5:15 PM - 5:30 PM
[S24-1-04] A new probabilistic framework to estimate the information content of industrial bottom-hole temperature data: A case study using the Australian OzTemp dataset
The industrial drilling for petroleum and minerals resources can provide extensive datasets of bottom-hole temperature (BHT) measurements. These represent a primary source of data for understanding the thermal state of the crust. Unfortunately, they are of notoriously low and varied quality. A new framework is therefore presented here to deal with these uncertainties in a probabilistic manner. The framework focuses on two major components of BHT uncertainties; provenance uncertainties, those related to measurement errors and subsequent correction, and intrinsic uncertainties, those related to data rounding and data entry error. By building an appreciation for the uncertainties inherit in BHT measurements, we are able to estimate the information provided by each datum on the true subsurface temperature.
A compilation of international, data-driven studies is used to identify the broad range of biases inherit in uncorrected BHT measurements, as well as the precision of corrected, drill string test, and equilibrium BHT measurements. To identify the degree and magnitude of data-rounding, a novel approached is developed based on a logical appraisal of the statistical characteristics expected of an unadulterated dataset. Deviations from these expectations reveal the scope of human intervention in the raw data. Synthetic testing suggests that it is possible for this method to accurately estimate both the degree and magnitude of potential data-rounding errors in BHT datasets.
Finally, the framework is applied to examine OzTemp, the Australian industrial BHT dataset. The results identify data rounding to the nearest 5* and 10*C, as well as to the nearest 2*, 5* and 10*F. Individual probability density functions can be created to describe the uncertainty between any BHT datum and the true virgin rock temperature. This framework thereby enables higher levels of precision and accuracy in the interpretation and analysis of BHT measurements than has been previously achieved to date.
A compilation of international, data-driven studies is used to identify the broad range of biases inherit in uncorrected BHT measurements, as well as the precision of corrected, drill string test, and equilibrium BHT measurements. To identify the degree and magnitude of data-rounding, a novel approached is developed based on a logical appraisal of the statistical characteristics expected of an unadulterated dataset. Deviations from these expectations reveal the scope of human intervention in the raw data. Synthetic testing suggests that it is possible for this method to accurately estimate both the degree and magnitude of potential data-rounding errors in BHT datasets.
Finally, the framework is applied to examine OzTemp, the Australian industrial BHT dataset. The results identify data rounding to the nearest 5* and 10*C, as well as to the nearest 2*, 5* and 10*F. Individual probability density functions can be created to describe the uncertainty between any BHT datum and the true virgin rock temperature. This framework thereby enables higher levels of precision and accuracy in the interpretation and analysis of BHT measurements than has been previously achieved to date.