一致性哈希算法在数据库集群上的拓展应用

在数据库集群的研究中,可扩展性是一个重要的性能指标。为实现在数据高速增长或部分集群服务器故障情况下,数据依然能够快速、可靠、安全的分布到新的集群服务器节点上的目的,就必须合理设置数据划分的策略。将Key-Value存储技术中使用的一致性哈希算法思想借鉴运用到并行分析型数据库集群中,提出针对大规模结构化类特殊数据的一致性哈希划分方法,并在MapReduce框架下设计具体的数据划分方案。最后,以TPC-DS作为测试基准,与同类系统进行性能对比测试,实验结果表明方案不仅有良好的划分性能,且扩展性也较好。     

一致性哈希算法在数据库集群上的拓展应用

在数据库集群的研究中,可扩展性是一个重要的性能指标.为实现在数据高速增长或部分集群服务器故障情况下,数据依然能够快速、可靠、安全的分布到新的集群服务器节点上的目的,就必须合理设置数据划分的策略.将Key-Value存储技术中使用的一致性哈希算法思想借鉴运用到并行分析型数据库集群中,提出针对大规模结构化类特殊数据的一致性哈希划分方法,并在MapReduce框架下设计具体的数据划分方案.最后,以TPC-DS作为测试基准,与同类系统进行性能对比测试,实验结果表明方案不仅有良好的划分性能,且扩展性也较好.     

城市创伤中心空间选址的层次分析模型

针对城市创伤中心空间选址问题,建立包含四大类、八小类影响因素的层次分析模型,邀请不同领域的专家对构造出的判断矩阵综合打分,得出各因素的相应权重后运用到深圳市创伤中心空间选址的实例研究上。模型中所涉及的空间数据均基于网络分析,更贴合实际情况。同时,考虑到深圳市创伤中心的依托医院相对集中这一特殊情况,通过聚类分析划分区域后再使用层次分析模型选址。优化后的结果将改善深圳市创伤医疗救治资源配置、提高创伤救治效率,为创伤救治机构规划部门提供参考意见。     

2008年1月贵州冻雨的数值模拟和层结结构分析

针对2008年初发生在贵州地区的严重冻雨过程,分别从环流背景、低空急流和水汽输送条件等方面分析了准静止锋维持的原因,并选取本次灾害最严重的第3次过程为典型个例,利用WRF模式针对准静止锋影响下的贵州冻雨进行数值模拟来研究冻雨的发生机制。模拟结果较好地反映出高低空环流形势场特征,强雨雪降水带的走向、落区,以及地面温度的分布,均与观测基本吻合。通过分析高分辨率模式的模拟结果,揭示了准静止锋上贵州地区冻雨的层结结构特征及云物质在冻雨区的分布特征。研究结果表明,贵州中部的冻雨区除一般的三层结构(包含冰晶层、暖层和冷层)外,还具有典型的两层结构特征,即:高空的固体降水粒子稀少,900~600 h Pa深厚的逆温层和0℃以上的暖层使中低空存在大量液态粒子,下落的液滴经过近地面的浅薄冷层,形成大量过冷却雨滴,而后降落至地面迅速冻结。     

基于双多普勒雷达及闪电资料分析一次强降水过程

为进一步了解多普勒雷达及闪电资料在中小尺度天气系统方面的应用以及降水系统动力结构的研究,利用合肥和阜阳两部S波段多普勒雷达资料,通过两步变分反演法反演双雷达,并采用模糊逻辑算法进行云类型识别,再结合垂直累计液态水含量VIL、降水和闪电分布特征,分析了2010年6月7-10日发生在安徽合肥和阜阳的一次持续性强降水过程。结果表明:此次持续性强降水过程与风场的低层辐合和高层辐散、风场的气旋性旋转以及水平风切变相关;此次降水过程是以层状云降水为主的积云混合云降水;另外,以降水为主的层状云,其闪电频数低且以负闪为主,闪电之后20 min左右出现强降水;此次暴雨的VIL值不是很大,强降水区对应的VIL值基本3 kg·m-2;云识别结果、降水区、VIL值与雷达回波强度之间的对应关系较好。     

Effects of prescribed initial cloud droplet spectra on convective cloud and precipitation developments under different thermodynamic conditions: A modeling and observational study

A two-dimensional cloud model with bin microphysics was used to investigate the effects of cloud condensation nuclei (CCN) concentrations and thermodynamic conditions on convective cloud and precipitation developments. Two different initial cloud droplet spectra were prescribed based on the total CCN concentrations of maritime (300 cm^− 3) and continental (1000 cm^− 3) air masses, and the model was run on eight thermodynamic conditions obtained from observational soundings. Six-hourly sounding data and 1-hourly precipitation data from two nearby weather stations in Korea were analyzed for the year 2002 to provide some observational support for the model results.For one small Convective Available Potential Energy (CAPE) ( 300 J kg^− 1) sounding, the maritime and continental differences were incomparably large. The crucial difference was the production of ice phase hydrometeors in the maritime cloud and only water drops in the continental cloud. Ice phase hydrometeors and intrinsically large cloud drops of the maritime cloud eventually lead to significant precipitation. Meanwhile negligible precipitation developed from the continental cloud. For the three other small CAPE soundings, generally weak convective clouds developed but the maritime and continental clouds were of the same phases (both warm or both cold) and their differences were relatively small.Model runs with the four large CAPE ( 3000 J kg^− 1) soundings demonstrated that the depth between the freezing level (FL) and the lifting condensation level (LCL) was crucial to determine whether a cloud becomes a cold cloud or not, which in turn was found to be a crucial factor to enhance cloud invigoration with the additional supply of freezing latent heat. For two large CAPE soundings, FL–LCL was so deep that penetration of FL was prohibitive, and precipitation was only mild in the maritime clouds and negligible in the continental clouds. Two other soundings of similarly large CAPE had small FL–LCL, and both the maritime and continental clouds became cold clouds. Precipitation was strong for both but much more so in the maritime clouds, while the maximum updraft velocity and the cloud top were slightly higher in continental clouds. Although limited to small CAPE cases, more precipitation for smaller FL–LCL for a selected group of precipitation and thermodynamic sounding data from Korea was in support of these model results in its tendency.These results clearly demonstrated that the CCN effects on cloud and precipitation developments critically depended on the given thermodynamic conditions and not just the CAPE but the entire structure of the thermodynamic profiles had to be taken into account.     

