Trec: An R package for trend estimation and classification to support integrated ecosystem assessment of the marine ecosystem and environmental factors

Abstract

Solvang and Planque [ICES Journal of Marine Science, 77, pp.2529–2540, (2020)] provided a trend estimation and classification (TREC) approach to estimating dominant common trends among multivariate time series observations. This approach was developed to improve communication among stakeholders like marine managers, industry representatives, non-governmental organizations, and governmental agencies as they investigate the common tendencies between a biological community in a marine ecosystem and the local environmental factors. The tasks of trend estimation and classification in the original computational procedure have been revised, and new features include an automatic icon assignment algorithm using a multinomial logistic discriminator. In this paper, we present R package trec. Implementation of this package involves three partitions corresponding to TREC1) estimating trends from observed time series data; TREC2) classifying two/three rough patterns; and TREC3) generating a table summarizing categories of common configurations (trends) and the automatic icon assignments to them. The proposed trec focuses on investigating mean non-stationary long-term trends of data, and it works for any length of time steps. It is not necessary to apply a stationary Gaussian assumption to the estimated trends to investigate the common trends, which are interpreted as common variations of biological and environmental data.