支持向量机方法在南京太阳总辐射推算中的应用

利用1981、1996和2001年逐日南京站太阳总辐射和日照时数观测资料,建立了基于支持向量机(support vector machine,SVM)方法的太阳总辐射推算模型,预测了1982、1997和2002年的太阳总辐射,并把推算结果和采用线性的气候学方法所得到的推算结果分别与实测值进行对比。采用线性方法得到的1982、1997和2002年的太阳总辐射预测值与实测值间基于1:1线的决定系数(R~2)分别为0.800、0.859和0.838,均方根误差(RMSE)分别为3.250、2.649和2.925 MJ·m~(-2)·d~(-1)。采用SVM方法得到的1982、1997和2002年的R~2分别为0.894、0.938和0.936,RMSE分别为2.353、1.726和1.804 MJ·m~(-2)·d~(-1)。SVM方法得到的太阳总辐射预测值与实测值之间的误差较小,预测精度高于线性方法,更适用于实际太阳总辐射的计算。     

Study on ann-based multi-step prediction model of short-term climatic variation

In the context of 1905-1995 series from Nanjing and Hangzhou, study is undertaken of establishing a predictive model of annual mean temperature in 1996-2005 to come over the Changjiang (Yangtze River) delta region through mean generating function and artificial neural network in combination. Results show that the established model yields mean error of 0.45℃ for their absolute values of annual mean temperature from 10 yearly independent samples (1986-1995) and the difference between the mean predictions and related measurements is 0.156℃. The developed model is found superior to a mean generating function regression model both in historical data fitting and independent sample prediction.     

Long Range Forecasting Indian Summer Monsoon Rainfallby a Hybrid Principal Component Neural Network Model

Summary  The existing methods based on statistical techniques for long range forecasts of Indian monsoon rainfall have shown reasonably accurate performance, for last 11 years. Because of the limitation of such statistical techniques, new techniques may have to be tried to obtain better results. In this paper, we discuss the results of an artificial neural network model by combining two different neural networks, one explaining assumed deterministic dynamics within the time series of Indian monsoon rainfall (Model I) and other using eight regional and global predictors (Model II). The model I has been developed by using the data of past 50 years (1901–50) and the data for recent period (1951–97) has been used for verification. The model II has been developed by using the 30 year (1958–87) data and the verification of this model has been carried out using the independent data of 10 year period (1988–97). In model II, instead of using eight parameters directly as inputs, we have carried out Principal Component Analysis (PCA) of the eight parameters with 30 years of data, 1958–87, and the first five principal components are included as input parameters. By combining model I and model II, a hybrid principal component neural network model (Model III) has been developed by using 30 year (1958–87) data as training period and recent 10 year period (1988–97) as verification period. Performance of the hybrid model (Model III) has been found the best among all three models developed. Rootmean square error (RMSE) of this hybrid model during the independent period (1988–97) is 4.93% as against 6.83%of the operational forecasts of the India Meteorological Department (IMD) using the 16 parameter Power Regression model. As this hybrid model is showing good results, it is now used by the IMD for experimental long-range forecasts of summer monsoon rainfall over India as a whole. Received August 20, 1998/Revised April 20, 1999     

A monitoring network for erythemally-effective solar ultraviolet radiation in Austria: determination of the measuring sites and visualisation of the spatial distribution

Summary In 1997 the Austrian ultraviolet radiation monitoring network was to built up to measure continuously erythemally-effective ultraviolet radiation. For this network the necessary measuring sites had to be selected by an objective method. Therefore a method was used, based on the de-correlation distances, calculated by the daily maximum of the global solar irradiance across the entire spectrum. For a correlation coefficient of 0.90, de-correlation distances were found to be in the order of 170 km (East-West) and 40 km (North-South). From this, 7 measuring sites wereselected to derive a total coverage of Austria. For the visualisation of the measured erythemally-effective solar radiation an optimum interpolation method was used to calculate the spatial distribution. The topography of Austria was used for height correction. The global solar radiation of 39 stations is also used to include the spatial distribution of cloud cover and the resulting attenuation of the erythemally-effective radiation. The combination of these two data sets provides an accuracy estimation of people’s exposure to erythemally-effective UV radiation within the mountainous country of Austria. Received September 12, 1999/Revised April 6, 2001     

