Nonparametric functional data estimation applied to ozone data: prediction and extreme value analysis.

Area of Science:

• Environmental Science
• Statistics
• Data Analysis

Background:

• Extreme value theory is commonly used for air pollution studies, typically employing parametric generalised extreme value (GEV) distributions.
• Parametric methods involve fitting a GEV distribution to extreme values to calculate return levels and other key metrics.
• There is a need for alternative statistical approaches to enhance the analysis of environmental data.

Purpose of the Study:

• To propose and evaluate nonparametric functional data methods as alternatives to classical parametric approaches for extreme value analysis in ozone data.
• To apply these nonparametric methods to real-world ozone concentration data from UK monitoring stations.
• To assess the predictive capabilities of functional data analysis for stratospheric ozone concentrations.

Main Methods:

• Utilized nonparametric functional data analysis, a statistical methodology for curve or multi-dimensional data.
• Employed nonparametric curve estimation techniques in conjunction with functional data analysis.
• Applied generalised extreme value (GEV) distribution for classical parametric comparison.
• Used Autoregressive Integrated Moving Average (ARIMA) models for time series analysis comparison.

Main Results:

• Nonparametric estimators demonstrated satisfactory performance in analyzing maximum ozone values from UK monitoring stations.
• The nonparametric methods outperformed classical parametric estimators in accuracy and behavior.
• Functional data analysis showed promise in predicting stratospheric ozone concentrations, with results comparable to ARIMA models.

Conclusions:

• Nonparametric functional data methods offer a robust and effective alternative to traditional parametric approaches for extreme value analysis of ozone data.
• These methods provide improved accuracy in estimating and predicting ozone levels, crucial for environmental monitoring.
• Functional data analysis presents a valuable tool for both environmental pollution studies and stratospheric ozone concentration prediction.