Trichomes are morphologically diverse and functionally important structures that vary within plant lineages in response to selection pressures across ecological gradients and evolutionary timescales. Classic hypotheses predict higher investment in trichomes in arid environments, at lower latitudes, and in long-lived species, as well as shifts in trichome production to reduce conflict between defense traits and mutualisms. However, tests of these hypotheses often neglect the rich diversity of trichome phenotypes and rely on aggregate metrics of trichome production, despite many plant species producing a variety of trichome morphologies across different organs. Here, we collected data on fine-scale patterns of trichome length, density, and type in 52 species of blazingstars (Mentzelia: Loasaceae) and tested whether individual metrics of trichome investment were consistent with existing adaptive hypotheses. Contrary to longstanding hypotheses, we found that Mentzelia species tend to display greater trichome investment in less arid environments and at higher latitudes. Sticky trichomes are significantly less common on the upper surface of the leaf, possibly reducing defense-pollination conflict. Species with larger petals (a proxy for reliance on insect pollinators) also shift investment away from insect-trapping hairs on the underside of the leaf. Examining trichomes types separately revealed that different morphologies show distinct responses to abiotic and biotic factors, demonstrating the need to consider multiple axes of diversity when testing adaptive hypotheses for complex traits.

These are the data and code to accompany Glos & Weber “Multiple metrics of trichome diversity support independent evolutionary hypotheses in blazingstars (Mentzelia: Loasaceae).” For this project, we collected data on fine-scale patterns of trichome length, density, and type in 53 taxa of blazingstars (Mentzelia: Loasaceae) and tested whether trichome investment was consistent with existing adaptive hypotheses across multiple trichome metrics.

Trichome measurements —  Data were obtained from up to 5 specimens (average of 3.38) per taxon from 52 species, including both subspecies of  Mentzelia laevicaulis (Douglas ex Hook.) Torr. & A.Gray, which were measured separately due to their morphological disparity. The resulting dataset spans 53 taxa (approximately half the genus) and 179 specimens. Specimens were imaged at 6.3x using an Olympus SZX16 stereomicroscope fitted with an Olympus DP28 camera and Olympus CellSens software (Evident Scientific, Inc., Massachusetts, USA). Updated nomenclature and synonymy were determined by querying specimen names against World Flora Online (www.worldfloraonline.org). Trichome measurements were performed in FIJI (Schindelin et al., 2012) using the ObjectJ plugin (https://sils.fnwi.uva.nl/bcb/objectj/). Trichome characteristics were measured near the midpoint of mature leaves, avoiding the leaf midvein where possible to facilitate consistency between specimens. Length and density of glochidiate and scabrid trichomes were recorded separately. Trichomes were considered glochidiate if they bore one or more whorls of recurved barbs (e.g. a trichome with a grappling, recurved tip but scabrid-like shaft was considered glochidiate). Density of each trichome type was quantified using a 1 mm² square superimposed in the center of the image, including trichomes that fell partially inside the square. Length measurements (recorded in um) were initiated in the same 1 mm², expanding outward or to additional images as necessary to measure ten trichomes per type per leaf surface per individual (up to 200 measurements per species). Unobserved trichome types represent functional, rather than absolute, absences; it is possible that a trichome type was not observed because it was present at such low densities as to not appear in our images, was restricted to the midvein, or was too small to be detectable.

Total trichome density was calculated by summing the scabrid and glochidiate density on each side of the leaf. Natural log transformations to the data were performed before species means and standard errors were calculated, to reduce approximation of standard errors (Garamszegi, 2014). Absent trichome types were recorded as 0s in the density data but marked “NA” in the length data to avoid biasing means toward zero-length (unobserved) trichomes. Before logging the density data, 1 was added to all density metrics to avoid creating undefined values from log(0). As the statistical methods we employed required that all traits had associated standard errors, it was necessary to assign standard errors for some species as follows: (a) in cases where we examined only one specimen for a species and standard error could not be calculated (12 cases), that species was assigned the median standard error for that trait and (b) in cases where multiple specimens were examined, but the calculated standard error was 0 because all specimens had identical measures for that trait (two cases: adaxial glochidiate density in Mentzelia chrysantha Engelm. and adaxial total density in Mentzelia pectinata Kellogg) that species was assigned the smallest non-zero standard error for that trait.

