The study site is located within a terra firme (plateau) forest area in the Amazonian rain forest, approximately 150 km northeast of Manaus, Brazil. The average annual rainfall is 2540 mm a-1 (de Ribeiro, 1984), reaching its monthly maximum of ∼ 335 mm in the wet (February to May) and its minimum of ∼ 47 mm in the dry season (August to November; Andreae et al., 2015). These main seasons are linked by transitional periods covering June and July after the wet and December and January after the dry season (Andreae et al., 2015; Martin et al., 2010; Pöhlker et al., 2016). The terra firme forest has an average growth height of ∼ 21 m and a tree density of ∼ 598 trees ha-1, and it harbors around 4590 tree species over an area of ∼ 3.78 × 106 km2, thus comprising a very high species richness compared to other forest types (McWilliam et al., 1993; ter Steege et al., 2013). Measurements were conducted at the research site ATTO (Amazon Tall Tower Observatory; 02∘08.602′ S, 59∘00.033′ W; 130 m a.s.l.), which has been fully described by Andreae et al. (2015). It comprises one walk-up tower and one mast of 80 m each, being operational since 2012, and a 325 m tower, which was erected in 2015. The ATTO research platform has been established to investigate the functioning of tropical forests within the earth system. It is operated to conduct basic research on greenhouse gas as well as reactive gas exchange between forests and the atmosphere and contributes to our understanding of climate interactions driven by carbon exchange, atmospheric chemistry, aerosol production, and cloud condensation.

The parameters temperature and light within and on top of the bryophyte communities and their WC were measured with a microclimate station installed along one evergreen tree of the species Buchenavia parvifolia (Combretaceae) in September 2014 (Fig. S1). The family regularly occurs in the Amazonian rain forest and represents a common genus in tropical America, growing on clayey soil in plateau environments. It presents flowers during the dry and fruits during the wet season (Stace, 2007). The sensors were placed along a vertical gradient at 1.5, 8, and 23 m above the ground on one tree, corresponding to the zones 1, 2, and 4 (i.e., at the base, the lower trunk, and the base of the crown) described by Oliveira and ter Steege (2015), to investigate the variation within the story structure of the forest.

It needs to be mentioned that not only one single species was measured by one sensor but usually several bryophyte species and also other cryptogams, such as lichenized and nonlichenized fungi and algae as well as heterotrophic fungi, bacteria, and archaea, which grow together, forming a cryptogamic community. Thus, the organisms mentioned throughout this paper were the dominating but not solitarily living species. The restriction of the measurements to one individual tree needs to be considered as a complete independence of the replicate sensors could not be assured. However, due to the large effort of such an installation within the rain forest, it was not possible to equip more trees with additional instruments. Thus, the data obtained from the measurements on this individual tree should be considered as exemplary.

The WC sensors were placed in four different bryophyte communities being heterogeneously distributed along three height levels. At 1.5 m height, the WC sensors were installed in communities dominated by Sematophyllum subsimplex (five sensors) and Leucobryum martianum (one sensor), at 8 m in Octoblepharum cocuiense (two sensors) and Symbiezidium barbiflorum (one sensor), and at 23 m in Symbiezidium barbiflorum (three sensors; Figs. S2, S3). The communities used for a placement of the sensors reflect the distribution of bryophytes among height zones in the Amazonian rain forest (Cornelissen and ter Steege, 1989; Oliveira, 2010; Oliveira and ter Steege, 2015; Pantoja et al., 2015). Studies describe that Lejeuneaceae (common liverwort family of the Amazon region comprising the genus Symbiezidium) are more diverse and abundant in the canopy area, while mosses are mainly concentrated at the tree base and trunk in a plateau ecosystem (Campos et al., 2019; Oliveira, 2010, 2018). The species identified by us (Table S1) have also been reported as being frequent at other tropical rain forest sites (Campos et al., 2015; Dislich et al., 2018; Gradstein and Salazar Allen, 1992; Mota de Oliveira et al., 2009; Pinheiro da Costa, 1999). They show different water-holding capacities, which are influenced by their life-form (Lakatos et al., 2006; Romero et al., 2006; Williams and Flanagan, 1996; Proctor, 1990). The liverwort Symbiezidium barbiflorum (Lejeuneaceae) has been described to have the life-form of mats (Batista and Santos, 2016; Mägdefrau, 1982; Valente et al., 2017), which are characterized by a high capillary retention of water, supporting the storage of condensed water. Mats also have an increased drought tolerance, being more adapted to dry conditions as well as to extreme changes (Gimingham and Birse, 1957). Sematophyllum subsimplex (Sematophyllaceae) and Leucobryum martianum (Dicranaceae) belong to the life-forms of wefts and turfs, respectively (Mägdefrau, 1982; Batista and Santos, 2016; Valente et al., 2017). Turfs show high capillary water conduction and are well known for special water-retaining cells, whereas wefts show high values of capillary water conduction but lower values of water retention (Mägdefrau, 1982), being characteristic for humid areas (Gimingham and Birse, 1957).

