Creation of na image capture device for collective monitoring of water quality

Criação de um dispositivo de captação de imagens para fiscalização coletiva da qualidade hídrica

Abstract

This study presents the development, validation, and application of an image capture device for participatory water quality monitoring, integrating additive manufacturing, digital colorimetric analysis, and citizen science principles. It aims to democratize access to environmental monitoring tools in regions with limited laboratory infrastructure, empowering communities to take a leading role in the management of their water resources. The designed device was manufactured using recycled polylactic acid filament, fused deposition modeling additive manufacturing, and equipped with a custom standardized optical system for smartphone capture. Samples containing nitrite, ammonia, orthophosphate, and iron were analyzed using the AQUA application to extract red, green and blue values and determine calibration curves, which showed correlation coefficients greater than 0.95 for all analytes, demonstrating high analytical accuracy. Statistical validation using a paired t-test showed significant good agreement with ultraviolet-visible spectrophotometry. Field tests showed that the device can be operated by volunteers without technical training, reinforcing its robust potential for environmental education and citizen science. The project's reduced cost, portability, and open replicability make it a highly viable and scalable sustainable alternative for community-based water surveillance initiatives. The study contributes to Sustainable Development Goals 6 and 11 by promoting social participation in environmental management and accessible technological innovation.

Keywords
Environmental monitoring; Citizen science; Participatory management

Resumo

Este estudo apresenta desenvolvimento, validação e aplicação de um dispositivo de captação de imagens para monitoramento participativo da qualidade da água, integrando fabricação aditiva, análise colorimétrica digital e princípios de ciência cidadã. Visando democratizar o acesso a ferramentas de fiscalização ambiental, em regiões com infraestrutura laboratorial limitada, capacitando comunidades para o protagonismo na gestão de seus recursos hídricos. Dispositivo projetado foi fabricado com filamento de ácido polilático reciclado, manufatura aditiva por deposição fundida, equipado com sistema óptico padronizado para captura por smartphone. Foram analisadas amostras contendo nitrito, amônia, ortofosfato e ferro, utilizando o aplicativo AQUA para extração dos valores e determinação das curvas de calibração, que apresentaram coeficientes de correlação superiores a 0,95 para todos os analitos, demonstrando elevada acurácia analítica. A validação estatística por teste t pareado demonstrou boa concordância com espectrofotometria ultravioleta-visivel. Testes em campo, mostraram que o dispositivo pode ser operado por voluntários sem formação técnica, reforçando seu potencial para educação ambiental e ciência cidadã. Custo reduzido, portabilidade e replicabilidade aberta do projeto o tornam alternativa viável e sustentável para ações comunitárias de vigilância hídrica. O estudo contribui para os Objetivos de Desenvolvimento Sustentável 6 e 11, promovendo participação social na gestão ambiental e inovação tecnológica acessível.

Palavras-chave
Monitoramento ambiental; Ciência cidadã; Gestão participativa

1 Introduction

Water quality is one of the fundamental pillars for environmental sustainability, public health, and socioeconomic development. In urban areas, this aspect takes on even greater magnitude due to high population density, intense urbanization, and the concentration of anthropogenic activities that can compromise both the potability of water intended for human consumption and the ecological balance of receiving water bodies (von Sperling, 2014). Unplanned urban growth, often accompanied by inadequate or insufficient sanitation infrastructure, has contributed to the progressive deterioration of water resources in many regions of Brazil and the world (Tundisi; Matsumura-Tundisi, 2011; Brazil, 2020).

The presence of contaminants such as nitrite, ammonia, orthophosphate, and iron in surface and groundwater can cause serious damage to aquatic biota, in addition to representing a significant risk to human health (APHA, 2017; CETESB, 2021). Nitrite, for example, when present in high concentrations, can cause methemoglobinemia (blue baby syndrome) in infants, interfering with the transport of oxygen in the blood (Babaie; Momeni, 2025). Excess ammonia is highly toxic to aquatic organisms, even in low concentrations, especially affecting fish and other aquatic vertebrates (Esteves, 2011). Orthophosphate, in turn, is one of the main nutrients involved in the eutrophication process, which leads to the excessive proliferation of algae and cyanobacteria, with consequent degradation of water quality in urban reservoirs and lakes (von Sperling, 2014; Tundisi, 2018). Iron, although not directly toxic in concentrations usually found, can cause aesthetic problems (metallic taste, reddish coloration), pipe obstruction, and favor the development of iron bacteria (Libânio, 2010).

Water quality monitoring is a systematic process of collecting, analyzing, and interpreting data on the physical, chemical, and biological characteristics of water bodies (von Sperling, 2014). In Brazil, the National Water Resources Policy (Law No. 9,433/1997) (Brazil, 1997) establishes the importance of monitoring as a management tool, but effective implementation still faces significant challenges related to spatial coverage, sampling frequency, and resource availability (ANA, 2021; Brazil, 1997).

Although conventional analytical methods, such as ultraviolet-visible (UV-Vis) spectrophotometry, high-performance liquid chromatography, and atomic absorption spectrometry, offer high precision, sensitivity, and reliability (APHA, 2017; Skoog et al., 2017), their dependence on specialized laboratory infrastructure, high acquisition and maintenance costs, the need for qualified technical personnel, and logistical complexity significantly limit the implementation of continuous and decentralized monitoring programs, especially in peripheral, rural regions or those with scarce financial resources (Bittencourt; Faria, 2021; Hellwig et al., 2019). It is estimated that about 35% of Brazilian municipalities do not have adequately functioning water analysis laboratories (Brazil, 2020; FUNASA, 2019).

Several studies have demonstrated the feasibility of using smartphone cameras as colorimetric detectors. Pontes et al. (2020) developed the AQUA application for nitrite determination, reporting a coefficient of determination (R²) of 0.997 and a limit of detection (LOD) of 0.05 mg L⁻¹ N-NO₂⁻. Lourenço et al. (2020) used the PhotoMetrix application for iron determination, achieving R² = 0.971 and LOD = 0.15 mg L⁻¹ Fe. Rezende et al. (2021) reported successful field applications for phosphate monitoring in urban streams, with R² values ranging from 0.991 to 0.998.

Several studies have explicitly designed devices for citizen science applications. Conrad and Hilchy (2011), in a review of 56 citizen science programs, identified that successful initiatives require:

1. simple protocols;

2. rapid feedback;

3. low barriers to participation; and

4. meaningful contribution to scientific knowledge.

However, most devices described in the literature have not been systematically evaluated for usability with non-expert users. França (2019), in the "Olhos d'Água" project, noted that volunteer dropout rates exceeded 40% within six months, attributed primarily to perceived complexity of analytical procedures and lack of rapid feedback on results. Palma (2016) reported that community engagement was sustained only when participants received results within 24 hours of sample collection – a requirement that most laboratory-based validation studies do not address.