Latitude and Climate data — Occurrences were obtained from GBIF using the R package rgbif (Global Biodiversity Information Facility, 2024), limiting the occurrences to preserved specimens with recorded collection dates. Additional filtering was performed using the “clean_coordinates” function in the r package CoordinateCleaner (Zizka et al., 2019). This step removed points around capitals, country centroids, biodiversity institutions, and the GBIF headquarters, as well as those with equal latitude and longitude coordinates, coordinates in the ocean, zeros, duplicates, and outliers. In an effort to reduce potential spatial autocorrelation due to geographic clustering in the GBIF data (Beck et al., 2014), we thinned the occurrence data to one observation per 1 km grid cell using the “thin_points” function in the r package geoThinneR (Mestre-Tomás, 2024). We chose a thinning distance of 1 km to approximately match the spatial resolution of the aridity index and WorldClim data (30 arc seconds, see below), reducing the chance that a given climate value will be over-weighted when calculating species means. This process yielded 14,550 occurrences for the 53 Mentzelia taxa in our dataset (mean: 274.5 occurrences per species, summarized in Table S1). Only four species – M. chrysantha Engelm., M. librina (K.H.Thorne & F.J.Sm.) J.J.Schenk & L.Hufford, M. flumensevera (N.H.Holmgren & P.K.Holmgren) J.J.Schenk & L.Hufford, and M. leucophylla Brandegee - were represented by fewer than ten occurrences. These four species have very narrow distributions (one or two US counties) and their occurrence data fell within their documented range (Hufford et al., 1993), so they were retained in the analysis.

We extracted the annual global aridity index from the “Global Aridity Index and Potential Evapotranspiration Database - Version 3” for each occurrence at a resolution of 30 arc seconds using the R package Raster and calculated the mean and standard error for each Mentzelia taxon in our dataset (R. Hijmans, 2024; Zomer et al., 2022). Aridity index is a single value, calculated as the ratio of precipitation to potential evapotranspiration, that provides a baseline measure of the moisture available for plant growth (Zomer et al., 2022). 
We calculated the mean and standard error of the absolute value of latitude for each species. Absolute values were taken so that our metric represented distance from the equator, accounting for the amphitropically disjunct populations of Mentzelia albescens (Gillies ex Arn.) Benth. & Hook.f. ex Griseb in Texas and South America (Hufford et al., 1993).

WorldClim data were associated with each GBIF occurrence at a resolution of 30 arc seconds (about 1 km) using the R packages geodata and terra (R. J. Hijmans et al., 2024, 2025). For each Mentzelia species, the mean of each climate variables was calculated and used in subsequent analyses. We selected eight of the 19 WorldClim variables a priori, excluding those that are quantified by quarter (rather than month) and by ranges rather than extremes or means (Pearse & Hipp, 2012). We employed the following eight variables: BIO1, mean annual temperature; BIO4, temperature seasonality, estimated as the standard deviation of temperature among months; BIO5, maximum temperature, mean temperature of the warmest month; BIO6, minimum temperature, mean temperature of the coldest month; BIO12, mean annual precipitation; BIO13, maximum precipitation, estimated as the mean precipitation of the wettest month; BIO14, minimum precipitation of the driest month; and BIO15, precipitation seasonality, estimated as the coefficient of variation in precipitation across months. 

Life History and Flower Size — Data on life history and flower size were collated first from the Flora of North America treatment of Mentzelia (Hufford et al., 1993) and then from additional species protologues as necessary to complete the dataset for species outside of North America (Darlington, 1934; Schenk et al., 2021; Urban & Gilg, 1900; Weigend, 2007). Life history was scored as annual, perennial, or biennial. As some Mentzelia species have variable life histories (Keeler, 1987), we assembled and analyzed two datasets: one where variable species were assigned the shorter life history and one where they were assigned the longer life history. To approximate flower size, we recorded minimum and maximum petal lengths (in mm, excluding extremes when given) from the literature as described above. Statistical analyses were performed on the mean of these ranges after logging the maximum and minimum values. Error (variability in petal length) was estimated as ¼ the range in floral size for each species. In cases where only single length measurements were available and no range in petal length could be calculated, that species was assigned the median error.