Tecomella undulata under threat: The impact of climate change on the distribution of a valuable tree species using a machine learning model

PLOS One | https://doi.org/10.1371/journal.pone.0326609 July 9, 2025 4 / 18 Zagros, extending from northwest to southeast, enclose the vast Iranian plateau, an arid expanse encompassing rugged deserts and salt flats known as Kavir [19]. Species occurrence record We obtained 52 occurrence points (presence-only) for T. Undulata from extensive field surveys conducted in 2021–2023, literature review, and the assistance of environmental staff, environmental activists, and local people. Permits were not required to collect species presence points at certain sites. The geographical coordinates of the records were checked and refined using Google Earth, and then saved in CSV format. The accuracy of the prediction is enhanced by the spatial autocorrelation of sampling effort between the training and test data [29]. Spatial filtering is employed to mitigate the impacts of sampling biases and model over-fitting [30,31]. Therefore, spatial thinning of species occurrence was conducted using the spThin package [32] in R software. In this pro- cess, multiple presence locations in a given grid with a spatial resolution of about 1 km² were eliminated. Only one record was retained per grid. Finally, in total 44 occurrence points (Fig 1) were employed to build the model. The maps were prepared in the open source QGIS software. Environmental data In this study, 19 bioclimatic factors (Table 1) at 30-arcsec resolution (~1 km2), derived from the WorldClim website, were used which include current climate conditions (www.worldclim.org). The current climate is the average data from 1970 to

2000 from the WorldClim version 2.1 [33]. Also, the Digital Elevation Model (DEM) derived from the SRTM elevation data was downloaded from the WorldClim website. We also downloaded bioclimatic variables from WorldClim for future climate scenarios (SSP24.5 and SSP58.5) in the 2030s (the average of period 2021-2040), 2050s (the average of period 2041- 2060), 2070s (the average of period 2061-2080) and 2090s (the average of period 2081-2100), using the same resolution and the same bioclimatic layers used in the current conditions. There are four main future climate paths known as shared socio-economic pathways (SSPs), which include SSP126, SSP245, SSP370, and SSP585. These SSPs help us understand how different social and economic developments might link to climate outcomes. According to these SSPs, we could see a global temperature rise of 3–5 °C by the year 2100. Specifically, the SSP126 envisions a sustainable, green future limiting warming to 3–3.5 °C; SSP245 takes a middle- of-the-road approach, with expected warming of 3.8–4.2 °C, as it continues current social, economic, and technological trends; SSP370 anticipates a future where countries prioritize regional development, leading to a temperature increase of 3.9–4.6 °C; finally, SSP585 predicts a scenario with unchecked economic and energy use growth, causing warming of 4.7–5.1 °C by the end of the century [34]. Between General Circulation Models (GCMs), we used ACCESS-ESM1–5, BCC-CSM2-MR, IPSL-CM6A-LR, and MRI-

ESM2–0. These specific GCMs were chosen for their demonstrated ability to accurately forecast temperature and precipi- tation patterns within Iran, as evidenced by comparisons with observational data from various synoptic stations throughout the country. Previous research has utilized a range of methods for evaluating GCM performance. Recognizing the success- ful application of these GCMs in prior studies, along with their established use in the field, we selected these four models and applied them in an ensemble approach by averaging four GCMs [25,35–38]. These GCMs use the SSP126, SSP245, SSP370, and SSP585 scenarios of the shared socioeconomic pathways (SSPs) adopted by the IPCC 6th Assessment. SSPs build on existing national and regional development plans and trends to explore potential future socioeconomic land- scapes. They are upgraded versions of representative concentration pathways (RCPs) that reflect the connection between socio-economic growth trends and climate change risks. [9,39]. We selected SSP245 and SSP585 to represent a range of plausible future climate pathways, from moderate to more extreme, allowing us to explore the potential impacts under different levels of greenhouse gas emissions and socioeconomic development. These scenarios are widely used in climate change research and provide a standardized framework for comparing results across studies [40–42].