Estimating UV erythemal irradiance by means of neural networks
Estimating UV erythemal irradiance by means of neural networks
复制标题
DOI：
10.1562/2004-03-12-ra-111.1
复制
发表时间：
2004-09-01
影响因子：
3.3
通讯作者：
Alados-Arboledas, L
中科院分区：
文献类型：
--
作者：
Alados, I;Mellado, JA;Alados-Arboledas, L
近年来，对紫外线辐射(UV)通量进行建模的尝试大幅增加。地表的紫外线辐照度是太阳天顶角、地表高度、云量、气溶胶负荷和光学性质、地表反照率和臭氧垂直廓线综合作用的结果。在这项研究中，我们提出了一个人工神经网络(ANN)模型，该模型可以根据光学空气质量、臭氧层柱含量、纬度、水平能见度数据和云信息(如类型、覆盖范围和高度)来估计太阳紫外线辐射。人工神经网络作为一种技术被广泛接受，它为解决复杂和定义模糊的问题提供了一种替代方法。它们可以从实例中学习，具有容错能力，能够处理噪声和不完整的数据，能够处理非线性问题，并且一旦经过训练，就可以高速进行预测和推广。在这项研究中，使用了一个由输入层、输出层和隐层组成的多层感知器网络。神经网络的训练采用贝叶斯规则反向传播算法。这项研究使用了伊比利亚半岛三个站的数据：2000年至2001年期间的马德里和穆尔西亚以及2001年期间的萨拉戈萨。为了训练和验证MPL神经网络，从每个站点的完整数据库中提取独立的数据子集。结果表明，利用光学空气质量、臭氧层柱含量、纬度和总云量建立的MLP神经网络对马德里、穆尔西亚和萨拉戈萨的平均偏差和均方根偏差分别为-0.1%和18.0%、1.6%和19.6%、0.1%和14.6%。尽管云的辐射效应依赖于云的类型，但使用额外的信息，如云的类型或云的海拔高度并没有改善这些结果。已开发的人工神经网络的性能已经就其估计紫外线指数(UVI)的能力进行了检查；结果表明，在超过95%的情况下，估计值与实测值之间的差异不超过一个单位的UVI。
In recent years, there has been a substantial increase in attempts to model the flux of ultraviolet radiation (UV). UV irradiance at surface level is a result of the combined effects of solar zenith angle, surface elevation, cloud cover, aerosol load and optical properties, surface albedo and the vertical profile of ozone. In this study, we present the development of an artificial neural network (ANN) model that can be used to estimate solar UV irradiance on the basis of optical air mass, ozone columnar content, latitude, horizontal visibility data and cloud information such as type, coverage and height. ANN are widely accepted as a technology offering an alternative way to tackle complex and ill-defined problems. They can learn from examples, are fault tolerant in the sense that they are able to handle noisy and incomplete data, are able to deal with nonlinear problems and, once trained, can perform prediction and generalization at high speed. In this study, a multilayer perceptron network (MLP) consisting of an input layer, an output layer and one hidden layer was used. Training of the neural network was done using the Bayesian regulation back propagation algorithm. The study was developed using data from three stations on the Iberian Peninsula: Madrid and Murcia during the period 2000-2001 and Zaragoza in 2001. To train and validate the MPL neural networks, independent subsets of data were extracted from the complete database at each station. The results suggest that a MLP neural network using optical air mass, ozone columnar content, latitude and total cloud coverage provides the best estimates, with mean bias deviation and root mean square deviation of -0.1% and 18.0%, 1.6% and 19.6%, 0.1% and 14.6% at Madrid, Murcia and Zaragoza, respectively. Despite the dependence of the cloud radiative effect on cloud type, the use of additional information such as cloud type or cloud elevation did not improve these results. The performance of the developed ANN has been checked regarding its ability to estimate the UV index (UVI); results indicate that in more than 95% of the cases, the difference between estimated and measured values does not exceed one unit of UVI.