Premium
Using climatically based random forests to downscale coarse‐grained potential natural vegetation maps in tropical Mexico
Author(s) -
Vaca Raúl A.,
Golicher Duncan J.,
Cayuela Luis
Publication year - 2011
Publication title -
applied vegetation science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.096
H-Index - 64
eISSN - 1654-109X
pISSN - 1402-2001
DOI - 10.1111/j.1654-109x.2011.01132.x
Subject(s) - downscaling , vegetation (pathology) , environmental science , climate change , scale (ratio) , random forest , climatology , vegetation classification , physical geography , geography , ecology , cartography , geology , computer science , machine learning , medicine , pathology , biology
Questions: Can the accuracy of coarse resolution potential vegetation maps be improved through downscaling to finer resolution climatic grids? Can output from random forests produce new insight into the climatic characteristics that are associated with the structural characteristics of the vegetation? Location: Southern Mexico. Methods: A potential vegetation map (National Atlas of Mexico) at a 1:4 000 000 scale, was downscaled to a 1 km 2 grid resolution using climatically based random forests models. The resulting map was then evaluated against 256 inventory plots sampled at the transition between different vegetation types in Southern Mexico. Results: Downscaling increased accuracy up to 0.40, as measured by the Kappa Index of Agreement. Multivariate analysis of the results allowed the association between Rzedowski's classification and climatic variation to be explored. This confirmed that many of the structural aspects of the vegetation that are used by the Rzedowski classification are closely associated with climate, but it also revealed weaknesses in the underlying basis of this classification system. Conclusions: Rzedowski's scheme for vegetation classification may require further modification in order to be an effective tool for research into vegetation–climate relationships.