Determining soil CO2 threshold levels by means of common forecasting methods as part of near-surface monitoring for carbon sequestration projects

Abstract

Abstract Throughout several pilot-projects for CO2 storage soil gas monitoring tied in with monitoring of other parameters have long been considered key elements to ensure safe operating and raise public acceptance. However, no real-life application of soil gas monitoring has been developed until now. We present a new monitoring approach that is based on a continuous comparison of actual soil CO2 concentration data against predicted target values including estimated confidence intervals which are considered as threshold values. Our new approach is based on time series forecasting methods. Since they are not commonplace in soil gas monitoring we give an introduction to simple approaches and more sophisticated modelling categories. These have been tested on different data sets from our soil gas monitoring data base. Finally, we apply the most robust overall model to real time monitoring data of the Hontomin carbon sequestration pilot site. The evaluation revealed that Auto-Regressive Integrated Moving Average (ARIMA) models give reliable estimates of future soil gas concentrations. On a practice-relevant time frame, the accuracy of predicted values and confidence intervals is better than compared to simple approaches like seasonal means. As a result, we present a practical, methodological approach that increases the technology readiness level of soil gas monitoring. Moreover, it has a high potential to be transferred to other environmental monitoring compartments (e.g. soil gas emissions). Since the applied workflow can be automated, integration in real time monitoring systems is straightforward. However, a minimum of two years of data is required to achieve an adequate forecasting precision.