Estimating net radiation at surface using artificial neural networks: a new approach

This study describes the results of artificial neural network (ANN) models to estimate net radiation (R _n), at surface. Three ANN models were developed based on meteorological data such as wind velocity and direction, surface and air temperature, relative humidity, and soil moisture and temperature. A comparison has been made between the R _n estimates provided by the neural models and two linear models (LM) that need solar incoming shortwave radiation measurements as input parameter. Both ANN and LM results were tested against in situ measured R _n. For the LM ones, the estimations showed a root mean square error (RMSE) between 34.10 and 39.48?W?m^?2 and correlation coefficient (R ²) between 0.96 and 0.97 considering both the developing and the testing phases of calculations. The estimates obtained by the ANN models showed RMSEs between 6.54 and 48.75?W?m^?2 and R ² between 0.92 and 0.98 considering both the training and the testing phases. The ANN estimates are shown to be similar or even better, in some cases, than those given by the LMs. According to the authors?? knowledge, the use of ANNs to estimate R _n has not been discussed earlier, and based on the results obtained, it represents a formidable potential tool for R _n prediction using commonly measured meteorological parameters.     

Precise estimation of hourly global solar radiation for micrometeorological analysis by using data classification and hourly sunshine

We have developed a method for estimating hourly global solar radiation (GSR) from hourly sunshine duration data. This procedure requires only hourly sunshine duration as the input data and utilizes hourly precipitation and daily snow cover as auxiliary data to classify time intervals into six cases according to weather conditions. To obtain hourly GSR using a simple algebraic form, a quadratic function of the solar elevation angle and the sunshine duration ratio is used. Daily GSR is given by a sum of hourly GSRs. We evaluated the performance of the newly developed method using data obtained at 67 meteorological stations and found that the estimated GSR is highly consistent with that observed. Hourly and daily root-mean-square misfits are approximately 0.2 MJ/m²/h (~55 W/m²) and 1.4 to 1.5 MJ/m²/day (~16 to 17 W/m²), respectively. Our classification of weather conditions is effective for reducing estimation errors, especially under cloudy skies. Since the sunshine duration is observed at more meteorological stations than GSR, the proposed new method is a powerful tool for obtaining solar radiation with hourly resolution and a dense geographical distribution. One of the proposed methods, GSRgrn, can be applicable to hourly GSR estimations at different observation sites by setting local parameters (the precipitable water, surface albedo, and atmospheric turbidity) suitable to the sites. The hourly GSR can be applied for various micrometeorological studies, such as the heat budget of crop fields.     

Dispersion Coefficients for Gaussian Puff Models

The Gaussian distribution is a good approximation for transient (instantaneously released) puff concentration distributions within a short period of time after release. Artificial neural network (ANN) models for puff dispersion coefficients were developed, based on observations from field experiments covering a wide range of meteorological conditions (in March, May, August and November). Their average predictions were in very good agreement with measurements, having high correlation coefficients (r > 0.99). A non-linear multi-variable regression model for dispersion coefficients was also developed, under the assumption that puff dispersion coefficients increase with time, and follow power laws. Both ANN-based and multi-regression non-linear models were able to use easily measured atmospheric parameters directly, without the necessity of predefining the Pasquill stability category. Predictions of ANN-based and multi-regression-based Gaussian puff models were compared with those of Gaussian puff models using Slade’s dispersion coefficients and COMBIC, a sophisticated model based on Gaussian distributions. Predictions from our two new models showed better agreement with concentration measurements than the other Gaussian puff models, by having a much higher fraction within a factor of two of measured values, and lower normalized mean square errors.     

The Influence of Selected Meteorological Parameters on the Concentration of Surface Ozone in the Central Region of Poland

