Influence of the bandwidth in the harmonic search to optimize the mixed univariate Gumbel function

Different regions of Mexico are affected temporarily by tropical cyclones whose evidence is reflected in the record of annual maximum flows of stream gages. The statistical analysis of these records presents two different trends represented by the Gumbel mixed distribution function. In order to achi...

Full description

Saved in:
Bibliographic Details
Published inTheoretical and applied climatology Vol. 139; no. 1-2; pp. 801 - 813
Main Authors Molina–Aguilar, Juan Pablo, Gutiérrez–López, M. Alfonso
Format Journal Article
LanguageEnglish
Published Vienna Springer Vienna 2020
Springer
Springer Nature B.V
Subjects
Analysis
Aquatic Pollution
Atmospheric Protection/Air Quality Control/Air Pollution
Atmospheric Sciences
Bandwidths
Climate science
Climatology
Cyclones
Distribution (Probability theory)
Distribution functions
Earth and Environmental Science
Earth Sciences
Forecasts and trends
Heuristic methods
Hurricanes
Logarithms
Optimization
Original Paper
Parameters
Probability theory
Records
Statistical analysis
Statistical methods
Tropical climate
Tropical cyclones
Waste Water Technology
Water Management
Water Pollution Control
Online AccessGet full text

Cover

More Information
Summary:Different regions of Mexico are affected temporarily by tropical cyclones whose evidence is reflected in the record of annual maximum flows of stream gages. The statistical analysis of these records presents two different trends represented by the Gumbel mixed distribution function. In order to achieve a reliable character of the historical information, it is necessary to determine the parameters of the function; therefore, the present work shows its optimization by means of the metaheuristic technique harmonic search verifying the influence of the bandwidth parameter in the solution. It was observed that the arithmetic expressions that define the bandwidth generate a better performance with respect to those that include exponential or logarithms. It is concluded that the technique achieves the optimization of the five parameters of the univariate Gumbel mixed function in an agile way, in particular the probability and the non-cyclonic elements, decreasing the adjustment error generated with respect to the classical methodologies, which improves upon having extensive historical records.
ISSN:0177-798X
1434-4483
DOI:10.1007/s00704-019-03001-9