7 November 2018 in ITM Web of Conferences
Kernel density estimation is a technique for estimation of probability density function that is a must-have enabling the user to better analyse the studied probability distribution than when using a traditional histogram. Unlike the histogram, the kernel technique produces smooth estimate of the pdf, uses all sample points' locations and more convincingly suggest multimodality. In its two-dimensional applications, kernel estimation is even better as the 2D histogram requires additionally to define the orientation of 2D bins. Two concepts play fundamental role in kernel estimation: kernel function shape and coefficient of smoothness, of which the latter is crucial to the method. Several real-life examples, both for univariate and bivariate applications, are shown.
1 April 1998 in Journal of Hydrology
1 February 2002 in Journal of Hydrology
1 April 2011 in Hydrology Research
This study discusses an application of heavy-tailed distributions to modelling of annual peak flows in general and of Polish data sets in particular. One- and two-shape parameter heavy-tailed distributions are obtained by transformations of random variables. The correct selection of a flood frequency model with emphasis on heavy-tailed distribution discrimination is then discussed. If a distribution is wrongly assumed, the error, in the upper quantile, arising as a result, depends on the method of parameter estimation and is shown analytically for three methods. Asymptotic and sampling values (got by simulation) were assessed for the pair log-Gumbel (LG) as a false distribution and log-normal (LN) as a true distribution. Comparing the upper quantiles of various distributions with the same values of moments, it is found that heavy-tailed distributions do not consistently provide higher flood frequency estimates than do soft-tailed distributions. Based on L-moment ratio diagrams and the test of linearity on log–log plots, it is concluded that Polish datasets of annual peak flows should be modelled using soft-tailed distributions, such as the three-parameter Inverse Gaussian, rather than heavy-tailed distributions.
16 December 2014 in International Journal of Climatology
The purpose of this paper is to describe the variability of sunshine duration and air temperature and to assess the relationship between these variables. The paper is based on daily means of air temperature and daily sunshine duration totals from the 1884–2012 period. The variability analyses of these variables for the whole period is supplemented by additional trend analyses for two separate time periods described in the literature as dimming (1958–1985) and brightening (1985–2012). All the data come from the Jagiellonian University Research Station in Krakow. In Krakow, there is a decrease in the number of sunshine hours in the 1950s to 1980s period followed by an increase in the last two decades of the 20th century. The long-term course of air temperature in Krakow confirms the results obtained in other places in the world and exhibits increasing warming of the climate. The temperature increase is visible in each month but is highest in winter (exceeding 1.5 °C per 100 years). Therefore, it may be assumed that the increase in air temperature, which has been particularly strong since the 1980s, showed the full effect of global warming in the brightening period that was previously masked by global dimming as shown in the sunshine duration records. The time course of air temperature for the whole study period (1884–2012) shows a statistically significant positive linear trend. The present warming is influenced probably by the air-quality improvement (brightening period) and an increase in sunshine duration in the years 1980–2012.
7 February 2006 in Stochastic Environmental Research and Risk Assessment