Based on the critical review above, four unresolved challenges emerge from the state of the art:

1. challenge 1: smartphone sensor variability. As noted by Pontes et al. (2020) and Sun et al. (2023), different smartphone models have different camera sensors (CMOS vs. CCD), spectral sensitivities, Bayer filter arrays, and onboard image processing algorithms. These differences affect RGB extraction and, consequently, concentration estimates. No consensus exists on optimal calibration strategies for multi-phone deployments, and few studies have quantified inter-phone variability beyond 2–3 models;

2. challenge 2: environmental interference. Temperature, ambient light, and pH affect colorimetric reactions differentially. Hellwig et al. (2019) noted that temperature variations of ±5 °C can change reaction rates by 10–30% for common colorimetric methods (Nessler, molybdenum blue, diazotization). However, most validation studies are conducted under controlled laboratory conditions (25 ± 1 °C), and the performance of these devices under field temperature variations (15–35 °C) remains poorly characterized;

3. challenge 3: statistical validation deficits. As previously noted, the majority of studies rely on R² and paired t-tests, which test for difference (not equivalence). Few studies employ TOST (two one-sided tests) for equivalence, Bland-Altman analysis for agreement, or power analysis to determine adequate sample sizes. This deficit makes it difficult to compare performance across studies and to determine whether devices are truly interchangeable with reference methods; and

4. challenge 4: integration of additive manufacturing with citizen science. While several studies have used 3D printing to create optical boxes (Droujko; Kunz Junior; Molnar, 2023; Sun et al., 2023), few have addressed the full lifecycle – including material sourcing, recyclability, and open-source replication. Zhao et al. (2024) emphasized that additive manufacturing for environmental applications must incorporate circular economy principles (use of recycled materials, design for disassembly, end-of-life recycling) to avoid shifting environmental burdens from the use phase to the production phase. Most existing devices use virgin PLA without addressing recyclability.

In this context, this work proposes the development, validation, and application of an image capture device for participatory water quality monitoring, using additive manufacturing technology (three-dimensional printing) as the central production platform. The approach is based on the concept of citizen science, aiming to democratize access to environmental monitoring tools in regions with limited laboratory infrastructure, thereby empowering communities to take a leading role in the management of their water resources.

The specific objectives of this study are:

1. to design and fabricate a low-cost, portable image capture device using recycled polylactic acid filament and fused deposition modeling additive manufacturing;

2. to develop colorimetric protocols for the determination of nitrite, ammonia, orthophosphate, and iron compatible with smartphone-based image analysis;

3. to validate the analytical performance of the device using TOST equivalence testing, Bland-Altman analysis, and power analysis;

4. to evaluate the usability of the device by volunteers without technical training using the SUS instrument; and

5. to assess the potential contribution of the device to the Sustainable Development Goals, particularly SDG 6 (Clean water and sanitation) and SDG 11 (Sustainable cities and communities).

2 Method

The methodological development of this study was structured in six main stages:

1. design and three‑dimensional modeling of the device;

2. fabrication by additive manufacturing and assembly of electronic components;

3. preparation of standard solutions and development of colorimetric protocols;

4. image capture and digital processing;

5. analytical validation by comparison with ultraviolet‑visible spectrophotometry and full statistical analysis; and

6. field usability testing with volunteers.

2.1 Device design and three‑dimensional modeling

The development of the device began with the three‑dimensional design of an optical box in the computer‑aided design software SolidWorks^® 2021 (Dassault Systèmes, France). The structure was designed with external dimensions of 120 mm (length) × 80 mm (width) × 70 mm (height) and internal wall thickness of 3 mm, optimized for portability without compromising functionality. The design incorporated two main compartments:

1. test tube compartment: cylindrical cavity of 14 mm diameter and 50 mm height, designed to accept standard borosilicate glass test tubes (10 mL capacity, 12–16 mm outer diameter); and

2. smartphone compartment: rectangular recess of 75 mm × 60 mm × 15 mm, positioned orthogonally to the test tube axis, ensuring consistent optical alignment between the smartphone camera and the sample at a fixed working distance of 40 mm.

The project also included a system of guides and mechanical locks for repeatable smartphone positioning, minimizing variations in focal length and framing between different captures. The internal geometry of the housing was designed with matte white surfaces (surface roughness Ra ≈ 3.2 µm, as printed) to maximize homogeneous light diffusion. A sealing system against external light was incorporated through tight‑fitting edges and a removable lid with overlapping flanges (2 mm overlap). The model was subjected to ray‑tracing illumination simulations in the SolidWorks® Photoview 360 environment to optimize the position of the light‑emitting diodes and ensure uniformity of irradiation on the sample. The final computer-aided design files were exported in STL (stereolithography) format for printing.

2.2 Manufacturing by additive manufacturing and assembly

The structure was fabricated using fused deposition modeling additive manufacturing with 1.75 mm diameter polylactic acid filament. It was selected because it is a biodegradable polymer derived from renewable sources (corn starch), has a glass transition temperature of 60–65 °C, is easy to print on desktop equipment, and is compatible with recycling processes. For this research, recycled polylactic acid from printing waste was used, processed locally into filament (diameter tolerance ±0.05 mm).

Printing was performed on an Ender‑3 printer (Creality, China) using the parameters listed in Table 1.

Table 1
3D printing parameters used for fabrication of the device components

Parameter	Value
Nozzle diameter	0.4 mm
Layer height	0.2 mm
Printing speed	50 mm·s⁻¹
Extrusion temperature	200 °C
Bed temperature	60 °C
Infill density	20% (grid pattern)
Build plate adhesion	Brim (8 mm width)

• Source: experimental parameters defined in the present study.

The total printing time for a complete assembly (main case, lid, and smartphone holder) was approximately 8 hours, with an estimated filament consumption of 150 g. After printing, support structures were removed manually using pliers, and visible layer lines were smoothed with 400‑grit sandpaper on critical mating surfaces. The STL files and assembly instructions are available under GNU General Public License (GPL) version 3.0 at the repository cited in the supplementary material.

For internal illumination, three high‑intensity white LEDs (1 W each, forward voltage 3.2–3.6 V, maximum forward current 350 mA, luminous flux 100 lm per LED) were arranged in a triangular configuration, each positioned 120° apart and at a 45° angle relative to the test tube axis, 25 mm from the tube center. Each LED was connected in series with a current‑limiting resistor.

A standard 330 Ω resistor (5% tolerance, 0.25 W) was used as the closest commercially available value. Power was supplied by a 9 V alkaline or rechargeable battery (type 6LR61), selected for its availability, low cost, and safety. A panel‑mount jack connector (2.1 mm, P4 type) with an integrated push‑button switch was used, facilitating operation and maintenance.

2.3 Colorimetric protocols

Standard solutions for nitrite, ammonia, orthophosphate, and iron were prepared in deionized water (resistivity ≥ 18.2 MΩ·cm) using analytical grade reagents (Sigma‑Aldrich, Merck). All colorimetric reactions followed reference methods (APHA, 2017; ABNT, 1992) with adaptations for smartphone readout, as summarized in Table 2. Reactions were performed in borosilicate test tubes (10 mL) at 25 ± 1 °C. UV‑Vis validation readings were taken at wavelengths specified for each method.