Comparison of model predicted cloud parameters and surface radiative fluxes with observations on the 100 km scale

Summary Cloud parameters and surface radiative fluxes predicted by regional atmospheric models are directly compared with observations for a 10-day period in late summer 1995 characterized by predominantly large-scale synoptic conditions. Observations of total cloud cover and vertical cloud structure are inferred from measurements with a ground-based network of Lidar ceilometers and IR-radiometers and from satellite observations on a 100 kilometer scale. Ground-based observations show that at altitudes below 3 km, implying liquid water clouds, there is a considerable portion of optically non-opaque clouds. Vertical distributions of cloud temperatures simultaneously inferred from the ground-based infrared radiometer network and from satellite can only be reconciled if the occurrence of optically thin cloud structures at mid- and high tropospheric levels is assumed to be frequent. Results of three regional atmospheric models, i.e. the GKSS-REMO, SMHI-HIRLAM, and KNMI-RACMO, are quantitatively compared with the observations. The main finding is that all models predict too much cloud amount at low altitude below 900 hPa, which is then compensated by an underestimation of cloud amount around 800 hPa. This is likely to be related with the finding that all models tend to underestimate the planetary boundary layer height. All models overpredict the high-level cloud amount albeit it is difficult to quantify to what extent due to the frequent presence of optically thin clouds. Whereas reasonably alike in cloud parameters, the models differ considerably in radiative fluxes. One model links a well matching incoming solar radiation to a radiatively transparent atmosphere over a too cool surface, another model underpredicts incoming solar radiation at the surface due to a too strong cloud feedback to radiation, the last model represents all surface radiative fluxes quite well on average, but underestimates the sensitivity of atmospheric transmissivity to cloud amount. Received August 31, 2000 Revised March 15, 2001     

基于改进EMD算法和BP神经网络的SST预测研究

海洋表面温度(Sea Surface Temperature,SST)具有非平稳、非线性的特征,直接将处理平稳数据序列的方法应用到非平稳非线性特征明显的序列上显然是不合适的,预测的误差将会很大。为了提高预测精度,更好地解决非平稳非线性序列预测的问题,本文以东北部太平洋(40°N~50°N、150°W~135°W)区域的月平均海洋表面距平温度为例,首先分别应用集合经验模态分解(EEMD)和互补集合经验模态分解(CEEMD)方法将SST分解为不同尺度的一系列模态分量(IMF),再运用BP(Back Propagation)神经网络模型对每一个模态分量进行分析预测,最后将各IMF预测结果进行重构得到SST的预测值。数值试验的结果表明,CEEMD分解精度比EEMD分解精度高,CEEMD提高了基于BP神经网络的预测精度。系列试验统计分析说明应用这种方法对SST的1年预测是有效的。     

Modelling daily global radiation at the forest floor

A simple model is developed which estimates daily global radiation at the floor of a non-homogeneous Eucalyptus forest. Model input parameters are easily derived from field measurements and consist of individual tree location, tree height, maximum canopy width and its corresponding height, height of the lowest branch and trunk thickness. In addition, the model requires values for global and diffuse irradiance in the open. The tree canopy is represented as a series of spheres containing leaves which are homogeneously spaced but are oriented in the vertical plane. This configuration closely approaches that of actual eucalyptus trees. A Monte-Carlo approach is used to estimate the albedo of the unit sphere as a function of solar zenith angle. At a given combination of solar zenith and azimuth angle, the model estimates the solar irradiance at a specific forest floor location.The model, when tested against pyranometer measurements, predicted daily solar irradiance with a correlation of 0.98 and a standard error of 0.98 MJ m^-2 day^-1. This good performance is attributed to the spatial averaging of the radiation fluxes over the entire day, and the relatively low sensitivity of the calculated solar irradiance to sphere albedo.     

