Comparative approach of multivariate statistics to decipher trade off in temporal datasets
Approche statistique multivariée pour décrire les données temporelles long terme (millénaire)
Abstract
Paleorecords are replete with examples of longterm biotic responses to past climate change, and disturbances such as fire, herbivores, insect outbreaks or ice storms. The now frequent challenges in palaeoecological studies are to better understand both the trends and the breaking points of biotic variables in response to disturbances. The aim of this article is to expand the capacity of new and existing methods to provide greater insight into ecosystem trajectories and functioning.
Our analyses focus on biomes where fire is a key structural disturbance for vegetation establishment and evolution. We aim to determine: 1)
explanative variances between biotic and abiotic variables at both short and millennial time scales, and 2) whether a temporal relationship exists between fire regimes and plant diversity change in both systems.
Our statistical approach focused on PCA (Principal Component Analysis), co-inertia, MFA (Multiple Factor Analysis) and Hill & Smith methods, which provide us new information about our data, and bring new conclusions to the former studies.
We advise to preferentially use the MFA when the environmental variables are numerous, while the co-inertia is a better tool to explain variance of the datasets.