Construction and Test of an Instrumented 2D Channel with Rainfall and Insolation Control

This paper presents the construction and testing of a large instrumented 2D channel for the simulation of the performance of compacted barriers under controlled conditions of insolation and rainfall. Details of the main apparatus devices and capabilities and the results of a long-term test performed on a capillary barrier (CB) are presented. The performed test aimed to simulate the behavior of the CB over 1 year in typical semi-arid conditions. The channel behavior was considered very promising. Boundary conditions in terms of rainfall and insolation were adequately properly imposed and suction values, moisture content, runoff and bottom drainage were measured throughout the test. As for the CB performance, the upper clayey layer of soil presented undesirable shrinkage cracks that impacted the CB performance, mainly at the end of the period of evaluation. The obtained results point to the need to use of thicker layers than used in the experiments in order to preserve the integrity of the clayey soil near the interface with the bottom coarse soil layer.

. They are normally composed of a granular soil layer superimposed by a clayey soil. Because of the contrast in their soil water retention curves (SWRC), the granular soil, which exhibits much higher water hydraulic conductivity compared to clayey soil when saturated, presents lower hydraulic conductivity for the suction levels normally obtained in the field (Koerner and Daniel 1997;Khire et al. 2000). The primary function of the clayey soil is to provisionally store the infiltration water, releasing it into the atmosphere in periods of no rain. Therefore, its thickness must consider the rainfall/evaporation patterns of the region of installation and the effective storage capacity of the clayey soil. The CB performance is highly dependent on the weather conditions of the landfill area, and its application must be preceded by studies that consider the weather patterns of each region.
Due to the growing interest in CB in recent years, much research work dealing with the CB soil-atmosphere interactions and modeling can be found. It can be said that the study of capillary barriers started with the work of Yeh et al. (1985), who carried out a stochastic analysis of the unsaturated flow in heterogeneous soils, using laboratory and field observations. The authors concluded that according to the experimental evidence they could take advantage of the existing contrast between the hydraulic properties of the materials (fine/coarse soils). Years later, Ross (1990) and Kämpf and Montenegro (1997) evaluated the performance of several CB on an inclined surface. Until then, the variables that intrigued the researchers were only the hydraulic conductivity functions of the materials involved in the layer and their laying angle. The influence of evaporation and evapotranspiration on the performance of capillary barriers has been studied by several authors (Stormont 1996;Webb 1997;Stormont and Anderson 1999;Khire et al. 2000). The obtained results have helped to consolidate CB as a cover alternative for waste disposal facilities, especially in arid and semi-arid regions. Considering more recent works, the laboratory study of CB performance has gained impetus, as illustrated in the works of Vieira (2005), Tidwell et al. (2003), Oliveira and Marinho (2007), Tami et al. (2004), Silva (2011) and Zhan et al. (2014). However, only few works focus on a 2D flow laboratory approach that considers the influence of the weather conditions on the expected CB performance and attempt to perform tests in environmental conditions that are close to the study area (see Table 1). The use of instrumented 2D flow channels can supply valuable information concerning CB performance under controlled conditions. They enable the visual observation of the experiments and the monitoring of important variables such as the soil moisture content, suction, and temperature. This technical note presents the development and testing of an instrumented 2D channel with inclination control which can simulate rainy events and periods of insolation. The barrier performance is evaluated considering weather data from the semi-arid region of Brazil over 1 year.

Study Area
Although the instrumented channel can be used to evaluate the behavior of distinct types of mineral barriers under different weather conditions, this research focused on the performance of capillary barriers under semi-arid climate conditions. Therefore, the weather conditions imposed to the channel in laboratory were based on surface meteorological measurements collected by INMET (National Meteorological Institute, Brazil) in the period of 2010-2018 in the cities of Barra, Irecê, Remanso and Petrolina. All these towns are located in the Brazilian semi-arid region which comprises a considerable portion of the Brazilian northeast states (see Fig. 1).
In this region medium to small towns predominate in which municipal solid waste is disposed of in dump sites and the presence of landfills is rare. The authors of this study believe that the research results can give impetus to the use of alternative and less expensive cover layers, which could provide adequate environmental safeguards and encourage the installation of landfills in the region.

