UMR 1224 Ecologie Comportementale et Biologie des Populations de Poissons
64310 Saint Pée sur Nivelle
What is Tampo ?
Tampo is a R package that facilitates analyzing microchemical profiles (both mono- and multielemental composition) from fish otoliths using a recursive partitioning approach (zoning algorithm) that can accommodate some form of user-specified constraints. It also allows extracting environmental histories from otolith based on typical elemental/isotopic sequences generated using methods such as LA-ICPMS. Microchemical analysis is widely used in fisheries management and fisheries biology to identify stocks and characterize fish movements but the provided functions may be considered from the more general perspective of the chronological clustering of multivariate time series using piecewise constant regressions.
The current version of the package is mostly dedicated to applying zoning algorithm on (multivariate) time series based on a recursive partitioning approach. Specific functions are also provided to get insights into the multivariate partitioning of the data.
Tampo was recently archived on CRAN due to its dependency to another package ('mvpart') which is no longer maintained. For Windows machines, one may try to automatically install the two packages using the "URLinstall.R" script. Alternatively, if it don't work, one can download the two files from the links mentioned in the script into your current directory and install them manually using "Packages -> Install package(s) from local zip files" from menu bar. In this case, you just need to install 'mvpart' first and 'Tampo' second. For other OS machines, source tarballs are available at :
– mvpart: http://cran.r-project.org/src/contrib/Archive/mvpart/
– Tampo: http://cran.r-project.org/src/contrib/Archive/Tampo/
The latest version should be found at the CRAN web site
Tampo and its supporting document are also available from this web site:
- Tampo Source package (curent version: 1.0) (tar.gz)
- Tampo Windows binaries (zip)
- Tampo Reference manual (pdf)
- Tutorial on how to use recursive partitioning (pdf, 43 pages)
- Script for the above tutorial (.R file)
- Brief tutorial on how to extract habitat use (coming soon)
- Mvpart Windows binaries (last version: 1.6-2) (zip)
- Mvpart source package (last version: 1.6-2, may not compile on Windows) (tar.gz)
Basic usage and examples
While the analysis of microchemical profiles from otoliths is of primary importance in fishery management and population ecology, the traditional approaches mostly fail to capture the multivariate nature of structural change in time series data. Tampo proposes a simple approach aiming at coping with the multi-elemental compositional otolith transect in a multi-scale quantitative manner. More specifically, the proposed recursive partitioning method try to bias the chronological clustering process by accommodating some form of user-specified constraints based on both intra- and inter-groups characteristics. This allows researchers to focus on biologically relevant information (i.e. the one researcher look for and can interpret). Chronological clustering based on recursive partition is easily interpretable and directly applicable to detection of hidden discontinuities in any multivariate time series. In a multi-proxy context, it also provides a powerful exploratory tool for assessing the relative importance of cross-correlated variables in structuring the environmental histories, as well as the congruency between variables.
The partition process can perform i) using an arbitrary number of clusters, ii) automatically using an optimization criteria based on cross-validation error, iii) in a constrained manner that accommodate specific user-defined criteria.
Fig 2. Univariate regression trees with different user-specified constraints
Fig.3 Algorithm applied to a multivariate profile
Fig.4 Profiles ordination for a multielemental isotopic signal
Fig.5 Temporal congruency among elements/isotopes