A low-cost system for calibrating methodologies used to measure GW-SW flow

Exchange fluxes across the groundwater–surface-water interface are critical components of hydrologic, climatic, and ecologic processes. We present details of the construction of a low-cost Controlled Flux Tank Calibration System (CFTCS). The construction of this low-cost ($318) easily replicable system allows for comparative testing of different groundwater flow measurement methodologies. This work presents an illustrated step-by-step construction guide and intends to contribute to the development of studies on the groundwater monitoring in coastal lagoons.


Introduction
Exchange fluxes between groundwater (GW) and surface water (SW) bodies, as well as fresh groundwater storage and its temporal distribution, are critical components of hydrologic, climatic, and ecologic processes (Gleeson et al. 2016;Tirado-Conde et al. 2019;Solomon et al. 2020).
Development of cost-effective and readily deployable methodologies for direct measurement of subsurface hydraulic properties is a current requirement to improve the rate and scale of these properties (Arcari et al. 2019;Mcmillan et al. 2019;Cremeans et al. 2020).Exchange fluxes between GW-SW bodies may be measured directly with seepage meters, for example, used for several decades to quantify exchange between GW-SW in wetlands, ponds, lakes, estuaries, and oceans (Rosenberry et al. 2020).Those provide local-scale measurements and are inexpensive to build and operate (Israelson and Reeve 1944;Lee 1977;Asbury 1990;Murdoch and Kelly 2003;Taniguchi et al. 2008;Rosenberry et al. 2008;Santos et al. 2009;Russoniello et al. 2013;Duque et al. 2018;Arcari et al. 2019;Cremeans et al. 2020).
Low-cost equipment, such as seepage meter, must be calibrated and validated to later be used in the field to monitor the flow between GW-SW.It is necessary to have a controlled flux tank calibration system (CFTCS), easily replicable, in which known infiltration flows between GW-SW are generated.The CFTCS makes it possible to carry out comparative tests and validation of different methodologies for quantifying groundwater flows (Rosenberry This article has been corrected with minor changes.These changes do not impact the academic content of the article.et al. 2020).This highlights the importance of academic articles, guides, and manuals that facilitate the understanding and replicability of low-cost CFTCS.This allows obtaining a correction factor (CF) that can be used to correct the flows measured in the field by different methodologies.This correction is necessary because, due to the pressure losses caused by the flow resistance in the equipment, the measured flow is always less than the real flow.
The first objective of this study develops a reproducible system tank constructed with only low-cost materials easily found on the market.The design was patterned after the Rosenberry and Menheer (2006a) paper and we improved upon it with materials that are far less expensive.It will allow researchers to test and calibrate different methodologies used to measure the flow between GW-SW, in a low-cost system tank.The second objective of this study is to verify the model efficiency and correction factor (CF) of the seepage meter design tested in the tank.

Material and methods
In the present study, the main focus is to develop a low-cost system based on the Rosenberry and Menheer (2006) paper.Overall, the tank setup allows for controlling the exact flow rate through the sediment-water interface, which in turn can be compared to the measured instrumentspecific seepage flow rate.From this, instrument-specific performance criteria can be derived.
Figure 1.Components controlled flux tank calibration system for the seepage meter.The system operates with two pumps to control and recirculate flow.The pumps operate allowing lifting and aspiration.When simulating positive flow (blue arrows), pump 01 (M1) that supplies the flux tank operates in lift mode and pump 02 (M2), which supplies the reservoir, operates in aspiration mode.To simulate negative flow (dotted red arrows) the pumps operate in reverse.To control the flow of the pump, a potentiometer was installed for each pump.The reservoir level is controlled by a counterbalance and float system.The flow in and out of the calibration tank must be the same.This process keeps the hydraulic difference between the reservoir and flux tank to be constant.A valve and a flowmeter were installed before the tank installation.The flowmeter sends real-time flow information to an Arduino datalogger.

Low cost experimental system construction
The CFTCS, where discharge through the sedimentwater interface is simulated, consists of a 2000 l capacity polyethylene water tank and is supplied by a smaller 35 l reservoir.The system was built using Rosenberry and Menheer (2006b) system as a basis, but here the entire system was built only with low-cost materials.The hydraulic system employs an inexpensive low-flow peristaltic pump.The CFTCS design scheme is shown in Figure 1.