On the use of a cloud modification factor for solar UV (290–385 nm) spectral range

Summary  Knowledge of ultraviolet radiation is necessary in different applications, in the absence of measurements, this radiometric flux must be estimated from available parameters. To compute this flux under all sky conditions one must consider the influence of clouds. Clouds are the largest modulators of the solar radiative flux reaching the Earth’s surface. The amount and type of cloud cover prevailing at a given time and location largely determines the amount and type of solar radiation received at the Earth’s surface. This cloud radiative effect is different for the different solar spectral bands. In this work, we analyse the cloud radiative effect over ultraviolet radiation (290–385 nm). This could be done by defining a cloud modification Factor. We have developed such cloud modification Factor considering two different types of clouds. The efficiency of the cloud radiative effect scheme has been tested in combination with a cloudless sky empirical model using independent data sets. The performance of the model has been tested in relation to its predictive capability of global ultraviolet radiation. For this purpose, data recorded at two radiometric stations are used. The first one is located at the University of Almería, a seashore location (36.83° N, 2.41° W, 20 m a.m.s.l.), while the second one is located at Granada (37.18° N, 3.58° W, 660 m a.m.s.l.), an inland location. The database includes hourly values of the relevant variables that cover the years 1993–94 in Almería and 1994–95 in Granada. Cloud cover information provided by the Spanish Meteorological Service has been include to compute the clouds radiative effect. After our study, it appears that the combination of an appropriate cloudless sky model with the cloud modification Factor scheme provides estimates of ultraviolet radiation with mean bias deviation of about 5% that is close to experimental errors. Comparisons with similar formulations of the cloud radiative effect over the whole solar spectrum provides evidence for the spectral dependency of the cloud radiative effect. Received November 15, 1999 Revised September 11, 2000     

Evaluating regional climate model estimates against site-specific observed data in the UK

This paper compares precipitation, maximum and minimum air temperature and solar radiation estimates from the Hadley Centre’s HadRM3 regional climate model (RCM), (50 × 50 km grid cells), with observed data from 15 meteorological station in the UK, for the period 1960–90. The aim was to investigate how well the HadRM3 is able to represent weather characteristics for a historical period (hindcast) for which validation data exist. The rationale was to determine if the HadRM3 data contain systematic errors and to investigate how suitable the data are for use in climate change impact studies at particular locations. Comparing modelled and observed data helps assess and quantify the uncertainty introduced to climate impact studies. The results show that the model performs very well for some locations and weather variable combinations, but poorly for others. Maximum temperature estimations are generally good, but minimum temperature is overestimated and extreme cold events are not represented well. For precipitation, the model produces too many small events leading to a serious under estimation of the number of dry days (zero precipitation), whilst also over- or underestimating the mean annual total. Estimates represent well the temporal distribution of precipitation events. The model systematically over-estimates solar radiation, but does produce good quality estimates at some locations. It is concluded that the HadRM3 data are unsuitable for detailed (i.e. daily time step simulation model based) site-specific impacts studies in their current form. However, the close similarity between modelled and observed data for the historical case raises the potential for using simple adjustment methods and applying these to future projection data.     

