Floral volatile organic compounds sampling and analysis: To analyse differences in flower VOC production across elevational bands, we sampled the headspace of single floral heads (or capitula) of natural populations of Haplopappus species occurring at three distinct elevational zones (2-3 species per elevational band, n=5-10), using polydimethylsiloxane (PDMS)-coated Twisters (Gerstel, 10mm length, 0.5mm film thickness). Sampling occurred during January 2018, between 11 AM and 2 PM, during sunny days. For each Haplopappus species, we randomly selected one floral head per plant with florets in the floral disk in full bloom and bagged it using 1L oven bags (Tangan No34, Turkey). Next, the headspace samples were pumped on the Twisters with a handpump set to 200 mL per minute. The pump was connected to a glass tube within which we inserted the adsorptive Twister, which was inserted into the bag and allowed to collect for two hrs. Additionally, we collected control samples (VOCs collected from the vegetative part of the plant but without the flowers inside the bag) using the same methodology as the floral samples. The hermetically sealed Twisters tubes were stored in a cooler in the field and then stored in a freezer at -20 °C within the same day, where they remained frozen until gas chromatography (GC) analysis. After including 1μL internal standard (5μg mL-1 naphthalene in dichloromethane) directly to each Volatile compounds were thermally desorbed using a Multipurpose Sampler MPS (Gerstel, Mülheim an der Ruhr, Germany), and injected onto a HP-5MS column, 30 m x 0.25mm x 0.25um at 40° C for 30 s. For VOCs separation, the GC oven temperature was increased at 5 °C per minute until 160 °C, which was held for 0.01 min before increasing the temperature again at 3 °C per minute until 200 °C, which was held for four minutes before a final temperature ramp at 100 °C per minute until reaching 250 °C for three minutes. VOCs were then detected by within an Agilent 5975C (Agilent, Santa Clara, CA, USA) mass spectrometer (MS). The mass detector in EI mode at 70 eV was used to scan over the mass range from m/z 33 to 350. To identify the molecules, we compared each peak found in the chromatograms with the NIST05 database. Compounds present in control samples were removed from our final analysis. Tentative compound identification was done using AMDIS and NIST libraries, as well as with comparisons with in- house libraries of mono- and sesquiterpenes mixtures.

Insect preference bioassay: To investigate whether floral heads from different elevation bands were differentially attractive to floral visitors, we measure individual preferences from the most common insect visitors, Dioxyna chilensis and from Trupanea sp. seed predators, towards Haplopappus floral heads growing at the two extremes of the elevation gradient (i.e., low- and high-elevation species). Both fly species have been described in (Soto Andrades, n.d.). We constructed seven arenas by taping a BugDorm mesh lid to the bottom of an inverted a plastic pot (Figure S2). We inserted a pair of inflorescences in each arena with H. foliosus from low elevation and H. scrobiculatus from high elevation. D. chilensis and Trupanea sp. flies emerging from inflorescences collected from all Haplopappus species along the elevation gradient were used for the choice bioassay. Each bioassay consisted of placing a pair of flies, a male and a female of the same species, in an arena for ten minutes and counting the number of landings on each inflorescence (n = 30 pairs for D. chilensis, and 20 pairs for Trupanea sp.). Flies were introduced in pairs to incentivize them to make a choice as both fly species mate in the flowers. Females were identified based on the presence of an ovipositor. In a few instances, one or both flies landed on a flower once for the entire duration of the bioassay. We converted landing counts to proportions so that a fly that could not decide between flowers would not outweigh a fly who made a clear choice from the start. Proportions of landing was estimated by dividing the number of landings on each floral head by the total number of landings. 

Predation vs. fertilized seed counts across species: To assess the potential positive or negative effects of floral visitors, we randomly sampled ten capitula that were at the mature stage across approximately 20 plants for each Haplopappus species (n = 18-34 plants per species, with a total of ~1500 capitula). Capitula were placed individually in plastic tubes with pierced lids, so to let them fully mature under controlled conditions in the laboratory. After a minimum of two weeks, each capitulum, was analysed under a stereo microscope for quantifying fertile (fully swollen) seed production, as well as the number of seeds that were visibly damaged by an herbivory event. In most cases, a small exit hole can be observed in the seed, meaning that herbivory by an insect larva had happened inside the seed. We opted to not quantify unfertile seeds since we could not assess the fertility of half-swollen seeds.