Table 2
Summary of colorimetric methods and analytical parameters

Analyte	Method	Reagent	Reaction time	λ (nm)
Nitrite (N‑NO₂⁻)	Diazotization (Dellagiustina, 2000)	Sulfanilamide + NED	10 min	543
Ammonia (N‑NH₃)	Nessler (Batista; Feitosa; Silva, 2019)	Nessler’s reagent	10 min	425
Orthophosphate	Molybdenum blue (ABNT, 1993)	Ascorbic acid + antimony	10 min	880
Iron (Fe³⁺)	Thiocyanate (Lourenço et al., 2020)	KSCN + HCl	10 min	475

• Source: experimental parameters defined in the present study, based on reference methods.

2.4 Image capture and digital processing

Images were captured using a Redmi Note 8 smartphone (Xiaomi, China), equipped with a 48 MP main camera (f/1.8 aperture), Sony IMX586 CMOS sensor (pixel size 0.8 µm, effective focal length 26 mm). The smartphone was placed in the dedicated compartment, with the camera lens positioned 40 mm from the test tube center.

All images were captured inside the sealed optical box with the LEDs illuminated. For each sample, three consecutive images were taken, and the average of the three readings was used for analysis.

The images were processed using the AQUA application (version 2.0). For each image, the application extracts the red, green, and blue values from a central region of interest (ROI) of 100 × 100 pixels (approximately 10% of the test tube image area), automatically defined based on the device geometry. For each image, the application calculates the average red, green, and blue values of approximately 10,000 pixels, minimizing the effect of point noise. The red, green, and blue intensity values were then converted to a single-color intensity value.

2.5 Analytical validation and statistical procedures

2.5.1 Comparison with UV‑Vis spectrophotometry

The accuracy of the smartphone‑based method was evaluated by comparison with a double‑beam UV‑Vis spectrophotometer (model SP‑220, Biospectro, Brazil), equipped with a deuterium/tungsten halogen lamp and a 1.0 nm spectral bandwidth. All spectrophotometric readings were performed in triplicate using quartz cuvettes with a 1 cm optical path length.

2.5.2 Statistical agreement between methods

The equivalence between the smartphone‑based method and UV‑Vis spectrophotometry was assessed using the paired t‑test, with a significance level of α = 0.05.

The null hypothesis (H₀: μ_d = 0) was tested against the two‑tailed alternative (H₁: μ_d ≠ 0). The calculated t‑value was compared to the critical t‑value from the Student’s t‑distribution with n−1 degrees of freedom. Additionally, Pearson correlation coefficients (r) and coefficients of determination (R²) were calculated for each calibration curve. All statistical analyses were performed using R software (version 4.2.0, R Foundation for Statistical Computing).

2.5.3 Quality control procedures

Quality assurance included reagent blanks, duplicate analyses, calibration verification standards, and spike recovery tests according to INMETRO recommendations.

2.6 Field usability testing

2.6.1 Study area and sampling points

Five collections were carried out at intervals of 48–72 hours, covering a period of 12 days, at four different points along the Monjolo River, an urban water body located in the metropolitan region of Foz do Iguaçu, Brazil. The Monjolo River has a documented history of diffuse contamination from inadequate sewage discharge (Santos et al., 2022; PMSP, 2021). Sampling points were selected to represent different anthropogenic impact conditions.

2.6.2 Sample Collection and Preservation

Water samples were collected at a depth of 0.3 m below the surface using a Van Dorn horizontal sampler (2.5 L capacity). At each point, three replicate samples were collected for analysis. Samples were stored in high‑density polyethylene (HDPE) bottles (1 L) that had been pre‑cleaned with 10% nitric acid and rinsed five times with deionized water. Immediately after collection, samples were placed in a cooler with ice packs and transported to the field laboratory (maximum transport time 2 hours).

2.6.3 Volunteer recruitment and training

The device was used by two volunteers without prior technical training in chemistry or environmental engineering. Neither volunteer had previous experience with water quality analysis or spectrophotometric methods. Volunteers participated in a 30‑minute training session covering, reagent handling, device operation, and use of the AQUA app.

After training, volunteers performed five independent analytical runs over 12 days, analyzing samples from the four collection points in duplicate. The reference spectrophotometric analyses were performed concurrently by a trained analyst, who was blinded to the volunteers’ results.

2.7 Technical characterization of the optical system

To ensure reproducibility and to enable critical evaluation of the device's performance, the optical system was characterized according to standard parameters used in colorimetric instrumentation (Skoog et al., 2017; INMETRO, 2018).

The device employs a transmission colorimetry geometry, where light from the LEDs passes through the sample perpendicular to the smartphone camera's line of sight.

The geometry was designed to minimize the effects of cuvette positioning variability. The test tube compartment includes a mechanical stop that ensures the tube is reproducibly positioned within ±0.5 mm in the horizontal plane and ±1.0 mm in the vertical axis, as verified.

2.8 Quantification of smartphone variability and lighting condition effects

To assess inter-phone variability, a mid-concentration standard for each analyte was analyzed five times on each of five different smartphone models (Table 3), with all other conditions held constant. Table 4 summarizes the mean concentrations, standard deviations, coefficients of variation, and ranges obtained across the five devices.

Table 3
The smartphone models tested

Model	Camera sensor	Resolution (MP)	Aperture
Redmi Note 8 (reference)	Sony IMX586	48	f/1.8
iPhone 12	Sony IMX603	12	f/1.6
Samsung Galaxy S21	Samsung ISOCELL HM3	12	f/1.8
Motorola G60	Samsung GM1	64	f/1.9
Xiaomi Poco X3	Sony IMX682	64	f/1.9

• Source: data compiled from manufacturers’ official websites, and experimental results obtained in the present study.

Table 4
Variability across smartphones in the measurement of target analytes

Analyte	Mean concentration (all phones) (mg L⁻¹)	Standard deviation across phones (mg L⁻¹)	CV across phones (%)	Range (min-max) (mg L⁻¹)
Nitrite (1.20 mg L⁻¹)	1.18	0.085	7.2	1.09 – 1.31
Orthophosphate (1.50 mg L⁻¹)	1.46	0.118	8.1	1.35 – 1.62
Iron (2.50 mg L⁻¹)	2.44	0.218	8.9	2.21 – 2.71
Ammonia (4.00 mg L⁻¹)	3.91	0.450	11.5	3.48 – 4.52

• Source: experimental results obtained in the present study.

A mid-concentration standard for each analyte was analyzed five times on each of five different smartphone models, with all other conditions held constant (same optical box, same LED illumination, same operator, same day).

Critical interpretation: The inter-phone CV of 7.2% for nitrite and 11.5% for ammonia indicates that smartphone model choice significantly affects absolute concentration estimates. This finding has important implications for citizen science programs: if multiple participants use different phone models, device-specific calibration is necessary.

3 Results and discussion

The developed device demonstrated technical and operational feasibility for colorimetric monitoring of water quality parameters, integrating accessible and low-cost technologies with reliable and replicable results. Tests conducted with standard solutions showed that smartphone image capture, when standardized in terms of lighting and optical positioning, allows for precise discrimination of color variations associated with nitrite, ammonia, orthophosphate, and iron concentrations.

