New Zealand Statistical Association 2024 Conference

Justin Murphy

Stats NZ

A tool to evaluate environmental trend estimation methods

This is joint work with Tiana Whitehead, Pubudu Senanayake

Accurately measuring environmental trends is crucial for informing the effectiveness of environmental management policies and identifying where resources need to be allocated. We were commissioned by the Ministry for the Environment (MfE) to assess current and alternative methods for environmental trend estimation. In order to evaluate trend estimation methods, we sought to create realistic synthetic data with known underlying trends, allowing us to determine the accuracy of various trend estimates. We created a Shiny tool that enables users to decompose real environmental time series into trend, seasonal, and noise components using STL decomposition. The tool allows users to replace the estimated trend component with a customizable trend, choosing from multiple patterns. This customizable trend is then recombined with the original seasonal and noise components to produce a new realistic synthetic time series. In addition, users can set custom time windows for trend estimation and scale the seasonal and noise components up and down. The Shiny tool then allows for the application of several trend estimation methods to the synthetic data, facilitating a comparison of these methods against the known trend component, and the sensitivity of the methods to time-window and the magnitude of the seasonal and noise components.