Integrating Genomic and Climate Data to Design Representative Seed Production Areas: A Pragmatic Workflow for Climate-Adjusted Provenancing

Abstract

Establishing genetically diverse ex situ collections, particularly seed production areas (SPAs), is essential not only for safeguarding biodiversity but also for generating high-quality and high-quantity germplasm material. However, practical tools for sourcing genetically representative material remain limited, especially for widespread, common species. Here, we present a flexible, data-driven workflow that integrates genomic data, future climate projections and real-world constraints to guide the design of representative SPAs. Using the widespread rainforest tree Neolitsea dealbata as a case study, we identified genetic neighbourhoods (GNs) across its range and used a climate-matching tool to pinpoint an external GN with a future climate analogous to a target restoration area (the Big Scrub). We evaluated how common allelic diversity is captured under three practitioner-defined decisions: (1) whether to minimise individuals or sites sampled, (2) whether to apply sampling constraints and (3) whether to sample randomly or optimally. To support the third decision, we developed a novel optimisation method that identifies combinations of individuals or sites using a down-projected site frequency spectrum (psfs), aiming to maximise allele representation in the final collection. These decisions were then implemented across three provenancing strategies: local, predictive and climate-adjusted. Our results show that multiple sampling approaches can capture over 90% of common alleles (a predefined threshold) for the local GN, even under various logistical and practical constraints. The same is feasible when including future climate-matched sources from an external GN, which nearly doubled allelic representation of the species in the SPA. This workflow is adaptable to practical limitations, such as site inaccessibility or reliance on existing collections. By balancing genetic resolution with practitioner flexibility, our approach supports scalable, evidence-based design of ex situ collections, such as SPAs, to maximise genetic representation under environmental change.

Keywords: climate‐adjusted provenancing; ex situ conservation; genetic diversity; sampling optimisation; seed production areas; seed sourcing.

FIGURE 1

Series of maps to illustrate (A) The distribution of Neolitsea dealbata (black dots); occurrence data from the Atlas of Living Australia (ALA 2025) and the sampling sites for genetic analysis, with each estimated GN indicated in a different colour. (B) The areas (yellow) currently experiencing the climate that the Target Restoration Area (green; Big Scrub) is predicted to experience in 2070 based on modelling using the climate‐matching tool. (C) The external Genetic Neighbourhood (external GN) in closest proximity to the Local Genetic Neighbourhood (local GN) of the Target Restoration Area where the sites that overlap with the modelled future climate occur within the yellow areas. (D) The extent of the local GN of the Target Restoration Area (green area). Further analyses supporting the identification of GNs presented here are outlined in Figure 3 and Figures S5 and S6 in Appendix 2.

FIGURE 2

Workflow for establishing a genetically representative germplasm collection across three provenancing strategies, outlining the key decision‐making steps guided by practitioner input to accommodate logistical challenges and sampling constraints. Processes coloured blue involves targeting the identified local genetic neighbourhood (local GN), while red involves the identified external GN.

FIGURE 3

Population genetic analyses highlighting the genetic neighbourhoods (GNs) across the extant distribution of Neolitsea dealbata. (A) Splitstree network with nodes coloured by GN. (B) LEA sNMF pie chart and bar plots of K = 5, with a zone of admixture highlighted (pink arrow). (C) Principal component analysis (PCA) of individuals coloured by GN. (D) Isolation by distance (IBD) plot coloured by pairwise comparisons between GNs (coral) and within GNs (teal). (including the zone of admixture), with significant mantel statistic provided.

FIGURE 4

Proportion of common (black) and rare (grey) alleles captured across randomised sampling combinations for both the local GN (blue border; A and B) and unique alleles found in the external GN (red border; C and D). Sampling approaches reflect alternative practitioner‐led decisions: Decision 1: Minimising either the total number of individuals (A and C) or the total number of sites (B and D); Decision 2: Filled boxplots represent unconstrained sampling, while unfilled boxplots represent constrained sampling (see constraints in Table S2 in Appendix 3); Decision 3: For each case, the random sampling combination capturing > 90% of common alleles with the smallest number of individuals or sites is indicated with an arrow. Allele proportions of optimised combinations are shown in orange or purple, either as fixed lines or as boxplots when variation arises from which individuals are represented for site combinations.