Landscape Genomics Workflow: Linking Population Structure, Environment, and Local Adaptation
Key takeaways
- Landscape genomics targets genotype–environment relationships under spatial structure, not "population genomics plus a climate spreadsheet."
- Start with a biological contrast + sampling hypothesis; environmental layers come later.
- Environmental coverage, replication, and confounding control often matter more than adding more samples in the same space.
- Treat structure, geography, and environment as competing explanations before interpreting any association as local adaptation.
- Genotype–environment association produces candidates, not validated adaptive loci; interpretation should be conservative unless evidence converges.
Why Landscape Genomics Is Not Just Population Genomics With Climate Layers
Landscape genomics becomes informative only when genomic variation, environmental gradients, and spatial structure are analyzed together. Population genomics describes how variation is distributed (structure, diversity, demography). Environmental layers describe what varies across space. Landscape genomics links the two and asks a narrower question: after accounting for structure and space, does environment still explain a defensible portion of genomic variation?
What Landscape Genomics Is Trying to Explain
A structure-aware workflow tries to identify environment-associated genomic patterns that persist after controlling neutral explanations. Done carefully, it can support candidate drivers of adaptive differentiation, shortlists of candidate regions for follow-up, and conservative vulnerability framing under environmental change.
Most studies still need a solid neutral baseline (diversity summaries, structure diagnostics) because it anchors interpretation. If you need a concise upstream reference for those baseline summaries, CD Genomics provides an overview of genetic diversity analysis.
Why Population Structure Alone Is Not Enough
Structure is necessary context, but not an adaptation result. Strong structure can arise from drift, barriers, colonization history, or range expansions. If you stop at "there are clusters," you cannot justify which parts of variation are environment-linked.
The deeper risk is confounding: structure is both signal and confounder. If a structure axis aligns with geography, it can align with climate too. Interpreting that overlap as selection without safeguards is often just relabeling demography.
Why Environmental Metadata Is Not Enough Either
Environmental variables are not selective pressures by default. Two pitfalls repeatedly weaken studies:
- ecological mismatch (predictors don't match how the organism experiences environment),
- design-induced confounding (environment and geography are so entangled in the sampled dataset that association becomes a spatial artifact).
Questions This Workflow Can Actually Answer
Landscape genomics can credibly answer questions about environment-associated variation and nominate candidate adaptive regions. It cannot replace validation experiments, and it cannot justify treating every environment-associated locus as a confirmed adaptive locus. Conservative language is a feature, not a weakness.
Start With the Research Question Before You Download Environmental Layers
The most useful landscape genomics projects begin with a biological contrast and a sampling hypothesis. If you define the question late, you'll end up fitting a model first and inventing the story afterward.
Questions About Local Adaptation vs Connectivity vs Vulnerability
Most projects sit in one (or more) of three goal families:
- Local adaptation / environmental association (GEA): which gradients are associated with genomic variation beyond neutral structure?
- Connectivity / gene flow: which landscape features shape movement and structure?
- Vulnerability / genomic offset: under changed environments, which populations are predicted to be most mismatched?
These are related but not interchangeable. Connectivity work requires replication across barriers; local adaptation work requires environmental contrasts that are not just "north vs south." Vulnerability metrics inherit upstream assumptions and therefore demand extra transparency.
When the Goal Is Discovery vs Decision Support
Discovery-oriented work can be broader, but it should label outputs as candidates and hypotheses. Decision-support work (restoration, assisted migration, conservation planning) needs clearer assumptions, uncertainty framing, and a defensible chain from design to inference.
Why the Wrong Environmental Question Weakens the Whole Study
If your question is "which variables are significant," significance is rarely scarce—especially with correlated predictors and spatial structure. The more defensible question is whether an interpretable variable set explains genomic patterns beyond neutral structure and whether the conclusion is stable under reasonable alternative variable definitions.
This sensitivity is especially visible for genomic offset. Uncertainty in offset forecasts can be driven by predictor choice and modeling components, as illustrated by Lachmuth et al. in an uncertainty-focused offset analysis (see Frontiers in Ecology and Evolution, 2023).
