 10 CFU/100 mL, High risk as 10- < 100 CFU/100 mL, and Very High risk as ≥100 CFU/100 mL E. coli) [20] for the wastewater dilutions. The true E. coli count was taken as the arithmetic average of the triplicate membrane filtration readings [21]. We compared MF readings against the average of the triplicate measurements for each sensor button on each device to determine suitability for predicting E. coli and the broader risk categories. The n = 199 samples varied in E. coli counts from non-detect to “too numerous to count” (TNTC noted as >300 CFU/100mL per individual plate), with count distribution shown in Table 1. The total set of readings comprised known-negatives (n = 990), known-positives (n = 1035), and other spiked samples (n = 6928). We prepared an equal number of serial dilutions each day, though this generated an unequal distribution of stock solutions in the various categories.
We evaluated false positives and false negatives, following the guidelines laid out by the UNICEF Target Product Profile – Rapid Water Quality Detection Tests guide [14]. We compared readings for these known negatives against the Lishtot score threshold of 81, above which “no significant contamination was detected” and therefore below which would indicate a false positive. Similarly, we compared known positive samples from the raw wastewater influent and compared them against the threshold of 81, above which would be considered a false negative.
We analyzed the data using Python and R. The analysis includes jitter plots, ANOVA significance tests with linear regression modeling (with significance at p = 0.05), and area under the curve (AUC) from Support Vector Classifiers (SVC) for each device and button [22].. Additionally, significant Shapiro-Wilk tests substantiated non-parametric distributions in the data, warranting a non-parametric evaluation using Spearman’s Rank-Order Correlation. We created regression models to evaluate the predictive power of sensor readings of each button on each device, and their combinations, for E. coli counts. We fit a basic linear model for each of the five devices and their three buttons, with their predictor being the aggregate of the three button press trials for the device, and the response being the actual E. coli count found through MF. We tested both the aggregate of individual buttons for a device (all Tap button measurements, all Bottle button measurements, and all Environment button measurements), and the aggregate of all button measurements for the device combined.
AUC is a method for estimating how well a model or device performs at identifying the difference between positives (E. coli presence) and negatives in observed data. We calculated the AUC measurement for each button (Tap, Bottle, and Environment) for each of the five sensors. We also calculated an aggregated AUC score across the various combinations of button and device. Within the AUC analysis, a score of 0.5 is considered equivalent to random categorization of presence/absence, with scores below this threshold indicating a device worse than random categorization. For this study, we use a standard heuristic of an AUC of at least 0.7 to indicate acceptable classification performance, with all else indicating poor or even detrimental performance [23]. A description of the AUC process and the resulting measure for individual buttons on each device is provided in the supplementary information. For Spearman’s Rank-Order Correlation, we considered a p-value of 0.05 statistically significant and evidence for a monotonic correlation.
Results
The sensor measurements ranged from 0 to 99 and are shown by WHO risk category in the jitter plot in Fig 1, with sensor score on the Y-axis. A breakdown of each device and button type (Tap, Bottle, Environment) is shown in Fig A in S1 Appendix. When discounting known positive and known negative samples, the remaining readings (n = 6928) had 22% and 44% of sensor scores returned 99 and 0, respectively. This is shown by the concentration of points at the poles of the jitter plot. The evaluation of DI water as known negative samples shows that 56% of readings resulted in a false positive reading based on the score cutoff of <81. A similar evaluation of raw wastewater influent as a known positive shows that 43% of readings resulted in a false negative reading based on the score cutoff of>=81. Samples in the Intermediate-High Risk categories showed a false negative percentage of 21%. The false positive and false negative rates exceed the UNICEF minimum-qualifying guideline of 15% [14]. All measurement data are available in S1 Data.
Fitted basic regression models did not result in any R2 value greater than 0.01. Additionally, no model for any device or button yielded a significant ANOVA F-statistic at p = 0.05, providing evidence of insignificant predictive power of the Lishtot score on E. coli counts for all buttons and devices. These values are listed below in Table 2.
Shapiro-Wilk tests on each of the devices’ aggregated sensor readings by button yielded p-values significant at 0.05, indicating the sensor reading values are very likely not normally distributed. Given the strong evidence for a non-parametric distribution of the data, Spearman’s Rank evaluation was used to determine any correlation between the sensor score and the averaged E. coli count, shown in Fig 2. Spearman’s rank correlation coefficient, or Spearman’s Rho (ρ), is -0.1493 with p = 0.0603. Since the ranking correlation is not significant at 0.05, there is not enough evidence to conclude that any monotonic correlation exists in the data between sensor readings and the true E. coli count. However, there is the possibility of noisy/skewed observations confounding results, specifically with samples with non-detectable E. coli counts. A Spearman’s Rank evaluation removing non-detect readings, which were clustered in the ranked ordering, indicates a Spearman’s Rho (ρ) of 0.0473 with a p-value of 0.6202, seen in Fig B in S1 Appendix. It is apparent that this result is not statistically meaningful, and there is not enough evidence to conclude that any monotonic correlation exists in our data despite removing this possible source of bias.
