Introduction
Plant genetic resources constitute the foundation of global food security, providing the raw material for crop improvement and adaptation to evolving environmental conditions. Phenotypic characterization is central to identifying useful sources of germplasm for breeding programmes, yet a critical conceptual limitation persists: the environments in which genotypes are evaluated rarely reflect the ecological conditions under which they evolved or are best adapted. This challenge is particularly acute in global plant germplasm collections, where accessions originate from diverse agro-ecological zones but are often assessed under a narrow set of experimental conditions. Recent analyses highlight that climate-smart breeding requires aligning genotypes with their agro-climatic zones, reinforcing the risks of evaluating diverse germplasm under uniform or standardized growing conditions (Garcia-Oliveira et al. Reference Garcia-Oliveira, Ortiz, Sarsu, Rasmussen, Agre, Asfaw, Kante and Chander2025). Phenotyping platforms, whether growth chambers, glasshouses or limited-location field trials, prioritize environmental uniformity to enhance reproducibility. While this reduces experimental noise, it introduces a more subtle but profound bias: environmental mismatch between genotype origin and evaluation conditions (Figure 1A).
Conceptual illustration of environmental mismatch in phenotypic evaluation of plant genetic resources. Germplasm collected from diverse agro-ecological zones often having adaptations to the collection site growing conditions (A). Genotypes adapted to contrasting environments are often screened under a same or standardized environment. Consequently, phenotypic rankings may reflect adaptation to the testing environment rather than genetic potential (B). Environment-aware phenotyping frameworks are needed to better capture adaptive diversity and identify germplasm relevant for climate-resilient crop improvement (C). The crop (here, five maize genotypes) and rankings are used as examples in this figure. The sources of environmental mismatch illustrated here are temperature, photoperiod, rainfall and moisture regimes, soil type and nutrients, radiation, natural selection and local management practices. The figure was created using BioRender (www.biorender.com).
Figure 1 Long description
The illustration is divided into three sections. Section A: It shows plant genetic resources originating from diverse agro-ecological zones with local adaptation. Each genotype (G) is best adapted to its collection site, labeled as G1, G2, G3, G4 and G5. Section B: It depicts the shift to a uniform or standardized evaluation environment, such as a greenhouse or limited field site. The observed performance in this uniform environment is shown with biased ranking for genotypes G1 to G5. A graph illustrates genotype by environment (G x E) interactions, showing performance across environments E1, E2, E3, to En. Section C: It focuses on environmental-aware phenotyping of plant genetic resources. It includes multi-environment trials, environmental metadata integration, reaction norms and plasticity analysis, international phenotyping networks and genomics, environmental and crop modeling.
The importance of genotype by environment (G × E) interactions in shaping phenotypic expression is well established (Cooper et al. Reference Cooper, Stucker, DeLacy and Harch1997; Des Marais et al. Reference Des Marais, Hernandez and Juenger2013). Ecological genomics further demonstrates that adaptive variation is often environment-specific and linked to local selective pressures (Lasky et al. Reference Lasky, Des Marais, Lowry, Povolotskaya, McKay, Richards, Keitt and Juenger2014; Exposito-Alonso et al. Reference Exposito-Alonso, Vasseur, Ding, Wang, Burbano and Weigel2018). Despite both, phenotyping assessments frequently assume that performance in limited test environments reflects general genetic value, an assumption that risks underestimating adaptive diversity (Figure 1B), particularly in crop wild relatives and landraces. It is important to note that the extent to which environmental mismatch affects trait evaluation depends on the genetic architecture of the trait. For qualitative traits controlled by major gene(s), such as quality traits, phenotypic evaluation tends to be more stable across environments and environmental mismatch is less of a concern. But for quantitative traits such as adaptation to abiotic or resistance to biotic stresses and yield with low heritability and strong G × E interactions, environmental mismatch is a major concern. Phenological traits such as flowering and maturity times are also particularly sensitive to environmental mismatch, as they are strongly regulated by photoperiod and temperature.
The importance of environmental mismatch in plant genetic resource evaluation is well recognized by plant breeders and crop scientists but remains frequently overlooked, particularly in genomics-centred approaches. Recent advances in high-throughput plant phenotyping platforms, crop genomes and pan-genomes, and data integration have greatly expanded the scale of plant genetic resource evaluation, yet most phenotypic screening still relies on limited-environment testing. Additionally, our understanding of G × E interactions and ecological genomics has made the consequences of environmental standardization more explicit and quantifiable than in earlier work. This combination of increased analytical capacity and stronger empirical evidence has sharpened awareness that current evaluation frameworks may systematically bias the assessment of adaptive variation. This paper aims to highlight environmental mismatch in phenotypic evaluation and outlines environment-aware approaches to better capture adaptive trait diversity for crop improvement.
