Abstract

Early Oligocene (Mahuva) pays hold significant potential for hydrocarbons in the southern part of the Tapti-Daman sector. Paleo-structural highs characterized by complex interplay of fluviodeltaic sandstone and shale deposits, influenced by marine conditions, have emerged as a proven hydrocarbon play for Early Oligocene sands in study area. The Mahuva Formation is subdivided into Upper and Lower Mahuva. The Lower Mahuva is predominantly shaly, while the Upper Mahuva contains hydrocarbon-bearing sandstone The study area is predominantly characterized by gas bearing sandstone reservoirs. During the Early Oligocene, a relative sea-level rise resulted in the deposition of thin limestone streaks at the base, succeeded by an intensified terrigenous influx that led to the development of thick sandstone reservoir facies in the upper section. These reservoirs are primarily composed of channel sands. To effectively delineate these channel sands, a multi seismic attribute approach was employed using unsupervised Machine Learning. Key seismic attributes utilized include cosine phase, instantaneous phase, Hilbert transform, relative acoustic impedance, RMS amplitude, signal envelope, sweetness and thin bed indicators. The characterization was conducted through Machine Learning using a self-growing neural network based on Growing Neural Gas (GNG) clustering algorithm (Fritzke, 1995) with optimized parameters (50 neurons, 15 iterations, and 5 output classes). The resulting seismic facies volume identified five distinct classes, with Class 5 reliably correlating with gas-bearing sands as validated by well data. An interval attribute map of Class 5 was generated between the Mahuva top and Mahuva top +120 ms, successfully delineating the geometry and spatial distribution of gas-bearing clastic reservoirs across the study area.