Further evaluation of the sensor scores as a presence/absence test was conducted with a classification model. The results in Fig 3 below show that the “Bottled Water” option has an Area Under the Curve (AUC) of 0.63 when aggregating the readings across all five devices, though this is beyond the recommendations of the manufacturer. A further analysis considering each and button (Tap, Bottle, and Environment separated) for each unique device results in a minimum and maximum AUC of 0.35 and 0.65 for a single device-button combination, respectively, as seen in Fig C inS1 Appendix.
Discussion
Across many samples, the Lishtot TestDrop Pro sensor scores do not appear to be associated with E. coli counts as estimated by the standard method and are a poor predictor of E. coli-based risk categories. The high number of readings at the measurement extremes of 0 and 99, which represented approximately 64% of all measurements, may lead to misinterpretation of water quality. Potential errors include the misclassification of both DI-blanks and raw wastewater influent as contaminated and clean, respectively. This misclassification was observed on all three button types across all five devices tested. The lack of statistically meaningful correlations in ranked data, shown by Spearman’s Rank evaluation, shows that the device is a poor predictor of E. coli counts. While the Lishtot TestDrop Pro meets UNICEF requirements regarding costs and throughput of samples, these benefits are moot when considering the rates of false negatives and false positives exceeding the defined 15% allowable threshold [14], which indicate that it is not a reliable method for measuring E. coli. The sensor is therefore inappropriate for indicating water safety.
It is possible that repeated measurements across many devices may result in an improved predictive ability for the presence/absence of E. coli. The distribution of AUC scores spans the 0.5 threshold (range: 0.35 to 0.65); however, equivalent to a random coin flip. Likewise, the high rates of false positive and false negative results indicate the lack of reliability in the generated results. At the time of this publication, no other peer-reviewed studies evaluating the sensor have been identified. One grey literature source suggests a correct presence/absence prediction of 76% of readings when measuring fecal contaminants, though this source seemed to suggest inconsistent readings within replicate samples [18]. This internal inconsistency appears similar to the observations of the current study.
Triboelectric effects have been used to directly measure some chemical compounds, such as mercury [24] and catechin [25], in agitated water systems. The success of triboelectric effect measurements for chemical contaminants has depended on selective surfaces, to which the target component is specifically attracted, and therefore surface electrification will occur more in the presence of the target compound. These measurements are typically taken against a baseline reading. The Lishtot TestDrop Pro manufacturer recommends using a generic plastic cup (PS, PP, or PET) but does not mention a selective surface [15]. The manufacturer notes that the sensor measures E. coli by measuring the proteins associated with the bacteria’s presence in water, though it does not mention a selective surface for these proteins or a protein in the marketing that would specifically identify pathogenic E. coli [15]. It is possible that the act of swirling the sample by hand does not allow for adequate buildup of surface electrification, or that the presence of E. coli does not create a unique electrification profile. The literature suggests that even DI water moving over a surface produces a triboelectric effect [26], with additional influences from Total Dissolved Solids [27], so the act of swirling should cause some triboelectric effect. The manufacturer does not specify the intensity of swirling needed, so the lack of standardization may also cause consistency issues. The use of diluted wastewater influent presented the ideal scenario for measuring any contamination in water through the Lishtot TestDrop pro. Wastewater influent will not only contain E. coli, but also other pathogens, chemicals, and organic matter. The lack of an observable trend between the Lishtot score and the dilution factor further suggests that the TestDrop Pro cannot reliably measure water contamination.
While the Lishtot TestDrop Pro sensor does not appear viable as a low-cost, rapid method of categorizing pathogenic risk, other novel methods exist that may prove appropriate. These include field-based biological methods like the Compartment Bag Test [28], Compact-Dry Plates [29], and the Aquatest method [29], which may be used without incubation [30], though these still require a delay between sample collection and results for decision-making. Similarly, the H2S test is a field-based method that may show microbial activity but cannot uniquely identify the presence of target microorganisms [31,32]. Emerging sensor technologies may offer advantages in throughput and time-to-result. There have been advances in the use of rapid proxy measurements like Tryptophan-like-Fluorescence (TLF) [33–35], though these may lack sensitivity or remain cost-prohibitive. These suggest that there are, in fact, novel methods which meet some of the UNICEF criteria, but there is no simple catch-all solution for all WQM scenarios. Sensors remain as high-throughput yet often costly methods, while microbial technologies are reliable but often leave a lag between sample collection and decision-worthy evidence generation. We recommend continued work by the sector to develop and rigorously examine novel techniques that may enable low-cost, timely, and reliable WQM evaluations.
