Using Kullback information of species composition for classification, monitoring, and modelling of forest vegetation with changing environments
Martin Jenssen
Last modified: 2008-09-13
Abstract
Assessing the impact of global change to forests requires dynamic vegetation modelling. The application of coupled species-based models, however, is seriously limited by the complexity of the vegetation system and by the large data amount required for model parameterization. More empirical approaches are based on statistical relationships between observed vegetation patterns and environmental conditions using a space for time substitution. However, new constellations of environmental conditions may lead to novel plant communities that had never been observed before.
The solution approach presented here is a linkage of a vegetation-type with an individual-species based approach. It starts from the previous finding* that structural information about species composition (calculated as Shannon information from species abundances) tends to a saturation value (saturation diversity) with increasing number of samples subject to similar ecological (edaphic, climatic, bio-geographic) and management constraints (e.g. tree species set by man). The species distributions corresponding to saturation diversities represent the vegetation potentials of the sites with given constraints. Kullback information measures the 'distance' between individual plant assemblages (relev?s) and the vegetation potentials of the sites. It is used to derive a numerical vegetation classification of North-Central European forests.
For more than 500 plant species, probability density functions (pdf) over edaphic and bioclimatic variables are derived from a comprehensive ecological data base. Indicator value models based upon these pdf's are used to construct a more-dimensional ecological state space from the Kullback information distances between adjacent vegetation types.
Time series from long-term vegetation monitoring as well as vegetation experiments with controlled ecological conditions are visualized and interpreted within the ecological state space. Finally, first results of predictive vegetation modeling within the more-dimensional ecological state space are presented.
*JENSSEN, M. (2007): Ecological potentials of biodiversity modelled from information entropies: Plant species diversity of North-Central European forests as an example. Ecological Informatics 2/4, 328-336. Online via ScienceDirect: http://dx.doi.org/10.1016/j.ecoinf.2007.06.003
The solution approach presented here is a linkage of a vegetation-type with an individual-species based approach. It starts from the previous finding* that structural information about species composition (calculated as Shannon information from species abundances) tends to a saturation value (saturation diversity) with increasing number of samples subject to similar ecological (edaphic, climatic, bio-geographic) and management constraints (e.g. tree species set by man). The species distributions corresponding to saturation diversities represent the vegetation potentials of the sites with given constraints. Kullback information measures the 'distance' between individual plant assemblages (relev?s) and the vegetation potentials of the sites. It is used to derive a numerical vegetation classification of North-Central European forests.
For more than 500 plant species, probability density functions (pdf) over edaphic and bioclimatic variables are derived from a comprehensive ecological data base. Indicator value models based upon these pdf's are used to construct a more-dimensional ecological state space from the Kullback information distances between adjacent vegetation types.
Time series from long-term vegetation monitoring as well as vegetation experiments with controlled ecological conditions are visualized and interpreted within the ecological state space. Finally, first results of predictive vegetation modeling within the more-dimensional ecological state space are presented.
*JENSSEN, M. (2007): Ecological potentials of biodiversity modelled from information entropies: Plant species diversity of North-Central European forests as an example. Ecological Informatics 2/4, 328-336. Online via ScienceDirect: http://dx.doi.org/10.1016/j.ecoinf.2007.06.003