预测有效波高的深度学习模型研究

研究基于RNN、LSTM、GRU深度学习模型,针对NOAA浮标数据集中的44013、44014、44017浮标的数据,通过斯皮尔曼相关性分析提高模型预测效果。实验结果表明,在进行相关性分析后,S-RNN、S-LSTM、 S-GRU的预测效果均比原始RNN、LSTM、GRU模型预测效果好。此外,提出一种基于LSTM的LSTM-Attention 波高预测模型,并进行相关实验,量化LSTM-Attention模型的预测效果,实验结果表明LSTM-Attention模型有更好的预测效果。为评估模型的泛化能力,研究还提出了一种采用邻近浮标数据进行学习,预测浮标缺失数据的方 法。实验结果表明,该方法的预测精度可以达到97.93%。本研究为海浪预测提供了新的方法和思路,也为未来深 度学习模型在海浪预测中的应用提供了参考。     

Seasonal change of ice algal and phytoplankton assemblages in the Nella Fjord near Zhongshan Station, East Antarctica

The ice algal and phytoplankton assemblages were studied from Nella Fjord near Zhongshan Station, East Antarctica from April 12 to December 30, 1992. Algal blooms occurred about 3 cm thick on the bottom of sea ice in late April and mid November to early December respectively, and a phytoplankton bloom appeared in the underlying surface water in mid December following the spring ice algal bloom. The biomass in ice bottom was 1 to 3 orders of magnitude higher than that of surface water. Amphiprora kjellmanii, Berkeleya sp., Navicula glaciei, Nitzschia barkelyi, N. cylindrus /N. curta, N. lecointei and Nitzschia sp. were common in the sea ice temporarily or throughout the study period. The biomass in a certain ice segment was decreased gradually and the dominant species were usually succeeded as the season went on. Nitzschia sublineata and Dactyliosolen antarctica were two seasonal dominant species only observed in underlying water column. The assemblages between bottom of ice and underlying surface water were different except when spring ice algae bloomed. The evidence shows that the ice algal blooms occurred mainly by in situ growth of ice algae, and the phytoplankton bloom was mostly caused by the release of ice algae.     

利用海洋温度剖面与海表盐度反演盐度剖面方法研究

为解决海洋中大量观测数据只含有温度剖面而缺乏盐度观测的问题, 基于历史观测的温盐剖面资料, 考虑到盐度卫星数据的发展, 采用回归分析方法, 在孟加拉湾建立了盐度与温度、经纬度、表层盐度的关系, 并对不同反演方法的反演结果进行检验评估。结果发现, 在不引入海表盐度(sea surface salinity, SSS)时, 最佳反演模型是温度、温度的二次项与经纬度确定的回归模型, 而SSS的引入则可以进一步优化反演结果。将反演结果与观测结果进行对比, 显示用反演的盐度剖面计算的比容海面高度误差超过2cm, 而引入SSS后的误差低于1.5cm。SSS的引入能够较为真实地反映海洋盐度场的垂直结构和内部变化特征, 既能够捕捉到对上混合层有重要影响的SSS信号, 又能够反映盐度在跃层上的季节内变化以及盐度障碍层的季节变化。水团分析显示, 与气候态相比, 盐度反演结果可以更好地表征海洋上层水团的变化特征。     

雪冰中的黑碳记录研究的历史回顾

黑碳(Blackcarbon)是一类大气气溶胶,也是重要的气候驱动因子之一,它与人类活动密切相关.迄今发现,雪冰是保存和记录历史时期黑碳变化的最好介质.根据前人对南极冰芯、格陵兰冰芯、中低纬度山地冰川冰芯以及雪坑样品中的黑碳记录的研究成果,讨论了不同研究地区雪冰中黑碳含量的变化,黑碳对气候环境突变、大气环流变化的响应,总结了人类活动与雪冰黑碳记录之间的关系.雪冰黑碳记录还可以获得如森林大火这样的特殊事件的信息.     

