1 This is the pre-print (non-reviewed) version of the article. The published version can be found here: 2 https://doi.org/10.1016/j.foreco.2021.119127 3 Dynamic seed zones to guide climate-smart seed sourcing for tropical dry 4 forest restoration in Colombia 5 Tobias Fremouta,b*, Evert Thomasb, Kelly Tatiana Bocanegra-Gonzálezc, Carolina Adriana Aguirre- 6 Moralesd, Anjuly Tatiana Morillo-Paze, Rachel Atkinsonb, Chris Kettlef,g, Roy González-M.h, Carolina 7 Alcázar-Caicedoh, Mailyn Adriana Gonzálezh, Carlos Gil-Tobóni, Janneth Patricia Gutiérrezj, Luis 8 Gonzalo Moscoso-Higuitak, Luis Augusto Becerra López-Lavallej, Dulcinéia de Carvalhol, Bart Muysa 9 a Division of Forest, Nature and Landscape, Department of Earth and Environmental Sciences, KU 10 Leuven, Celestijnenlaan 200E, 3001 Leuven, Belgium 11 b Bioversity International, Av. La Molina 1895, Apartado Postal 1558, Peru 12 c Molecular Plant Sciences, University of Edinburgh, Old College, South Bridge, Edinburgh EH8 13 9YL, United Kingdom 14 d Facultad de Ciencias Agropecuarias, Universidad Nacional de Colombia, Cra. 32 ##12-00, Palmira, 15 Valle del Cauca, Colombia 16 e Facultad de Medio Ambiente y Recursos Naturales, Universidad Distrital Francisco José de Caldas, 17 Cra. 7 ##40b-53, Bogotá, Colombia 18 f Bioversity International, Viale Tre Denari 472, 00054 Maccarese-Stazione RM, Italy 19 g Department of Environmental System Science, ETH Zürich, Rämistrasse 101, 8092 Zürich, 20 Switzerland 21 h Instituto de Investigación de Recursos Biológicos Alexander von Humboldt, Cl. 28a #15-09, Bogotá, 22 Colombia 23 i Universidad de Antioquia, Cl. 67 ##53-108, Medellín, Antioquia, Colombia 24 j CIAT (International Centre of Tropical Agriculture), Recta Cali-Palmira km 17, Valle del Cauca, 25 Colombia 26 k Forestpa SAS, Cra. 38 #9 A 27 Oficina 203, Medellín, Antioquia, Colombia 27 l Laboratório de Conservação Genética de Espécies Florestais, Universidade Federal de Lavras, Aquenta 28 Sol, Lavras - State of Minas Gerais, 37200-900, Brazil 29 * corresponding author; tobias.fremout@gmail.com; Division of Forest, Nature and Landscape, 30 Department of Earth and Environmental Sciences, Celestijnenlaan 200E, 3001 Leuven, Belgium 31 e-mail addresses: Tobias Fremout (tobias.fremout@kuleuven.be); Evert Thomas: e.thomas@cgiar.org; 32 Kelly Tatiana Bocanegra-González: ktbocanegrag@gmail.com; Carolina Adriana Aguirre-Morales: 33 acaguirrem@gmail.com; Anjuly Tatiana Morillo-Paz: anjulymorillo@gmail.com; Rachel Atkinson 34 (r.atkinson@cgiar.org); Chris Kettle (c.kettle@cgiar.org); Roy González-M. 35 (rgonzalez@humboldt.org.co); Carolina Alcázar-Caicedo (alcazarcaicedo@gmail.com); Mailyn 36 Adriana González (gonzalez.mailyn@gmail.com); Carlos Gil-Tobón (carlosgiltobon@gmail.com); 37 Janneth Patricia Gutiérrez (j.gutierrez@cgiar.org); Luis Gonzalo Moscoso-Higuita 38 (moscosoluisgonzalo@yahoo.es); Luis Augusto Becerra López-Lavalle (l.a.becerra@cgiar.org); 39 Dulcinéia de Carvalho (dul.car@hotmail.com); Bart Muys (bart.muys@kuleuven.be) 40 Abstract 41 Tree-based forest landscape restoration interventions require knowledge on the suitability and origin of 42 seed sources and planting material. A common recommendation is to select locally sourced material 43 based on the assumption that it is well adapted to local environmental conditions and to avoid 44 introduction of maladapted genes. However, faced with accelerating climate change, it may be prudent 45 to supplement local provenances with ‘climate-matched’ provenances, i.e. where current climate 46 conditions are similar to those anticipated in the future at the planting site. Restoration practitioners 47 usually do not have access to the necessary information to implement such climate-smart seed sourcing. 48 Here, we combine genetic data of 11 socio-economically important tree species of the tropical dry 49 forests of Colombia with spatial environmental data to inform the delineation of dynamic seed zones 50 for the restoration of this highly threatened ecosystem. Analysis of Molecular Variance (AMOVA) 51 indicates significant population genetic differentiation within all 11 species. We fitted linear mixed 52 effects models to evaluate if the genetic distance between trees was mainly related to geographic 53 distance (i.e. isolation by distance; IBD), environmental distance (i.e. isolation by environment; IBE), 54 or both. Observed scales of genetic differentiation were best explained by the model including both 55 geographic and environmental distance (IBD + IBE) for 6 out of 11 species, and by the IBE model for 56 the remaining species, suggesting that the observed differentiation is at least partly driven by adaptive 57 processes. Aiming at capturing as much as possible of the observed genetic differentiation, we propose 58 a set of 36 provisional seed zones that are applicable across species and dynamic under climate change, 59 based on the clustering of environmental data and geographical coordinates. We project these seed 60 zones to future climate conditions using five general circulation models and two emission scenarios, 61 and discuss how they can be used to implement different climate-smart seed sourcing strategies in a 62 pragmatic way. The seed zone maps are made available in a user-friendly online tool. 63 1 Introduction 64 In times of unprecedented anthropogenic pressure on ecosystems worldwide, the restoration of 65 degraded lands has become a global priority (Aronson and Alexander, 2013; FAO, 2020; Suding et al., 66 2015). Restoration efforts often include tree planting, in which case decisions need to be made on the 67 species to be planted, but also on the provenance of the planting material (also called forest reproductive 68 material; typically seeds but may also consist of cuttings, stakes, wildlings, etc.. we use ‘seeds’ here for 69 simplicity). Many provenance trials, also called common garden trials, have found evidence of adaption 70 of tree species populations to local environmental conditions, although most have been carried out in 71 temperate regions (e.g., Isaac-Renton et al., 2018; Kreyling et al., 2014; St. Clair, 2006; Vitasse et al., 72 2009) and knowledge on adaptive variation in tropical tree species remains scarce (Alberto et al., 2013, 73 but see for example Barton et al., 2020). As a result of these findings, it is frequently recommended to 74 source seeds or seedling planting material for restoration locally to avoid maladaptation, while also 75 reducing the risk of outbreeding depression or erosion of intraspecific genetic diversity (Bischoff et al., 76 2010; Hufford and Mazer, 2003; Krauss and He, 2006; McKay et al., 2005; Montalvo and Ellstrand, 77 2001; Vander Mijnsbrugge et al., 2010). 78 The importance of using local seed for tree planting activities remains a crucial principle to guide seed 79 sourcing decisions (Pedrini and Dixon, 2020). However, implementing a seed sourcing strategy purely 80 focussed on local sources may be problematic for a number of reasons. Populations closest to a planting 81 site may be small and fragmented, resulting in inbred seeds of low genetic diversity (Aguilar et al., 82 2006; Breed et al., 2012; Vranckx et al., 2012), making it difficult or impossible to capture sufficient 83 genetic diversity to establish self-sustaining populations (Broadhurst et al., 2008). In addition, several 84 studies have found evidence that the geographic distance between individuals or populations is not 85 always the best indicator of their genetic dissimilarity, which is sometimes more correlated to 86 environmental distance (i.e. isolation by environment; IBE) than to geographic distance (i.e. isolation 87 by distance; IDB) (Montalvo and Ellstrand, 2001; Sexton et al., 2014). Furthermore, the magnitude and 88 rate of climate change is increasingly raising concerns about the ability of local populations to adapt to 89 future climate conditions (Aitken et al., 2008; Hancock and Hughes, 2014; Vitt et al., 2010) and has 90 sparked an intense debate on what constitutes an appropriate seed sourcing strategy. 91 Several alternative ‘climate-smart’ or ‘climate-proof’ seed sourcing approaches have been proposed, 92 often involving the collection of at least a part of the seeds in areas where current climate conditions 93 are most similar to those anticipated at the planting site under climate change, i.e. ‘climate matching’ 94 or ‘predictive provenancing’ (e.g., Booth, 2016; Broadmeadow et al., 2005; Crowe and Parker, 2008; 95 Gray and Hamann, 2011; Harrison et al., 2017; Shryock et al., 2018; Thomson et al., 2010). Other 96 alternative approaches put more focus on increasing genetic diversity and thereby adaptive potential. 97 These include the ‘composite provenancing’ approach, in which seeds from several local provenances 98 are mixed with progressively smaller amounts of seeds from more distant provenances (Broadhurst et 99 al., 2008), the ‘admixture provenancing’ approach, in which seeds from different populations are mixed 100 without considering the geographic or environmental distance to the planting site (Breed et al., 2013), 101 the ‘regional admixture’ approach, in which seeds are sourced from different populations within a 102 biogeographic region (or seed zone) (Bucharova et al., 2018), and the ‘climate-adjusted provenancing’ 103 approach, in which seeds are collected in several localities along a gradient coinciding with the direction 104 of predicted climatic changes (Prober et al., 2015). Although the debate continues, several authors have 105 provided guidance on how to select the most appropriate strategy depending on species characteristics, 106 the expected impacts of climate change and the local context (Breed et al., 2013; Havens et al., 2015; 107 Ramalho et al., 2017; Sgrò et al., 2011). 108 Most of previously proposed climate-smart seed sourcing strategies developed based on temperate 109 species are not easily implemented in a tropical context, given the large local species pools, scarcity of 110 data on scales of adaptive variation, generally limited logistic capacity of restoration practitioners and 111 lack of government incentives to obtain appropriate planting material (Atkinson et al., 2018; Jalonen et 112 al., 2018). To support climate-smart seed sourcing for a tropical context, we propose a more pragmatic 113 approach using dynamic seed zones to guide seed sourcing decisions. Seed zones (also called seed 114 transfer zones, seed provenance zones, or breeding zones) are geographic areas in which planting 115 material can be moved freely while minimizing the risk for a loss of population fitness and the disruption 116 of population genetic patterns (Hufford and Mazer, 2003; Miller et al., 2011). Ideally, they should be 117 informed by insights on intraspecific adaptation patterns (Hufford and Mazer, 2003), traditionally 118 obtained using provenance trials (Hamann et al., 2000; Kramer et al., 2015; Miller et al., 2011). While 119 provenance trials remain crucial to inform seed sourcing decisions (Brancalion et al., 2015; Breed et 120 al., 2018), they do not exist for most species used in restoration, especially in the tropics, prompting 121 researchers to delineate provisional seed zones using proxies for local adaptation. 122 Based on the notion that seed zones are assumed to reflect genetic differentiation between populations, 123 several studies have used genetic markers to inform seed sourcing decisions, including both adaptive 124 markers (e.g., Hufford et al. 2016, Shryock et al. 2017), which are directly under natural selection, and 125 neutral markers (e.g., Jorgensen et al. 2016, Durka et al. 2017, Listl et al. 2018). While neutral markers 126 are not directly linked to adaptation, meta-analyses have found that population differentiation in neutral 127 markers tends to be correlated with differentiation in adaptive markers (Leinonen et al., 2008; Merilä 128 and Crnokrak, 2001). Other proxies purely based on environmental characterization data have also been 129 used, such as ecoregions (e.g., Kramer et al., 2015; Miller et al., 2011), selected climate variables (e.g., 130 Bower et al. 2014, Castellanos-Acuña et al. 2018), or a wide range of environmental variables (e.g., 131 Potter and Hargrove 2012, Crow et al. 2018). These have the advantage that they do not rely on detailed 132 species-specific data, but it remains uncertain how representative environmentally delimited seed zones 133 are for mapping genetic and adaptive variability across a species range. In addition, they are not likely 134 to accurately reflect differences in biotic interactions, which may also drive local adaptation, especially 135 in the tropics (Hargreaves et al., 2020). Nonetheless, one great advantage of environmentally-based 136 delimitation of seed zones is that it permits to predict how zones are expected to shift under climate 137 change, i.e. ‘dynamic’ seed zones (Kramer and Havens, 2009; Vitt et al., 2010). Hence, the combination 138 of environmentally-based limitation of seed zones and species-specific estimates of population genetic 139 differentiation based on genetic marker data seems a promising approach to develop such dynamic seed 140 zones. 