Dependence of global radiation on cloudiness and surface albedo in Tartu, Estonia

Summary ?The dependence of global and diffuse radiation on surface albedo due to multiple reflection of radiation between the surface and the atmosphere (base of clouds) is found on the basis of data obtained at the Tartu–T?ravere Actinometric Station over the period 1955–2000. It is found that the monthly totals of global radiation increase by up to 1.38–1.88 times, particularly in the winter half-year between November and March, when snow cover albedo may be high. A semi-empirical formula is derived for calculating with sufficient accuracy the monthly totals of global radiation, considering the amount of cloudiness and the surface albedo. In the time series of the monthly total by global radiation a downward trend occurs in winter months. A decrease in global radiation by up to 20% in the past 46 years can be explained primarily by a relatively high negative trend in the snow cover duration and surface albedo (up to − 0.24). As a result, days are growing darker, a new phenomenon associated with climate change, which undoubtedly affects human mood to some extent. Received November 8, 2001; revised January 24, 2002; accepted February 2, 2002     

Predicting Indian monsoon rainfall: a neural network approach

The summer monsoon rainfall over India is predicted by using neural networks. These computational structures are used as a nonlinear method to correlate preseason predictors to rainfall data, and as an algorithm for reconstruction of the rainfall time-series intrinsic dynamics. A combined approach is developed which captures the information built into both the stochastic approach based on suitable predictors and the deterministic dynamical model of the time series. The hierarchical network so obtained has forecasting capabilities remarkably improved with respect to conventional methods.     

All India summer monsoon rainfall prediction using an artificial neural network

The prediction of Indian summer monsoon rainfall (ISMR) on a seasonal time scales has been attempted by various research groups using different techniques including artificial neural networks. The prediction of ISMR on monthly and seasonal time scales is not only scientifically challenging but is also important for planning and devising agricultural strategies. This article describes the artificial neural network (ANN) technique with error- back-propagation algorithm to provide prediction (hindcast) of ISMR on monthly and seasonal time scales. The ANN technique is applied to the five time series of June, July, August, September monthly means and seasonal mean (June + July + August + September) rainfall from 1871 to 1994 based on Parthasarathy data set. The previous five years values from all the five time-series were used to train the ANN to predict for the next year. The details of the models used are discussed. Various statistics are calculated to examine the performance of the models and it is found that the models could be used as a forecasting tool on seasonal and monthly time scales. It is observed by various researchers that with the passage of time the relationships between various predictors and Indian monsoon are changing, leading to changes in monsoon predictability. This issue is discussed and it is found that the monsoon system inherently has a decadal scale variation in predictability. Received: 13 March 1999 / Accepted: 31 August 1999     

Comparison of the performance of net radiation calculation models

Daily values of net radiation are used in many applications of crop-growth modeling and agricultural water management. Measurements of net radiation are not part of the routine measurement program at many weather stations and are commonly estimated based on other meteorological parameters. Daily values of net radiation were calculated using three net outgoing long-wave radiation models and compared to measured values. Four meteorological datasets representing two climate regimes, a sub-humid, high-latitude environment and a semi-arid mid-latitude environment, were used to test the models. The long-wave radiation models included a physically based model, an empirical model from the literature, and a new empirical model. Both empirical models used only solar radiation as required for meteorological input. The long-wave radiation models were used with model calibration coefficients from the literature and with locally calibrated ones. A measured, average albedo value of 0.25 was used at the high-latitude sites. A fixed albedo value of 0.25 resulted in less bias and scatter at the mid-latitude sites compared to other albedo values. When used with model coefficients calibrated locally or developed for specific climate regimes, the predictions of the physically based model had slightly lower bias and scatter than the empirical models. When used with their original model coefficients, the physically based model had a higher bias than the measurement error of the net radiation instruments used. The performance of the empirical models was nearly identical at all sites. Since the empirical models were easier to use and simpler to calibrate than the physically based models, the results indicate that the empirical models can be used as a good substitute for the physically based ones when available meteorological input data is limited. Model predictions were found to have a higher bias and scatter when using summed calculated hourly time steps compared to using daily input data.     