Additionally, at each height level two temperature and two light sensors (except for 1.5 m, with only one light sensor) were installed in or on top of the bryophyte communities located on the approximately 26 m high tree (Fig. S2, Table S1). The temperature sensors were installed in the same communities as the WC sensors, and the light sensors were installed adjacent to them on ∼ 5 cm long sticks (Fig. S1). As the morphology of the different species affects their overall WC, different maximum WC values were observed (Tables S1, S2). At 1.5 and 8 m the sensors were installed vertically along the trunk, whereas at 23 m they were placed at the upper side of a horizontal branch. Thus, also the orientation at the stem may influence the WC of the bryophyte communities, not only the species and the height above ground. Furthermore, sample properties such as their thickness and density might play a relevant role for their WC as samples at 1.5 m height tended to be more loose and thinner as compared to the ones at the upper height levels (Fig. S4). Since the installation, automatic measurements at 5 min intervals were taken with a data logger (CR1000; Campbell Scientific, Logan, Utah, USA) equipped with a relay multiplexer (AM16/32; Campbell Scientific, Bremen, Germany) and two interfaces.

The WC sensors, initially developed for biological soil crust research (Tucker et al., 2017; Weber et al., 2016), were optimized for measurements in epiphytic bryophyte communities by a straight-lined construction and with outer pins of 25 mm length, serving as an effective holdfast. However, during stormy episodes and/or physical friction, some WC and temperature sensors fell out of the bryophyte samples and required a reinstallation. Additionally, during some episodes the sensors showed unreliable data, which had to be removed from the overall dataset. All data that could be used for data analysis are shown in Fig. S5.

The electrical conductivity (EC) values, on which the WC calculations were based (see Sect. 2.3 below), showed some unexplained oscillation, causing an inaccuracy corresponding to approximately 20 mV. Besides the specific position in the substrate, the EC also depends on the texture of the sample material, its ion concentration, and the temperature. Because of all these factors influencing the sensor readings, the provided values of the WC should be considered as the best possible estimates and not as exact values. For the temperature measurements, thermocouples (Conatex, St. Wendel, Germany) with a tip length of 80 mm and a measurement accuracy of ± 0.5 ∘C were used. For the light sensors, GaAsP photodiodes (G1118, Hamamatsu Photonics Deutschland GmbH, Herrsching, Germany) were placed in a housing covered by a convex translucent polytetrafluoroethylene (PTFE) cap and calibrated against a PAR (photosynthetically active radiation) quantum sensor (SKP215; Skye Instruments, Llandrindod Wells, Powys, UK).

The average daily PAR values were calculated from the data collected during daytime, i.e., 06:00 to 18:00 LT (local time), while PARmax represents the daily maximum value. The 5 min readings obtained from the light sensors fluctuated by approximately ± 10 µmol m-2 s-1 photosynthetic photon flux density (PPFD). To smoothen the microclimate data (i.e., the PPFD values obtained from light sensors, the temperature values measured within the bryophytes, and water content values obtained from electrical conductivity sensors), 30 min averages were calculated and used for all further calculations. During measurements, the light sensors were regularly checked for algal growth and cleaned accordingly.

The WC sensors measure the electrical conductivity in the field (ECt), which is influenced by temperature; consequently, a temperature correction was performed according to Eq. (1), analogous to Weber et al. (2016). 1EC25=fT⋅ECt, with EC25 as EC at 25 ∘C, T as bryophyte temperature (∘C), and the temperature conversion factor fT: 2fT=0.447+1.4034e-T/26.815.