A Simple Rule for Aligning the Question With the Design
Write your question as a contrast you can sample and defend: "Do populations in low-moisture sites show environment-associated genomic variation beyond neutral structure?" If you can't express a contrast, you'll struggle to justify variables, sampling allocation, and interpretation.
Design Sampling to Cover Environmental and Geographic Space
Sampling design matters most when it captures environmental contrasts and spatial replication needed to distinguish adaptation from neutral structure.
Why Environmental Coverage Matters More Than Sheer Sample Count
More samples help, but coverage and replication often buy more than adding N within a narrow space. Simulation-based optimization work shows that environmentally representative sampling can increase power to detect true genotype–environment relationships compared with geographically clustered designs (see Dubois et al. 2019: Sampling strategy optimization…). Recent optimization work also emphasizes that many common analyses are relatively robust when sampling covers sufficient environmental and geographic space (see Bishop 2024: Optimising Sampling Design for Landscape Genomics).
Population-Based vs Individual-Based Sampling
Population-based designs (multiple individuals per site) are often easier to defend when outputs are population-level (units, differentiation, management recommendations). Individual-based designs can be valuable when individuals are the decision unit or when fine-scale heterogeneity matters, but they raise the bar for relatedness control and metadata consistency.
Replication Across Similar Environments
Replication is what makes an association look less like geography. If you can sample similar environments in multiple regions, you can test whether the same environment association appears under different spatial histories.
How to Avoid Geographic and Environmental Confounding
Perfect separation is uncommon, but you can reduce predictable confounding by avoiding sampling paths where environment changes only along one geographic axis, adding replication where feasible, and being explicit about remaining overlap so interpretation stays conservative.
A Practical Sampling Checklist Before Genotyping Starts
Before genotyping, you should be able to state your unit of inference, the gradient you intend to test, where replication exists, and where environment and geography are inseparable.
Which Sampling Design Fits Your Landscape Genomics Study?
| Study goal | Sampling unit | Environmental coverage needed | Geographic replication | Main strength | Main risk | Best-fit scenario |
|---|---|---|---|---|---|---|
| Candidate local adaptation gradients (GEA discovery) | Individuals or populations | High | Medium–high | Identifies candidate drivers and regions | Confounding if replication is weak | Broad-range species; hypothesis generation |
| Delineating conservation/management units | Populations | Medium | Medium | Clear population-level interpretation | Misses fine-scale heterogeneity | Projects needing defensible units |
| Connectivity and barrier inference | Individuals + populations | Low–medium | High | Strong spatial movement inference | Mistaking selection for resistance | Fragmented landscapes; corridor planning |
| Vulnerability screening (genomic offset) | Individuals or populations | High | Medium | Structured mismatch-risk framing | Overinterpreting offset as fitness | Change-risk screening with explicit uncertainty |
| Restoration / assisted migration donor selection | Populations | High | High | Better transferability logic | Governance/ethics constraints | Projects with monitoring or validation plans |
Choose Environmental Variables That Can Defend a Biological Story
Environmental predictors become useful only when they are interpretable, non-redundant, and aligned with organism ecology and scale.
Key terms (quick definitions)
- GEA (genotype–environment association): Statistical tests/models that relate genetic variation to environmental predictors while attempting to control confounding from neutral structure.
- Collinearity: Strong correlation among environmental predictors that can inflate apparent significance and destabilize interpretation.
- IBD (isolation by distance): A pattern where genetic similarity declines with geographic distance due to limited dispersal, which can mimic selection if not modeled.
- Spatial autocorrelation: Nearby samples tend to be more similar (genetically and environmentally) than distant ones; ignoring it can create spatial artifacts.
- Genomic offset: A model-based measure of predicted mismatch between current genetic–environment relationships and future (or alternative) environments; best treated as a relative risk indicator, not direct fitness.
Climate, Soil, Elevation, and Land-Use Variables
A defensible set usually includes predictors tied to plausible stress or exposure: temperature extremes, moisture balance, seasonality, topography, substrate/soil proxies, and land-use or habitat features when they plausibly affect survival or dispersal.
Why Correlated Layers Create False Confidence
Collinearity can make many predictors look significant and makes causal interpretation weaker. If several correlated predictors emerge, a defensible claim is often "this genomic pattern aligns with a correlated environmental complex," not "we identified multiple independent selective drivers."