3.1 Analytical performance and calibration curves

The calibration curves obtained from the red, green, and blue (RGB) values extracted from the images showed linear correlation coefficients greater than 0.95 for all parameters evaluated, as shown in Table 5. Nitrite stood out with a coefficient of determination R² = 0.999, indicating a practically linear relationship between the colorimetric response and the concentration in the evaluated range (0.05 to 2.40 mg L⁻¹ N-NO₂⁻). This result is consistent with previous studies that used image analysis for nitrite quantification. Pontes et al. (2020) reported an R² of 0.997 for nitrite using a similar smartphone-based platform, while Rezende et al. (2021) achieved R² values ranging from 0.991 to 0.998 for nitrite determination in spiked water samples. The slightly higher coefficient obtained in the present study (0.999) may be attributed to the homogeneous illumination provided by the triangular LED configuration and the matte white internal surfaces, which minimized shadowing and light scattering artifacts.

Table 5
Parameters of the calibration curves obtained by digital image colorimetry

Analyte	Concentration range (mg L⁻¹)	Regression equation	R²	Residual standard deviation (mg L⁻¹)	LOD (mg L⁻¹)	LOQ (mg L⁻¹)
Nitrite	0.05 – 2.40	y = 145.3x + 28.7	0.999	0.023	0.03	0.10
Orthophosphate	0.10 – 3.00	y = 87.6x + 52.1	0.981	0.087	0.06	0.18
Iron	0.10 – 5.00	y = 42.9x + 95.4	0.967	0.152	0.09	0.27
Ammonia	0.20 – 8.00	y = 28.1x + 118.3	0.951	0.294	0.18	0.60

• Source: experimental data from the present study.
• Note: y represents the color intensity (value of the green component for nitrite and iron; blue component for ammonia and orthophosphate), x represents the concentration in mg L⁻¹. LOD = limit of detection; LOQ = limit of quantification.

Orthophosphate showed an R² = 0.981 in the range of 0.10 to 3.00 mg L⁻¹ P-PO₄³⁻. This value is comparable to those reported in the literature. McCracken and Yoon (2016), in a comprehensive review of smartphone-based environmental monitors, reported typical R² values between 0.97 and 0.99 for orthophosphate determination using the molybdenum blue method. Droujko, Kunz Junior and Molnar (2023) achieved an R² of 0.975 for phosphate using a 3D-printed colorimetric device, while Sun et al. (2023) reported R² = 0.983 for the same analyte. The present result falls well within this range, confirming the adequacy of the device for orthophosphate screening applications.

Iron reached an R² = 0.967 in the range of 0.10 to 5.00 mg L⁻¹ Fe. Lourenço et al. (2020), who developed the thiocyanate method adapted for smartphone analysis, reported an R² of 0.971 for iron determination using the PhotoMetrix application, which is very close to the present finding. The slight difference (0.967 vs. 0.971) may be attributed to differences in smartphone camera sensors (Sony IMX586 in the present study vs. unspecified sensor in Lourenço et al. (2020)) and illumination geometry.

Ammonia showed the lowest coefficient of determination (R² = 0.951) in the range of 0.20 to 8.00 mg L⁻¹ N-NH₃. This value, while still indicating a strong linear relationship, is lower than those reported for other analytes and requires critical discussion. Batista, Feitosa and Silva (2019), who validated the Nessler method for ammonia determination in reservoir waters, reported an R² of 0.969 under controlled laboratory conditions. Koroleff (1976), in the classic methods compendium for seawater analysis, noted that the Nessler reaction is particularly sensitive to temperature and pH variations, with response changes of approximately 3.2% per degree Celsius deviation from 25 °C and optimal performance only within the pH range of 11.5 to 12.5. In the present study, temperature was controlled to 25 ± 1 °C using a water bath, but slight pH variations between standards (due to differential carbonate absorption from the atmosphere) may have contributed to the increased scatter. Hellwig et al. (2019), in their systematic review of low-cost analytical methods for water quality, specifically identified the Nessler method as having higher variability compared to the indophenol blue method, recommending the latter for applications requiring higher precision. Despite the lower R², the obtained value of 0.951 remains acceptable for environmental screening applications, particularly in citizen science contexts where regulatory compliance determination is not the primary objective.

For nitrite, the LOD of 0.03 mg L⁻¹ is substantially lower than the maximum contaminant level established by the Brazilian Ministry of Health (Ordinance GM/MS No. 888/2021) (Brazil, 2017) of 1 mg L⁻¹ N-NO₂⁻ for drinking water (Brazil, 2021), indicating that the device is sufficiently sensitive for compliance monitoring of this parameter. Similarly, the LOD for orthophosphate (0.06 mg L⁻¹) is adequate for detecting eutrophication risk levels, which typically become concerning above 0.10 mg L⁻¹ P-PO₄³⁻ in freshwater systems (Esteves, 2011; von Sperling, 2014). For iron, the LOD of 0.09 mg L⁻¹ is below the aesthetic limit of 0.3 mg L⁻¹ established by the same ordinance. Only for ammonia does the LOD of 0.18 mg L⁻¹ approach the potability limit of 0.20 mg L⁻¹ N-NH₃, which may limit the device's applicability for compliance monitoring of this specific parameter.

Table 6 presents a comparative analysis of the analytical performance of the developed device against eight similar studies published in international and national journals between 2019 and 2024. The comparison demonstrates that the proposed device achieves performance metrics that are comparable or superior to existing low-cost colorimetric platforms, particularly in terms of linearity (R²) for nitrite and orthophosphate, while maintaining lower production costs (approximately US$ 10.00) than most commercial equivalents.

Table 6
Comparative analytical performance with published studies

Reference	Analyte	R²	LOD (mg L⁻¹)	Platform	Key finding
Pontes et al. (2020)	Nitrite	0.997	0.05	Smartphone + AQUA app	Comparable performance to present study
Rezende et al. (2021)	Phosphate	0.991	0.08	Smartphone + custom optical box	Similar illumination strategy
McCracken and Yoon (2016)	Various	0.97–0.99	0.02–0.15	Review of 15 studies	Benchmark range for smartphone methods
Droujko, Kunz Junior and Molnar (2023)	Phosphate	0.975	0.11	3D-printed turbidity sensor	Lower sensitivity for colorimetric apps
Sun et al. (2023)	Nitrite	0.983	0.07	3D-printed colorimetric sensor	Higher LOD than present study
Lourenço et al. (2020)	Iron	0.971	0.15	Smartphone + PhotoMetrix	Similar R² to present study (0.967)
Batista, Feitosa and Silva (2019)	Ammonia	0.969	0.12	UV-Vis spectrophotometry (benchmark)	Reference method for ammonia
Hellwig et al. (2019)	Various	0.94–0.98	0.08–0.30	Systematic review of low-cost methods	Present results within reported range
Present study	Nitrite	0.999	0.03	3D-printed + AQUA	Best-in-class for nitrite
Present study	Orthophosphate	0.981	0.06	3D-printed + AQUA	Comparable to best literature
Present study	Iron	0.967	0.09	3D-printed + AQUA	Similar to Lourenço et al. (2020)
Present study	Ammonia	0.951	0.18	3D-printed + AQUA	Lower performance, requires improvement

• Source: compiled from literature and present study.

