The Yanacocha Reserve, located about 15 km northwest of Quito, the capital of Ecuador (Figure 1A), was created by the Jocotoco Foundation, a World Land Trust Partner spanning 1,200 ha (960 ha of montane evergreen forest and 240 ha of páramo). The study area, embedded in the eastern flank of the Pichincha Volcanic Complex, is located in the western mountain range of the Andes in Ecuador. The western mountain range is found in Ecuador’s north and central regions and is geologically more recent than the eastern mountain range (Hughes and Pilatasig, 2002). It is influenced climatically by air masses that originate in the Pacific Ocean (Vuille et al., 2000). Precipitation varies seasonally—there is more precipitation between January and May and October to December, whereas June through September is the dry season (Navarro-Serrano et al., 2020). The average annual precipitation is 1,500 mm year^–1 (Rugel, 2019) and less than 250 mm for the dry season (Sklenář et al., 2008). Therefore, a decrease in monthly rainfall during the dry season, coupled with increased UV radiation due to cloud-free skies leads to augmented thermal convection, resulting in higher maximum and minimum daily temperatures. Soils within the Yanacocha Reserve study area are Andosols, typical of the high Andes (WRB, 2015) with well-defined layers of volcanic ash intercalated with the organic layers. The topsoil layer in both forest and páramo varies from 15 to 18 cm in depth containing more than 7% of soil organic matter (Calderón-Loor et al., 2020).

The páramo vegetation in the Pichincha Volcanic Complex, dominated by shrubs like Monnina crassifolia (Bonpl.) Kunth and Brachyotum ledifolium (Desr.) Triana, herbs like Hieracium frigidum Wedd. and Lachemilla orbiculata (Ruiz & Pav.) Rydb., and tussock grasses like Calamagrostis intermedia (J. Presl) Steud., starts at ca. 3,900 m asl. At 4,200 m asl—the location of the experimental plots—the dominant growth-forms shift to mainly tussock grasses with interspersed sclerophyllous shrubs [including Chuquiraga jussieui J.F. Gmel. and Loricaria thuyoides (Lam.) Sch. Bip.], and prostrate herbs, acaulescent rosettes and cushions (including Cerastium floccosum Benth., Werneria nubigena Kunth, and Plantago rigida Kunth). Our study area lies in the transition zone between grass páramo and superpáramo.

In 2012, the Yanacocha Reserve became the location of a long-term ecological research site to monitor vegetation changes at different elevations and under different scenarios (Teunissen van Manen et al., 2019, 2020; Calderón-Loor et al., 2020; Llerena-Zambrano et al., 2021). For example, the Yanacocha Reserve houses two of the GLORIA monitoring summits for the Pichincha Volcanic Complex (Cuesta et al., 2020; Tovar et al., 2020). In July of 2012, a passive warming experiment was established. It started only with control plots using a hierarchical block design and later, after the installation of the OTCs in 2012 (Figure 1B), the monitoring of the experimental plots started in 2013.

Five monitoring blocks—installed at 4,200 m asl (−0.13518 Lat, −78.57458 Long)—contained ten units of 6 m × 6 m each, with four sub-units of 3 m × 3 m for every unit. Two sub-units contained the control and experimental plots: (1) one biodiversity control plot of 1 m × 1 m and two control AGB plots of 1 m × 1 m each; and (2) one experimental biodiversity plot of 1 m × 1 m and two 1 m × 0.1 m experimental AGB plots (Figure 1C). Furthermore, OTCs were placed randomly in four units per block, covering the biodiversity and AGB experimental plots. Therefore, four units per block had an OTC sub-unit, making a total of 20 OTC sub-units and 50 control sub-units among all five blocks (Báez et al., 2013). The blocks were placed at a minimum distance of 5 m from each other to reduce spatial autocorrelation. Through incidence-based species accumulation and rarefaction curves—using the functions specaccum and rarecurve both from the Vegan package in R (Oksanen et al., 2020)—the saturation of the curves showed an appropriate sampling effort (Supplementary Figure 1).

