Calcium content prediction and lithology classification from well logs in sand-rich intervals of the Xihu depression, East China Sea Basin

In the sand-rich intervals of the Xihu Depression, identifying potential sealing beds is important for evaluating reservoir–seal relationships, particularly where continuous mudstone barriers may be limited. Calcareous sandstone interbeds may act as local barriers, but their distribution is difficult to trace because measured elemental calcium is sparse along the wells. This study combines well-log-based calcium-content prediction with lithology classification to identify calcareous sandstone and compare its distribution between wells. Calcium-content prediction from conventional logs achieved an R² of 0.89 and an RMSE of 0.52 percentage points. Predicted calcium was then combined with conventional logs and vertical features to identify six target lithologies. Among five classifiers compared on an internal test set of approximately 2, 000 records, AdaBoost achieved an accuracy of 0.88, balanced accuracy of 0.81, and macro-averaged F1 of 0.81. After model selection, the classifier was evaluated on 1, 900 labeled records from independent Well W4, achieving an accuracy of 0.82 and macro-averaged F1 of 0.76. The results support the identification of calcareous sandstone and interwell comparison of calcareous intervals. Some calcareous sandstone interbeds and mudstone overlie gas-bearing intervals, suggesting a potential reservoir–seal association; confirming effective sealing requires further petrophysical evidence.