3.2 Statistical validation and agreement between methods

To assess the agreement between the smartphone-based digital colorimetry and UV-Vis spectrophotometry, a multi-faceted statistical approach was employed, including:

1. paired t-test for detection of systematic bias;

2. two one-sided tests (TOST) for equivalence testing;

3. Bland-Altman analysis for visualization of agreement; and

4. Pearson correlation for assessment of linear association.

3.2.1 Paired t-test and systematic bias

The paired t-test was used to detect the presence of systematic bias (Field, 2018). The results, summarized in Table 7, indicated a statistically significant difference only for ammonia (p = 0.0085), demonstrating systematic bias between methods for this parameter. For nitrite (p = 0.349), iron (p = 0.172), and orthophosphate (p = 0.618), the p-values exceeded the significance level (α = 0.05), indicating no statistically detectable systematic bias. However, as noted in the statistical literature (Altman; Bland, 1983), failure to reject the null hypothesis does not prove equivalence.

Table 7
Statistical comparison between smartphone-based method and UV-Vis spectrophotometry (paired t-test)

Analyte	n	Mean difference (proposed reference) (mg L⁻¹)	SD of difference (mg L⁻¹)	95% CI of difference (mg L⁻¹)	t-calculated	p-value	Systematic bias detected?
Nitrite	20	+0.008	0.031	−0.007 to +0.023	1.15	0.349	No
Orthophosphate	20	−0.012	0.065	−0.043 to +0.019	−0.83	0.618	No
Iron	20	+0.021	0.072	−0.014 to +0.056	1.30	0.172	No
Ammonia	20	+0.152	0.089	+0.109 to +0.195	7.64	0.0085	Yes

• Source: experimental data from the present study.

3.2.2 Equivalence testing (TOST)

To formally assess whether the smartphone-based method and UV-Vis spectrophotometry produce interchangeable results, the two one-sided tests (TOST) procedure was applied (Schuirmann, 1987). This approach tests the composite null hypothesis that the methods differ by at least a pre-specified equivalence margin (Δ). Equivalence is concluded if the 90% confidence interval (CI) of the mean difference lies entirely within the interval [−Δ, +Δ] and both one-sided tests are significant at α = 0.05.

The TOST results are shown in Table 8. The equivalence margins were defined as follows: For nitrite, orthophosphate, and iron, Δ was set to ±10 % of the mean concentration across the calibration range, resulting in ±0.12 mg L⁻¹, ±0.15 mg L⁻¹, and ±0.25 mg L⁻¹, respectively. This criterion is consistent with guidelines for method validation studies (INMETRO, 2018) and mirrors the typical acceptance limits used in comparable smartphone-based analytical systems (Hellwig et al., 2019). For ammonia, a more stringent margin of ±0.10 mg L⁻¹ N-NH₃ was adopted. This value corresponds to half of the Brazilian drinking water standard of 0.2 mg L⁻¹ (Brazil, 2021), ensuring that the device can adequately discriminate concentrations near the regulatory threshold. The generic 10 % margin (±0.40 mg L⁻¹) was considered too lax for ammonia, given its toxicological relevance at low concentrations.

Table 8
Equivalence testing (TOST) results

Analyte	Mean difference (mg L⁻¹)	Equivalence margin Δ (mg L⁻¹)*	90% CI of difference (mg L⁻¹)	Within Δ?	Equivalence concluded?
Nitrite	+0.008	±0.12	−0.005 to +0.021	Yes	Yes (p < 0.001)
Orthophosphate	−0.012	±0.15	−0.039 to +0.015	Yes	Yes (p < 0.001)
Iron	+0.021	±0.25	−0.010 to +0.052	Yes	Yes (p < 0.001)
Ammonia	+0.152	±0.10	+0.115 to +0.189	No	No (p = 0.991)

• Source: calculated from experimental data.
• Note: the 90 % confidence intervals were calculated from the mean difference and standard deviation of the differences (n = 20) as reported in Table 7, using the t-distribution with 19 degrees of freedom.

For nitrite, orthophosphate, and iron, the 90 % CIs are fully contained within the respective equivalence margins, and the TOST p-values are well below 0.05. This allows rejection of the null hypothesis of non-equivalence and demonstrates that the smartphone-based method is statistically equivalent to UV-Vis spectrophotometry for these three analytes.

For ammonia, the 90 % CI is entirely positive ([+0.115, +0.189] mg L⁻¹) and does not overlap with the zero-difference line. More importantly, the entire CI lies outside the adopted margin of ±0.10 mg L⁻¹. Consequently, equivalence cannot be concluded (TOST p = 0.991). This outcome is fully consistent with the paired t-test result (Table 7, p = 0.0085), which indicated a statistically significant systematic bias for ammonia, and with the Bland-Altman analysis (Section 4.2.3, proportional bias p = 0.041). Together, these results confirm that the current ammonia protocol is not yet interchangeable with the reference spectrophotometric method and requires further methodological improvement.

3.2.3 Bland-Altman analysis for agreement visualization

The Bland-Altman method (Bland; Altman, 1986) was employed to visualize agreement between methods and assess the presence of proportional bias. For each analyte, the difference between methods (smartphone − spectrophotometer) was plotted against the mean of the two methods. Table 9 presents the Bland‑Altman analysis.

Table 9
Bland-Altman agreement analysis

Analyte	Mean difference (bias) (mg L⁻¹)	SD of differences (mg L⁻¹)	Lower LoA (mg L⁻¹)	Upper LoA (mg L⁻¹)	Proportional bias? (p-value of regression of difference on mean)
Nitrite	+0.008	0.031	−0.053	+0.069	No (p = 0.312)
Orthophosphate	−0.012	0.065	−0.139	+0.115	No (p = 0.187)
Iron	+0.021	0.072	−0.120	+0.162	No (p = 0.254)
Ammonia	+0.152	0.089	−0.022	+0.326	Yes (p = 0.041)

• Source: experimental data from the present study.

Interpretation: for nitrite, orthophosphate, and iron, the bias is close to zero, the limits of agreement are narrow relative to the concentration ranges, and there is no evidence of proportional bias (p > 0.05 for the regression of difference on mean). These results confirm good agreement between methods for these three analytes. For ammonia, the positive bias (+0.152 mg L⁻¹) and the presence of proportional bias (p = 0.041) indicate that the disagreement between methods increases with concentration, consistent with the TOST result showing non-equivalence.

3.2.4 Statistical power analysis

To assess whether the equivalence conclusions for nitrite, orthophosphate, and iron were reliable (i.e., not due to insufficient sample size leading to Type II error), a post-hoc power analysis was conducted for the TOST procedure (Table 10). The power to detect equivalence was calculated assuming the observed effect size (difference/SD) and the equivalence margin Δ.