141 Tropical dry forests (TDFs), also called seasonally dry tropical forests (SDTFs), are among the most 142 threatened of ecosystems worldwide (Fremout et al., 2020; Hoekstra et al., 2005; Janzen, 1988). In the 143 Americas, only around one-third of the original TDF cover remains (Portillo-Quintero and Sánchez- 144 Azofeifa, 2010). Despite this, TDFs receive less scientific attention than more humid tropical forests 145 and there is a pressing need to step up science-based restoration and conservation efforts in this 146 ecosystem (Pennington et al., 2018; Schröder et al., 2021). The situation is especially critical in 147 Colombia, where only around 8% of the original TDF cover remains (García et al., 2014), making it a 148 national priority for restoration and conservation (Norden et al., 2020; Vargas and Ramírez, 2014). 149 In this study, we used neutral genetic marker data of 11 socio-economically important tree species of 150 the TDFs of Colombia combined with spatial environmental data to inform the delineation of dynamic 151 seed zones for guiding climate-smart provenance decisions for the restoration of this highly threatened 152 ecosystem. First, we use these markers to evaluate if tree species populations are genetically 153 differentiated, and, whether genetic differences between trees are driven by isolation by distance (IBD), 154 isolation by environment (IBE), or both. Next, we propose a set of seed zones based on the clustering 155 of environmental data and geographical coordinates, and project these seed zones to future climate 156 conditions using different climate models and emission scenarios. Finally, we discuss how these seed 157 zones can be used to implement several previously proposed climate-smart seed sourcing strategies in 158 a pragmatic way. 159 2 Methods 160 2.1 Study region 161 The study region comprises the potential distribution of TDF in Colombia, i.e. all areas that would be 162 covered by TDF in the absence of human disturbance, as delimited by the Instituto Alexander von 163 Humboldt (Rodríguez-Buritacá et al., 2016), building on the work of Etter et al. (2008). The Colombian 164 TDFs are located mainly along the Caribbean coast and the Inter-Andean valleys of the Cauca, 165 Magdalena, Chicamocha and Patía rivers. Some authors also include a number of deciduous or semi- 166 deciduous forest types embedded in a mosaic of savannah and grassland of the Los Llanos region in the 167 Orinoquia department of north-eastern Colombia, which show similarity with other Colombian TDFs 168 due to drought stress caused by topographic and soil conditions (e.g., rocky outcrops and calcareous 169 soils) (Pizano et al., 2016), but these forests are not included here. The Caribbean and Inter-Andean 170 TDFs can be considered as floristically distinct from each other (Banda et al., 2016; González-M et al., 171 2018). 172 To avoid any possible omission errors, we slightly extended the potential TDF distribution to include 173 all areas falling within the climatic definition of TDF, for which we considered a maximum annual 174 precipitation of 1600 mm, at least 5 months with less than 100 mm precipitation, and a mean annual 175 temperature of at least 17°C (Murphy and Lugo, 1986; Pennington et al., 2000). Average precipitation 176 and temperature data were obtained from the WorldClim database at a resolution of 30 arcsec (ca. 0.9 177 km at the equator) (Hijmans et al., 2005). We did not set a lower limit for precipitation, which resulted 178 in the inclusion of the entire Guajira peninsula, the driest part of the Colombian TDFs, which in reality 179 consists of a mosaic of TDF, shrublands, and desert. The reason for this is that populations from the 180 Guajira peninsula may be useful for climate-smart seed sourcing (see below) to restore areas that are 181 expected to become drier in the future. 182 2.2 Study species 183 The study species consist of 11 socio-economically important tree species native to the TDFs of 184 Colombia, comprising 5 species mostly known for their high-quality timber (Astronium graveolens, 185 Aspidosperma polyneuron, Caesalpinia ebano, Cedrela odorata, Platymiscium pinnatum) and 6 186 multipurpose tree species (Albizia saman, Bursera simaruba, Ceiba pentandra, Enterolobium 187 cyclocarpum, Hura crepitans, and Hymenaea courbaril). Information on their life history traits 188 (dispersal, pollination, sexual system, and mating system) is given in Table 1. All 11 species are diploid. 189 Table 1: Study species with their life history traits. Note that ‘monoecious’ is used in the strict sense here, only 190 referring to species with separate male and female flowers on the same plant. Species Dispersal Pollination Sexual system Mating system Albizia saman (Jacq.) F. Terrestrial Outcrossing (self- Bees and moths Hermaphrodite Muell. (Fabaceae) mammals* incompatible) 1 Aspidosperma Wind (samara Moths and Mostly outcrossing polyneuron Müll.Arg. Hermaphrodite fruit) butterflies 2 (Apocynaceae) Astronium graveolens Wind (samara- Outcrossing Small bees Dioecious Jacq. (Anacardiaceae) like fruit) (dioecious) 3 Bursera simaruba (L.) Bees, wasps and Outcrossing Birds, bats Dioecious Sarg. (Burseraceae) wind (dioecious) 1 Caesalpinia ebano Terrestrial Insects Hermaphrodite Unknown 4 H.Karst (Fabaceae) mammals* Cedrela odorata L. Wind (samara Mostly outcrossing Bees and moths Monoecious (Meliaceae) fruit) 5 Ceiba pentandra (L.) Wind (cottony Bats Hermaphrodite Mixed 6 Gaertn. (Malvaceae) seeds) Enterolobium Terrestrial Outcrossing (self- cyclocarpum (Jacq.) Bees and moths Hermaphrodite mammals* incompatible) 1 Griseb. (Fabaceae) Hura crepitans L. Mostly outcrossing Autochory Bats Monoecious (Euphorbiaceae) 6 Hymenaea courbaril L. Terrestrial Outcrossing (self- Bats Hermaphrodite (Fabaceae) mammals* incompatible) 1 Platymiscium pinnatum Wind (samara Bees Hermaphrodite Unknown (Fabaceae) fruit) 191 * A. saman, E. cyclocarpum, and H. courbaril were primarily dispersed by now extinct megafauna (Janzen and 192 Martin, 1982), the same is possibly true for C. ebano, which fruits are also hard pods. H. courbaril and E. 193 cyclocarpum were part of the diet of Precolumbian societies (Iriarte et al., 2020; Zizumbo-Villarreal et al., 2016) 194 and may have undergone considerable human dispersal as a consequence, the same is possibly true for A. saman 195 which also has edible fruits. Similarly, A. saman and E. cyclocarpum have been used traditionally as pasture trees 196 for centuries (Aguirre-Morales et al., 2020; Thomas et al., 2016). 197 1 (Bawa, 1974) 198 2 (Chaves et al., 2016) 199 3 (Sanchez-Gomez et al., 2020) 200 4 No info on C. ebano was found. Bullock (1985) report mating systems of several Caesalpinia species in Mexico, 201 but these include entirely outcrossing, mostly outcrossing and mixed reproduction (both selfing and outcrossing). 202 5 (James et al., 1998) 203 6 (Murawski and Hamrick, 1992) 204 2.3 Field sampling 205 Field sampling was carried out across the TDFs of Colombia in a total of 18 localities (Figure 1; Table 206 S1), covering all Colombian TDF regions except the TDFs of the Orinoquia department. Within each 207 locality, adult trees were sampled within a circular area with radius of ca. 10 km, with selected trees 208 standing at least 50 meters apart to avoid sampling closely related individuals. The number of sampled 209 populations per species varied between 4 (C. ebano) and 12 (C. pentandra). A total of 725 trees was 210 sampled, with the number of sampled trees per species varying between 37 (H. courbaril) and 96 (A. 211 saman, A. graveolens, C. pentandra) (Table 2). All biological materials were collected in collaboration 212 with the Instituto Alexander von Humboldt following the Colombian Decreto 302 of 2003. 213 Table 2: Number of sampled trees per population for the tree species included in this study. The species codes 214 refer to the following species: Asa: Albizia saman; Apo: Aspidosperma polyneuron; Agr: Astronium graveolens; 215 Bsi: Bursera simaruba, Ceb: Caesalpinia ebano, Cod: Cedrela odorata, Cpe: Ceiba pentandra; Ecy: 216 Enterolobium cyclocarpum; Hcr: Hura crepitans; Hco: Hymenaea courbaril, Ppi: Platymiscium pinnatum. The 217 central coordinates and average climatic conditions of the populations are given in Table A.1 (Supplementary 218 material). Species Region Population Asa Apo Agr Bsi Ceb Cod Cpe Ecy Hcr Hco Ppi Caribbean Santa Marta N 3 12 9 9 4 5 9 Coast Santa Marta S 7 1 9 10 12 4 9 7 El Guamo 10 13 14 10 10 10 Zambrano 14 9 10 8 11 11 Colosó 9 9 11 10 8 3 15 3 Cauca Ituango 10 9 9 10 10 Valley Santa Fe de 7 12 11 10 13 9 11 Antioquia La Pintada 12 9 11 11 11 16 10 10 La Paila 12 12 3 3 Mata de 10 1 4 3 1 Guadua CIAT 2 3 2 Jamundí 7 Magdalena Tolima N 5 4 Valley Tolima S 9 1 2 Tatacoa N 10 10 Tatacoa S 6 3 Patía Valley Patía 10 8 8 Chicamocha Chicamocha 5 9 10 13 12 10 10 Valley N° of 11 6 10 9 4 5 11 9 10 5 6 populations Total n° trees 96 45 96 74 42 40 96 65 85 37 49 219 220 Figure 1: Sampling locations for all species (first panel 1) and individual species (remaining panels). The potential 221 tropical dry forest (TDF) distribution (see section 2.1) is shown in green. Populations are indicated with red 222 triangles and are numbered as follows: 1. Santa Marta Norte; 2. Santa Marta Sur; 3. El Guamo; 4. Zambrano; 5. 223 Colosó; 6. Ituango; 7. Santa Fe de Antioquia; 8. La Pintada; 9. La Paila; 10. Mata de Guadua; 11. Centro 224 Internacional de Agricultura Tropical (CIAT); 12. Jamundí; 13. Tolima Norte; 14. Tolima Sur; 15. Tatacoa Norte; 225 16. Tatacoa Sur; 17. Patía; 18. Chicamocha. Populations 1-5 are located along the Carribean coast, localities 6-12 226 in the Cauca Valley, localities 13-16 in the Magdalena Valley, locality 17 in the Patía Valley, and locality 18 in 227 the Chicamocha Valley. 228 2.4 Genetic markers 229 Total genomic DNA was extracted from leaf material using the CTAB procedure (Doyle, J. J. and 230 Doyle, 1987), with modifications described by Novaes et al. (2009), and was amplified using 231 polymerase chain reactions (PCR). Seven of the study species (A. saman, A. polyneuron, A. graveolens, 232 C. odorata, C. pentandra, E. cyclocarpum and H. courbaril) were characterized using simple sequence 233 repeat markers (SSRs), also called microsatellite markers, with the number of loci per species ranging 234 from 7 (A. graveolens) to 12 (A. saman) (Table A.2, Supplementary material), whereas the four 235 remaining species (B. simaruba, C. ebano, H. crepitans, P. pinnatum), were characterized using inter- 236 simple sequence repeat markers (ISSRs), with the number of loci per species ranging from 44 (H. 237 crepitans) to 67 (B. simaruba) (Table A.2, Supplementary material). SSR markers are codominantly 238 inherited whereas ISSR markers are dominantly inherited, which means that ISSR markers do not allow 239 to distinguish between heterozygosity and homozygosity. The data of the species characterized with 240 ISSR markers have been published by Bocanegra-González et al. (2019), those of A. saman by Aguirre- 241 Morales et al. (2020), those of A. graveolens by Morillo-Paz (2019), those of C. odorata by Aguirre- 242 Morales (2017), those of C. pentandra by Bocanegra-González et al. (2018), and those of E. 243 cyclocarpum by Thomas et al. (2016). Those of A. polyneuron and H. courbaril have not yet been 244 published. Although similar, SSR and ISSR markers may result in different estimates of population 245 genetic differentiation (Ganopoulos et al., 2011; Li et al., 2017; Rawat et al., 2014), but comparisons 246 between species were not a primary objective of this study. 247 2.5 Environmental variables 248 A set of 30 climate and soil variables were selected to evaluate IBE (see section 2.6) and to construct 249 seed zones (see section 2.7) (Table A.1, Supplementary material). These comprise 19 bioclimatic 250 variables from the WorldClim v1.4 database (Hijmans et al., 2005), aridity (i.e. annual precipitation 251 divided by annual potential evapotranspiration), which was calculated using WorldClim temperature 252 and precipitation data and the Hargreaves formula for estimating potential evapotranspiration 253 (Hargreaves and Allen, 2003), and 9 continuous soil variables from the ISRIC SoilGrids250m v1 254 database (Hengl et al., 2017). All variables were used at a spatial resolution of 30 arcsec (ca. 0.9 km at 255 the equator). 256 Future climate data were downloaded from the CCAFS Climate Data Portal (http://www.ccafs- 257 climate.org/). Data were downloaded for the 2050s and 2070s time horizons and two greenhouse gas 258 emission scenarios, using the representative concentration pathway RCP8.5 as the worst-case scenario 259 and RCP4.5 as a more optimistic scenario, as forecasted by five general circulation models (GCMs). 