First Break - Volume 43, Issue 7, 2025

Accurate pore pressure prediction is critical for safe and efficient drilling operations. However, conventional methods face challenges due to insufficient quality of seismic data and limited wellbore availability. This study explores the application of machine learning algorithms (decision tree, random forest and XGBoost) to predict pore pressure at the offshore Nile Delta Basin, especially in the Sapphire Field, which consists of a complex geological environment with vertically stacked reservoirs. The XGBoost model achieved the best performance with 99% accuracy on training data and 97% on test data, using 53,711 recorded data points from seven drilling records in four wells. The model has been adjusted to accurately predict pore pressure using existing drilling logs in the absence of direct measurements. Error analysis demonstrated a strong correlation with actual data. This approach has the potential to enhance pressure prediction operations and mitigate drilling risk.

Understanding pre-seismic gases and surface anomalies is crucial for improving earthquake risk assessment. This study analyses remote sensing data from Sentinel-5P, Aqua-AIRS, and Sentinel-1 satellites to detect potential pre-seismic anomalies associated with seven earthquakes (M > 6) between 2017 and 2023. Atmospheric variables (CO, CH4, CO2, O3, surface temperature, water vapour, and aerosol optical depth) and surface deformation patterns were evaluated two to five months before the main events. Results reveal consistent trends, particularly in gases, although differences between events suggest the influence of tectonic setting and seismic preparation processes. These findings support integrating multi-parameter remote sensing data into pre-earthquake monitoring strategies.

A porous low-impedance gas-saturated sandstone, can cause P-wave seismic signals to experience dispersion and attenuation. This phenomenon is commonly referred to as Wave-Induced Fluid Flow (WIFF). This paper discusses a methodology to determine gas saturation in the low impedance sandstone using dispersion and attenuation attributes. The Zoeppritz (AVO) approximation by Shuey is used to calculate the attributes from the partial angle stack seismic data. The methodology is applied to the seismic data from Sadewa Deepwater Field in, East Kalimantan, Indonesia. As it is observed that the density parameter also has a good relationship with the porosity of gas-saturated rocks, then the combination of dispersion, attenuation, and density attributes is used to determine the gas saturation. Based on cross-plot analysis at the Sadewa well, it can be concluded that the dispersion and attenuation attributes calculated from near-angle gather and mid-angle stack seismic data have a good correlation with the gas saturation log values. A more detail analysis shows that the best correlation is obtained when a combination attribute calculated by the equation of (AVFDp)^0.5 / Dn is used, where AVF is the attenuation attribute, Dp is the dispersion attribute, and Dn is density. Gas saturation values in this study were computed using the linear regression equation derived from the cross-plot analysis between the (AVFDp)^0.5 / Dn attribute and gas saturation (Sg), which is expressed as .

Onshore seismic data is often challenged by high noise levels and limited resolution, even after thorough processing. In this study, we showcase the effectiveness of a novel bandwidth extension technique applied to a newly reprocessed 3D dataset from onshore USA. This approach delivers higher-resolution seismic images and more geologically consistent inversion results. The enhanced data-set improves the accuracy of landing zone selection and supports better planning and execution of the lateral drilling path, ultimately reducing out-of-zone drilling and increasing production rates.

Before the bandwidth extension process, targeted data conditioning was carried out to minimise noise as much as possible. The bandwidth extension itself was achieved using the Multiscale Fourier Transform of the seismic trace, which performs time-frequency analysis across a range of window lengths. This variation in window length helps to capture both local and global amplitude relationships between events, enabling reflectivity inversion that is independent of the seismic wavelet’s amplitude spectrum. Since the temporal and spatial variation of the seismic wavelet in reflection data is often poorly understood, this method offers several advantages over traditional seismic reflectivity inversion. No wavelet extraction is required, meaning the inversion process can proceed without relying on well data, seismic ties, or time-depth functions. Additionally, the inversion is sparse and doesn’t require a starting model. Since no wavelet is necessary, the method can be applied directly to depth-migrated data.

This inversion technique is particularly valuable for multi-client seismic datasets because it doesn’t rely on well data and can be applied over large areas. It enhances the value of existing datasets and revitalises older ones, making them suitable for detailed interpretation and well planning. Well data can then be used to validate the results.