Article pubs.acs.org/est
Adaptation of an Osmotically Pumped Continuous in Situ Water Sampler for Application in Riverine Environments A. Gkritzalis-Papadopoulos,*,† M. R. Palmer,‡ and M. C. Mowlem† †
Underwater Systems laboratory, Centre for Marine MicroSystems, National Oceanography Centre, European Way, Southampton SO14 3ZH, U.K. ‡ School of Ocean and Earth Science, NOCS, University of Southampton, European Way, Southampton SO14 3ZH, U.K. S Supporting Information *
ABSTRACT: We present the design of an osmotic water sampler that is adapted to and validated in freshwater. The sample is drawn into and stored in a continuous narrow bore tube. This geometry and slow pump rate (which is temperature dependent: 0.8 mL/d at 4 °C to 2.0 mL/d at 28 °C) minimizes sample dispersion. We have implemented in situ time-stamping which enables accurate study of pump rates and sample time defining procedures in field deployments and comparison with laboratory measurements. Temperature variations are common in rivers, and without an accurate time-stamping, or other defining procedure, time of sampling is ambiguous. The sampler was deployed for one month in a river, and its performance was evaluated by comparison with manually collected samples. Samples were analyzed for major ions using Ion Chromatography and collision reaction Inductively Couple Mass Spectrometry. Despite the differences of the two sampling methods (osmotic sampler averages, while manual samples provide snapshots), the two data sets show good agreement (average R2 ≈ 0.7), indicating the reliability of the sampler and at the same time highlighting the advantages of high frequency sampling in dynamic environments.
■
INTRODUCTION Rivers have a significant role in global biogeochemical cycles, with large river basins having been given particular attention. Areas of interest include the relationship of their geochemical characteristics with physicochemical processes (e.g., denudation rate, sediment transport and temperature variations),1 the effects of anthropogenic activities (e.g., responses to raised atmospheric CO2),2 and pollution events from domestic agricultural and industrial sources.3 Water parameters such as temperature, conductivity, flow rate, optical fluorescence, and optical backscatter can be monitored routinely in situ, which allows data acquisition with high spatial and high temporal resolution.4−8 Devices for in situ measurement of some biogeochemical species have also been developed,9−13 but conventional water sampling and analytical techniques are preferred for high frequency monitoring of small and accessible riverine ecosystems.14−17 Water Samplers for River Applications. Long-term, autonomous river water sampling is usually carried out using samplers based on a pump and a carousel with initially air filled bottles where sample is stored until recovery and transfer to the laboratory.5,6 These samplers are easy to use, robust and portable, but depend on an external power supply, collect discrete “snapshot” in time samples only, and are limited by the number of bottles. Alternatively samples can be acquired using Diffusive Gradients in Thin films (DGT) samplers.18−21 Here the chemical species of interest diffuse through gel mediums and are subsequently captured and concentrated on a resin © 2012 American Chemical Society
layer. After deployment the resin is analyzed yielding the time average concentration of the target in the environment. DGT samplers can only provide one sample over the deployment period, and the resin are usually optimized to specific chemical compounds, which limits their widespread application. One promising development has been the design and application of continuous water samplers based on sample storage in a single continuous tube connected to a pump utilizing osmotic pressure to provide low flow rates.22,23 Typical pump rates of ∼1 mL/day yield a sample/tube length of ∼1 m/ day, hence a 180 m long tube can provide storage of samples taken continuously over a period of six months. Pump flow rates range from 0.1 mL/d to 12 mL/d.23 For analysis, the sample tube is split into segments, and the discrete samples are transferred into vials for later analysis. Osmosamplers can produce a continuous record over extended periods resulting in daily, or near daily, discrete measurements of water constituents. The pumps do not depend on electrical power and do not contain moving parts making them very reliable.23 OsmoSamplers have been deployed in remote environments, such as boreholes and submarine hydrothermal vents, and produced time series records of a number of dissolved elements (1−2 days Received: Revised: Accepted: Published: 7293