260 These GCMs were selected using the same approach as in Fremout et al. (2020), first selecting all GCMs 261 that perform better than the median GCM performance (against observed temperature and precipitation 262 values) following Knutti et al. (2013) and that are available at the CCAFS Climate Data Portal and next 263 maximizing the dissimilarity between GCMs by selecting the GCM with the best performance in each 264 node of the GCM family tree of Knutti et al. (2013), determined after cutting the tree at level 16 265 (Schlaepfer et al., 2017). This procedure resulted in the selection of 5 GCMs: CESM1(CAM5), GFDL- 266 CM3, HADGEM2-ES, MIROC5, and MPI-ESM-LR. 267 2.6 Population genetic structure and isolation by distance and environment 268 To assess whether populations were genetically differentiated, we first calculated the genetic distance 269 between sampled trees. The simple matching coefficient was used to express genetic distance for the 270 species characterized by ISSR markers, whereas the Kosman-Leonard distance was used for the species 271 characterized by the codominant SSR markers, for which traditional genetic distance measures are not 272 appropriate (Kosman and Leonard, 2005). We calculated these genetic distances using the ‘ade4’ and 273 ‘mmod’ packages for R (Dray and Dufour, 2007; Winter, 2012). Next, these distances were used to test 274 for genetic differentiation between the sampled populations by carrying out an Analysis of Molecular 275 Variance (AMOVA; Excoffier et al. 1992) for each of the species, using the ‘adegenet’ and ‘ade4’ 276 packages for R (Dray and Dufour, 2007; Jombart and Ahmed, 2011). AMOVA, which results in ΦST 277 statistics analogous to FST statistics (Wright, 1949), accommodates both dominant and codominant 278 marker data and works relatively well with limited numbers of loci (Nelson and Anderson, 2013). 279 Statistical significance of population genetic differentiation was assessed with a permutation test (n = 280 1000) using the ‘randtest’ function of the ‘ade4’ package, which involves the random permutation of 281 the rows of the genetic distance matrix (Excoffier et al., 1992). When overall differentiation was 282 significant, pairwise AMOVA analyses were carried out to evaluate differentiation between individual 283 populations. 284 To evaluate IBD and IBE, we compared the pairwise genetic distances between trees described above 285 with the corresponding geographical and environmental distances. To remove multicollinearity between 286 environmental variables, environmental distances between trees were calculated as the Euclidean 287 distance in the multi-dimensional space made up by the first n PCs (principal components) derived from 288 a principal component analysis (PCA) on the entire set of environmental variables, with n the number 289 of PCs having an eigenvalue higher than one (i.e. the Kaiser-Guttman criterion; Guttman, 1954). 290 Geographical distance was calculated as the Euclidean distance between tree locations. Next, we used 291 linear mixed effects models (LMMs) to regress genetic distance against geographic distance and 292 environmental distance. LMMs have been found to outperform other regression methods for landscape 293 genetic analysis (Shirk et al., 2018), including Mantel tests, which have been criticized in the context 294 of spatial analysis (Guillot and Rousset, 2013; Legendre et al., 2015). Following the maximum 295 likelihood population effects framework (MLPE; Clarke et al. 2002), we accounted for the non- 296 independence of pairwise comparisons using random effects, with tree identity and population as nested 297 random effects. Geographic distance and environmental distance were included as fixed effects. 298 For each species, three LMMs were fit using the ‘lme4’ package for R (Bates et al., 2015), using (i) 299 geographic distance, (ii) environmental distance, and (iii) both geographic and environmental distance 300 as fixed effects. Each of these LMMs was then compared with a null model including only random 301 effects, using the Akaike Information Criterion (AIC) corrected for small sample sizes (Hurvich and 302 Tsai, 1989). Models were with fit with maximum likelihood (ML) rather than REML (restricted 303 maximum likelihood) to obtain valid AIC scores (Clarke et al., 2002; Shirk et al., 2018). Models with 304 AIC scores less than ten units lower than the AIC score of the null model were considered to have 305 insufficient support (Burnham and Anderson, 2004). For each species, the best performing model (IBD, 306 IBE, or IBD + IBE) was selected based on the AIC scores. In addition, the marginal R-squared (i.e. the 307 proportion of variance explained by the fixed effects) was estimated for each of the fitted LMMs 308 following Nakagawa et al. (2017), using the ‘MuMIn’ package for R (Barton, 2016). 309 2.7 Seed zones 310 To guide seed sourcing decision in the TDFs of Colombia, we aimed at constructing seed zones that are 311 (i) applicable across tree species, (ii) dynamic in the light of ongoing climate change, and (iii) able to 312 capture as much as possible of the observed population genetic differentiation of the study species, but 313 large enough to keep it practically feasible to implement the seed zones on the ground. We constructed 314 geo-environmental seed zones with the CLARA (Clustering Large Applications) approach, an extension 315 for large datasets of the partitioning-around-medoids (PAM) algorithm, which represents a more robust 316 alternative of the k-means algorithm (Kaufman and Rousseeuw, 1990). To reflect both environmental 317 conditions and geographical location, the environmental variables, which were represented by the same 318 PCs described in the previous section, were clustered along with longitude and latitude. Longitude and 319 latitude were rescaled to have a variance equal to the average variance of the first six PCs. The number 320 of clusters (i.e. number of seed zones) was varied between 5 and 50. To evaluate the performance of 321 the resulting seed zone maps, we carried out additional AMOVAs to estimate the percentage of 322 molecular variance explained by each of these seed zone scenarios. The optimal number of seed zones 323 for each of the species was estimated as the number that explained most of the molecular variance. In 324 order to obtain a seed zone map applicable across species, we selected the seed zone map with the 325 highest number of zones among these optima. To reduce the occurrence of isolated patches in this 326 proposed seed zone map, it was subject to a modal filter with a window of 3x3 grid cells (i.e. each grid 327 cell was assigned to the most common seed zone in the 3x3 window). 328 To project the current seed zones to future climate conditions, the predicted future environmental 329 conditions of every grid cell within the potential TDF distribution were projected to the same PCs as 330 those used for the clustering of the current seed zones (i.e. the variables describing the future 331 environmental conditions were recombined using the same linear combinations of variables as defined 332 by the PCA that was carried out based on the current environmental variables), after which each grid 333 cell was assigned to the closest cluster medoid, using the ‘cl_predict’ function of the ‘clue’ package for 334 R (Hornik, 2019). In this way, the current seed zones were projected to future climate conditions as 335 predicted by each of the selected GCMs under emission scenarios RCP4.5 and RCP8.5 for the 2050s 336 and 2070s (section 2.5). 337 3 Results 338 We found significant genetic differentiation between populations for all 11 species, with the percentage 339 of genetic variance between populations ranging from only 4% for E. cyclocarpum to 28% for P. 340 pinnatum. Overall population differentiation was highly significant (P<0.001) for all species except for 341 E. cyclocarpum (P=0.02). For most species, at least half of pairwise comparisons between populations 342 was significant, except for E. cyclocarpum (5 out of 36 comparisons) and A. saman (23 out of 55 343 comparisons). C. ebano was the only species for which all pairwise combinations of populations were 344 significantly differentiated (6 out of 6 comparisons) (Table 3). 345 Table 3: AMOVA (Analysis of Molecular Variance) partitioning of genetic variance for the tree species included 346 in this study, showing the results of the overall AMOVA in the second column and results of the pairwise AMOVA 347 analyses in the third column. All overall AMOVA analyses were statistically significant (P<0.05). SSR and ISSR 348 stand for simple sequence repeats and inter-simple sequence repeats, respectively. % Variance between populations Number of significant pairwise Species (ΦST) comparisons SSR markers Albizia saman (B) 14 24/55 Aspidosperma polyneuron (A) 15 16/28 Astronium graveolens (A) 7 26/45 Cedrela odorata (A) 18 5/10 Ceiba pentandra (A) 10 38/66 Enterolobium cyclocarpum (B) 4 6/36 Hymenea courbaril (B) 10 5/10 ISSR markers Bursera simaruba (B) 19 18/21 Caesalpinia ebano (B) 19 6/6 Hura crepitans (A) 25 31/45 Platymiscium pinnatum (A) 27 11/15 349 Table 4 presents the results of the pairwise AMOVA analyses in more detail. Genetic differentiation 350 not only occurred between the five biogeographic regions where sampling was carried out but also 351 within these regions, with 47% of the pairwise population combinations within regions being 352 significantly differentiated. For example, the Santa Fe de Antioquia and Ituango populations, despite 353 being separated by only ca. 100 km and being both located in the Cauca river valley, were genetically 354 differentiated for 3 out of 5 species (Table 4). This is further illustrated by the significant differentiation 355 between all four populations of C. ebano, which are all located in the same region (Caribbean coast). 356 Further, Table 4 shows that the Patía population was clearly the most differentiated from the other 357 sampled populations, with significant pairwise differentiation for all but two of the pairwise 358 comparisons, and high percentages of genetic variance between populations, as high as 50% and higher 359 (Table 4). While genetic differentiation between the five sampling regions was often significant, this 360 was not always the case. For example, the populations in the south of the Magdalena valley (TAN: 361 Tatacoa N; TAS: Tatacoa) were only differentiated from the populations in the Cauca valley in 50% of 362 the comparisons. 363 Table 4: Pairwise AMOVA (Analysis of Molecular Variance) results. The lower triangular matrix shows the number of species for which each of the pairwise comparisons 364 indicated significant genetic differentiation (P<0.05), as the fraction of the total number of species for which pairwise comparisons were made. The upper triangular matrix 365 indicates the average percentage of variance situated between the two populations (calculated from all the pairwise comparisons, not only those that were statistically significant). 366 Dashes (“-”) indicate pairs of populations for which no comparisons could be made. Populations are abbreviated as follows: SMN: Santa Marta N; SMS: Santa Marta S; GUA: 367 El Guamo; ZAM: Zambrano; COL: Colosó; ITU: Ituango; SFE: Santa Fe de Antioquia; PIN: La Pintada; PAI: La Paila; MAT: Mata Guadua; CIA: Centro Internacional de 368 Agricultura Tropical; JAM: Jamundí; TON: Tolima N; TOS: Tolima S; TAN: Tatacoa N; TAS: Tatacoa S; PAT: Patía; CHI: Chicamocha. Biogeographic SMN SMS GUA ZAM COL ITU SFE PIN PAI MAT CIA JAM TON TOS TAN TAS PAT CHI region Carribean SMN 11 12 15 15 17 8 14 21 35 0 - 27 22 2 4 59 20 coast SMS 3/7 10 10 13 15 11 11 9 15 0 - 23 11 10 5 49 19 GUA 2/5 3/5 10 5 13 2 11 18 13 9 - - 26 0 1 35 13 ZAM 4/5 2/5 4/4 11 15 10 12 15 16 16 - 22 22 4 1 44 18 COL 2/5 4/6 1/5 2/4 21 8 14 13 6 3 - 18 12 7 0 28 19 ITU 2/4 4/5 2/3 3/3 3/4 13 11 17 14 0 - - - 19 11 50 25 Cauca valley SFE 0/4 3/5 0/4 2/4 2/5 3/5 7 11 -7 5 15 - 17 4 -3 29 15 PIN 2/4 2/5 3/4 3/3 3/6 1/4 2/6 14 12 9 20 5 13 14 5 31 18 PAI 1/3 1/3 2/3 3/3 2/3 3/3 2/4 3/3 0 0 - - 24 - 12 45 16 MAT 1/2 0/2 1/2 2/3 1/2 1/2 1/4 0/3 0/3 0 35 - 21 - 13 25 15 CIA 0/1 0/1 1/2 1/1 0/3 0/1 0/2 0/3 0/2 0/1 - 2 18 - 10 34 7 JAM - - - - - - 1/1 1/1 - 0/1 - - - - - 12 - Magdalena TON 1/1 1/1 - 1/1 2/2 - - 0/1 - - 0/1 6 - - - 13 valley TOS 1/1 0/1 1/1 2/2 1/2 - 1/1 0/1 1/1 1/1 0/1 - 0/1 29 21 29 20 TAN 0/1 1/2 0/1 0/1 1/2 2/2 0/2 1/2 - - - - - - - - 8 TAS 0/1 0/1 0/1 0/2 0/1 1/1 0/2 1/2 2/2 0/2 0/1 - - 1/1 - 50 7 Patía valley PAT 1/1 1/1 1/1 2/2 1/1 1/1 2/3 3/3 2/2 2/3 1/1 1/1 - 1/1 - 2/2 43 Chicamocha CHI 2/4 4/5 3/4 3/4 4/6 3/4 4/5 4/6 3/3 1/3 0/2 - 1/1 0/1 1/2 0/2 2/2 valley 369 370 The first six PCs resulting from the PCA of the environmental variables (n = 30) together explained 371 85% of the total variance. Already 60% was explained by the first two PCs only, mainly reflecting 372 variation in temperature and precipitation, respectively. The loadings of the environmental variables of 373 the first six PCs are given in Table A.4 and the scree plot is given in Figure A.1 (Supplementary 374 material). 