Analysis Of Uncertainty Factors For Species Distribution Modeling Of Invasive Alien Plants

With the development of global trade and rapid climate change,the risk of plant invasion increases obviously.Plant invasion has a large potential to damage the stability of invaded ecosystem and biodiversity.Hence,it is urgent to prevent and control plant invasion worldwide.Species distribution modelings(SDMs)can predict the distribution pattern of plant species based on the occurrence records and environmental variables.SDMs have been widely used in the evaluation of plant invasion risk.However,the uncertainty on the distribution models of invasive alien plants may decrease the robustness and effectiveness of SDMs.Here,we analyzed the uncertainty factors on the species distribution modeling of invasive alien plants,and put forward the effective suggestions to improve the performance on the distribution modeling of invasive alien plants.We conducted the analysis of uncertainty factors for species distribution modeling of invasive alien plants based on four aspects:1)the ecological niche shifts of invasive plants;2)the transferability of the distribution modeling of invasive plants;3)the reasonable sets for the distribution modeling of invasive plants;4)the distribution modeling of invasive plants under future climate change.Specially,1)We explored the potential factors affecting the performance on the distribution modeling of invasive plants.2)We examined the ecological niche shifts of invasive plants including terrestrial and aquatic species between native and invasive ranges,and explored the impacts of such shifts on the transferability of the distribution modeling of invasive plants.3)We explored the impacts of spatial scales,environmental variables,and model complexity on the the distribution modeling of invasive plants.4)We used spatial analysis to explore the impacts of future climate change on the distribution pattern of invasive plants,and determine the risk hotspots of plant invasion.5)When finishing the analysis of uncertainty factors on the distribution modeling of invasive plants,the effective and feasible suggestions were proposed.The main conclusions and suggestions were as follows:1)The number of occurrence records and environmental variables,and spatial scales were important for the distribution modeling of invasive plants.The response of these model sets to AUC was a single peak curve.The presence-only SDMs may have the good performance to model the distribution pattern of invasive plants on the global scale.2)The ecological niche shifts of terrestrial and aquatic invasive plants between native and invasive ranges could exist widely,and such shifts would affect the transferability of SDMs.Hence,we suggested to model the distributions of invasive plants based on both occurrence records of the native and invasive ranges at global scales.3)With the increasing spatial scales of climate data the probabilities of invasive plants could increase.Hence,the large spatial scales of climate data may over-estimate the SDM results.Our suggestion was to use 5.0 arc-minutes to model the distribution pattern of invasive plants.4)The selection of model complexity was the process of balance between performance and transferability of SDMs.The increasing model complexity could enhance the performance of SDMs,and weaken the model transferability.We suggested to integrate the occurrence records of native and invasive ranges into the distribution modeling of invasive plants,and to use the regularization multiplier values of two through four as the sets of MaxEnt modeling.5)Although the climatic factors are the primary drivers of distributions of invasive plants,non-climatic factors including soil variables and human footprint could be extremely important for the distribution modeling of invasive plants.Soil variables and human footprint may increase the probabilities of invasive plants in some regions,particularly,Montane Grasslands and Shrublands.Furthermore,soil variables and human footprint may increase the performance of distribution modeling of invasive plants compared with climatic variables only.We suggested to integrate climate,soil and human footprint variables into SDMs for modeling the distributions of invasive plants.6)Climate change could not increase the expansion areas of invasive plants,but the probabilities of expansion risk hotspots would increase.The risk hotspots of invasive plants were mainly distributed in Europe,southwestern Australia,New Zealand and Madagascar.Invasive plants,particularly herbaceous and woody ones,might still expand into the ecoregions of these risk hotspots in the high concentration scenarios.It might be still a high risk of plant invasion in critical or endangered ecoregions under climate change.