The WC sensor has a fixed distance between the sensor pins, which ensures that in all sensors the resistance is equal. This guarantees that the electrical voltage, being the inverse resistance, is proportional to the electrical conductivity. The values of the sensors were recorded as electrical voltage in millivolts and by calibration transformed into the WC of the samples, given as dry weight percentage, as explained below.

To determine the maximum water content of the different bryophyte communities, samples of them were collected in the forest area surrounding the ATTO site. They were removed from the stem with a pocket knife and stored in paper bags in an air-conditioned lab container until calibration (a few hours after collection). Prior to the calibration, the samples were cleaned from adhering material using forceps. The weight of the bryophytes was determined when they were moistened until saturation (temperature 30 ∘C; relative humidity, RH, 100 %) and again after drying in a dryer overnight (temperature 40 ∘C, RH 30 %) to simulate the natural range of the WC under controlled temperature and RH conditions. The dry weight (DW) was determined after drying at 60 ∘C until weight consistency was reached (Caesar et al., 2018). The WC of the sample was calculated in analogy to Eq. 3 in Weber et al. (2016): 3WC%DW=(FW-DW)DW⋅100%, with FW being the fresh weight (g) and DW the dry weight (g) of the sample.

In a previous approach, calibration curves were established under controlled conditions, logging the electrical conductivity values and the corresponding weight and water content of samples of the different bryophyte species during drying, analogous to Weber et al. (2016). However, the variability of electrical conductivity values between samples and even at different spots within one sample turned out to be too large, and thus this was not a feasible approach to calibrate the sensors. On the other hand, the electrical conductivity values decreased in a linear way with decreasing sample weight, demonstrating that a linear relationship between both factors could be assumed (except for water content close to saturation).

In the current approach, the calibration of the water content was performed based on the maximum and minimum values of electrical conductivity reached in the field and the maximum WC reached during the laboratory measurements. We assumed that the maximum electrical conductivity achieved in the field corresponds to the maximum WC achieved in the laboratory due to water saturation of the samples during the laboratory measurement. The minimum electrical conductivity achieved in the field was used as an approximation (simplification) of a water content of 0 %. We are aware of the fact that bryophytes do not reach a water content of 0 % under field conditions, but they can be expected to dry to a water content of a few percent, which we assume to be close enough for this model assumption in a tropical rain forest environment. Accordingly, the WC was calculated as follows: 4WC[%DW]=WCmaxECmax-ECmin⋅ECi-ECmin, where WCmax corresponds to the maximum WC measured in the laboratory, ECi is the electrical conductivity, ECmin is the minimum electrical conductivity, and ECmax is the maximum electrical conductivity measured in the field.

The measured electrical conductivity values showed short-time oscillations, which might be caused by the fact that the bryophyte cushions have some air spaces in-between as we observed that these oscillations are less pronounced in denser substrate. Nevertheless, the overall functionality of the sensors is still ensured also in less dense material, and the short-term fluctuations could be removed with a 30 min smoothing algorithm. Thus, for all calculations the 30 min averages were used. The electrical conductivity data of replicate samples at the same height were combined to obtain average values for each height.

For the purpose of long-term monitoring, a set of meteorological parameters has been being measured within the ATTO project since 2012. In our study we used rainfall data measured at 81 m (mm min-1; rain gauge TB4, Hydrological Services Pty. Ltd., Australia), RH measured at 26 m (%), air temperature measured at 26 m (∘C; thermohygrometer CS215, Rotronic Measurement Solutions,UK), and photosynthetically active radiation (PAR) measured at 75 m height above the ground (µmol m-2 s-1 PPFD; Quantum sensor PAR LITE, Kipp & Zonen, Netherlands). All data were recorded at 1 min intervals with data loggers (CR3000 and CR1000, Campbell Scientific, Logan, Utah, USA) on the walk-up tower (Andreae et al., 2015).

For calculation of the average light intensities per month, season, or year (PARavg month, PARavg season, PARavg year), only values during daytime (06:00–18:00 LT) were considered. Rainfall data are presented as accumulated values in millimeters per month, season, or year, which were calculated by an integration of 30 min intervals. As there were gaps in the data record of the rain detection, additional information from the WC sensors was used to calculate the number of days with rain events. The sensors at 1.5 m height were found to react reliably to rain events. Thus, the gaps in the rain detection were corrected with the information received from these sensors. Furthermore, the amount of rain within each month was corrected by assuming that during the missing days there were the same amounts as during the rest of the month. Overall, a malfunction of the rain detection was observed on only 6 % of the days (Table S3).

