The River Wye catchment covers an area of approximately 4130 km² (Bussi et al., 2018), spanning both England and Wales. The catchment has an active and growing community of engaged citizen scientists who regularly monitor water quality using low-cost colorimetric test kits and handheld digital sensors. In the year-long period between March 2021 and February 2022, volunteers across the Wye catchment analysed a total of 7,439 water samples at 339 volunteer river monitoring sites. Thirty-four of these volunteer monitoring sites were selected for comparative laboratory analysis, based on where citizen scientist volunteers were available to collect samples and carry out field tests on specific dates (Figure 1). Monitoring sites that contributed towards a good representation of the geographical range across the catchment and of different volunteer monitoring groups were also prioritised.

Volunteers from the Wye Salmon Association (WSA), Friends of the Upper Wye (FOUW), Friends of the Lugg (FOL) and Campaign to Protect Rural England (Herefordshire) (CPRE), participated in data collection for this study. These groups agreed a set of monitoring tools and co-developed shared protocols outlining how citizen science water quality sampling and analysis should be conducted across the River Wye catchment (known as the Wye Alliance). This approach aims to ensure that the citizen science datasets produced are comparable. Groups train and equip their own volunteers with the tools required to safely and effectively carry out water quality testing at agreed fixed sites across the catchment either twice weekly, weekly or fortnightly. Samples were collected in plastic or glass containers, provided by volunteers, that were triple-rinsed on site with river water before use. Volunteers followed detailed step-by-step guidance on how to perform each test in-situ.

Nutrient concentrations were analysed by citizen scientists using two phosphate and one nitrate method: Hach Water Quality Test Strips for Nitrate and Nitrite (‘Hach Nitrate test strips’, Hach Lange, Berlin, Germany), and La Motte Insta Test low-range test strips for phosphate (‘La Motte Phosphate test strips’, La Motte Company, Maryland, United States) and the Hanna Phosphate Low Range Handheld Digital Colorimeter (‘Hanna Phosphate Checker’ or ‘Hanna Checker’, Hanna Instruments, Rhode Island, United States) (Table 1). Users followed the protocols for reaction times given by the manufacturers, and recorded supporting observations of environmental conditions and physiochemical water quality indicators including temperature and electrical conductivity (measured using a HM Digital EC-3 handheld probe (HM Digital Inc., California, United States)), turbidity (using a standard Secchi tube, also known as a transparency tube or turbidity tube (Dahlgren et al., 2004)) and estimated water level.

Samples were collected during three separate ‘mass sampling events’, where volunteers collected measurements and water samples from different sites on the same day, similar to other citizen science studies (Muenich et al., 2016; Stankiewicz et al., 2023). This resulted in a ‘snapshot in time’ of water quality status across the catchment on 21st September (A), 19th October (B) and 16th November (C) 2021. In total, 491 samples were collected (141, 155 and 185 on each event, respectively), and tested by volunteer citizen scientists from their regular WSA, FOUW, FOL or CPRE monitoring sites. Statutory bodies (the Environment Agency in England and Natural Resources Wales in Wales) also aligned their routine sampling across the catchment to take place on the same days. A subset of volunteers were asked to collect a second sample (or ‘paired sample’) at the same time and from the same location for comparative analysis using laboratory methods. In total, 70 samples were collected by citizen scientists for paired laboratory analysis at the same time they collected samples for immediate in situ analysis using the low-cost test kits. Of these, 56 compared laboratory analysis with the Hanna Phosphate Checker; 51 with the La Motte Phosphate Test Strips; and 50 with Hach Nitrate test strips.

Samples for accredited laboratory analysis were collected by volunteers in a pre-prepared 500 mL PET bottle (inorganic nutrients: nitrate, nitrite, phosphate), provided by the United Kingdom Accreditation Service (UKAS) accredited Dŵr Cymru Welsh Water laboratory. Bottles were triple-rinsed with river water prior to sample collection and stored in cool bags for transit to the laboratory within 8 hours. Samples were filtered on arrival at the laboratory, prior to analysis within 24 h. Inorganic orthophosphate as P (PO₄-P), total oxidised nitrogen as N (TOxN-N) and nitrite as N (NO₂-N) were analysed using a discrete analyser (Aquakem 600). Nitrate as N was calculated from nitrite and TOxN-N concentrations measured (TOxN-N–NO₂-N). The laboratory measured phosphate as P in the range of 0.03–1.2 mg/L with a precision of 0.001 mg/L, while nitrate as N ranged from 0.48–60 mg/L with a precision of 0.01 mg/L. Uncertainty of measurement (UoM) for analysis of these nutrients was 7.5% for phosphate and 8.7% for nitrate.