Matching Variable Resolution to Organism Biology
Resolution should match how the organism experiences environment and the spacing of your samples. Microclimate proxies can help when sampling density supports them and can mislead when they do not.
When High-Resolution Proxies Add Value
High-resolution proxies add value when you can defend the mechanism and demonstrate stability against simpler alternatives. Otherwise, they mostly add flexibility.
How to Build a Smaller, Stronger Variable Set
A conservative workflow starts from ecology, reduces redundancy, keeps variables you can interpret, and documents sensitivity checks. For offset-style metrics, quantitative work emphasizes that offset inherits assumptions from the fitted genotype–environment model (see Gain et al. 2023: A Quantitative Theory for Genomic Offset Statistics).
Build a Workflow That Controls Structure Before Calling Local Adaptation
Landscape genomics becomes more credible when structure, geography, and environment are treated as competing explanations.
Step 1: QC and Cohort Definition
QC is not only missingness and marker filters; it is where you define the cohort, identify relatedness or batch risks, and decide whether cryptic lineages or admixture require explicit handling.
Step 2: Population Structure and Relatedness Review
Structure review must happen before GEA. The goal is to learn whether structure axes align with geography or environment and whether outliers will dominate models.
For a compact overview of structure-first deliverables and interpretation, CD Genomics summarizes population structure & evolution analysis. If your dataset is built from dense SNP calling, an upstream option is whole-genome re-sequencing for population genetics.
Step 3: Environmental Variable Filtering
Filter predictors before association testing and report the rationale (redundancy control, scale choices, and ecological meaning).
Step 4: Genotype–Environment Association Testing
GEA is the core association step, but credibility depends on modeling structure and space explicitly rather than "correcting later."
Step 5: Cross-Checking With Differentiation or Selection Signals
GEA is stronger when candidates are cross-checked against an independent evidence line (differentiation, selection scans where appropriate, ecological plausibility). Cross-checking does not prove adaptation, but it reduces predictable false positives.
Step 6: Prioritizing Candidate Regions and Variables
Prioritization converts many associations into a short list you can justify. Candidates that mirror structure axes or depend on a single correlated predictor should be treated as lower-confidence unless other evidence supports them.
Control the Five Risks That Most Often Create False Signals
Most overclaims trace back to the same five risks.
Collinearity Among Environmental Predictors
Correlated predictors spread apparent association across many layers and destabilize interpretation. The defensible response is redundancy control and sensitivity checks.
Spatial Autocorrelation and Isolation by Distance
Nearby samples are often genetically similar and environmentally similar. Without treating space as a competing explanation, isolation by distance can masquerade as adaptation.
Structure–Environment Overlap
When structure and environment overlap, your workflow should force conservative interpretation and lean more heavily on replication and cross-evidence.
Uneven Sampling Across Habitats
Sampling imbalance can inflate false positives and produce associations driven by clustered subsets. Reporting allocation and coverage is part of risk control.
Variables That Are Statistically Convenient but Biologically Weak
Predictors that are hard to defend mechanistically should not anchor adaptation claims even if they improve model fit.
Before You Call It Local Adaptation, Check These Five Risks: collinearity, spatial autocorrelation, structure–environment overlap, uneven habitat sampling, and weak proxies.
When connectivity and movement are central to interpreting structure and spatial risk, gene flow outputs are often part of the evidentiary package; CD Genomics provides an overview of deliverables for gene flow analysis.
Interpret Candidate Signals Without Overclaiming Adaptation
A landscape genomics result is more convincing when genomic, environmental, and biological evidence converge.
Environment-Associated Loci vs Confirmed Adaptive Loci
Environment-associated loci are statistical signals. Confirmed adaptive loci require converging evidence and ideally validation. Treating association as confirmation is a credibility failure.
How to Use Population Differentiation as Supporting Evidence
Differentiation can support interpretation when candidates align with ecological boundaries and when patterns are consistent across analyses. When candidates reproduce a structure axis, the conservative interpretation is that structure dominates.
When Functional Annotation Helps
Annotation can improve plausibility and help prioritize follow-up, but it should not be treated as proof.