The information on fog events was provided by visibility measurements using an optical fog sensor (OFS; Eigenbrodt GmbH, Königsmoor, Germany) installed at 50 m height. Fog was defined to occur at visibility values below 2000 m.

In order to assess the potential water input by condensation, we calculated the dew point temperature, at which saturated air humidity levels are reached. If the temperature drops below the dew point, condensation might occur and form liquid water. The dew point spread is the temperature difference between a surface temperature and the dew point of the surrounding air and can be used to assess potential condensation processes at surfaces. If the difference between the surface temperature Ts and the dew point of the surrounding air Td is negative, water vapor is able to condense at the colder surface. The calculations were performed according to Eqs. (5) and (6): 5ΔTd=Ts-Td and 6Td=241.2ln⁡φ100+4222.03716ϑ241.2+ϑ17.5043-ln⁡φ100-17.5043241.2+ϑ, where ΔTd is the dew point spread (∘C), Ts is the surface temperature (bryophytes sensors, 23 m or 1.5 m; ∘C), Td is the dew point temperature air (∘C), ϑ is the temperature (∘C), and φ is the relative humidity (%). The dew point spread calculation was performed for the levels of 23 and 1.5 m height. As a data source for air temperature and humidity, meteorology data assessed at the walk-up tower at 26 and 1.5 m were used. For surface temperature (Ts), the temperature data measured within the bryophyte communities at 23 and 1.5 m height were applied. To analyze if condensation might influence the WC in the bryophytes, a correlation between events with negative ΔTd and the change in the WC in the 4 h before these events was calculated. The potential effect of fog was analyzed by calculating the average change in water content (ΔWC) from the beginning of the fog event until 1 h later. The dew point and fog calculations and correlations were performed with R version 3.6.1 (5 July 2019).

Time readings are always presented as UTC (universal coordinated time) values, except for diurnal cycles, where local time (i.e., UTC-4) is shown, as labeled in the figures.

The physiological activity of bryophytes – and of cryptogams in general – is primarily controlled by water and light, whereas temperature plays a secondary role, at least in the environment of the central Amazon (Lösch et al., 1994; Wagner et al., 2013). While the availability of water determines the overall time of physiological activity, the light intensity regulates whether net photosynthesis (NP) or dark respiration (DR) dominates the overall metabolic balance. Furthermore, high nighttime temperatures cause increased carbon losses due to high respiration rates, as previously shown for lichens (Lange et al., 1998, 2000). For tropical bryophytes along an altitudinal gradient in Panama, however, it has been shown that respiration loss during the night might not play a determining role for an overall positive net carbon balance as species acclimatized to elevated temperatures but that the restricted time for photosynthesis was a decisive factor (Wagner et al., 2013).

To assess the potential physiological activity of bryophyte communities, the water and light conditions as major drivers of the metabolism were investigated in somewhat greater detail. The lower water compensation point (WCP) presents the minimum WC that allows positive net photosynthesis. For the tropical liverwort Symbiezidium spp., occurring in the lowlands near sea level in Panama, WCP values in the range between ∼ 30 % and ∼ 80 % have been determined (Wagner et al., 2013; Table S4).

The lower light compensation point (LCPl) represents the minimum light intensity that allows a positive primary production; it ranges between ∼ 3 and ∼ 12 µmol m-2 s-1 for bryophytes (based on measurements of Ectropothecium cf. perrottii, Frullania spec, Neckera spec., Plagiochila divergens, Plagiochila squamulosa, Porothamnium stipitatum, Porotrichum molliculum, Racopilum tomentosum, Radula boryana, Rhizogonium spiniforme) occurring in African tropical lowland rain forests (Lösch et al., 1994). The epiphytic bryophytes grew in an upper lowland rain forest in the Kahuzi-Biega National Park (Zaire) at about 800 m a.s.l. Microclimatic conditions inside the forest were similar to the conditions at the ATTO site as RH ranged from around 60 % to 70 % during sunny days, and temperatures remained above 20 ∘C during night and day. At light intensities below the LCPl and WC below the WCP, respiration rates are higher than NP rates, causing overall net respiration to occur.

