Improving Interpretation Outcomes: quantifying biases and designing workflows for better seismic interpretation

地下地质模型的发展主要依赖于遥感数据的解释。我们提出了一个知识交流计划，共享现有信息，并尝试新的方法来确定人类偏见、锚定和信心对用于构建地质模型的数据解释的影响。通过这种知识交流和创新，我们将创建和促进解释的最佳工作流程，从而最大限度地降低石油和天然气行业解释不确定性的风险。碳氢化合物的地质勘探和生产以及地质储层中二氧化碳的储存都需要建立地下的三维图像。该图像由遥感信息（如分辨率较差的地震反射数据）和一维点源（如取样相对较少的感兴趣的地下体积的井眼）组成。改善这些数据集解释的工作主要集中在技术改进上，以改进遥感数据的成像和处理，以更好地阐明地下结构。但是，即使使用改进的技术来解释数据，以及随后创建的模型也是不确定的。这种不确定性等同于勘探和生产风险。这种风险是由于缺乏对数据的约束，无法创建“确定”的预测模型，并且由于在解释有限数据集期间应用的已知偏差而被放大。该知识交流提案旨在：量化已知偏差对地震反射数据集解释的影响，并建立一个工作流，以最大限度地减少行业可以部署的解释偏差。我们将与行业和行业数据集合作，在学术界和相关合作伙伴公司之间以及与硕士和博士学生之间交流行业工作流程和人为偏见影响的知识。在这种交流的基础上，我们将通过一系列实验来研究和量化锚定对解释的影响，从而创造新的知识。通过在实验中加入额外的数据，我们将测试个人如何处理与他们原来的解释相一致或相反的新信息；以及在面对相互矛盾的证据时，人们能坚持最初的预测多久。我们将比较有阶段访问不同数据的人群和一开始就有所有数据的人群。在整个过程中，我们将通过专家启发过程来衡量个人对其口译的信心。利用这些新知识，我们将量化人类偏见对解释不确定性的影响，并确定地震解释的最佳工作流程。从我们现有的和共同生成的知识中，我们将创建一系列的产品来促进这个工作流程，以及相关的知识，以及它们所基于的NERC科学。其中包括通过现有的虚拟地震地图集部署的数字视频片段的在线资源，每月有8,000-10,000名用户访问，以及作为NERC石油和天然气博士培训中心的一部分，为工业和早期职业科学家提供一系列培训包。

Development of geological models of the sub-surface relies on the interpretation of largely remotely sensed data. We propose a program of knowledge exchange that shares existing information and trials new methods for determining the impact of human biases, anchoring and confidence, on the interpretation of data used to build geological models. From this knowledge exchange and creation we will create and promote optimal workflows for interpretation that minimize risk in the oil and gas industry from interpretational uncertainty. The geological exploration and production of hydrocarbons and the storage of CO2 in geological reservoirs requires a 3D picture to be built of the sub-surface. This picture is made up of remotely sensed information like seismic reflection data with poor resolution, and 1D point sources such as well bores which sample a relatively small amount of the sub-surface volume of interest. Work on improving interpretation of these datasets has mainly focused on technological improvements to refine the imaging and processing of the remotely sensed data to better illuminate the sub-surface architecture. But even with improved techniques interpretations of the data, and the subsequent models created are uncertain. This uncertainty equates to exploration and production risk. The risk results from the lack of constraint from the data to create a 'certain' predictive model, and is amplified by known biases that are applied during interpretation of limited datasets. This knowledge exchange proposal aims to: quantify the effect of known biases on interpretation of seismic reflection datasets and to build a workflow that minimizes biases in interpretation that industry can deploy.We will work with industry, and on industry datasets, to exchange knowledge of industry workflows and the effects of human bias between the academics and partner companies involved, as well as with MSc and PhD students. Building on this exchange we will create new knowledge through a series of experiments to investigate and quantify the influence of anchoring on interpretation. By building into the experiment release of additional data we will test how individual's deal with new information that either confirms, or is contrary, to their original interpretation; and for how long individuals remain anchored to an original prediction in the face of contradictory evidence. We will compare cohorts of individuals with staged access to different data against those with all the data at the outset. Throughout the process we will gauge an individual's perception of confidence in their interpretation through an expert elicitation process. Using this new knowledge we will quantify the impact of human biases on interpretational uncertainty and determine an optimal workflow for seismic interpretation.From our combined existing and co-generated knowledge we will create a series of products to promote this workflow, and the associated knowledge, as well as the NERC science on which they are based. These will include an online resource of digital video footage deployed through the existing Virtual Seismic Atlas, accessed by 8,000-10,000 users monthly, and a series of training packages for industry and early career scientists undertaking PhDs as part of the NERC Oil and Gas Centre for Doctoral Training.