安徽铜陵凤凰山夕卡岩型铜矿床中辉钼矿Re-Os同位素年龄及其地质意义

铜陵矿集区是我国长江中下游Cu-Au-Fe-Mo成矿带中最重要的有色金属基地之一,凤凰山矿床是铜陵矿集区的重要组成部分,为一个典型的夕卡岩型铜矿床。本文利用Re-Os同位素定年方法对凤凰山铜矿床进行了成矿时代测定,获得了辉钼矿的Re-Os同位素模式年龄范围为139.1±2.4～142.0±2.2Ma,等时线年龄为141.1±1.4Ma,与矿区内石英二长闪长岩和花岗闪长岩SHRIMP锆石U-Pb年龄(144.2±2.3Ma)相吻合,也与铜陵地区其他矿田的成矿时代基本一致,可能为岩石圈减薄事件的成矿响应。     

武汉市不同站域建成环境与轨道交通站点客流特征关系分析

中国已经成为全世界城市轨道交通建设里程最长、建设速度最快的国家,合理识别城市轨道站点类别与影响不同类别站点客流特征的建成环境因素对轨道交通的建设具有十分重要的作用。因此,本文以武汉市为例,通过轨道交通刷卡数据,运用引入客流特征的EM聚类方法,将轨道交通站点分为职住错位型、居住导向型、就业导向型、居住综合型、就业综合型、综合型6类。并在此基础上,建立无序多分类logistic回归模型,定量分析站点客流吸引范围内建成环境因素对不同类型站点轨道交通客流特征的影响。结果表明：以综合型站点为对照组,路网密度和交叉口密度对所有类型站点的客流特征有显著影响,低路网密度高交叉口密度更有利于职住平衡;公交车站密度与就业导向型站点的客流特征存在负相关;商务用地占比对职住错位型、居住导向型、居住综合型站点有负向关系;服务业设施用地占比与职住错位型、居住导向型和就业综合型站点的客流特征负相关;科研教育用地占比与居住导向型和就业导向型站点客流特征负相关;土地利用混合度与居住导向型和就业综合型站点的客流特征负相关。研究结论将对武汉市轨道交通建设、轨道交通和土地利用协调发展等具有重要意义。     

Relationship between surface pollen assemblages and vegetation in Luonan Basin,Eastern Qinling Mountains,Central China

The catchment of South Luohe River in Central China is an important region for investigating modern pollen-environment relationship, because it is located in the transitional zone between south and north China, an environment which is sensitive to climate changes. In this study, 40 surface samples under ten vegetation types were collected to reveal the relationship between pollen assemblages and vegetation. The results show that the surface pollen assemblages reflect the vegetation quite well. In forest topsoils, the average of arboreal pollen content is greater than 40%, and the Selaginella sinensis spore is high. As to sparse forest grassland and shrub community, the average arboreal pollen is 13.2% and 16.6% respectively, and the shrub pollen is relatively higher than that of grassland samples. The grassland and farmland are characterized by low percentage of tree and shrub pollen (<10% and <1%), and high percentage of herbs (>80%). Pinus, Quercus and some other arboreal pollen can indicate the regional vegetation because of their dispersal ability. Quercus pollen is under-representative and so is Pinus. Artemisia pollen is significantly over-represented, has poor correlation with the plant coverage, and may reflect human disturbance. Gramineae can indicate plant quite well, but with low representation. High content of Chenopodiaceae probably suggests human impact. Predominant Selaginella sinensis can be used as an indicator of forest environment. Cluster analysis and principal components analysis of pollen assemblages can distinguish forest and non-forest vegetation well. The former method is better at separating pine and mixed forests, while the latter is more stable and could better differentiate farmland and other non-forest area. The first axis of PCA mainly reflects the humidity.     

Using a Bayesian estimator to combine information from a cluster analysis and remote sensing data to estimate high-resolution data for agricultural production in Germany

In Germany, a county-resolution data set that consists of 35 land-use and animal-stock categories has been used extensively to assess the impact of agriculture on the environment. However, because such environmental effects as emission or nutrient surplus depend on the location, even a county resolution might produce misleading results. The aim of this article is to propose a Bayesian approach which combines two sorts of information, with one being treated as defining the prior and the other the data to form a posterior, used to estimate a data set at a municipality resolution. We define the joint prior density function based on (i) remote sensing data, thus accounting for differences in county data and missing data at the municipality level, and (ii) the results of a cluster analysis that was previously applied to the micro-census, whereas the data are defined by official statistics at the county level. This approach results in a fairly accurate data set at the municipality level. The results, using the proposed method, are validated by the national research data centre by comparing the estimates to actual observations. The test statistics presented here demonstrate that the proposed approach adequately estimates the production activities.