Herbarium specimens reliably track plant phenological responses to climate change in understudied montane biomes

This study demonstrates that herbarium specimens are reliable resources for tracking plant phenological responses to climate change in montane systems, showing consistent flowering time estimates and sensitivity to snow density with field observations, though they may slightly underestimate temperature sensitivity and benefit from focusing on early-flowering individuals to capture community-level trends.

The Gist

Imagine you are trying to understand how a city's traffic patterns have changed over the last 50 years. You have two ways to get this information:

1. The Gold Standard: You have a team of traffic officers standing on every corner, counting cars every single day for 50 years. This is perfect, but it only covers a few specific streets.
2. The "Snapshot" Method: You have a giant, dusty photo album containing millions of random snapshots of traffic taken by tourists over the last 50 years. These photos cover the whole city, but they are messy. Some are blurry, some are taken at noon, some at 3 PM, and some people might have missed the morning rush hour entirely.

This paper is about asking: "Can we trust the messy photo album to tell us the same story as the traffic officers?"

Here is the breakdown of the study, translated into everyday language:

The Big Question

Scientists want to know how plants in the mountains are reacting to climate change. Specifically, are they blooming earlier because the snow is melting sooner?

• The "Gold Standard" data comes from the Rocky Mountain Biological Laboratory (RMBL), where scientists have been counting flowers in specific meadows every week for 50 years.
• The "Snapshot" data comes from herbarium specimens. These are dried plants pressed onto paper and glued into museum collections. They are like the "photos" of the plant world. They cover a much wider area, but they were collected randomly by botanists over decades, not systematically.

The researchers wanted to know: If we only look at the museum specimens, will we get the right answer about how climate change is affecting plants?

The Experiment

The team took 45 different mountain plant species. They compared the "museum snapshots" against the "traffic officer" records from the same area. They looked at two main things:

1. Snow: Does the amount of snow in May affect when the flowers bloom?
2. Temperature: Does the warmth of spring affect when they bloom?

What They Found (The Results)

1. The "When" is mostly right, but the "Details" are missing.
The museum specimens gave a pretty good average idea of when plants bloom. If you asked, "When do these flowers usually bloom?" the museum data said "Mid-July," and the field data said "Mid-July."

• The Catch: The field data showed that sometimes, a few flowers bloom super early (right after the snow melts), and then there's a second wave of blooms later. The museum data missed these "early birds" and the "late waves." It was like the tourists in our photo analogy only took pictures during the middle of the day, so they missed the morning rush and the evening commute.

2. Snow is easy to track; Temperature is tricky.

• Snow: The museum data was great at showing how plants react to snow. If the snow melts early, the flowers bloom early. The museum records matched the field records almost perfectly here.
• Temperature: This is where the museum data got a little "lazy." When it got warmer, the field data showed plants blooming much earlier. The museum data showed them blooming earlier too, but not as much.
  • Why? Think of it like a fashion show. Botanists collecting plants for museums usually wait until the flowers are fully open and looking their best (the "peak" of the show). They rarely catch the very first buds opening. Because they missed the very early, temperature-sensitive start of the blooming season, the museum data made it look like plants are less sensitive to heat than they actually are.

3. The "Early Bird" Trick
The researchers found a clever fix. If you look at the earliest flowering plants in the museum collections (the "early birds"), their reaction to temperature actually matches the field data very well. It's just that the "average" museum plant (which is usually the mid-blooming one) hides this sensitivity.

The Bottom Line

Yes, museum specimens are reliable, but you have to know how to read them.

• They are a superpower for filling gaps: We can't have traffic officers on every mountain in the world. Museums give us data from places we've never visited.
• They have blind spots: They tend to miss the very first and very last moments of a season.
• The Verdict: If you want to know the general trend of how plants are changing with the climate, the museum "snapshots" are a fantastic tool. Just remember that they might slightly underestimate how fast plants are reacting to warming temperatures, because they often miss the very first bloomers.

In short: The museum data is like a great summary of a movie, but if you want to see the very first scene and the very last scene, you might need to go back to the theater (the field) to see the full picture. But for understanding the plot? The summary works perfectly.

Technical Summary

1. Problem Statement

Long-term field observations are considered the "gold standard" for studying plant phenology (timing of biological events like flowering) in montane systems. However, these observations are spatially limited, often restricted to specific research sites like the Rocky Mountain Biological Laboratory (RMBL). Conversely, herbarium specimens offer vast spatial and temporal coverage but are suspected to contain biases that may compromise their utility in montane environments.

Specific challenges in using herbarium data for montane biomes include:

• Sampling Bias: Collections are often concentrated at lower, more accessible elevations and occur during a narrow window (late spring/summer), potentially missing early-flowering individuals.
• Environmental Complexity: Montane phenology is heavily driven by snowmelt timing and micro-topography, which may not be accurately captured by coarse climate data associated with herbarium collection points.
• Validation Gap: It remains largely untested whether herbarium-derived phenological data can faithfully capture montane phenological responses to climate change compared to continuous field observations.

