This digital dataset represents the modelled outputs of the three-dimensional distribution of sandiness in the Paskapoo Formation, Alberta, Canada. Sand-abundance values were determined from log analysis of water-well lithologs and petroleum exploration downhole-geophysical logs, specifically gamma-ray logs. We derived the sandiness values by calculating the average sandiness in 25 m thick slices above the base of the Paskapoo Formation in a borehole and by assigning the average value (P50) to the midpoint of each slice interval. Alberta Geological Survey Bulletin 66 provides detailed descriptions of the methodology.
We tested the borehole sand-abundance values in a 3-D variogram prior to kriging in a regular block model consisting of cells with dimensions of 1000 m on the x and y planes and 25 m in the z (elevation) direction. Values of modelled sandiness range from 0 (absence of sand) to 1 (100% sand) and are assigned elevations based on two datums:
1) a stratigraphic level corresponding to the height of the midpoint of each respective 25 m thick slice interval above the base of Paskapoo Formation, and
2) the elevation of the midpoint of the slice above sea level.
Importing the modelled results of sandiness into model viewing software, such as Voxler™, permits the user to generate three-dimensional isovalue plots illustrating the differences in distribution and geometry of aquifers as different cutoff or threshold criteria are applied. The dataset also enables one to delineate the regional lithostratigraphic units nested within stratigraphically complex rock formations.
We extracted the water-well lithologs from Environment and Sustainable Resource Development’s water-well database. These records lack rigorous location information other than approximations of x-y coordinates based on the Dominion Land Survey system, which we converted to geographic coordinates. To determine water-well collar elevations, we used ArcGIS software to extract values from the Shuttle Radar Topography Mission digital elevation model.
Alberta Geological Survey Bulletin 66 details how the data were collected, screened, culled and compiled to generate this dataset.
The main data used to determine and evaluate sandiness of the Paskapoo Formation were borehole logs from two major sources: water-well drillers’ lithologs and gamma-ray logs from the petroleum exploration industry. We also used ground-resistivity datasets derived from airborne electromagnetic surveys to verify modelled results of sandiness from borehole-log analyses. Expert-based decision methods coupled with numerous digital and geostatistical analysis methods were applied to generate estimates of sandiness within a borehole prior to 3-D modelling. With water-well drillers’ logs, more than 1500 unique water-well drillers’ descriptor terms for geology were summarized into two categories: sandy sediment (coded as ‘1’) and non-sandy, clayey sediment (coded as ‘0’).
For gamma-ray logs, all intervals with a <=75 API value were considered to be sand or permeable, based on air-permeameter tests on core. Sand-abundance values were calculated in 25 m thick intervals initially using two datums. We calculated the sandiness for 25 m thick slice intervals constructed parallel to the modelled surface of the base of the Paskapoo Formation, up to a maximum height of 750 m above the base. The analysis of sand abundance in 25 m thick slice intervals from about 2,400 gamma-ray logs yielded 14,566 point values, each of which was assigned a depth value corresponding to the midpoint of its respective slice above the base of the Paskapoo Formation.
For water-well lithologs, we calculated sandiness for 25 m thick slices constructed parallel to the bedrock topography to a depth of 150 m below top of bedrock. As with the gamma-ray log analysis, the percent-sand value (n) was assigned to the midpoint (z) of the height of the slice above the base of the Paskapoo Formation. We repeated this process for about 33,500 water-well lithologs, creating a dataset consisting of 48,162 three-dimensional x-y-z-n data, which, combined with the gamma-ray log dataset, resulted in about 62,725 sand-abundance values as raw data for modelling the three-dimensional distribution.
We applied several geostatistical procedures to evaluate vertical and horizontal trends in the data prior to modelling, including vertical-trend analysis and variography. Alberta Geological Survey Bulletin 66 outlines the details of these procedures, which included
1) normal score transformation of the data,
2) variogram calculation and modelling,
3) cross-validation to check the kriging model,
4) simple kriging of the data, and
5) back-transformation to original units.
This is a standard geostatistical modelling workflow to estimate values of a variable (sandiness) at unsampled locations and to assess the uncertainty in the estimates.
Both water-well and gamma-ray log datasets had the same shape of spatial structure, but we relied on the spatial distribution of the gamma-ray log data as they had less randomness. We designed a regional scale, 3-D block model with the following dimensions to model the sandiness values from the borehole datasets: cell size 1000 m (x-axis) x 1000 m (y-axis) x 25 m (z-axis) for a total number of cells of about 6,468,000.
Just slightly more than 10% of the cells (684,269) were populated with true values. We assigned the remainder with a null value of ‘-9999’. Modelled results, which are the basis of this dataset, were imported into Voxler™ 3-D software to visualize the results and assist in interpreting hydrostratigraphic units.