Controlled flux tank calibration system construction
The base of the tank was formed by a 1.60 m diameter and 5 mm thick PVC diffuser plate, which had a grid of holes of 4 mm diameter, spaced in a 6 cm vs 5 cm mesh, to allow uniform distribution of water flow through the plate (Figure 2 on the left).The base had supported by PVC boxes, originally designed for electrical applications, used to support the plate and the weight of sand and water that rests on the plate (Figure 2 on the right).
The tank has walls that are not vertical (Figure 1), this creates non-linear flow through the sand, because the area increases with distance above the diffuser plate.The tank walls have an inclination of approximately 15%, such inclination results in a negligible source of error, but this cannot be disregarded.
The diffuser plate was fitted into the bottom of the controlled flux tank and then caulking paste and duct tape were used to seal the edge of the plate to the tank wall to prevent any preferential flow around the edge of the diffuser plate (Figure 3(c)).
To prevent overlying sand from clogging the diffuser plate holes, a Bidim RT 07 Geotextile Blanket was installed over the plate.Bidim RT 07 Geotextile Blanket is a 100% polyester continuous filament non-woven, with 0.212 mm apparent opening used as a blanket to drainage, protection, and reinforcement of the soil.Due to its high permeability (0.4 cm/s) and permittivity (2.5 s −1 ), it allows the free passage of infiltration water to the geotextile blanket, ensuring that the diffuser plate holes do not clog with fine-grain sediment.
The sand layer used in the flux tank has d10 equal to 0.10 mm and d50 equal to 0.19 mm, and this material can be found in building supply store.The flux tank has three layers, first 3 cm of medium gravel, then 40 cm layer of sand and then 40 cm layer of water.Gravel, sand, and water thicknesses have been selected to maximize the sand thickness, ensuring adequate water depth to completely submerge a variety of seepage meter types and sizes (Figure 3).
Figure 3 photographs the components of the CFTCS.
2.1.1.1.Hydraulic system The hydraulic system flux tank operates in two ways.When the flow is positive, the Figure 2. On the left, the project with the dimensions of the diffuser base of the flux tank, positioning of the seepages in relation to the water distribution system.To the right, diffuser base ready with PVC support couplings and water distribution system.Each PVC boxe (47 mm tall), were fixed with a special glue for PVC and positioned so that no holes in the diffuser plate were blocked.The water was distributed through a system composed of 32 mm perforated tubes installed below the diffuser plate.The tube holes were 5 mm in diameter and uniformly distributed every 2 cm.
water is directed by gravity or via a mini-pump connected to flexible plastic (PVC Crystal), with a diameter of 8 mm, the reservoir adjacent to the base of the flux tank.To simulate negative flow the pumps operate in reverse.The 12V mini peristaltic pump (RS-385) is capable of delivering between 1.5-2.0l/min with a maximum lift of 3 m and suction height up to 2 m.The diameter of the tank's infiltration area is 160 cm, which results in an infiltration area of 20.096 cm 2 , thus the pump flow range is equivalent to a discharge of 107 cm/day to 143 cm/day.Water returns from the controlled flux tank to the reservoir through an identical second peristaltic pump.The pumps are connected to a power source through a switch.The general scheme of the circuit is presented in (Figure 4).The simulated flows must be constant.To do this, without needing a second flow sensor, the water levels in both tanks must be kept virtually constant (the level in the smaller diameter reservoir has been changed by only 1.5 mm).The average return flow should be equal to the driven flow from the reservoir to the tank.The most inexpensive peristaltic pumps do not pump at a constant rate but instead the pumping rate slowly increases or decreases with time.To solve this problem, the pump flow was manually regulated by the potentiometer over time in order to keep the upper reservoir level always the same.The reservoir level is controlled by a counterbalance and float system (Figure 3).And time-average the pumping rate during the time a seepage bag was attached to the seepage meter inside the tank.