Variations in Vegetation Net Primary Production in the Qinghai-Xizang Plateau, China, from 1982 to 1999

Vegetation net primary production (NPP) derived from a carbon model (Carnegie–Ames–Stanford Approach, CASA) and its interannual change in the Qinghai-Xizang (Tibetan) Plateau were investigated in this study using 1982–1999 time series data sets of normalized difference vegetation index (NDVI) and paired ground-based information on vegetation, climate, soil, and solar radiation. The 18-year averaged annual NPP over the plateau was 125 g C m⁻² yr⁻¹, decreasing from the southeast to the northwest, consistent with precipitation and temperature patterns. Total annual NPP was estimated between 0.183 and 0.244 Pg C over the 18 years, with an average of 0.212 Pg C (1 Pg = 10¹⁵ g). Two distinct periods (1982–1990 and 1991–1999) of NPP variation were observed, separated by a sharp reduction during 1990–1991. From 1982 to 1990, annual NPP did not show a significant trend, while from 1991 to 1999 a marked increase of 0.007 Pg C yr⁻² was observed. NPP trends for most vegetation types resembled that of the whole plateau. The largest annual NPP increase during 1991–1999 appeared in alpine meadows, accounting for 32.3% of the increment of the whole region. Changes in solar radiation and temperature significantly influenced NPP variation, suggesting that solar radiation may be one of the major factors associated with changes in NPP.     

Surface flux response to interannual tropical Pacific sea surface temperature variability in AMIP models

 A systematic comparison of observed and modeled atmospheric surface heat and momentum fluxes related to sea surface temperature (SST) variability on interannual time scales in the tropical Pacific is conducted. This is done to examine the ability of atmospheric general circulation models (AGCMs) in the Atmospheric Model Intercomparison Project (AMIP) to simulate the surface fluxes important for driving the ocean on interannual time scales. In order to estimate the model and observed response to such SST variability, various regression calculations are made between a time series representing observed ENSO SST variability in the tropical Pacific and the resulting surface flux anomalies. The models exhibit a range of differences from the observations. Overall the zonal wind stress anomalies are most accurately simulated while the solar radiation anomalies are the least accurately depicted. The deficiencies in the solar radiation are closely related to errors in cloudiness. The total heat flux shows some cancellation of the errors in its components particularly in the central Pacific. The performance of the GCMs in simulating the surface flux anomalies seems to be resolution dependent and low-resolution models tend to exhibit weaker flux responses. The simulated responses in the western Pacific are more variable than those of the central and eastern Pacific but in the west the observed estimates are less robust as well. Further improvements in atmospheric GCM flux simulation through better physical parametrization is clearly required if such models are to be used to their full potential in coupled modeling and climate forecasting. Received: 24 August 1999 / Accepted: 11 September 2000     

Gap Filling and Quality Assessment of CO2 and Water Vapour Fluxes above an Urban Area with Radial Basis Function Neural Networks

Vertical turbulent fluxes of water vapour, carbon dioxide, and sensible heat were measured from 16 August to the 28 September 2006 near the city centre of Münster in north-west Germany. In comparison to results of measurements above homogeneous ecosystem sites, the CO₂ fluxes above the urban investigation area showed more peaks and higher variances during the course of a day, probably caused by traffic and other varying, anthropogenic sources. The main goal of this study is the introduction and establishment of a new gap filling procedure using radial basis function (RBF) neural networks, which is also applicable under complex environmental conditions. We applied adapted RBF neural networks within a combined modular expert system of neural networks as an innovative approach to fill data gaps in micrometeorological flux time series. We found that RBF networks are superior to multi-layer perceptron (MLP) neural networks in the reproduction of the highly variable turbulent fluxes. In addition, we enhanced the methodology in the field of quality assessment for eddy covariance data. An RBF neural network mapping system was used to identify conditions of a turbulence regime that allows reliable quantification of turbulent fluxes through finding an acceptable minimum of the friction velocity. For the data analysed in this study, the minimum acceptable friction velocity was found to be 0.15 m s⁻¹. The obtained CO₂ fluxes, measured on a tower at 65 m a.g.l., reached average values of 12 μmol m⁻² s⁻¹ and fell to nighttime minimum values of 3 μmol m ⁻² s⁻¹. Mean daily CO₂ emissions of 21 g CO₂ m⁻²d ⁻¹ were obtained during our 6-week experiment. Hence, the city centre of Münster appeared to be a significant source of CO₂. The half-hourly average values of water vapour fluxes ranged between 0.062 and 0.989 mmol m⁻² s⁻¹and showed lower variances than the simultaneously measured fluxes of CO₂.     