With regard to temperature, a range for optimum NP (Topt) and an upper compensation point, where NP equals DR (TCP), can be defined. For tropical bryophytes (i. e., the species Octoblepharum pulvinatum, Orthostichopsis tetragona, Plagiochila sp. 1, Stictolejeunea squamata, Symbiezidium spp., Zelometeorium patulum), Topt ranges between 24 and 27 ∘C, and the TCP ranges between 30 and 36 ∘C, as described by Wagner and coauthors (Wagner et al., 2013). For long-term survival and growth, the bryophytes need to be predominantly exposed to temperatures below the upper compensation point, at least under humid conditions. The measurements performed by Wagner et al. (2013) were conducted at a study site (BT) in a lowland rain forest in western Panama on the Bocas del Toro archipelago, located approximately at sea level. The mean temperature was 25 ∘C (26 ∘C during day, 24 ∘C during night), thus slightly warmer than the temperatures measured at ATTO. With 3300 mm a-1 of rain, BT is in a similar range as the ATTO site (2540 mm a-1). Unfortunately, literature data on the compensation points are rare, facilitating only a first approximate assessment of the physiological processes (Lösch et al., 1994; Wagner et al., 2013).

A WC above the compensation point allows NP if both light intensity and temperature are above the lower compensation point. If WC is above the compensation point, but light intensities are too low, or if temperatures are above the upper compensation point, net DR occurs. There is also a narrow span of low WC when samples are activated already, but despite sufficient light intensities only net respiration can be measured. As this span of WC is narrow and respiration rates are low, it has been neglected in the current calculations. The compensation points for the different parameters are also to some extent interrelated; e.g., the water compensation point of lichens has been shown to slightly increase with increasing temperature (Lange, 1980), but this can be neglected in such a first qualitative approach. Finally, also inter- and intraspecific variation in compensation points could not be considered in the current study.

All data processing steps and analyses were performed with the software Igor Pro (Igor Pro 6.37, WaveMetrics. Inc, Lake Oswego, Oregon, USA). For the average values obtained at the different height levels, the data of the individual sensors were pooled.

Whereas overall light intensities at the upper two height levels were rather similar, with values below 108 and 147 µmol m-2 s-1 (at 8 and 23 m height) 90 % of the time, the values at 1.5 m height remained below 10 µmol m-2 s-1 during the same time fraction (Fig. 4). In contrast to that, maximum light intensities were similarly high, reaching 1550 (1.5 m), 1500 (8 m), 1040 (18 m), and 950 µmol m-2 s-1 (23 m). If we assume a low light compensation point (LCPl) ranging between 3 and 12 µmol m-2 s-1 (Lösch et al., 1994), the understory samples (1.5 m) exceeded that range only less than 20 % of the time during both the wet and the dry season, whereas at the uppermost height level the bryophytes exceeded these values one-third to almost half of the time (Table 3).

The temperatures inside the moss stands at different height levels mainly ranged between 22 and ∼ 28 ∘C during the wet and between 23 and ∼ 33 ∘C during the dry season (Fig. 4). For tropical lowland regions, the optimum temperatures for bryophytes (Topt) have been suggested to range between 24.0 and 27.0 ∘C (Wagner et al., 2013). If we assume this range for our study, the temperatures in the understory (at 1.5 m height) remained within these limits 91 % of the time during the wet season and 76 % of the time during the dry season (Table 3). In the canopy (at 23 m height), temperatures remained in this range 70 % of the time during the wet and 46 % of the time during the dry season (Table 3). For an upper temperature compensation point (TCP), above which respiration exceeds photosynthesis, Wagner and coauthors suggested 30.0–36.0 ∘C (Wagner et al., 2013). In the understory, this TCP was never surpassed during the wet season and only rarely during the dry season. Similarly, at the uppermost canopy level, the upper TCP was surpassed 0 %–3 % of the time during the wet and 0 %–27 % of the time during the dry season. Overall, the highest temperatures were reached when the bryophytes were relatively dry and most probably inactive (Fig. S13). The WC of bryophytes differed along the vertical profile, with smaller values in the understory at 1.5 m height than in the canopy (8 m and 23 m), particularly during the dry season (Fig. 4). Considering a lower water compensation point (WCP) between 30 % and 80 % according to the literature (Wagner et al., 2013), bryophytes at the uppermost level surpassed these values 3 %–80 % of the time during the wet and 6 %–64 % of the time during the dry season. Contrastingly, at 1.5 and 8 m height there were larger differences between seasons as the assumed upper range of the WCP was only rarely reached during the dry but ∼ 40% of the time during the wet season (Fig. 4, Table 3).