Abstract?This paper presents the results of measurements of the concentration of surface ozone and concurrent standard meteorological parameters: total solar radiation, temperature, relative humidity, pressure, wind speed, and vertical and horizontal components of the wind. The data were collected from 2005 to 2010 at stations located in central Poland (Mazowieckie voivodeship): Warszawa (urban), Legionowo (suburban), Granica and Belsk (rural). Furthermore, Granica is situated in the forested area of Kampinoski National Park. Continuously measured surface ozone concentrations demonstrated the well-known diurnal cycle of surface ozone concentration with a maximum in the afternoon and a minimum in the early morning hours. The averaged diurnal variations over six years reveal that the highest concentrations appear at rural stations (Belsk: 55?µg?m^?3 and Granica: 50?µg?m^?3) and the lowest at the urban station (Warszawa: 41?µg?m^?3). The threshold for high levels of surface ozone (120?µg?m^?3 per 8?h) was exceeded most often at Granica and Belsk. The occurrence of the ozone “weekend effect,” especially at urban stations, has been identified. The difference between weekend and weekday surface ozone concentrations at urban and rural stations was as high as 6.5?µg?m^?3 and approximately 2?µg?m^?3, respectively. Using appropriate statistical tools, it has been shown that meteorological conditions have a significant influence on ozone concentration. High correlation coefficients were found between ozone concentration and solar radiation, temperature, relative humidity, and wind speed. The forward stepwise regression model explains up to 75% of the variations in daily surface ozone concentration in terms of meteorological variability in summer and up to 70% in winter. At the same time, a multilayer perceptron neural network model was used to reconstruct the concentration of surface ozone. High correlation coefficients (up to 0.89) indicate that, on the basis of standard meteorological parameters and NO₂ concentration, we can determine ozone concentration with high accuracy.     

Comparative analysis of regression and artificial neural network models for wind speed prediction

In this study, wind speed was modeled by linear regression (LR), nonlinear regression (NLR) and artificial neural network (ANN) methods. A three-layer feedforward artificial neural network structure was constructed and a backpropagation algorithm was used for the training of ANNs. To get a successful simulation, firstly, the correlation coefficients between all of the meteorological variables (wind speed, ambient temperature, atmospheric pressure, relative humidity and rainfall) were calculated taking two variables in turn for each calculation. All independent variables were added to the simple regression model. Then, the method of stepwise multiple regression was applied for the selection of the “best” regression equation (model). Thus, the best independent variables were selected for the LR and NLR models and also used in the input layer of the ANN. The results obtained by all methods were compared to each other. Finally, the ANN method was found to provide better performance than the LR and NLR methods.     

Prediction of soil temperature using regression and artificial neural network models

In this study, monthly soil temperature was modeled by linear regression (LR), nonlinear regression (NLR) and artificial neural network (ANN) methods. The soil temperature and other meteorological parameters, which have been taken from Adana meteorological station, were observed between the years of 2000 and 2007 by the Turkish State Meteorological Service (TSMS). The soil temperatures were measured at depths of 5, 10, 20, 50 and 100 cm below the ground level. A three-layer feed-forward ANN structure was constructed and a back-propagation algorithm was used for the training of ANNs. In order to get a successful simulation, the correlation coefficients between all of the meteorological variables (soil temperature, atmospheric temperature, atmospheric pressure, relative humidity, wind speed, rainfall, global solar radiation and sunshine duration) were calculated taking them two by two. First, all independent variables were split into two time periods such as cold and warm seasons. They were added to the enter regression model. Then, the method of stepwise multiple regression was applied for the selection of the “best” regression equation (model). Thus, the best independent variables were selected for the LR and NLR models and they were also used in the input layer of the ANN method. Results of these methods were compared to each other. Finally, the ANN method was found to provide better performance than the LR and NLR methods.     

Retrieving daily global solar radiation from routine climate variables

Solar radiation is an important variable for studies related to solar energy applications, meteorology, climatology, hydrology, and agricultural meteorology. However, solar radiation is not routinely measured at meteorological stations; therefore, it is often required to estimate it using other techniques such as retrieving from satellite data or estimating using other geophysical variables. Over the years, many models have been developed to estimate solar radiation from other geophysical variables such as temperature, rainfall, and sunshine duration. The aim of this study was to evaluate six of these models using data measured at four independent worldwide networks. The dataset included 13 stations from Australia, 25 stations from Germany, 12 stations from Saudi Arabia, and 48 stations from the USA. The models require either sunshine duration hours (Ångstrom) or daily range of air temperature (Bristow and Campbell, Donatelli and Bellocchi, Donatelli and Campbell, Hargreaves, and Hargreaves and Samani) as input. According to the statistical parameters, Ångstrom and Bristow and Campbell indicated a better performance than the other models. The bias and root mean square error for the Ångstrom model were less than 0.25 MJ m² day^?1 and 2.25 MJ m² day^?1, respectively, and the correlation coefficient was always greater than 95 %. Statistical analysis using Student’s t test indicated that the residuals for Ångstrom, Bristow and Campbell, Hargreaves, and Hargreaves and Samani are not statistically significant at the 5 % level. In other words, the estimated values by these models are statistically consistent with the measured data. Overall, given the simplicity and performance, the Ångstrom model is the best choice for estimating solar radiation when sunshine duration measurements are available; otherwise, Bristow and Campbell can be used to estimate solar radiation using daily range of air temperature.     