Eletronic system
For both positive and negative flow, the flow rate between the reservoir and the calibration tank was monitored with a Hall effect flow sensor (OF-201).The rotor, which rotates at a rate proportional to the speed of water flowing through the meter hole, sends an electrical pulse to the data logger (arduino) for each revolution.The Arduino design diagram in conjunction with a hall effect flow sensor and an LCD reader is shown in Figure 5.
Is necessary determining the flow coefficient K to convert the flowmeter pulses to a volumetric flow, with the required accuracy (Equation ( 1)).It may be a particular line for each meter and dependent on the viscosity of the fluid.
where V: Volumetric flow (ml/min); f : Pulse frequency (pulse/min); K: constant k of the flow sensor (ml/pulse); Calibration was performed on a tank-independent system.To carry out the calibration, the sensor, an auxiliary reservoir, hose, graduated cylinder and a stopwatch were used.
The pulses are electrical signals measured by the Hall's effect sensor and sent into the programed Arduino algorithm (the Arduino script can be found in supplements).The data acquisition system adds the pulses at each 60-second increment.This interval minimizes the error and represents significant samples for the calculation of the discharge test.To enable the calculation of the average pulses per minute, each test lasted 4 min, so 4 values of pulses per minute were acquired for the same constant flow rate.To keep the flow constant, the level of the upper reservoir was kept the same during the 4 min of each test.
The linear calibration curve was estimated by the relationship between the actual volumetric flow rate, measure with a graduated cylinder, and the pulses measured by the electrical signals.Three calibrations were performed for the sensor.
First, to determine the operating range of the flow sensor, an analysis of the field-measured discharges was performed to determine the rate of simulated discharges in the tank.Having verified this, a study was performed to verify which flow meters were compatible with the Arduino system and contemplated the desired flow rate.Unfortunately, no sensors were found at an affordable price for flow rates below ∼ 350 ml/min, which is equivalent to a 20 cm/day discharge for the flow-controlled tank watersediment interface area (20.096 cm 2 ).After the theoretical approval of the sensor with the best cost-benefit, it was purchased and calibrated to the required range.

Seepage meter construction
In this work, a specific model of seepage meter was tested, which is also used in the field by the laboratory.To quantify flow rates between sediment-water interface, seepage meters were made based on the methodologies described by (Lee 1977) and (Rosenberry 2008).
The basic concept of a seepage meter is to enclose and isolate an area of the sediment-water surface interface with a cylinder that is open at its base and vented at the top to bag collector (Figure 6).The change in the volume of water in the collection bag over a measured time interval is used to determine the direction and rate of flux between SW-GW (φ).However, due to head losses caused by the resistance of flow in the equipment, the real natural flux (φR) is generally underrepresented.Thus, a correction factor (CF) needs to be applied in the flux measured by seepage meters (φSP).The CF can be defined as the equipment efficiency and is the resultant of the inverse of the φSP/φR ratio.
As convention, the seepage meter flow direction in the sediment-water interface is defined as positive flux when the flow occurs from groundwater (sediment) to surface water (GW-SW).The negative flux is defined when the flow happens from surface water to groundwater (SW-GW).
The components used in the construction of the seepage meters are listed below: • Clear NPS 1 PVC pipe (Figure 6 The bag collector is the most important piece of equipment, as well the largest source of measurement errors.The bag material should be waterproof and resistant but needs to be flexible to enable water flow with low resistance.A polyethylene bag, 5 l of capacity, with 0.6 mm thickness was chosen, with 25 cm × 35 cm dimensions.The bag should be tied to the connecting pipe by a rubber band so the bag is equally distributed around the perimeter of the tube, otherwise will change the measured infiltration rates. The hole where the tube and plastic bag are coupled should be made at the top or the edge of the chamber.When installed, the side of the seepage meter with the hole should be slightly raised, allowing any sediment gases to escape freely.The bag was pre-filling with a known volume of water (1000 ml).The bag and seepage-meter cylinder must remain fully submerged, to ensure that the total hydraulic head in the seepage meter is the same as head at the sediment-water interface.
A seepage meter should be deployed in a location free of vegetation, debris, and large rocks.In addition, the accumulation of organic matter generates the production of gases that must be avoided.The softer the soil where the seepage meter is installed, more it needs to be buried, until proper sealing is achieved.In relatively compact soils, a depth of 10 cm is considered adequate (Woessner 2020).After installing the barrel we wait 24 h to ensure that the sediments are in equilibrium before starting the measurements.

