Predicting the effect of climate change on the spatiotemporal distribution of two endangered plant species, Silene leucophylla Boiss. and Silene schimperiana Boiss., using machine learning, in Saint Catherine Protected Area, Egypt

Abstract

Abstract Background Climate change significantly influences the geographical distribution of plant species worldwide, especially endemics. Endemic species are plants that live in limited distribution ranges of unique ecology and, thus, are the most vulnerable species to climate change. Therefore, understanding the impacts of climate change on the distribution of these species can assist in developing appropriate plans for their conservation. In this study, we aimed to apply various species distribution models (SDMs) to predict the current potential distributions of two endangered plant species, Silene leucophylla (S. leucophylla, endemic) and Silene schimperiana (S. schimperiana, near-endemic), in Saint Catherine protected area (St. Catherine PA), Egypt. Then, using the best-fit model to project their future distribution under the maximum climate emission scenario (Representative Concentration Pathway 8.5 (RCP8.5)). Six different SDMs were constructed using different geospatial raster imagery sets of environmental factors. For each model, five machine learning (ML) algorithms were used. The results of these ML algorithms were then ensembled by calculating the weighted average of their predictions. Results Based on the analysis of digital geospatial imageries produced by the best-fitting model, the predicted suitable areas of S. leucophylla and S. schimperiana were 23.1 km2 and 125 km2, respectively. These sites are located mainly in the high-elevation middle northern part of the study area. Annual precipitation, mean temperature of the driest quarter, altitude, and precipitation seasonality were the essential predictors of the distributions of both species. Future predictions of both species indicated opposing results between the studied species. Predictions in the 2050 and 2070 future conditions revealed significant range contraction for the distribution of S. leucophylla. For S. schimperiana, a range shift is predicted, with both range contraction and range expansion of its current suitable habitats, for the same future projections. Unfortunately, in 2080 predictions, both species could be projected to a complete loss from the entire area. Conclusion This study highlights the importance of including diverse types of environmental variables in SDMs to produce more accurate predictions, rather than relying only on one variable type. It also revealed the potential negative impacts of future climate change on the distributions of two endangered plant species, S. leucophylla and S. schimperiana, inhabiting St. Catherine PA. Consequently, we urgently recommend the initiation of different plans and strategies seeking their conservation.

Keywords

• Environmental variables
• SDMs
• Ensemble models
• Distribution change detection
• Range contraction
• Range expansion