NONPARAMETRIC INTERVAL ESTIMATION OF QUANTILES OF EMPIRICAL DISTRIBUTIONS IN THE PROBLEM OF PRONOSIS OF CHARACTERISTIC VALUES OF SNOW LOADS

Abstract

The establishment of characteristic values of climatic loads, in particular snow loads, is carried out by statistical estimation of available empirical data. Known estimation methods based on approximating the right “tail” part of empirical data series with different types of distributions of extreme values do not have a certain level of statistical confidence of the resulting estimate. A new technique for nonparametric interval estimation of quantiles of an empirical distribution with five-point alignment of empirical data is proposed, which allows one to set the required level of statistical confidence of the result. The effectiveness of the methodology is shown by the example of estimating the median quantile level of 0.98 (corresponding to a return period of 50 years) when solving the problem of predicting the characteristic values of snow load, during which the territory of the Republic of Belarus was zoned according to snow loads, taking into account the identified dependencies of the load on the terrain height.