Comparison of artificial intelligence techniques for prediction of soil temperatures in Turkey

Soil temperature is a meteorological data directly affecting the formation and development of plants of all kinds. Soil temperatures are usually estimated with various models including the artificial neural networks (ANNs), adaptive neuro-fuzzy inference system (ANFIS), and multiple linear regression (MLR) models. Soil temperatures along with other climate data are recorded by the Turkish State Meteorological Service (MGM) at specific locations all over Turkey. Soil temperatures are commonly measured at 5-, 10-, 20-, 50-, and 100-cm depths below the soil surface. In this study, the soil temperature data in monthly units measured at 261 stations in Turkey having records of at least 20 years were used to develop relevant models. Different input combinations were tested in the ANN and ANFIS models to estimate soil temperatures, and the best combination of significant explanatory variables turns out to be monthly minimum and maximum air temperatures, calendar month number, depth of soil, and monthly precipitation. Next, three standard error terms (mean absolute error (MAE, °C), root mean squared error (RMSE, °C), and determination coefficient (R ²)) were employed to check the reliability of the test data results obtained through the ANN, ANFIS, and MLR models. ANFIS (RMSE 1.99; MAE 1.09; R ² 0.98) is found to outperform both ANN and MLR (RMSE 5.80, 8.89; MAE 1.89, 2.36; R ² 0.93, 0.91) in estimating soil temperature in Turkey.

     

Estimation of daytime downward longwave radiation under clear and cloudy skies conditions over a sub-humid region

Downward longwave radiation (LW _↓ ) is a relevant variable for meteorological and climatic studies. Good estimates of this term are vitally important in correct determining of the net radiation, which, in turn, modulates the magnitude of the terms in the surface energy budget (e.g., evaporation). In remote sensing applications, the determination of daytime LW_↓ is required for estimation of the net radiation using satellite data. LW_↓ is not directly measured in weather stations and then is estimated using models with surface air temperature and humidity as input. In this paper, we identify the best models to estimate daytime downward longwave radiation from meteorological data in the sub-humid Pampean region. Several well-known models to estimate LW_↓ under clear and cloudy skies were tested. We use downward radiation components and meteorological data registered at Tandil (Argentina) from 2006 to 2010 (840 days). In addition, we propose two multiple linear regression models (MLRM-1 and MLRM-2) to estimate LW_↓ at the surface for all sky conditions. The new equations show better performance than the others models tested with root mean square errors between 12 and 16 W m^?2, bias close to zero and best agreements with measured data (r ²?≥?0.85).     

Hourly air temperature driven using multi-layer perceptron and radial basis function networks in arid and semi-arid regions

This study employed two artificial neural network (ANN) models, including multi-layer perceptron (MLP) and radial basis function (RBF), as data-driven methods of hourly air temperature at three meteorological stations in Fars province, Iran. MLP was optimized using the Levenberg–Marquardt (MLP_LM) training algorithm with a tangent sigmoid transfer function. Both time series (TS) and randomized (RZ) data were used for training and testing of ANNs. Daily maximum and minimum air temperatures (MM) and antecedent daily maximum and minimum air temperatures (AMM) constituted the input for ANNs. The ANN models were evaluated using the root mean square error (RMSE), the coefficient of determination (R ²) and the mean absolute error. The use of AMM led to a more accurate estimation of hourly temperature compared with the use of MM. The MLP-ANN seemed to have a higher estimation efficiency than the RBF ANN. Furthermore, the ANN testing using randomized data showed more accurate estimation. The RMSE values for MLP with RZ data using daily maximum and minimum air temperatures for testing phase were equal to 1.2°C, 1.8°C, and 1.7°C, respectively, at Arsanjan, Bajgah, and Kooshkak stations. The results of this study showed that hourly air temperature driven using ANNs (proposed models) had less error than the empirical equation.     

