基于梯度提升算法的岩性识别方法

传统的岩性识别方法如岩屑录井、钻井取心及测井资料解释等技术,对录井质量的依赖程度较高,识别精度与效率低,泛化能力差。随着计算机技术的迅速发展,将测井资料与计算机技术相结合开展岩性研究已成为岩性识别的有效手段。本文提出了一种基于梯度提升算法XGBoost和LightGBM的岩性识别方法。以苏里格气田苏东41-33区块下碳酸盐岩储层为例进行测试验证,采用该方法结合测井资料中的声波时差、自然伽马、光电吸收截面指数、密度、深侧向电阻率和补偿中子等6种参数进行岩性识别,并与KNN （K近邻分类器）、朴素贝叶斯和支持向量机等传统算法进行对比,结果表明,3种传统算法的岩性识别准确率分别为78.45%、74.43%和78.72%,基于梯度提升算法XGBoost和LightGBM的识别准确率分别达到了98.90%和98.72%,远高于传统算法。     

CONDITIONAL NONLINEAR OPTIMAL PERTURBATIONS: APPLICATION IN TROPICAL CYCLONE FORECASTS

Considering the feature of tropical cyclones (TCs) that strong positive vorticity exists in the lower layers of troposphere, this study proposed to use vorticity at 850 hPa as cost function to find the conditional nonlinear optimal perturbation (CNOP), which was largely different from those previous studies using total energy of perturbed forecast variables. The CNOP was obtained by an ensemble-based approach. All of the sensitive areas determined by CNOP with vorticity at 850 hPa as cost function for the three cases were located over the TC core region and its vicinity. The impact of the CNOP-based adaptive observations on TC forecasts was evaluated with three cases via observational system simulation experiments (OSSEs). Results showed obvious improvements in TC intensity or track forecasts due to the CNOP-based adaptive observations, which were related to the main error source of the verification area, i.e., intensity error or location error.     

定量研究雅安地形坡向坡度对降水分布的影响

利用四川省雅安市30 m分辨率基础高程数据,提取栅格的坡向和坡度参数,将雅安307个区域自动站在2017年汛期(6—9月)共50次的降水天气个例,分为16次大尺度降水和34次中小尺度降水,使用对应时次的欧洲中心细网格0.25°×0.25°再分析风场资料,根据不同的站点地形高度将风场合成平均风场,和各站点地形的坡向和坡度计算出其动力抬升作用,同时使用当天日照和天文太阳辐射值来计算地形的热力抬升作用,与对应降水过程的降水分布进行多元线性回归,根据回归的标准系数的大小确定各自变量对降水分布的影响,得出以下结论:(1)中小尺度降水中,地形的热力抬升作用对降水分布的影响作用最大,其次是海拔高度,地形的动力抬升作用在三者中对降水分布的影响最小;(2)在大尺度降水中刚好相反,地形的动力抬升对降水的分布影响作用最大,海拔次之,热力抬升作用在三者中影响作用最小;(3)日降水量最大值的站点海拔高度基本位于1 000 m左右,与抬升凝结高度对应较好;(4)从长期的统计来看,地形的动力作用和地表的植被情况对降水分布的影响最大。在实际预报工作用,根据不同的降水类型,关注不同的动力和热力作用对于判断降水分布大值区的位置有较好的参考作用。     

Uncertainty of the Linear Trend in the Zonal SST Gradient Across the Equatorial Pacific Since 1881

The change in the zonal sea surface temperature gradient (ZSSTG) across the equatorial Pacific plays an important role in the global climate system. However, there has not yet been a consensual conclusion about the changing ZSSTG at either a short-term (from 20 to 90 years) or a long-term time scale (longer than 90 years) in the literature. In this study, the uncertainty of the trend in ZSSTG for different sub-periods since 1881 was examined using four interpolated datasets and four un-interpolated datasets. It was found that the trend in ZSSTG on the short-term time scale could be significantly influenced by internal variability such as the El Niño–Southern Oscillation and the Pacific Decadal Oscillation. On the long-term time scale, the sign of the ZSSTG trend depends on the dataset used. In particular, it was not possible to draw a uniform conclusion about the secular trends in ZSSTG in recent history, given the high sensitivity of the ZSSTG trends to the period, dataset, and regions used to calculate the trends. Our results imply that it may not be possible to detect the response of ZSSTG to global warming until a longer data record becomes available in the future.     

System identification for nonlinear maneuvering of large tankers using artificial neural network

This paper deals with the application of nonparametric system identification to a nonlinear maneuvering model for large tankers using artificial neural network method. The three coupled maneuvering equations in this model for large tankers contain linear and nonlinear terms and instead of attempting to determine the parameters (i.e. hydrodynamic derivatives) associated with nonlinear terms, all nonlinear terms are clubbed together to form one unknown time function per equation which are sought to be represented by the neural network coefficients. The time series used in training the network are obtained from simulated data of zigzag maneuvers and the proposed method has been applied to these data. The neural network scheme adopted in this work has one middle or hidden layer of neurons and it employs the Levenberg–Marquardt algorithm. Using the best choices for the number of hidden layer neurons, length of training data, convergence tolerance etc., the performance of the proposed neural network model has been investigated and conclusions drawn.     