375 The results of the LMMs that were fit to evaluate IBD and IBE are given in Table 5. For 6 out of 11 376 species, the model with the lowest AIC was the model including both geographic distance and 377 environmental distance (IBD + IBE) as explanatory variables, whereas the IBE model had the lowest 378 AIC for the 5 remaining species (Table 5). On average, environmental distance was a better predictor 379 of genetic distance, according to both the AIC and R2 values (Table 5). 380 Table 5: Results of the linear mixed effects models (LMMs) evaluating the relationship between genetic distances 381 between trees and corresponding geographical distances (isolation by distance; IBD) and environmental distances 382 (isolation by environment; IBE). LMMs were fitted within a maximum likelihood population effects (MLPE) 383 framework, using trees and population as nested random effects and geographical and environmental distance as 384 fixed effects. The reported R2 values are marginal R2 values only reflecting the variance explained by the fixed 385 effects. ∆AIC scores refer to the difference between the AIC score of the fitted model and the AIC score of a null 386 model with only random effects. All AIC scores were corrected for small sample sizes. Models with ∆AIC scores 387 higher than -10 were not considered to have sufficient support and are not shown. For each species, the best model 388 (last column) was identified based on the lowest ∆AIC. Isolation by Isolation by environment IBD + IBE Best model distance (IBD) (IBE) ∆AIC R2 ∆AIC R2 ∆AIC R2 SSR markers Albizia saman -118.7 0.03 -12.3 0.00 -156.1 0.05 IBD + IBE Aspidosperma polyneuron -79.0 0.10 -20.2 0.03 -79.2 0.10 IBD + IBE Astronium graveolens -11.3 0.00 -13.3 0.00 IBD + IBE Cedrela odorata -36.5 0.06 -77.3 0.11 -75.7 0.11 IBE Ceiba pentandra -101.7 0.02 -12.3 0.00 -113.1 0.03 IBD + IBE Enterolobium cyclocarpum -38.1 0.02 -34.0 0.02 -41.8 0.02 IBD + IBE Hymenea courbaril -18.8 0.03 -17.5 0.03 IBE ISSR markers Bursera simaruba -11.4 0.01 -198.5 0.17 -196.9 0.17 IBE Caesalpinia ebano -60.4 0.11 -65.2 0.13 -64.2 0.13 IBE Hura crepitans -62.4 0.04 -118.7 0.07 -116.8 0.07 IBE Platymiscium pinnatum -85.3 0.19 -93.0 0.20 -98.0 0.22 IBD + IBE Average -55.7 0.05 -60.0 0.07 -88.4 0.08 389 The performance of the seed zone maps with different number of seed zones was evaluated by means 390 of AMOVAs (Table 6; full results in Table A.5 in Supplementary material). The optimal number of 391 seed zones, determined as the number of seed zones explaining most of the genetic variance, ranged 392 from 5 (A. polyneuron) to 36 (C. pentandra). In order to obtain a conservative, risk-minimizing seed 393 zone map applicable across species, we selected the base map with 36 zones for further analysis. In 394 addition, this approach has the advantage that seed zones could be merged for species with known 395 genetic variance which can be explained by fewer zones. 396 Table 6: Optimal number of seed zones per species as estimated by the AMOVAs, also showing the percentage 397 of genetic variance situated between zones at these numbers, and the percentage of genetic variance situated 398 between populations for comparison. % variance between % variance between Optimal n° seed zones seed zones populations SSR markers Albizia saman 26 17 14 Aspidosperma polyneuron 5 16 15 Astronium graveolens 27 7 7 Cedrela odorata 8 18 18 Ceiba pentandra 36 11 10 Enterolobium cyclocarpum 26 5 4 Hymenea courbaril 20 10 10 ISSR markers Bursera simaruba 24 19 19 Caesalpinia ebano 19 18 19 Hura crepitans 27 23 25 Platymiscium pinnatum 27 27 27 399 The seed zones under present climatic conditions are shown in Figure 2. Panel b, depicting the seed 400 zones with colours reflecting environmental similarity, shows that some seed zones located far apart 401 have similar environmental conditions, as illustrated for example by the purple colors in both the central 402 part of the Caribbean TDFs and the Magdalena valley. However, by adding longitude and latitude as 403 input variables in the clustering, these areas were assigned to different seed zones, as would be expected 404 under an IBD scenario. Figure 3 depicts the seed zones under both present (panel a) and future climate 405 conditions (2050s; panels b and c), showing the expected expansion and contraction of seed zones under 406 future climatic conditions, as illustrated for example by the predicted southward expansion of the pink 407 seed zone of the Guajira peninsula in the northernmost part of Colombia. 408 Figure 2: Seed zones (n = 36) for the tropical dry forests (TDFs) of Colombia under present climate conditions. The colours of the seed zones in panel a were randomly chosen, whereas the colours in panel b reflect the environmental similarity between seed zones, using a red-green-blue colouring scheme determined by the average scores of the three first principal components (PCs) of the environmental variables, as proposed by Hargrove and Hoffman (2005). 409 410 Figure 3: Seed zones (n = 36) for the tropical dry forests (TDFs) of Colombia under present climate conditions 411 (panel a) and future climate conditions for the 2050s under the representative concentration pathways RCP4.5 and 412 RCP8.5 (panel b and c). The future seed zones shown here are those predicted by the HadGEM2-ES model, one 413 of the five selected GCMs (section 2.5), but note that the tool in which the seeds zones are integrated 414 (www.diversityforrestoration.org/) combines the predictions of all five GCMs. 415 4 Discussion 416 4.1 Population genetic differentiation 417 All of the 11 study species showed significant genetic differentiation between populations, with the 418 percentage of variance situated between populations (ΦST) ranging from 4% to 28% (Table 3). These 419 figures are similar to other neutral marker characterizations of Neotropical tree species at similar spatial 420 scales (e.g., Chase et al. 1995, Lacerda et al. 2001, Lowe et al. 2003, Cerón-Souza et al. 2005) and 421 within the range of expected values for outcrossing long-lived species, which tend to retain most of the 422 genetic variability within populations (Hamrick and Godt, 1996; Nybom, 2004; Reisch and Bernhardt- 423 Römermann, 2014). The marked differences between species, also between species characterized by 424 the same type of markers, suggest that species-specific seed zones may be more appropriate than general 425 seed zones. However, collecting genetic data on all the species used in restoration of the Colombian 426 TDFs, which are home to several hundreds of tree species, would be a daunting task, and it may be 427 unrealistic to expect restoration practitioners to implement different seed zones for each of the species 428 used, given the generally limited logistic capacity and lack of incentives to obtain appropriate planting 429 material (Atkinson et al., 2018; Jalonen et al., 2018). Trait-based generalizations provide one possible 430 solution, but previous research has shown that predicting the degree of population genetic 431 differentiation using life history traits is far from an easy undertaking, as relationships between both are 432 generally weak (Duminil et al., 2007). We therefore adopted a more pragmatic approach and aimed to 433 construct a single set of seed zones applicable across species, drawing on genetic marker data of 11 434 species with differing life history traits (Table 1). 435 The current difference in population genetic structure between the study species are the result of a 436 complex interplay of different factors. Past climatic changes are likely to have led to different degrees 437 of range contraction and expansion and possibly convergence of different genetic groups in some areas 438 (Aguirre-Morales et al., 2020; Bocanegra-González et al., 2018; Thomas et al., 2016). Using Bayesian 439 genetic clustering, these studies furthermore suggest that exchange of tree genetic resources across the 440 Inter-Andean valleys may have taken place during or prior to the Last Glacial Maximum (LGM) 441 (Aguirre-Morales et al., 2020; Bocanegra-González et al., 2018), which may explain why genetic 442 differentiation between populations in the Magdalena valley and those in the Cauca valley was often 443 not significant (Table 4). 444 Other important factors influencing the genetic differentiation between populations are species’ seed 445 dispersal modes, pollination modes, and mating systems (Ballesteros-Mejia et al., 2016; Hamrick et al., 446 1993; Lowe et al., 2018), which result in differences in the degree of gene flow. For example, our 447 findings partly followed the expected pattern that populations of animal-dispersed species tend to be 448 less genetically differentiated (average ΦST =13.2%) than those of abiotically dispersed species (e.g. 449 wind, autochory; average ΦST = 15.0%) as a consequence of gene flow over larger distances in animal- 450 dispersed species (Hamrick et al. 1993, Lowe et al. 2018, but see Duminil et al. 2007), although some 451 wind-dispersed seeds can be dispersed over extremely large distances (e.g., C. pentandra; Dick et al., 452 2007). The species with the lowest degree of differentiation, E. cyclocarpum (Table 3), used to be 453 dispersed by Pleistocene horses (Janzen and Martin, 1982), but has long been widely planted in pasture 454 lands and is now mainly dispersed by introduced cattle and horses (Gonzales et al., 2010; Janzen, 1982), 455 resulting in large distance dispersal when these animals are moved around by humans. Human-mediated 456 dispersal may also have become the dominant dispersal mode of other species originally dispersed by 457 megafauna, including H. courbaril and A. saman (Janzen and Martin, 1982). E. cyclocarpum and H. 458 courbaril also have a long history as human food (Iriarte et al., 2020; Zizumbo-Villarreal et al., 2016) 459 and the same is possibly true for A. saman. However, as each of the factors mentioned above may have 460 affected the genetic structure of each of the species differently, it is difficult to explain the findings of 461 the AMOVAs (Table 3). In addition, the differences between species in our study should be interpreted 462 with care because not all species were present in all sampling sites and because two different marker 463 types were used. Rigorous hypothesis testing on the relationship between life history traits and 464 population genetic structure was therefore not part of the scope of the present study. Rather, the value 465 of the selected set of species lies in the broad range of life history traits they represent, increasing the 466 likelihood that the obtained results are applicable across many tree species of the Colombian TDFs. 467 For more than half (6 out of 11) of the study species, genetic differences between trees was best 468 explained using both geographical and environmental distance, while they were best explained by 469 environmental distance for the remaining species (Table 5), suggesting that the observed genetic 470 differentiation is a consequence of both neutral and adaptive processes, in line with similar findings 471 elsewhere (Sexton et al. 2014). The obtained R2 values were relatively low for most species, which was 472 expected since the studied species retain the majority of genetic variability within populations, within 473 which both geographic and environmental distances are limited. As discussed by Sexton et al. (2014), 474 IBE results from reduced gene flow between populations growing under divergent environmental 475 conditions, which may be the consequence of natural selection, with maladapted immigrants or 476 offspring of local and immigrant parents failing to germinate, survive or reproduce, and hence failing 477 to introduce their genes in the local population. It may also be caused by nonrandom mating resulting 478 from different environmental conditions, both as a consequence of genetic adaptation (e.g., flowering 479 period mismatch between immigrants and locals) or phenotypic plasticity (e.g., plants growing at 480 different altitudes flowering at different times). Without provenance trials, it is difficult to determine 481 the relative importance of these different mechanisms (Sexton et al., 2014). 482 Neutral markers have been used extensively to inform provenancing decisions (Durka et al., 2017; 483 Jorgensen et al., 2016; Krauss et al., 2013; Krauss and He, 2006; Listl et al., 2018; Malaval et al., 2010; 484 Massatti et al., 2020). However, they are not directly linked to local adaptation (Holderegger et al., 485 2006), which has led to criticism on their use in guiding seed provenancing (McKay et al., 2005). 