February 13, 2012 May 29, 2012 May 30, 2012 May 30, 2012 dx.doi.org/10.1021/es300226y | Environ. Sci. Technol. 2012, 46, 7293−7300
Environmental Science & Technology
Article
and must be validated for successful use of the Osmosamplers in environments with large temperature variations (such as rivers where temperatures typically vary by 15 °C annually 28 and 4 °C daily 29). The resolution of this uncertainty is the focus of this study. National Oceanography Centre, Southampton (NOCS) Modified Osmotic Sampler. We have developed a system which time-stamps the sample within the sample tube to independently define sample age, thus allowing us to study the effect of temperature variations on the accuracy of the sample age. The principle is based on the periodic injection of an inert tracer (Rhodamine 6G (Sigma Aldrich UK), 0.5 mg/L) close to the sample inlet, which allows for more accurate segmentation of the sampling tube (see Figure 1 in the Supporting Information). Rhodamine is readily available and nontoxic and does not interfere with the laboratory analysis of major cations, anions, or trace metals. It has a molar absorptivity of 105 L·mol−1·cm−1 at 540 nm and strong fluorescence at ∼560 nm, enabling detection at trace concentrations. It also forms a strong color in the visible range in solution at 0.5 mg/L which can be readily observed through the FEP sample tube. This tracer was stored in collapsible blood bags (Baxter Co., UK). The tracer was injected using a solenoid micropump (Lee Co. Ltd. model: LPLA1210550L) which has a displacement volume of 50 μL per actuation into an inlet of a PEEK T-piece (Omnifit) installed in the sample tube 15 cm from its inlet. The sample tube and osmotic pump chamber has low compressibility and presents a high resistance to the dye addition, whereas there is a low resistance path to the open ended inlet. Hence, the injected dye replaces the most recently acquired sample rather than displacing or causing motion in the previously acquired sample. The pump was controlled by a custom-made electronic circuit based on a PIC microcontroller (PIC 18LF6722 Microchip Ltd.) with software written in C programming language. Each rhodamine injection initially occupied 6.5 cm in the sampling tube, but this length increases to ∼17 cm after one month due to dispersion. The system was powered by batteries (6x 1.5 V (9 V DC) AA Duracell batteries, USA), which were placed in a custom-built, water-tight, pressure-tolerant case (material: Delrin). The sampler’s pump has twelve (12) Alzet2ML1 osmotic pumps, operating in parallel. These were modified by removing the internal drug delivery bag, which enables continuous flow of large volumes over an extended period. The modified Alzet pumps were positioned on a PolyMethyl MethylAcrylate (PMMA) base and held in place using a retaining plate (also made from PMMA) that was placed on the base with stainless steel screws (see Figures 2 and 3 in the Supporting Information). Silicone O-rings were used to seal the individual pump elements and the two compartments (supersaturated NaCl brine and DIW). The o-rings were placed approximately 2 mm below the top part of the membrane, in order to allow maximum exposure of the membranes surface. This reduces the surface area by 3%, but both laboratory tests and field deployments showed that this setup is effective and we did not experience loss in flow rate. The selected materials (especially PMMA) have similar thermal expansion coefficients to water, which means that for the temperature gradients the sampler was exposed to thermal expansion and did not have a significant effect on the tube segmentation (less than 1.8 μL/°C (approx 1.8 mm/°C)). The spool for the sampling tube and the supporting plates were made from acetal plastic. The sampling
