Inferring injection returns from chloride monitoring data

Abstract

Reservoir chloride concentration and injection rate data are used to examine injector-producer connectivity in two reservoirs, Dixie Valley, Nevada and Palinpinon-I, Philippines. General trends of chloride and injection rate with time are isolated from their respective short-term variations using the wavelet transformation approach. Multiple regression techniques are then used to correlate the isolated short-term variations in chloride with corresponding short-term fluctuations in injection rates and subsequently to quantify the degree of connectivity between injectors and producers. Communication between specific injector-producer pairs, as implied by analysis of data from Palinpinon-I, was also verified by comparison with tracer test data and qualitative field observations. Results of analysis of data from Dixie Valley demonstrate that multilinear modeling is not suitable for analyzing data sets that lack sufficient time variability. In contrast, adequate time variability is observed in data from Palinpinon-I, and qualitative field observations and tracer test data agreed best with the results of regression on changes in chloride concentration over four-month periods (wavelet detail level 3). Improvements in the analysis could result from increased data collection frequency of both chloride and injection rate as well as accounting for the nonlinearity of chloride with injection rate.