KEYWORDS: Autoregressive models, Data modeling, Statistical analysis, Time series analysis, Systems modeling, Statistical modeling, Deep learning, Visualization, Machine learning, Interpolation
A scheme and a basic set of software experiments for time series forecasting using an integrated autoregressive-moving average model (Box-Jenkins model) are presented. The model is based on the assumption that there is some relationship between neighboring values of a time series. In particular, the hypothesis is accepted that the time series contains three components: autoregressive, integrated and moving average. The application of the ARIMA model for forecasting time series using the statistical modelling language R - from the stage of data loading and preprocessing to the prediction of future values - is presented.
A practical study of statistical modelling language packages R has been carried out using regularization algorithms, more precisely one of the algorithms called the Extreme Learning Machine (ELM). Due to its simple implementation, ELM requires less researcher intervention in setting its parameters. At the same time, the generalization performance of ELM is not sensitive to the dimensionality of the feature space (the number of hidden nodes). Even on a medium-power personal computer, this class of neural networks has made it possible to perform numerous experiments on model building, forecasting and identifying cause-effect relationships in meteorological time series, downloaded from the climate monitoring system of IMCES SB RAS in a reasonable amount of time.
KEYWORDS: Artificial neural networks, Data modeling, Education and training, Atmospheric modeling, Air temperature, Meteorology, Environmental monitoring, Climatology, Ultrasonics, Systems modeling
An implementation of a week-ahead air temperature and atmospheric pressure forecast using a multilayer perceptron is presented (MLP). According to the specified meteorological parameters, data preparation, implementation and performance evaluation were performed for two MLP models. The MLP architecture was a s upervised feed -forward neural network with five hidden nodes and twenty iterations (repetitions). The obtained values of the ris k function (in this case, the standard deviation of the MSE) in both implementations are quite large.
KEYWORDS: Digital filtering, Optical filters, Bandpass filters, Signal to noise ratio, Signal processing, Nonlinear filtering, Linear filtering, Binary data, Ultrasonics, Temperature metrology
The results of a study noise filtering in temperature time series measuring channels using basic frequency filtering algorithms are presented. The study was carried out from the perspective of high frequency random noise suppression. Temperature time series filtering was compared using sliding window procedures with adaptive bandwidth selection, median sliding window, and Baxter-King band-pass approximation. Experimental comparative analysis of filtering efficiency was carried out using the statistical programming language R and open libraries with sliding regression methods. The recorders of surface atmosphere parameters were ultrasonic meteorological stations located at the test site of IMKES SB RAS.
The results of correlation and regression analysis of meteorological data from an ultrasonic weather station are presented. Automatic weather station AMK-03 was equipped with three units for measuring meteorological parameters installed on the radio mast at heights of 2, 8 and 28 meters. Discreteness of measurements was 1 min. Time series with temperature, wind speed, atmospheric pressure and relative humidity obtained for the period from 01.01.2017 to 31.12.2017 were processed in software experiments. The methodology of correlation-regression analysis meteorological data, additionally, was used for diagnostics of weather station operation. Inaccuracies in the measurement of some parameters surface layer of the atmosphere have been revealed. The conducted control measurements and processing allowed to improve the measuring system of automatic weather station AMK-03.
The results of the nonlinearity analysis of the temperature time series with different sampling steps are presented. The values of statistical functions describing the amount of information contained in one time series relative to another are calculated. The graphs of the Lyapunov exponent plotted. The nonlinearity of the studied time series was checked using White's test and Terasvirt's test. In both tests, Fisher's criteria and chi-square's criteria reject the null hypothesis, i.e. confirm the nonlinearity nature of the time series under study.
The results of mathematical and software development for experimental studies of the surface atmospheric turbulence structure are presented. High-frequency measurements of ultrasonic thermoanemometer are used as data. Data processing and calculation of atmospheric turbulence parameters were carried out according to the schemes of sampling and scaling of meteorological parameters under study using semi-empirical methods of describing heat, moisture, amount of motion in the surface layer and Monin-Obukhov's theory of similarity. The calculated prognostic values of parameters are sufficient to estimate the dynamic regime of turbulence in the surface layer of the atmosphere, including estimates of the possibility of temperature inversion formation in the atmosphere and determination of the stability class of atmospheric stratification.