2-D Flow Channel
The developed channel has the following internal dimensions: 4.0 m long, 0.40 m high, and 0.15 m wide. Figure 2 presents a schematic view of the channel with the instrumentation and the main apparatus used. Moisture content and suction probes were installed at depths of 35 cm (sand layer) and 5 cm, 15 cm and 25 cm (clayey soil layer). The instrumentation is described in Sect. 2.6.1. As indicated in Fig. 2, at the lower end of the channel, two discharge exits measure run-off drainage (ROD) and, eventually, the amount of water that passes through the barrier is collected by the bottom layer drainage (BLD). In addition to allowing different slope configurations, a galvanized steel structure was installed in the channel top so as to simulate different weather conditions with variations in the intensity of solar radiance and rainfall (see Fig. 3). Both systems are discussed in detail later and are also indicated in Fig. 2.

The Rainfall Simulator and the Rainfall Regime
The development of the rainfall simulator (RS) was inspired by two previous studies (Nissen et al. 2000;Vieira 2005). The RS has about 1500 needles, the same channel cross section and an internal height of 10 cm. Rain intensity (I) was calibrated as a function of the water head inside the RS, which varied with the inflow rate. The accumulated rainfall was calculated as the integral of the rain intensity curves (e.g. Fig. 4a, b).
Several experiments were carried out to calibrate the rainfall produced by RS for different combinations of water head and channel water feed duration. Figure 4a, b show the obtained rains for water heads of 2 and 8 mm and water supply time of 1 h. As can be observed, the produced rainfall is characterized by two transient portions located at the beginning and in the final part of the event, with a nearly  Fig. 4, which can be considered as moderated/high to very high intensities according to the semi-arid rainfall pattern (see Table 2 and Fig. 5). In each graph in this figure the presented results correspond to six different tests performed under the same conditions to evaluate the repeatability of the produced rain. The central line represents the mean results while the upper and lower bounds represent their standard deviation.
As the focus of the experiment was to evaluate the CB performance under weather conditions similar to those found in Brazil's northeast semi-arid region, an analysis of the rainfall of the towns cited above was performed. In the INMET dataset, hourly rainfall is available for Barra, Irecê, Remanso and Petrolina. Days with an accumulated rainfall of at least 1 mm were considered rainy days and their data were used in the statistical analysis. Data from all the four towns were treated without distinction as they are in an area with the same climatic characteristics. Furthermore, statistical analysis of the main rainfall parameters proved to be similar, as presented in Fig. 5.
The rainfall data (hourly basis) can also be seen in Table 2. In this case, some intensity intervals of hourly rains are shown along with the percentage of the annual accumulated rainfall they represent. As can be observed, about 75% of the yearly accumulated rainfall is composed of rains with an hourly rainfall of less than 18.2 mm. Table 3 presents the average rainfall for the different months of the year in the period analyzed. Also shown is the expected number of rainy days in each month. In the last four columns of Table 3 the rain distribution adopted for each month following the data presented in Table 2 is detailed. As an illustration, in January a total rainfall of 91 mm is expected in the region, with 8 rainy days. This rainfall was then simulated using 10 hly rainy events of 5 mm (about 50% of the expected rainfall for these months) and one hourly event of 14 mm and 25 mm each, totaling 89 mm. The adopted hourly rainfalls correspond to the middle of the intervals indicated in Table 2 and are marked as I 5 , I 14 , I 25 and I 45 in Table 3. Only one rainy event of 45 mm/h (representing the 37.2-68.2 mm interval) was adopted in the simulations, following the frequency observed in the field for this rain intensity.
Rather than starting on the equivalent first day of the year, the simulations began by considering the lowest rainfall period in the study area, which embraces the months of May-September. Simulations in the channel proceeded until stationary values were reached in the suction and moisture probes. After that, the rainy season, which corresponds to the months of October-April in the field, started to be  simulated. This procedure enabled the researchers to save about two months in the required test duration.