486 Nevertheless, meta-analyses have found that differentiation in adaptive and neutral markers tends to be 487 correlated (Leinonen et al., 2008; Merilä and Crnokrak, 2001). Divergent natural selection (i.e. selection 488 acting in different directions on different populations) can not only lead to population differentiation by 489 acting on specific loci or those physically associated with them, but can also promote barriers to gene 490 flow as discussed above. This results in genome-wide neutral divergence via genetic drift, which is 491 readily detectable by neutral markers (Nosil et al., 2009). 492 4.2 Seed zones 493 The population genetic differentiation we found in all the study species (Table 3) supports the need for 494 establishing seed zones to promote adaptedness of seeds used in tree planting activities and to avoid 495 disruptions of genetic patterns. Pairwise genetic differentiation between populations within the five 496 sampled biogeographic regions was often significant (47% of within-region comparisons were 497 significant; Table 4), indicating that population differentiation occurs at a finer scale than these five 498 regions, which should be reflected in a sufficiently high number of seed zones. Further, as our results 499 indicate that the genetic differences between trees are related to both geographic distance and 500 environmental differences (Table 5), we included longitude and latitude along with the PCs derived 501 from environmental variables in the PAM clustering, seeking to construct seed zones that are 502 environmentally homogeneous while also avoiding large geographic distances between locations within 503 the same seed zone. This reduces the probability that isolated areas with similar environmental 504 conditions would be grouped in the same seed zone, which is a disadvantage of seed zones based on 505 clustering environmental variables only (Potter and Hargrove, 2012). 506 Non-hierarchical clustering methods such as k-means clustering or PAM have been used before to 507 delineate seed zones (Potter and Hargrove, 2012; Shryock et al., 2018) or ecoregions (Hargrove and 508 Hoffman, 2005). Compared to hierarchical clustering, these methods have the advantage that they result 509 in similar environmental heterogeneity within different seed zones (Hargrove and Hoffman, 2005). We 510 followed Shryock et al. (2018) by submitting the environmental variables to a PCA prior to clustering, 511 which is more robust against the inclusion of correlated variables, and does not require prior knowledge 512 about which environmental variables are most closely linked to local adaptation of tree species 513 populations. We generated seed zone maps with differing number of seed zones, ranging from 5 to 50, 514 and determined the optimal number of zones by carrying out an AMOVA for each of these numbers of 515 zones. However, it should be noted that the sampling density was relatively low, and that the number 516 of seed zones proposed here should be seen as the optimal scenario based on currently available 517 information. In order to more thoroughly study the spatial scale of genetic differentiation, a more dense 518 sampling scheme is recommended, in analogy with Malaval et al. (2010) or Michalski and Durka 519 (2012). 520 We projected the proposed set of seed zones to the climate conditions anticipated in the 2050s and 521 2070s by assigning the future conditions of each of the grid cells to the closest cluster medoid resulting 522 from the CLARA algorithm using the present environmental conditions as an input. As such, a grid cell 523 is assigned to another seed zone under future climate conditions only if its distance in PCA space to the 524 cluster medoid of that other seed zone is shorter than to the medoid of the present seed zone. By also 525 including longitude and latitude in the clustering, the probability that grid cells are assigned to another 526 seed zone under future climate conditions decreases with the geographic distance to the grid cell 527 representing the cluster medoid of the latter zone. Grid cells changing from one seed zone to another 528 result in the contraction and expansion of seed zones under future climate conditions, for example seed 529 zones that are currently already characterized by high temperatures are likely to expand under future 530 conditions. 531 Our approach to predict shifts in seed zones from present to future climate condition is similar to the 532 multivariate spatiotemporal clustering (MSTC) approach proposed by Hargrove and Hoffman (2005), 533 with the difference that in MSTC, predicted future environmental conditions are used directly as input 534 variables in the clustering along with the present environmental conditions. Consequently, when some 535 of the predicted future conditions are very dissimilar to any of the current conditions, some clusters in 536 MSTC may only exist under future conditions (i.e. represent novel climates). While such novel climates 537 are likely to occur in the future, in the context climate-smart seed sourcing, we considered it more 538 important to identify the current seed zone where conditions are most similar to the expected novel 539 climate. 540 The aim of the seed zones presented here is to promote the use of planting material that is well-adapted 541 to present and future climate conditions in the TDFs of Colombia. To facilitate their use, the zones have 542 been integrated into a map-based online decision-support tool (DiversityForRestoration, available at 543 www.diversityforrestoration.org; Thomas et al. 2017a). The tool recommends to supplement planting 544 material from the present seed zone with material from areas currently located in the seed zone that is 545 anticipated at the planting site under future climate conditions, i.e. combining local provenancing with 546 predictive provenancing. While the tool recommends a 50/50 ratio, some restoration planners may 547 consider the risk of introducing 50% material from a non-local seed zone too high, and may opt for only 548 introducing only a smaller amount of non-local seeds, coinciding with the ‘genetic enrichment’ 549 approach proposed by Lefèvre et al. (2013). Both local and predictive provenancing may include one 550 or more source populations (more is better to increase adaptive potential, but logistical costs may be a 551 constraint). 552 While all five GCMs predict marked increases in temperature in the TDFs of Colombia, the direction 553 of predicted precipitation changes is not always consistent. As a result, we found that different GCMs 554 did not always coincide in the future seed zone projections, in which case the tool recommends sourcing 555 proportional parts of the planting material in each of the future seed zones as predicted by different 556 GCMs. This approach coincides with the risk-minimizing portfolio approach proposed by Crowe and 557 Parker (2008), directly incorporating the uncertainty of future climate predictions, as illustrated in 558 Figure 4. 559 560 Figure 4: Illustration of the risk-minimizing seed sourcing strategy (‘portfolio approach’) proposed in the 561 DiversityForRestoration decision support tool (www.diversityforrestoration.org), which involves sourcing 50% 562 of the seeds in the present seed zone, and 50% in the future seed zones as predicted by different General 563 Circulation Models (GCMs), recommending to source from different future seed zones if different GCMs do not 564 coincide in their predictions of future seed zones. This is illustrated here with two GCMs for simplicity, but note 565 that 5 GCMs were used (section 2.5). 566 4.3 Final considerations and prospects 567 Climate-smart seed sourcing is a complex issue and the ‘local + predictive’ strategy we adopted in the 568 DiversityForRestoration tool is only one of several possible strategies. The seed zones presented here 569 can be used to implement other strategies too, as illustrated in Figure 5. While the inclusion of longitude 570 and latitude in the clustering ensures that seed zones do not expand over very large geographic distances, 571 limiting the distances over which predictive provenancing is recommended, some restoration planners 572 may prefer to not carry out any assisted gene flow at all, as experimental evidence of its potential 573 remains limited and some concerns remain about the risk of outbreeding depression and mismatches in 574 biotic interactions (Aitken and Whitlock, 2013; Bucharova, 2017; Bucharova et al., 2018). To increase 575 adaptive capacity, these restoration planners may opt to only focus on the collection of genetically 576 diverse and locally adapted planting material instead. In this case, the proposed seed zones can be used 577 as boundaries within which seeds of several populations may be mixed, coinciding with the regional 578 admixture approach proposed by Bucharova et al. (2018), which we termed ‘zonal admixture’ for 579 consistency with the seed zones terminology. Similarly, seed zones can be used as boundaries for 580 composite provenancing (Breed et al., 2013), which we termed ‘zonal composite’ provenancing, while 581 combining seeds from different seed zones coincides with an admixture provenancing strategy 582 (Broadhurst et al., 2008). Lastly, when sourcing seeds for restoration in the long term (many restoration 583 objectives will typically go beyond the 2050s; e.g. biodiversity conservation, slow-growing timber 584 species), an approach similar to the climate-adjusted provenancing approach (Prober et al., 2015; 585 Ramalho et al., 2017) could be implemented by sourcing from both the seed zone predicted for the 586 2050s and the seed zone predicted for the 2070s, if different (Figure 5). In this way, seed zones provide 587 a useful tool to implement different previously proposed climate-smart seed sourcing strategies in a 588 pragmatic way. It is important to note that restoration should not be considered a onetime activity, and 589 that adaptive management may include assisted migration only at a later stage, for example when it has 590 become more clear that local populations are not able to adapt fast enough, or when the uncertainty 591 about future climatic changes has become lower. Other climate-smart adaptive management strategies 592 are discussed in Lefèvre et al. (2013). 593 594 Figure 5: Use of dynamic seed zones for different seed sourcing strategies, in relation to the trade-off between local adaptation 595 vs. genetic variability, modified from Bucharova et al. (2018) (climate gradient replaced by seed zones and some seed sourcing 596 strategies added or modified). The sizes of the grey circles reflect the relative contributions of different seed source populations. 597 The ‘regional admixture’ approach proposed by Bucharova et al. (2018) is termed ‘zonal admixture’ here to fit with the seed 598 zone terminology. The climate-adjusted provenancing follows Ramalho et al. (2017) by making the amount of seed smaller 599 further along the climate change gradient and was therefore placed higher on the local adaptation axis than in Bucharova et al. 600 (2018). Note that with ‘local’ seed sourcing we refer to any seed source population within the local seed zone under current 601 climate conditions. We assume that all populations in the local seed zone are equally well adapted to local conditions; the local 602 seed sourcing strategy is therefore put at the same level of local adaptation as the ‘zonal admixture’ and ‘zonal composite’ 603 seed sourcing strategies. Similarly, we assume that the ‘local + predictive’ and the ‘zonal admixture’ approaches result in a 604 similar level of total genetic variability (the ‘local + predictive’ approach consists of a lower number of provenances but from 605 more different environmental conditions). Note that the seed sourcing strategies shown here are not exhaustive and that more 606 variations are possible, for example the predictive provenancing approach may involve sourcing seeds from more than one 607 source population. 608 Once restoration practitioners have identified the appropriate seed zones, they need to identify at least 609 one seed source or seed provider in each of these zones. Hence, it is clear that any seed sourcing strategy 610 should go hand in hand with a strategy for the conservation of viable seed sources. The seed zone 611 approach we presented here can also be used as a proxy for the identification of ‘management units’ or 612 ‘evolutionary significant units’ that should be subject of gene conservation efforts (Azpilicueta et al., 613 2013; Potter and Hargrove, 2012; Soliani et al., 2017). Identifying protected areas in each of the seed 614 zones provides one way forward, but only 5% of the remaining Colombian TDFs is currently protected 615 (i.e. less than 1% of the original TDF cover; García et al., 2014). While this underlines the pressing 616 need to step up conservation efforts, it also indicates that protected areas alone will not be sufficient to 617 underpin an efficient and climate-smart national seed sourcing strategy in Colombia. Experiences from 618 the restoration of the Atlantic Forest in Brazil have shown that involving local communities and private 619 landowners in seed collection is a promising way forward for large-scale seed collection (Brancalion et 620 al., 2012; Schmidt et al., 2019). This can provide local communities and private landowners with 621 alternative income sources and serve as an economic incentive for conserving local tree populations. 