When Genomic Offset or Vulnerability Models Add Value
Offset metrics can add value when you need a structured way to discuss relative mismatch risk under environmental change, but interpretation should remain conservative. Quantitative and critical discussions emphasize that genomic offset inherits assumptions from the genotype–environment model and is often overinterpreted as fitness; see Gain et al. 2023 (linked above) and Lotterhos et al. 2024: Interpretation issues with "genomic vulnerability".
A Conservative Rule for Candidate Adaptive Regions
If you can't describe a candidate in one sentence that states the environment contrast, the structure context, and why the signal is not a geography proxy, it is not ready to be labeled adaptive.
For complementary evidence lines alongside GEA candidates, CD Genomics provides overviews for selective sweep analysis service and CNV analysis.
What Good Landscape Genomics Figures and Tables Look Like
Reviewer-trusted reporting shows sampling context, environmental contrasts, structure control, and candidate-region evidence rather than only a map and a Manhattan plot.
Sampling and Environmental Coverage Maps
A strong sampling map makes replication and gaps visible and shows whether the design supports the contrast you claim.
Structure-Aware Environment Summaries
Environment summaries are more persuasive when paired with structure diagnostics, especially when overlap is present.
GEA Result Plots and Candidate-Region Tables
Useful tables and figures help readers understand why candidates are credible: which predictors they align with, whether signals are stable, and what cross-evidence exists.
Variable Selection and Collinearity Summaries
Transparent reporting of variable selection and redundancy control prevents the "you tried everything until it worked" impression.
A Short Reporting Checklist for Methods and Supplementary Files
A reviewer-ready package typically includes sampling logic and metadata standards, structure diagnostics, predictor definitions and filtering steps, candidate criteria, and uncertainty framing for any vulnerability outputs.
What Real Landscape Genomics Studies Reveal About Design Trade-Offs
Published studies are most useful when they clarify design trade-offs rather than when they list significant loci.
Case Example 1: A Crop or Landrace Study Linking Climate and Adaptive Differentiation
In landraces, broad environmental gradients and plausible stress hypotheses can make study design cleaner than in highly structured wild systems. The general lesson is that the strongest claims come from structure-first analysis, interpretable predictors, and conservative candidate prioritization.
Case Example 2: A Conservation-Genomics Study Showing the Value of Individual-Based Sampling
Individual-based designs can capture heterogeneity that population averages miss and can be decision-relevant when individuals are the management unit. The trade-off is stricter relatedness control, careful spatial balance, and conservative interpretation when structure–environment overlap is strong.
Case Example 3: What High-Resolution Environmental Proxies Add and What They Do Not
High-resolution proxies can sharpen inference when they match organism exposure and sampling density. They can also inflate confidence when they mainly add correlated detail. The safest practice is to treat them as sensitivity-tested hypotheses, not automatic upgrades.
When to Use a Service Instead of Building Every Step In-House
Landscape genomics projects are often worth external support when sampling design, environmental curation, structure-aware analysis, and reviewer-ready interpretation are the real bottlenecks.
CD Genomics can support research-use landscape genomics projects (RUO) when teams need help with sampling-aware study design, environmental variable curation, structure-conscious association analysis, and candidate-region interpretation. For scoping, see the landscape genomics solution page and the broader menu of bioinformatics analysis for population genomics.
FAQs
Population genomics focuses on structure, diversity, and demographic history using genomic variation, while landscape genomics focuses on genotype–environment relationships under spatial structure to ask whether environmental gradients explain variation beyond neutral processes. Landscape genomics is not an "extra layer" after PCA; it is a joint inference problem that must treat space, structure, and environment together.
Include as many as you can defend biologically and report transparently, not as many as you can download. A smaller set with clear ecological meaning, controlled redundancy, and sensitivity checks is usually more reviewer-trusted than a large correlated set that produces fragile interpretations.
More sampling helps, but coverage and replication often matter more than raw counts. If the design does not span the gradients you claim to test, or if environment and geography are perfectly confounded, adding more samples in the same narrow space may not improve interpretability.
