Integration of seismic first arrival times to estimate water-air contact and initial water saturation Ting Sun a, * , Yuanyuan Shuai b , Christopher White c , Juan Lorenzo d , Shannon Chollett d a China University of Petroleum, Beijing, China b Shell Oil Company, Houston, TX, USA c Tulane University, New Orleans, LA, USA d Louisiana State University, Baton Rouge, LA, USA article info Article history: Received 20 December 2016 Received in revised form 30 April 2017 Accepted 7 May 2017 Available online 27 May 2017 Keywords: Data integration Seismic first arrival times Trust region method Water-air contact Initial water saturation abstract History matched models have helped reservoir engineers obtain a better understanding of the reservoir and optimize the future productions. Recent years, seismic data is more often to be integrated to improve the reservoir models, however, seismic data has to be inverted to seismic impedance values first. In this study, we introduce a new way to improve the initial reservoir models -integrating the seismic first arrival times to estimate the fluid contacts and initial water saturation. The key aspect of this method that distinguishes it from other methods is that seismic data is integrated directly. As a result, traditional seismic inversion step is not needed, and the data integration are through the comparison of observed first arrival times and simulated first arrival times (raytracing results). In our paper, the feasibility of using seismic first arrival times to improve initial reservoir models is discussed, and the method is verified by using sand tank experiment data. The observed seismic first arrival times were collected during the experiment, and the simulated first arrival times were calculated by seismic raytracing program. Trust region method was used to adjust water-air contact depth and initial water saturation to minimize the difference between simulated and observed data. At last, a good estimation of initial condition is achieved, and it demonstrates a potential of integrating seismic data to improve the reservoir models without an inversion step. © 2017 Elsevier B.V. All rights reserved. 1. Introduction Improved modeling has helped to optimize reservoir develop- ment. To reduce the inherent uncertainty of the models and obtain more predictive simulations, more data should be integrated into the geomodels. Production data, such as bottom hole pressure, gas oil ratio and water oil ratio have been widely used in history matching. Interwell tracer tests and well testing also provide additional sources of data (Thulin et al., 2007; Li et al., 2009; Valestrand et al., 2010). However, sometime we still could not have a good estimation of reservoir properties with these data, since they can provide high resolution estimation around well lo- cations, but the properties in regions far from wells remain poorly constrained. To reduce uncertainty in estimation, seismic data can be integrated with production data to provide denser information across whole field. Many investigators have addressed seismic data integration and proved that seismic data can help improve the reservoir models. Dong (2005) applied the ensemble Kalman filter (EnKF) method to rapidly update the estimation of the model var- iables in a small synthetic case which shows that it is possible to integrate both time–lapse seismic impedance data and production data. Emerick, de Moraes, and Rodrigues (2007) integrated time– lapse seismic attributes into a derivative-based assisted history matching tool; their optimization algorithm was based on a trust- region quasi-Newton method to minimize the mismatch between observed and simulated data from production and seismic. A method based on the combination of EnKF and ensemble Kalman smoother (EnKS) (Skjervheim et al., 2007) used a combination of production data and 4D seismic data. Their method was tested on a synthetic case and a real North Sea field case. For both the synthetic and field case, a better permeability estimate was obtained by including both seismic data and production data. For the 2D syn- thetic problem, better estimates of the permeability were obtained by integrating inverted seismic data at the time they were measured instead of using 4D data. Zhao, Reynolds, and Li (2008) proposed integrating seismic data (acoustic impedance data) at * Corresponding author. E-mail address: upcxiaos@gmail.com (T. Sun). Contents lists available at ScienceDirect Journal of Natural Gas Science and Engineering journal homepage: www.elsevier.com/locate/jngse http://dx.doi.org/10.1016/j.jngse.2017.05.011 1875-5100/© 2017 Elsevier B.V. All rights reserved. Journal of Natural Gas Science and Engineering 45 (2017) 391e398