论文部分内容阅读
土壤湿度的实测数据量因地面观测网的稀疏不均和观测时间的不连续而十分有限,而其作为气象、农业、水文、环境等学科领域的重要研究内容,数据量的匮乏直接影响了研究工作的顺利进行。地面常规气象数据的观测频率较高(逐日/时),提供了丰富的大气及土壤状态信息,从地气交互作用的普遍性出发,对气象要素与土壤湿度之间的作用关系进行研究,拟借助人工神经网络良好的函数模拟能力,建立以多气象要素为网络输入、以土壤湿度为网络输出的BP神经网络。通过主成分分析筛选特征要素、选择训练函数、确定合理的隐层神经元个数等来精细化网络。以甘肃省2008年8、9月份的AB报(土壤湿度数据)和A报(气象观测数据)资料进行了实验,建立BP神经网络,最终获得了较好的土壤湿度预测结果。
The measured data of soil moisture is very limited due to the sparse unevenness of the observational network and the discontinuity of observation time. As an important research subject in the field of meteorology, agriculture, hydrology and environment, the lack of data has a direct impact on the research Work smoothly. Conventional meteorological data on the ground have a high frequency of observation (daily / hour) and provide abundant information on the state of the atmosphere and soil. Based on the universality of the interaction between air and ground, the relationship between meteorological elements and soil moisture is studied. With the good function of artificial neural network, a BP neural network with multi-meteorological elements as network input and soil moisture as network output is established. Through the principal component analysis screening feature elements, select the training function to determine the number of hidden neurons and other reasonable to refine the network. Experiments were carried out in August and September 2008 in Gansu Province by AB (soil moisture data) and A (meteorological observation data) data to establish a BP neural network and finally to obtain better soil moisture prediction results.