Coupling ordination techniques and GAM to spatially predict vegetation assemblages along a climatic gradient in an ENSO‐affected region of extremely high climate variability

Abstract

AbstractQuestionsEl Niño Southern Oscillation (ENSO) is a strong driver of climatic and ecosystem variability in coastal NW Peru. La Niña amplifies the already dry local conditions, and led to depleted ecosystems in 2011. However, the 2012 La Niña event triggered rainfall far above the average. (1) Did plant species diversity, primary productivity and vegetation assemblages change along a climatic gradient between two climatologically different La Niña years; (2) Is there a difference in explanatory power of environmental predictors between the 2 yr; and (iii) is it possible to predict the observed vegetation patterns spatially?LocationTransect along a climatic gradient in the Sechura Desert of Piura, NW Peru (corresponds to the terrestrial part of the El Niño region 1 + 2) – a region of extremely high climatic variability.MethodsWe visited 50 30 m × 30 m randomly sampled plots in 2011 and 2012. A Procrustes analysis of two non‐metric multidimensional scaling (NMDS) ordinations provided information on the temporal change of species assemblages. Variation partitioning revealed the differences in explanatory power of the predictors. We employed a generalized additive model (GAM) to fit the scores of the first ordination axis with a floristic gradient map as a result.ResultsGenerally, higher rainfall resulted in a positive feedback when considering biodiversity, productivity and vegetation assemblages. The floristic gradient map resulting from the GAM displayed the spatial distribution of the three main assemblages along the climatic gradient. Edaphic variables added no independent portion to the explanation of the vegetation assemblages, but explained in conjunction with topography and NDVI a considerable amount of the variance.ConclusionsStrong Atlantic easterly winds crossing the Andes can boost plant growth even during a La Niña situation. This underscores the need for a deeper understanding of ENSO‐related climate variability of ENSO. Combining vegetation maps with accurate predictions of such climatic anomalies would aid the effective execution of conservation and recovery strategies. Additionally, coupling an unconstrained ordination with a GAM appears to be a promising tool for vegetation mapping, especially in the presence of a non‐linear gradient.