Conclusion
In its current form, the Lishtot TestDrop Pro sensor does not seem to be a reliable tool in filling the capacity gap for WQM based on this study’s evaluation and the high prevalence of misclassifications. This work has served to highlight the need for evaluation of novel technologies related to low-cost WQM. The sector must find and implement solutions that adequately meet the health and safety needs of populations in a timely manner while remaining economically viable. Such solutions may rely on novel technologies, which should be properly evaluated before implementation. It may be useful to replicate this effort with other products and marketed WQM tests.
Supporting information
S1 Appendix. Description of Statistical Analyses.
https://doi.org/10.1371/journal.pwat.0000442.s001
(DOCX)
S1 Data. Sensor Test Scores and E coli measurements.
https://doi.org/10.1371/journal.pwat.0000442.s002
(XLSX)
Acknowledgments
The authors would like to thank the laboratory members at the Orange County Water and Sewerage Authority who supported in sample collection and preparation.
References
- 1. Mac Kenzie WR, Hoxie NJ, Proctor ME, Gradus MS, Blair KA, Peterson DE, et al. A massive outbreak in Milwaukee of cryptosporidium infection transmitted through the public water supply. N Engl J Med. 1994;331(3):161–7.
- 2. Wolf J, Johnston RB, Ambelu A, Arnold BF, Bain R, Brauer M, et al. Burden of disease attributable to unsafe drinking water, sanitation, and hygiene in domestic settings: a global analysis for selected adverse health outcomes. Lancet. 2023;401(10393):2060–71.
- 3. Bartram J, Ballance R. Water quality monitoring: A practical guide to the design and implementation of freshwater quality studies and monitoring programmes [Internet]. 1996 [cited 2026 Feb 8]. Available from: https://www.who.int/publications/i/item/0419217304
- 4. Fisher MB, Purvis T, Seidu Z, Roy S, Cawley M, Baldwin-SoRelle C, et al. Lead (Pb) contamination in drinking water in low- and middle-income countries: a systematic review and meta-analysis. 2025 [cited 2026 Feb 8]. Available from: https://eartharxiv.org/repository/view/9455/
- 5. Syafrudin M, Kristanti RA, Yuniarto A, Hadibarata T, Rhee J, Al-onazi WA, et al. Pesticides in Drinking Water—A Review. IJERPH. 2021;18(2):468.
- 6. Kalita I, Kamilaris A, Havinga P, Reva I. Assessing the health impact of disinfection byproducts in drinking water. ACS ES&T Water. 2024;4(4):1564–78.
- 7. Baluja R, Guo B, Howden W, Langer A, Lemoine D. PFAS-contaminated drinking water harms infants. Proc Natl Acad Sci U S A. 2025;122(50):e2509801122. pmid:41359852
- 8.
Water Safety Plan Manual: Step-By-step Risk Management for Drinking-water Suppliers [Internet]. World Health Organization; 2023 [cited 2026 Feb 8]. Available from: https://www.who.int/publications/i/item/9789240067691
- 9. US Environmental Protection Agency. SW-846 Test Method 9131: Total Coliform: Multiple Tube Fermentation Technique | US EPA [Internet]. 1986 [cited 2026 Feb 8]. Available from: https://www.epa.gov/hw-sw846/sw-846-test-method-9131-total-coliform-multiple-tube-fermentation-technique
- 10.
US Environmental Protection Agency. Method 1604. Total coliforms and Escherichia coli in water by membrane filtration using a simultaneous detection technique (MI medium). Washington DC: US Environmental Protection Agency; 2002.
- 11. Cao Y, Raith MR, Griffith JF. Droplet digital PCR for simultaneous quantification of general and human-associated fecal indicators for water quality assessment. Water Res. 2015;70:337–49. pmid:25543243
- 12. Crocker J, Bartram J. Comparison and cost analysis of drinking water quality monitoring requirements versus practice in seven developing countries. Int J Environ Res Public Health. 2014;11(7):7333–46. pmid:25046632
- 13. Kirschke S, Avellán T, Bärlund I, Bogardi JJ, Carvalho L, Chapman D, et al. Capacity challenges in water quality monitoring: understanding the role of human development. Environ Monit Assess. 2020;192(5):298. pmid:32307607
- 14.