To confirm that temperature caused the effects observed in the OTCs and that any change in biomass or diversity would not be attributed to a possible reduction in herbivory—caused by the physical blocking of the OTC walls (Marion et al., 1997)—six herbivore exclusion units were installed among the five blocks in 2012. The exclusions consisted of 1-meter-tall fences buried 30 cm below ground and covered the area of a whole unit within a block. Before installing the exclusion, we cleared all small mammals: genera Thomasomys, Cryptotis, Sylvilagus, and Microryzomys. To prepare the area, we trapped and then released them outside of the exclusions.

Control and OTC biodiversity was monitored once a year between 2014 and 2019 (Table 1) using permanent the 1 m × 1 m plots, sub-divided in 10 cm × 10 cm squares to determine species cover (percent) and composition. In addition, we collected control AGB samples from one of the two designated AGB plots located within one of the sub-units, following Calderón-Loor et al. (2020), once per year between 2012 and 2019 (Table 1 and Figure 1C)—through two strips of 10 cm × 1 m at the top and the bottom of the plot. In the OTC sub-units, we collected AGB from the outer limit of the biodiversity plot, where 1 m × 10 cm strips were harvested outward instead of inward, as was the case for the control plots (Figure 1C). AGB samples were tagged, bagged, and placed inside plastic coolers to preserve the integrity of the collected samples before laboratory processing. In sum, the monitoring included assessing the effect of temperature on the following variables over time: AGB, which refers explicitly to live plus dead plant material collected above ground level, and vegetation composition.

Thermochron ibuttons^® (DS 1922L-F5#) were installed in the soil (at 10 cm in depth) and at 10 cm above the ground to measure soil and air temperature, respectively, in 18 plots (9 for both control and OTCs), making a total of 36 loggers, within blocks 1 through 3. The loggers recorded the temperature every hour. We also placed Hygrochron^® temperature and humidity data loggers (DS 1923-F5#) when installing the blocks, but they failed after the second year of sampling. Therefore, for this study, we focused on temperature data only.

We first weighed the AGB samples in the field with a manual scale to obtain fresh weight. Then we transferred the samples to a laboratory where we sorted them between above ground necromass and live mass. We distinguished between necromass and live mass based on the color and rigidity of the plants—yellow/brown leaves for necromass and rigid green leaves for live mass (Calderón et al., 2013). Samples were then dried at 60°C for 72 h in an oven until they reached a stable weight. Once the samples were dried, we weighed them again separately. This study’s statistical analysis used the sum of the dry aboveground necromass and live mass.

At the QCA Herbarium, we processed and deposited vouchers of plants identified in the field and confirmed the plant identifications (Supplementary Table 1). Identification of plant vouchers were done in the QCA Herbarium where samples were also deposited. We identified 51 out of 70 morphospecies to species level. The rest of the 19 morphospecies were left at their most substantiated level and grouped into family groups. For example, the “Apiaceae group” housed individuals that look like species within the Oreomyrrhis and Niphogeton genera but were not identified up to species. The “Asteraceae group” included individuals that look like species from the Aphanactis, Gamochaeta, and Achyrocline genera. The “Poaceae group” included most species from the Poaceae family except C. intermedia and Calamagrostis fibrovaginata Lægaard, grouped in their own Calamagrostis group (Supplementary Table 1).

We obtained the growth form data for each species and morphospecies from the GLORIA-Andes database (Muriel et al., unpublished data), which uses a revised version of the Ramsay and Oxley classification (Ramsay and Oxley, 1997). We used the following growth form categories: cushion, herb, rosette, shrub, and tussocks.

Compared to controls, daily mean air temperature increased in OTCs by 0.16°C (Table 2). During the dry season, the OTCs had the most increase in air temperature for the daily average maximum by 2.33°C (Table 2). At night, the mean temperature increased for OTCs compared to controls by 0.33°C, and the highest increase was in the average maximum temperature by 1.31°C (Table 2). On the contrary, the OTCs significantly decreased the daily mean soil temperature by −0.08°C (Table 2). The highest increase in soil temperature was for the daily average minimum by 0.32°C (Table 2).

