Article pubs.acs.org/est
Spatial and Temporal pCO2 Marine Monitoring Near Panarea Island (Italy) Using Multiple Low-Cost GasPro Sensors Stefano Graziani, Stan E. Beaubien,* Sabina Bigi, and Salvatore Lombardi Dipartimento di Scienze della Terra, Università di Roma “La Sapienza”, Piazzale Aldo Moro 5, 00185, Rome, Italy S Supporting Information *
ABSTRACT: The present paper describes the GasPro probe, a small, low-cost unit for in situ, continuous pCO2 monitoring. Laboratory tests defining its performance characteristics are reported, as are the results from a 60 h water-column deployment of 20 such units near a natural CO2 seep site off the coast of Panarea Island (Italy). The spatial-temporal evolution of dissolved CO2 movement is presented and possible origins and controlling mechanisms discussed. Results highlight the potential for this technology to be used for better understanding various dynamic physical and biochemical processes in marine environments, and for marine environmental monitoring of off-shore industrial sites. These experiments have allowed us to assess the advantages and disadvantages of the present GasPro prototype and to define areas for ongoing improvement.
■
INTRODUCTION The need for continuous and autonomous chemical monitoring of the marine environment has been widely acknowledged1 due to its dynamic nature as well as the logistical difficulties and high costs of intermittent, manual, ship-based sampling. Such tools have a wide spectrum of possible applications, ranging from the study of coastal biological systems, natural seep sites, or ocean acidification, to the environmental monitoring of industrial sites like hydrocarbon installations or CO2 capture and storage (CCS) sites. The need for continuous, in situ monitoring is particularly acute for the carbonate system.2 Recent developments in this area include innovative tools for the monitoring of pH,3 dissolved inorganic carbon (DIC),4,5 and pCO2. Most of the pCO2 methods use a gas permeable membrane to equilibrate the surrounding seawater with either a chemical solution or a gas volume. The most common chemical approach is to measure spectrophotometrically the pH change in a reference solution caused by the diffusing CO2. Both experimental systems6,7 and widely deployed units such as the highly precise and accurate SAMI-CO28 and CARIOCA9,10 have been developed. For gas-phase equilibration, CO2 can be measured directly using mass spectrometers11 or nondispersive infrared (NDIR) analyzers.12 Based on quoted sensitivity, power consumption, and schematic drawings, the NDIR sensors used in commercially available units like the Contros HydroC12 or Pro-Oceanus CO2−Pro are assumed to be full-size units and not the miniature sensors described here. Although precise, rapidly responding, and often equipped with autozero methods to adjust for sensor drift, these various instruments are typically large and costly, and their deployment © 2014 American Chemical Society
time may be restricted due to power requirements. Because elevated costs can limit the number of units deployed at one time, temporal monitoring is usually only conducted at a single point. To address these issues we have developed small, lowpower consuming, low-cost, pCO2 gas probes (“GasPro”), based on the principle of a miniature NDIR sensor located behind a gas permeable membrane. This extends our previous work for both groundwater13 and marine14,15 monitoring. The goal of this development has always been to minimize costs so that multiple units can be deployed simultaneously, thus providing both temporal and spatial data. Here we describe the technical characteristics of the GasPropCO2 sensor and then report the results obtained during a simultaneous deployment of 20 such units along a water column transect near natural CO2 seeps off the coast of Panarea Island (Italy). Field results are used to illustrate the potential of the technology and to infer gas origin and mixing processes at the site. To our knowledge this is the first reporting of the deployment of a large number of sensors to monitor the spatialtemporal evolution of pCO2 distribution at a real marine site.
