Fig 1.
The distribution of cultivated (Bactris gasipaes var. gasipaes, triangles) and wild (var. chichagui, dots) peach palm observation points.
The points used only for modelling and gathered from sources other than this study are reported in blue and green, whereas the points referring to samples used in this study are shown in red or yellow, depending on the additional analyses carried out on each (genetic or phenotypic characterization). The red, blue and green polygons show the approximate distribution of Bactris wild types [20]; the grey lines divide the different regions of Amazonia (NWA: north-western Amazonia, GS: Guyana shield, EA: eastern Amazonia, CA: central Amazonia, SA: southern Amazonia, SWA: south-western Amazonia; after [34]).
Fig 2.
Principal component analysis of peach palm wild (var. chichagui) and domesticated (var. gasipaes) populations.
As in Fig 1, green poins are the wild form (var. chichagui) and blue triangles are the cultivated form (var. gasipaes). The two components shown here explain 94% of the overall variation in the data. The polygon represents the convex hull area constructed around all var. chichagui points, extended with a 3% buffer of the largest hull axis.
Fig 3.
Spatial distribution of allelic richness (A), locally common alleles (B), observed heterozygosity (C) and variation in standardized phenotypic diversity (D), measured as coefficient of variation (st dev/mean) in our Bactris gasipaes var. gasipaes dataset.
The blue, green and red polygons in Fig 3d indicate areas of occurrence of different peach palm landraces [20] (see discussion). Blue polygons enclose the mesocarpa landraces (20–75 gr) Rama (1), Útilis (2), Cauca (3), Pampa Hermosa (7), Tigre (8), Pastaza (9) and Inirida (10); green areas include the microcarpa landraces (< 20 gr) Tembe (4), Juruá (5) and Pará (6); and red polygons refer to the macrocarpa landraces (75–200 gr) Putumayo (including Solimões, 11) and Vaupés (12). It is important to note that the many locations for which only one accession was included in the phenotypical characterization are not included in the Fig 3d because the coefficient of variance can only be calculated for two or more individuals.
Fig 4.
Spatial distribution of allelic richness (A) and locally common alleles (B) based on the cultivated samples (~20 individuals per population) from the Bactris gasipaes var. gasipaes dataset of Hernández et al [37].
Table 1.
Metrics of model calibrations and evaluation under current environmental conditions for projections to past and future climate conditions.
Fig 5.
Putative distribution of suitable habitat of peach palm during the Last Glacial Maximum.
A: results of model calibration undertaken based on the wild form only (Bactris gasipaes var. chichagui); B: results of calibration based on both wild and cultivated (Bactris gasipaes var. gasipaes) trees pertaining to the ecological niche occupied by var. chichagui.
Fig 6.
Distribution of potential LGM refugia of peach palm (green areas) and distribution of rescaled locally common alleles (A) and rescaled allelic richness (B).
The genetic data visualized is based on a spatial combination of the results from the present study and that of Hernández et al[37]. The blue, green and red polygons indicate areas of occurrence of different peach palm landraces (see discussion) [20]. Blue polygons enclose the mesocarpa landraces (20–75 gr) Rama (1), Útilis (2), Cauca (3), Pampa Hermosa (7), Tigre (8), Pastaza (9) and Inirida (10); green areas include the microcarpa landraces (< 20 gr) Tembe (4), Juruá (5) and Pará (6); and red polygons refer to the macrocarpa landraces (75–200 gr) Putumayo (including Solimões, 11) and Vaupés (12). Several LGM suitable areas are not visible as they graphically coincide with the extent of the circular neighbourboods of the genetic data. This is the case in particular for the circular neighborhoods overlapping with the polygons describing the distribution of the Pampa Hermosa and Tigre landraces.
Fig 7.
Expected changes in habitat suitability for cultivated Bactris gasipaes var. gasipaes in the future compared with hotspots of allelic richness based on a combination of our genetic dataset with that of Hernández et al [37].
The suitable areas shown here largely correspond to the overall area for which the modelling ensemble predicts habitat suitability for at least 15 out of the 19 different future climate models considered.
Fig 8.
Priorities for on-farm conservation of areas rich in genetic and phenotypic diversity and likely to remain suitable so in the future (in red and yellow, respectively).
Dotted polygons represent areas likely to be occupied by indigenous peoples (based on language maps from the Ethnologue database [47]), suggesting sites where on-farm conservation may be already carried out by local communities or could be strengthened through their involvement.