A basin atlas and offset wells: tops, thickness and the next prognosis

Every well prognosis starts with two questions: what rock will the bit meet, and at what depth? A basin atlas answers the first for the whole basin; the offset wells answer the second for one location. This article builds the prognosis for a new gas well on a ridge from 13 offsets: the structure grids and isopachs, a prognosis checked by hand, an honest uncertainty for each top, and the update once the bit picks its first deep top.

What the basin atlas gives, and what it doesn't

A basin atlas is the regional picture: the stratigraphic column with the age and lithology of each unit, its typical depth and thickness ranges, its role as reservoir, seal or source, its drilling hazards, and the basin's structural style. It is where a planner learns that the Triassic shale here thins onto basement ridges, or that a coal-bearing interval swells and slows the bit.

What it cannot give is a depth for one location. Basin-wide ranges run to hundreds of metres, because the same unit sits shallow on a ridge and deep in a trough. The atlas sets the order of the units, their character and the concepts to test. The depths come from the wells around the new location, converted into surfaces and thicknesses, and from how well those surfaces predict a well they have not seen.

From offset tops to surfaces

The first step is a common datum. Each offset's tops are converted from measured depth to true vertical depth with its survey, then to depth below sea level (TVDSS) by subtracting its rotary-table elevation, so wells drilled from different rigs line up structurally.

A surface through a handful of wells needs a method that honours each well and behaves sensibly between and beyond them. A common one fits a plane through the tops by least squares, for the regional dip, and spreads each well's departure from the plane by inverse distance squared:

a₀, a_E, a_N

the plane: depth at the wells' centre x̄, ȳ, and its gradient east and north

r_i

well i's residual: its top less the plane at its location

d_i

horizontal distance from well i to the point being estimated

t

an isopach: the gridded thickness between a shallow top and a deeper one

The second line is the isopach method. Deep tops have fewer wells than shallow ones, so instead of gridding a deep top from its own few wells, grid the better-controlled shallow top and add the thickness below it. With the same wells for both and a linear gridder like this one, the two ways give the same answer to the centimetre; the isopach method pays only when the shallow top has more control.

The uncertainty comes from leave-one-out cross-validation: take each offset out in turn, grid the rest, and see how far the surface misses the well it has not seen. The root-mean-square of those misses is a fair ± for a new well placed among the same offsets.

The example

The example is set in conditions typical of the Cooper Basin, Australia: a Permian–Triassic gas province under a Jurassic–Cretaceous cover, on a flat desert plain. The target is a Permian fluvial sandstone and coal sequence on a north-east-trending ridge, under a Triassic shale that thins over the ridge. The new well is vertical, on the ridge axis, with a rotary table 46 m above sea level.

Four tops are mapped: the Jurassic sandstone, the top of the Triassic shale, the top of the Upper Permian coal measures (the target) and the top of the Lower Permian sandstones. The shallow oil wells are the reason the isopach method matters here: they put 13 wells on the Triassic top, against 7 on the Permian tops.

The Triassic top by hand

The prognosis for the top of the Triassic shale at the new well, worked through as the engine does it.

1. The plane. Fitted through the 13 tops, it sits at 1,964.25 m TVDSS at the wells' centre and deepens 4.46 m per km east and 7.87 m per km north: 9.04 m per km towards 030°, a dip of 0.52°.
2. The plane at the new well. The new well is 54 m west and 123 m south of the centre: 1,964.25 − 0.24 − 0.97 = 1,963.04 m.
3. The residuals. Offset 9, 0.92 km away, sits 13.6 m above the plane; Offset 10, 1.03 km away, 17.7 m above it. With weights of 1/d², those two carry 38 % and 30 % of the weight. The other 11 wells, 2.3–5.6 km away, share the remaining 32 %.
4. The weighted residual. Σ wr / Σ w = −10.2 m.
5. The prognosis. 1,963.0 − 10.2 = 1,952.8 m TVDSS, 1,998.8 m below the rotary table.

DrillIQ's engine gives 1,952.8 m. The new well found the top at 1,947.5 m, 5.3 m high. The step that matters is the third one: two shallow wells about a kilometre away decide the answer, and both sit on the crest. Without them the ridge would not be in the Triassic grid at all.

The section across the ridge

Across the ridge the Triassic top is well controlled: the shallow wells near the new well pull it up to a crest, and the new well lands 5.3 m above it. The Permian tops are another matter. Gridded from the seven deep wells alone (dashed), the target top is nearly flat across the crest at 2,165 m, because the only deep wells close to the section are on the flanks. Built from the Triassic grid plus the gridded isopach (solid), it follows the crest and comes up to 2,148 m.

