Hou Xianlong, Hodges Ben R, Feng Dongyu, Liu Qixiao
Institute of Advanced Computing and Digital Engineering, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, Guangdong 518055, PR China.
Dept. of Civil, Architectural and Environmental Engineering, Univ. of Texas at Austin, 1 University Station C1786, Austin, TX 78712, USA.
Mar Pollut Bull. 2017 Mar 15;116(1-2):420-433. doi: 10.1016/j.marpolbul.2017.01.038. Epub 2017 Jan 23.
As oil transport increasing in the Texas bays, greater risks of ship collisions will become a challenge, yielding oil spill accidents as a consequence. To minimize the ecological damage and optimize rapid response, emergency managers need to be informed with how fast and where oil will spread as soon as possible after a spill. The state-of-the-art operational oil spill forecast modeling system improves the oil spill response into a new stage. However uncertainty due to predicted data inputs often elicits compromise on the reliability of the forecast result, leading to misdirection in contingency planning. Thus understanding the forecast uncertainty and reliability become significant. In this paper, Monte Carlo simulation is implemented to provide parameters to generate forecast probability maps. The oil spill forecast uncertainty is thus quantified by comparing the forecast probability map and the associated hindcast simulation. A HyosPy-based simple statistic model is developed to assess the reliability of an oil spill forecast in term of belief degree. The technologies developed in this study create a prototype for uncertainty and reliability analysis in numerical oil spill forecast modeling system, providing emergency managers to improve the capability of real time operational oil spill response and impact assessment.
随着德克萨斯湾石油运输量的增加，船舶碰撞风险加大将成为一项挑战，可能导致石油泄漏事故。为了将生态损害降至最低并优化快速反应，应急管理人员需要在溢油事故发生后尽快了解石油扩散的速度和地点。先进的操作性溢油预测建模系统将溢油应急响应提升到了一个新阶段。然而，由于预测数据输入存在不确定性，往往会影响预测结果的可靠性，导致应急计划出现偏差。因此，了解预测的不确定性和可靠性变得至关重要。本文采用蒙特卡洛模拟来提供参数，生成预测概率图。通过比较预测概率图和相关的后报模拟，对溢油预测的不确定性进行量化。开发了一个基于HyosPy的简单统计模型，从置信度的角度评估溢油预测的可靠性。本研究中开发的技术为数值溢油预测建模系统中的不确定性和可靠性分析创建了一个原型，为应急管理人员提高实时操作性溢油应急响应和影响评估能力提供了支持。
