Identification of Gas Chimney in the Krishna-Godavari basin, eastern Indian margin Rowtu Ramu 1* and Kalachand Sain 1 Presenting Author email: *rowturamu@gmail.com 1 CSIR-National Geophysical Research Institute, Uppal Road, Hyderabad-500007, India Keywords Gas Chimney, Artificial Neural Network, Seismic attributes, KG Basin, Interpretation, Dip Steered Median Filter. Summary The chimney analysis tool is very useful for identifying fluid migration paths from source through the reservoir to the surface. We apply this tool to the multi-channel seismic data in the Krishna-Godavari (KG) offshore basin with a view to imaging the chimneys. After having conditioned the seismic data, multiple attributes such as the frequency washout, energy, dip variance, similarity have been computed and then merged using a non-linear Multi-Layer Perceptron (MLP) to derive a meta attribute, defined as the chimney attribute. This study helps in better interpretation of seismic data in terms of understanding the petroleum system of KG basin, and risk assessment in future drilling. Introduction The main aim of studying chimneys is to identify the hydrocarbon migration pathways. Chimneys are the vertical chaotic disordered features having low reflection strength. These are the spatial link between source, reservoir and cap rocks, spill-point and shallow fuel anomalies. Chimneys are interpreted on a seismic section as upward migrating gas seepages, which show a clear signature of migration from the bottom up to near the seabed. Seismic attributes based on directionality principle help in improving the perceptibility and mapping effectiveness of certain geological features like faults, chimneys, folds, reflections etc. Selection of attributes in the best possible way plays an important role in enhancing a particular geological feature (Meldahl et al., 2001). The similarity, strength, dip variance, frequency attributes are very sensitive in enlightening the chaotic activities from the environment (Brouwer et al., 2008), and can be used to create a new attribute through a supervised Neural Network in which multi-trace attributes are steered in a user determined or data determined manner. Mapping of gas chimneys from the seismic section helps in understanding the hydrocarbon seepage history from the source rock to the shallower prospects (Heggland, 1998). These geological features also provide an appraisal of pre- drilling shallow gas vulnerability or geo-hazard. Here we employ this concept to the multi-channel seismic data in Krishna-Godavari (KG) basin (Fig.1) in the eastern margin of India where disordered vertical zones have been observed on seismic section. To conform these features as seismic chimneys, we implement the neural network training to the directionality attributes at every chimney and non- chimney locations chosen on the seismic data. Data The present study is carried out on 2D post stack time migrated seismic data (Fig.2) in the KG basin. The data was acquired using a streamer of 4500 meter (360 channels) at a nominal towing depth of 8 meters by CSIR-NGRI (Sain et al., 2012) for the exploration of gas hydrates at water depths varying between 500 m to 2500 m (Fig.1). Fig.1. Location of seismic lines covering an area of 3369.249 km 2 in the southeast part of KG basin. The seismic line (KG-02), marked by red, is studied here.