Determination of correction factor and efficiency of seepage meter
Three seepage meters were installed inside the tank.The tank has the capacity to generate known flows through the sediment-water interface ( ∼ 20 cm/day to ∼ 100 cm/day).Each test lasts 15 min, the time necessary to significantly change the volume of the bag collector, for the sampled flow rate, without exceeding the bag's storage limit.The flow intensity in each test was varied randomly.Multiple ways to couple the bag collector to the meter chamber and several ranges of fluxes are some test examples that can be done in the tank.
A simple linear regression with intercept of zero model was used to fit the tank flux and the seepage meter flow.The fit slope was taken to be the efficiency of the seepage meter.And, CF is the inverse of efficiency.
The volume measurement in the collectors was measured before and after the sampling period using a graduated 1.000 ml beaker.Seepage flux in the collectors was where φSP: Seepage flux (cm/day); V: Sampled initial and final volume (ml); T: Sampling period (min); A: Seepage meter area (2.659 cm 2 ).
Nineteen tests were performed on the calibration tank, 10 in the positive upward direction (GW-SW) and 9 in the negative downward direction (SW-GW).The tests included a range of flows in both direction from 24 cm/d to 46 cm/d.

Low-cost experimental system
All the materials used in the construction of the experimental system can be easily found in the Brazilian market.The total cost of the experimental system was US$ 318, which is considered a favorable result.The CFTCS's hydraulic system, which also includes the cost of the two peristaltic pumps (RS-385), had a total cost of US$ 253, with the most expensive part being the 2000 l tank, which cost US$ 121.The complete electronic system, which includes the flowmeter (OF-201), costs US$ 65.More expensive flowmeters can be found on the market, which are able to detect lower flow rates.But, for this study we aimed for the cheapest available system.The seepage meter and collector bag set have a unit value of US$ 11.An overview list of all materials used is presented in Table 1.

Hydraulic and electronic system calibration
The flowmeter was capable of measuring flow rates ranging from 295 to 658 milliliters per minute (ml/min) with high precision.This translated to seepage meter fluxes in the flux tank ranging from 21 to 47 centimeters per day (cm/d) considering the infiltration area of the tank equal to 20.096 cm 2 .The calibration model of flow sensor showed a determination coefficient R 2 greater than 0.99 in all cases (Figure 7).
The measurements were performed in random order.It helps to avoid any patterns or trends that could lead to systematic errors influencing the results, resulting in a more accurate and reliable calibration of the flow sensor.

