Modeling Phenotypic Trait Variation and Plasticity in Elymus elymoides to Guide Climate‐Informed Seed Transfer

TL;DR

This study models how climate affects traits in a grass species to improve seed transfer for restoration in the western USA.

Contribution

A novel hierarchical modeling approach captures complex trait-climate associations and practical seed transfer constraints.

Findings

Seed zones were divided into three major zones with 13 subzones, reflecting distinct trait-climate relationships.

Populations from milder climates showed higher trait plasticity, except for seed maturation in warmer, drier climates.

The modeling approach improves precision by capturing nuanced trait-climate associations missed by standard methods.

Abstract

Information on climate‐associated phenotypic variation is essential for sourcing seed that matches restoration site conditions. Spatially explicit seed transfer models can effectively deliver this information. However, standard modeling approaches often do not provide flexibility for practical considerations and may not capture highly complex trait‐climate associations. We characterized climate‐associated variation in growth, reproduction, morphology, phenology, and survival across 98 source populations at 3 common gardens for the grass
Elymus elymoides
(bottlebrush squirreltail), an important restoration species in the Intermountain Region of the western USA. We developed fixed‐boundary seed zones and focal‐point seed transfer models using non‐standard methods (regression trees and random forests). In general, source populations with larger plant sizes and later flowering originated from cooler and wetter or milder climates than those with smaller sizes and earlier flowering, though some associations were more complex. Populations from milder climates also had higher trait plasticity than populations from other climates, except for plasticity in seed maturation, which was highest in populations from warmer and drier climates. Seed zones identified through our approach consisted of three major zones with 2–7 subzones each (13 seed zones in total). Two subspecies groups had distinct trait‐climate associations, and separate seed zone models were developed for each. Our modeling approach provides a hierarchical structure that partitions predictor variables based on their importance. This doubles as a prioritization framework that assists in navigating trade‐offs between risk avoidance and practical constraints by explicitly defining how zones can be combined or subdivided in response to user needs. Our approach also captures trait‐climate association nuances missed by standard approaches, increasing the precision of our focal‐point seed transfer zones. Our findings emphasize the multifaceted nature of trait‐climate associations and highlight the importance of seed transfer modeling to seed‐sourcing decisions in a time of global change.

Figures