FIELD PERFORMANCE OF NDIR-BASED SENSORS FOR THE CONTINUOUS MEASUREMENT OF DISSOLVED CO2 IN KARST SPRINGS

The ability to monitor significant short-term changes in CO₂ concentrations in surface and near-surface aquatic environments is necessary prior to the implementation of advanced technologies based on the deep geologic sequestration of CO₂. Previous research in our laboratory has shown that non-dispersive infrared (NDIR) CO₂ sensors are useful in determining instantaneous dissolved CO₂ in groundwater monitoring wells and surface springs, but less is known about their robustness in the field during long-term monitoring activities. In this study, NDIR-based sensors were deployed for over a year in karst springs in central Pennsylvania. These springs are part of an on-going study where numerous water quality variables are continuously monitored to better understand complex regional water flow and recharge paths during storm events.

NDIR sensors measure CO₂ concentrations up to 20% in humid air using a single-beam dual-wavelength light source and a silicon-based sensor enclosed in a waterproof poly-tetrafluoroethylene (PTFE) sleeve that is highly-permeable to CO₂. Data were collected hourly using waterproof loggers, and the systems were powered by marine storage batteries. Three different modifications of this sensor design were examined during this study: analog and digital versions of the membrane-modified NDIR sensor, and a commercial temperature/pressure-compensated sensor designed for marine research. Issues addressed included comparison of the accuracy of the CO₂ data measured by the three sensors relative to alternative methods (geochemical and volumetric), and their relative ease of use in the field. In addition, the influences of hydrostatic pressure and temperature on sensor operation and the potential influence of biofouling on CO₂ diffusion across the sensor membranes were examined. Measured dissolved CO₂ revealed short-term changes throughout the entire rise and fall of selected storm hydrographs, providing much greater detail than could be achieved with discrete grab samples.