基于MCMC模拟的MGPD模型及其在地质灾害风险度量中的应用

基于马尔科夫链蒙特卡洛(简记为MCMC)模拟的参数贝叶斯估计,对改进的广义帕累托分布(简记为MGPD)模型进行了优化,并利用该模型得到了地质灾害损失的在险损失值(简记为VaR)和条件损失值(简记为CVaR).以湖南娄底市地质灾害损失数据进行实证分析及模型适应性检验,结果表明:优化后的模型不仅具有很好的极值数据描述能力,而且具有较强的适用性.     

中国农业面源污染排放的空间差异及其动态演变

以2003—2014年为研究时段,在测算农业面源污染排放强度的基础上,综合运用基尼系数和非参数估计方法,研究我国农业面源污染的空间差异及其动态演变,结果表明:1)我国农业面源污染排放总体下降,且表现出明显的空间差异,东部和中部地区的排放强度较高,西部和东北地区则相对较低。2)2003—2014年,中国农业面源污染排放强度的区域差异略微扩大,地区间差异是其总体差异的主要来源。3)核密度估计结果表明,中国农业面源污染排放总体差异表现为"下降-上升-下降"的波动变化趋势。4)马尔科夫链分析表明,中国农业面源污染在不同类型间的相互流动较为微弱,但从长期来看,存在向两极分化发展的趋势特征。     

具有交易费用和马氏调制的均值-方差组合选择

研究了均值-方差准则下，最优投资组合选择问题.投资者为了增加财富它可以在金融市场上投资.金融市场由一个无风险资产和n个带跳的风险资产组成，并假设金融市场具有马氏调制，买卖风险资产时，考虑交易费用.目标是，在终值财富的均值等于d的限制下，使终值财富的方差最小，即均值-方差组合选择问题.应用随机控制的理论解决该问题，获得了最优的投资策略和有效边界.     

A.A. Marcov Forecast Method of the System Ration in Estate Business Study

A. A. Markov Rolling Forecast Method is presented in this paper according to the dynamic characteristics of the estate business system. Test results reveal that the system quantitative forecast could be realized on the basis of previous samples, which provide a direction of the rational management in estate business. Concerning sample molding analysis and Markov Rolling Forecast Method, this paper gives a corresponding conclusion of analytic research quantitatively in order to demonstrate the formation of factors working in estate business activities and its systematic logistic connotation. Markov Rolling Forecast Method can be applied not only to short-term prediction but long term one as well, provided when the market is rather stable with no significant change during a regular period of time and when the rolling times of the state transition matrix are fairly enough.     

城市工业用水量的灰色马尔可夫预测模型

灰色GM(1,1)是预测城市工业用水量的模型,这种模型不适合长期的、随机和波动性较大的数据序列预测,但是马尔可夫模型适合描述随机波动性较大的预测问题.可以将这两种模型结合,构建灰色马尔可夫预测模型.按特定的状态划分方法,先用灰色GM(1,1)预测模型进行预测,再用马尔可夫模型预测结果进行优化,使预测精度大大提高.最后以抚顺市为例,预测结果证明了该模型的优势.     

基于GIS与马尔可夫模型的渭河盆地景观动态变化研究

基于1988年与2003年两期TM影像,综合遥感与GIS技术以及景观生态学理论,深入分析了渭河盆地景观变化的数量和空间特征以及由此所引起的一系列生态环境效应,主要表现在:耕地、园地与城镇建设用地面积显著增加,林地、水体和未利用地面积呈减少趋势,城镇建设用地的增加是以侵占大量农田为代价的,而园地与耕地的增加则主要是通过开荒的方式。景观形状趋于复杂,破碎度降低,异质性下降,景观类型有向单一化或非均衡化方向发展的趋势,表明人类的开发活动增强,这样必然导致景观稳定性下降,因而降低了景观抗干扰的能力。最后采用马尔可夫模型对景观的变化趋势进行了预测。     

松嫩平原西部荒漠化景观动态及驱动力研究-以吉林省大安市为研究区域

以松嫩平原西部荒漠化的典型区域大安市为研究对象,在GIS平台下,根据1986年,1996年和2000年TM影像解译获取的景观动态数据,分析近15年大安市荒漠化土地动态变化规律。并应用Markov模型预测大安市景观格局动态,结果表明如不采取合理有效的措施,2050年荒漠化土地将占土地总面积的42.30%,高-中覆盖度草地,水域,沼泽将大面积退化,分别仅占1986年水平的48.69%,24.71%和22.74%。在此基础之上,运用灰色关联方法分析了土地荒漠化的主要驱动因子。近15年来筑路、开荒等不合理的社会经济活动是土地荒漠化的主要驱动因子;气候因子的变动有利于土地荒漠化过程的发展,是土地荒漠化过程的背景因子;水文因子迭加了人类的无序开发、利用,是土地荒漠化过程的重要影响因子。     

Testing assumptions of cost surface analysis—a tool for invasive species management

Applied ecology could benefit from new tools that identify potential movement pathways of invasive species, particularly where data are sparse. Cost surface analysis (CSA) estimates the permeability (friction) across a landscape and can be applied to dispersal modelling. Increasingly used in a diversity of applications, several fundamental assumptions that might influence the outputs of CSA (cost surfaces and least-cost pathways) have yet to be systematically examined. Thus, we explore two issues: the presumed relationship between habitat preferences and dispersal behaviour as well as the degree of landscape fragmentation through which an organism moves by modelling a total of 18 sensitivity and dispersal scenarios. We explored the effect of fragmentation by altering the friction values (generally assigned using expert opinion) associated with patch and linear features. We compared these sensitivity scenarios in two sites that differed in fragmentation. We also used eastern grey squirrels (Sciurus carolinensis) as an example invading species and compared diffusion models and two contrasting cost surface dispersal scenarios. The diffusion model underestimated spread because squirrels did not move randomly through the landscape. Despite contrasting assumptions regarding dispersal behaviour, the two cost surfaces were strikingly similar while the least-cost paths differed. Furthermore, while the cost surfaces were insensitive to changes in friction values for linear features, they were sensitive to assumptions made for patch features. Our results suggest that movement in fragmented landscapes may be more sensitive to assumptions regarding friction values than contiguous landscapes. Thus, the reliability of CSA may depend not only on the range of friction values used for patches but also the degree of contiguity in the landscape.     

PA序列下非参数回归函数估计的相合性

设{εi,i≥1}为PA序列，Eεi=0，supE（ε^2j）〈∞，对某个r〉2及占〉δ，supE｜εj｜^r＋δ〈∞，u（n）=O（n^（r-2）（r＋δ）/2δ，在PA序列误差下，讨论了非参数回归函数加权核估计的相合性．     

Markov积分半群的逼近

通过证明Markov积分Q半群的逼近证明了l∞上一个Markov积分半群G(t)可以由一族一致收敛的积分半群逼近.