BP神经网络在长期天气过程预报中的应用试验

采用误差反传前向网络(简称BP网络)方法,以日、月相概率作为输入因子,建立长期天气预报模型。结果表明,模型的业务预报试验效果比较理想,对较大降水和升(降)温过程均有一定预报能力,相对于传统的单纯运用日、月相概率预报长期天气过程的方法,BP神经网络方法具有预报较客观、准确率较高等特点,在目前长期天气预报理论和数值预报模式尚不能用于实际业务的情况下具有较大的应用价值。     

用卡尔曼滤波制作河南省冬春季沙尘天气短期预报

在分析河南省冬春季沙尘天气气候特征的基础上，筛选出T213数值预报产品中与风速和能见度相关性较好的预报因子，采用卡尔曼滤波方法，分站建立了河南省冬、春季的风速和能见度短期预报方程，并利用相应的沙尘天气分级量化标准，在一定程度上实现了河南省冬、春季沙尘天气的短期分站、分级预报。业务运行时，又增加了两条判别指标进行后处理，有效地消除了液滴对能见度预报的影响，提高了沙尘天气预报的准确率。     

交叉熵神经网络及其在闽北大雨以上降水预报中的应用

基于误差平方和最小化准则的BP神经网络(ANN-MSE)并不适合解决小概率天气事件的预报问题,引进一种改进的以交叉熵函数为目标函数的神经网络方法(ANN-CE),该法是一个三层反向传播神经网络,其输出层只用一个节点.利用2003-2008年的ECMWF预报场资料,把该法用于福建省南平市4-6月部分大雨或以上降水96h预报中,分别用原始因子和PCA降维后的主因子建立了ANN-CE预报模型和ANN-MSE预报模型,用这些模型对2009-2010年独立样本进行了试报.测试结果显示主因子预报模型TS评分比原始因子预报模型高且漏报次数少,其中,主因子ANN-CE预报模型的TS评分和漏报率分别是0.51和0.17,其性能是所有模型中最好且最为稳定的,是一种适合于小概率事件预报的方法.     

北京地区沙尘天气监测预报预警业务系统

在北京地区沙尘天气发生、发展的天气学分析研究和沙尘数值模式预报研究以及卫星遥感监测沙尘等方法的应用基础上 ,对相关的研究成果进行应用开发、集成 ,形成一套完整的业务系统。该系统由沙尘天气历史数据库、沙尘天气动态监测、沙尘天气概念模型、沙尘数值预报和沙尘天气预警 5个子模块组成。在近两年的北京市沙尘天气预报服务中 ,本系统得到充分的应用 ,提高了沙尘天气预报服务的效率 ,在实际业务中发挥了重要作用 ,可应用于气象业务部门及相关部门。     

大连沙尘天气及预报模型分析

通过对位于东北亚沙尘暴天气下游的大连近7年的沙尘天气统计分析,以及应用天气学方法进行预报模型分析,揭示出大连地区沙尘的主要特征和预报着眼点.大连沙尘天气主要出现在春季,变化趋势同全国基本一致;沙尘天气的源地有4个,有三条影响路径.沙尘影响时,平均相对湿度都较低,以西北风为主.西北路径是影响大连地区产生沙尘天气最多的路径.在北路系统影响下,有三种天气类型;西路系统和西北路系统也分别有三种和四种天气类型.大连的沙尘天气预报着眼点首先要关注其上游的起沙情况,然后结合三路影响系统的预报模型进行分析;三路影响系统有不同的起沙关注区和不同的天气系统位置和移动方向.     

不确定性理论——集对分析在MOS概率天气预报中的应用

概率天气预报的关键问题之一是天气预报的不确定性,而集对分析(SPA)为研究天气预报的不确定性问题提供了新的思路和途径。本文针对概率天气预报的特点,应用SPA中联系度的概念,对数值预报产品的预报因子设立中相关概率区间,进行同异反分析,建立了基于SPA的MOS概率天气预报模型。试验表明,联系度概念的引进增强了概率天气预报模型结构的合理性,提高了用数值预报产品制作概率天气预报的准确率。     

全国各省市气象科技期刊文摘⑨

１非参数估计概率天气预报方法的概率基础《军事气象》２００１年第４期赵瑞星总参气象中心１０００８１摘要非参数估计概率天气预报方法是直接基于天气样本数据建立概率天气预报模型的方法，它避免了概率密度估计误差的影响。最近邻相似概率天气预报方法和Ｋ近邻概率天气预报方法是这类预报方法中的两种，实际上窗函数估计概率天气预报方法也属此类方法。许多文献中已较详细地叙述了它们的原理和计算方法，但仍没有说明其理论基础是什么。为此，我们从Ｂａｙｅｓ后验概率公式出发，用频率和窗函数的方法对最近邻和Ｋ近邻权函数概率天气…     

