MRI Track 1: Acquisition of High Performance Computing to Model Coastal Responses to a Changing Environment

由于气候和人为因素的综合影响，缅因湾正在经历前所未有的变化。从某些方面来看，缅因湾是全球海洋变暖最快的地区之一。这种变暖趋势与过去一个世纪极端事件增加73%同时发生，极端事件被定义为前1%的风暴。对于一个与海洋资源和基础设施紧密相连的经济体来说，这些变化的潜在影响是巨大的。从2005年到2013年，缅因州的水产养殖业（主要包括鲑鱼和贝类）的价值翻了一番，现在仅次于该州的龙虾产业。2014年，缅因州的商业渔业价值达到创纪录的5.85亿美元。对沿海地区自然资源和基础设施的有效管理需要建模工具，这些工具不仅有助于我们了解自然和人为引起的环境驱动因素变化的生态和物理后果，而且还有助于我们预测这些变化并权衡适应或缓解的成本和效益。准确、及时地预报强风暴和沿海洪水，对确保海上活动安全、保护沿海生命财产安全至关重要。为了进一步发展缅因州沿海经济的这些建模工具，缅因大学将大幅增加其高性能计算能力，以满足沿海地区面临的严峻问题。具体来说，缅因大学提高计算能力将使沿海建模者能够提高模型的空间分辨率，从而为当地尺度的决策提供信息（例如，水产养殖租赁），并显著增加本科生和研究生使用高性能计算的机会。因此，可以培训熟悉先进计算的新一代科学家，以开发不断变化的沿海生态系统所需的环境决策支持基础设施。缅因大学在海洋科学研究和教育方面是国际和国内的领导者，因此，依赖沿海资源和基础设施的社区希望大学能够告知他们即将发生的变化。缅因州在地理和经济上都处于独特的位置，容易受到气候变化的影响。该州坐落在世界上温度垂直梯度最大的地区之一，拥有全国最长的海岸线之一。沿海渔业和水产养殖系统的未来是难以预测的。然而，预测未来潜在变化的多种情景的能力将使管理人员和决策者有能力适应。投资于高性能计算有四个原因：(1)空间分辨率的提高将使目前正在开发的模型能够解析更精细的尺度特征，并为小于100米的尺度上做出的许多决策提供信息；(2)当前的高性能计算集群在接近满容量的情况下运行，减少了研究生和本科生的教育机会；(3)与使用更简单的方法相比，使用动态降尺度将气候变化预测与区域尺度联系起来需要更多的处理能力。(4)运行未来变化的多种情景的能力将更充分地表征不确定性。集群系统将有28个节点，每个节点都有两个12核英特尔至强2.6 GHz CPU，这是最新的Haswell v3芯片。这将包括672个内核，并将产生大约28个峰值TFlops。这几乎是缅因大学高性能计算能力的三倍。

The Gulf of Maine is undergoing unprecedented change due to a combination of influences from climatic and anthropogenic sources. By some measures, the Gulf of Maine is among the fastest warming regions of the world's oceans. This warming trend has been concurrent with a 73% increase in extreme events, defined as the top 1% of storms, over the past century. The potential impact of these changes are significant for an economy tightly linked to marine resources and infrastructure. Maine's aquaculture industry (comprising mostly salmon and shellfish) has doubled in value from 2005 to 2013 and is now second only to the lobster industry in the state. Maine's commercial fisheries were valued at a record $585 million dollars in 2014. The effective management of natural resources and infrastructure in the coastal zone requires modeling tools that not only help us understand the ecological and physical consequences of natural and human-induced changes in environmental drivers, but also help us predict those changes and weigh the costs and benefits of adaptation or mitigation. The accurate and timely forecasts of severe storms and coastal inundation are critical to ensure safe maritime activities, protect life and property along the coast. To further the development of these modeling tools for the coastal economy of Maine, the University of Maine will significantly increase its high performance computing to meet the rigors of problems faced by the coastal zone. Specifically, increasing the computing capacity at the University of Maine will allow coastal modelers to increase the spatial resolution of models to inform decisions made on local scales (e.g., aquaculture leases) and significantly increase access to high performance computing for undergraduate and graduate students. Consequently, a new generation of scientists familiar with advanced computing can be trained to develop the environmental decision support infrastructure needed in changing coastal ecosystems. The University of Maine is an established international and national leader in marine science research and education and as such, the community that relies on coastal resources and infrastructure looks to the University to inform them of impending changes. Maine is uniquely positioned physically and economically to be affected by climate change. The state sits on one of the sharpest latitudinal gradients in temperature in the world and has one of the longest coastlines in the country. What the future holds for fisheries and aquaculture systems along the coast is difficult to predict. However, the capacity to forecast multiple scenarios of potential future change will allow managers and decision makers the ability to adapt. There are four reasons for an investment in increased high performance computing: (1) an increase in spatial resolution will allow models currently under development to resolve finer scale features and inform many decisions made on the scale of less than 100 m, (2) the current high performance computing cluster runs at near full capacity, reducing educational opportunities for graduate and undergraduate students, (3) linking climate change projections to the regional scale using dynamic downscaling as opposed to simpler methods requires more processing capacity, and (4) the ability to run multiple scenarios of future change will more adequately characterize uncertainty. The cluster system will have 28 nodes, each with two 12 core Intel Xeon 2.6 GHz CPU's, the latest Haswell v3 chips. This will comprise 672 cores and will yield about 28 peak TFlops. This nearly triples the high performance computing power at the University of Maine.