The results of the study of meteorological series of observations using the methodology of singular spectral analysis are presented. High-frequency measurements of ultrasonic weather stations located at the testing ground of IMKES SO RAS were used as data. The processing of meteorological data included two complementary stages - decomposition and reconstruction. At the decomposition stage, the meteorological series of observations were transformed into a multidimensional series by forming a trajectory matrix and its decomposition into singular vectors - sets of additive components. At the reconstruction stage, various groups of components formed reconstructed rows, interpreted as trend, harmonic and noise components of the meteorological series structure.
The paper describes program experiments on noise filtering by using recursive filters and wavelet-filtering frequency algorithms during processing and systematization of natural pulsed electromagnetic noises of Earth registered by ground multichannel geophysical loggers. The results on iterative use of Kalman filter are included.
The paper describes the analysis of correlation dependences between meteorological parameters for a series of observations obtained at station with a synoptic index of 29430. It was shown that the parameters with strong correlation dependences have virtually unchanging correlation coefficient for every time scale of the sample. In the case of meteorological parameters with weak correlation coefficient, there are non-significant jumps in the correlation coefficient values. For meteorological parameters with moderate correlation coefficient, increasing the sample time scale leads to stabilization of correlation coefficient values.
KEYWORDS: Data processing, Computing systems, Distributed computing, Data acquisition, Meteorology, Parallel computing, Data storage, Climatology, Computer science, Algorithm development
The paper describes a software complex for designing a horizontally-scalable distributed information-calculation platform with loosely coupled calculation nodes. The platform is intended information support for parallel processing of multi-dimensional data and large time series. The technological scheme for platform design and deployment includes a cluster of processing nodes and a cluster of storage nodes which provide their services if requested by researcher. The main node of each cluster is the command center. Storage management center coordinates functional data processing according to instructions received from researchers. Applications are designed as jnlp-files which ensures their functionality on research terminals.
The article describes an iterative parallel phase grouping algorithm for temperature field classification. The algorithm is based on modified method of structure forming by using analytic signal. The developed method allows to solve tasks of climate classification as well as climatic zoning for any time or spatial scale. When used to surface temperature measurement series, the developed algorithm allows to find climatic structures with correlated changes of temperature field, to make conclusion on climate uniformity in a given area and to overview climate changes over time by analyzing offset in type groups. The information on climate type groups specific for selected geographical areas is expanded by genetic scheme of class distribution depending on change in mutual correlation level between ground temperature monthly average.
Basing upon example of Wolf number series synchronous analysis and temperature values from 818 meteorological stations in the Northern hemisphere (1955-2010) it is shown that, for studied series, the components that differ by distinctive features matching and mismatching display extreme properties. The histograms of the primary temperature series coincide with histograms of their components except for the range of ± 3°С. Second initial moments of Wolf numbers’ components match climate geography and end up in two zones with width and difference in-between equal to about a third of possible change amount. Correlating synchronization features of geosphere processes initiated by external influence, with the use of physical-geographical hierarchy, allows to solve the classification task for temperature field; i.e. it allows to decompose initial sets into subsets containing strongly connected components. There were no discrepancies with known ideas about climate processes.
The general trend of modern ecological geophysics is changing priorities towards rapid assessment, management and prediction of ecological and engineering soil stability as well as developing brand new geophysical technologies. The article describes researches conducted by using multi-canal geophysical logger MGR-01 (developed by IMCES SB RAS), which allows to measure flux density of very low-frequency electromagnetic radiation. It is shown that natural pulsed electromagnetic fields of the earthen lithosphere can be a source of new information on Earth’s crust and processes in it, including earthquakes. The device is intended for logging electromagnetic processes in Earth’s crust, geophysical exploration, finding structural and lithological inhomogeneities, monitoring the geodynamic movement of Earth’s crust, express assessment of seismic hazards. The data is gathered automatically from observation point network in Siberia
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.