南极威德尔海冬季冰脊形拖曳力

利用2006年南极威德尔海冬季科学考察期间由机载激光高度计测得的海冰上表面形态数据,以拖曳分割理论为基础研究了冰脊形拖曳力及其对冰-气总拖曳力的贡献以及中性条件下对应10 m高度处风速的冰-气拖曳系数C_dn(10)。结果显示,在密集冰区,冰脊形拖曳力及其对总拖曳力的贡献随冰脊强度(脊高/脊距)的增大呈递增趋势,而随粗糙长度的增大而减小,对应于威德尔海冬季典型冰脊强度和粗糙长度,占总拖曳力的35%,说明了冰脊形拖曳力在冰-气界面动量交换中的重要作用。C_dn(10)随冰脊强度的增大而增大,但冰脊强度较小时,C_dn(10)随粗糙长度增大而增大;而对较大的冰脊强度,C_dn(10)则随粗糙长度增大而减小。     

影响南极海洋石油烃低温降解菌希瓦氏菌NJ49生长和降解率的环境因素研究

以柴油为唯一碳源和能源,从南极海水海冰微生物资源库中筛选到一株石油烃低温降解菌希瓦氏菌NJ49,并对影响其生长和降解率的环境因素(pH、温度、盐度、营养盐和表面活性剂)进行了初步研究。结果表明:希瓦氏菌NJ49可作为低温海域石油烃污染生物修复的菌源,其生长和降解的最适条件为:初始pH7.5,温度15℃,盐度6%,摇瓶装量80ml,最佳氮源硝酸铵,最佳磷源为磷酸二氢钾和磷酸氢二钾的混合物,添加表面活性剂可促进希瓦氏菌NJ49的生长和生物降解率。     

浑江末前寒武纪丝状藻类及其环境意义

本文初步探讨了吉林浑江晚前寒武纪丝状藻类与沉积环境的关系。这些丝状藻类产自晚前寒武纪不同层位,大多属蓝绿藻类。通过对丝状藻类整体面貌和个体丰度的分析,笔者将其划分为三类即:藻席型、孤立生存型和其间的过渡类型。根据对现代丝状藻类生态及环境的研究,采取比较古生物学的方法,具体讨论了浑江地区晚前寒武纪丝状藻类的指相意义。根据沉积学与事件地层记录分析的沉积环境模式与对丝状藻类环境分析的结果是一致的。     

强震作用下冰水域桥墩动力响应研究分析

地震作用下冰水域桥墩受到动水压力与动冰荷载的的双重影响。基于Morison动水理论和Croteau动冰力模型,提出一种适用于冰水域桥墩动力时程反应分析的简便计算模型,通过振动台试验验证了其有效性,并分析了海冰质量和水深对冰海域桥墩动力响应的影响。结果表明,强震作用下有冰时墩底最大曲率比无冰时最大增加8.93倍,桥墩被自由冰包围时,墩底截面出现最大曲率时对应的海冰质量随着水深的增加而变大;随着冰质量的持续增大,墩身最大位移易超越极限位移,造成桥梁垮塌,在桥梁设计中应重点考虑。     

On the stable growth of segregated ice in freezing soil under negligible overburden pressure

The stable growth condition of a segregated ice layer is studied by the use of the principle of mass and heat conservation. This condition evidently depends upon the properties of a thin transitional zone, which is believed to exist between the boundary of an ice layer and a 0[°C] isotherm. All probable models of the transitional zone are classified and the conditions for each model is derived. The effect of the small amount of soil minerals contained in an ice layer is also studied.     

A numerical model of the partitioning of trace chemical solutes during drop freezing

Partitioning of volatile chemicals among the gas, liquid, and solid phases during freezing of liquid water in clouds can impact trace chemical distributions in the troposphere and in precipitation. We describe here a numerical model of this partitioning during the freezing of a supercooled liquid drop. Our model includes the time-dependent calculation of the coupled processes of crystallization kinetics, heat transport, and solute mass transport, for a freezing hydrometeor particle. We demonstrate the model for tracer partitioning during the freezing of a 1000 μm radius drop on a 100 μm ice substrate, under a few ambient condition scenarios. The model effectively simulates particle freezing and solute transport, yielding results that are qualitatively and quantitatively consistent with previous experimental and theoretical work. Results suggest that the ice shell formation time is governed by heat loss to air and not by dendrite propagation, and that the location of ice nucleation is not important to freezing times or the effective partitioning of chemical solutes. Even for the case of nucleation at the center of the drop, we found that dendrites propagated rapidly to form surface ice. Freezing then proceeded from the outside in. Results also indicate that the solid-liquid interfacial surface area is not important to freezing times or the effective partitioning of chemical solutes, and that the rate aspects of trapping are more important than equilibrium solid-liquid partitioning to the effective partitioning resulting from freezing.