UNICEF. UNICEF Target Product Profile Rapid Water Quality Detection Tests 4.0. 2023.
- 15.
Lishtot. Lishtot - TestDrop Pro - Test Your Tap Water!. 2025.
- 16. Edberg SC, Rice EW, Karlin RJ, Allen MJ. Escherichia coli: the best biological drinking water indicator for public health protection. J Appl Microbiol. 2000;88(S1):106S-116S.
- 17. Burgo TAL, Galembeck F, Pollack GH. Where is water in the triboelectric series? J Electrostat. 2016;80:30–3.
- 18.
Murcott S, Van Hemel A, Urquhart I. Lishtot TestDrop Pro Test Protocol and Interpretation. 2019.
- 19. Purvis T. An Evaluation Of Low-Cost, Novel Methods For Risk-Based Monitoring Of Water Quality For Fecal Contaminant [Internet]. [Chapel Hill]. 2022 [cited 2026 Feb 8]. Available from: https://cdr.lib.unc.edu/concern/masters_papers/76537b367?locale=en
- 20. World Health Organization. Guidelines for drinking-water quality: small water supplies. 2024. Available from: https://www.who.int/publications/i/item/9789240088740
- 21. Haas CN. How to average microbial densities to characterize risk. Water Res. 1996;30(4):1036–8.
- 22. Cortes C, Vapnik V. Support-vector networks. Mach Learn. 1995;20:273–97.
- 23.
Hosmer DW, Lemeshow S, Sturdivant RX. Applied Logistic Regression: Third Edition. Wiley; 2013. p. 1–510.
- 24. Lin ZH, Zhu G, Zhou YS, Yang Y, Bai P, Chen J, et al. A Self-Powered Triboelectric Nanosensor for Mercury Ion Detection. Angew Chem Int Ed. 2013;52(19):5065–9.
- 25. Chatterjee S, Saha S, Barman SR, Khan I, Pao Y-P, Lee S, et al. Enhanced sensing performance of triboelectric nanosensors by solid-liquid contact electrification. Nano Energy. 2020;77:105093.
- 26. Ravelo B, Duval F, Kane S, Nsom B. Demonstration of the triboelectricity effect by the flow of liquid water in the insulating pipe. J Electrostat. 2011;69(6):473–8.
- 27. Mao W, Yang S, Xu G, Liu P, Zhong T, Dong J, et al. The effect of ions on liquid-solid interface investigated by charge sensitive direct current droplet-based electricity generator. DeCarbon. 2024;4.
- 28. Stauber C, Miller C, Cantrell B, Kroell K. Evaluation of the compartment bag test for the detection of Escherichia coli in water. J Microbiol Methods. 2014;99:66–70. pmid:24566129
- 29. Brown J, Bir A, Bain RES. Novel methods for global water safety monitoring: comparative analysis of low-cost, field-ready E. coli assays. npj Clean Water. 2020;3(1):1–6.
- 30. Brown J, Stauber C, Murphy JL, Khan A, Mu T, Elliott M, et al. Ambient-temperature incubation for the field detection of Escherichia coli in drinking water. J Appl Microbiol. 2011;110(4):915–23. pmid:21214694
- 31. Manja KS, Maurya MS, Rao KM. A simple field test for the detection of faecal pollution in drinking water. Bull World Health Organ. 1982;60(5):797–801. pmid:6983930
- 32. Khush RS, Arnold BF, Srikanth P, Sudharsanam S, Ramaswamy P, Durairaj N, et al. H2S as an indicator of water supply vulnerability and health risk in low-resource settings: a prospective cohort study. Am J Trop Med Hyg. 2013;89(2):251.
- 33. Bedell E, Harmon O, Fankhauser K, Shivers Z, Thomas E. A continuous, in-situ, near-time fluorescence sensor coupled with a machine learning model for detection of fecal contamination risk in drinking water: Design, characterization and field validation. Water Res. 2022;220:118644. pmid:35667167
- 34. Bedell E, Sharpe T, Purvis T, Brown J, Thomas E. Demonstration of Tryptophan-Like Fluorescence Sensor Concepts for Fecal Exposure Detection in Drinking Water in Remote and Resource Constrained Settings. Sustainability. 2020;12(9):3768.
- 35. Sorensen JPR, Lapworth DJ, Marchant BP, Nkhuwa DCW, Pedley S, Stuart ME, et al. In-situ tryptophan-like fluorescence: A real-time indicator of faecal contamination in drinking water supplies. Water Res. 2015;81:38–46. pmid:26026711