The OTCs significantly increased the mean monthly air temperature by an average of 0.24°C except in September and October, at the end of the dry season (Figure 2A). In contrast, the monthly mean and maximum soil temperature in OTCs were lower than controls from July to October and April to December, respectively (Figure 2B). The OTCs attenuated soil monthly minimums and maximums in most of the year except for the height of the wet season in February and March (Figure 2B).

There were significantly fewer freezing events and of less intensity in OTCs than in control in soil and air during the dry season, but OTCs and control had a similar number of freezing events in the air during the whole year (Supplementary Tables 3, 4). Temperature loggers in the soil recorded zero freezing events during the dry season (daily and nightly; Supplementary Table 3).

Over the entire study period, mean AGB was significantly higher within OTCs (mean = 1.11 kg/m²; sd = ± 0.13) than in control plots (mean = 0.93 kg/m²; sd = ± 0.13) by 9% (W = 9,423; p = 0.029; Figure 3A; Supplementary Table 5). The biggest difference in AGB within the same year between OTC and control was in 2019 with 27.9% (p = 0.007). However, AGB decreased over time in both OTCs and control plots as shown by the negative values of the mixed-effects model coefficients (Figure 3B and Supplementary Table 5). Exclusion plots (mean = 1.24 kg/m²; sd = ± 0.16) did not show significant differences with OTC plots (mean = 1.11 kg/m²; sd = ± 0.16) or control plots (mean = 0.93 kg/m²; sd = ± 0.16; Figure 3B and Supplementary Table 6).

Species diversity (Hill index) in OTCs (mean = 3.66; sd = ± 1.9) was significantly lower than in control over the entire study period (mean = 4.47; sd = ± 1.9; W = 3,499; p = 0.041; Figure 4A). However, this difference is not significant when accounting by sampling time in the mixed-effects model (t = −1.85; p = 0.065). In the mixed-effects model the years 2016, 2017, and 2019 had lower diversity values than 2015 (Figure 4B and Supplementary Tables 7, 8).

Evenness in the OTCs (mean = 0.46; sd = ± 0.17) was significantly lower than in control plots (mean = 0.54; sd = ± 0.16) over the entire study period (W = 3,434; p = 0.0088; Figure 4A). This pattern was further confirmed when time was held as a fixed factor in the mixed model, and evenness for 2016, 2017, and 2019 was lower than in 2015 (Supplementary Table 8 and Figure 4B).

There was no significant difference in species richness between control plots and OTCs (Figure 4A). Only in 2019, species richness was significantly lower in OTCs (mean = 9.6; sd = ± 2.4) than in control (mean = 12.19; sd = ± 2.7) as was suggested by the Wilcoxon test (p = 0.015) and visualized by the mixed-effects model (Supplementary Table 7 and Figure 4B). Nevertheless, by 2019, five out of the 57 recorded species [e.g., Astragalus geminiflorus Bonpl., Azorella pedunculata (Spreng.) Mathias & Constance] were no longer present inside the OTCs, and six more demonstrated a progressive reduction in the warmed plots (Supplementary Table 1). On the other hand, only two species appeared in the OTC plots as the warming experiment progressed without being present in the control plots, Bartsia laticrenata Benth. and G. acostae Cuatrec.

Growth form diversity was significantly lower (p = 0.0026) in the OTC plots (mean = 2.36; sd = ± 0.7) than in control plots (mean = 2.73; sd = ± 0.71) but decreased in both treatments over time (Figure 5A). This difference holds after accounting by sampling time in the mixed-effects model (Figure 5B and Supplementary Table 9). Growth form evenness was also significantly lower (p = 0.003) in OTC plots (mean = 0.56; sd = ± 0.2) than in control plots (mean = 0.66; sd = ± 0.17) over the entire study period and decreased over time in both (Figure 5B). However, after accounting by time in the mixed-effects model, evenness in OTCs was not significantly different from control, and all times had lower evenness compared to 2014 (Figure 5B and Supplementary Table 9).

Percent cover of cushion growth form was significantly lower in OTCs than in control plots, while tussock cover was significantly higher over the entire study period (Figure 6A and Supplementary Table 10). Over time, cushion species cover decreased in the OTCs but increased in control plots (Figure 6B and Supplementary Table 10). Meanwhile, percent cover of tussock species (e.g., Calamagrostis group) increased over time in OTCs by 16.5% but kept similar cover values in control plots (Figure 6B and Supplementary Table 10).