2. Methodology

The study conducted a rigorous validation by comparing phenological data from digitized herbarium specimens against 50 years of continuous field observations (1975–2022) from the RMBL in the Southern Rocky Mountains, USA.

Data Sources:

• Field Observations (RMBL): 35 permanent plots across meadow and forest habitats. Researchers counted flowers/umbels/capitula three times weekly. Data included 14 environmental variables (temperature, precipitation, snow density, etc.) derived from local weather stations and PRISM climate data.
• Herbarium Specimens: 29,909 digitized specimens from the SEINet Portal Network.
  • Filtering: Specimens were filtered to include only those within the same "climatic space" (EPA Level IV ecoregion and PCA-derived climate envelope) as the RMBL plots.
  • Final Dataset: 1,120 specimens representing 45 species with recorded flower counts.
  • Crowdsourcing: Flower counts on specimen sheets were extracted using Amazon Mechanical Turk (MTurk) via the CrowdCurio interface, with reliability scores calculated based on duplicate image scoring.

Statistical Analysis:

• Phenological Distribution: Weibull distributions were fitted to flowering Day of Year (DOY) to estimate early (10th), median (50th), and late (90th) quantiles for both datasets.
• Climate Modeling:
  • Snow Density: A Random Forest (RF) model was trained on local RMBL climate data to predict monthly snow density based on temperature, precipitation, and temperature range. This model was applied to both herbarium and field data.
  • Sensitivity Analysis: Linear Mixed Models (LMMs) were used to estimate the slope of flowering DOY in response to:
    1. Snow Density: Specifically May snow density (identified as the best predictor via AIC).
    2. Spring Temperature: Mean temperature from April to June.
• Comparisons: The study compared mean slopes, species-specific slopes, and quantile regression slopes between the three data sources:
  1. Herbarium specimens + PRISM climate data.
  2. RMBL observations + Local climate data (Prather et al.).
  3. RMBL observations + PRISM climate data.

3. Key Contributions

• Validation of Herbarium Data in Montane Systems: This is one of the first studies to rigorously validate herbarium phenology against long-term field data specifically in high-elevation, snow-driven ecosystems.
• Quantification of Sampling Bias: The study quantifies how herbarium data differs from field data (e.g., missing early peaks, underestimating temperature sensitivity) rather than just stating they are different.
• Methodological Framework: It provides a robust protocol for integrating crowdsourced phenological data from herbaria with high-resolution local climate data to study climate change impacts in data-poor regions.
• Insight into "Early" vs. "Mean" Responses: The paper demonstrates that for herbarium data, focusing on "early" flowering individuals (10th quantile) may better approximate community-level responses to temperature than using mean values.

4. Key Results

• Flowering Time Distribution:
  • Consistency: Mean flowering times (median/50th quantile) inferred from herbarium specimens were generally consistent with RMBL observations (approx. DOY 200).
  • Missing Peaks: Herbarium data failed to capture secondary community-level flowering peaks observed in the field (e.g., an early peak at DOY 130–150). This suggests herbarium collections miss the very onset of the growing season, likely due to collection bias toward peak flowering.
• Sensitivity to Snow Density:
  • High Agreement: Estimates of phenological sensitivity to May snow density were consistent between herbarium specimens and field observations. Both datasets showed that higher snow density leads to later flowering.
  • Correlation: Species-specific slopes were moderately correlated ($r = 0.5$) between the two datasets.
• Sensitivity to Spring Temperature:
  • Underestimation: Herbarium specimens yielded significantly shallower (more conservative) slopes for temperature sensitivity compared to field observations.
  • Cause: This is attributed to the bias of collecting specimens at peak flowering, which smooths out the rapid shifts in flowering time that occur in response to early post-snowmelt temperature fluctuations.
  • Quantile Insight: The 10th quantile (early flowering) slopes from herbarium data were most similar to the 50th quantile (median) slopes from field data, suggesting that "early" herbarium specimens are the best proxy for population-level temperature responses.
• Robustness: Results remained robust when accounting for non-independence of repeated observations and varying sample sizes.

5. Significance

• Bridging Data Gaps: The study confirms that despite inherent biases, herbarium specimens are a reliable resource for closing knowledge gaps in understudied montane biomes where long-term field data is scarce.
• Refining Interpretation: Researchers can use herbarium data to track broad-scale phenological shifts, but must interpret temperature sensitivity with caution. Specifically, they should account for the "smoothing" effect of collection timing or focus on early-flowering individuals to capture true population responses.
• Future Directions: The findings support scaling up phenological assessments across broader elevational gradients using herbarium data, provided that statistical methods (like quantile regression) are used to correct for sampling biases.
• Conservation Implications: Accurate tracking of montane phenology is critical for understanding how climate change affects gene flow, pollinator interactions, and community composition in these sensitive ecosystems. Herbaria provide the only historical baseline for many of these remote regions.