6 th International Conference & Exposition on Petroleum Geophysics “Kolkata 2006” (1045) Introduction Some authors (Vernik et. al., 2002, Engelmark, 2004) define net-to-gross (N/G) as the fractional volume of sand in the entire reservoir. It may be useful to estimate the fractional volume of hydrocarbon reserves (pay) in the reservoir. In this paper, we estimate two quantities – (1) the volume fraction of sand in the reservoir (V sand or N/G), (2) the volume fraction hydrocarbon reserves (pay) in the reservoir (V Pay or PVF). In each block, or “seismic pixel” within the reservoir, the volume fraction of hydrocarbon is given by the following equation: ( 29 W Pay S v - = 1 φ (1) In the above equation, φ represents porosity, and S W represents water saturation. To compute the volume fraction of hydrocarbon in the reservoir, we compute v Pay at each seismic pixel that indicates sand, and then sum over the entire reservoir. Sources of Uncertainty The major sources of uncertainty in estimating net- to-gross and pay volume fraction are as follows: (1) Measurements – The measured data, both seismic and well-log data are noisy. The seismic signal-to- noise is generally lower than the signal-to-noise of the well-log data, because of acquisition geometries. Net-to-gross from Seismic P and S Impedances: Estimation and Uncertainty Analysis using Bayesian Statistics Madhumita Sengupta*, Ran Bachrach, Niranjan Banik, WesternGeco. Summary Net-to-gross (N/G ) is a measure of the amount of sand or pay in the overall reservoir and is used to appraise reservoir quality and the economics associated with reservoir development. As seismic based reservoir characterization technology is advancing in many cases lithology and porosity information derived from seismic inversion can be used to derive an estimate of N/G. In many cases this problem is non unique due to accuracy of seismic inversion and the rock physics nature of the lithology and porosity prediction. Thus as N/G estimates are often used for economic decision-making it is important to associate expected risk or confidence associated with the prediction. In this paper we we present a workflow to compute quantitative estimates of N/G, along with associated uncertainties, from well-log calibrated pre-stack seismic inversion attributes. The main tools we use in this workflow are Seismic (AVO) Inversion, Rock Physics, and Bayesian Statistics. We estimate the N/G from the seismically derived rock properties. We derive the uncertainty in net-to-gross from uncertainties in seismic inversion, reservoir properties, and geologic interpretation. Especially at the higher frequencies, the seismic data quality is poorer than the corresponding log data quality. In calibrating the seismic to the well-log data, we account for the log-to-seismic discrepancy in data quality. (2) Inversion – Apart from the measurement errors in the seismic data, there are errors in the inversion results, These source of these errors can be categorized as follows: A. Errors resulting from various processing sequences (e.g., gather flattening or multiple attenuation before AVO inversion). B. Model approximations (e.g., a 2 vs 3 term AVO is or approximate inversion process that ignores some second-order wave-propagation effects such as mode conversion etc.). and C. errors associated with filling of the inversion null space (this is specifically important when using convolutional forward model as the basis for the AVO inversion, which explicitly place all the low frequencies in the null space of the inversion process. Low frequency compensation is then needed to fill up the low frequency part of the reflectivity spectrum). We note that although these uncertainties are not easy to quantify, one should try to include a measure of these uncertainties resulting from inversion when estimating the uncertainty in net- to-gross from seismic. (3) Rock Properties – Rock properties such as clay content, porosity, and fluid saturation are often