Milan Gocic

Soft computing approaches for forecasting reference evapotranspiration

The GP, SVM-FFA, ANN and SVM-Wavelet modeling of ET0 was reported.SVM-Wavelet had the smallest RMSE of 0.233mmday-1 in testing phase.The ANN model had the largest RMSE of 0.450mmday-1.SVM-Wavelet model was found to perform better than the GP, SVM-FFA and ANN models. Accurate estimation of reference evapotranspiration (ET0) is needed for planning and managing water resources and agricultural production. The FAO-56 Penman-Monteith equation is used to determinate ET0 based on the data collected during the period 1980-2010 in Serbia. In order to forecast ET0, four soft computing methods were analyzed: genetic programming (GP), support vector machine-firefly algorithm (SVM-FFA), artificial neural network (ANN), and support vector machine-wavelet (SVM-Wavelet). The reliability of these computational models was analyzed based on simulation results and using five statistical tests including Pearson correlation coefficient, coefficient of determination, root-mean-square error, absolute percentage error, and mean absolute error. The end-point result indicates that SVM-Wavelet is the best methodology for ET0 prediction, whereas SVM-Wavelet and SVM-FFA models have higher correlation coefficient as compared to ANN and GP computational methods.

Determination of the most influential weather parameters on reference evapotranspiration by adaptive neuro-fuzzy methodology

The monthly ET0 data were obtained by the Penman-Monteith method.ANFIS was applied for selection of the most influential ET0 parameters.Tmin, ea and sunshine hours are the most influential for ET0 estimation.Variables selection with ANFIS improves ET0 predictive accuracies.The ANFIS model can be used for ET0 estimation with high reliability. The adaptive neuro-fuzzy inference system (ANFIS) is applied for selection of the most influential reference evapotranspiration (ET0) parameters. This procedure is typically called variable selection. It is identical to finding a subset of the full set of recorded variables that illustrates good predictive abilities. The full weather datasets for seven meteorological parameters were obtained from twelve weather stations in Serbia during the period 1980-2010. The monthly ET0 data are obtained by the Penman-Monteith method, which is proposed by Food and Agriculture Organization of the United Nations as the standard method for the estimation of ET0. As the performance evaluation criteria of the ANFIS models the following statistical indicators were used: the root mean squared error (RMSE), Pearson correlation coefficient (r) and coefficient of determination (R2). Sunshine hours are the most influential single parameter for ET0 estimation (RMSE=0.4398mm/day). The obtained results indicate that among the input variables sunshine hours, actual vapor pressure and minimum air temperature, are the most influential for ET0 estimation. The maximum relative humidity and maximum air temperature are the most influential optimal combination of two parameters (RMSE=0.2583mm/day).

Extreme learning machine based prediction of daily dew point temperature

An ELM-based model is proposed to predict daily dew point temperature.Weather data for two Iranian stations with different climate conditions were used.ELM model enjoys much greater predictions capability than SVM and ANN.Application of the proposed ELM model would be highly promising and appealing. The dew point temperature is a significant element particularly required in various hydrological, climatological and agronomical related researches. This study proposes an extreme learning machine (ELM)-based model for prediction of daily dew point temperature. As case studies, daily averaged measured weather data collected for two Iranian stations of Bandar Abass and Tabass, which enjoy different climate conditions, were used. The merit of the ELM model is evaluated against support vector machine (SVM) and artificial neural network (ANN) techniques. The findings from this research work demonstrate that the proposed ELM model enjoys much greater prediction capability than the SVM and ANN models so that it is capable of predicting daily dew point temperature with very favorable accuracy. For Tabass station, the mean absolute bias error (MABE), root mean square error (RMSE) and correlation coefficient (R) achieved for the ELM model are 0.3240?C, 0.5662?C and 0.9933, respectively, while for the SVM model the values are 0.7561?C, 1.0086?C and 0.9784, respectively and for the ANN model are 1.0324?C, 1.2589?C and 0.9663, respectively. For Bandar Abass station, the MABE, RMSE and R for the ELM model are 0.5203?C, 0.6709?C and 0.9877, respectively whereas for the SVM model the values are 1.0413?C, 1.2105?C and 0.9733, and for the ANN model are 1.3205?C, 1.5530?C and 0.9617, respectively. The study results convincingly advocate that ELM can be employed as an efficient method to predict daily dew point temperature with much higher precision than the SVM and ANN techniques.