To understand the performance of low-cost methods under varying temperature and reaction time, phosphate standards were prepared and analysed using both the La Motte Phosphate test strips and the Hanna Phosphate Checker under controlled laboratory conditions, and by a Thermo Scientific Gallery Plus Discrete Photometric Analyser at the University of Bristol’s School of Geographical Sciences laboratory. The Gallery Plus discrete analyser reports phosphate as P from 0.002 to 2.5 mg/L (if both the low and high range assays are used) with a resolution of 0.00001 mg/L. Tests were carried out in triplicate to ascertain the Standard Error of the Mean (SEM), which was calculated as 1.1% and used as an estimate of the UoM for phosphate analysis using the Gallery Plus discrete analyser. Eleven phosphate standards were made up from 5 mg/L phosphate stock solution, ranging in concentration from approximately 0.1–2.5 mg/L PO₄ ^3-. Standards were tested using the low-cost methods at three different controlled temperature ranges set by water baths: 5 °C ± 1 °C, 11 °C ± 1 °C and 21 °C ± 1 °C, to reflect the typical range of average temperatures observed in United Kingdom rivers annually (Johnson et al., 2009; Jonkers and Sharkey, 2016). After samples were brought to the correct experimental temperature, tests were conducted at room temperature for the 11 °C and 21 °C samples, and for the 5 °C samples, tests were carried out in a cold room set at 5 °C to mimic winter conditions.

For La Motte Phosphate test strips, the procedure involved: (i) filling a plastic test tube with 10 mL of water sample (to the indicated mark); (ii) inserting a strip containing the regent into the test tube cap; (iii) placing the cap on the test tube; and, (iv) inverting the sample five times. The manufacturer’s stated method (followed by volunteers in this study) indicates that the colour of the reacted sample should be compared to the colour chart and a concentration determined immediately after the test procedure has been carried out. However, to examine the hypothesis that this method does not allow the reagent sufficient time to combine with the phosphate in colder weather, we compared the manufacturer’s method with an amended timescale. In a supplementary set of tests at 11 °C ± 1 °C, we tested samples immediately after the test procedure had been carried out (time = 0 min) and then left each sample for an additional 3 and 10 min before determining the closest colour chart match. For the Hanna Checker, we increased the reaction time of the 5 °C ± 1 °C samples by re-evaluating the apparent concentration of the original sample on the handheld colorimeter after a time interval of between 10 and 30 min after the first test result was recorded.

All citizen science groups involved in this study used Epicollect, a free online data gathering platform, to record and store their water quality data. For each dataset downloaded from Epicollect the following steps were taken: (1) redundant metadata were removed; (2) headings were standardised; and, (3) site names were replaced with sites unique identifying code and fixed site location (coordinates) from each group’s master site log. Erroneous data points were removed against the following criteria: for test strips, any data points that did not match a value defined by the corresponding test strip colour chart were excluded; for Hanna Phosphate Checkers, any data that was reported either outside the instrument range or beyond its limit of sensitivity were discarded. Lastly, all phosphate data collected by volunteers (using the Hanna Phosphate Checker or the La Motte test strips) were converted to phosphate as P (PO₄-P) in mg/L to match data reported by the accredited laboratory and commonly used by statutory bodies. Hach Nitrate test strips were reported as NO₃-N in mg/L.

The data provided by test strips is categorical (approximate, banded measurements, also described as semi-quantitative, as displayed in Table 1), making a direct comparison with digital data (numerical measurements, also described as quantitative) - produced by professional methods of analysis - challenging. To address this, we categorised or ‘binned’ digital data from laboratory methods of analysis to enable a direct comparison with the test strip data, as has been done in other similar studies (e.g., Muenich et al., 2016). The binned ranges used for nitrate and phosphate are shown in Table 2. For digital data provided by the Hanna Checkers, we were able to perform regression analysis in R (R Core Team, 2024) to directly compare the volunteer’s measurements with accredited laboratory data.