Controlled environments: Precision without ecological realism
Controlled-environment phenotyping platforms enable precise manipulation of environmental variables and high-throughput and precise data collection. However, their assumed environmental uniformity also constitutes a major limitation. Standardized environmental growing conditions enable meaningful comparisons among genotypes. However, by applying the same test environment across all genotypes, these systems implicitly assume that all genotypes respond optimally under the same settings, an assumption that is rarely valid for diverse germplasm, particularly for traits with strong G × E interactions. Genotypes from distinct ecological regions are adapted to specific combinations of photoperiod, temperature regimes, radiation intensity and soil characteristics. When evaluated under generic controlled conditions, such genotypes may exhibit reduced performance due to environmental maladaptation rather than genetic potential.
Controlled environments often fail to reproduce the ecological complexity of field conditions (Poorter et al. Reference Poorter, Fiorani, Stitt, Schurr, Finck, Gibon, Usadel, Munns, Atkin, Tardieu and Pons2012; Langstroff et al. Reference Langstroff, Heuermann, Stahl and Junker2022). For example, previous studies have shown that phenotypes measured in controlled settings frequently diverge from those expressed in natural environments (Passioura Reference Passioura2006; Tardieu et al. Reference Tardieu, Simonneau and Muller2018; Gomulkiewicz and Stinchcombe Reference Gomulkiewicz and Stinchcombe2022), especially for stress-adaptive traits (Fiorani and Schurr Reference Fiorani and Schurr2013). Thus, while controlled environments provide precision, they may lack ecological validity, an important distinction when evaluating genetic resources intended for diverse agricultural systems.
Field trials: Ecological relevance with inherent bias
Field-based phenotyping is often considered the gold standard due to its ecological relevance. However, field trials are not immune to environmental mismatch. Large germplasm panels are typically evaluated at a limited number of locations – sometimes for logistical convenience rather than ecological representativeness. As a result, only genotypes adapted to the local environment are assessed under near-optimal conditions, while other genotypes, particularly those from non-related environments, are evaluated under suboptimal conditions. This leads to location-specific selection bias, where genotype performance reflects environmental compatibility rather than intrinsic genetic potential.
Multi-environment trials (METs) consistently reveal substantial crossover G × E interactions (Cooper et al. Reference Cooper, Stucker, DeLacy and Harch1997; Crossa et al. Reference Crossa, Pérez-Rodríguez, Cuevas, Montesinos-López, Jarquín, De Los Campos, Burgueño, González-Camacho, Pérez-Elizalde, Beyene, Dreisigacker, Singh, Zhang, Gowda, Roorkiwal, Rutkoski and Varshney2017). In crops, genotypes frequently change rank across environments, especially for drought, heat and nutrient-use efficiency traits (Varshney et al. Reference Varshney, Thudi, Nayak, Gaur, Kashiwagi, Krishnamurthy, Jaganathan, Koppolu, Bohra, Tripathi, Rathore, Jukanti, Jayalakshmi, Vemula, Singh, Yasin, Sheshshayee and Viswanatha2014; Annicchiarico et al. Reference Annicchiarico, Nazzicari, Pecetti, Romani, Ferrari, Wei and Brummer2017). Consequently, single- or limited-test environment trials risk favouring broadly adapted ‘generalist’ genotypes while overlooking specialized adaptations critical for resilience under specific stress conditions. It should be noted, however, that single test environment trials conducted over multiple years may capture environmental variation, as year-to-year fluctuations in biotic and abiotic pressure at a given site may reveal different growing conditions. Yet they often remain geographically constrained and may fail to represent the broader range of agro-ecological conditions from which diverse germplasm originates.
Environmental mismatch and plant genetic resource utilization
The effects of controlled-environment and field-based biases have important implications, summarized in Table 1. Of these, the misclassification of genotypes and the narrowing of the genetic base (Allaby et al. Reference Allaby, Ware and Kistler2018) are perhaps the most consequential for long-term breeding progress. Once valuable diversity is overlooked in early screening, it rarely re-enters breeding pipelines, and the cumulative effect across generations can permanently reduce the adaptive potential available to plant breeders. This risk is particularly acute for crop wild relatives and landraces, which harbour specialized adaptations that broadly adapted elite lines cannot provide, yet are most likely to be undervalued under standardized evaluation conditions.
