“…More recently, machine learning models have also been used [6][7][8] as well as regression kriging [9]. Rudiyanto et al [6] (2016) and Young et al [9] (2018) built models using a relatively limited amount of environmental data (i.e., elevation, slope, aspect, System of Automated Geoscientific Analyses Wetness Index (SAGAWI) and nearest distance to river for the first study, and elevation, slope, aspect, vegetation type and soil for the latter). Rudiyanto et al [8] (2018) and Aitkenhead [7] (2017) used a wide array of environmental covariates: topography (i.e., elevation, vegetation-corrected elevation, and two derived terrain attributes-the Multi-Resolution Index of Valley Bottom Flatness (MRVBF) and SAGAWI-Euclidean distances to rivers, seas and combined rivers and seas, radar images (i.e., Sentinal-1A and ALOS-PALSAR) and vegetation (i.e., seven Landsat raw bands and the normalized difference vegetation index) in Rudiyanto et al [8] (2018), and topography (i.e., elevation and seven derived terrain attributes), climate (i.e., 24 different meteorological layers), soil (i.e., land cover, geology and soil maps) and vegetation (i.e., Landsat raw bands and derived vegetation indices) in Aitkenhead [7] (2017).…”