622 However, it is important that seed providers are trained in proper seed collection practices (Basey et al., 623 2015; Thomas et al., 2017), to ensure that the planting material has a sufficient broad genetic basis 624 promoting the evolutionary potential of established populations (Broadhurst et al., 2008; Thomas et al., 625 2014). 626 Acknowledgements 627 We thank Alvaro Vásquez-Peinado for providing the map depicting the potential distribution of tropical 628 dry forest in Colombia constructed by the Instituto Alexander von Humboldt. 629 Funding 630 This work was supported by the Flemish Interuniversity Council (VLIR-UOS; grant nr. 631 NDOC2016PR002), the German Federal Ministry for Economic Cooperation and Development (BMZ; 632 contract nr. 8121944), the CGIAR Fund Donors (https://www.cgiar.org/funders/), the Colombian 633 companies Ecopetrol and Empresas Públicas de Medellin, and the Government of the Colombian 634 department of Antioquia. The funders had no role in study design, data collection and analysis, or 635 preparation of the manuscript. 636 References 637 Aguilar, R., Ashworth, L., Galetto, L., Aizen, M.A., 2006. Plant reproductive susceptibility to habitat 638 fragmentation: Review and synthesis through a meta-analysis. Ecol. Lett. 9, 968–980. 639 https://doi.org/10.1111/j.1461-0248.2006.00927.x 640 Aguirre-Morales, A.C., 2017. Patrones de distribución y estructura genética de Cedrela odorata y Albizia saman 641 para la conservación y restauración ecológica del bosque seco tropical en Colombia. MSc thesis. 642 Departamento de Ciencias Biológicas, Universidad Nacional de Colombia. 643 Aguirre-Morales, C.A., Thomas, E., Cardozo, C.I., Gutiérrez, J., Alcázar Caicedo, C., Moscoso Higuita, L.G., 644 Becerra López-Lavalle, L.A., González, M.A., 2020. Genetic diversity of the rain tree (Albizia saman) in 645 Colombian seasonally dry tropical forest for informing conservation and restoration interventions. Ecol. 646 Evol. 10, 1905–1916. https://doi.org/10.1002/ece3.6005 647 Aitken, S.N., Whitlock, M.C., 2013. Assisted gene flow to facilitate local adaptation to climate change. Annu. 648 Rev. Ecol. Evol. Syst. 44, 367–388. https://doi.org/10.1146/annurev-ecolsys-110512-135747 649 Aitken, S.N., Yeaman, S., Holliday, J.A., Wang, T., Curtis-McLane, S., 2008. Adaptation, migration or 650 extirpation: climate change outcomes for tree populations. Evol. Appl. 1, 95–111. 651 https://doi.org/10.1111/j.1752-4571.2007.00013.x 652 Alberto, F.J., Aitken, S.N., Alía, R., González-Martínez, S.C., Hänninen, H., Kremer, A., Lefèvre, F., 653 Lenormand, T., Yeaman, S., Whetten, R., Savolainen, O., 2013. Potential for evolutionary responses to 654 climate change – evidence from tree populations. Glob. Chang. Biol. 19, 1645–1661. 655 https://doi.org/10.1111/gcb.12181 656 Aronson, J., Alexander, S., 2013. Ecosystem restoration is now a global priority: Time to roll up our sleeves. 657 Restor. Ecol. 21, 293–296. https://doi.org/10.1111/rec.12011 658 Atkinson, R., Thomas, E., Cornelius, J., Zamora, R., Chuaire, M.F., 2018. Fit-for-purpose seed supply systems 659 for the implementation of landscape restoration under Initiative 20x20:0: An analysis of national seed 660 systems in Mexico, Guatemala, Costa Rica, Colombia, Peru, Chile and Argentina. Bioversity International 661 and World Agroforestry, Lima, Peru. 662 Azpilicueta, M.M., Gallo, L.A., van Zonneveld, M., Thomas, E., Moreno, C., Marchelli, P., 2013. Management 663 of Nothofagus genetic resources: Definition of genetic zones based on a combination of nuclear and 664 chloroplast marker data. For. Ecol. Manage. 302, 414–424. https://doi.org/10.1016/j.foreco.2013.03.037 665 Ballesteros-Mejia, L., Lima, N.E., Lima-Ribeiro, M.S., Collevatti, R.G., 2016. Pollination mode and mating 666 system explain patterns in genetic differentiation in neotropical plants. PLoS One 11, e0158660. 667 https://doi.org/10.1371/journal.pone.0158660 668 Banda, K.R., Delgado-Salinas, A., Dexter, K.G., Linares-Palomino, R., Oliveira-Filho, A., Prado, D., Pullan, 669 M., Quintana, C., Riina, R., Rodríguez, G.M., Weintritt, J., Acevedo-Rodríguez, P., Adarve, J., Álvarez, 670 E., Aranguren, A.B., Arteaga, J.C., Aymard, G., Castaño, A., Ceballos-Mago, N., Cogollo, Á., Cuadros, 671 H., Delgado, F., Devia, W., Dueñas, H., Fajardo, L., Fernández, Á., Fernández, M.Á., Franklin, J., Freid, 672 E.H., Galetti, L.A., Gonto, R., González, R.M., Graveson, R., Helmer, E.H., Idárraga, Á., López, R., 673 Marcano-Vega, H., Martínez, O.G., Maturo, H.M., McDonald, M., McLaren, K., Melo, O., Mijares, F., 674 Mogni, V., Molina, D., Moreno, N.D.P., Nassar, J.M., Neves, D.M., Oakley, L.J., Oatham, M., Olvera- 675 Luna, A.R., Pezzini, F.F., Dominguez, O.J.R., Ríos, M.E., Rivera, O., Rodríguez, N., Rojas, A., Särkinen, 676 T., Sánchez, R., Smith, M., Vargas, C., Villanueva, B., Pennington, R.T., 2016. Plant diversity patterns in 677 neotropical dry forests and their conservation implications. Science (80-. ). 353. 678 https://doi.org/10.1126/science.aaf5080 679 Barton, K., 2016. MuMIn: Multi-Model Inference. R package version 1.15.6. 680 Barton, K.E., Jones, C., Edwards, K.F., Shiels, A.B., Knight, T., 2020. Local adaptation constrains drought 681 tolerance in a tropical foundation tree. J. Ecol. 108, 1540–1552. https://doi.org/10.1111/1365-2745.13354 682 Basey, A.C., Fant, J.B., Kramer, A.T., 2015. Producing native plant materials for restoration: 10 rules to collect 683 and maintain genetic diversity. Nativ. Plants J. 16, 37–53. https://doi.org/10.3368/npj.16.1.37 684 Bates, D., Maechler, M., Bolker, B., Walker, S., 2015. Fitting Linear Mixed-Effects Models Using lme4. J. Stat. 685 Softw. 67, 1–48. 686 Bawa, K.S., 1974. Breeding Systems of Tree Species of a Lowland Tropical Community. Evolution (N. Y). 28, 687 85–92. https://doi.org/10.2307/2407241 688 Bischoff, A., Steinger, T., M̈ller-Schärer, H., 2010. The importance of plant provenance and genotypic diversity 689 of seed material used for ecological restoration. Restor. Ecol. 18, 338–348. https://doi.org/10.1111/j.1526- 690 100X.2008.00454.x 691 Bocanegra-González, K.T., Thomas, E., Guillemin, M.-L., Alcázar Caicedo, C., Moscoso Higuita, L.G., 692 Gonzalez, M.A., De Carvalho, D., 2019. Diversidad y estructura genética de cuatro especies arbóreas 693 clave del Bosque Seco Tropical en Colombia. Caldasia 41, 78–91. 694 https://doi.org/10.15446/caldasia.v41n1.71327 695 Bocanegra-González, K.T., Thomas, E., Guillemin, M.L., de Carvalho, D., Gutiérrez, J.P., Alcázar Caicedo, C., 696 Moscoso Higuita, L.G., Becerra, L.A., González, M.A., 2018. Genetic diversity of Ceiba pentandra in 697 Colombian seasonally dry tropical forest: Implications for conservation and management. Biol. Conserv. 698 227, 29–37. https://doi.org/10.1016/j.biocon.2018.08.021 699 Booth, T.H., 2016. Identifying particular areas for potential seed collections for restoration plantings under 700 climate change. Ecol. Manag. Restor. 17, 228–234. https://doi.org/10.1111/emr.12219 701 Bower, A.D., St. Clair, J.B., Erickson, V., 2014. Generalized provisional seed zones for native plants. Ecol. 702 Appl. 24, 913–919. https://doi.org/10.1890/13-0285.1 703 Brancalion, P.H.S., Rodrigues, R.R., Oliveira, G.C.X., 2015. When and how could common gardens be useful in 704 the ecological restoration of long-lived tropical plants as an aid to the selection of seed sources? Plant 705 Ecol. Divers. 8, 81–90. https://doi.org/10.1080/17550874.2013.879941 706 Brancalion, P.H.S., Viani, R.A.G., Aronson, J., Rodrigues, R.R., 2012. Improving Planting Stocks for the 707 Brazilian Atlantic Forest Restoration through Community-Based Seed Harvesting Strategies 20, 704–711. 708 https://doi.org/10.1111/j.1526-100X.2011.00839.x 709 Breed, M.F., Gardner, M.G., Ottewell, K.M., Navarro, C.M., Lowe, A.J., 2012. Shifts in reproductive assurance 710 strategies and inbreeding costs associated with habitat fragmentation in Central American mahogany. 711 Ecol. Lett. 15, 444–452. https://doi.org/10.1111/j.1461-0248.2012.01752.x 712 Breed, M.F., Harrison, P.A., Bischoff, A., Durruty, P., Gellie, N.J.C., Gonzales, E.K., Havens, K., Karmann, M., 713 Kilkenny, F.F., Krauss, S.L., Lowe, A.J., Marques, P., Nevill, P.G., Vitt, P.L., Bucharova, A., 2018. 714 Priority actions to improve provenance decision-making. Bioscience 68, 510–516. 715 https://doi.org/10.1093/biosci/biy050 716 Breed, M.F., Stead, M.G., Ottewell, K.M., Gardner, M.G., Lowe, A.J., 2013. Which provenance and where? 717 Seed sourcing strategies for revegetation in a changing environment. Conserv. Genet. 14, 1–10. 718 https://doi.org/10.1007/s10592-012-0425-z 719 Broadhurst, L.M., Lowe, A., Coates, D.J., Cunningham, S.A., McDonald, M., Vesk, P.A., Yates, C., 2008. Seed 720 supply for broadscale restoration: Maximizing evolutionary potential. Evol. Appl. 1, 587–597. 721 https://doi.org/10.1111/j.1752-4571.2008.00045.x 722 Broadmeadow, M.S.J., Ray, D., Samuel, C.J.A., 2005. Climate change and the future for broadleaved tree 723 species in Britain. Forestry 78, 145–161. https://doi.org/10.1093/forestry/cpi014 724 Bucharova, A., 2017. Assisted migration within species range ignores biotic interactions and lacks evidence. 725 Restor. Ecol. 25, 14–18. https://doi.org/10.1111/rec.12457 726 Bucharova, A., Bossdorf, O., Hölzel, N., Kollmann, J., Prasse, R., Durka, W., 2018. Mix and match: regional 727 admixture provenancing strikes a balance among different seed-sourcing strategies for ecological 728 restoration. Conserv. Genet. 20, 7–17. https://doi.org/10.1007/s10592-018-1067-6 729 Bullock, S.H., 1985. Breeding Systems in the Flora of a Tropical Deciduous Forest in Mexico. Biotropica 17, 730 287. https://doi.org/10.2307/2388591 731 Burnham, K.P., Anderson, D.R., 2004. Multimodel inference: Understanding AIC and BIC in model selection. 732 Sociol. Methods Res. 33, 261–304. https://doi.org/10.1177/0049124104268644 733 Castellanos-Acuña, D., Vance-Borland, K.W., St. Clair, J.B., Hamann, A., López-Upton, J., Gómez-Pineda, E., 734 Ortega-Rodríguez, J.M., Sáenz-Romero, C., 2018. Climate-based seed zones for Mexico: guiding 735 reforestation under observed and projected climate change. New For. 49, 297–309. 736 https://doi.org/10.1007/s11056-017-9620-6 737 Cerón-Souza, I., Toro-Perea, N., Cárdenas-Henao, H., 2005. Population genetic structure of neotropical 738 mangrove species on the Colombian Pacific Coast: Avicennia germinans (Avicenniaceae). Biotropica 37, 739 258–265. https://doi.org/10.1111/j.1744-7429.2005.00035.x 740 Chase, M.R., Boshier, D.H., Bawa, K.S., 1995. Population genetics of Cordia alliodora (Boraginaceae), a 741 neotropical tree. 1. Genetic variation in natural populations. Am. J. Bot. 82, 468–475. 742 https://doi.org/10.2307/2445693 743 Chaves, C.L., Sebbenn, A.M., Baranoski, A., Goez, B.D., Gaino, A.P.S.C., Ruas, C.F., Ruas, E., Ruas, P.M., 744 2016. Gene dispersal via seeds and pollen and their effects on genetic structure in the facultative- 745 apomictic Neotropical tree Aspidosperma polyneuron. Silvae Genet. 65, 46–57. 746 https://doi.org/10.1515/sg-2016-0016 747 Clarke, R.T., Rothery, P., Raybould, A.F., 2002. Confidence limits for regression relationships between distance 748 matrices: Estimating gene flow with distance. J. Agric. Biol. Environ. Stat. 7, 361–372. 749 https://doi.org/10.1198/108571102320 750 Crow, T.M., Albeke, S.E., Buerkle, C.A., Hufford, K.M., 2018. Provisional methods to guide species-specific 751 seed transfer in ecological restoration. Ecosphere 9. https://doi.org/10.1002/ecs2.2059 752 Crowe, K.A., Parker, W.H., 2008. Using portfolio theory to guide reforestation and restoration under climate 753 change scenarios. Clim. Change 89, 355–370. https://doi.org/10.1007/s10584-007-9373-x 754 Dick, C.W., Bermingham, E., Lemes, M.R., Gribel, R., 2007. Extreme long-distance dispersal of the lowland 755 tropical rainforest tree Ceiba pentandra L. (Malvaceae) in Africa and the Neotropics. Mol. Ecol. 16, 3039– 756 3049. https://doi.org/10.1111/j.1365-294X.2007.03341.x 757 Doyle, J. J. and Doyle, J.L., 1987. A rapid DNA isolation procedure for small quantities of fresh leaf tissue. 758 Phytochem. Bull. 19, 11–15. 759 Dray, S., Dufour, A.B., 2007. The ade4 package: Implementing the duality diagram for ecologists. J. Stat. 760 Softw. 22. https://doi.org/10.18637/jss.v022.i04 761 Duminil, J., Fineschi, S., Hampe, A., Jordano, P., Salvini, D., Vendramin, G.G., Petit, R.J., 2007. Can 762 population genetic structure be predicted from life-history traits? Am. Nat. 169. 