The decline of alpine lichen heaths generates atmospheric heating but subsurface cooling during the growing season

. Lichen heaths are declining in abundance in alpine and arctic areas partly due to an increasing competition with shrubs. This shift in vegetation types might have important consequences for the microclimate and climate on a larger scale. The aim of our study is to measure the difference in microclimatic conditions between lichen heaths and shrub vegetation during the growing season. With a paired plot design, we measured the net radiation, soil heat flux, soil temperature, and soil moisture on an alpine mountain area in south Norway during the summer of 2018 and 2019. We determined that the daily net 15 radiation of lichens was on average 3.15 MJ (26%) lower than for shrubs during the growing season. This was mainly due to a higher albedo of the lichen heaths, but also due to a larger longwave radiation loss. Subsequently, we estimate that a shift from a lichen heath to shrub vegetation leads to an average increase in atmospheric heating of 3.35 MJ per day during the growing season. Surprisingly, the soil heat flux and soil temperature were higher below lichens than below shrubs during days with high air temperatures. This implies that the relatively high albedo of lichens does not lead to a cooler soil compared to 20 shrubs during the growing season. We hypothesize that the thicker litter layer, the presence of soil shading, and a higher evapotranspiration rate at shrub vegetation are far more important factors in explaining the variation in soil temperature between lichens and shrubs. Our study shows that a shift from lichen heaths to shrub vegetation in alpine and arctic areas will lead to atmospheric heating, but has a cooling effect on the subsurface during the growing season, soil underneath the lichens has a higher soil temperature and a higher soil heat flux than the soil below shrubs, especially during warm days. This implies that the relatively high albedo of lichens affects the radiation balance, but not the subsurface microclimate. Potential reasons for this could be the thicker litter layer, shading by the canopy or more evapotranspiration in the shrub plots. We conclude that the decline of lichens due to shrub expansion will lead to atmospheric heating (i.e. higher latent + sensible heat flux), but has a cooling effect on the subsurface during the growing season. Future studies should focus on the quantification of the effect of lichen decline on the climate on a regional and possibly on the scale of the arctic.