To quantify the probable error of the test kits investigated, we first assumed the ‘true value’ to be the concentration of phosphate or nitrate determined in the laboratory (whilst acknowledging that these reported values are subject to stated uncertainty of 7.5% for phosphate and 8.7% for nitrate). Absolute error, as shown in Equation 1, was calculated each of the low-cost field tests using the field test result obtained by volunteers as the ‘measured value’ and the paired laboratory result as the ‘expected value’. However, due to the semi-quantitative nature of phosphate and nitrate test strips, a modified method (shown in Figure 2 and summarised in Equation 2) was used for those tests to account for the range of values each test strip result represents (shown in Table 2). This method produced values indicating absolute error outside the expected range for test strip readings. Mean absolute error outside the expected range was then calculated by summing all the values calculated and dividing by total number of measurements, including those assigned an error of zero. Absolute error = measured value  –  expected value (1) Absolute error  o u t s i d e   t h e   e x p e c t e d   r a n g e = test strip result  binned range   –  lab result (2)

The bias of each test was assessed by quantifying whether the results that did not match the expected value or range underestimated or overestimated the ‘true value’, as defined by lab analysis of paired samples.

The total number of samples analysed by each of the low-cost test kits and paired professional laboratory techniques during the mass sampling events are summarised in Table 3. Test strip performance was variable (Figure 3), though nitrate tests generally showed better agreement with laboratory data than phosphate tests, particularly when considering higher concentrations of nitrate (above 1.5 mg/L NO₃-N). Fifty-four percent of the nitrate test strip results from mass sampling events were in agreement with comparative laboratory analysis, which improved to 66% including laboratory uncertainty (UoM). Furthermore, no nitrate test strip results fell more than one binned category outside the expected range based on laboratory analysis (Figure 4a). In contrast, just 14% of phosphate test strip results from the mass sampling events were in agreement with the binned range determined by paired laboratory analysis of river water samples, improving to only 17% when the UoM in the laboratory results was account for. Correspondingly, over 50% of the phosphate test strip data deviated from laboratory results by two or more binned categories (Figure 4a). Nitrate test strip results identified as not being in agreement with the binned range determined by laboratory analysis of paired samples showed a bias towards overestimation of nitrate whilst, conversely, the phosphate test strip results which were not in agreement with the binned range determined by paired laboratory analysis showed a strong bias towards underestimation of phosphate (Figure 4b).

Although over 50 paired samples were analysed for the phosphate test strips across all three sampling events (Table 3), a large proportion (22/51) of samples produced laboratory test results below the lower limit of detection (LoD) for phosphate as P on the Aquakem 600 (<0.03 mg/L). As it is not possible to determine which of the two lowest binned ranges (0–0.016 or 0.017–0.049 mg/L PO₄-P) laboratory data below the LoD should be classified as, and thus correctly identify a binned range match or mismatch with the corresponding phosphate test strip data, those paired samples were excluded from further analysis in this study. Similarly, for the Hanna Checker, a large proportion (18/56) of the paired samples analysed fell below the LoD for phosphate as P on the Aquakem 600, as shown in Table 3. These data were also removed from further analysis in this study as they would skew the relationship between Hanna Checker and laboratory determined P concentrations at the lower range. The 50 nitrate samples analysed during mass sampling events all fell above the laboratory limit of detection for nitrate as N (0.48 mg/L) so all were included in further analysis.

The results from comparative analysis of prepared phosphate standards under controlled laboratory conditions are displayed in Figure 5. Testing the performance of La Motte phosphate strips in this way demonstrated that lower concentration solutions (<0.4 mg/L PO₄-P) tended to give test strip results closer to the true concentration of phosphate in most cases, but false zeros were also more common in these samples. Conversely, higher concentrations (>0.4 mg/L PO₄-P) gave more consistent readings above zero, but failed to accurately indicate the difference in concentration of samples in this range and consistently underreported phosphate concentrations.

