Resumen

CAMPOS-ARANDA, Daniel Francisco. Simple Probability Model for Frequencies Analysis in Extreme Hydrological Records with Trend. Tecnol. cienc. agua [online]. 2016, vol.7, n.3, pp.171-186. ISSN 2007-2422.

Due to climate changes, land use mainly by urbanization and hydraulic works, the records of extreme hydrological data are no longer stationary, that is that their statistical properties are changing over time. When an annual record of floods or maximum rainfall shows a trend statistically different from zero, the frequencies analysis aimed at estimating their design events should take into account such non-random behavior. The simplest probabilistic model that can be established to process non-stationary records relies on the lognormal distribution with two fitting parameters, with time-varying mean as explanatory covariate, according to a logarithmic linear regression. This model allows estimating the impact, for example at the end of a decade in the future, of the upward (downward) trend on the obtained predictions at the end of a record. The latter through two indicators: the predictions magnification (diminution) factors and the reduction (increase) in the recurrence of an event of a given return period. Non-stationary frequencies analysis with such a model is applied to two records of annual floods of the Hydrological Region núm. 10 Sinaloa and one annual daily maximum precipitation of the Hydrological Region núm. 37 El Salado. Conclusions highlight the simplicity and utility of the described model as an initial approach to the non-stationary frequencies analyses.

Palabras llave : Annual floods; annual maximum daily precipitation; log-normal distribution; linear regression; standard error of fit; predictions magnification factors.