Introduction
Lichen heaths are one of the most dominant vegetation types across alpine and arctic areas (Cornelissen et al., 2001). For example, lichen heaths cover up to 6% of Norway (Bryn et al., 2018). Besides their extensive abundance, lichens are important forage for reindeer during winter (Heggberget et al., 2002;Vistnes and Nellemann, 2008). However, the lichen cover has decreased in alpine and arctic areas during the last decades (Cornelissen et al., 2001;Joly et al., 2009;Elmendorf et al., 2012;30 Lang et al., 2012;Fraser et al., 2014;Maliniemi et al., 2018). For instance, Fraser et al. (2014) estimated that lichen cover decreased by 24% in the western Canadian Arctic between 1980 and 2013. Also, large continuous lichen mats are rarely observed anymore in this region, while they were common 40 years ago (Fraser et al., 2014). Similar declining trends have been observed throughout the alpine and arctic areas. The lichen decline is attributed to an increased competition with vascular plants that benefit from climate change, especially shrubs (Cornelissen et al., 2001;Fraser et al., 2014;Moffat et al., 2016;35 Vuorinen et al., 2017;Chagnon and Boudreau, 2019). Experimental warming studies show that this lichen decline has the potential to proceed with the ongoing temperature increase (Walker et al., 2006;Elmendorf et al., 2012). Therefore, it is important to study the consequences of the lichen decline on alpine and arctic ecosystems.
Shrubs benefit from recent climate change, since the higher temperatures and longer growing seasons are in favor of their growing conditions (Myers-Smith et al., 2011;. Indeed, many studies found an increase in shrub 40 cover, biomass and abundance in alpine and arctic areas over the past decades (Sturm et al., 2001b;Hallinger et al., 2010; see Myers-Smith et al. (2011) for a review). Such an increase of shrubs alters the vegetation composition in these areas (Pajunen et al., 2011;Boscutti et al., 2018). For example, multiple studies have reported a negative relationship between shrubs and lichen occurrence (Cornelissen et al., 2001;Pajunen et al., 2011;Maliniemi et al., 2018). Moreover, Chagnon and Boudreau (2019) found a lower lichen abundance and diversity below shrubs compared to areas without shrubs. These studies imply that 45 shrub vegetation outcompetes the lichens heaths in the long run. This might alter the alpine and arctic environment in various ways, since lichens and shrubs have distinct characteristics. For example, Aartsma et al. (2020) measured an average albedo of 0.255 for lichen heaths, while the average albedo of shrubs was 0.132. Therefore the shift from lichen-dominated areas to shrub-dominated areas might have, among others, important consequences for the microclimate and the large-scale climate of alpine and arctic areas. 50 Extensive studies have shown that shrub expansion has a substantial impact on microclimatic conditions, including surface albedo, soil temperature and permafrost stability (Myers-Smith et al., 2011;Loranty et al., 2018). Chapin et al. (2005) estimated that a shift from tundra to a complete shrub environment has the potential to increase the atmospheric heating substantially. Contrary to this increase in atmospheric warming, shrubs can have a cooling effect on the subsurface due to shading by the canopy. Myers-Smith and Hik (2013) found that summer soil temperatures were 2 °C lower below a shrub 55 cover than below shrub-free patches due to shading of the soil by the shrub canopy. The shading effect also reduces permafrost thaw below shrubs (Blok et al., 2010). However, it is expected that the large scale increase in atmospheric heating due to shrub expansion will overwhelm the cooling effect of shading and soil temperature will increase below shrubs in the long-term https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License. (Lawrence and Swenson, 2011;Bonfils et al., 2012). Therefore, the general accepted view is that shrub expansion has a positive feedback on climate warming (Pearson et al., 2013), although, some uncertainties still exist (Loranty and Goetz, 2012). 60 While we have a good understanding of how shrubs affect the climate, the impact of lichens on the micro-and largescale climate has not been studied in a thorough way yet. It is anticipated that the lichens' insulating properties and their high albedo will have a cooling effect on the micro-and large-scale climate (Bernier et al., 2011;Porada et al., 2016). For example, Odland et al. (2017) found a negative correlation between lichen abundance and soil temperature on Norwegian mountain summits. Also, Porada et al. (2016) modelled the impact of lichens and bryophytes on the soil temperature at high latitudes. 65 They estimated that lichens and bryophytes lower the soil temperature on average by 2.7 °C compared to an environment without lichens and bryophytes. However, they considered only the insulating properties of the two vegetation types and not the high albedo of lichens. Therefore, lichens might decrease the soil temperature even more. Most of the field measurements on the influence of lichens on the subsurface microclimate are based on differences between lichens and bare soil or disturbed lichens (e.g. Fauria et al., 2008;Nystuen et al., 2019;Van Zuijlen et al., 2020). However, constructive field measurements on 70 the difference in soil temperature between lichens and shrubs are lacking and therefore it is uncertain how the observed shift from lichen-dominated areas to shrub-dominated areas will change the micro-and large-scale climate in alpine and arctic areas.
To address this issue, we have set up a study to measure the difference in microclimatic conditions between lichen heaths and shrub vegetation at a mountain site in Norway. Our study design follows recommendations to apply a vegetationspecific approach to come to more detailed conclusions on the impact of shrub expansion and lichen decline (Stoy et al., 2012;75 Juszak et al., 2016;Williamson et al., 2016;Loranty et al., 2018). We focus on four microclimatic variables: net radiation, soil heat flux, soil temperature, and soil moisture. We used a paired plot design to measure these variables simultaneously at lichen and shrub plots in a Norwegian mountain area during two summers. Due to the paired plot design, we ensured that the lichen and paired shrub plots face similar background weather conditions, topographical characteristics and parent material. We hypothesize that lichens will have a lower net radiation because of their high albedo. Due to this lower net radiation and the 80 insulating properties of lichens (Porada et al., 2016), we further hypothesize that the soil heat flux and soil temperature will be lower under lichens than under shrubs. With this study, we advance the knowledge on the impact of lichens on the microclimate during the growing season, which is important to answer the question how a future vegetation shift from lichen-dominated areas towards shrub-rich environments might alter the micro-and large-scale climate.