■
STUDY SITE Panarea is the smallest island within the 200 km long Aeolian Arc, located off the north coast of Sicily.16 Panarea, and the small islets situated 3 km to the east, are the emergent parts of a 1200 m high, 20 km wide submarine stratovolcano.17 The Received: Revised: Accepted: Published: 12126
February 9, 2014 September 19, 2014 September 25, 2014 September 25, 2014 dx.doi.org/10.1021/es500666u | Environ. Sci. Technol. 2014, 48, 12126−12133
Environmental Science & Technology
Article
The membrane, supported by a porous brass disk, is a 35 mm diameter, 38 μm thick disk of Teflon AF 2400 (Biogeneral Inc., San Diego, CA). This material is one of the most permeable polymers available for CO227 and for this reason it is also used in many of the instruments described above.5−7 A 20 mm diameter, 16 mm tall NDIR sensor (model IRC-A1, Alphasense) is used for CO2 analysis. Although numerous manufacturers produce 20 mm diameter NDIR cells, the IRCA1 was chosen for its quoted specifications, including a sensor resolution of 1 ppm at zero CO2 and 100 ppm at a full scale of 5% (the maximum calibration span for this study). According to the manufacturer this NDIR sensor has minimal short-term drift and long-term drift caused by filament degradation can be decreased by reducing lamp warm-up time, lamp voltages, and analyzing less frequently for long-term deployments. The GasPro is programmed to make measurements at predetermined time intervals depending on the study needs and predicted gas exchange rate across the membrane. Battery life is on the order of 1 month for a lamp warm-up time of 2 min and a sampling frequency of once every 10 min (as used during this field campaign) and memory is essentially nonlimiting with a 2Gb SD card. GasPro Calibration. The following is a summary of the more detailed explanation given in File S1 (SI). An initial two-step laboratory, gas-phase calibration procedure was applied. First, the inversely proportional relationship between NDIR response and gas concentration was measured for a range of standards and the results fit to a power equation to obtain the necessary linearization coefficients. Note that over wider concentration ranges there will be a greater error in the low concentration values. Although manufacturer-provided coefficients can be used, we calculated them for each individual probe over the range of 0−5% CO2. Second, the probes were zeroed under pure N2 and then a span was applied at full scale (5% CO2 in this case); both of these parameters must be compensated for temperature due to its complex effect on sensor response. Again the required coefficients can be defined by the user or the generic manufacturer values can be used; the manufacturer values were used here. Subsequently, a small pCO2 correction (+5%) was applied to both laboratory and field probe results based on laboratory tests in a seawater tank. In this work, the response from a single GasPro probe (P) and associated headspace (HS) gas chromatograph analysis was compared for a series of standards between 0.4 and 4.6 matm pCO2. The regression of this data yielded the linear formula HSpCO2 = 1.05 × PpCO2, with an r2 of 0.9998 (Figure S4b; SI). At the time of this study it was not possible to calibrate all probes in water, and thus this correction factor was applied to adjust the gas-phase calibration for water conditions. For the probes deployed along the Panarea transect a final additional correction was applied to standardize their response and to minimize potential errors at low values caused by calibrating over such a wide concentration range, as described above. This involved deploying all 20 probes in a single block within the pockmark (about 5 m away from an area of diffuse bubbling) for a 6 h period, and then using the regression line formula between each individual probe and a single probe in the center of the distribution to narrow the spread of measurements for a given low pCO2 value (Figure S5, SI). The temperature sensors were also standardized by performing a simple shift of the block deployment data sets, to correct for
eastern islets are arranged in a 1 km-wide, subcircular pattern on a shallow plateau whose water depth ranges from 0 to 30 m (Figure S1b, Supporting Information (SI)). The primary fractures trend NE-SW and NW-SE, and it is along these discontinuities that natural, deep-origin CO2 is leaking from the sea floor into the overlying water column.18 Gas leakage is relatively stable in terms of composition (about 98% CO2, 1.7% H2S plus other trace gases) and flux rates ((7−9) × 106 l/d),19 aside from short-term outburst events like that which occurred in November of 2002.20 During this event flux increased by at least an order of magnitude for a three month period and compositional changes indicated a transitory magmatic input into the hydrothermal system.19 The outburst was focused near the islet of Bottaro, with the formation of a NW-SE trending, 10 m wide by 30 m long pockmark (Figure S1c, SI). Gas is still leaking from this pockmark, although at a much reduced rate. Other gas leaks have been mapped on the platform to the N and NW and in deeper waters on the volcano slope to the S and SE.21 Although gas leakage off Panarea has been studied since the 1980s in terms of chemical volcanology,22 the outburst event focused subsequent research on issues like seawater dynamics around bubble plumes20,23,24 and the potential impact of the leaking CO2 on the local ecosystem.25 Continuous monitoring has also been conducted at this site by other researchers, including the multiyear monitoring of a hydrothermal vent for acoustic noise (as a proxy for bubble flux);26 these results indicate that the bubble leakage rate may have both a tidal and a seasonal periodicity.