Potential sources of bias due to environmental mismatch in phenotypic evaluation of plant genetic resources and potential solutions. References are provided for each row to support the concepts summarized in the table
Table 1 Long description
Columns list potential bias, mechanism, germplasm most affected, potential solution, and references. Misclassification of genotypes is linked to performance outside the collection site that underestimates adaptive potential, affecting crop wild relatives, landraces, and breeding lines adapted for a specific climate, with multi environment testing linked to environmental metadata. Narrowing of the genetic base is described as selection that may favor widely adapted germplasm, affecting diverse and underrepresented genotypes, with diversity aware selection indices and allele frequency monitoring. Failure to detect adaptive traits is tied to standardized growing conditions that suppress environment specific traits such as photoperiod sensitivity, heat stress, and drought escape, affecting genotypes with specific stress adaptation, with targeted phenotyping under conditions that trigger adaptive traits and stress response. Underutilisation of climate resilient genetic resources is linked to niche specific adaptations remaining undetected under uniform or standardized screening, affecting locally adapted ecotypes and landraces, with integration of climate projections with genomic and phenotypic data.
Relevance under climate change
Climate change will increase environmental variability, including temperature extremes, irregular rainfall and shifting pest pressures. Under such conditions, local adaptation and niche-specific traits become increasingly valuable. Ecological genomics shows that adaptive alleles are often associated with environmental gradients (Lasky et al. Reference Lasky, Des Marais, Lowry, Povolotskaya, McKay, Richards, Keitt and Juenger2014; Exposito-Alonso et al. Reference Exposito-Alonso, Vasseur, Ding, Wang, Burbano and Weigel2018). Predictive models integrating genomic and climatic data demonstrate that genotype fitness under future climates depends strongly on ecological matching (Hancock et al. Reference Hancock, Brachi, Faure, Horton, Jarymowycz, Sperone, Toomajian, Roux and Bergelson2011; Cortés and Blair Reference Cortés and Blair2018). These findings underscore the utility of evaluating genetic resources in environments that reflect their ecological origins. Recent work further emphasizes that genebanks remain underutilized unless genomic and environmental data are jointly considered to identify climate-adapted germplasm (Campbell et al. Reference Campbell, Castañeda-Álvarez, Domingo, Von Wettberg, Runck, Nandkangré, Halpin-Mccormick, Fumia, Neyhart, Kilian, Wambugu, Nyamongo, Hübner, Sitar, Thompson, Rieseberg, Gore and Kantar2025; Cortés Reference Cortés2025). Beyond the breeding pipeline, increasing environmental mismatch under climate change may also compromise the reliability of variety trial data on which plant breeders, governments, NGOs, agricultural advisory services and development donors base their recommendations and investment priorities.
Strategies for environment-aware phenotyping
Addressing environmental mismatch requires environment-aware phenotyping approaches that better capture the agro-ecological diversity of both germplasm origins and target environments (Figure 1C). Key strategies include the following.
Multi-environment trials (METs)
METs enable explicit quantification of G × E interactions and identification of both broadly and specifically adapted genotypes (Crossa et al. Reference Crossa, Pérez-Rodríguez, Cuevas, Montesinos-López, Jarquín, De Los Campos, Burgueño, González-Camacho, Pérez-Elizalde, Beyene, Dreisigacker, Singh, Zhang, Gowda, Roorkiwal, Rutkoski and Varshney2017). For plant genetic resource assessment, evaluation across a minimum of two environments, whether two locations within the same season or the same location across two or more seasons, should be considered a baseline requirement rather than an optional enhancement. While two environments provide only a partial view of adaptive potential, they enable at least preliminary estimation of G × E interactions and reduce the risk of systematic misclassification of germplasm.
Integration of environmental metadata
Linking genotypes to their environments of origin (climate, soil, altitude) improves interpretation of phenotypic data and supports targeted testing (Hijmans et al. Reference Hijmans, Garrett, Huamán, Zhang, Schreuder and Bonierbale2000; Meyer and Purugganan Reference Meyer and Purugganan2013). Despite its importance, environmental metadata remains incomplete or inconsistent across many genebank collections, limiting its practical utility. Systematic recording and integration of such data into genebank information systems, alongside phenotypic and genomic data, is therefore a priority for improving the utilization of plant genetic resources (Volk et al. Reference Volk, Byrne, Coyne, Flint-Garcia, Reeves and Richards2021; Ghamkhar et al. Reference Ghamkhar, Hay, Engbers, Dempewolf and Schurr2025).
Reaction-norm and plasticity analysis
Reaction norms capture genotype responses across environmental gradients, providing a more comprehensive view of adaptive potential (Des Marais et al. Reference Des Marais, Hernandez and Juenger2013; Tardieu et al. Reference Tardieu, Simonneau and Muller2017). Plasticity varies in its forms and adaptive value among traits, genotypes and environments (Schneider Reference Schneider2022; Napier et al. Reference Napier, Heckman and Juenger2023). For plant breeders, two forms of phenotypic plasticity are particularly relevant. First, cryptic plasticity refers to cases where genotypes harbour environmentally responsive traits that are not expressed under standard evaluation conditions but become evident in ecologically matched environments. As a result, potentially valuable germplasm may be overlooked in conventional screening. Second, distinguishing between adaptive and maladaptive plasticity is essential: when genotypes are evaluated outside their native environments, observed plastic responses may not reflect performance in target conditions, leading to biased selection decisions. Reaction norm approaches, which quantify genotype-specific responses across environmental gradients, provide a robust framework for identifying both cryptic and potentially maladaptive plastic responses in diverse germplasm.