763 https://doi.org/10.1086/513490 764 Durka, W., Michalski, S.G., Berendzen, K.W., Bossdorf, O., Bucharova, A., Hermann, J.M., Hölzel, N., 765 Kollmann, J., 2017. Genetic differentiation within multiple common grassland plants supports seed 766 transfer zones for ecological restoration. J. Appl. Ecol. 54, 116–126. https://doi.org/10.1111/1365- 767 2664.12636 768 Etter, A., McAlpine, C., Possingham, H., 2008. Historical patterns and drivers of landscape change in Colombia 769 since 1500: A regionalized spatial approach. Ann. Assoc. Am. Geogr. 98, 2–23. 770 https://doi.org/10.1080/00045600701733911 771 Excoffier, L., Smouse, P.E., Quattro, J.M., 1992. Analysis of molecular variance inferred from metric distances 772 among DNA haplotypes: Application to human mitochondrial DNA restriction data. Genetics 131, 479– 773 491. https://doi.org/10.5962/bhl.title.86657 774 FAO, 2020. Restoring the Earth - The next decade. Unasylva 71. 775 Fremout, T., Thomas, E., Gaisberger, H., Van Meerbeek, K., Muenchow, J., Briers, S., Gutierrez-Miranda, C.E., 776 Marcelo-Peña, J.L., Kindt, R., Atkinson, R., Cabrera, O., Espinosa, C.I., Aguirre-Mendoza, Z., Muys, B., 777 2020. Mapping tree species vulnerability to multiple threats as a guide to restoration and conservation of 778 tropical dry forests. Glob. Chang. Biol. 26. https://doi.org/10.1111/gcb.15028 779 Ganopoulos, I., Aravanopoulos, F.A., Argiriou, A., Kalivas, A., Tsaftaris, A., 2011. Is the genetic diversity of 780 small scattered forest tree populations at the southern limits of their range more prone to stochastic 781 events? A wild cherry case study by microsatellite-based markers. Tree Genet. Genomes 7, 1299–1313. 782 https://doi.org/10.1007/s11295-011-0414-2 783 García, Hernando, Corzo, G., Isaacs, P., Etter, A., 2014. Distribución y estado actual de los remanentes del 784 Bioma de Bosque Seco Tropical en Colombia: insumos para su gestión, in: Pizano, C., García, H (Eds.), 785 El Bosque Seco Tropical En Colombia. Instituto de Investigación de Recursos Biológicos Alexander von 786 Humboldt (IAvH), Bogotá, Colombia, pp. 228–251. 787 Gonzales, E., Hamrick, J.L., Smouse, P.E., Trapnell, D.W., Peakall, R., 2010. The impact of landscape 788 disturbance on spatial genetic structure in the guanacaste tree, Enterolobium cyclocarpum (Fabaceae). J. 789 Hered. 101, 133–143. https://doi.org/10.1093/jhered/esp101 790 González-M, R., García, H., Isaacs, P., Cuadros, H., López-Camacho, R., Rodríguez, N., Pérez, K., Mijares, F., 791 Castaño-Naranjo, A., Jurado, R.D., Idárraga-Piedrahíta, A., Rojas, A., Vergara, H., Pizano, C., 2018. 792 Disentangling the environmental heterogeneity, floristic distinctiveness and current threats of tropical dry 793 forests in Colombia. Environ. Res. Lett. 13, 045007. https://doi.org/10.1088/1748-9326/aaad74 794 Gray, L.K., Hamann, A., 2011. Strategies for reforestation under uncertain future climates: Guidelines for 795 Alberta, Canada. PLoS One 6, e22977. https://doi.org/10.1371/journal.pone.0022977 796 Guillot, G., Rousset, F., 2013. Dismantling the Mantel tests. Methods Ecol. Evol. 4, 336–344. 797 https://doi.org/10.1111/2041-210x.12018 798 Guttman, L., 1954. Some necessary conditions for common-factor analysis. Psychometrika 19, 149–161. 799 https://doi.org/10.1007/BF02289162 800 Hamann, A., Koshy, M.P., Namkoong, G., Ying, C.C., 2000. Genotype x environment interactions in Alnus 801 rubra: developing seed zones and seed-transfer guidelines with spatial statistics and GIS. For. Ecol. 802 Manage. 136, 107–119. 803 Hamrick, J.L., Godt, M.J.W., 1996. Effects of life history traits on genetic diversity in plant species. Philos. 804 Trans. R. Soc. B Biol. Sci. 351, 1291–1298. https://doi.org/10.1098/rstb.1996.0112 805 Hamrick, J.L., Murawski, D.A., Nason, J.D., 1993. The influence of seed dispersal mechanisms on the genetic 806 structure of tropical tree populations. Vegetatio 107, 281–297. https://doi.org/10.1007/BF00052230 807 Hancock, N., Hughes, L., 2014. Turning up the heat on the provenance debate: Testing the “local is best” 808 paradigm under heatwave conditions. Austral Ecol. 39, 600–611. https://doi.org/10.1111/aec.12122 809 Hargreaves, A.L., Germain, R.M., Bontrager, M., Persi, J., Angert, A.L., 2020. Local Adaptation to Biotic 810 Interactions : A Meta-analysis across Latitudes. Am. Nat. 195, 395–411. https://doi.org/10.1086/707323 811 Hargreaves, G.H., Allen, R.G., 2003. History and evaluation of Hargreaves evapotranspiration equation. J. Irrig. 812 Drain. Eng. 129, 53–63. https://doi.org/10.1061/(ASCE)0733-9437(2003)129:1(53) 813 Hargrove, W.W., Hoffman, F.M., 2005. Potential of multivariate quantitative methods for delineation and 814 visualization of ecoregions. Environ. Manage. 34, S39–S60. https://doi.org/10.1007/s00267-003-1084-0 815 Harrison, P.A., Vaillancourt, R.E., Harris, R.M.B., Potts, B.M., 2017. Integrating climate change and habitat 816 fragmentation to identify candidate seed sources for ecological restoration. Restor. Ecol. 25, 1–8. 817 https://doi.org/10.1111/rec.12488 818 Havens, K., Vitt, P., Still, S., Kramer, A.T., Fant, J.B., Schatz, K., 2015. Seed Sourcing for Restoration in an Era 819 of Climate Change. Nat. Areas J. 35, 122–133. https://doi.org/10.3375/043.035.0116 820 Hengl, T., Mendes de Jesus, J., Heuvelink, G.B.M., Ruiperez Gonzalez, M., Kilibarda, M., Blagotić, A., 821 Shangguan, W., Wright, M.N., Geng, X., Bauer-Marschallinger, B., Guevara, M.A., Vargas, R., 822 MacMillan, R.A., Batjes, N.H., Leenaars, J.G.B., Ribeiro, E., Wheeler, I., Mantel, S., Kempen, B., 2017. 823 SoilGrids250m: Global gridded soil information based on machine learning. PLoS One 12. 824 https://doi.org/10.1371/journal.pone.0169748 825 Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution interpolated 826 climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978. https://doi.org/10.1002/joc.1276 827 Hoekstra, J.M., Boucher, T.M., Ricketts, T.H., Roberts, C., 2005. Confronting a biome crisis: Global disparities 828 of habitat loss and protection. Ecol. Lett. 8, 23–29. https://doi.org/10.1111/j.1461-0248.2004.00686.x 829 Holderegger, R., Kamm, U., Gugerli, F., 2006. Adaptive vs. neutral genetic diversity: Implications for landscape 830 genetics. Landsc. Ecol. 21, 797–807. https://doi.org/10.1007/s10980-005-5245-9 831 Hornik, K., 2019. clue: Cluster ensembles. R package version 0.3-57. 832 Hufford, K.M., Mazer, S.J., 2003. Plant ecotypes: Genetic differentiation in the age of ecological restoration. 833 Trends Ecol. Evol. 18, 147–155. https://doi.org/10.1016/S0169-5347(03)00002-8 834 Hufford, K.M., Veneklaas, E.J., Lambers, H., Krauss, S.L., 2016. Genetic delineation of local provenance 835 defines seed collection zones along a climate gradient. AoB Plants 8, 1–12. 836 https://doi.org/10.1093/aobpla/plv149 837 Hurvich, C.M., Tsai, C.L., 1989. Regression and time series model selection in small samples. Biometrika. 838 https://doi.org/10.1093/biomet/76.2.297 839 Iriarte, J., Elliott, S., Maezumi, S.Y., Alves, D., Gonda, R., Robinson, M., Gregorio, J., Souza, D., Watling, J., 840 Handley, J., 2020. The origins of Amazonian landscapes : Plant cultivation , domestication and the spread 841 of food production in tropical South America. Quat. Sci. Rev. 248, 106582. 842 https://doi.org/10.1016/j.quascirev.2020.106582 843 Isaac-Renton, M., Montwé, D., Hamann, A., Spiecker, H., Cherubini, P., Treydte, K., 2018. Northern forest tree 844 populations are physiologically maladapted to drought. Nat. Commun. 9, 1–9. 845 https://doi.org/10.1038/s41467-018-07701-0 846 Jalonen, R., Valette, M., Boshier, D., Duminil, J., Thomas, E., 2018. Forest and landscape restoration severely 847 constrained by a lack of attention to the quantity and quality of tree seed: Insights from a global survey. 848 Conserv. Lett. 11, 1–9. https://doi.org/10.1111/conl.12424 849 James, T., Vege, S., Aldrich, P., Hamrick, J.L., 1998. Mating systems of three tropical dry forest tree species. 850 Biotropica 30, 587–594. https://doi.org/10.1111/j.1744-7429.1998.tb00098.x 851 Janzen, D.H., 1988. Tropical Dry Forests The Most Endangered Major Tropical Ecosystem, in: Wilson, E.O., 852 Peter, F.M. (Eds.), Biodiversity. National Academies Press, Washington D.C. 853 Janzen, D.H., 1982. Differential Seed Survival and Passage Rates in Cows and Horses, Surrogate Pleistocene 854 Dispersal Agents. Oikos 38, 150–156. https://doi.org/10.2307/3544014 855 Janzen, D.H., Martin, P.S., 1982. Neotropical anachronisms: The fruits the gomphotheres ate. Science (80-. ). 856 215, 19–27. https://doi.org/10.1126/science.215.4528.19 857 Jombart, T., Ahmed, I., 2011. adegenet 1.3-1: New tools for the analysis of genome-wide SNP data. 858 Bioinformatics. https://doi.org/10.1093/bioinformatics/btr521 859 Jorgensen, M.H., Elameen, A., Hofman, N., Klemsdal, S., Malaval, S., Fjellheim, S., 2016. What’s the meaning 860 of local? Using molecular markers to define seed transfer zones for ecological restoration in Norway. 861 Evol. Appl. 9, 673–684. https://doi.org/10.1111/eva.12378 862 Kaufman, L., Rousseeuw, P.J., 1990. Finding Groups in Data: An Introduction to Cluster Analysis. Wiley, New 863 York. https://doi.org/10.1007/s13398-014-0173-7.2 864 Knutti, R., Masson, D., Gettelman, A., 2013. Climate model genealogy: Generation CMIP5 and how we got 865 there. Geophys. Res. Lett. 40, 1194–1199. https://doi.org/10.1002/grl.50256 866 Kosman, E., Leonard, K.J., 2005. Similarity coefficients for molecular markers in studies of genetic 867 relationships between individuals for haploid, diploid, and polyploid species. Mol. Ecol. 14, 415–424. 868 https://doi.org/10.1111/j.1365-294X.2005.02416.x 869 Kramer, A.T., Havens, K., 2009. Plant conservation genetics in a changing world. Trends Plant Sci. 14, 599– 870 607. https://doi.org/10.1016/j.tplants.2009.08.005 871 Kramer, A.T., Larkin, D.J., Fant, J.B., 2015. Assessing Potential Seed Transfer Zones for Five Forb Species 872 from the Great Basin Floristic Region, USA. Nat. Areas J. 35, 174–188. 873 https://doi.org/10.3375/043.035.0119 874 Krauss, S.L., He, T.H., 2006. Rapid genetic identification of local provenance seed collection zones for 875 ecological restoration and biodiversity conservation. J. Nat. Conserv. 14, 190–199. 876 https://doi.org/10.1016/j.jnc.2006.05.002 877 Krauss, S.L., Sinclair, E.A., Bussell, J.D., Hobbs, R.J., 2013. An ecological genetic delineation of local seed- 878 source provenance for ecological restoration. Ecol. Evol. 3, 2138–2149. https://doi.org/10.1002/ece3.595 879 Kreyling, J., Buhk, C., Backhaus, S., Hallinger, M., Huber, G., Huber, L., Jentsch, A., Konnert, M., Thiel, D., 880 Wilmking, M., Beierkuhnlein, C., 2014. Local adaptations to frost in marginal and central populations of 881 the dominant forest tree Fagus sylvatica L. as affected by temperature and extreme drought in common 882 garden experiments. Ecol. Evol. 4, 594–605. https://doi.org/10.1002/ece3.971 883 Lacerda, D.R., Acedo, M.D.P., Lemos Filho, J.P., Lovato, M.B., 2001. Genetic diversity and structure of natural 884 populations of Plathymenia reticulata (Mimosoideae), a tropical tree from the Brazilian Cerrado. Mol. 885 Ecol. 10, 1143–1152. https://doi.org/10.1046/j.1365-294X.2001.01264.x 886 Lefèvre, F., Boivin, T., Bontemps, A., Courbet, F., Davi, H., Durand-Gillmann, M., Fady, B., Gauzere, J., 887 Gidoin, C., Karam, M.J., Lalagüe, H., Oddou-Muratorio, S., Pichot, C., 2013. Considering evolutionary 888 processes in adaptive forestry. Ann. For. Sci. 71, 723–739. https://doi.org/10.1007/s13595-013-0272-1 889 Legendre, P., Fortin, M.J., Borcard, D., 2015. Should the Mantel test be used in spatial analysis? Methods Ecol. 890 Evol. 6, 1239–1247. https://doi.org/10.1111/2041-210X.12425 891 Leinonen, T., O’Hara, R.B., Cano, J.M., Merilä, J., 2008. Comparative studies of quantitative trait and neutral 892 marker divergence: A meta-analysis. J. Evol. Biol. 21, 1–17. https://doi.org/10.1111/j.1420- 893 9101.2007.01445.x 894 Li, X., Fu, X., Shang, X., Yang, W., Fang, S., 2017. Natural population structure and genetic differentiation for 895 heterodicogamous plant: Cyclocarya paliurus (Batal.) Iljinskaja (Juglandaceae). Tree Genet. Genomes 13. 896 https://doi.org/10.1007/s11295-017-1157-5 897 Listl, D., Poschlod, P., Reisch, C., 2018. Do seed transfer zones for ecological restoration reflect the spatial 898 genetic variation of the common grassland species Lathyrus pratensis? Restor. Ecol. 26, 667–676. 899 https://doi.org/10.1111/rec.12613 900 Lowe, A.J., Breed, M.F., Caron, H., Colpaert, N., Dick, C., Finegan, B., Gardner, M., Gheysen, G., Gribel, R., 901 Harris, J.B.C., Kremer, A., Lemes, M.R., Margis, R., Navarro, C.M., Salgueiro, F., Villalobos-Barrantes, 902 H.M., Cavers, S., 2018. Standardized genetic diversity-life history correlates for improved genetic 903 resource management of Neotropical trees. Divers. Distrib. 24, 730–741. 904 https://doi.org/10.1111/ddi.12716 905 Lowe, A.J., Jourde, B., Breyne, P., Colpaert, N., Navarro, C., Wilson, J., Cavers, S., 2003. Fine-scale genetic 906 structure and gene flow within Costa Rican populations of mahogany (Swietenia macrophylla). Heredity 907 (Edinb). 