Towards constraining climate sensitivity by linear analysis of feedback patterns in thousands of perturbed-physics GCM simulations

A linear analysis is applied to a multi-thousand member “perturbed physics" GCM ensemble to identify the dominant physical processes responsible for variation in climate sensitivity across the ensemble. Model simulations are provided by the distributed computing project, climate prediction.net . A principal component analysis of model radiative response reveals two dominant independent feedback processes, each largely controlled by a single parameter change. The leading EOF was well correlated with the value of the entrainment coefficient—a parameter in the model’s atmospheric convection scheme. Reducing this parameter increases high vertical level moisture causing an enhanced clear sky greenhouse effect both in the control simulation and in the response to greenhouse gas forcing. This effect is compensated by an increase in reflected solar radiation from low level cloud upon warming. A set of ‘secondary’ cloud formation parameters partly modulate the degree of shortwave compensation from low cloud formation. The second EOF was correlated with the scaling of ice fall speed in clouds which affects the extent of cloud cover in the control simulation. The most prominent feature in the EOF was an increase in longwave cloud forcing. The two leading EOFs account for 70% of the ensemble variance in λ—the global feedback parameter. Linear predictors of feedback strength from model climatology are applied to observational datasets to estimate real world values of the overall climate feedback parameter. The predictors are found using correlations across the ensemble. Differences between predictions are largely due to the differences in observational estimates for top of atmosphere shortwave fluxes. Our validation does not rule out all the strong tropical convective feedbacks leading to a large climate sensitivity.     

Analysis of UVB values in the centre of the Iberian Peninsula

Records of UV data started in Madrid at the beginning of the 90's decade. After some quality control on the data a seven year period from 1996 to 2002 was selected to perform an analysis of daily and seasonal variability of solar ultraviolet B (UVB) values at the centre of the Iberian Peninsula. Instruments used are a Brewer MKIV spectrophotometer and a YES UVB-1 broadband pyranometer. Both instruments provided integrated daily values according with the Diffey spectrum although they have different measurement procedures. Data statistics from each instrument for the same period and statistical relationships between daily values from both instruments are shown. As broadband YES provided a record with fewer time gaps, analysis of UVI extreme values is performed using that instrument. The relationship between UVB values with ozone, cloudiness and visible radiation has been described in several papers for different places in the world. In this paper, those relationships are shown for Madrid. A discussion about specific problems found when trying to isolate every effective factor is also included. Some of those relationships could be helpful to retrieve UVB values from other ancillary data as visible radiation and cloudiness. Finally, a detailed study for 10 days in June 1997 when the maximum record of daily values from the present time series was obtained, highlight the close correlation between total ozone content and the daily variability of UVB for similar amounts of incoming radiation.     

我国太阳总辐射月总量资料的均一性检验及订正

采用建站在30a以上的月总辐射资料，以相关性最好的月份为代表月份，以经过均一性检验的日照资料为参考序列，用Potter法进行总辐射资料的均一性检验。结果表明，存在1个或2个间断点的测站占总数的58．6％；无间断点的测站占总数的34．3％。出现间断点的年份以20世纪70年代最多，其次是90年代。90年代初全国辐射仪器的换型对辐射资料没有普遍明显的影响。最后用回归法对存在间断点的绝大多数测站进行了均一性订正，订正后辐射和目照的年平均相关系拊兽谝增大平均增大近5％．