Solar Radiance Simulator
A solar radiance simulator was developed to simulate the incidence of solar radiance on the soil surface. This sought to represent the equivalent average daily radiance of the meteorological stations studied, based on data obtained from INMET. Data for the period 2010-2018 was used to calculate the average daily radiance for each month. To perform the calibration of the solar radiance simulator, the apparatus depicted in Fig. 6 was used. It consists of a rectangular light fixture, a Philips™ HPI-400 W metal-halide lamp, a pyranometer model CMP-3 by Kipp and Zonen, with a sensitivity of 12.43 × 10 −6 V∕Wm −2 , and a device to change the elevation and horizontal position of the pyranometer. Figure 7a presents the solar radiance intensity behavior as a function of the distance lamp/pyranometer. In this case, the measurements took place in the focus of the light fixture. As expected, there is a decrease in the radiance intensity as the distance to the pyranometer increases. Considering an average radiance intensity of 440 W∕m 2 (average radiance intensity for a period of 14hs of insolation in the semi-arid region obtained from the weather stations data), a distance of 27 cm was adopted between the lamps and the CB soil surface. Figure 7b presents the surface distribution of the radiance intensity for a distance lamp/pyranometer of 27 cm. As can be observed, although the occurrence of solar radiance decrease at the edges of the system, the solar radiance distribution was considered to be satisfactorily distributed and the distance between the light fixtures was adjusted to avoid this border problem in the experiments.
Based on the obtained radiance distribution on the surface and in order minimize the occurrence of shadow areas, a horizontal distance between light fixtures of 15 cm was adopted. 8 metal-halide lamps were used in total. As the channel was installed in a closed room, the periods when the lamps were off  were considered equivalent to night periods in the field.

Tests Performed with Temperature and Relative Humidity Controls
Although the daily average temperatures of the experiments (25.6 • C) could be considered close to those in the field in the target area (25.9 • C), the experiments were performed in higher relative humidity (RH) values (71.0-85.5%) compared to field conditions (36.4-69.7%). In order to evaluate the RH influence on the obtained results, a second test is currently running with temperature and RH control. A dual inverter air conditioner and a pair of humidifier/dehumidifiers are used to keep the room weather conditions close to the field conditions day and night. Values of 24 • C ≤ T ≤ 30 • C and 35% ≤ RH ≤ 55% were set for the air-conditioned room during the second test. Therefore, the second test will be performed adopting similar temperatures and a drier atmosphere, which is expected to apply higher suction values on the top surface of the channel, inducing higher values of suction on the suction probes and a sharper decrease in the moisture content values over time compared to the first test. Furthermore, a more intense shrinkage crack pattern is also expected, with the occurrence of cracks in the early stages of the second test.

Capillary Barrier Composition
The upper part of the studied capillary barrier is made up of a residual clayey soil (USCS classification: MH) from metamorphic rocks (granulite/gneiss). The drainage layer of the barrier consists of a uniform fine to medium sand (0.06 − 0.6 mm) (dune sand, USCS classification: SP). Table 4 presents the main characteristics of the soils used whereas their SWRC are presented in Fig. 8. Two methods were used to determine the clayey soil SWRC. For high suctions levels the dew point method was used using the WP4CⓇ apparatus (Meter Group, USA). For low suction values, a small Richard's chamber (Machado and Dourado 2001) was employed, following the methodology proposed by Fourie and Papageorgian (1995).
In the case of the dune sand, continuous vaporization technique was used in conjunction with   Sousa et al. (2011): after saturation, samples were kept exposed to the atmosphere, and the water mass loss and suction increase were monitored. SWRC were obtained by fitting experimental data using Eq. 1 (van Genuchten 1980): where r , s and are the volumetric residual and saturated water contents and the volumetric water content for a given suction, , respectively. is related to air entry suction of the soil and n is a parameter linked to the soil pore size distribution. The physical and hydraulic parameters of the soils are presented in Table 5.

