Comparison of Time Series Trajectories at Multiple Spatiotemporal Resolutions: A Case Study of Soybeans in Western Bahia, Brazil

Abstract

Analyzing the change during a time series requires methods to summarize the change patterns meaningfully while allowing scientists to ignore unnecessary details. This paper presents developed methods that analyze change trajectories of a time series and extends the concept to multiple-resolution analysis. I analyze six trajectories and three components of change involving soybeans at multiple resolutions during 32 annual time intervals from 1989−2021. Results show that the largest component of change is Alternation, while the smallest component of change is Exchange. Alternation derives from pairs of loss and gain at the same location. The multiple-resolution analysis shows how Alternation and Exchange shrink as temporal and spatial resolutions coarsen. The trajectories of change describe change patterns during sequential time intervals in a way that allows scientists to focus on meaningful change patterns. The methods can help to detect hotspots of change and regions of suspected map error. My timeseriesTrajectories R package, available for free at https://github.com/bilintoh/timeseriesTrajectories performs the analysis.