90, 268–275. https://doi.org/10.1038/sj.hdy.6800247 908 Malaval, S., Lauga, B., Regnault-Roger, C., Largier, G., 2010. Combined definition of seed transfer guidelines 909 for ecological restoration in the French Pyrenees. Appl. Veg. Sci. 13, 113–124. 910 https://doi.org/10.1111/j.1654-109X.2009.01055.x 911 Massatti, R., Shriver, R.K., Winkler, D.E., Richardson, B.A., Bradford, J.B., 2020. Assessment of population 912 genetics and climatic variability can refine climate-informed seed transfer guidelines. Restor. Ecol. 28, 913 485–493. https://doi.org/10.1111/rec.13142 914 McKay, J.K., Christian, C.E., Harrison, S., Rice, K.J., 2005. “How Local Is Local?”A Review of Practical and 915 Conceptual Issues in the Genetics of Restoration. Restor. Ecol. 13, 432–440. 916 https://doi.org/10.1111/j.1526-100X.2005.00058.x 917 Merilä, J., Crnokrak, P., 2001. Comparison of genetic differentiation at marker loci and quantitative traits. J. 918 Evol. Biol. 14, 892–903. https://doi.org/10.1046/j.1420-9101.2001.00348.x 919 Michalski, S.G., Durka, W., 2012. Assessment of provenance delineation by genetic differentiation patterns and 920 estimates of gene flow in the common grassland plant Geranium pratense. Conserv. Genet. 13, 581–592. 921 https://doi.org/10.1007/s10592-011-0309-7 922 Miller, S.A., Bartow, A., Gisler, M., Ward, K., Young, A.S., Kaye, T.N., 2011. Can an Ecoregion Serve as a 923 Seed Transfer Zone? Evidence from a Common Garden Study with Five Native Species. Restor. Ecol. 19, 924 268–276. https://doi.org/10.1111/j.1526-100X.2010.00702.x 925 Montalvo, A.M., Ellstrand, N.C., 2001. Nonlocal transplantation and outbreeding depression in the subshrub 926 Lotus scoparius (Fabaceae). Am. J. Bot. 88, 258–269. https://doi.org/10.2307/2657017 927 Morillo-Paz, A.M., 2019. Distribución de la diversidad genética de Diomate (Astronium graveolens Jacq) en el 928 bosque seco tropical de Colombia. MSc thesis. Facultad de Medio Ambiente y Recursos Naturales, 929 Universidad Distrital Francisco José de Caldas. 930 Murawski, D.A., Hamrick, J.L., 1992. Mating system and phenology of Ceiba pentandra (Bombacaeae) in 931 Central Panama. J. Hered. 83, 401–404. https://doi.org/10.1093/oxfordjournals.jhered.a111241 932 Murphy, P.G., Lugo, a E., 1986. Ecology of Tropical Dry Forest. Annu. Rev. Ecol. Syst. 17, 67–88. 933 https://doi.org/https://doi.org/10.1146/annurev.es.17.110186.000435 934 Nakagawa, S., Johnson, P.C.D., Schielzeth, H., 2017. The coefficient of determination R2 and intra-class 935 correlation coefficient from generalized linear mixed-effects models revisited and expanded. J. R. Soc. 936 Interface 14, 20170213. https://doi.org/10.1098/rsif.2017.0213 937 Nelson, M.F., Anderson, N.O., 2013. How many marker loci are necessary? Analysis of dominant marker data 938 sets using two popular population genetic algorithms. Ecol. Evol. 3, 3455–3470. 939 https://doi.org/10.1002/ece3.725 940 Norden, N., González‐M., R., Avella‐M., A., Salgado‐Negret, B., Alcázar, C., Rodríguez‐Buriticá, S., Aguilar‐ 941 Cano, J.R., Castellanos‐Castro, C., Calderón, J.J., Caycedo‐Rosales, P., Cuadros, H., Díaz‐Pulido, A., 942 Fajardo, Z., Franke‐Ante, R., García, D.H., González, M.A., Hernández‐Jaramillo, A., Idárraga‐Piedrahita, 943 Á., López‐Camacho, R., Martínez‐Callejas, S.J., Nieto, J., Pizano, C., Rodríguez, G., Torres, A.M., 944 Vergara, H., García, H., 2020. Building a socio‐ecological monitoring platform for the comprehensive 945 management of tropical dry forests. Plants, People, Planet. https://doi.org/10.1002/ppp3.10113 946 Nosil, P., Funk, D.J., Ortiz-Barrientos, D., 2009. Divergent selection and heterogeneous genomic divergence. 947 Mol. Ecol. 18, 375–402. https://doi.org/10.1111/j.1365-294X.2008.03946.x 948 Novaes, R.M.L., Rodrigues, J.G., Lovato, M.B., 2009. An efficient protocol for tissue sampling and DNA 949 isolation from the stem bark of Leguminosae trees. Genet. Mol. Res. 8, 86–96. 950 https://doi.org/10.4238/vol8-1gmr542 951 Nybom, H., 2004. Comparison of different nuclear DNA markers for estimating intraspecific genetic diversity 952 in plants. Mol. Ecol. 13, 1143–1155. https://doi.org/10.1111/j.1365-294X.2004.02141.x 953 Pedrini, S., Dixon, K.W., 2020. International principles and standards for native seeds in restoration. Restor. 954 Ecol. 28, S286–S303. https://doi.org/10.1111/rec.13155 955 Pennington, R.T., Lehmann, C.E.R., Rowland, L.M., 2018. Tropical savannas and dry forests. Curr. Biol. 28, 956 541–545. https://doi.org/10.1016/j.cub.2006.11.063 957 Pennington, R.T., Pradot, D.E., Pendry, C.A., 2000. Neotropical seasonally dry forests and Quaternary 958 vegetation changes 27, 261–273. https://doi.org/https://doi.org/10.1046/j.1365-2699.2000.00397.x 959 Pizano, C., González-M., R., López, R., Jurado, R.D., Cuadros, H., Castaño-Naranjo, A., Rojas, A., Peréz, K., 960 Vergara-Varela, H., Idárraga, Á., Isaacs, P., García, H., 2016. El bosque seco tropical en Colombia, in: 961 Biodiversidad 2015. Estado y Tendencias de La Biodiversidad Continental de Colombia. 962 https://doi.org/10.21068/B001.2015.202 963 Portillo-Quintero, C.A., Sánchez-Azofeifa, G.A., 2010. Extent and conservation of tropical dry forests in the 964 Americas. Biol. Conserv. 143, 144–155. https://doi.org/10.1016/j.biocon.2009.09.020 965 Potter, K.M., Hargrove, W.W., 2012. Determining suitable locations for seed transfer under climate change: A 966 global quantitative method. New For. 43, 581–599. https://doi.org/10.1007/s11056-012-9322-z 967 Prober, S.M., Byrne, M., McLean, E.H., Steane, D.A., Potts, B.M., Vaillancourt, R.E., Stock, W.D., 2015. 968 Climate-adjusted provenancing: a strategy for climate-resilient ecological restoration. Front. Ecol. Evol. 3, 969 65. https://doi.org/10.3389/fevo.2015.00065 970 Ramalho, C.E., Byrne, M., Yates, C.J., 2017. A climate-oriented approach to support decision-making for seed 971 provenance in ecological restoration. Front. Ecol. Evol. 5, 95. https://doi.org/10.3389/fevo.2017.00095 972 Rawat, A., Barthwal, S., Ginwal, H.S., 2014. Comparative assessment of SSR, ISSR and AFLP markers for 973 characterization of selected genotypes of Himalayan Chir pine (Pinus roxburghii Sarg.) based on resin 974 yield. Silvae Genet. 63, 94–109. https://doi.org/10.1515/sg-2014-0013 975 Reisch, C., Bernhardt-Römermann, M., 2014. The impact of study design and life history traits on genetic 976 variation of plants determined with AFLPs. Plant Ecol. 215, 1493–1511. https://doi.org/10.1007/s11258- 977 014-0409-9 978 Rodríguez-Buritacá, S., Corzo, G., García, H., Córdoba, D., Isaacs, P., Etter, A., 2016. Haciendo visible lo 979 invisible., in: Gomez, M.F., Moreno, L.A., Andrade, G.I., Rueda, C. (Eds.), Estado y Tendencias de La 980 Biodiversidad Continental de Colombia. Instituto Alexander Von Humboldt, Bogotá, Colombia. 981 Sanchez-Gomez, K.F., Cristóbal-Pérez, E.J., Harvey, N., Quesada, M., 2020. Isolation and characterization of 982 microsatellite loci in Astronium graveolens (Anacardiaceae) and cross amplification in related species. 983 Mol. Biol. Rep. 47, 4003–4007. https://doi.org/10.1007/s11033-020-05397-4 984 Schlaepfer, D.R., Bradford, J.B., Lauenroth, W.K., Munson, S.M., Tietjen, B., Hall, S.A., Wilson, S.D., 985 Duniway, M.C., Jia, G., Pyke, D.A., Lkhagva, A., Jamiyansharav, K., 2017. Climate change reduces 986 extent of temperate drylands and intensifies drought in deep soils. Nat. Commun. 8, 1–9. 987 https://doi.org/10.1038/ncomms14196 988 Schmidt, I.B., de Urzedo, D.I., Piña-Rodrigues, F.C.M., Vieira, D.L.M., de Rezende, G.M., Sampaio, A.B., 989 Junqueira, R.G.P., 2019. Community-based native seed production for restoration in Brazil – the role of 990 science and policy. Plant Biol. 21, 389–397. https://doi.org/10.1111/plb.12842 991 Schröder, J.M., Ávila-Rodríguez, L.P., Günter, S., 2021. Tropical dry forests: The neglected research agenda ? 992 For. Policy Econ. 122. https://doi.org/10.1016/j.forpol.2020.102333 993 Sexton, J.P., Hangartner, S.B., Hoffmann, A.A., 2014. Genetic isolation by environment or distance: Which 994 pattern of gene flow is most common? Evolution (N. Y). 68, 1–15. https://doi.org/10.1111/evo.12258 995 Sgrò, C.M., Lowe, A.J., Hoffmann, A.A., Sgro, C.M., 2011. Building evolutionary resilience for conserving 996 biodiversity under climate change. Evol. Appl. 4, 326–337. https://doi.org/10.1111/j.1752- 997 4571.2010.00157.x 998 Shirk, A.J., Landguth, E.L., Cushman, S.A., 2018. A comparison of regression methods for model selection in 999 individual-based landscape genetic analysis. Mol. Ecol. Resour. 18, 55–67. https://doi.org/10.1111/1755- 1000 0998.12709 1001 Shryock, D.F., DeFalco, L.A., Esque, T.C., 2018. Spatial decision-support tools to guide restoration and seed- 1002 sourcing in the Desert Southwest. Ecosphere 9, e02453. https://doi.org/10.1002/ecs2.2453 1003 Shryock, D.F., Havrilla, C.A., DeFalco, L.A., Esque, T.C., Custer, N.A., Wood, T.E., 2017. Landscape genetic 1004 approaches to guide native plant restoration in the Mojave Desert. Ecol. Appl. 27, 429–445. 1005 https://doi.org/10.1002/eap.1447 1006 Soliani, C., Umaña, F., Mondino, V., Thomas, E., Pastorino, M., Gallo, L., Marchelli, P., 2017. Zonas genéticas 1007 de Lenga y Ñire en Argentina y su aplicación en la conservación y manejo de los recusros forestales. 1008 INTA Ediciones, Bariloche, Argentina. 1009 St. Clair, J.B., 2006. Genetic variation in fall cold hardiness in coastal Douglas-fir in western Oregon and 1010 Washington. Can. J. Bot. 84, 1110–1121. https://doi.org/10.1139/b06-084 1011 Suding, K., Higgs, E., Palmer, M., Callicott, J.B., Anderson, C.B., Baker, M., Gutrich, J.J., Hondula, K.L., 1012 LaFevor, M.C., Larson, B.M.H., Randall, A., Ruhl, J.B., Schwartz, K.Z.S., 2015. Committing to 1013 ecological restoration. Science (80-. ). 348, 638–640. https://doi.org/10.1126/science.aaa4216 1014 Thomas, E., Alcazar, C., Moscoso H., L.G., Vasquez, A., Osorio, L.F., Salgado-Negret, B., Gonzalez, M., 1015 Bozzano, M., Loo, J., Jalonen, R., Ramirez, W., 2017. The importance of species selection and seed 1016 sourcing in forest restoration for enhancing adaptive potential to climate change: Colombian tropical dry 1017 forest as a model. CBD Tech. Ser. Biodivers. Clim. Chang. 122–132. 1018 Thomas, E., Gil Tobón, C., Gutiérrez, J.P., Alcázar Caicedo, C., Moscoso Higuita, L.G., Becerra, L.A., Loo, J., 1019 González, M.A., 2016. Genetic diversity of Enterolobium cyclocarpum in Colombian seasonally dry 1020 tropical forest: implications for conservation and restoration. Biodivers. Conserv. 26, 825–842. 1021 https://doi.org/10.1007/s10531-016-1274-8 1022 Thomas, E., Jalonen, R., Loo, J., Boshier, D., Gallo, L., Cavers, S., Bordács, S., Smith, P., Bozzano, M., 2014. 1023 Genetic considerations in ecosystem restoration using native tree species. For. Ecol. Manage. 333, 66–75. 1024 https://doi.org/10.1016/j.foreco.2014.07.015 1025 Thomson, A.M., Crowe, K.A., Parker, W.H., 2010. Optimal white spruce breeding zones for Ontario under 1026 current and future climates. Can. J. For. Res. 40, 1576–1587. https://doi.org/10.1139/X10-112 1027 Vander Mijnsbrugge, K., Bischoff, A., Smith, B., 2010. A question of origin: Where and how to collect seed for 1028 ecological restoration. Basic Appl. Ecol. 11, 300–311. https://doi.org/10.1016/j.baae.2009.09.002 1029 Vargas, W., Ramírez, W., 2014. Lineamientos generales para la restauración del bosque seco tropical en 1030 Colombia, in: El Bosque Seco Tropical En Colombia. https://doi.org/10.1017/CBO9781107415324.004 1031 Vitasse, Y., Delzon, S., Bresson, C.C., Michalet, R., Kremer, A., 2009. Altitudinal differentiation in growth and 1032 phenology among populations of temperate-zone tree species growing in a common garden. Can. J. For. 1033 Res. 39. https://doi.org/10.1139/x09-054 1034 Vitt, P., Havens, K., Kramer, A.T., Sollenberger, D., Yates, E., 2010. Assisted migration of plants: Changes in 1035 latitudes, changes in attitudes. Biol. Conserv. 143, 18–27. https://doi.org/10.1016/j.biocon.2009.08.015 1036 Vranckx, G., Jacquemyn, H., Muys, B., Honnay, O., 2012. Meta-Analysis of Susceptibility of Woody Plants to 1037 Loss of Genetic Diversity through Habitat Fragmentation. Conserv. Biol. 26, 228–237. 1038 https://doi.org/10.1111/j.1523-1739.2011.01778.x 1039 Winter, D.J., 2012. mmod: an R library for the calculation of population differentiation statistics. Mol. Ecol. 1040 Resour. 12, 1158–1160. https://doi.org/https://doi.org/10.1111/j.1755-0998.2012.03174.x 1041 Wright, S., 1949. The genetical structure of populations. Ann. Eugen. 15, 323–354. 1042 https://doi.org/10.1111/j.1469-1809.1949.tb02451.x 1043 Zizumbo-Villarreal, D., Colunga-GarcíaMarín, P., Flores-Silva, A., 2016. Pre-Columbian Food System in West 1044 Mesoamerica, in: Lira, R., Casas, A., Blancas, J. (Eds.), Ethnobotany of Mexico: Interactions of People 1045 and Plants in Mesoamerica. Springer New York, New York, NY, pp. 67–82. https://doi.org/10.1007/978- 1046 1-4614-6669-7_4 1047