Design of sensor networks for long term monitoring of geological sequestration

Analysis of geologic sequestration on a wide scale with the aid of integrated assessment models showed earlier that overall leakage rates from the storage sites must be less than 0.1% of stored volume per year if the strategy is to be effective for long term control of atmospheric CO2 concentrations. Early detection and characterization of a nascent leak significantly increases the probability that a timely solution can be found. Because potential leakage pathways are not necessarily known a priori, monitoring must be done across a region as large as 100 km(2) in the vicinity of a CO2 injection well, which represents the area of review of the supercritical CO2 bubble under typical injection scenarios. The relatively high background levels of CO2 present in the atmosphere and soil, coupled with their seasonal and diurnal modulation, make immediate detection of a small CO2 leak difficult. Co-injection of tracers during sequestration eliminates the ambiguity and delay in positive identification of a leak. We report here on the design of sensor networks, as a part of the development of a prototypical, field deployable sensor/tracer technology for monitoring the rate of leakage of geologically sequestered CO2. The primary focus is on conservative tracer technology for immediate leak detection. In so doing, we are laying a foundation from which reactive tracer technology for pathway characterization can be developed. A semi-analytical model developed earlier for modeling the injection, fate and transport of sequestered CO2 was adapted for sensor network design. The design simulations are developed to answer several key questions about the sensor network layout. Simulation results indicate that the key parameters governing the monitoring network design are the sensor node spacing (S), sampling density D (no. of sensors/km(2)). minimum detection limit of sensor (L), injected tracer concentration (C-o) and dispersed tracer concentration (C-t) in the environment. Environmental dispersal modeling shows that C-t is a strong function of environmental conditions such as wind velocity. Analysis of the trade-offs indicates that a low detection limit allows effective monitoring with a minimal sensor sampling density and low tracer C-o. (c) 2005 Elsevier Ltd. All rights reserved.