■
EXPERIMENTAL SECTION GasPro Description. The GasPro units tested in the laboratory and deployed at Panarea were the internal data logger version for shallow water monitoring of dissolved CO2 (Figure 1; Figure S2a, SI). Each probe is housed in a 200 mm
Figure 1. Schematic drawing of a GasPro-pCO2 probe; see photograph in Figure S2, SI.
long, 78 mm diameter Plexiglas cylinder, weighs 0.7 kg in air, and consumes less than 40 mA during the 2 min warm-up/ analysis period. Measurement is based on equilibration of a small-volume headspace, containing a miniature nondispersive infrared (NDIR) detector, with the surrounding water via diffusion through a gas permeable membrane. Equilibration is completely passive and no pumps are used to minimize power consumption. A larger chamber, physically isolated and located behind the sensor chamber, contains control electronics, memory, and two 3.67 V batteries in series (model LS 26500, Saft). All probes were also equipped with an external water temperature sensor while one probe had a pressure sensor to monitor tidal fluctuations. 12127
dx.doi.org/10.1021/es500666u | Environ. Sci. Technol. 2014, 48, 12126−12133
Environmental Science & Technology
Article
Figure 2. Laboratory tests of GasPro signal stability (a) and sensitivity (b). Assessment of field based sensitivity using two Panarea probes deployed 5 m apart at 2 m depth (c). Response times for four different flow rates at 21 and 18.5 °C (d); T90 is the time to attain 90% of the final value and 10′ is the 10 min sampling interval used in the field deployment.
Five lines (5 m horizontal spacing) with four probes each (2 m vertical spacing) were held vertically in the water column using heavy ballast and 5 L subsurface buoys (schematic diagram in Figure S1d and photo in Figure S2, SI). The probes were mounted with membranes facing upward to prevent direct contact with bubbles. For reference the transect results are presented looking north, with line 1 to the WSW toward Panarea Island and line 5 to the ENE toward Bottaro Islet. The top probe on each line was positioned 2 m below the water surface, while the membrane of the bottom probes were about 50 cm above the sediment surface. The probe with the pressure sensor was placed at the bottom of line 5. All sensors were programmed for one analysis every 10 min with a warm up time of 2 min. The total deployment period was about 2.5 days (May 27−30, 2013).
small offsets resulting from the use of the manufacturer’s calibration coefficients. Laboratory Experiments. All laboratory experiments were conducted using filtered Mediterranean seawater. Signal stability was assessed by monitoring pCO2 in water under a constant atmospheric CO2 concentration, under both stable and changing temperatures. The system was first left at an ambient temperature of about 19.5 °C, and then the temperature was allowed to rise slowly. The probe was programed to collect data once every 10 min with a 2 min lamp warm up time. Sensitivity was assessed by equilibrating a probe in seawater with 0.7 matm pCO2 at 18.6 °C, followed by the stepwise addition of progressively smaller aliquots (from 1.0 to 0.2 L) of seawater with 5.2 matm pCO2 to determine the point where the induced change in response is too small to be separated from the signal noise. Response time was assessed in the laboratory by transferring a single probe from the air into a 20 L tank filled with seawater having a pCO2 value of about 3.6 matm. Experiments were conducted at different flow rates to assess the influence of hydrodynamics on the response time of these passive sensors, including no flow, membrane-parallel flow at 37 and 63 mL s−1 (equivalent to current velocities of about 12 and 21 cm s−1), and direct flow at 83 mL s−1. Parallel flow was created by pumping tank water into one end of a thin Plexiglas chamber mounted on the probe face while direct flow was created by pointing the pump outlet directly onto the membrane. For the response time and sensitivity experiments the NDIR lamp was always on and data was collected once per second. Field Deployment. As stated above, all 20 probes were first deployed together for 6 h in a single block for intercomparison and probe standardization. The next day, all 20 GasPro units were deployed along a 20 m long and 6 m high transect located approximately 3 m SE of the pockmark. The transect was oriented ENE-WSW, perpendicular to the long axis of the pockmark and the main local current direction (Figure S1c, SI).