沙尘天气的数值预报

沙尘天气预报的核心问题是要理解沙尘天气形成的物理机制,并定量地描述风沙的分布、强度、输送与沉降.沙尘天气的形成是一个复杂的物理过程,涉及到大气运动与地表状况等诸多因素.作者介绍了一个以数值预报为核心、以卫星观测与地理信息系统为资料背景的沙尘天气集成预报系统,特别是其中的起沙模式.作者运用该系统预报了2002年春季的沙尘天气,并用天气观测和采样观测资料对预报效果进行了鉴定.分析表明,预报系统不仅能定性预报沙尘天气的空间分布与时间演变,而且能定量预报沙尘源地、大气中的沙尘浓度、沙尘的输送以及沉降.模式72 h的预报仍然有较高的可信度.另外讨论了存在的问题和今后的研究方向.       

库尔勒地区沙尘天气的分析与预报

通过对库尔勒机场1991—2000年出现沙尘天气的统计分析，总结出了库尔勒地区沙尘天气的天气、气候特征及预报指标；确定了预报沙尘天气的隶属函数，并对其进行了历史回报。     

内蒙古乌海市沙尘天气潜势预报方法研究

利用1990—2015年内蒙古乌海市逐日地面、高空气象观测资料和NCEP/NCAR FNL再分析资料,分析沙尘天气发生时各预报因子的分布特征,确定预报因子的阈值及消空指标,建立基于PP法和指标叠套法的乌海市沙尘天气潜势预报模型。通过2011—2018年逐日回报检验显示,其TS评分为0.24,准确率为89%,说明该模型对乌海市沙尘天气有较好的判别效果和较好的业务应用价值。     

Rate constants for the reactions of the NO3 radical with HCOOH/HCOO− and CH3COOH/CH3COO− in aqueous solution between 278 and 328 K

The kinetics of the aqueous phase reactions of NO₃ radicals with HCOOH/HCOO^– and CH₃COOH/CH₃COO^– have been investigated using a laser photolysis/long-path laser absorption technique. NO₃ was produced via excimer laser photolysis of peroxodisulfate anions (S₂O ₈ ^2– ) at 351 nm followed by the reactions of sulfate radicals (SO ₄ ^– ) with excess nitrate. The time-resolved detection of NO₃ was achieved by long-path laser absorption at 632.8 nm. For the reactions of NO₃ with formic acid (1) and formate (2) rate coefficients ofk ₁=(3.3±1.0)×10⁵ l mol^–1 s^–1 andk ₂=(5.0±0.4)×10⁷ l mol^–1 s^–1 were found atT=298 K andI=0.19 mol/l. The following Arrhenius expressions were derived:k ₁(T)=(3.4±0.3)×10¹⁰ exp[–(3400±600)/T] l mol^–1 s^–1 andk ₂(T)=(8.2±0.8)×10¹⁰ exp[–(2200±700)/T] l mol^–1 s^–1. The rate coefficients for the reactions of NO₃ with acetic acid (3) and acetate (4) atT=298 K andI=0.19 mol/l were determined as:k ₃=(1.3±0.3)×10⁴ l mol^–1 s^–1 andk ₄=(2.3±0.4)×10⁶ l mol^–1 s^–1. The temperature dependences for these reactions are described by:k ₃(T)=(4.9±0.5)×10⁹ exp[–(3800±700)/T] l mol^–1 s^–1 andk ₄(T)=(1.0±0.2)×10¹² exp[–(3800±1200)/T] l mol^–1 s^–1. The differences in reactivity of the anions HCOO^– and CH₃COO^– compared to their corresponding acids HCOOH and CH₃COOH are explained by the higher reactivity of NO₃ in charge transfer processes compared to H atom abstraction. From a comparison of NO₃ reactions with various droplets constituents it is concluded that the reaction of NO₃ with HCOO^– may present a dominant loss reaction of NO₃ in atmospheric droplets.     

Characteristics of ionic components in precipitation in Kitakyushu City,Japan

Precipitation samples were collected by filtrating bulk sampler in Kitakyushu City, Japan, from January 1988 to December 1990. Volume weighted annual mean of pH was 4.93, but the pH distribution indicated that most probable value lay in the range pH 6.0–6.4. Volume weighted annual mean concentrations of major ionic components were as follows; SO ₄ ^2– : 84.2, NO ₃ ^– : 28.1, Cl^–: 86.3, NH ₄ ⁺ : 45.5, Ca²⁺: 63.3, Mg²⁺: 27.0, K⁺: 3.4, Na⁺: 69.0 µ eq l^–1. The highest concentrations of these ionic components were observed in winter and the lowest occurred in the rainy season. The ratio of ex-SO ₄ ^2– /NO ₃ ^– exhibited the lowest ratio in summer, and the highest ratio in winter. Good correlations were obtained between Cl^– and Na⁺, ex-SO ₄ ²⁺ and ex-Ca²⁺, NO ₃ ^– and ex-Ca²⁺, and NH ₄ ⁺ and ex-SO ₄ ^2– , respectively. However, no correlation between Cl^– and Na⁺ with Ca²⁺ was observed. The relationship of H⁺ with (ex-SO ₄ ^2– + NO ₃ ^– ) - (ex-Ca²⁺ + NH ₄ ⁺ ) indicated positive correlation.     