In our study, the OTCs changed the microthermal conditions, potentially leading to shifts in AGB, plant community composition, and growth form diversity. Generally, air temperature and soil temperatures had different responses to the OTCs, where air temperature on average increased and soil temperature on average decreased. Specifically, the thermal conditions changed by first, a general increase in daily and monthly air temperature with the highest increase in means occurring for night temperatures. Second, the dry season accentuated the differences in microclimatic conditions between control and OTC plots, for both air and soil temperatures, where soil temperature in OTCs was mainly lower than control, and air temperature in OTCs had its highest increase in temperature compared to controls. Third, extreme temperatures were attenuated, including reducing freezing events.

Daily (+0.16°C) and monthly (+0.24°C) mean air temperature increased but in a lower amount than what we hypothesized and less than observed in tundra and subtropical grassland environments. For subtropical grasslands, Buhrmann et al. (2016) reported an increase of 1.6 to 2.4°C in air temperature during the day in OTCs compared to controls and 0.0 to 0.6°C during the night across all four seasons. In our study, night temperatures in OTCs were significantly different from control and showed higher mean air temperatures by an average of +0.33°C (p < 0.001), which falls within the range of change in temperate regions. Our results show that in equatorial regions, larger changes occur for night temperatures than from daily temperatures, different from temperate and subtropical alpine areas. Additionally, lower values of temperature change than Lasso et al. (2021) in the Colombian páramos may result from differences in OTC design, such as the type of fiberglass used in their warming experiment.

Our study found smaller differences in the daily mean soil temperature between OTCs and control than in the mean air temperature. This result was similar to the results reported by Pancotto et al. (2020) in southern Patagonia. However, a meta-analysis of 32 experimental sites across four temperate biomes: tundra, low tundra, grassland, and forest found a broader range, of 0.3 to 6.0°C, for mean soil warming temperature increases (Rustad et al., 2001). The studies analyzed in that meta-analysis used different warming methods, not only OTCs, which, along with latitudinal differences, may contribute to the differences from our results. For example, in subtropical grasslands, the absolute maximum temperature in OTCs increased by 0.9 to 1.8°C while the absolute minimum was increased by 0.3 to 0.8°C (Buhrmann et al., 2016). Meanwhile, we obtained a mean maximum temperature increase of 0.55°C and an increase in mean minimum temperature of 0.4°C.

The OTCs stabilized soil temperatures by keeping the monthly minimums and maximums consistent throughout the year with maximum temperatures lower than those in control plots. This reduction in the soil thermal range may reduce plants’ ability to tolerate extreme events under normal conditions (Sierra-Almeida and Cavieres, 2010). Further, soil temperature attenuation also reduces the frequency and intensity of freezing events (Supplementary Table 3). In this study, the mean monthly soil minimum temperature for OTCs was above freezing. This attenuation could be caused by greater evapotranspiration during the day, which could generate a buffer that reduces heat transfer through convection during the night (Sierra-Almeida and Cavieres, 2010). Consequently, the thermal amplitude experienced by plants is reduced, which would alter their physiological response to extreme conditions (Sklenář et al., 2016).

The dry season analyses revealed an important pattern in the temperature trends observed in the OTCs, including the intensification of temperature and shifts in trends around the start and end of the dry season. In our results, at the beginning of the dry season (June), air temperatures in OTCs intensified, obtaining their maximum difference from controls in July by 0.7°C. During June and July, the OTC trapped higher amounts of UV radiation, causing this peak in temperature (Marion and Pidgeon, 1992). Additionally, monthly maximum air temperature in OTCs only fell below the control plots in January and February, during the wet season, where days have higher humidity, cloud cover, and lower radiation (Vuille et al., 2003; Letts and Mulligan, 2005). Studies in tundra found that passive warming treatments increased mean growing season air temperature by 1 to 3°C, a warming range predicted and observed for tundra regions (Walker et al., 2006). Therefore, variations in monthly precipitation and UV radiation over a year-long period are important in describing OTC temperature trends.