You rarely separate them perfectly; you reduce confounding by designing replication across similar environments, treating structure and space as competing explanations in models, and interpreting candidates conservatively when structure axes overlap environmental gradients. Cross-checking GEA candidates with differentiation signals and ecological plausibility helps avoid overclaiming.
Genomic offset is worth adding when you need a structured way to discuss relative mismatch risk under environmental change and when the upstream genotype–environment model is already defensible and structure-aware. It should be interpreted as a relative risk indicator with explicit uncertainty, not as a direct measurement of future fitness loss.
A good brief states the biological contrast, unit of inference, and publication or management goal; includes complete sample metadata; specifies which environmental gradients are hypothesized to matter; and defines actionable deliverables such as a structure-aware workflow report, a filtered predictor set with rationale, candidate-region tables with cross-evidence notes, and reviewer-ready figures.
Dr. Yang H.
Senior Scientist at CD Genomics
Dr. Yang H. contributes scientific content on genomics methods, sample strategy, and project planning for research teams working in biodiversity, population genetics, and related fields. His writing focuses on helping readers make clearer technical decisions before starting or outsourcing complex research workflows.
References
- Aitken, S. N. (2024). Conserving evolutionary potential: Combining landscape genomics with climate adaptation applications. Annual Review of Plant Biology. https://www.annualreviews.org/content/journals/10.1146/annurev-arplant-070523-044239
- Alvarado, A. H., Bossu, C. M., Harrigan, R. J., Bay, R. A., Nelson, A. R. P., Smith, T. B., & Ruegg, K. C. (2022). Genotype–environment associations across spatial scales reveal the importance of putative adaptive genetic variation in divergence. Evolutionary Applications, 15(10), 1823–1841. https://doi.org/10.1111/eva.13444
- Bishop, T. R. (2024). Optimising sampling design for landscape genomics. Molecular Ecology Resources. https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14052
- Caye, K., Jumentier, B., Lepeule, J., & Francois, O. (2019). LFMM 2: Fast and accurate inference of gene-environment associations in genome-wide studies. Molecular Biology and Evolution, 36(4), 852–860. https://doi.org/10.1093/molbev/msz008
- Dubois, M.-P., et al. (2019). Sampling strategy optimization to increase statistical power in landscape genomics: A simulation-based approach. Molecular Ecology Resources. https://pmc.ncbi.nlm.nih.gov/articles/PMC6972490/
- Flanagan, S. P., et al. (2017). Guidelines for planning genomic assessment and monitoring of locally adaptive variation to inform species conservation. Frontiers in Genetics. https://pmc.ncbi.nlm.nih.gov/articles/PMC6050180/
- Gain, C., et al. (2023). A quantitative theory for genomic offset statistics. Molecular Biology and Evolution, 40(6), msad140. https://doi.org/10.1093/molbev/msad140
- Hoban, S., et al. (2016). Finding the genomic basis of local adaptation: Pitfalls, practical solutions, and future directions. The American Naturalist, 188(4), 379–397. https://doi.org/10.1086/688018
- Lachmuth, S., et al. (2023). Uncertainty in genomic offset analyses under environmental change. Frontiers in Ecology and Evolution. https://www.frontiersin.org/journals/ecology-and-evolution/articles/10.3389/fevo.2023.1155783/full
- Lotterhos, K. E., & Whitlock, M. C. (2014). Evaluation of demographic history and neutral parameterization on the performance of FST outlier tests. Molecular Ecology, 23(9), 2178–2192.
- Lotterhos, K. E. (2023). The paradox of adaptive trait clines with nonclinal patterns: Limits of genotype–environment association methods. PNAS, 120(11). https://www.pnas.org/doi/10.1073/pnas.2220313120
- Lotterhos, K. E., et al. (2024). Interpretation issues with "genomic vulnerability". Proceedings of the National Academy of Sciences / commentary and related discussion. https://pmc.ncbi.nlm.nih.gov/articles/PMC11134465/
- Salloum, P. M., Gardner, M. G., Bertozzi, T., & Weeks, A. R. (2022). Finding the adaptive needles in a population-structured haystack: Landscape genomics of a threatened freshwater fish. Molecular Ecology Resources. https://pmc.ncbi.nlm.nih.gov/articles/PMC9311215/
Related
Services
Download