An increase in the reported concentration of PO₄-P with increasing temperature of the sample is observed when the mean differences and t-values are calculated for the same standards measured at different temperatures (Table 4). However, increasing the temperature of samples from 5 °C to 11 °C or from 11 °C to 21 °C did not yield a statistically significant effect on the outcome of the test strip results (p-values greater than 0.05 and confidence intervals that included zero). Only in comparing samples at 5 °C and 21 °C, did the paired t-test indicate a statistically significant (95%) probability of a slight increase in PO₄-P concentration with temperature (Table 4).

When the strips were left to develop for longer time periods (3 or 10 min), the reported phosphate concentration generally increased, often by at least one colour chart category after 3 min or 10 min (Figure 6). Paired t-test results indicate a statistically significant increase in phosphate concentration reported between 0 and 3 min, and between 0 and 10 min (p < 0.05). However, no significant difference was observed between 3 and 10 min (Table 4).

The relationship between phosphate results obtained from the Hanna Phosphate Checker and from professional laboratory analysis of paired samples taken across all three mass sampling events is displayed in Figure 7. The line of best fit shown is a simple linear regression (y = 0.9362x - 0.0034, r ² = 0.65, p < 0.0001, T = 7.67). Further analysis of the data (Table 6) shows that 62% of the 38 comparable results reported by citizen scientists using the Hanna Phosphate Checker were in agreement with laboratory results when the total known error was accounted for. Results in agreement are displayed in black in Figure 7.

Although the Hanna Checker has a detection limit of zero, manufacturers guidance indicates an accuracy equivalent to ±0.013 mg/L + 4% PO₄-P (at 25 °C), hence the working detection limit on these devices is likely to be 0.013 mg/L PO₄-P. A substantial number 18/56 (∼32%) of samples analysed for phosphate using the Hanna Phosphate Checker were determined to be below the lower limit of detection for the laboratory (0.03 mg/L PO₄-P). Although not included in the analysis presented in this section, it is important to consider that a high proportion (over 80%) of those paired samples recorded as below the lower detection limit by laboratory analysis were also reported as values of 0.03 mg/L or less PO₄-P by Hanna Phosphate Checker users. Hence, there is still a good general agreement at the lower range but the assessment becomes semi-quantitative, as with the test strips, and it is not possible to state the specific accuracy, precision, bias or error on these results.

Comparison of the Hanna Phosphate Checker against accredited laboratory methods under controlled environmental conditions showed very good agreement at all temperature ranges (Figure 8; Table 5 (R ² = 0.98, p < 0.0001)). At lower temperatures the Hanna Checker slightly underestimated phosphate concentrations: the mean difference between the value measured by the Hanna Checker and the value measured by the Gallery Plus analyser at 5 °C was −0.046 mg/L PO₄-P, improving to −0.0076 mg/L PO₄-P at 11 °C. However, the absolute difference between values was much greater for higher concentration P solutions.

The agreement between PO₄-P concentration measured by the Hanna Checker and the Gallery analyser improved as temperature increased (Figure 8): the biggest improvement in relationship was between 5 °C and 11 °C as mean absolute error improved from ±0.048 mg/L at 5 °C to ±0.011 mg/L at 11 °C, and remained ±0.011 mg/L at 21 °C (Table 6). For 11 °C and 21 °C samples, 100% and 92% of results respectively were in agreement with laboratory analysis when the manufacture stated equipment error on the Hanna Checkers and the error on laboratory analysis is accounted for, whereas this is limited to 42% of paired values at 5 °C (Table 6). At the lowest controlled temperature range (5 °C), a subset of samples were left for a longer reaction time (10–30 min before re-assessing the concentration using the colorimeter), which resulted in an increase in measured concentration and better agreement with the laboratory analyses (Figure 8).

Data from the mass sampling events demonstrated that the Hach nitrate test strips performed well in the field, with good accuracy when compared to professional laboratory analysis of NO₃-N: 54% of results were in direct agreement with laboratory analysis when data was binned, and 66% of results were in agreement once the uncertainty on laboratory analysis was also accounted for. Furthermore, no results were more than one test strip category (binned range) outside the expected result. These findings are in agreement with a study carried out as part of the Wabash River Sampling Blitz (Muenich et al., 2016) where the same Hach nitrate test strips were assessed. Similarly, Muenich et al. found that 55% of nitrate + nitrite results reported by volunteers matched the range determined by laboratory analysis, and that 84% of test strip results were no more than one binned category outside the expected results based on laboratory analysis of paired samples.