Soil Compaction and Instrumentation of the Capillary Barrier
Due to the very distinct nature of the soils used and to the fact that sheepsfoot rollers are most suitable for fine soils or coarse soils with more than 20% of fines, distinct compaction approaches were adopted for bottom and top layers. The CB bottom layer (dune sand) was compacted using a static strip foot until reaching a dry bulk density of d = 1.72 (1) = r + s − r [1 + ( ) n ] m g cm −3 using the moist tamping technique (1% moisture) (Diambra et al. 2010). After release into the channel, leveled in a loose state, the soil layers with initial heights of about 2 cm were subjected to light pressure until reaching the desired dry bulk density. A final height of about 10 cm was adopted for the bottom sand layer.
The upper clayey layer was compacted using a mini sheepsfoot (Fig. 9b) to achieve a d = 1.41 g cm −3 (maximum dry density of the Normal Proctor energy). First, the soil was released into the channel and leveled to reach about 1 cm in thickness; then the strip foot was used to gently compact the layer for the final compaction using the mini sheepsfoot roller. Figure 9a, b illustrate the compaction process.
The flow channel enables the monitoring of the soil moisture content and suction at different points in the barrier (see Fig. 10). Temperature, electrical conductivity and moisture content were monitored using 5 TE™ moisture sensors (Meter Group, USA) in both soils. Moisture sensors were calibrated specifically for the soils used prior to the tests. Matric suction in the clayey layer was monitored using MPS-6™ (Meter Group, USA) suction sensors, whereas low capacity tensiometers (LCT) were used in the dune sand layer. The MPS-6™ sensors are dielectric water potential sensors for measuring soil water potential and temperature. According to Decagon Devices™ , the accuracy of the probes is about 10% of the reading for the suction range reached in the experiments. The LCT has a pressure range of 0-100 kPa (absolute pressure) and porous stone tips with air entry value of 100 kPa (Model 0604D04-B01M1, Soil moisture™ Equipment Corp.) and were calibrated using water columns in the laboratory (Sousa et al. 2011). In order to avoid porous stone tip desaturation, LCT were removed, saturated and reinserted in the channel on a regular basis and before each rainy event.  Table 5 Fitting parameters of SWRC (Eq. 1) and saturated hydraulic conductivity of the employed soils k s is the soil saturated hydraulic conductivity and R 2 is the determination coefficient of fitting Eq. 1 to the SWRCs