Correction factor and efficiency
For positive upward direction (Figure 8(a)), the adjusted R 2 found is equal to 0.991, while for negative downward direction (Figure 8(b)) the R 2 is equal to 0.996.In the scatter plot, it is also possible to observe that seepage meter 02 (SP2), for both directions, presents values far from the regression line.SP2 measured volumes for all tests were lower than SP1 and SP3 measured volumes.
There are several possible causes for reduced flow at SP2.It could be an obstruction in some roles attached to the diffuser plane due to a miss positioning of the PVC couplings or miss inclination of the pipe where the water flows.It could also be both former hypotheses mix together.It is also possible that it is a problem with uniform flow through the sand, just as it occurs in a natural environment.(Belanger and Montgomery 1992;Rosenberry and Menheer 2006b;Rosenberry et al. 2020) in his studies found that seepage meter measurements made in both flux tanks indicate that substantial spatial heterogeneity in seepage exists even in homogeneously distributed sand.
The Correction Factor (CF), defined as the equipment efficiency is the resultant of the inverse of the φSP/φR ratio.CFs have been reported in literature from 1.05 to 1.74 (e.g.Asbury 1990;Belanger and Montgomery 1992;Cherkauer and Mcbride 1988;Dorrance 1989;Erickson 1981;Murdoch and Kelly 2003;Rosenberry 2005).According to Rosenberry (2005), the seepage meters inefficiency has declined over time with improvements in the meter design.Rosenberry (2005), obtained a correction factor (CF) for the positive flow of 1.05 through the mean of 26 tests with seepage rates from 5 to 37 cm/day.While Figure 7. Linear regression between the sum of the pulses emitted by the flowmeter and the measured volume (water fluid) in one minute.The has three straight lines, Line 01 (red), Line 02 (blue) and Line 03 (green), with a 95% confidence interval represented by dotted lines.As p value is zero, the null hypothesis is rejected, so the slope is not null for any significance level.The R 2 for both lines are greater than 0.99, that is, 99% of the flow variation is explained by the regression line.
Figure 8. Scatterplot chart of the known tank flow (cm/day) against the flow measured in seepage meters (cm/day).SP1 is represented in blue, SP2 in red, and SP3 in black.The gray area presents the 99% confidence interval of the line.Belanger and Montgomery (1992), found a CF of 1.3 with rates from 9.6 to 69.1 cm/day.The correction factors found were 1.09 for the positive flow (GW-SW) and 1.03 for the negative flow (SW-GW).
Table 2 presents the different regression models analyzed.The first is the general model, which considers data from all seepages installed in the calibration tank, encompassing all simulated flows in both the positive and negative directions.Then, regression models were generated considering only the positive flow direction and only the negative flow direction for the following cases: considering all 3 seepage meters, considering each meter individually, and considering only the seepage meters with greater efficiency, SP1 and SP3.
It is possible to observe (Table 2) that for both negative and positive flow the regression model of the SP2 drum is outside the confidence interval of the models with all seepage meters, showing the difference in the behavior of the drum in relation to the others.However, even with SP2 presenting flow rates lower than the other two seepage meters, the flow values measured in the three seepage meters were kept for statistical analysis and determination of CF in this study.
The global involving both flow directions, was 0.94 ± 0.01 and the CF is equal to 1.06.For the positive flux (GW-SW), overall efficiency is equal to 0.92 ± 0.02, with a 99% confidence interval from 0.87 to 0.97.The CF is equal to 1.09, and confidence limits from 1.03 to 1.14.For the negative flow (SW-GW), overall efficiency is equal to 0.97 ± 0.02, with a 99% confidence interval, from 0.95 to 0.99.The CF is equal to 1.03, and confidence limits from 1.00 to 1.07.Russoniello and Michael (2015) performed experiments in a calibration tank to determine the effects of mechanical resistance on measurement efficiency and occurrence of directional asymmetry in seepage meters.For them, the efficiency of the seepage meter was 0.93.Such information confirms the robustness of the low-cost CFTCS and also of the seepage meters used.
Based on the results, this study considers that the correction of groundwater discharges measured in the field, the CFs are derived from the regression models that consider all seepage meters to be adequate in fine sandy setups.The regression model is considered adequate, as it contains the flow measured under different conditions and has high efficiency.The regression models obtained represent the variations existing in nature due to preferential flows of groundwater discharges, difference in particle size, among other reasons.
In recent decades, a growing number of developed methods have contributed to advancements in our capabilities to identify and quantify exchange between groundwater and surface water.However, there still exist many uncertainties and assumptions that reveal an incomplete understanding of these processes, including the lack of studies in many regions of the world and insufficient sharing of practical methodologies between scientific disciplines (Duque and Rosenberry 2022).Currently, Rosenberry et al. (2020) have discovered that the effectiveness of seepage meters, which were previously believed to have a consistent characteristic determined by their design, actually varies in highly permeable sediments.The study indicates an inverse relationship between seepage meter efficiency and hydraulic conductivity in highly permeable sediments.Furthermore, research shows that automated seepage meters have been essential for both understanding the exchange of fluxes in coastal areas and assessing chemical fluxes and ocean composition (Duque et al. 2020).the cost of these devices has prevented most studies.The CFTCS can facilitate the development of low-cost automatic devices, just as the calibration of seepage meter models for different soil types and flow rates.Enabling a better understanding of subsurface discharge at the sediment-water interface in different regions, such as coastal areas, providing greater reliability to the field equipment used.