Spatio-temporal patterns of precipitation in Serbia

The monthly precipitation data from 29 synoptic stations for the period 1946–2012 were analyzed using a number of different multivariate statistical analysis methods to investigate the spatial variability and temporal patterns of precipitation across Serbia. R-mode principal component analysis was used to study the spatial variability of the precipitation. Three distinct sub-regions were identified by applying the agglomerative hierarchical cluster analysis to the two component scores: C1 includes the north and the northeast part of Serbia, while C2 includes the western part of Central Serbia and southwestern part of Serbia and C3 includes central, east, south and southeast part of Serbia. The analysis of the identified sub-regions indicated that the monthly and seasonal precipitation in sub-region C2 had the values above average, while C1 and C3 had the precipitation values under average. The analysis of the linear trend of the mean annual precipitation showed an increasing trend for the stations located in Serbia and three sub-regions. From the result of this analysis, one can plan land use, water resources and agricultural production in the region.

Precipitation Estimation Using Support Vector Machine with Discrete Wavelet Transform

Precipitation prediction is of dispensable importance in many hydrological applications. In this study, monthly precipitation data sets from Serbia for the period 1946–2012 were used to estimate precipitation. To fulfil this objective, three mathematical techniques named artificial neural network (ANN), genetic programming (GP) and support vector machine with wavelet transform algorithm (WT-SVM) were applied. The mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), Pearson correlation coefficient (r) and coefficient of determination (R2) were used to evaluate the performance of the WT-SVM, GP and ANN models. The achieved results demonstrate that the WT-SVM outperforms the GP and ANN models for estimating monthly precipitation.

Soft-Computing Methodologies for Precipitation Estimation: A Case Study

The current paper presents an investigation of the accuracy of soft-computing techniques in precipitation estimation. The monthly precipitation data from 29 synoptic stations in Serbia from 1946 to 2012 are used as a case study. Despite a number of mathematical functions having been proposed for modeling precipitation estimation, the models still have disadvantages such as being very demanding in terms of calculation time. Soft computing can be used as an alternative to the analytical approach, as it offers advantages such as no required knowledge of internal system parameters, compact solutions for multivariable problems, and fast calculation. Because precipitation prediction is a crucial problem, a process which simulates precipitation with two soft-computing techniques was constructed and presented in this paper, namely, adaptive neurofuzzy inference (ANFIS) and support vector regression (SVR). In the current study, polynomial, linear, and radial basis function (RBF) are applied as the kernel function of the SVR to estimate the probability of precipitation. The performance of the proposed optimizers is confirmed with the simulation results. The SVR results are also compared with the ANFIS results. According to the experimental results, enhanced predictive accuracy and capability of generalization can be achieved with the ANFIS approach compared to SVR estimation. The simulation results verify the effectiveness of the proposed optimization strategies.

Analysis of trends in reference evapotranspiration data in a humid climate

Abstract Statistically significant FAO-56 Penman-Monteith (FAO-56 PM) and adjusted Hargreaves (AHARG) reference evapotranspiration (ET0) trends at monthly, seasonal and annual time scales were analysed by using linear regression, Mann-Kendall and Spearman’s Rho tests at the 1 and 5% significance levels. Meteorological data were used from 12 meteorological stations in Serbia, which has a humid climate, for the period 1980–2010. Web-based software for conducting the trend analyses was developed. All of the trends significant at the 1 and 5% significance levels were increasing. The FAO-56 PM ET0 trends were almost similar to the AHARG trends. On the seasonal time scale, for the majority of stations significant increasing trends occurred in summer, while no significant positive or negative trends were detected by the trend tests in autumn for the AHARG series. Moreover, 70% of the stations were characterized by significant increasing trends for both annual ET0 series. Editor Z.W. Kundzewicz; Associate editor S. Grimaldi Citation Gocic, M. and Trajkovic, S., 2013. Analysis of trends in reference evapotranspiration data in a humid climate. Hydrological Sciences Journal, 59 (1), 165–180.