In contrast, the La Motte phosphate test strips did not perform well in the field: only 17% of comparable test strip readings reported were in agreement with the expected binned results based on laboratory analysis, after the uncertainty on laboratory results was accounted for, and 52% of test strip results were more than one bin outside the expected range determined by laboratory analysis of paired samples. Although we have not found any other studies which carry out a similar comparative analysis on the La Motte Insta test strips for phosphate, the reliability of alternative commercially-available, low-cost test kits for phosphate analysis of freshwater samples have been reviewed by others: Muenich et al. (2016) found that the Hach Aquachek test strips (Cat. 27,571–50) (0–50 ppm PO₄) produced results which matched the binned range determined by laboratory analysis in just 33% of samples but which fell within one bin of the laboratory result in 99% of samples on paired analysis undertaken during the Wabash River Sampling Blitz. However, the Hach test strips are designed to detect a much wider range of P concentrations compared to the La Motte test strips and typical concentrations of phosphate found in rivers (Meybeck, 1982). Correspondingly, they have low sensitivities, reflected in the large colour chart increments (0, 5, 15, 30, 50 mg/L PO₄ ^3-). Thus, as most of the samples measured in the study by Muenich et al. (2016) were below 5 mg/L PO₄ ^3-, they were categorised into either of the two lowest binned ranges (0–2.5 mg/L or 2.6–10 mg/L), demonstrating the limited practical application of the Hach test strips for monitoring phosphate in river water.

Quinlivan et al. (2020) assessed the performance of an alternative low-cost phosphate test, the Kyoritsu PACKTEST (Kyoritsu Chemical-Check Lab, Corp., Tokyo, Japan) (0–2 ppm PO₄), and found that 53% and 36% of results produced were in agreement with binned ranges determined by laboratory analysis at each of two different river sites in South-West Ireland. These tests have a similar range (0–2 mg/L PO₄ ^3-) and sensitivity (colour chart increments: 0.05, 0.1, 0.5, 1, 2 mg/L PO₄ ³) to the La Motte phosphate test strips, but the results of the study by Quinlivan et al. (2020) suggest the Kyoritsu PACKTEST produce more accurate results. However, while 60 tests were undertaken at each site, only two sites were analysed. Lévesque et al., 2017 also assessed the Kyoritsu PACKTEST and found a strong agreement with professional laboratory analysis: 81% of the 111 paired samples assessed using the kits in the same binned range as laboratory results. However, all samples in their study were determined by laboratory analysis to be in the same (lowest) binned range of <0.02 mg/L, which again limits our knowledge of the low-cost test kit’s performance across a range of environmental conditions and concentrations. Ultimately, the results from our study and others (Quinlivan et al., 2020; Muenich et al., 2016; Lévesque et al., 2017) indicate that there is a substantial challenge in finding low-cost phosphate tests that provide both the sensitivity of test required to define meaningful differences in concentration in freshwater and a good level of accuracy within these ranges.

Further analysis of the phosphate test strips under controlled laboratory conditions using phosphate standards, rather than river water samples tested in the field by volunteers, demonstrated a marked improvement in their performance. Overall 35% of test strip results from the controlled laboratory tests were in direct agreement with professional analysis of samples (compared to 17% of field results) and 79% were no more than one binned category outside the expected range (compared to 48% of field results). These improvements in performance could be due to a range of factors, including: the use of phosphate standard solutions instead of river water samples that remove interferences; controlled light conditions that reduce perceived differences in the colour change produced by the reaction; consistency in technique as a result of one person rather than multiple individuals performing tests; and consistency in interpretation due to differences in individual’s unique perception of colour. Temperature was unlikely to be a substantial influence in this test since samples assessed in controlled laboratory conditions were tested at a similar temperature to the river water samples (average 11 °C compared to 12 °C). However, temperature did have a significant influence on La Motte test strip results when samples were increased from 5 °C to 21 °C. Although further analysis with a larger sample size is required to confirm this, it is likely that with colder river water samples (<10 °C) warming them (to >10 °C) could lead to small improvements in the accuracy of results reported by volunteers.