Environmental limitations on recruitment from seed in invasive Spartina densiflora on a southern European salt marsh

The South American cordgrass, Spartina densiflora, has invaded a range of different habitats that can support different native species assemblages on salt marshes in the Gulf of Cadiz, Spain. Little is known about the mechanisms of invasion. We examined the potential for seed germination and recruitment in a field transplant experiment, representing a wide range of environments, on elevational gradients across marshes with muddy and sandy sediments. The biotic resistance of native, perennial vegetation (where present) to recruitment of the alien was also investigated. Spartina densiflora seeds were able to germinate over a greater than 2-m range of elevation in the tidal frame. Germination success on unvegetated muddy sediments was related to sediment redox potential, with poor germination at strongly negative redox potentials on the lower sites. On sandy, well-drained sediments, germination was apparently constrained by water availability at the highest elevations. Comparison of vegetated and cleared plots on the upper marsh showed that there was a negative relationship between the presence of Atriplex portulacoides and germination on the muddy sediments. Recruitment (survival of seedlings for 12 weeks) was seen only on unvegetated muddy sediments at the highest elevation. Hence the invasive success and wide elevational tolerance of S. densiflora on the marshes of the Gulf of Cadiz are not reflected in its short-term ability to become established from its prolific seed production. Colonization of sub-optimal habitats may be largely by vegetative propagules and clonal growth.     

Dissolved inorganic carbon (DIC) and its carbon isotopic composition in sediment pore waters from the Shenhu area,northern South China Sea

The Shenhu area is one of the most favorable places for the occurrence of gas hydrates in the northern continental slope of the South China Sea. Pore water samples were collected in two piston cores (SH-A and SH-B) from this area, and the concentrations of sulfate and dissolved inorganic carbon (DIC) and its carbon isotopic composition were measured. The data revealed large DIC variations and very negative δ ¹³C-DIC values. Two reaction zones, 0–3 mbsf and below 3 mbsf, are identified in the sediment system. At site SH-A, the upper zone (0–3 mbsf) shows relatively constant sulfate and DIC concentrations and δ ¹³C-DIC values, possibly due to bioturbation and fluid advection. The lower zone (below 3 mbsf) displays good linear gradients for sulfate and DIC concentrations, and δ ¹³C-DIC values. At site SH-B, both zones show linear gradients, but the decreasing gradients for δ ¹³C-DIC and SO₄ ²⁻ in the lower zone below 3 mbsf are greater than those from the upper zone, 0–3 mbsf. The calculated sulfate-methane interface (SMI) depths of the two cores are 10.0 m and 11.1 m, respectively. The depth profiles of both DIC and δ ¹³C-DIC showed similar characteristics as those in other gas hydrate locations in the world oceans, such as the Blake Ridge. Overall, our results indicate an anaerobic methane oxidation (AMO) process in the sediments with large methane flux from depth in the studied area, which might be linked to the formation of gas hydrates in this area.     

An approximated method for the solution of elliptic problems in thin domains: Application to nonlinear internal waves

Realistic numerical simulations of nonlinear internal waves (NLIWs) have been hampered by the need to use computationally expensive nonhydrostatic models. In this paper, we show that the solution to the elliptic problem arising from the incompressibility condition can be successfully approximated by a few terms (three at most) of an expansion in powers of the ratio (horizontal grid spacing)/(total depth). For an n dimensional problem, each term in the expansion is the sum of a function that satisfies a one-dimensional second-order ODE in the vertical direction plus, depending on the surface boundary condition, the solution to an n-1 dimension elliptic problem, an evident saving over having to solve the original n-dimensional elliptic problem. This approximation provides the physically correct amount of dispersion necessary to counteract the nonlinear steepening tendency of NLIWs. Experiments with different types of NLIWs validate the approach. Unlike other methods, no ad hoc artificial dispersion needs to be introduced.     

“源-汇”景观格局的热岛效应研究——以武汉市为例

论文基于2017年Landsat 8遥感影像,通过GIS技术选取城市地表温度、源汇景观密度、源汇景观贡献度和景观效应指数,结合地理空间分析和数理统计方法,分析了武汉市局部源汇景观格局对城市地表温度的贡献及其效应。结果表明：① 武汉市城区热岛(urban heat island, UHI)季节变化明显,夏季热岛强度大、范围广,呈“大”字型分布,工业集聚区和经济开发区产生的UHI更显著;② 武汉市城区平均地表温度(land surface temperature, LST)在距市中心约8 km处达到峰值后沿城乡梯度逐渐下降,呈现“倒钩型”特征,这一现象与景观密度变化密切相关;③ LST与源(正向)、汇(负向)景观密度存在极强相关性,不透水表面是城区热岛的主要贡献因子;④ 相同密度的水体和绿地,水体更具有削弱城市热岛效应的功能;⑤ 一般当汇-源景观面积比>0.89(汇区密度>0.47)时,局部景观格局对城市热岛有缓解作用,且局部汇区越集中缓解效果越显著。     

基于共轭梯度搜索的病态问题处理方法

在最小二乘平差准则基础上,把病态平差问题转化为无约束的二次规划问题,并利用优化理论分析病态对平差解的影响。通过共轭梯度搜索算法在可行域中寻找最优步长因子,自动寻找最速下降方向,并给出迭代初值的设置方法。分析近似计算中病态问题与局部最优解的关系,讨论局部最优解的快速迭代方法,并通过实例验证算法的有效性,计算迭代的速度。由于整个过程没有对法方程系数矩阵进行求逆计算,该算法可用于处理大规模系数矩阵高病态的平差问题。