基于遗传神经网络的采暖季PM10浓度预测

用GA算法和BP算法相结合的算法预测西安市PM10污染浓度,首先采用GA算法优化BP神经网络模型的初始权重,再用BP算法进行精确训练,在此基础上进行浓度预报。实例表明GA-BP神经网络解决了BP神经网络收敛速度慢和易于陷入极小值的问题,提高了预测精度。     

基于支持向量机的农村用电量需求预测

提出了一种基于支持向量机的农村用电量需求预测方按。该方法充分发挥支持向量机在解决非线性和有限样本问题方面体现出的优势。保证了预测模型的泛化能力,具有较高的预测精度。实例验证了该方法的有效性。     

吉安市冻雨灾害动态预测模型

用灰色GM(1,1)模型对吉安市7个历史时期数据进行建模,预测下一次该地区冻雨灾害可能的发生时期。     

基于神经网络的林区道路林产品货运量预测方法分析

根据近年来的交通现状来分析和阐释林区道路货运量预测方法的优缺点,运用神经网络的相关理论对林区道路林产品货运量进行预测分析,运用MATLAB进行训练,预测出林区道路林产品货运量,并与实际数据相比较,对此种方法检验准确度。结果表明,基于神经网络的林区道路货运量的预测简单有效,准确并且容易掌握。     

Phenology and pest status of Agrotis segetum in a changing climate

Insects are strongly dependent on climatic factors, especially temperature. For this reason, changes in insect phenology are predicted to be one of the many effects of climate change. We analysed the effect of climate warming on Agrotis segetum (Denis & Schiffermüller) in Denmark. Agrotis segetum is an agricultural pest that causes root crop damage. Extensive registers of root crop damages by A. segetum in Denmark were kept between 1905 and 1980. These records show a correlation between A. segetum outbreaks and hot, dry summers. The last outbreak in Denmark, in 1976, prompted the implementation of pheromone trap monitoring. Data from these pheromone traps, for the periods between 1981–1991 and 1997–2009, show a decline in the number of A. segetum individuals during 2000–2009. We used this pheromone trapping data to test the hypothesis that temporal changes in phenology and abundance are correlated with temperature and precipitation patterns. As phenological fixed points, we use the time of capture (first, 10%, 50%, 90% and last) of 1st generation A. segetum together with the 10%–90% capture period and the total capture per location for a statistical comparison of the period 1981–1990 versus 2000–2009. We also analysed the effect of temperature on the timing of fixed points and the effect of precipitation on the mean total capture from the summer and the preceding winter. Our analyses showed an earlier occurrence of A. segetum during 2000–2009 than during 1981–1990: all phenological fixed points advanced by 6–9 days. Trap captures of the 1st generation roughly halved from the 1981–1990 decade to the 2000–2009 decade, and a much larger 2nd generation of moths developed in the latest decade. When a second generation occurs, populations in following years decrease because the larval offspring of the 2nd generation cannot successfully overwinter. Our analyses further indicated that increased precipitation during the summers and mild winters contributed to an overall decline in the populations of this pest. The presence of A. segetum remains sufficient to be a potential threat to yields. Forecasting an economically important attack of A. segetum is challenging because of the early emergence of A. segetum and more variable local weather conditions than 20 years ago; forecasting relies on the timely and correct delivery of monitoring results. Farmers are a vital source of information for monitoring, for estimating control requests and for timing the emergence of A. segtum. However, the growing frequency of no-risk forecasts has reduced the interest and lowered the motivation of farmers to participate in monitoring programs.     

榆绥高速公路水土流失的动态防治

通过对榆绥高速公路建设过程中水土流失的实时监测,预测了因工程建设给环境可能带来的损害,依据预测结果提出了需要治理的重点区域,进而在治理实践中把水土流失防治和改善周边生态环境完美地结合起来.研究结果为生态脆弱区实现生产建设与生态环境保护的双赢提供了参考.     

冬小麦种植条件下土壤水盐运移特征的数值模拟与预报

种植作物条件下水盐在土壤中的转化和运移构成了一个非常复杂的物理-化学-生物系统。研究种植作物条件下土壤水盐运移的动态规律和运行特征,建立水盐运移的数学模型对于指导盐渍土的灌溉管理及劣质水利用、土壤盐渍化与持续农业和生态环境之间的相互作用及土壤盐渍化预测等方面具有重要的意义。本研究首先提出了自主开发的土壤水盐运移的数学模型SWSTM(Soil Water and Salt Transport Mod-el),然后对冬小麦种植条件下土壤水盐的运动规律和特征进行了数值模拟,最后对不同地下水位和不同气象条件下的土壤水盐运移规律进行了数值预测,以期从土壤水盐运动的规律出发,提出一种应用数值模拟方法来预报土壤水盐动态的途径,同时为种植作物条件下田间大面积土壤水盐动态预测预报提供参考。     

犍为县城镇化发展趋势及其用地需求分析

分析了四川省犍为县的社会经济环境和小城镇发展现状,采用综合增长率预测法、时间序列预测法、灰色数列预测法对2010和2020年犍为县总人口和县城聚集人口进行预测,2010年和2020年犍为县总人口和积聚人口分别为559773、550773和123009、136356人;并以人口数据为基础,利用定性分析法、人口预测法、联合国法等方法预测了2010和2020年犍为县的城镇化水平,2010和2020年犍为县城镇化水平分别为37.74%、48.82%;最后推算出2010和2020年城镇建设所需用地分别为1636.8448hm2和2490.474hm2.以此为区域规划者和领导决策层提供依据.     

辽宁省玉米热量指数预测模型研究

通过分析辽宁省玉米热量指数分布情况,为辽宁省玉米延迟型低温冷害的预测提供基础方法。利用1961—2010 年的逐旬气温资料、作物发育期数据和大气环流资料,建立了分区域玉米热量指数的滚动预测模型。结果表明：辽东、辽南、辽西、辽北及中部地区2000—2010 年的预测数据与实况数据相关系数分别为97.3%、95.4%、98.2%、97.5%、98.5%。模型的准确率较高,且稳定性较好,能较为准确地预测辽宁省玉米生长发育期间的热量状况。     

辽西地区松毛虫发生期气象预报方法

为了研究松毛虫的生活习性与环境的关系,更有效地对松毛虫灾害进行预测和防治,利用气象因子与松毛虫发生期的关系,采用期距法、多元回归分析法对松毛虫各发生始见期进行预报。结果表明：利用期距法研究单个气象要素与松毛虫发生期的关系,幼虫上树到羽化始见期温度是影响松毛虫发生的主要因素,而降水是造成松毛虫下树的主要原因,利用期距法预报各虫态发生始见期偏差6天以内的比例为72.5%;利用多元回归分析法研究多个气象要素与松毛虫发生期的关系,幼虫上树到羽化始见期与日照时数、日较差关系更为密切,下树始见期的出现与温度和降水有关,利用多元回归分析法预报各虫态始见期偏差6天以内的比例为67.5%。