Predictive distribution for Salvia aethiopis (Lamiaceae) in Middle Asian Region based on climatic modelling

Abstract

Geospatial investigation of distribution Salvia aethiopis L. (Lamiaceae) on the eastern limits of the range is performed using climatic modeling by MAXENT approach. Climatic conditions for the 33 local populations of the species as well as territories neighbouring to them were examined in detail, using concepts of nestle-cells and a small polygone. According to obtained results, the most suitable climatic conditions are in only two small polygons (SP) around local populations with coordinates 69.96E:42.48N and 70.03E:42.47N in central part of the area. They have 7.7 average points per a cell of 9.0 maximum. Other seven SP have average points per a cell 7.1–7.3 (quite suitable climatic conditions), all in the western part of the area, with coordinates from 66.69E to 67.69E and from 38.21N to 39.26N. High values of raw prediction are in the locations with coordinates 70.97E:42.48N (0.123; Zhualyn District in Zhambyl Region, Kazakhstan), 70.03E:42.47N (0.063; Keltemashat, Tulkibas District, Turkestan Region, Kazakhstan), 67.09E:38.34 (0.063; Yuqori Machay, Surkhandarya Region, Uzbekistan), and 69.96E:42.48N (0.054; Antonovka, Tulkibas District, Turkestan Region, Kazakhstan). These local populations of S. aethiopis should be assessed as the most likely secondary centers for further expansion of the species in the Middle Asian Region.

Comparison of different models in predicting habitat suitability of rare species in Uzbekistan: 8 rare Tulipa species case-study

“…Using SDMs for studying rare plant species in Uzbekistan is in the initial stage now. Only a few studies were carried out (Baikov et al 2021; Khujanov, 2021; Mavlanov et al 2021) and most of them were conducted using MaxEnt (Philips et al 2006), which is considered the most powerful machine learning algorithm with an uncomplicated graphical user interface, which made MaxEnt the most comfortable and popular technique for researchers in last years (Philips et al 2017). However, numerous studies demonstrated that Random Forest and other machine learning techniques in some cases are more efficient than MaxEnt (Williams et al 2009; Duan et al 2014; Mi et al 2017).…”