Conceptual uncertainty in the interpretation of geological data: statistical analysis of factors influencing interpetation and associated risk

该项目将研究和开发新的方法，以便能够评估因不同的解释概念和范式对同一地质数据集所带来的偏见而产生的风险。石油工业的一项关键任务是评估与碳氢化合物估算相关的风险。这对于达到或超过产量峰值的盆地（如北海）的采收率最大化至关重要。在更复杂的地质情况下查明新储量也必须在可接受的商业风险范围内进行。数据测量、收集和处理中的不确定性至少可以粗略地加以解释。然而，一个更根本的困难是如何正确分配由于解释偏见而产生的风险。地下地质模型是根据三维地震数据、井眼地球物理测井和遥感数据等数据集创建的。这些数据在有限的时间和空间分辨率下采样了有限的地下体积，因此最终的模型高度依赖于解释者的概念框架。通常情况下，来自不同教育背景的口译员，或者具有不同油田环境经验的口译员，可能会对相同的数据得出非常不同的结果。地下地质模型是数据受限的自然系统，“诊断技能”和统计不确定性具有重大的社会和经济影响。这个问题将通过使用一些解释概念来确定从合成地震数据集导出的模型的变异性来研究。合成地震数据将从Midland Valley软件中创建的完全定义的地质模型中创建。然后对数据集进行解释，并将应用于没有事先信息的数据的概念与有事先信息的对照组进行比较。结果将被分析与“真实”地质的差异，以及控制组和非控制组之间的差异。差异的量化将用于评估由于解释偏差造成的不确定性程度。合成地震数据的工作将得到实地研究的补充，自然受限于3D曝光。在一个二氧化碳自然泄漏的地区，断层网络地图将用于创建基于不同几何断层连接概念的多种结构模型。Midland Valley新开发的软件4DMove将用于验证和评估与收集到的不同解释概念相关的不确定性，包括合成地震和现场数据。4DMove中的分析将通过对问卷参与者所收集信息的多元回归分析来支持，以评估经验、教育和培训等其他因素的影响。在软件软件TrapTester中，将评估划分，从而评估油气藏或二氧化碳储存潜力，并对不同结构模型进行量化，以突出地下结构框架对油藏连通性和潜力的关键影响。目前，北海油井的成功率在35-40%之间，每口井的成本约为1000万美元，在超深水区，每口井的成本高达5000万美元，错误的定位是对石油公司资源的浪费。任何能够减少地质概念不确定性的工具都将对行业产生重大影响。对于地下废物处理和二氧化碳储存的社会和经济影响，也可以提出类似的论点。该项目的目标是：-开发技术和方法来评估影响概念不确定性的因素-量化解释误差-量化不同模型对前景的影响（使用4DMove） -创建一个过程来减少石油勘探和废物储存中与结构模型相关的不确定性。

This project will research and develop new methods to enable the assessment of risk arising from the bias introduced by alternative interpretational concepts and paradigms to the same geological data set. A crucial task for the petroleum industry is assessing the risk associated with estimates of hydrocarbons. This is of increasing importance for maximising recovery in basins at or past peak production, such as the North Sea. The identification of new reserves in more complex geological situations must also occur within an acceptable commercial risk envelope. Uncertainty in data measurement, collection and processing can be accounted for, at least crudely. A more fundamental difficulty, however, is correctly assigning risk due to interpretational bias. Models of sub-surface geology are created from data sets such as 3D seismic data, well bore geophysical logs, and remote sensing data. These data sample a limited volume of the subsurface and at a limited resolution in time and space, therefore the final model is highly dependent on the interpreter's conceptual framework. It is often the case that interpreters from different educational backgrounds, or with experience in different oil field settings, can come up with very different results for the same data. Models of sub-surface geology are data-under-constrained natural systems and 'diagnostic skill' and statistical uncertainty have significant social and economic impact. This problem will be examined by determining the variability of models derived from synthetic seismic data sets using a number of interpretational concepts. Synthetic seismic data will be created from a fully defined geological model created in Midland Valley's software. The data sets will then be subject to interpretation, and the concepts applied to the data without prior information will be compared to control groups, given prior information. The results will be analysed for variation from the 'real' geology, and for variability between the control and non-control groups. Quantification of the differences will be used to assess the degree of uncertainty due to interpretational bias. The work on synthetic seismic data will be complimented by a field based study, naturally limited in 3D by exposure. Mapped fault networks will be used to create multiple structural models based on different concepts for geometrical fault linkages, in an area that has natural leakage of CO2. Midland Valley's newly developed software 4DMove will be used to validate and assess the uncertainties related to the different interpretational concepts collected, both for the synthetic seismic and the field data. Analysis in 4DMove will be supported by polytomous regression analysis of information captured from participants in questionnaires to assess influences from other factors such as: experience, education and training. Compartmentalisation, and hence hydrocarbon reservoir or CO2 storage potential will be assessed in the software package TrapTester and quantified for different structural models, to highlight the critical impact of sub-surface structural frameworks on reservoir connectivity and hence potential. The current success rate of wells in the North Sea stands at between 35-40%, and with the cost per well c. 10 million dollars, increasing to c.50 million dollars in ultra deep water, erroneous well positioning is a waste of oil company resources. Any tool that reduces geological concept uncertainty will have a large impact within the industry. Similar arguments, for social and economic impact can be made for sub-surface waste disposal and CO2 storage. The projects objectives are: -to develop techniques and methodologies to assess factors influencing conceptual uncertainty -to quantify interpretational error -to quantify the impact on prospectivity of different models (using 4DMove) -to create a process to reduce the uncertainty associated with the structural model in petroleum exploration and waste storage.