Comparative analysis of reference evapotranspiration equations modelling by extreme learning machine

Forecasting ET0 is important for agricultural production and irrigation scheduling.Differences of performance between compared ELM models are not very significant.Results showed that ELM ET0,AHARG can be applied to forecast ET0 effectively. This study presents an extreme learning machine (ELM) approach, for estimating monthly reference evapotranspiration (ET0) in two weather stations in Serbia (Nis and Belgrade stations), for a 31-year period (1980-2010). The data set including minimum and maximum air temperatures, actual vapour pressure, wind speed and sunshine hours was employed for modelling ET0 using the adjusted Hargreaves (ET0,AHARG), Priestley-Taylor (ET0,PT) and Turc (ET0,T) equations. The reliability of the computational model was accessed based on simulation results and using five statistical tests including mean absolute percentage error (MAPE), mean absolute deviation (MAD), root-mean-square error (RMSE), Pearson correlation coefficient (r) and coefficient of determination (R2). The validity of ELM modelled ET0 are compared with the FAO-56 Penman-Monteith equation (ET0,PM) which is used as the reference model. For the Belgrade and Nis stations, the ET0,AHARG ELM model with MAPE=9.353 and 10.299%, MAD=0.142 and 0.151mm/day, RMSE=0.180 and 0.192mm/day, r=0.994 and 0.992, R2=0.988 and 0.984 in testing period, was found to be superior in modelling monthly ET0 than the other models, respectively.

Long-Term Precipitation Analysis and Estimation of Precipitation Concentration Index Using Three Support Vector Machine Methods

The monthly precipitation data from 29 stations in Serbia during the period of 1946–2012 were considered. Precipitation trends were calculated using linear regression method. Three CLINO periods (1961–1990, 1971–2000, and 1981–2010) in three subregions were analysed. The CLINO 1981–2010 period had a significant increasing trend. Spatial pattern of the precipitation concentration index (PCI) was presented. For the purpose of PCI prediction, three Support Vector Machine (SVM) models, namely, SVM coupled with the discrete wavelet transform (SVM-Wavelet), the firefly algorithm (SVM-FFA), and using the radial basis function (SVM-RBF), were developed and used. The estimation and prediction results of these models were compared with each other using three statistical indicators, that is, root mean square error, coefficient of determination, and coefficient of efficiency. The experimental results showed that an improvement in predictive accuracy and capability of generalization can be achieved by the SVM-Wavelet approach. Moreover, the results indicated the proposed SVM-Wavelet model can adequately predict the PCI.

Drought Characterisation Based on Water Surplus Variability Index

Drought assessment, characterisation and monitoring increasingly requires considering not only precipitation but also the other meteorological parameters such as an evapotranspiration. Thus, some new drought indices based on precipitation and evapotranspiration have been developed. This study introduces a new drought index named the water surplus variability index (WSVI). The procedure to estimate the index involves accumulation water surplus at different time scales. To approve the proposed procedure, the WSVI is compared with the standardized precipitation index (SPI), the reconnaissance drought index (RDI) and the standardized precipitation evapotranspiration index (SPEI) based on 1-, 3-, 6- and 12-month timescales using data from several weather stations located in regions with different aridity index. Near perfect agreement (d ~ 1) between WSVI and SPI, RDI and SPEI was indicated in humid and sub-humid locations. The results also showed that the correlation coefficients between WSVI and SPI, RDI and SPEI were higher for semi-arid stations than for arid ones.