Conclusions
The need to join efforts to improve the understanding of groundwater flows in coastal areas is a reality, especially in developing countries.Above all, it is necessary to develop and use low-cost instruments that enable studies that require data collected in the field, due to the current difficulty in acquiring financial resources and the high costs of these types of equipment in the conventional market.But, that instrument biases need to be known in order to derive good estimates on GW-SW interaction, and this can be obtained in laboratory also with low-cost systems.Thus, this study presented a do-it-yourself approach, using lowcost instrumentation, and hopes to encourage researchers to replicate it, facilitating and expanding research that requires data acquisition in the field.
The CFTCS presented was built only with equipment of low cost and easy to acquire in countries with development conditions similar to those in Brazil.The total cost of building the CFTCS was $318.00.The constructive method can be easily replicated directly or with adaptations, as the article presents illustrations and a detailed description of the components used.Its relevance in the scientific community is highlighted, as it enables testing, calibration, and comparison of different methodologies for monitoring groundwater flow.It can be used with punctual measurement methodologies (like seepages meter), and with continuous and automatic measurement methodologies (such as automatic mini piezometers).
This study took advantage of multidisciplinarity to explore several areas, from the use of low-cost flow sensors and arduinos, to the use of materials used in civil construction, for the creation of tools that help hydrological monitoring.Among the difficulties encountered, attention was finding a sensor with adequate precision for the required flow range.As well as prior knowledge of the granulometry of the sediment in the field, so that the CFTCS corresponds to that to be seen in the field.
The CFTCS, built with low-cost flow sensors and arduino, was effective and allowed the determination of correction factors (CF) for seepages, with high efficiency of the models.With the use of seepages and the CF determined in this study, measurements of groundwater flow will be acquired in a coastal lagoon region with anthropogenic influence, in southern Brazil.It is expected that the application of the methodologies developed in this study will help to understand the water balance in ungauged hydrographic basins of coastal lagoons.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Notes on contributor
Priscilla Kern is currently a PhD student at the Graduate Program of Environmental Engineering, Federal University of Santa Catarina (UFSC), Florianopolis, Brazil, at Marine Hydraulics Laboratory (LaHiMar).She has experience in areas of Urban Drainage, Hydrology, Hydrogeology, Hydraulics, Hydrometry, Water Quality, and Instrumentation.The doctoral research seeks to elucidate the relevance of the groundwater component in balance of a coastal lagoon.

Figure 3 .
Figure 3. Photos of the CFTCS project executed.a: Positioning of the seepage meters installed in the flux tank and M1 mini peristaltic pump and adjacent reservoir.b: adjacent reservoir, with the level meter (float and counterweight) and the M2 pump.c: Diffuser base fitted to the bottom of the flow tank, with caulking compound and 'duct tape' tape to seal the rim.d: flow sensor (F) that measures the flow into and out of the flux tank (GW-SW and SW-GW, respectively).

Figure 4 .
Figure 4. General diagram of the pump circuit with the photo of its respective components.To control and reduce the flow of the pump a 50K potentiometer (upper) was installed on 12 V mini peristaltic pump, RS-385, (bottom).For the potentiometer exceed its capacity, a transistor TIP122 (right) was installed to act as an amplifier, this way it was possible to reduce the flow of the pumps to 20 cm/day.
(a)); • Ball valve (PVC NPS 1 ) to attach the chamber to the tube collector, allowing the seepage bag collector to be disconnected without losing water (Figure 6(a)) • Rubber band to attach the bag collector to the tube (Figure 6(b)); • Plastic bag collector (Figure 6(b)) The size of the bag depends on the seepage meter rate and the data sampling period; • Half of 200-liter cylindrical barrel (High-Density Polyethylene), with 58.2 cm internal diameter, 85 cm high (Figure 6(c)); • Flange assembly (PVC), with sealing O-ring (Figure 6(c)).

Figure 5 .
Figure 5. Diagram of the Arduino project in conjunction with a hall effect flow sensor and an LCD reader.In this project, the Arduino UNO R3 is used, which has an 8-bit or 32-bit ATmega328 microcontroller with a clock speed of 16 MHz.The microcontroller can be powered by a laptop with a USB cable or by a DC Adapter or battery from 7 to 12 V and operates with a voltage of 5 V.The necessary connections for communication between the Arduino and the hall effect flow sensor are performed by jumpers.The LCD plate allows the visualization of the flow in real time.calculated by Equation (2):

Figure 6 .
Figure 6.Materials used in the confection of seepage meters.Seepage meter equipment components.a: pipes and valves that compose the connecting pipe; b: connecting pipe with plastic bag and rubber band; c: barrel with flange installed for the connecting pipe coupling (b); d: field installed equipment.

Table 1 .
Complete list of materials used in making the CFTCS (approximate dollar price).

Table 2 .
Regression models without intercepts, between the flow measured by the sensor and flow measured by the seepage meters.The angular coefficient (B) in the regression represents the efficiency (φSP/φR) of the seepage group meters.