Limitations of Species Distribution Models Based on Available Climate Change Data: A Case Study in the Azorean Forest

Abstract

Climate change is gaining attention as a major threat to biodiversity. It is expected to further expand the risk of plant invasion through ecosystem disturbance. Particularly, island ecosystems are under pressure, and climate change may threaten forest-dependent species. However, scientific and societal unknowns make it difficult to predict how climate change and biological invasions will affect species interactions and ecosystem processes. The purpose of this study was to identify possible limitations when making species distribution model projections based on predicted climate change. We aimed to know if climatic variables alone were good predictors of habitat suitability, ensuring reliable projections. In particular, we compared the performance of generalized linear models, generalized additive models, and a selection of machine learning techniques (BIOMOD 2) when modelling the distribution of forest species in the Azores, according to the climatic changes predicted to 2100. Some limitations seem to exist when modelling the effect of climate change on species distributions, since the best models also included topographic variables, making modelling based on climate alone less reliable, with model fit varying among modelling approaches, and random forest often providing the best results. Our results emphasize the adoption of a careful study design and algorithm selection process. The uncertainties associated with climate change effect on plant communities as a whole, including their indigenous and invasive components, highlight a pressing need for integrated modelling, monitoring, and experimental work to better realize the consequences of climate change, in order to ensure the resilience of forest ecosystems in a changing world.

Taxonomic, structural diversity and carbon stocks in a gradient of island forests

“…The structure of PF and EW was mainly influenced by the dominant species Cryptomeria japonica and Pittosporum undulatum, with the highest basal area and tree density. The structural dominance of P. undulatum in exotic woodland confirms the potential of this species to originate almost pure stands[40][41][42]69,70,[85][86][87].…”

“…Higher values of precipitation and lower values of temperature were found for NF, and an inverse situation for EW, since the remaining NF are mostly found at high elevation sites, at sloped terrain [51], while EW is generally found at low elevation and at places with higher temperatures [69][70][71][72].…”

Predicting suitable habitats of Melia azedarach L. in China using data mining

“…Akpoti et al used BRT, GLM, MAXNT and RF algorithms to predict rice production suitability and the results showed that RF has better generalizability 28 . Silva et al found the highest model quality for the RF and GAM algorithms when assessing the limitations of different species distribution models using the Azorean Forest as an example 29 . The RF is an ensemble machine-learning model that could handle data with multi-dimensional, non-linear relationships, high-order correlations, and missing values 30 .…”

Production and Health Management from Grazing to Confinement Systems of Largest Dairy Bovine Farms in Azores: A Farmers’ Perspective

“…The survey addressed dairy farmers of São Miguel Island (Azores). Located in the middle of the Atlantic Ocean, the Azores is an archipelago with 9 islands, with mild temperatures (minimum and maximum temperature: 12.0–18.4 °C, respectively), humidity (minimum and maximal relative humidity: 89.0–97.4%) [ 26 , 27 ], and an abundant rainfall climate with precipitation of 960.6 ± 201 mm per year, with 75% of the precipitation falling between October and March [ 26 , 27 ].…”