Multi-step inversion using electromagnetic measurements

1. A method of evaluating a subterranean earth formation comprising:

running a process with a logging tool residing in a borehole in the earth formation to collect shallow measurements of a property of the formation, defined to be measurements of the property of the formation adjacent the borehole and adjacent the logging tool, and deep measurements of the property of the formation, defined to be measurement of the property of the formation farther away from the logging tool than the shallow measurements;

performing an inversion on the shallow measurements to produce a group of possible formation models that fit the shallow measurements;

applying a machine-learning algorithm to estimate a shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure;

performing an inversion on the deep measurements to produce a group of possible formation models that fit the deep measurements, wherein performing the inversion on the deep measurements includes performing an inversion on one or more of the shallow measurements and the deep measurements and includes starting the inversion with the group of possible formation models that fit the shallow measurements;

expanding the shallow formation structure using the group of possible formation models that fit the deep measurements to produce a deep formation structure;

combining the shallow formation structure and the deep formation structure to produce a combined formation structure; and

prospecting for hydrocarbons, planning well drilling, controlling well drilling, performing production analysis, or performing reservoir analysis based on the combined formation structure.

2. The method of claim 1 further comprising, after expanding the fixed shallow formation structure using the group of possible formation models:

determining that deeper measurements of the property of the formation, defined to be measurements by the logging tool of the property of the formation farther away from the logging tool than the deep measurements, are available and, in response:

performing an inversion on one or more of the shallow measurements, the deep measurements, and the deeper measurements to produce a group of possible formation models that fit the deeper measurements, and

expanding the deep formation structure using the group of possible formation models that fit the deeper measurements to produce a deeper formation structure.

3. The method of claim 2 wherein performing the inversion on one or more of the shallow measurements, the deep measurements, and the deeper measurements includes starting the inversion with one or more of the group of possible formation models that fit the shallow measurements and the group of possible formation models that fit the deep measurements.

4. The method of claim 2 wherein the response to determining that deeper measurements of the property of the formation are available includes, before performing the inversion on the deeper measurements:

re-applying the machine-learning algorithm to estimate the shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure.

5. The method of claim 2 wherein expanding the fixed shallow formation structure comprises:

fixing the formation inversion results at the shallow formation structure for a shallow radial interval from a wall of the borehole to a shallow distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deep formation structure, using the group of possible formation models that fit the deep measurements, to produce a deep formation structure;

fixing the formation inversion results at the deep formation structure for a deep radial interval from the shallow distance from the wall of the borehole to a deep distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deeper formation structure, using the group of possible formation models that fit the deeper measurements, to produce a deeper formation structure; and

fixing the formation inversion results at the deeper formation structure for a deeper radial interval from the deep distance from the wall of the borehole to a deeper distance from the wall of the borehole.

6. The method of claim 1 wherein expanding the fixed shallow formation structure comprises:

fixing the formation inversion results at the shallow formation structure for a shallow radial interval from a wall of the borehole to a shallow distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deep formation structure, using the group of possible formation models that fit the deep measurements, to produce a deep formation structure; and

fixing the formation inversion results at the deep formation structure for a deep radial interval from the shallow distance from the wall of the borehole to a deep distance from the wall of the borehole.

7. A system for evaluating a subterranean earth formation, comprising:

a logging tool residing in a borehole intersecting the earth formation to collect shallow measurements of a property of the formation, defined to be measurements of the property of the formation adjacent the borehole and adjacent the logging tool, and deep measurements of the property of the formation, defined to be measurements of the property of the formation farther away from the logging tool than the shallow measurements; and

a processor in communication with the logging tool and operable to:

perform an inversion on the shallow measurements to produce a group of possible formation models that fit the shallow measurements;

apply a machine-learning algorithm to estimate a shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure;

perform an inversion on the deep measurements to produce a group of possible formation models that fit the deep measurements, wherein performing the inversion on the deep measurements includes using the shallow measurements in the inversion in addition to the deep measurements and includes starting the inversion with the group of possible formation models that fit the shallow measurements;

expand the shallow formation structure using the group of possible formation models that fit the deep measurements to produce a deep formation structure;

combine the shallow formation structure and the deep formation structure to produce a combined formation structure; and

prospect for hydrocarbons, plan well drilling, control well drilling, perform production analysis, or perform reservoir analysis based on the combined formation structure.

8. The system of claim 7 wherein the processor is further operable to, after expanding the fixed shallow formation structure using the group of possible formation models:

determine that deeper measurements of the property of the formation, defined to be measurements by the logging tool of the property of the formation farther away from the logging tool than the deep measurements, are available and, in response:

perform an inversion on one or more of the shallow measurements, the deep measurements, and the deeper measurements to produce a group of possible formation models that fit the deeper measurements, and

expand the deep formation structure using the group of possible formation models that fit the deeper measurements to produce a deeper formation structure.

