Prediction of CO2 occurrence in natural gases is difficult owing to the int
erplay among basinal sources (e.g. volcanic emanations, carbonate hydrolysi
s and metamorphism, organic diagenesis and maturation), reservoir processes
(e.g. mineralogical transformations), and poorly understood physicochemica
l phenomena (e.g. differential solubility of gas components, carbonate buff
ering, phase segregation). Published approaches to assessing CO2 risk are b
ased either on circumstantial evidence of basinal influences or statistical
analysis of reservoir attributes. The present study presents the results o
f empirical, statistical and neural network analysis of CO2 abundance in na
tural gas and discusses potentially influential basinal and reservoir attri
butes to quantify risk factors for anomalous CO2 occurrences in Southeast A
sia. The results of this study indicate that tectonic setting (including ba
sement crustal age), basement fault density, reservoir temperature and rese
rvoir pressure are key elements controlling CO2 abundance. Assignment of CO
2 risk at the basin level can be roughly approximated using the results of
this study. More accurate assessments of CO2 risk at the basin, play, field
and reservoir levels require more comprehensive geological and geochemical
data as well as consideration of potentially influential physico-chemical
processes. (C) 1998 Elsevier Science Ltd. All rights reserved.