ПРОГНОЗУВАННЯ ПОЛІВ ГІДРОМЕТЕОРОЛОГІЧНИХ ВЕЛИЧИН МЕТОДОМ СЕЛЕКТИВНОЇ ЛОКАЛЬНОЇ АПРОКСИМАЦІЇ

Abstract

Aim of the investigations this estimation further possibilities of the forecast natural (hydrometeorological) fields by means of chaotical dynamic methods. Method of Local Approximation of the zero order (LA∅-method) was used. Raw data this mean – month fields of minimal and maximal temperature and air pressure (ground level for all this fields). At raw data beforehand were not filtered. Quality of the forecast defined by way of the comparison forecasting values and true values. True and forecasting fields are present itself matrix. So fields transformed to vector – columns and then found linear factor to correlation between them. Herewith suppose that linear factor to correlation measures quality of the forecast (outside of dependencies from probabilistic law of the raw datas). For conservation of the goodness of the forecast prediction itself run for one step ahead (one month ahead). For current vector of the state found only one analogue (for all considered in article examples). Variety of the method was designed to Local Approximation which authors have named as Selective Local Approximation. Selective Local Approximation method at the average enlarges factor to correlation between real and forecasting fields on 10% in contrast with classical method of the Local Approximation. Results of the forecasting hydrometeorological fields by Selective Local Approximation method were presented on each month for corresponding years. Special criteria quality forecast were incorporated (two groups of such criterion). Indirect criteria consider correspondence to between “Sample” and “Analog (ue)”. Direct criteria consider correspondence to between real field and forecasting field. It was found that between direct and indirect criteria is present enough strong intercoupling (than better forecasting quality with position of the indirect criterion that usually better and quality most forecast). Also quality of the forecast is considered with position nonparametric analog of the standard deviation (scattering forecasting results). In article is show that quality of the forecast with position of the average factor of correlations is found as inverse value with average of scatter data. With position of the forecasting for public facilities is offered corresponding to categorization quality forecast. In offered categorizations quality forecast is defined numbers of the months (from 12)of which forecast unsatisfactory. Researches have shown that for all considered fields forecasting grows worse for a period of begin “springtimes … a midde summer”. Is it specifically chosen very large difference between extremly hard criteria given to article and criteria (quality of the forecasting) used other authors.