Study area
The study was conducted at Imingfjell (60.1901° N, 8.5724° E), a mountain area in southern Norway with an elevation ranging from 1100 to 1350 m a.s.l. The vegetation is typical low alpine zone vegetation. Windswept ridgetops are covered with lichen heaths (see Appendix A for a picture of the area). Most common lichen species are of the genera Cladonia, Flavocetraria, Alectoria and Cetraria (Aartsma et al., 2020). The most abundant shrub species in the area is Betula nana, mainly located on 90 https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License. the midslope and ridgetop positions. The parent material of the soils in the study area consists of metarhyolitic moraine material (NGU, 2020). The nearest weather station (Dagali, 25 km from the study site; 828 m a.s.l., MET Norway (2019), Station nr. 29790) reported an average yearly temperature of 0.5 °C with an average July temperature of 11 °C for the period 1988-2007.
The average yearly precipitation during this period was 550 mm.

Data collection 95
We selected a study site of 2.5 km along a county road and 200 meters from this road into the field, resulting in an area of ca. 50 ha. In this study site, we delineated the lichen heaths using areal images of Geonorge (2018) in ArcMap (ESRI, 2019). The delineated lichen patches had a total area of 15 ha. Within these patches, we randomly selected ten locations. Subsequently, we used the criteria in Table 1 to select a lichen-dominated plot and a shrub-dominated plot near each of these ten locations ( Fig. 1). We measured the climatic variables simultaneously and in an identical way in one of the paired lichen and shrub plot 100 at the time for two days. After these two days, we moved the sensors to the next paired plots. We conducted the measurements on these plots between 4 July and 13 August 2018. Days with a precipitation duration of more than 30 min were excluded to minimize the effect of precipitation on the radiation measurements.  The terms of the net radiation, the soil heat flux, the soil temperature and the soil moisture were measured in a similar way on similar positions in each plot (Fig. 2). We measured the incoming shortwave radiation, reflected shortwave radiation, incoming longwave radiation and outgoing longwave radiation in W m -2 with one Kipp & Zonen CNR4 net radiometer per 110 plot. We placed the radiometer 30 cm above the canopy, which led to a measurement radius of 112 cm. With this height, we ensured that all the measured reflected shortwave radiation was reflected by the studied plot. The radiometer measured every 5 s and the data loggers (Kipp & Zonen Logbox SE) collected 5 min averages. We measured the soil heat flux at two positions per plot with Hukseflux HFP01SC self-calibrating heat flux sensors. We placed the heat flux sensors at 5 cm depth below the soil surface and measurements were done every 5 min. These measurements were recorded with Campbell Scientific CR800 115 data loggers. We measured the soil temperature on three positions per plot and at each of these positions on two depths (1 and 5 cm below the soil surface) with LogTag TRIX-8 temperature loggers. The temperature loggers measured the soil temperature every 5 min. We measured soil moisture at the same three positions as the soil temperature with ECH2O 5TM soil moisture sensors at 5 cm below the soil surface. These sensors measured the soil moisture every 5 min and the measurements were recorded with Em50 data loggers. We measured the reference air temperature at 1 m height at one location in the study area 120 ( Fig. 1) with an UTL-3 Temperature Datalogger placed in a Stevenson screen throughout the field season. In addition, we measured the precipitation manually with a regular rain gauge.
We measured the vegetation height in every plot at 10 cm intervals along a North-South and an East-West transect.
This led to 49 height measurements per plot. The thickness of the litter layer was determined at each of the five positions where soil temperature or soil heat flux were measured in each plot. We drilled one hole in the middle of each plot with a soil auger 125 (Ø 4 cm) and described the soil using the FAO guidelines (WRB, 2006). In each plot, we took three soil samples of the upper https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License. rock during insertion into the soil, we took a new sample in close vicinity. The two sub-samples were mixed and dried at 105 °C for 24 h. Subsequently we weighed the samples and calculated the bulk density per sample. In addition, we measured the organic matter percentage using the loss on ignition method (Heiri et al., 2001, ignition conditions: 550 °C for 3 h) and 130 measured the particle size distribution by dry sieving using an Endecott E.F.L. 1 MK11 sieve shaker.
The two-day measurements performed in 2018 were complemented in 2019 with measurements for 6 subsequent days in three additional paired lichen and shrub plots. In contrast to the plots of 2018, we selected the locations of the paired plots of 2019 subjectively (Fig. 1). However, the plots of 2019 also fulfilled the criteria of Table 1. We measured one paired plot at the end of June, one paired plot at the end of July and one paired plot in mid-August. To monitor the background weather 135 conditions in a more thorough way than during the 2018 field season, we placed a HOBO RX3000 remote weather station at the study site for the 2019 field season (see Appendix B for a list of sensors of the weather station).