■
RESULTS AND DISCUSSION Probe Characterization. Stability tests under a constant temperature of 19.5 °C yielded an average pCO2 value of 0.313 matm and a 3σ of about 0.01 matm (10 μatm), giving an initial estimate of the detection limit of the probe (Figure 2a). This 3σ value is the same as the quoted zero repeatability of ±10 ppm for the IRC-A1 NDIR, indicating that the added control electronics have not introduced a significant amount of noise to the sensor. During the subsequent slow rise in water temperature the probe showed a similar level of stability. Initial tests had shown a temporary rise in probe-measured pCO2 values in correspondence with increasing temperature, but this was an artifact related to the changing headspace-water equilibrium caused by the temperature dependent Henry’s constant. Maximizing exchange with the surrounding atmosphere helped address this problem, however the fact that our present equipment does not allow for a precise control of ambient temperature may have introduced some minor artifacts in these results. Sensitivity was tested by adding progressively smaller amounts of dissolved CO2 to a probe-monitored water tank 12128
dx.doi.org/10.1021/es500666u | Environ. Sci. Technol. 2014, 48, 12126−12133
Environmental Science & Technology
Article
Figure 3. pCO2 and temperature data from all four GasPro sensors (GP0−3) deployed along line 5 (note water depths in legend), and pressure data from the deepest probe. Lettered events and specific starred peaks (e.g., a*) are discussed in the text.
coefficient (see discussion in File S1, SI); typically measurements of low concentration standards are within 9 matm). Event “c” immediately follows “b”, and is very similar except for having lower anomalies (best seen in the video file S3, SI). Event “f” enters from the ENE, migrates across the transect as a narrower anomaly (about 10 m wide), and then exits to the WSW about 2 h later. Interestingly this event is preceded by a short period that has the lowest measured bottom water temperatures and greatest stratification but only low pCO2 values (“e” in Figure 3), in contrast to the cold water−high pCO2 association observed for events “b” and “f”. This indicates different providence and shows that the measured pCO2 anomalies are not a temperature-induced artifact of the GasPro sensors. The animated video (File S2, SI) clearly shows that events “b” and “f” were formed by the ingress of colder bottom waters. Considering that the nearby pockmark is not capable of generating such a temperature anomaly, it is assumed that the observed CO2 originates from another leakage source in deeper, colder waters. In the absence of current data, knowledge of the mapped gas leakage points and the bathymetry of the subaqueous platform from the literature18,21 can narrow possible source locations. The shallowing of the seafloor to the east and northeast toward Bottaro Islet and to the southwest toward the exposed shoal of Lisca Nera excludes these two directions. Instead the platform slopes gently downward to the N and NW and to the south, whereas a more marked drop occurs to the SE (see bathymetry published by Anzidei and co-workers21). Gas leakage has been mapped both to the NW and SE,21 whereas colder currents coming from the SE have been experienced by divers (A. Fogliuzzi, Pers. comm.). Based on this evidence we tentatively interpret the cold-water, high pCO2 anomalies to originate from the south to southeast of the study site. Although the transport mechanism is unknown, it may be related to the upwelling of deeper waters via tidally related bottom water currents, similar 12130
dx.doi.org/10.1021/es500666u | Environ. Sci. Technol. 2014, 48, 12126−12133
Environmental Science & Technology
Article
Figure 4. Spatial distribution of temperature (top) and pCO2 (bottom) data along the transect: (a) pCO2 plume development in bottom waters during stratified conditions (point “b*” in Figure 3); (b) migration of dissolved CO2 toward the atmosphere during nonstratified conditions. Note different color scales. Each black dot represents a GasPro sensor.
importance of this type of data has been illustrated with the use of 20 such probes at the Panarea site, which defined small-scale transient events that would have been difficult to observe otherwise. In particular these field data have illustrated the difficulty of capturing transient events using discrete ship-based sampling, have shown how leaked, dissolved CO2 can be transferred toward the atmosphere, and have shown how seafloor CO2 seepage may expose benthic and pelagic biota to a highly variable range of pCO2 (and pH) values. These issues are of particular importance, for example, to the monitoring and risk assessment of offshore CCS sites. While the probes were prepared specifically for the expected wide concentration range at the Panarea seep site, laboratory tests have shown the potential for this technology to be applied in settings where smaller, more subtle pCO2 anomalies may be encountered. Although not yet attaining the precision needed for open ocean studies, other applications, such as the study of plume dynamics, coastal biological processes, or offshore industrial site monitoring, are within its capability given a more rigorous temperature compensation and narrower calibration range.
to the transient low pH, hypoxic water anomalies observed along the California coast.28,29 It should be noted that this plume of elevated pCO2 values was often restricted to the bottom waters of only two or three adjacent lines, resulting in a small (
