Understanding diagenetic heterogeneity in tight sandstone reservoirs is vital for hydrocarbon exploration. As a typical tight sandstone reservoir, the seventh unit of the Upper Triassic Yanchang Formation in the Ordos Basin (Chang 7 unit), central China, is an important oil-producing interval. Results of helium porosity and permeability and petrographic assessment from thin sections, X-ray diffraction, scanning electron microscopy and cathodoluminescence analysis demonstrate that the sandstones have encountered various diagenetic processes encompassing mechanical and chemical compaction, cementation by carbonate, quartz, clay minerals, and dissolution of feldspar and lithic fragments. The sandstones comprise silt-to medium-grained lithic arkoses to feldspathic litharenites and litharenites, which have low porosity (0.5%-13.6%, with an average of 6.8%) and low permeability (0.009 x 10(-3) mu m(2) to 1.818 x 10(-3) mu m(2), with an average of 0.106 x 10(-3) mu m(2)). This study suggests that diagenetic facies identified from petrographic observations can be up-scaled by correlation with wire-line log responses, which can facilitate prediction of reservoir quality at a field scale. Four diagenetic facies are determined based on petrographic features including intensity of compaction, cement types and amounts, and degree of dissolution. Unstable and labile components of sandstones can be identified by low bulk density and low gamma ray log values, and those sandstones show the highest reservoir quality. Tightly compacted sandstones/siltstones, which tend to have high gamma ray readings and relatively high bulk density values, show the poorest reservoir quality. A model based on principal component analysis (PCA) is built and show better prediction of diagenetic facies than biplots of well logs. The model is validated by blind testing log-predicted diagenetic fades against petrographic features from core samples of the Upper Triassic Yanchang Formation in the Ordos Basin, which indicates it is a helpful predictive model. (C) 2017 Elsevier Ltd. All rights reserved.