In the current study we measured the microclimatic conditions experienced by epiphytic bryophyte communities along a vertical gradient over the course of more than 2 years. In previous studies, microclimatic data on the light, temperature, and air humidity have been assessed at different height levels within the forest (Chazdon and Fetcher, 1984; Lösch et al., 1994; Romero et al., 2006), but long-term measurements of the water content and the light and temperature on top and inside the cryptogamic communities have been missing up to now.

The microclimatic conditions experienced by epiphytic bryophyte communities along a height gradient at the ATTO site followed the meteorological parameters to some extent, but they also revealed microsite-specific properties regarding annual, seasonal, and diel microclimate patterns. Whereas the water content and the temperature mostly followed the patterns of the meteorological parameters precipitation and temperature, the light intensities were clearly altered, particularly in the understory, due to the local canopy structure.

Within one height level, the small-scale environmental conditions (such as radiation and shading), water conditions, and wind velocity vary, depending on the specific habitat conditions, such as, e.g., exposition, tree foliage, and inclination of the substrate (Barkman, 1958; Campos et al., 2019; Cornelissen and ter Steege, 1989; Oliveira and Oliveira, 2016; Sierra et al., 2018). These small-scale patterns also explain the variability observed within one height level.

Over the measurement period, the monthly averages of above-canopy light conditions (PARavg) were rather stable (Fig. 1, Table S3). Within the canopy, the monthly PARavg values at 23 m height tended to be higher during the dry seasons, whereas patterns were less clear at 8 m height, and there was hardly any seasonal variation at 1.5 m height. This was most probably an effect of the canopy structure, cushion orientation, and shading. The sensors at 1.5 and 8 m were installed vertically along the trunk, whereas at 23 m they were positioned on the upper side of a horizontal branch. As the light sensors at 23 m height were located within the canopy, newly growing leaves may have periodically shaded the organisms, which may explain the lower monthly PARavg values at this height level compared to the values at the lower levels, where sunbeams could come through the canopy of neighboring trees and open space.

The diel patterns of PARavg are expected to show a decreasing gradient from the canopy to the understory as the canopy receives the most solar radiation, while the understory vegetation is expected to be shaded by foliage and branches. During the dry season this general pattern was indeed observed, whereas during the wet season mean light intensities were often higher at 8 m than at 23 m, probably also caused by canopy shading effects at the upper two height levels (Fig. 2). High light intensities above 1000 µmol s-2 s-1 occurred in the understory only as small light spots of short duration and thus were only observed 0.008 % of the time. For the understory of a rain forest in Costa Rica, light intensities were reported to range from 10 to 1000 µmol m-2 s-1, and more than in 50 % of the total amount of light resulted from sun flecks (Chazdon and Fetcher, 1984). Bryophyte and lichen taxa in the understory are known to be adapted to these low-light conditions and are able to make efficient use of the rather short periods of high light intensities (Lakatos et al., 2006; Lange et al., 2000; Wagner et al., 2014).

The temperatures measured inside the bryophyte communities followed the above-canopy temperature at all height levels, with a mostly increasing gradient from the understory towards the canopy, probably caused by a reduced shading effect towards the canopy (Fig. 1, Tables 1, 2). At the uppermost height level, mean temperatures inside the bryophyte communities were often even higher than the mean above-canopy temperatures. During the wet season, the overall temperature conditions were more buffered due to reduced incoming radiation caused by clouds and a frequent mixing of the air masses during rain events (von Arx et al., 2012; Gaudio et al., 2017; Thompson and Pinker, 1975).

The microclimatic mean temperature differences measured inside the bryophyte stands between the understory (1.5 m) and the canopy (23 m) were 1.5 ∘C in the dry and only 0.5 ∘C in the wet season. Compared to these results, a temperature difference of 4.0 ∘C was determined during the dry season in a tropical evergreen forest in Thailand, while in the wet season it was below 1.0 ∘C, thus corresponding quite well to our results (Thompson and Pinker, 1975; Table 2). The daily amplitude of the temperature was about twice as large in the canopy as compared to the understory (Fig. 3). This could be caused by the exposure to strong solar radiation and higher wind velocity in the canopy compared to the sheltered understory (Kruijt et al., 2000).

