基于互信息量与BP神经网络的中长期径流预报方法研究

针对中长期径流预报因子的选择问题,采用互信息量方法筛选预报模型输入因子,在BP神经网络模型中,分别用均方误差和互信息量作为目标函数,衡量因子复合相关关系,优化选择最终预报因子并应用于碧流河汛期径流预报中。结果表明,基于互信息量筛选的预报因子与BP神经网络模型相结合,可有效识别多个预报因子与预报量间的复合相关性,对中长期径流预报因子的选择有很好参考价值。     

CORS系统在GPSC级控制网测量中的应用

简要介绍了CORS系统的发展情况,分析了GPS C级控制网测量的精度,就GPS网的布设、基线解算、GPS网平差及检核进行有益的探讨,旨在发挥CORS的优势,解决GPS控制网测量中控制点缺乏的问题,提高CORS的应用水平。     

新疆哈拉奇地区地面高精度磁测异常及解释简

新疆哈拉奇地区位于西南天山与塔里木盆地接触部位,属于航空磁法测量的边缘区,未开展过地面物探工作。此次开展地面高精度磁测工作,划分了6个区域磁异常区,圈定了3处次级高磁异常和1处局部高磁异常。推断区域磁异常由隐伏基性侵入岩、基底引起,次级高磁异常由构造引起,局部异常由矿体引起。磁异常解释为地质研究、寻找磁性矿产提供了参考依据。     

A second derivative method of resistivity sounding

The second vertical derivatives of gravity and magnetic potential are widely used in geophysical prospecting because of their better resolution. On the same basis an attempt has been made to obtain the expressions for the second vertical derivative of the electrical potential and to compute its nature for comparison. Derivative responses over a two-layered earth and also over an anticlinal structure have been computed and it is shown that the second vertical derivative sounding could be employed for greater accuracy in finding out the thickness of such beds or the inclination of the sides of the anticline and its depth when compared with normal resistivity sounding     

改进的灰色预测模型在地面沉降预测中的应用

在地面沉降这一复杂系统中,既含有已知的又古有未知的或非确定的信息,可以作为一个灰色系统来研究.本文针对地面沉降的下沉曲线非线性特征,提出用一种基于残差灰色预测模型对地面沉降量时间序列进行研究.结果表明,通过改进后的灰色预测模型,预测精度得到了提高,在沉降量比较大和水准点比较稀少的地区,利用此模型预测地面沉降可减少地面沉降监测经费,并能实时提供地面沉降预警信息.     

Limitations of real-time models for forecasting river flooding from monsoon rainfall

Very intense rainfall during the southwest and northeast monsoons causes severe river flooding in India. Some traditional techniques used for real-time forecasting of flooding involve the relationship between effective rainfall and direct surface runoff, which simplifies the complex interactions between rainfall and runoff processes. There are, however, serious problems in deducing these variables in real time, so it is highly desirable to have a real-time flood forecasting model that would directly relate the observed discharge hydrograph to the observed rainfall. The storage routing model described by Baba and Hoshi (1997), Tanaka et al. (1997), and Baba et al. (2000), and a simplified version of this model, have been used to compute observed river discharge directly from observed hourly rainfall. This method has been used to study rainfall–runoff data of the Ajay River Basin in eastern India. Five intense rainfall events of this basin were studied. Our results showed that the Nash–Sutcliffe efficiency of discharge prediction for these five events was 98.6%, 94.3%, 86.9%, 85.6%, and 67%. The hindcast for the first two events is regarded as completely satisfactory whereas for the next two events it is deemed reasonable and for the fifth it is unsatisfactory. It seems the models will yield accurate hindcast if the rainfall is uniform over the drainage basin. When the rainfall is not uniform the performance of the model is unsatisfactory. In future this problem can, in principle, be corrected by using a weighted amount if rainfall is based upon multiple rain-gauge observations over the drainage basin. This would provide some measure of the dispersion in the rainfall. The model also seems unable to simulate flooding events with multiple peaks.     

潞西盆地浅层地下水脆弱性预测分析

运用DRASTIC方法，评价潞西盆地浅层地下水脆弱性，结合区内人类工程活动因子。对潞西盆地远期地下水脆弱性进行预测分析，成果可为区内土地规划、备用水源地的选择、垃圾填埋场的选址及实现水资源可持续利用提供合理依据。     

宁镇地区地质灾害预报模型——以镇江润州区为例

宁镇地区是长江中下游地质灾害最严重的地区之一。镇江润州区虽然仅是宁镇地区的一个局部区域,但其气候和地质环境特征具有典型意义,所提出的地质灾害气象预警预报模型同样适用于整个宁镇地区,对长江中下游地区亦有借鉴作用。气候环境、降雨尤其是连续降雨或强降雨是诱发地质灾害的重要因素。润州区地质灾害主要与梅雨期总降雨量有关,其次与台汛期台风带来的降雨量有关,而与台汛期总降雨量无关。地质灾害预测预警方程应针对不同时期采用不同的预警模型：非梅雨期的预警方程采用预报日降雨量结合前5日降水之和的综合模型,梅雨期的预警方程采用梅雨期降雨总量模型。提出地质灾害气象预报预警等级应根据《国家突发公共事件总体应急预案》将等级统一划分为4级。该模型可作为完善我国现有地质灾害气象预警预报系统的参考。     

变权组合预测模型在地下水水位预测中的应用

为探讨变权组合预测模型在地下水水位预测中的应用,以内蒙古通辽市东郊为研究区,利用该区1997-2008年的地下水水位埋深数据,采用基于指数预测法、灰色预测法以及线性回归预测法的变权组合预测方法进行水位埋深的模拟和预测。经计算,以上3种单项预测方法的预测值以及组合预测方法的预测值与实际值的误差平方和分别为5.116 1、5.080 1、4.914 8、4.672 3。这表明,利用变权组合预测方法比单纯运用某一种预测方法预测精度更高。并且,由预测值可以看出,到2015年该研究区水位埋深将达到15.39 m,相比2008年递增幅度将达到20%。     

基于改进型前馈神经网络的流域产流预报模型的研究

在分析流域产流机制、影响因素和现行产流计算方法的基础上,首次取前期影响雨量、主产流历时、全过程面平均雨量和４个代表雨强计７个因子作为神经网络输入,直接以流域产流深作为神经网络输出,并针对传统ＢＰ算法的固有缺陷,采用混合ＧＮ－ＢＦＧＳ算法训练网络。实例验证了所建模型及算法的有效性和可行性。还对神经网络隐层单位数等进行了初步研究。