resolution) for periods of up to two years. Such records would be difficult, to obtain using other techniques or sensors.22,24 The osmotic pump pressure created when two liquids with different ionic concentrations (typically supersaturated sodium chloride solution and deionized water (DIW)) are separated by an osmotic membrane. Typically this is either a reverse osmosis membrane or a modified drug delivery system (Alzet osmotic pump).23,25 The pump rate depends on temperature, with a typical gradient of 3%/°C. The sampling tube (typical internal diameter is 0.8 mm to 2 mm) is initially filled with DIW. One end is connected to a DIW reservoir (on one side of the membrane) with sufficient volume to accommodate the total volume of sample to be acquired. The sampling tube is usually from PTFE or FEP and in some cases copper.5,25−27 Samples collected with the polymer tubes have been analyzed for dissolved elements (e.g., Fe, Mg, Mn), while samples from copper tubes could be analyzed for dissolved gases as well.5 One of the main features of this technology is that the collected samples are not physically separated, and therefore diffusion and dispersion can lead to mixing of distinct samples in the tube. However, Jannasch et al. (2004) demonstrated that careful design of flow rate and sample tube dimensions enabled them to predict and observe 99.9% of any step change in sample concentration with a time resolution of 1.5 days (2 m in sample length) when the sampler was deployed for 1 year.23 The variation of pump rate with time, and internal diameter of the sample tube with distance, must also be accounted for to enable unambiguous identification of the age of the sample at any given location in the sample tube. The flow across the osmotic membrane, and hence pump rate, depends on the membrane’s surface characteristics, any modifications during pump assembly, the ionic concentration difference across the membrane and the temperature of the two solutions. Sufficient volume in the DIW reservoir and overcapacity in the supersaturated salt reservoir ensure that it is only the temperature that has a significant time dependent effect on pump rate. For Alzet osmotic drug delivery system, the manufacturer of the osmotic pumps (Alzet) provides the following equation to describe the flow rate of each osmotic pump Q = Q 0·(0.141·e 0.051·T − 0.007·π + 0.12)
(1)
where Q0 is the osmotic pump specific flow rate calculated at 37 °C (10 μL/h), T is the temperature (°C), and π is the osmolality (bar) of the solution outside the membrane (for this case π = 0). Therefore, together with a continuous record of pump temperature this equation can be used to predict pump flow rate as a function of time and hence sample age as a function of location in the sample tube. The accuracy of the prediction depends on how accurately this equation predicts flow rate. Accuracy can be improved if the total length of sample is known, and this is used to adjust the coefficients in eq 1. In salt water deployments this length can be defined by the distance between the sample inlet and the sharp concentration gradient between the sample and the DIW initially used to fill the sample tube predeployment. In fresh water environments we have determined this by placing a dye mark at the inlet to the sample tube at the start of the deployment (“T0”) and measuring the distance between this mark and the final sample on retrieval. This we term a “T0 and end point calibration”. However, the accuracy of this approach has not been assessed 7294
dx.doi.org/10.1021/es300226y | Environ. Sci. Technol. 2012, 46, 7293−7300
Environmental Science & Technology
Article
Figure 1. Flow rate evaluation of the sampler. The solid line is the nominal/theoretical flow rate which is calculated from eq 1 for unmodified alzet osmotic pumps. The dotted lines are the ±10% tolerance quoted by the manufacturer’s data sheet. • are flow rate measurements on July 2008; X are measurements of the flow rate on December 2005. Error bars are 2σ of the measured values.