Rainfall amounts and relative air humidity values differed between the seasons and also between the years as they were considerably higher between October 2014 and February 2015 as compared to the following year. This was most probably due to an El Niño event, which caused air humidity and WC of bryophytes to be substantially lower compared to the previous dry season (Fig. 1, Table 1). Generally, the moisture conditions, including rain, condensation of dew and fog, and RH, differed between seasons, resulting in different WC patterns of bryophytes. A higher frequency of rain during the wet season particularly affected the bryophyte communities at the lower levels (1.5 and 8 m), whereas those in the canopy showed similar water content, during all seasons (Figs. 2, S7a, Table 2).

The data also suggest that the position of the measured communities and the tree foliage played a crucial role for rainwater absorption and the subsequent drying process. Whereas the sensors at 1.5 and 8 m height responded quite reliably to rain events during all seasons, those at 23 m responded quite reliably during the dry but only rarely during the wet seasons. This might be caused by dense foliage during the wet season, shading the communities at 23 m from direct inundation, whereas during the dry season the canopy foliage changes a lot, and intense leave shedding takes place before new leaves develop, which seems to allow rain to also reach the samples directly below the canopy (Lopes et al., 2016). Apart from the effect of growing foliage, one has to keep in mind that there was a reinstallation of the sensors around July 2015, which could cause differences in the sensor readings before and after this event.

During the wet season, the WC of bryophytes in the understory and at 8 m height responded strongly to rain events, and subsequently, the water was lost gradually, with bryophytes staying wet and active over prolonged time spans, indicating that large amounts of water were taken up during prolonged rain events (Figs. 2, S7). In contrast to that, during the dry season the drying of the samples occurred quite rapidly after the rain. Most rain events in the central Amazon occur in the early afternoon (12:00–14:00 LT), and more than 75 % of them are weak events of less than 10 mm (Cuartas et al., 2007), which often cause no complete water saturation of the bryophytes. Consequently, the organisms tend to dry much quicker than after strong rain events. Also here, a potential effect of the sensor reinstallation around July 2015 has to be kept in mind.

In this rain forest environment, epiphytes growing in different parts of the tree and along the stem can benefit from different sources of water. The gross precipitation, as the main water source, can be converted into throughfall, stemflow, water storage, and water vapor (Mendieta-Leiva et al., 2020). Thus, rainwater can influence the bryophytes in various ways depending on its redistribution and the microenvironmental conditions: at the canopy level, direct interception of precipitation can be used for water storage, whereas in the understory stemflow is more pronounced and contributes to the water supply of the bryophytes. It has been estimated that in tropical forests, up to 4 % of the annual rainfall amount could be converted into stemflow (Lloyd and Marques, 1988; Marin et al., 2000; van Stan and Gordon, 2018), corresponding to maximum values of 68 and 75 mm in the years 2015 and 2016, respectively, at the ATTO site.

The water content data at 8 and 23 m height showed diel fluctuations, which were particularly regular during the dry season due to rare interfering rain events (Fig. S5c). They showed a parallel behavior to the RH data, with the highest values reached during the morning hours (Fig. S11). It is well known from the literature that moist bryophytes and many cryptogams could utilize high air humidity as a source of water (Lange et al., 2001; Raggio et al., 2017), and this likely also occurs here.

In addition, also condensation and fog need to be considered as potential additional sources of water for epiphytic covers as well as for near-stem vegetation at the forest floor (Lakatos et al., 2012; van Stan and Gordon, 2018.; León-Vargas et al., 2006). Our data show that the necessary conditions for condensation were regularly met and occurred most frequently during the wet season at 23 m height (Fig. S11). Cases with a negative dew point spread, when condensation could occur, were related to increasing water content of the bryophytes, supporting the calculated condensation data. During the occurrence of fog, an increase in water content could not be directly proven, suggesting that fog does not represent a major water source for the bryophytes.

