A provenance management framework for reservoir engineering

Provenance is metadata that pertains to the history of data products. It is useful information in tracing the audit trails of data, determining the resource usage, and estimating data quality and data reliability, in addition to many other uses. It is challenging to collect and utilize provenance information in reservoir engineering mainly due to two reasons: (1) data products in reservoir engineering are heterogeneous, both syntactically and semantically, and; (2) the lack of workflow orchestration framework makes it challenging to collect and represent provenance in an integrated manner.;In this dissertation, we present a framework that models, collects and represents provenance information in reservoir engineering. We first present a methodology that collects provenance information from legacy logs. In order to collect provenance from such low level and fine-grained data, our approach is built upon two components: semantic rich workflow model and automatic workflow instance detection. The semantic rich workflow model maps low-level log entries onto high-level information and captures semantic information such as user intentions. Given such workflow models, we design a pattern matching algorithm to automatically detect workflow instances from legacy logs. Provenance information is collected within the context of workflow instances. Experimental results based on multiple synthetic data sets demonstrate the efficiency of our approach.;We present provenance models to represent provenance information collected from multiple workflow models as an integrated provenance graph. To address information heterogeneity, we also present a semantic provenance model, which annotates the collected provenance entities using a domain ontology. The domain ontology provides a shared representation of the concepts and their relationships in the reservoir engineering domain. It improves interoperability with third-party software applications.