Use of Image Analysis to Develop New Benchmarking Datasets for Variable Density Flow Scenarios

Image analysis (IA) techniques are used to measure properties such as concentration and water content in porous media experiments. Measurements made by IA techniques provide extensive spatial information on a temporal scale. However, a robust analysis of the accuracy of IA techniques has not been presented in previous publications. Therefore, results from experiments using IA techniques are not widely accepted as benchmarking datasets. The conventional method for testing the accuracy of IA techniques is the computation of global mass balance error. We demonstrate the limitations of quantifying IA errors based on this conventional method. We also introduce an alternate statistics-based method for estimating errors. The entire discussion is presented using a theoretical test problem. We also present the results of a physical laboratory experiment processed using the IA technique. The dataset is from a sinking plume experiment simulated by injecting saltwater into a freshwater aquifer. The experimental results are also compared with numerical modeling results generated using SEAWAT.