tube had a 1 mm ID and was made from Fluorinated Ethylene Propylene (FEP). For all the fluidic connections we used zero dead volume PEEK connectors (Omnifit, DibA Industries Ltd., UK). In order to prevent large particles entering the sampling tube we used an in-line filter with a 60 μm pore size (GENP012, 60 μm Filter − with a 1/4−28 UNF tapped thread, EnviroTech Instruments LLC, USA), reduced in size in order to decrease the dead volume from 4 mL to less than 80 μL. Such filters increase the dead volume for the sampler (80 μL occupy approximately 8 cm of tube length and represent approximately 2 h of sampling time) and can potentially support bacterial growth. However it is documented30 that sampling tube of osmotic sampling devices have been clogged, and as our deployment arrangement was to deploy the sampler close to the river bed (approximately 20 cm above), we wanted to prevent suspended matter entering the tube (especially when the river flow was low). The results from this study did not indicate that the filter hindered the performance of our sampler. Determination of Flow Rate under Laboratory Conditions. The flow rate of the sampler pump was evaluated over a range of temperatures (Figure 1). Equation 1 predicts that a device with 12 osmotic pumps connected in parallel will yield a flow rate ranging from 0.7 to 2.1 mL/d (temperature range 2−30 °C). The pump assembly was placed in a temperature controlled water bath (±0.1 °C), and the temperature was changed every 5 days. Measurements of flow rate were made by connecting the inlet to a titration buret (±0.01 mL, negligible observed evaporation), and the volume change was recorded every 24 h. Test Deployment Site. The deployment site (SZ 4318 1019) is located in the Lymington River catchment (see Figure 2 in the Supporting Information). The catchment has an area of 127 km2 and lies in the New Forest National Park, ∼10 km west of Southampton. The maximum elevation of the catchment is ∼100 m. The site is located 5 km up stream of the maximum tidal reach and lies in a rural area, consisting of open moorland, mixed forests, and minor enclosed grazing. The geology of the area consists of postglacial deposits of gravel and brick earth, overlaying sands, silts, and clays. The most
common type of soil in the catchment is classed as surface water gley.31 These soils are relatively impermeable, which results in large river run offs during heavy rainfall. This is a general characteristic of New Forest catchments; hence the Environment Agency has developed the New Forest Catchment Flood Management Plan (CFMP), part of which is the gauging station, where the sampler was deployed (Environment Agency, 2009).
■
METHODOLOGY AND DEPLOYMENT SETTING The sampler was deployed at the sampling site from 14/05 until 12/06/2008 with its inlet placed 30 cm from the river bed (maximum water depth at the sampling point was 180 cm), and the entire device was always immersed below the water surface. The performance of the sampler was evaluated against spot samples collected manually ∼20 cm below the water surface at noon every day. All collected samples were filtered on-site through 0.45 μm cellulose acetate filters and analyzed for a suite of anions and cations, as described below. Part of each filtered sample was transferred to DI-rinsed and dried polyethylene bottles and analyzed by ion chromatography, and part was transferred to an acid cleaned polyethylene bottle, acidified to pH < 2 with triple distilled 6 M HNO3 and analyzed with ICP-MS. Sampling bottles used to collect samples for ICP-MS analysis were soaked overnight with 10% HNO3, rinsed 3 times with DIW, and kept in zip lock plastic bags. The in situ conductivity and temperature were also measured at the time of collection (conductivity meter: WTW LF320, sensor-conductivity cell: WTW TetraCon 325 ± 0.1 μS/cm for conductivity, ± 0.15 °C for temperature). Samples from the Osmotic Sampler. Immediately after recovery the sampling tube was disconnected from the main pump, and the two ends were sealed using blanking plugs (P316 PFA Plug, Upchurch Scientific, USA). On return to the laboratory, the sampling tube was segmented with a ceramic knife at the center of the rhodamine time-stamp (based on visual inspection). Each sample was expelled from the tube segments into a cleaned centrifuge vial using filtered compressed air. Samples where then diluted 10-fold with DIW. Part 7295
dx.doi.org/10.1021/es300226y | Environ. Sci. Technol. 2012, 46, 7293−7300
Environmental Science & Technology
Article
Environment Agency and the precipitation data by the Met Office (www.metoffice.gov.uk). Both data sets have gone through the quality control procedures of the issuing bodies which is adequate for the purpose of this study (our main interest is to identify relationships between the relative changes of the dissolved load, the river’s flow rate, and precipitation). Between 15/05 and 25/05/2008 there was no precipitation in the area, which is reflected in the low flow rate of the river (0.3−0.5 m3/s) and high conductivity (290−330 μS/cm). In the early hours of 26/05/08 a heavy storm led to a rapid increase of the river’s flow rate from ∼0.3 m3/s at 03:00 BST (British Summer Time, which is GMT+01:00) to a peak of 27 m 3/s at 18:30 BST (and a large associated drop in conductivity), with the flow rate then subsiding to
