Analysis of PM10 hourly concentration via Dynamic Quantile Regression

  • Patricio, S. C. (Examiner)
  • Fabio A. Fajardo Molinares (Examiner)
  • Alessandro J. Q. Sarnaglia (Supervisor)

Activity: Examination and external supervisionExamination

Description

The bachelor thesis evaluates the effect of pollutants and meteorological variables on hourly concentrations of particulate matter with a diameter smaller than 10 micrometers (PM10). Specifically, hourly data collected in 2018 in Cariacica, Brazil, were analyzed. For data adjustment, dynamic quantile models were used, and the analysis was performed from a Bayesian perspective. The effect of each predictor variable at different response quantiles was evaluated to investigate changes in its importance at different response levels. The hourly concentrations of the predictors showed statistical significance with the response levels in almost all observed quantiles. In general, it was found that the effect of covariates tends to be greater in the upper tail of the PM10 distribution. The predictive performance was also investigated in data discarded from the model estimation step. These results show that the proposed model showed excellent predictive capacity, and the predicted quantiles were practically equal to the nominal values.
PeriodOct 2022