Abstract:
This study uses extreme value theory (EVT) to assess the use of extreme value distributions of the
Gumbel and Generalized Extreme value (GEV) models on the maximum quarter-hour, half-hour,
one-hour and eight-hour average of concentrations of carbon monoxide (CO) data, measured
between February 2009 and February 2010 in Akure, Ondo State, Nigeria. CO was measuredusing
Lascar EL-USB-CO portable data logger with a sampling frequency of 30 seconds, installed above
head height on a lamp post in the middle of the road at Oja-Oba on Oba Adesida road. Data analysis
was achieved using the Extreme Package in the R statistical software. Parameters for all the
distributions are estimated using the Maximum Likelihood Estimator (MLE). Performance
indicators namely: the coefficient of variation and Root Mean Square Error (RMSE), are used to
find thegoodness-of-fit of the distributions. The best distribution is selected based on the
coefficient of variation and the smallest error measures. The results showed that the GEV is the
best fit for quarter-hour, half-hour, one-hour maximum concentrations, while Gumbel was best for
eight-hour maximum concentration. The result of this study suggests that extreme value theory is
a useful tool for urban air quality management decision making.
