Despite the several thousand horizontal wells drilled in the Austin Chalk, regional productivity analyses are limited, the most recent by Pearson (2010), Martin et al. (2011), and Pitman et al. (2020). Two of these three studies predate modern hydrofracture treatments and increased drilling in the Karnes trough region. This study incorporates recent production trends and includes recent completion techniques within the Austin Chalk play in Texas. Furthermore, we have investigated the Karnes trough fault and fractured trend to determine potential influence on productivity.

We performed an exploratory analysis to investigate the impact of these faults on well production. Following this, we used completion and fault-distance data to build a gradient-boosting model predicting per-well productivity. The gradient-boosting model was tuned through cross-validated Bayesian optimization. Finally, we interpreted the gradient boosting model by generating Shapley additive explanations values to explain the factors that influence model predictions.

The best wells are drilled downdip within 2 mi of major faults, but the variance in productivity is high in these areas. Analysis of the machine learning results shows that the most important fault parameters are sinuosity and the closest distance between a well and its nearest fault. Fault bearing and length are of secondary importance. Apart from fault properties, early-producing gas-oil ratios and well spacing are key to productivity.