The new well found the target at 2,107.8 m TVDSS, 40 m above even the better prognosis. The Triassic shale is 160 m thick at the new well, against 195 m gridded from the offsets. The isopach is the problem, not the structure: four of the seven deep wells are on the flanks, where the shale is 212–234 m thick, and inverse distance cannot know that it thins onto the crest.

The uncertainty of each top

Leave-one-out puts numbers on that. The Jurassic sandstone, 13 wells, comes back within 8.5 m RMS; the Triassic top within 21 m. The two Permian tops, with seven wells each, miss by 76–105 m RMS, and by up to 179 m at the corner wells. The isopach method improves them only a little, by 6 m. The largest misses are the crestal wells predicted from the flanks and the flank wells predicted across the crest: exactly the case of the new well.

Every found depth lies inside its ±. That is the point of the exercise: a prognosis of 2,148 ± 76 m is honest, and one of 2,148 m on its own is not. The casing point above the target, the coring point and the mud weight entering the Permian all have to be planned for the whole range, not the single number.

Thickness from the right analogue

The atlas concept narrows the range. If the Triassic shale thins onto ridges, the thickness at a crestal location should come from a crestal well, not an average of all of them. Offset 5 sits on the ridge axis, 2.8 km along it, with 165.4 m of Triassic shale and 262.8 m of Upper Permian. Added to the 13-well Triassic grid, its thicknesses put the target at 2,118.2 m and the Lower Permian at 2,381.0 m: 10.4 m and 9.1 m deep, against 40–76 m for the grids.

The analogue is a geological choice, and it can be wrong; it is right here because the new well and Offset 5 share a structural position. Present it as the base case with the gridded prognoses as the deep side of the range, and say which wells it rests on.

Updating while drilling

Once the new well picks the Triassic top, the Permian tops can be revised from the pick. DrillIQ's look-ahead keeps a thickness below the last pick and adds it to the pick's depth. With the planned isopach it expects the target at 2,188.8 m MD; with the average thickness of the seven deep offsets, 196.7 m, at 2,190.2 m; with the nearest deep offset, Offset 5, at 2,158.9 m. The bit found it at 2,153.8 m MD: the nearest offset was 5.1 m deep, the average 36.4 m.

The lesson repeats: an average of offsets in different structural positions is not a better estimate than one offset in the right one. For the cuttings and drill-break side of picking that top, see cuttings, lithology and the mud log; for steering within it once landed, geosteering with a typewell.

A checklist for the well team

• Read the basin atlas first: the order of the units, their character, and how they change across highs and lows.
• Put every offset's tops on one datum, TVDSS, from its own survey and rotary-table elevation.
• Count the wells behind each top; grid deep tops from the better-controlled shallow top plus isopachs.
• Run leave-one-out on every surface and quote each top with its ±.
• Check whether the offsets near the new well share its structural position, and choose thickness analogues accordingly.
• Plan the casing point, coring point and mud weight for the whole range, not the single prognosis.
• After each pick, revise the tops ahead and say which thickness the revision used.
• After the well, add its tops and regrid before the next one.

How DrillIQ handles it

StratDrill 3D is DrillIQ's basin atlas: 97 basins with stratigraphic columns carrying typical depth and thickness ranges, lithology, petroleum roles, hazards and a confidence grade for every row, structural frameworks, a 3D subsurface block and cross-sections, and a pre-spud brief listing the units and hazards to a planned TD. It works offline. The structure grids above were made with DrillIQ's structure-from-tops engine in the Geosteering module, which grids each marker from the offsets' penetrations and samples it along the well path; Geology & Cutting Analysis correlates the offsets, hangs them on a chosen top, and revises the tops ahead from the prognosis, the offsets' average or the nearest offset while drilling. The numbers above come from those engines and the hand check shown.

Book a demo → See all capabilities

References

1. Tearpock, D.J. and Bischke, R.E. Applied Subsurface Geological Mapping with Structural Methods, 2nd ed. Prentice Hall, 2003.
2. Davis, J.C. Statistics and Data Analysis in Geology, 3rd ed. Wiley, 2002.
3. Shepard, D. A two-dimensional interpolation function for irregularly-spaced data. Proceedings of the 23rd ACM National Conference, 1968.
4. Isaaks, E.H. and Srivastava, R.M. An Introduction to Applied Geostatistics. Oxford University Press, 1989.

About this example. The well in this article is a representative scenario set in typical Cooper Basin, Australia conditions and worked through with DrillIQ's engines. It is not a client's well, and its figures are not field results.