Results and Analysis
The experimental CB performance evaluation started on September 9th, 2020 and after reaching equilibrium conditions with the environment, indicated by constant readings of matric suction and moisture content, the simulation of the rainy period was initiated (first rainy day occurred on day 145, February 9th, 2021).
Concerning the clayey layer behavior during the dry season, the appearance of cracks was observed. This had been expected, however, some cracks extended far enough to reach the bottom sand layer. Figure 11 presents the main cracks observed in the clayey layer. Figure 11a, b were submitted to filters to improve crack visibility. The observed results indicate the need for the use of thicker clayey layers in the field mainly if the occurrence of differential settlements is Fig. 9 Compaction procedures adopted for a bottom dune sand and b top clayey soil layers Fig. 10 Channel instrumentation. a low capacity tensiometer installed in the sand layer, b moisture and c suction probes used in the clayey soil layer and d comparison between the different tensiometers considered. Another aspect to be considered concerns the geotechnical nature of the soil used. Due to the observed shrinkage cracks, it is plausible to speculate that a more adequate volumetric behavior could be obtained if silty or low plasticity soils are used in the upper CB layer, although some loss of performance is expected in terms of hydraulic conductivity. Figure 12 illustrates the CB performance in the period of testing in the monitoring profiles 1 and 2 indicated in Fig. 2. As expected, shallower probes (5 cm depth) presented higher and more abrupt variations in moisture content and matric suction over time. Considering the monitoring profile 1, probes located at 25 cm depth presented a smooth transition and nearly constant values of moisture content throughout the whole period of the test, indicating that the CB clayey layer was able to temporarily store the infiltration water above this point. Concerning the probes located in the middle of the clayey layer (15 cm depth), a suction decrease can be observed after day 170, indicating the accumulation of water at this depth. With regard to the probes located in the sand layer, it is possible to observe an increase in the sand moisture content after day 248. Since the moisture content at the 25 cm depth moisture probe remained almost constant, the authors believe that there was a lateral supply of water coming from the crack presented in Fig. 11a. This was supported by visual inspections of the sand layer. One of the main shrinkage cracks, presented Fig. 11b, was formed close to the soil monitoring profile 2. The authors believe that this was the main reason for the differentiated behavior presented in the two studied profiles. In the case of the monitoring profile 2, all the probes present accentuated variations after rainy events, indicating a poor performance of the barrier in this region. The CB bottom drainage was activated on the days 127, 145, 193, 238, 243, 246 and 263. The total amount of water that leaked to the bottom corresponded to 5 mm of rain or about 1.89% of the total rainfall applied to the barrier in the period.
After day 248 in monitoring profile 1 and day 235 in monitoring profile 2, the suction probes indicated an increase in the soil moisture content, despite the observation of some drainage periods. Concomitantly and as expected, the suction probes indicate a decrease in the recorded values of matric suction. However, the suction probes were not able to reproduce the expected variations in suction due to moisture changes. The authors believe the following could have contributed to this inconsistency: SWRC hysteresis, since during this period intermediate drying/wetting branches of the SWRC are used and the nature of the dune sand retention curve, since in the moisture interval 10% ≤ ≤ 20% the expected suction variations, considering the main drying branch, lie in the narrow interval of 3.7-6.4 kPa.

Conclusions
This study involved the construction and testing of a flow channel apparatus to evaluate the performance of mineral cover barriers under different weather conditions. Concerning the apparatus developed, it can be said that a good overall performance was obtained, with the flow channel able to reproduce most of the main weather variables and monitor the CB performance in different profile positions. The use of suction measurement devices with different working principles and suction measurement ranges proved to be adequate to monitor two such different materials in terms of hydraulic properties. The employed moisture probes performed well both in coarse and clayey soils and the rain and the solar radiance simulators were able to achieve the desired performance, producing nearly homogeneous conditions for the whole length of the channel superficial area.
The different compaction protocols proved to be efficient in providing the desired values of moisture content and dry density to the soil and a satisfactory level of homogeneity in the layers hydraulic properties. No lateral spreading was observed in the water flow in the contacts between the compacted layers, which is particularly important in the case of the CB upper clayey soil. Furthermore, the shrinkage cracks formation as well as any flow anomaly can be observed visually through the channels glass walls.
The flow channel demonstrated its great potential use in laboratory simulations of field conditions. A temperature and relative humidity controlled room will allow the device to simulate and control all the main environment variables that influence the performance of mineral barriers in the field. Therefore, the flow channel can be considered as an important tool in the refinement of soil hydraulic properties obtained in laboratory, allowing that numerical simulations of CB barriers be performed with much more confidence.
With regard to the capillary barrier performance, the study has demonstrated that thickness of clayey layer was insufficient to provide an acceptable long term performance of the cover layer, preventing the shrinkage cracks from reaching the bottom sand layer. Although only about 1.89% of the total rainfall applied to the capillary barrier leaked to the bottom layer, which can be considered a good overall performance indicator, the tendency is that the CB performance continues to deteriorate in the following years, increasing the leachate generation in the landfill by infiltration.