Modeling monthly mean air temperature for Brazil

Air temperature is one of the main weather variables influencing agriculture around the world. Its availability, however, is a concern, mainly in Brazil where the weather stations are more concentrated on the coastal regions of the country. Therefore, the present study had as an objective to develop models for estimating monthly and annual mean air temperature for the Brazilian territory using multiple regression and geographic information system techniques. Temperature data from 2,400 stations distributed across the Brazilian territory were used, 1,800 to develop the equations and 600 for validating them, as well as their geographical coordinates and altitude as independent variables for the models. A total of 39 models were developed, relating the dependent variables maximum, mean, and minimum air temperatures (monthly and annual) to the independent variables latitude, longitude, altitude, and their combinations. All regression models were statistically significant (α?≤?0.01). The monthly and annual temperature models presented determination coefficients between 0.54 and 0.96. We obtained an overall spatial correlation higher than 0.9 between the models proposed and the 16 major models already published for some Brazilian regions, considering a total of 3.67?×?10⁸?pixels evaluated. Our national temperature models are recommended to predict air temperature in all Brazilian territories.     

Estimation of relative humidity based on artificial neural network approach in the Aegean Region of Turkey

The aim of this study is to estimate the monthly mean relative humidity (MRH) values in the Aegean Region of Turkey with the help of the topographical and meteorological parameters based on artificial neural network (ANN) approach. The monthly MRH values were calculated from the measurement in the meteorological observing stations established in Izmir, Mugla, Aydin, Denizli, Usak, Manisa, Kutahya and Afyonkarahisar provinces between 2000 and 2006. Latitude, longitude, altitude, precipitation and months of the year were used in the input layer of the ANN network, while the MRH was used in output layer of the network. The ANN model was developed using MATLAB software, and then actual values were compared with those obtained by ANN and multi-linear regression methods. It seemed that the obtained values were in the acceptable error limits. It is concluded that the determination of relative humidity values is possible at any target point of the region where the measurement cannot be performed.     

华南低能见度天气特征及客观预报研究

利用2000—2016年华南219个县级气象观测站的地面、高空气象观测资料以及对应站点的再分析资料,统计发生低能见度天气的天气形势和特征,归纳低能见度天气的预报指标。将与能见度以及能见度变化相关的气象要素输入神经网络进行训练,利用EC集合预报数据集获得能见度集合预报结果,通过对其离散度的统计分析以及经验公式最终获得具有泛用性、可靠性的神经网络模型的参数集。通过输入EC确定场数据,获得华南219县级站长时效精细化能见度预报结果,2017年上半年的能见度预报试验显示,模型预报结果的误差与TS评分均优于CUACE模式能见度预报。     

江苏省太阳能资源评估

采用1：25万DEM数据和常规气象站观测资料,实现了江苏省100mX100m分辨率太阳总辐射量分布式模拟,并分析了江苏省太阳总辐射量的时空分布规律。结果表明：江苏省气候平均太阳总辐射量为4749MJ／m2,呈现由西南向东北递增的特点,连云港市最高（5063MJ／m2）,无锡市最低（4514MJ／m2）。太阳总辐射量在年内变化特点为,5月最高,12月最低。结合常规气象站日照时数观测资料,从年日照时数、年日照时数i〉6h的天数以及日照时数〉16h的最多天数月份与最少天数月份的天数的比值分析了江苏省太阳能资源的稳定度特征,其总体规律依然是西南至东北走向,即江苏省东北部地区太阳能资源开发利用优势最高。     

随机森林回归法在冬季路面温度预报中的应用

基于宁宿徐高速公路三个交通气象站2015—2018年冬季逐10 min实时观测资料,使用随机森林回归模型预报这三个站的未来1 h冬季路面温度,分析了该模型在冬季路面温度预报中的可行性和适用性。研究结果表明:随机森林回归法可以被用来预报高速公路冬季路面温度,不同类型的交通气象站点的特征输入方案和参数调试标准存在差异;与简单特征相比,引入的复合特征能更好地补充解释交通气象站所处的环境和气象要素,且其对普通路面交通气象站和靠近桥梁、水体的交通气象站的区分度更高,故引入复合特征的随机森林回归模型可以被用来预报高速公路冬季路面温度,且其在对普通路面交通气象站和靠近水体、桥梁的交通气象站的预报效果较好,而对服务区交通气象站的预报效果略差;袋外误差率的降低并不代表预报精度的提高;引入复合特征的随机森林回归模型不论在何种天气状况下,均可用于各不同类型交通气象站冬季路面温度的预报,雨雪天时的预报效果最佳,阴天其次,晴天略差。     

