A space-time skew-t model for threshold exceedances.

Area of Science:

• Environmental science
• Statistics
• Atmospheric chemistry

Background:

• Regulatory compliance necessitates monitoring air quality, specifically ozone levels, against set thresholds.
• Accurate prediction of ozone exceedances is crucial for environmental protection and public health.

Purpose of the Study:

• To introduce a novel space-time statistical model for predicting threshold exceedances in air quality data.
• To improve the accuracy of ozone exceedance predictions using a skew-t process.

Main Methods:

• Developed a space-time model incorporating a random partition for spatial dependence and thresholding for tail analysis.
• Introduced a transformed autoregressive AR(1) time-series component for temporal dependence.
• Enabled high-dimensional Bayesian inference with computational efficiency comparable to geostatistical methods.

Main Results:

• The proposed skew-t process model demonstrated superior performance in predicting high-level ozone exceedances.
• The model improved prediction accuracy over traditional Gaussian and max-stable statistical methods.
• Applied successfully to ozone data analysis for July 2005.

Conclusions:

• The new skew-t based space-time model offers a more accurate approach for assessing air quality regulatory compliance.
• The model's ability to handle spatial and temporal dependencies and focus on extreme values enhances prediction of ozone exceedances.
• This method provides a computationally efficient tool for large-scale environmental data analysis.