基于深度学习和雷达观测的华南短临预报精度评估

利用最新的深度学习算法,即卷积长短期记忆(Convolution Long-Short Term Memory)神经网络,构建基于深度学习的人工智能短临预报系统,以广州地区2019年3-5月雷达观测的数据为输入进行训练,然后进行短期1h内的降水预报。利用常用的统计评分指标(探测率POD、误报率FAR、临界成功指数CSI,相关系数CC)检验模型。结果表明,预报结果与实际观测的相关系数在1h内预报均保持在0.6以上,在1h内预报探测率均保持在80%以上,临界成功指数在降水强度为10mm·h^-1时,基本保持在60%,误报率均小于40%。     

A mirage at Regina

The Dirt Hills, located roughly forty miles to the southwest of the Regina Weather Office, at Regina, Saskatchewan, were clearly visible on the horizon in the early morning of the 29th day of August 1972 from the Regina Weather Office. This mirage was documented and a picture of it was taken.     

基于MonoRTM模型的微波辐射计反演方法研究

基于辐射传输模型MonoRTM计算天空亮温度,使用多元线性统计回归方法和BP神经网络方法分别对大气温度和水汽密度廓线进行了反演,检验并分析了两种方法的反演精度.结果表明,多元线性回归方法反演的温度偏差总体不大于6K,反演的水汽密度偏差小于4 g/m3;神经网络方法反演的温度偏差小于2K,反演的水汽密度误差总体不大于2 g/m3.与探空数据的对比表明,对于大气温度和水汽密度反演,BP神经网络方法的反演结果都要比多元线性回归方法的反演结果更接近探空资料值.     

基于中尺度模式与神经网络的风电功率预测

将中尺度数值天气预报模式与BP神经网络模型相结合用于风电功率预测,以WRF模式回算了2008年6月至2009年6月试验风电场的气象要素,精度检验结果显示风速预报值与对应实测值之间的相关系数达到0.72,风向、气温、湿度、气压的预报也比较准确,满足建立BP神经网络预报模型的需要.逐一建立试验风电场40台风电机组输出功率的BP神经网络预报模型,分析了数据标准化方法、隐含层神经元数对预报精度的影响.进行了26天实效为24 h的逐10 min预报试验,并以独立样本进行预报精度检验,结果显示单台风电机组输出功率相对均方根误差在24.8％～32.6％之间,预报值与实测值之间的相关系数现在0.45～0.68之间;风电场整体相对均方根误差为19.5％,预报值与实测值之间的相关系数为0.74.研究结果表明该方法可以用于实际的风电功率预测.     

粒子群神经网络及其在非线性系统辨识中的应用

介绍了一种新的神经网络权值优化算法——粒子群优（Particle Swarm Optimization，PSO）算法，提出了用粒子群神经网络对非线性系统进行系统辨识的构思。仿真实验结果表明，粒子群算法具有比BP算法更强的非线性系统辨识能力和更好的泛化能力。     

Towards automation of palynology 2: the use of texture measures and neural network analysis for automated identification of optical images of pollen grains

The automation of palynology (the identification and counting of pollen grains and spores) will be a small step for image recognition, but a giant stride for palynology. Here we show the first successful automated identification, with 100% accuracy, of a realistic number of taxa. The technique used involves a neural network classifier applied to surface texture data from light microscope images. A further significance of the technique is that it could be adapted for the identification of a wide range of biological objects, both microscopic and macroscopic. Copyright © 2004 John Wiley & Sons, Ltd.     

Evaluation and forecasting of daily groundwater outflow in a small chalky watershed

The purpose of this research is to include expert knowledge as one part of the modelling system and therefore offer the chance to create a productive interaction system between expert, mathematical model (MMO8) and artificial neural networks (ANNs). In the present project, the first objective is to determine some parameters by the MMO8 model, introduced as ANN input parameters to forecast spring outflow. The second objective is first to investigate the effect of temporal information by taking current and past data sets and then to forecast spring outflow. The good results obtained reveal the merit of the ANNs–MMO8 combination, and specifically multilayer perceptron (MLP) models. This methodology, for a network with lower, lag and number hidden layer, consistently produced better performance. Copyright © 2003 John Wiley & Sons, Ltd.     

四川旱涝震灾害的人工神经网络外推预测

选取旱、涝、震灾害之间的一步转移概率作为Ｂ－Ｐ人工神经网络训练样本的输入信息，建立了四川旱、涝、震的人工神经网络灾型预测模型。Ｂ－Ｐ网络模型应用于实例预测结果与用主分量分析法的趋势预测结果精度接近。     

TREE INVASION IN MEADOWS OF LASSEN VOLCANIC NATIONAL PARK,CALIFORNIA*

The causes of tree invasion into meadows in Lassen Volcanic National Park were investigated by dating invasive trees and associating age structure patterns with regional variations in fire history, grazing, and climate change. Massive invasion was associated primarily with cessation of livestock grazing and burning between 1905 and 1933, but climatic change may also have contributed. Maintenance of meadows may require continued use of unnatural processes due to uncertainties about pre-European fire regimes and other resource considerations.     

Neural network and wavelet conjunction model for modelling monthly level fluctuations in Turkey

This paper proposes the application of a conjunction model (neuro‐wavelet) for forecasting monthly lake levels. The neuro‐wavelet (NW) conjunction model is improved combining two methods, discrete wavelet transform and artificial neural networks. The application of the methodology is presented for the Lake Van, which is the biggest lake in Turkey, and Lake Egirdir. The accuracy of the NW model is investigated for 1‐ and 6‐month‐ahead lake level forecasting. The root mean square errors, mean absolute relative errors and determination coefficient statistics are used for evaluating the accuracy of NW models. The results of the proposed models are compared with those of the neural networks. In the 1‐month‐ahead lake level forecasting, the NW conjunction model reduced the root mean square errors and mean absolute relative errors by 87–34% and 86–31% for the Van and Egirdir lakes, respectively. In the 6‐month‐ahead lake level forecasting, the NW conjunction model reduced the root mean square errors and mean absolute relative errors by 34–48% and 30‐46% for the Van and Egirdir lakes, respectively. The comparison results indicate that the suggested model could significantly increase the short‐ and long‐term forecast accuracy. Copyright © 2009 John Wiley & Sons, Ltd.