Over the study period, OTCs showed higher values of AGB but lower values of species and growth form diversity and evenness than control plots. This pattern seems to be linked to an increase in the vegetation cover of tussocks and a decrease of cushion and herb species in the OTCs, leading to a less diverse community. While AGB reduced in both control and OTCs over the study period, AGB was highest in exclusion plots followed by OTCs, and the lowest means of AGB were found in controls.

By the end of our 7-year experiment, OTCs had 27% more AGB than control plots, as we hypothesized, and which is consistent with other experiments in the alpine meadows of the Hedguan mountains (Yang et al., 2018) and Patagonia (Pancotto et al., 2020). However, other studies have reported no warming impact or decreases in AGB. For example, Fu et al. (2013) found that AGB significantly decreased with experimental warming at 4,313 m asl. This study used different methods than ours and had a much shorter experimental time of 1 year, potentially explaining the difference from our results. Although OTCs generally had more AGB than controls, we observed an overall decrease of AGB for both controls and OTCs. This contrasts with results reported by Calderón-Loor et al. (2020) in the same páramos of the Yanacocha Reserve for the period 2012–2014 in which AGB increased, despite showing similar values of AGB to ours for 2014. Finally, the higher amount of AGB in exclusion plots than OTCs shows that the OTCs did not entirely exclude herbivory by small mammals, as was also observed by Barrio et al. (2021). However, herbivory intensity is dependent on site and OTC design. In our study, given the visual evidence of herbivory in OTCs, more AGB in OTCs may be more likely related to the changes in thermal conditions rather than the potential exclusion effect of herbivores by the OTCs. However, further research on herbivory intensity related to OTCs in tropical alpine regions would help clarify this topic.

The higher AGB in OTCs (by 9%) in our study area is mainly related to increased vegetation cover of tussocks (e.g., C. intermedia), which increased by 16.5% by the end of the experiment and it is a pattern that has been observed in other studies as well. For example, in subtropical alpine grasslands, annual graminoid biomass production significantly increased by nearly 20% in OTCs (Buhrmann et al., 2016). Similarly, Na et al. (2011) found an increase in AGB inside OTCs after 2 years of warming in the Tibetan Plateau, where sedges were responsible for 90% of the total biomass gain. Lastly, in the páramos of Colombia, graminoid and shrub coverage increased inside OTCs in one out of the two study sites (Lasso et al., 2021).

In our study, an increase in tussocks in OTCs could be due to three reasons: (1) tussocks have specific adaptations that allow them to become dominant, (2) the drier environments created by OTCs favor tussocks, and (3) at this point, tussocks are favored but may be replaced later by shrubs and woody plants. First, tussock grasses have broad thermal niches (Cuesta et al., 2020) and are highly tolerant to freezing (Márquez et al., 2006; Rada et al., 2019). Second, they compete by being taller, shading out shade-intolerant species, which leads to local/plot extinctions (Wong, 1991), dispersing by wind (Tovar et al., 2020), and by being photosynthetically efficient (i.e., reduced leaf area index and higher leaf dry matter content (Leegood, 2002; Cuesta, unpublished data). Next, a secondary effect of the OTCs is that they create drier environments (e.g., Lasso et al., 2021), and tussocks are resistant to cold and resistant to drought (Monteiro et al., 2011; Rada et al., 2019). Therefore, tussock grasses seem suited to take advantage of conditions created by increased temperatures coupled with increased water limitation. Lastly, we hypothesized that both graminoid and shrub coverage would increase but only the increase in graminoid cover was significant. In their monitoring studies, Porro et al. (2019) and Hamid et al. (2020) found that the vegetation cover of shrubs and graminoids increased with warmer temperatures. However, in our study, we found that shrubs as a group did not increase, but that individual species, such as C. jussieui, showed an increase in cover. The magnitude of this trend was small and may depend on more prolonged warming exposure periods and higher warming temperatures. Previous studies on the tropical upper forest line suggest that woody plants cannot grow adequately at soil temperatures below 5 or 6°C (Hoch and Körner, 2003; Körner and Paulsen, 2004; Hoch and Korner, 2005; Wiley and Helliker, 2012). Therefore, in the long term, as the world continues to get warmer in the future, shrubs [e.g., Hypericum laricifolium Juss., Hesperomeles ferruginea (Pers.) Benth.] and trees (e.g., Polylepis pauta Hieron., Gynoxys acostae) might outcompete tussock grasses, and there is evidence for this in the paleo records (Flantua et al., 2019). Here our hypothesis was partially met, so given our results, tussock dominance could be a temporary state on the way to a forested landscape.