9. The system of claim 8 wherein performing the inversion on one or more of the shallow measurements, the deep measurements, and the deeper measurements includes starting the inversion with one or more of the group of possible formation models that fit the shallow measurements and the group of possible formation models that fit the deep measurements.

10. The system of claim 8 wherein the response to determining that deeper measurements of the property of the formation are available includes, before performing the inversion on the deeper measurements:

re-applying the machine-learning algorithm to estimate the shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure.

11. The system of claim 8 wherein expanding the fixed shallow formation structure comprises:

fixing the formation inversion results at the shallow formation structure for a shallow radial interval from a wall of the borehole to a shallow distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deep formation structure, using the group of possible formation models that fit the deep measurements, to produce a deep formation structure;

fixing the formation inversion results at the deep formation structure for a deep radial interval from the shallow distance from the wall of the borehole to a deep distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deeper formation structure, using the group of possible formation models that fit the deeper measurements, to produce a deeper formation structure; and

fixing the formation inversion results at the deeper formation structure for a deeper radial interval from the deep distance from the wall of the borehole to a deeper distance from the wall of the borehole.

12. The system of claim 7 wherein expanding the fixed shallow formation structure comprises:

fixing the formation inversion results at the shallow formation structure for a shallow radial interval from a wall of the borehole to a shallow distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deep formation structure, using the group of possible formation models that fit the deep measurements, to produce a deep formation structure; and

fixing the formation inversion results at the deep formation structure for a deep radial interval from the shallow distance from the wall of the borehole to a deep distance from the wall of the borehole.

13. A non-transitory computer-readable medium on which is recorded a computer program, the computer program comprising executable instructions, that, when executed, perform a method of evaluating a subterranean earth formation comprising:

running a process with a logging tool residing in a borehole in the earth formation to collect shallow measurements of a property of the formation, defined to be measurements of the property of the formation adjacent the borehole and adjacent the logging tool, and deep measurements of the property of the formation, defined to be measurement of the property of the formation farther away from the logging tool than the shallow measurements;

performing an inversion on the shallow measurements to produce a group of possible formation models that fit the shallow measurements;

applying a machine-learning algorithm to estimate a shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure;

performing an inversion on the deep measurements to produce a group of possible formation models that fit the deep measurements, wherein performing the inversion on the deep measurements includes performing an inversion on the shallow measurements and the deep measurements and includes starting the inversion with the group of possible formation models that fit the shallow measurements;

expanding the shallow formation structure using the group of possible formation models that fit the deep measurements to produce a deep formation structure;

combining the shallow formation structure and the deep formation structure to produce a combined formation structure; and

prospecting for hydrocarbons, planning well drilling, controlling well drilling, performing production analysis, or performing reservoir analysis based on the combined formation structure.

14. The computer-readable medium of claim 13 wherein the method further comprises, after expanding the fixed shallow formation structure using the group of possible formation models:

determining that deeper measurements of the property of the formation, defined to be measurements by the logging tool of the property of the formation farther away from the logging tool than the deep measurements, are available and, in response:

performing an inversion on one or more of the shallow measurement, the deep measurements, and the deeper measurements to produce a group of possible formation models that fit the deeper measurements, and

expanding the deep formation structure using the group of possible formation models that fit the deeper measurements to produce a deeper formation structure.

15. The computer-readable medium of claim 14 wherein performing the inversion on one or more of the shallow measurements, the deep measurements, and the deeper measurements includes starting the inversion with one or more of the group of possible formation models that fit the shallow measurements and the group of possible formation models that fit the deep measurements.

16. The computer-readable medium of claim 14 wherein the response to determining that deeper measurements of the property of the formation are available includes, before performing the inversion on the deeper measurements:

re-applying the machine-learning algorithm to estimate the shallow formation structure, using the group of possible formation models that fit the shallow measurements, to produce the shallow formation structure.

17. The computer-readable medium of claim 14 wherein expanding the fixed shallow formation structure comprises:

fixing the formation inversion results at the shallow formation structure for a shallow radial interval from a wall of the borehole to a shallow distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deep formation structure, using the group of possible formation models that fit the deep measurements, to produce a deep formation structure;

fixing the formation inversion results at the deep formation structure for a deep radial interval from the shallow distance from the wall of the borehole to a deep distance from the wall of the borehole;

applying the machine-learning algorithm to estimate the deeper formation structure, using the group of possible formation models that fit the deeper measurements, to produce a deeper formation structure; and

fixing the formation inversion results at the deeper formation structure for a deeper radial interval from the deep distance from the wall of the borehole to a deeper distance from the wall of the borehole.