The La Motte phosphate test strips appear to under-report concentrations of phosphate in both field and under controlled laboratory conditions, which volunteers suggested with either because the reaction time allowed by the test strip method may be insufficient for the reagent to fully react with the phosphate in the sample, resulting in limited colour development, or because the colour charts used for comparison may be difficult to interpret. Our controlled laboratory analysis of the phosphate tests showed that waiting longer time intervals (3 or 10 min) after carrying out the test procedure and before assessing the test result from the colour chart improved the accuracy of the results. Only in a very limited number of cases did an increase in reaction time result in an overestimation of PO₄-P compared to professional standard laboratory analysis of paired samples. These findings support those of Jayawardane et al. (2012) who explored the effect of reaction time on colour intensity using similar low-cost colormetric phosphate tests. They observed a colour intensity increase up to 40 min following the initial reaction and after that colour intensity would begin to reduce (as the products degraded), and suggested that assessing colour intensity of the reaction after 10 min balanced allowing sufficient colour development to permit an accurate reading, without making the time taken to assess samples in the field onerously long. Our study suggests that as the change in concentration between 3 and 10 min is not statistically significant, a shorter time period still–somewhere between 3 and 10 min - might obtain optimum results for the La Motte phosphate tests.

Handheld digital colorimeters, such as the Hanna Phosphate Checker, are higher cost and more complicated to use than the test strips but can report concentrations with better precision and, critically, remove colour perception subjectivity in the reporting of results. The Hanna Phosphate Checker performed well in this study, producing results that generally provided good agreement with paired professional laboratory analysis, both when samples were analysed by citizen scientists in the field (62%) and very good agreement with paired laboratory analysis when phosphate standards were analysed under controlled laboratory conditions (up to 100%). The lower degree of accuracy demonstrated in field tests than under controlled laboratory conditions using the Hanna Checker could be due to a range of factors, including: human error, field conditions, sample interferences, reporting errors, storing and transporting samples etc. It should also be noted that our mass sampling events were in autumn, when river temperatures ranged from 10.0 °C to 15.7 °C, which is below the manufacturer’s recommended operating temperature (25 °C).

Our results clearly demonstrate the effect of temperature on results produced by the Hanna Checker; the improved performance between 5 °C and 11 °C is substantial but any further improved performance between 11 °C and 21 °C is minimal. Under controlled laboratory conditions, Hanna Checker results also indicate that if increased reaction times between the sample and the reagent are allowed at lower temperatures, the accuracy of results improves. This suggests that at lower temperatures (<10 °C), the accuracy of results could be improved by either (i) leaving the sample to react for longer than stated by the manufacture before assessing concentration of PO₄-P, or, (ii) bringing samples to room temperature before analysis. Nevertheless, the Hanna Checker performs well in “real-world” conditions, assessed down to 0.03 mg/L PO₄-P in this study, and can make a valuable contribution to in situ monitoring of phosphate.

Whilst quantification of the performance of the Hanna Phosphate Checker at lower concentrations of river water samples (<0.03 mg/L PO₄-P) would be advantageous for a full understanding of biogeochemical processes, analysis was carried out down to 0.005 mg/L on controlled laboratory analysis of phosphate standards. Furthermore, if concentrations of P are below the detectable threshold used for comparative analysis in this study (<0.03 mg/L) then they are generally within regulatory limits that classify the ecological status of a river as ‘good’, as defined by United Kingdom Technical Advisory Group (TAG) under the WFD (Fones et al., 2020). Arguably, higher concentrations (or nutrient ‘hotspots’) are of most interest in assessing river water quality and identifying samples as <0.03 mg/L PO₄-P, even if the accuracy of tests below this limit are poor, would allow the Hanna Checkers to act a useful screening tool to detect low concentration samples.

