Assessing the Value of Information of Geophysical Data For Groundwater Management
Abstract
Effective groundwater management requires hydrogeologic models informed by various data sources. The long-term goal of our research is to develop methodologies that quantify the value of information (VOI) of geophysical data for water managers. We present an initial sensitivity study on assessing the reliability of airborne electro-magnetic (EM) data for detecting channel orientation. The reliability results are used to calculate VOI regarding decisions of artificial recharge to mitigate seawater intrusion. To demonstrate how a hydrogeologic problem can be framed in decision analysis terms, a hypothetical example is built, where water managers are considering artificial recharge to remediate seawater intrusion. Is the cost of recharge justified given the large uncertainty of subsurface heterogeneity that may interfere in a successful recharge? Thus, the decision is should recharge be performed, and if yes, where should recharge wells be located? This decision is difficult because of the large uncertainty of the aquifer heterogeneity that influences flow. The expected value of all possible outcomes to the decision without gathering additional EM information is the prior value VPRIOR. The value of information (VOI) is calculated as the expected gain in value after including the relevant new information, or the difference between the value after a free experiment (VFE) and the value prior (VPRIOR): VOI = VFE - VPRIOR Airborne EM has been used to detect confining clay layers and flow barriers. However, geophysical information rarely identifies the subsurface perfectly. Many challenges impact data quality and the resulting models (interpretation uncertainty). To evaluate how well airborne EM data detect the orientation of subsurface channel systems, 125 alternative binary, fluvial lithology models are generated, each categorized into one of three subsurface scenarios: northwest, southwest and mixed channel orientation. Using rock property relations, the lithology models are converted into electrical resistivity models for EM forward modeling, to generate time-domain EM data. Noise is added to the late times of the EM data to better represent typical airborne acquisition. Inversions are performed to obtain 125 inverted resistivity images. From the images, we calculate the angle of maximum spatial correlation at every cell, and compare it with the truth - the original lithology model. These synthetic models serve as a proxy to estimate misclassification probabilities of channel orientation from actual EM data. The misclassification probabilities are then used in the VOI calculations. Results are presented demonstrating how the reliability measure and the pumping schedule can impact VOI. Lastly, reliability and VOI are calculated and compared for land-based EM data, which has different spatial sampling and resolution than air-borne data.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2008
- Bibcode:
- 2008AGUFM.H43J..02T
- Keywords:
-
- 0684 Transient and time domain;
- 1816 Estimation and forecasting;
- 1869 Stochastic hydrology;
- 1880 Water management (6334);
- 6334 Regional planning (1880)