Parameterization and mapping of solar radiation in data sparse regions

Knowledge of temporal and spatial variation of solar radiation is essential for many applications. In this work, a simple and feasible procedure is conducted to map the daily solar radiation for Liaoning province, one of the most important agricultural areas in China, but with sparsely measured solar radiation data. The daily sunshine duration are interpolated to the whole area, subsequently, solar radiation are calculated by ?ngstr?m-Prescott model, the generic parameters of which are determined by least square to minimize the overall fitting residual between the ratio of actual to potential sunshine duration and the ratio of actual to extra-terrestrial solar radiation of the sites where solar radiation are available. In other local regions with sparse data, mapping of the solar radiation could be done following the simple procedure. In the present study area, using the interpolated daily sunshine duration data by ANUSPLIN, ?ngstr?m-Prescott model with the generic parameters (a = 0.505, and b = 0.204) returns reasonable results, with the overall RMSE of 2.255 MJ m^?2, and RRMSE of 16.54%. The daily solar radiation varies between 5.26 in December and 22.74 MJ m^?2 in May, and shows an obviously spatial variation which is mainly contributed to the climate and topography. The substitution of solar radiation from nearby station is preferred to estimation by ?ngstr?m-Prescott model if the distance between the stations falls below the threshold of 135 ± 15 km. The RMSE of such substitution increases by approximately 0.157 MJ m^?2 per 10 km.     

Error assessment of climate variables for FAO-56 reference evapotranspiration

Meteorological stations, which measure all the required meteorological parameters to estimate reference evapotranspiration (ET_o) using the Food and Agriculture Organization Penman?CMonteith (FAO56-PM) method, are limited in Korea. In this study, alternative methods were applied to estimate these parameters, and the applicability of these methods for ET_o estimation was evaluated by comparison with a complete meteorological dataset collected in 2008 in Korea. Despite differences between the estimation and observation of radiation and wind speed, the comparison of ET_o showed small differences [i.e., mean bias error (MBE) varying ?0.22 to 0.25?mm?day^?1 and root-mean-square-error (RMSE) varying 0.06?C0.50?mm?day^?1]. The estimated vapor pressure differed considerably from the observed, resulting in a larger discrepancy in ET_o (i.e., MBE of ?0.50?mm?day^?1 and RMSE of 0.60?C0.73?mm?day^?1). Estimated ET_o showed different sensitivity to variations of the meteorological parameters??in order of vapor pressure?>?wind speed?>?radiation. It is clear that the FAO56-PM method is applicable for reasonable ET_o estimation at a daily time scale especially in data-limited regions in Korea.     

电厂空冷系统环境气象场要素分析中的几个关键技术

以山西省左权、王曲电厂等为期一年的铁塔气象观测资料和各邻近地面气象站同期观测资料为例,说明如何选取典型年以及相关性较好的对比气象站,通过对电厂空冷梯度的主要气象要素分析,结合选取的对比气象站长时间序列的逐时气象资料,采用相关统计分析并进行回归检验,重建厂址区域风、温场资料,并针对风资料转换中存在的问题进行了探讨.结果表明:在两地风资料相关较差时.利用条件概率结合线性回归以及风矢量相关等方法补充订正厂址区域风资料效果较好.其结果对风、温场历史资料的重建有一定的指导意义.     

中国区域性大气加权平均温度线性模型精度评估

以欧洲中期天气预报中心的再分析资料ERA5为参考数据,评估由探空数据建立的中国区域88个单站大气加权平均温度（T_m）与地表气温（T_s）线性关系模型的精度.各站T_m-T_s线性模型计算的T_m（计算值）与ERA5 气压层数据积分所得的T_m（参考值）间偏差均方根值（RMSE）为1.8~5.5 K.不同站模型计算值与参考值间存在-1.22~4.54 K 的系统性偏差,且绝大多数测站（82个站）系统性偏差为正值,即模型计算值总体上大于参考值.补偿各站系统性偏差后,模型计算值与参考值间RMSE降为1.5~3.5 K.与使用中国区域统一模型相比,使用单站模型平均能提高0.6 K的T_m计算精度,尤其在中国西部、西北和内蒙区域,精度提高可达1~3.9 K.对所有测站模型计算值和参考值间偏差时序进行分析,发现超过半数测站的偏差存在明显季节性变化.