REGRESSION MAXIMUM AND ITS USE TO HYDROECOLOGICAL RESEARCH

Abstract

One of the actual problems of hydrochemistry there is influence to solar activity on condition chemical substance in water environment. The goal of the work it is possible estimation relationship some hydrochemical factors and solar activity (Wolf's Numbers W). To this effect the research was given for concentration 〖NH〗_4^+, 〖NO〗_2^- , 〖NO〗_3^- in water large flat river of Ukraine (1991 … 2010 years). The ion concentrations denominated in mg on litre. Were they also analysed given over Wolf's numbers for this period (1991 … 2010 years). Since these dependencies for the best reveals itself at a rate of extreme values that for each value of the Wolf's number selected only maximum to concentrations of the ion. Accordingly building to regressions executed not on all raw data’s but for maximum values concentration only. For this regression execute in two steps: STEP 1: Selected only one (maximum) value concentration of the ion for each unique value of Wolf's number. STEP 2: To prepared by specified way data is used for regression. Experiment has shown that linear regression in this instance to use it is impossible (got the horizontal line). Experiment has also shown adequacy of the use polynomial quantile regression (so it was used this type of regression). It was used median egression (quantile regression for Q=0,5). Median regression demonstrates the observable reduction a concentration of 〖NH〗_4^+, 〖NO〗_2^-,〖 NO〗_3^- when increasing of values Wolf's Numbers. Offered in article “regression for maximus” important for decision of the practical problems of hydroecology. The general trend in behavioyr of Wolf's Numbers can be forecasting on times of the ten years onward. This means as the general trend in behavioyr of 〖NH〗_4^+, 〖NO〗_2^- , 〖NO〗_3^- also can be forecasting on times of ten of years onward (the factor of the purity of natural water). The natural water more pure (the minimum concentration of 〖NH〗_4^+, 〖NO〗_2^- , 〖NO〗_3^-) under high solar activity (high values of Wolf's Numbers).