Using coupled modeling approaches to quantify hydrologic prediction uncertainty and to design effective monitoring networks.

摘要

Designing monitoring networks that can discriminate among competing conceptual models is a key challenge for hydrologists. This issue is examined by considering the impact of network design on the utility of measurements for constraining hydrologic prediction uncertainty. Specifically, a three-staged approach was developed and is presented as a set of modeling case studies. The first case study presents a sensitivity analysis that examines conditions under which the proposed measurement method is likely to detect observations associated with the hydrologic process and properties of interest. This application is focused on the use of geomorphic information to estimate infiltration on arid alluvial fans.; The second stage is an assessment of the likely utility of the measurement method to determine whether proposed measurements are likely to be useful for identifying hydraulic properties or hydrologic processes. This objective screening approach could reduce the number of unsuccessful uses of geophysical and other indirect measurement methods. A hypothetical site assessment examines whether the measurement method, temporal gravity change, is likely to detect signals associated with drawdown in an unconfined aquifer that occurs in response to pumping. Also, the utility of these measurements for identifying hydraulic conductivity and specific yield was considered.; The third stage, an analysis of optimal network design, compares the projected measurement costs with the expected benefits of constraining hydrologic prediction uncertainty. The final case study presents a network design approach for a feasibility assessment of a proposed artificial recharge site. Predefined sets of proposed measurements of temporal gravity change were considered for various measurement times. An ensemble approach was used to assess the likely impact of measurement error on prediction error and uncertainty for different combinations of measurement sets. The ensemble of prediction errors was translated to probability-weighted performance costs for each measurement set using a cost function. Total cost was calculated as the sum of the performance and measurement costs. The optimal measurement set, defined as the set with the lowest total cost, depends on the prediction of interest, the per measurement cost, the maximum risk-based cost associated with the hydrologic prediction, and the treatment of uncertainty in defining performance costs.