Microclimate calculations
For each plot, we calculated the net radiation (Q*) with the four terms of the radiation balance using Eq. (1) (Oke, 2002) in which SWin is the incoming shortwave radiation, SWout is the reflected shortwave radiation, LWin is the incoming longwave radiation, and LWout is the outgoing longwave radiation.
We corrected the measurements of the soil heat flux for heat storage above the heat flux plates using Eq. (2) (Oke, 2002): in which QG0 is the soil heat flux at the soil surface, QGz is the measured soil heat flux at depth z, CS is the heat capacity of the 145 soil above the plate and ΔT/Δt is the change in temperature of the soil above the plate. For this correction, we converted the 5 min measurements to hourly averages and used the soil temperature that was measured at 1 cm depth to calculate ΔT/Δt. We determined Cs using Eq. (3) (De Vries, 1963): in which xmin, xorg and xw are the volume fractions of the mineral soil, organic matter and water, respectively. We obtained the 150 volume fractions of the mineral soil and organic matter with the organic matter and bulk density measurements. At two plots, the QG0 could not be calculated at one position due to the loss of a xorg measurement and malfunctioning of a soil temperature sensor at 1 cm depth.

Data analysis
For every plot of the field season of 2018, we calculated an average vegetation height, thickness of the litter layer, percentage 155 of soil organic matter, bulk density, and silt-and clay fraction. To test if these variables and the thickness of the A-horizon (obtained during the soil description) differ significantly between the lichen and shrub plots, we performed a Wilcoxon signed rank test.
To test for differences in microclimatic conditions between the lichen and shrub plots of 2018, we used linear mixed models for the net radiation, soil heat flux, soil temperature, and soil moisture. In the mixed models, we utilized vegetation 160 type (lichen or shrub) and the reference air temperature with interaction as fixed effects and day of measurement nested into plot number as random effect to account for the paired sample design. We added the reference air temperature as fixed effect since we expected that it affects the response variables directly (soil temperature and soil heat flux) or indirectly by being a proxy for the general weather conditions (net radiation and soil moisture). Per microclimatic variable, we constructed separate models for daily measurements, daytime measurements (08:00-22:00 LT) and nighttime measurements (22:00-8:00 LT). 165 Therefore, we converted the 5 min measurements of the net radiation and hourly measurements of QG0 into daily, daytime and https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License. nighttime totals (in megajoule) and the 5 min measurements of soil temperature and soil moisture into daily, daytime and nighttime averages. Only the soil temperature measured at 5 cm depth was used for this analysis.
The longer period of measurements per paired plot during the field season of 2019 allowed us to study the difference in microclimatic conditions between lichens and shrubs over a longer time period. We constructed time series of the hourly 170 averages of the reference air temperature, net radiation, soil heat flux and soil temperature for the three paired plots to gain more insight in the specific dynamics of the variables. In addition, we analyzed the difference in microclimatic conditions between lichens and shrubs during a warm, sunny day and a cold, cloudy day. As a basis for this analysis, we selected from one paired plot measurements from a distinct warm, sunny day and measurements from a distinct cold, cloudy day, and constructed time series of the reference air temperature, net radiation, soil heat flux and soil temperature. 175 All statistical analyses were made using R version 4.0.2 (R Core Team, 2020). The mixed models were constructed with the package nlme (Pinheiro et al., 2011).