Whilst citizen science data are unlikely to perform to the same standards (sensitivities, limit of detection) as data acquired by professionals using more expensive field or laboratory methods of analysis, this study has shown improved understanding of the limitations and characteristics of low-cost methods of analysis means that reliable data of a consistent and known quality (accuracy, bias, precision) can be gathered by citizen scientists. It is, however, critical that the aims and objectives of a citizen science monitoring programme are set out with test kit performance in mind in order to gather meaningful data.

Colorimetric test strips, such as those assessed in this study, are designed primarily as education tools (Kosmala et al., 2016) or for specific non-environmental monitoring purposes, such as for home aquariums, hot tubs or swimming pools, so their applicability to a wider range of freshwater environments is not necessarily intended. First, the temperatures of these recreational water systems are typically distinctly warmer than United Kingdom rivers, which is likely to affect reaction times and test results (in this study, average water temperature across all 67 sites assessed: 12.8 °C, min = 10.0 °C, max = 15.7 °C). Second, low ionic strength freshwaters often challenge field analytical methods (Bagshaw et al., 2016) and performance may be affected by interference from other chemical species present in high concentrations in natural water bodies, such as silicate (Jayawardane et al., 2012). Third, degradation of reactants on the test strips may also occur when they are exposed to water or UV for example, (Jayawardane et al., 2012), so correct storage and adhering to expiry dates is very important and could easily be overlooked. Finally, the overall expected ranges of phosphate in freshwater are very low compared to nitrate, usually approximately an order of magnitude less (Meybeck, 1982), so methods suitable for some analytes of interest may not be applicable to all. Producing low-cost colorimetric tests that are sensitive enough to detect subtle step changes at such low concentrations whilst making these changes clearly visible to the naked eye is a substantial challenge facing reliable citizen science phosphate monitoring using these tools.

Low-cost test kits that do not rely on subjective interpretation by the user and produce results comparable to professional laboratory analysis are highly desirable. Whilst they may be more difficult to use than test strips, and therefore require more comprehensive training or leave more room for human error in test procedure, the Hanna Checker demonstrated its ability to produce results very close to professional laboratory analysis. Those designing citizen science monitoring programmes need to consider whether tests that are more expensive and more involved for volunteers to use (such as the Hanna Checker) are required to meet the objectives or the monitoring programmes or whether simpler, lower cost alternatives (such as colorimetric test strips) are sufficient: is the improved data quality (in terms of the degree of accuracy and precision provided) required to reliably answer the scientific research questions posed, or, is a lesser but known quality of data sufficient?

Our assessment shows that lower cost test strips for nitrate perform well within the sensitivity ranges provided by the manufacturer (±0.06 mg/L on average). Conversely, phosphate test strips appear to be better suited to indicating the relative (low, medium or high) concentrations of phosphate rather than the more refined categories presented in the color chart provided by the manufacturer. The ongoing use of test strips across the concentration range of interest needs to be carefully considered. Our data demonstrate that when water temperatures are <20 °C, increasing the reaction time of the La Motte strips from the manufacturer’s recommended 15 s up to 3 min can improve the accuracy of data reported, and when water temperatures are <10 °C, warming to >11 °C could also deliver improvement in both La Motte test strip and Hanna Checker colorimeter assessments.

There are other low-cost methods available for these parameters, although we note that the range covered by the Hach test strip for phosphate is generally too wide for most river water quality monitoring applications (Muenich et al., 2016). Novel low-cost colorimetric tests, similar to test strips, are under development (Richardson et al., 2021; Cinti et al., 2016; Jayawardane et al., 2012), which may have utility in river water monitoring. However, we emphasise that data gathered using low-cost tests have the potential to indicate contamination hotspots or detect critical changes in nutrient concentrations in freshwater across large geographical areas, especially when the quality of data can be assessed. Their value lies in volunteers’ ability to use these low-cost tests to acquire large quantities of high spatial and temporal resolution data compared to professional grab sampling approaches. It is essential that volunteer training in best practice is ongoing and that methods are amended if new evidence emerges to suggest improvements can be achieved in a range of real world conditions. This approach will ensure the quality of data that can be achieved with low-cost test kits is optimised and that data gathered by groups using the same tools remains both reliable (within the known limits of test sensitivity and accuracy) and comparable across different volunteer monitoring groups.