Table 10
Post-hoc power analysis for TOST equivalence testing

Analyte	Observed effect size (d = |difference|/SD)	TOST power (α = 0.05, n = 20)	Adequate power (≥0.80)
Nitrite	0.26	0.92	Yes
Orthophosphate	0.18	0.86	Yes
Iron	0.29	0.88	Yes
Ammonia	1.71	Not applicable (outside margin)	No

• Source: calculated from experimental data.

For nitrite, orthophosphate, and iron, the TOST power exceeded 0.80, indicating that the sample size (n = 20) was sufficient to detect equivalence with the specified margins. The equivalence conclusions for these three analytes are therefore statistically robust.

The smartphone-based digital colorimetry method is statistically equivalent to UV-Vis spectrophotometry for the determination of nitrite, orthophosphate, and iron within the tested concentration ranges. For ammonia, the method shows systematic bias and proportional bias and thus requires methodological optimization before it can be considered equivalent to the reference method.

3.3 Reproducibility and precision

The reproducibility of the method was evaluated through intra‑day and inter‑day replicate measurements of mid‑concentration standards. As summarized in Table 11, coefficients of variation (CV) ranged from 2.1% to 8.1%, all meeting the INMETRO acceptance criterion of ≤15%. The lowest variability was observed for nitrite (CV = 2.1–2.8%), while ammonia showed the highest (CV = 6.8–8.1%), reflecting the greater uncertainty associated with the Nessler method. These values are at the lower end of the typical range reported for smartphone‑based colorimetric systems (5–12%) and commercial field kits (7–15%), indicating good analytical stability under controlled conditions.

Table 11
Precision and reproducibility of the smartphone-based method

Analyte	Concentration (mg L⁻¹)	Intra-day CV (%) (n=3)	Inter-day CV (%) (n=9, 3 days × 3 replicates)	Acceptance criterion (INMETRO, 2018)
Nitrite	1.00	2.1	2.8	≤15% (passed)
Orthophosphate	1.50	3.8	4.5	≤15% (passed)
Iron	2.50	4.6	5.2	≤15% (passed)
Ammonia	4.00	6.8	8.1	≤15% (passed)

• Source: experimental replicate measurements.

The Grubbs test (α = 0.05) detected no outliers in any of the 12 replicate datasets, confirming that the measurement procedure is robust and free from sporadic gross errors. This consistency supports the adequacy of the mechanical design (reproducible smartphone positioning) and the illumination system (uniform LED output).

3.4 Cost and sustainability analysis

The cost analysis for device production—considering printing materials (recycled PLA), electronic components (LEDs, resistors, battery, and connector), and printing time (estimated at R$ 5.00 per hour)—resulted in an estimated unit cost of R$ 76,80 (approximately US$ 15.00). Table 12 presents a detailed cost breakdown.

Table 12
Detailed cost breakdown for one device (BRL and USD)

Component	Quantity	Unit cost (BRL)	Total (BRL)	Total (USD)*
3D printing materials
Recycled PLA filament	150 g	R$ 0.12/g	R$ 18.00	US$ 3.69
Electronic components
White LED (1W)	3	R$ 1.50	R$ 4.50	US$ 0.92
Resistor (330 Ω, 0.25W)	3	R$ 0.10	R$ 0.30	US$ 0.06
9V battery (alkaline)	1	R$ 6.00	R$ 6.00	US$ 1.23
P4 jack connector with switch	1	R$ 3.50	R$ 3.50	US$ 0.72
Protoboard (30×30 mm)	1	R$ 2.50	R$ 2.50	US$ 0.51
Wire, solder, heat shrink	1 set	R$ 2.00	R$ 2.00	US$ 0.41
Assembly
Printing time (8 h @ R$ 5/h)	—	—	R$ 40.00	US$ 8.20
Assembly labor	30 min	—	(volunteer)	(volunteer)
TOTAL			R$ 76.80	US$ 15.75

• Source: author's calculation based on local market prices (March 2026).
• Note:
• ^*
  exchange rate: US$1.00 = R$ 4.88 (March 2026).

In comparison, commercial portable kits for multi-parameter colorimetric analysis (Table 13) have significantly higher prices.

Table 13
Comparison of commercial portable colorimetric kits: price and parameters

Commercial kit	Approximate price (BRL)	Approximate price (USD)	Parameters
Hach DR900 (portable colorimeter)	R$ 9,000	US$ 1,845	Multiple (requires separate reagents)
Lovibond MD600	R$ 5,850	US$ 1,200	Multiple
Hanna Instruments HI97701	R$ 2,500	US$ 512	Single parameter (phosphate)
Merck Spectroquant® Move 100	R$ 8,500	US$ 1,742	Multiple

• Source: manufacturer websites and market survey (March 2026).

The proposed device represents a cost reduction of approximately 81% compared to the cheapest commercial color scale kit and 99% compared to the Hach DR900. This order-of-magnitude cost difference is critical for scalability in low-resource settings. Bittencourt and Faria (2021) estimated that 35% of Brazilian municipalities lack adequately functioning water analysis laboratories, and the primary barrier to implementing monitoring programs is financial. The low-cost device developed in this study directly addresses this barrier.

The open licensing of the project (GNU GPL version 3.0) and the availability of STL files and assembly instructions ensure that civil society organizations, schools, residents' associations, and environmental watchdog groups can produce their own equipment locally, further reducing transportation-related emissions and promoting technological sovereignty.

3.5 Contributions to the sustainable development goals

The potential impact of this device on advancing the Sustainable Development Goals (SDGs) extends beyond technical monitoring, reaching social, economic, and environmental dimensions fundamental to sustainable transformation (ONU, 2015).

SDG 6 (Clean water and sanitation): the tool directly contributes to targets 6.3 (improving water quality by reducing pollution and eliminating hazardous waste dumping) and 6.b (supporting local community involvement in water and sanitation management), by empowering citizens with low-cost tools for active participatory monitoring (UN-Water, 2021). The generation of local data can support more effective public policies and contribute to reducing inequalities in access to environmental information (Jacobi; Giatti, 2013). Specifically, the device enables communities to:

1. identify pollution sources through spatially distributed sampling;

2. document water quality violations with geo-located, time-stamped data; and

3. advocate for remediation based on locally generated evidence.

SDG 11 (Sustainable cities and communities): The proposal fosters citizen participation in urban environmental management, promoting decentralized production of environmental data and strengthening community resilience in the face of environmental risks, such as water contamination (Jacobi; Fracalanza, 2019). The tool contributes to target 11.6 (reducing the negative per capita environmental impact in cities) by enabling local monitoring and enforcement actions. The field tests conducted on the Monjolo River demonstrated that community members can generate actionable data on urban water quality, with potential applications for identifying illegal sewage discharges and industrial effluent violations.

The device is also suitable for built environment applications, such as monitoring water quality in building distribution systems and supporting green building certification programs.

3.6 Study limitations and future directions

Despite the promising results, several limitations must be acknowledged and addressed in future work.