Decentralized phenotyping networks
While multi-environment trials provide the analytical framework for capturing G × E interactions, their effectiveness depends on both germplasm diversity and environmental coverage. Decentralized national and international phenotyping networks help address this limitation by distributing evaluation across a broad range of geographically and ecologically distinct sites. Unlike centralized trials designed primarily for experimental control, these networks are structured to capture the full breadth of agro-ecological variation relevant to target environments. Examples include those implemented by the CGIAR multi-location trials (Dhondt et al. Reference Dhondt, Dickmann, Fahrner, Pieruschka, Hoyos-villegas, Robinson and Von Gillhaussen2025) that enable broader environmental variation to be captured. However, such large-scale networks depend on substantial and sustained financial investment that is not available to many research institutions or universities. In resource-limited settings, more accessible approaches, such as regional collaborations among breeding programmes and genebanks, coordinated evaluation across two or more representative environments and the sharing of phenotypic and environmental data through collaborative networks, can provide practical and scalable alternatives for improving the assessment of G × E interactions.
Simulation and modelling approaches
Crop models and genomic prediction frameworks incorporating environmental covariates improve prediction accuracy across environments (Cooper et al. Reference Cooper, Messina, Podlich, Totir, Baumgarten, Hausmann, Wright and Graham2014; Millet et al. Reference Millet, Kruijer, Coupel-Ledru, Alvarez Prado, Cabrera-Bosquet, Lacube, Charcosset, Welcker, Van Eeuwijk and Tardieu2019). However, when genomic selection is an appropriate breeding strategy, environment-aware genomic selection remains a significant challenge, particularly for complex quantitative traits under changing climatic conditions (Yan and Wang Reference Yan and Wang2026). For traits with simple genetic architectures or where conventional phenotypic selection is sufficient, genomic selection may provide limited additional benefit. Together, these approaches align phenotyping strategies with the ecological complexity of plant agricultural systems.
Taken together, these strategies provide a practical framework for aligning phenotypic evaluation with the agro-ecological diversity of crop germplasm collections and the environments in which selected genotypes will ultimately be deployed. In practice, these approaches are most effective when used in combination and adapted to the specific breeding context. Realizing this transition will require coordinated investment across national and international genebanks, breeding programmes and researchers.
Conclusions
Phenotypic evaluation remains a cornerstone of plant genetic resource research, yet current approaches are constrained by a fundamental limitation: the mismatch between evaluation environments and the ecological contexts of genotypes. Both controlled-environment and field-based screening systems can introduce bias that obscures true genetic potential. Recognizing and addressing this limitation is essential for improving genetic resource characterization, enhancing breeding efficiency and unlocking adaptive diversity for climate resilience. Several practical recommendations follow from this critical and empirical analysis for greater precision. First, evaluation of plant genetic resources should span a minimum of two environments as a baseline for capturing G × E interactions. Second, environmental metadata describing genotype origins should be systematically linked to phenotypic records in genebank databases to support targeted evaluation. Third, breeding programmes may explicitly account for local adaptations present in crop wild relatives and landraces. Fourth, the integration of genomic and pre-breeding tools (including GenBank genomics, pan-genomes, genomics-assisted breeding, multi-omics approaches and the use of backcross-derived materials generated by crossing plant genetic resources with modern varieties) with phenotypic evaluation will improve the accuracy of identifying adaptive diversity across diverse environments. Fifth, investment in international decentralized phenotyping networks will be essential for capturing the full breadth of agro-ecological variation relevant to diverse germplasm. A shift towards environment-aware phenotyping that incorporates ecological and evolutionary principles will be critical for maximizing the value of plant genetic resources in the coming decades.
Acknowledgements
We thank Prof. Petr Smýkal (Palacký University Olomouc) and Dr Pedro Aphalo (University of Helsinki) for valuable discussions and suggestions on this topic.
Author contributions
H.K. and R.O. did the conceptualization, writing – original draft and writing – editing. H.K. did visualization.
Data availability statement
No datasets were generated or analysed during the current study.
Funding statement
H.K. was supported by the Research Council of Finland, Academy projects, funding decision 363375 (Fabagen).
Competing interests
The authors declare no competing interests.
Ethical standards
As this is a critical review article, no ethics approval is required.
Open access