The microenvironmental conditions influence the WC of epiphytic bryophyte communities, but the ability to deal with these conditions differs among species (interspecific variability), being determined by morphological and physiological features. Apart from the interspecific variability, the performance of a single species under differing microenvironmental conditions can also be modulated by short-term acclimation and long-term adaptation processes, with the latter being driven by environmental exposure, genetic variation among populations, and plasticity, as, for example, shown for bryophytes and lichens (Cornelissen et al., 2007; Marks et al., 2019; Pardow et al., 2010). These aspects help to understand the occurrence of bryophytes under widely varying microclimatic conditions within the canopy. During our study, we measured the microenvironmental conditions of epiphytic bryophytes and observed bryophyte taxa to vary depending on these. Additionally, we estimated the potential ranges of physiological activity based on the compensation points for light, temperature, and WC, which have been reported from other studies in tropical forests (Lösch et al., 1994; Wagner et al., 2013).

In the canopy it is essential for the cryptogams to be adapted to high-light conditions and UV radiation in order to avoid cell damage by radiation (Green et al., 2005; Pardow and Lakatos, 2013; Sinha and Häder, 2008; Westberg and Kärnefelt, 1998). As high-light conditions mainly occur as short light flecks in the understory, the organisms need to react rapidly and efficiently to changing light conditions to reach overall positive net photosynthesis rates. Furthermore, understory bryophytes and lichens show higher rates of net photosynthesis at low-light conditions as compared to canopy species (Kangas et al., 2014; Lakatos et al., 2006; Wagner et al., 2013). Epiphytic organisms growing under low-light conditions in the understory are also known to have lower LCPl values compared to the ones in the canopy, as documented for epiphytic lichens in French Guiana (Lakatos et al., 2006).

The temperature regulates the overall velocity of metabolic processes. While it has a strong impact on the respiration, the photosynthetic light reaction is by far less affected by it (Elbert et al., 2012; Green and Proctor, 2016; Lange et al., 1998). As the measured net photosynthesis rates are the sum of simultaneously occurring photosynthesis and respiration processes, positive net photosynthesis may still be reached at higher temperatures if the photosynthetic capacity is high enough, whereas during the night, high temperatures could cause a major loss of carbon due to high respiration rates (Lange et al., 2000). In the course of our study, the lowest temperatures predominantly occurred during the night, contributing to lower respiration rates, and values were mostly below the upper TCP. Thus, the temperature did not seem to be a limiting factor for the physiological activity of epiphytic bryophytes in this environment (Fig. S11). Similarly, Wagner and coauthors (Wagner et al., 2013) stated that the temperature likely was not a limiting factor for the overall carbon balance of the bryophytes investigated in a low- and highland rain forest in Panama.

Utilizing the compensation points of water, light, and temperature taken from the literature, one can make rough estimates of the time fractions when NP and DR occur at the different height levels (Table 3).

These data suggest that at the upper height level, NP occurred 1 %–30 % and DR 2 %–52 % of the time during the wet season and 1 %–24 % (NP) and 4 %–45 % (DR) of the time during the dry season, respectively (Table 3). These estimates suggest that the duration of DR was about twice as long as that of NP. For the samples at 8 m height, the results were similar, whereas for those in the understory, the duration of DR was about 5- to 30-fold higher than the duration of NP. The large discrepancy between the time ranges for NP and DR calculated for the bryophytes in the understory gives reason to expect the LCPl and the WCP to be at the lower end of the range (3 µmol m-2 s-1, 30 %) for the bryophytes at the lowest height level and to be at the upper end of the range (12 µmol m-2 s-1, 80 %) for the bryophytes at the two upper height levels. For other habitats, LCPl values as low as 1 µmol m-2 s-1 have been defined for lichens (Green et al., 1991), and thus it could be possible that the bryophyte communities in the understory exhibit similarly low LCPl values.

In the environment being studied, the acclimation of the organisms to the environmental conditions is also crucial for their survival. Thus, the time ranges of metabolic activity are only rough estimates, depending on the actual compensation points, which are influenced by inter- and intraspecific variation. There are also some differences between groups as, for example, lichens tend to perform photosynthesis at lower WC than bryophytes, and chlorolichens (with green algae as photobionts) may utilize high air humidity, whereas cyanolichens (cyanobacteria as photobiont) need liquid water (Green et al., 2011; Lange and Kilian, 1985; Raggio et al., 2017). Furthermore, there are also differences between the bryophyte divisions of mosses and liverworts, and also within one division the interspecific variability can be large.