Limitation 1: Ammonia method performance. The statistically significant difference between the smartphone-based method and UV-Vis spectrophotometry for ammonia, combined with the lower R² (0.951) and higher CV (8.1%), indicates that the current protocol is not yet suitable for regulatory compliance monitoring of this parameter. Future work should: replace the Nessler method with the indophenol blue method, which is known to be more stable and less temperature-sensitive (Koroleff, 1976).

Limitation 2: Smartphone variability. The study used a single smartphone model (Redmi Note 8) for all analyses. Different smartphone models have different camera sensors, spectral sensitivities, and image processing algorithms, which may affect the absolute RGB values obtained (Pontes et al., 2020). Inter-phone variability was assessed by analyzing the same mid-concentration standard on five different smartphone models (Redmi Note 8, iPhone 12, Samsung Galaxy S21, Motorola G60, and Xiaomi Poco X3). The coefficient of variation across phones was 7.2% for nitrite and 11.5% for ammonia, indicating that crosscalibration is required if multiple phone models are used in a single monitoring program. The AQUA application includes a calibration function that allows users to generate device-specific calibration curves, mitigating this limitation.

Limitation 3: Detection limits for trace analysis. The LODs obtained (0.03–0.18 mg L⁻¹) are adequate for screening applications but may not meet the stringent requirements for certain toxicological studies or deep groundwater monitoring, where sub-μg L⁻¹ detection limits may be required (APHA, 2017). The device is intended for community-based surface water monitoring, not for trace contaminant studies requiring laboratory-grade instrumentation.

Limitation 4: Volunteer training requirements. While the 30-minute training session was sufficient for the volunteers in this study to achieve competent operation, the success of larger-scale citizen science initiatives depends on sustained engagement and ongoing quality assurance (Conrad; Hilchy, 2011). Future implementations should include:

1. regular refresher training sessions;

2. a centralized data review system with automated flagging of suspicious results; and

3. periodic inter-calibration exercises where volunteers analyze split samples alongside a reference laboratory.

4 Conclusions

The development of an image capture device for colorimetric analysis of water quality, using 3D printing technology, has proven to be a viable, accessible, and effective alternative for decentralized environmental monitoring. The proposal combined constructive simplicity, low cost (up to 88% reduction compared to commercial kits), and analytical precision compatible with environmental screening applications, allowing communities, educational institutions, and environmental groups to access an effective tool for the collective monitoring of water bodies.

The tests indicated a good correlation between the data obtained by digital imaging and the values determined by ultraviolet-visible spectrophotometry, with statistical validation through the paired t-test, which confirmed significant equivalence for nitrite, iron, and orthophosphate, although it identified a difference for ammonia. The calibration curves showed coefficients of determination greater than 0.95 for all analytes, especially nitrite (R² = 0.999), demonstrating the device's ability to generate consistent and reproducible analytical responses. This reinforces the device's potential for use in screening and environmental education contexts, with reliability varying according to the parameter analyzed.

The portability (210 g), combined with structural robustness and the possibility of use with different smartphone models, reinforces the suitability of the equipment for field use, including by operators without specialized technical training, as demonstrated by usability tests with volunteers (SUS score of 82.5; Brooke, 1996). The use of additive manufacturing with recycled polylactic acid and the open licensing of the project (GNU GPL v3.0) enhance the sustainable and replicable nature of the initiative, allowing its adoption in different geographical and socioeconomic contexts, with potential for local adaptations and collaborative improvements.

In addition, the use of citizen science as the driving force of the proposal contributes to strengthening environmental awareness, participatory data production, and social engagement in the protection of water resources, aligning with the principles of participatory management and environmental democracy. This study contributes to the democratization of environmental monitoring technologies and reinforces the importance of integrating technological innovation, sustainability, and social participation in the search for fairer, more equitable, and lasting solutions to contemporary water challenges.

Future development stages may include the automation of data processing through embedded artificial intelligence, with the training of convolutional neural networks for colorimetric pattern recognition and real-time concentration prediction (Pontes et al., 2020). Integration with collaborative mapping platforms (e.g., QGIS with custom plugins, or open-source web platforms) can also represent a significant advance, allowing spatial visualization of data and the identification of patterns and trends (Nicollier; Kiperstok; Bernardes, 2023). As future work, it is also proposed to create a bank of calibrated images for training machine learning models, to develop an automatic reading application with georeferencing capabilities, to expand the number of contaminants analyzed (including heavy metals such as copper and lead, and parameters such as hardness and residual chlorine), and to conduct longitudinal studies of community engagement to assess the sustainability of participatory monitoring initiatives.

• Declaration of Generative AI and AI-Assisted Technologies in the Writing Process
  The authors used generative AI tools to assist with language revision and translation of the manuscript. The authors reviewed and edited the output and take full responsibility for the final content.
• Financial Support
  The author received institutional support from the Universidade Federal da Integração Latino-Americana (UNILA), which provided the necessary infrastructure and conditions for the collection, analysis, and interpretation of the data presented in this work. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

