Trend detection of atmospheric time series

This paper is aimed at atmospheric scientists without formal training in statistical theory. Its goal to (1) provide a critical review of the rationale for trend analysis time series typically encountered field chemistry, (2) describe range trend-detection methods, and (3) demonstrate effective means conveying results general audience. Trend detections chemical composition data are often challenged by variety sources uncertainty, which behave differently other environmental phenomena such as temperature, precipitation rate, or stream flow, may require specific methods depending on science questions be addressed. Some uncertainty can explicitly included model specification, autocorrelation seasonality, but some inherent uncertainties difficult quantify, heterogeneity measurement due combined effect short long term natural variability, instrumental stability, aggregation from sparse sampling frequency. Failure account these might result an inappropriate inference trends their estimation errors. On hand, variation extreme events interesting different scientific questions, example, frequency extremely high surface ozone relevance human health. In this study we aim detection addressing levels complexity species, that incorporation scientifically interpretable covariates outperform pure numerical curve fitting techniques terms reduction improved predictability, illustrate based quantiles insight beyond standard mean median estimates, (4) present advanced method quantifying regional inter-site correlations multisite data. All demonstrations observed trace gases relevant applied sets.