Nitrogen and sulfur species in Antarctic aerosols at Mawson,Palmer Station,and Marsh (King George Island)

High volume bulk aerosol samples were collected continuously at three Antarctic sites: Mawson (67.60° S, 62.50° E) from 20 February 1987 to 6 January 1992; Palmer Station (64.77° S, 64.06° W) from 3 April 1990 to 15 June 1991; and Marsh (62.18° S, 58.30° W) from 28 March 1990, to 1 May 1991. All samples were analyzed for Na⁺, SO ₄ ^2– , NO ₃ ^– , methanesulfonate (MSA), NH ₄ ⁺ ,²¹⁰Pb, and⁷Be. At Mawson for which we have a multiple year data set, the annual mean concentration of each species sometimes vary significantly from one year to the next: Na⁺, 68–151 ng m^–3; NO ₃ ^– , 25–30 ng m^–3; nss SO ₄ ^2– , 81–97 ng m^–3; MSA, 19–28 ng m^–3; NH ₄ ⁺ , 16–21 ng m^–3;²¹⁰Pb, 0.75–0.86 fCi m^–3. Results from multiple variable regression of non-sea-salt (nss) SO ₄ ^2– with MSA and NO ₃ ^– as the independent variables indicates that, at Mawson, the nss SO ₄ ^2– /MSA ratio resulting from the oxidation of dimethylsulfide (DMS) is 2.80±0.13, about 13% lower than our earlier estimate (3.22) that was based on 2.5 years of data. A similar analysis indicates that the ratio at Palmer is about 40% lower, 1.71±0.10, and more comparable to previous results over the southern oceans. These results when combined with previously published data suggest that the differences in the ratio may reflect a more rapid loss of MSA relative to nss SO ₄ ^2– during transport over Antarctica from the oceanic source region. The mean²¹⁰Pb concentrations at Palmer and Marsh and the mean NO ₃ ^– concentration at Palmer are about a factor of two lower than those at Mawson. The²¹⁰Pb distributions are consistent with a²¹⁰Pb minimum in the marine boundary layer in the region of 40°–60° S. These features and the similar seasonalities of NO ₃ ^– and²¹⁰Pb at Mawson support the conclusion that the primary source regions for NO ₃ ^– are continental. In contrast, the mean concentrations of MSA, nss SO ₄ ^2– , and NH ₄ ⁺ at Palmer are all higher than those at Mawson: MSA by a factor of 2; nss SO ₄ ^2– by 10%; and NH ₄ ⁺ by more than 50%. However, the factor differences exhibit substantial seasonal variability; the largest differences generally occur during the austral summer when the concentrations of most of the species are highest. NH ₄ ⁺ /(nss SO ₄ ^2– +MSA) equivalent ratios indicate that NH₃ neutralizes about 60% of the sulfur acids during December at both Mawson and Palmer, but only about 30% at Mawson during February and March.     

Processing of oxidised nitrogen compounds by passage through winter-time orographic cloud

Four case studies are described, from a three-site field experiment in October/November 1991 using the Great Dun Fell flow-through reactor hill cap cloud in rural Northern England. Measurements of total odd-nitrogen nitrogen oxides (NO _y ) made on either side of the hill, before and after the air flowed through the cloud, showed that 10 to 50% of the NO _y , called NO _z , was neither NO nor NO₂. This NO _z failed to exhibit a diurnal variation and was often higher after passage through cloud than before. No evidence of conversion of NO _z to NO₃ ^- in cloud was found. A simple box model of gas-phase chemistry in air before it reached the cloud, including scavenging of NO₃ and N₂O₅ by aerosol of surface area proportional to the NO₂ mixing ratio, shows that NO₃ and N₂O₅ may build up in the boundary layer by night only if stable stratification insulates the air from emissions of NO. This may explain the lack of evidence for N₂O₅ forming NO₃ ^- in cloud under well-mixed conditions in 1991, in contrast with observations under stably stratified conditions during previous experiments when evidence of N₂O₅ was found. Inside the cloud, some variations in the calculated total atmospheric loading of HNO₂ and the cloud liquid water content were related to each other. Also, indications of conversion of NO _x to NO _z were found. To explain these observations, scavenging of NO _x and HNO₂ by cloud droplets and/or aqueous-phase oxidation of NO₂ ^- by nitrate radicals are considered. When cloud acidity was being produced by aqueous-phase oxidation of NO _x or SO₂, NO₃ ^- which had entered the cloud as aerosol particles was liberated as HNO₃ vapour. When no aqueous-phase production of acidity was occurring, the reverse, conversion of scavenged HNO₃ to particulate NO₃ ^-, was observed.     