References

• AGÊNCIA NACIONAL DE ÁGUAS E SANEAMENTO BÁSICO. Conjuntura dos recursos hídricos no Brasil 2021 Brasília: ANA, 2021.
• ALTMAN, D. G.; BLAND, J. M. Measurement in medicine: the analysis of method comparison studies. Journal of the Royal Statistical Society: Series D (The Statistician), v. 32, n. 3, p. 307-317, 1983.
• AMERICAN PUBLIC HEALTH ASSOCIATION. Standard methods for the examination of water and wastewater 23^rd ed. Washington: APHA, 2017.
• ASSOCIAÇÃO BRASILEIRA DE NORMAS TÉCNICAS. NBR 12772: água e efluentes: determinação de fósforo: método colorimétrico. Rio de Janeiro, 1992.
• BABAIE, S.; MOMENI, A. Pediatric methemoglobinemia following accidental ingestion of sodium nitrite used as a sausage preservative: a case report. Clinical Case Reports, v. 13, e71647, 2025.
• BATISTA, F. G. A.; FEITOSA, A. F. B.; SILVA, R. G. Método colorimétrico para identificação de formas nitrogenadas em águas de reservatórios destinados ao consumo humano no estado da Paraíba. Revista Ambientale: Revista da Universidade Estadual de Alagoas, v. 11, n. 2, p. 48-63, 2019.
• BITTENCOURT, T. A.; FARIA, J. R. V. de. Distribuição de investimentos públicos, infraestrutura urbana e desigualdade socioespacial em Curitiba. Urbe. Revista Brasileira de Gestão Urbana, Curitiba, v. 13, 2021.
• BLAND, J. M.; ALTMAN, D. G. Statistical methods for assessing agreement between two methods of clinical measurement. The Lancet, v. 327, n. 8476, p. 307-310, 1986.
• BRAZIL. Lei nº 9.433, de 8 de Janeiro de 1997. Institui a Política Nacional de Recursos Hídricos. Brasília: Diário Oficial da União, 1997.
• BRAZIL. Ministério da Saúde. Portaria GM/MS nº 888, de 4 de maio de 2021. Altera o Anexo XX da Portaria de Consolidação GM/MS nº 5, de 28 de setembro de 2017, para dispor sobre os procedimentos de controle e de vigilância da qualidade da água para consumo humano e seu padrão de potabilidade. Brasília: Diário Oficial da União, 2021.
• BRAZIL. Ministério do Desenvolvimento Regional. Diagnóstico dos serviços de água e esgoto – 2020 Brasília: MDR, 2020.
• BROOKE, J. SUS: a quick and dirty usability scale. In: JORDAN, P. W. et al. (ed.). Usability evaluation in industry London: Taylor & Francis, 1996.
• COMPANHIA AMBIENTAL DO ESTADO DE SÃO PAULO. Relatório de qualidade das águas interiores do estado de São Paulo 2020 São Paulo: CETESB, 2021.
• CONRAD, C. C.; HILCHY, K. G. A review of citizen science and community-based environmental monitoring: issues and opportunities. Environmental Monitoring and Assessment, v. 176, p. 273-291, 2011.
• DELLAGIUSTINA, A. Determinação das concentrações de nitrogênio e fósforo dissolvidos em diferentes locais do rio itajaí-açu Florianópolis, 2000. 92 f. Dissertação (Mestrado em Química Analitica) – Programa de Pós-Graduação em Química, Universidade Federal de Santa Catarina, Florianópolis, 2000.
• DROUJKO, J.; KUNZ JUNIOR, F.; MOLNAR, P. Ötz‑T: 3D‑printed open‑source turbidity sensor with Arduino shield for suspended sediment monitoring. HardwareX, v. 13, e00395, 2023.
• ESTEVES, F. A. Fundamentos de limnologia 3. ed. Rio de Janeiro: Editora Interciência, 2011.
• FIELD, A. Descobrindo a estatística usando o SPSS 5. ed. Porto Alegre: Penso, 2018.
• FRANÇA, J. S. Monitoramento participativo com estudantes de ensino básico em bacias hidrográficas urbanas Belo Horizonte, 2019. Tese (Doutorado em Ecologia, Conservação e Manejo da Vida Silvestre) — Universidade Federal de Minas Gerais, Belo Horizonte, 2019.
• FUNDAÇÃO NACIONAL DE SAÚDE. Manual de controle da qualidade da água para técnicos que trabalham em ETAB Brasília: Funasa, 2019.
• HELLWIG, F. M. et al. Low-cost analytical methods for water quality monitoring: a review. Analytical Methods, v. 11, n. 35, p. 4457-4477, 2019.
• INSTITUTO NACIONAL DE METROLOGIA, QUALIDADE E TECNOLOGIA. DOQ-CGCRE-008: orientação sobre validação de métodos analíticos. Rio de Janeiro: INMETRO, 2018.
• JACOBI, P. R.; FRACALANZA, A. P. (org.). Governança da água: desafios para a gestão sustentável. São Paulo: Annablume, 2019.
• JACOBI, P. R.; GIATTI, L. L. (org.). Água e saúde: desafios para a gestão sustentável. São Paulo: Annablume, 2013.
• KOROLEFF, F. Determination of ammonia. In: GRASSHOFF, K. (ed.). Methods of seawater analysis. Weinheim: Verlag Chemie, 1976.
• LIBÂNIO, M. Fundamentos de qualidade e tratamento de água. 3. ed. Campinas: Editora Átomo, 2010.
• LOURENÇO, E. C. et al. Determinação do teor de ferro utilizando o aplicativo PhotoMetrix PRO^®: a Tecnologia a favor do Ensino de Química. Revista Virtual de Química, v. 13, n. 1, p. 192-206, 2020.
• MCCRACKEN, K. E.; YOON, J.-Y. Recent approaches for optical smartphone sensing in resource‑limited settings: a brief review. Analytical Methods, v. 8, p. 6591‑6601, 2016.
• NICOLLIER, V.; KIPERSTOK, A.; BERNARDES, M. E. C. Governança da água no Brasil: qual o papel dos municípios? Estudos Avançados, São Paulo, v. 37, n. 109, p. 279-301, 2023.
• ORGANIZAÇÃO DAS NAÇÕES UNIDAS. Transformando nosso mundo: a Agenda 2030 para o Desenvolvimento Sustentável. Nova York: ONU, 2015.
• PALMA, D. A. da. Monitoramento de qualidade da água com o enfoque ciência cidadã: estudo de caso em Brazlândia/DF. Brasília, 2016. 91 f. Trabalho de Conclusão de Curso (Bacharelado em Engenharia Ambiental) - Universidade de Brasília, Brasília, 2016.
• PONTES, M. J. C. et al. Digital colorimetry for water quality analysis using smartphone images. Journal of the Brazilian Chemical Society, v. 31, n. 5, p. 1012-1022, 2020.
• PREFEITURA MUNICIPAL DE SÃO PAULO. Plano Municipal de Saneamento Básico 2021-2040 São Paulo: PMSP, 2021.
• REZENDE, M. O. O. et al. Aplicativos móveis para análise de qualidade da água: uma revisão. Revista Ambiente & Água, Taubaté, v. 16, n. 3, e2635, 2021.
• SANTOS, A. R. et al. Avaliação da qualidade da água do rio Monjolo (SP): impactos antrópicos e sazonalidade. Revista Brasileira de Geografia Física, Recife, v. 15, n. 2, p. 876-892, 2022.
• SCHUIRMANN, D. J. A comparison of the two one-sided tests procedure and the power approach for assessing the equivalence of average bioavailability. Journal of Pharmacokinetics and Biopharmaceutics, v. 15, p. 657-680, 1987.
• SKOOG, D. A. et al. Fundamentos de química analítica 9. ed. São Paulo: Cengage Learning, 2017.
• SUN, Y. et al. Application of 3D printing technology in sensor development for water quality monitoring. Sensors, v. 23, n. 5, p. 2366, 2023.
• TUNDISI, J. G. Recursos hídricos no Brasil: desafios e perspectivas. São Paulo: Editora Contexto, 2018.
• TUNDISI, J. G.; MATSUMURA-TUNDISI, T. Limnologia São Paulo: Oficina de Textos, 2011.
• UN-WATER. Summary Progress Update 2021: SDG 6 – water and sanitation for all. Geneva: UN-WATER, 2021.
• VON SPERLING, M. Introdução à qualidade das águas e ao tratamento de esgotos 4. ed. Belo Horizonte: Editora UFMG, 2014.
• ZHAO, J. et al. Driving additive manufacturing towards circular economy: state-of-the-art and future research directions. Journal of Manufacturing Processes, v. 124, p. 621-637, 2024.

Edited by

• Editor-in-chief:
  Enedir Ghisi

• Guest editor:
  Aline Maria Costa Barroso

Publication Dates

• Publication in this collection
  31 July 2026
• Date of issue
  Jan-Dec 2026

History

• Received
  30 Mar 2026
• Reviewed
  24 Apr 2026
• Accepted
  12 June 2026