The increase in percent coverage of tussock species and some shrubs (e.g., C. intermedia, C. jussieui) in OTCs is related to the downward trend observed for diversity and evenness of species and growth form, as we hypothesized. This downward trend is consistent with observations made in a 14-year monitoring study in the Apennines (Porro et al., 2019). Furthermore, Walker et al. (2006) found that warming increased the cover of deciduous shrubs and graminoids and decreased species diversity and evenness. While our results for species evenness and diversity and growth form diversity were significant, species richness was not, yet the downward trend for species richness was similar to that found in previous studies in alpine/tundra regions (Klein et al., 2004; Yang et al., 2018). For example, after the fourth sampling time, five out of the 57 recorded species (e.g., A. geminiflorus, A. pedunculata) were no longer present inside the OTCs, and six more evidenced a progressive reduction in the warmed plots (mainly herbs and cushions). This result was contrary to that of Steinbauer et al. (2018), who found increased species richness in nine regions across Europe over 145 years, mostly related to temperature change. However, after 20 years of monitoring, Steinbauer et al. (2020) showed that previously recorded species richness increases in the Alps were slowing down due to the disappearance of species, specifically cold-adapted species. In our study, decreasing species richness was more in line with this trend. Yet, the slower rate of decrease compared to other OTC studies may be due to the OTCs raising the daily mean temperature less than the increased mean temperature in those studies.

Besides the increase in tussock percent cover, we also observed decreased cushion (e.g., Plantago sericea Ruiz & Pav.), herb (e.g., Geranium humboldtii Willd. ex Spreng.), and rosette cover in the OTCs. Cushion plants are stress-tolerant, which allows them to be a dominant growth form at the highest elevations, along with rosettes, which decreased in both control and OTC plots in our study (Sklenář et al., 2010; Anthelme et al., 2017; Rada et al., 2019). However, when subjected to increased temperatures, Pancotto et al. (2020) found that their study subject Astelia pumila (G. Forst.) R. Br., a cushion plant, decreased their photosynthetic efficiency. Thus, the increase in temperature could be a potential explanation for the decrease in cushion plant coverage in our site. The reduction in cushion plant cover, together with the overall decrease of rosettes, which in tropical alpine ecosystems are characteristic of high elevations (Sklenář and Balslev, 2007; Rada et al., 2019), in OTC and control plots over time, may suggest that a macro-scale process may be already occurring in the Yanacocha Reserve. One potential explanation for this pattern is that rosettes are already migrating upward. On the Antisana volcano, about 63 km southeast of our study site, Moret et al. (2019) compared Humboldt’s collections to a resurvey of the same collection route. This study reported upward shifts for three rosettes [W. nubigena, Valeriana alypifolia Kunth, and Senecio nivalis (Kunth) Cuatrec.] and one herb [Phlegmariurus crassus (Humb. & Bonpl. ex Willd.) B. Øllg.] where S. nivalis expanded upslope 216 to 266 m in 215 years.

Time is an essential variable in warming experiments. OTCs accelerate the rate to visualize change compared to prolonged comparative studies (Porro et al., 2019; Steinbauer et al., 2020; Anthelme et al., 2021), and our study found significant changes after 7 years of sampling. However, studies in non-tropical ecosystems found differences in a much shorter timeframe; for example, studies in tundra found changes in plant communities after two growing seasons (Walker et al., 2006). The distinct seasonality found in tundra and sub-tropical ecosystems, contrasting from the tropics, may make the tropics more resilient to temperature changes and may reveal the effects of climate warming over more time. Furthermore, those studies reported much higher temperature changes than what we obtained in our OTCs, which could also be contributing to the longer response times.