Measurement of the photodissociation coefficient of NO2 in the atmosphere: II,stratospheric measurements

The photodissociation coefficient of NO₂, J _{NO 2}, has been measured from a balloon platform in the stratosphere. Results from two balloon flights are reported. High Sun values of J _{NO 2} measured were 10.5±0.3 and 10.3±0.3×10^-3 s^-1 at 24 and 32 km respectively. The decrease in J _{NO 2} at sunset was monitored in both flights. The measurements are found to be in good agreement with calculations of J _{NO 2} using a simplified isotropic multiple scattering computer routine.     

Kinetics of the reactions of hydroxyl radicals with aliphatic ethers studied under simulated atmospheric conditions: Temperature dependences of the rate coefficients

Rate coefficients have been measured for the reactions of hydroxyl radicals with five aliphatic ethers over the temperature range 242–328 K. Competitive studies were carried out in an atmospheric flow reactor in which the hydroxyl radicals were generated by the photolysis of methyl nitrite in the presence of air containing nitric oxide. The reaction of OH with 2,3-dimethyl-butane was used as the reference reaction and the following Arrhenius parameters have been obtained for the reactions: OH+RORproducts:

The influence of light intensity,SO2, NOx and C3H6 on sulfate aerosol formation

The photochemical oxidation of SO₂ in the presence of NO and C₃H₆ was studied in a 18.2 liter pyrex reactor. When light intensity, irradiation time and SO₂ concentration were constant, SO₄ ^2- concentration, derived from the total volume of aerosol produced, peaked when [C₃H₆]/[NO] was approximately 6.0. Another increase im SO₄ ^2- formation was reached at very high ratios (>50). The experimental observations are consistent with the two SO₂ oxidation mechanisms. At low [C₃H₆]/[NO] ratios, the processes proceed via the HO–SO₂ reaction, while at high ratios the O₃–C₃H₆ adduct is assumed to oxidize SO₂ to produce SO₄ ^2- aerosols.     

Kinetics of the Gas-Phase Reactions of OH Radicals, NO3 Radicals and O3 with Three C7-Carbonyls Formed From The Atmospheric Reactions of Myrcene, Ocimene and Terpinolene

Rate constants for the gas-phase reactions of OH radicals, NO₃ radicals and O₃ with the C₇-carbonyl compounds 4-methylenehex-5-enal [CH₂=CHC(=CH₂)CH₂CH₂CHO], (3Z)- and (3E)-4-methylhexa-3,5-dienal [CH₂=CHC(CH₃)=CHCH₂CHO] and 4-methylcyclohex-3-en-1-one, which are products of the atmospheric degradations of myrcene, Z- and E-ocimene and terpinolene, respectively, have been measured at 296 ± 2 K and atmospheric pressure of air using relative rate methods. The rate constants obtained (in cm³ molecule^–1 s^–1 units) were: for 4-methylenehex-5-enal, (1.55 ± 0.15) × 10^–10, (4.75 ± 0.35) × 10^–13 and (1.46 ± 0.12) × 10^–17 for the OH radical, NO₃ radical and O₃ reactions, respectively; for (3Z)-4-methylhexa-3,5-dienal: (1.61 ± 0.35) × 10^–10, (2.17 ± 0.30) × 10^–12, and (4.13 ± 0.81) × 10^–17 for the OH radical, NO₃ radical and O₃ reactions, respectively; for (3E)-4-methylhexa-3,5-dienal: (2.52 ± 0.65) × 10^–10, (1.75 ± 0.27) × 10^–12, and (5.36 ± 0.28) × 10^–17 for the OH radical, NO₃ radical and O₃ reactions, respectively; and for 4-methylcyclohex-3-en-1-one: (1.10 ± 0.19) × 10^–10, (1.81 ± 0.35) × 10^–12, and (6.98 ± 0.40) × 10^–17 for the OH radical, NO₃ radical and O₃ reactions, respectively. These carbonyl compounds are all reactive in the troposphere, with daytime reaction with the OH radical and nighttime reaction with the NO₃ radical being predicted to dominate as loss processes and with estimated lifetimes of about an hour or less.