We present results from direct seismic scale porosity and lithology prediction work in the Norwegian North Sea, targeting the Paleocene, Cretaceous and Jurassic. Firstly machine learning models were built at well scale using measurements and descriptions from core as target labels and conditioned wireline logs across 44 wells in order to predict effective porosity and lithology logs across the entire well track from Paleocene to Jurassic. Well scale prediction used Random Forests and MLPs achieving blind test accuracies of 74%.

Subsequently, a seismic survey with partial stack data was selected and a deep convolutional neural network was used to predict porosity and lithology based on upscaled version of the predicted logs. A subset of 11 wells were used in a blind cross validation training scheme allowing network hyperparameters to be tuned. We achieve overall R2 scores of 56%-62% for porosity at blind well locations. Over the extent of the survey we find that responses conform very well to the major interval boundaries. Maps generated using the predicted 3D data conform to known field boundaries and distribution of lithologies were consistent with depositional environment expected from traditional stratigraphic analysis and experience.


Article metrics loading...

Loading full text...

Full text loading...


  1. Russel, B.
    , HampsonD. and BankheadB. [2006] An inversion premier. CSEG Recorder, Special issue, 96–103.
    [Google Scholar]
  2. Noh, H., HongS. and HanB
    . [2015] Learning Deconvolution Network for Semantic Segmentation. ICCV ‘15 Proceedings, 1520–1528.
    [Google Scholar]

Data & Media loading...

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error