Canopy and soil variables
We found a significant difference in vegetation height, thickness of litter layer and thickness of A-horizon between the lichen 180 and shrub plots (Fig. 3). Almost no plant litter was present under the lichen plots, while we measured an average (± SE) thickness of 7.1 (± 0.2) cm under the shrub plots. We found no significant difference in soil organic matter, bulk density, and silt and clay fraction between the lichen and shrub plots. Moreover, there was no clear difference in soil type between the lichen and shrub plots. All soils were classified as podzols or showed clear signs of podzolisation.

Microclimatic conditions throughout the field season 185
The daily total net radiation, daily total soil heat flux and daily average soil temperature differed significantly between the lichen and shrub plots of 2018 (Table 2, Fig. 4, Fig. 5). The shrub plots had a higher net radiation than the lichen plots during the entire field season (Fig. 4b). This difference in net radiation was mainly initiated by a difference in SW* (SWin -SWout) between the vegetation types (Fig. 6), governed by the higher albedo of the lichens compared to the shrubs, since SWin values were the same. On average, the daily net radiation was 3.15 MJ (26%) lower for the lichen plots than for the shrub plots. The 190 daily total soil heat flux and daily average soil temperature were higher under lichens than under shrubs for a substantial amount of days during the field season (Fig. 4c,d) and this difference was significant when air temperatures are relatively high (Fig. 5b,c). There was no significant difference in soil moisture between the lichen and shrub plots ( Table 2, Fig. 4e, Fig. 5d).

Day vs night
The difference in daily total net radiation between the lichen and shrub plot arose during daytime (Fig. 7a, Fig. 8b). The higher 205 albedo of lichens compared to shrubs will have its effect on the net radiation only during the day due to the absence of shortwave radiation at night. The soil heat flux below lichens was larger than below shrubs during daytime, while it was smaller or even negative below lichens during nighttime (Fig. 7b, Fig. 8c). As a consequence, the daily amplitude of the soil heat flux was larger for a lichen plot than for a shrub plot (Fig. 8c). The daily amplitude of the soil temperature was also larger for lichens than for shrubs (Fig. 8d), but the soil temperature differed only significantly between lichens and shrubs during daytime 210 with higher air temperatures for our measurements of 2018 (Fig. 7c).

Warm and sunny day vs cold and cloudy day
The difference in the microclimatic variables between the lichen and shrub plots were more pronounced during a warm, sunny day than during a cold, cloudy day (Fig. 9). The difference in net radiation was larger during a sunny day, since the incoming shortwave radiation is relatively high and therefore the higher albedo of lichens played a more dominant role in the net 215 radiation. As for the net radiation, the difference in soil heat flux between the lichen and shrub plot was larger during a warm, sunny day. However, the soil heat flux was higher below the lichen plot than below the shrub plot. https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.

Radiation balance 220
The higher net radiation of the shrub plots compared to the lichen plots was in line with our hypothesis. This difference is mainly initiated by the higher albedo of the lichen plots, as SW* is higher for shrubs than for lichens while the difference in LW* (LWin -LWout) is marginal (Fig. 6). Moreover, previously we measured an average difference in albedo of 0.124 between the lichen and shrub plots of 2018 (Aartsma et al., 2020). The daily average net radiation of our shrub plots was 3.15 MJ higher than of our lichen plots. Chapin et al. (2005) reported an increase in atmospheric heating of 0.55 MJ per day when the alpine 225 tundra shifts into shrubs. Using their definition of atmospheric heating (sensible + latent heat flux, i.e. net radiationsoil heat flux), we measured an average difference in atmospheric heating of 3.35 MJ per day between our lichen and shrub plots. This difference is more than six times larger than estimated by the study of Chapin et al. (2005). However, Chapin et al. (2005) assumed that the albedo of shrubs is 0.02 higher than the albedo of alpine tundra, which is substantially lower than the difference in albedo between our lichen and shrub plots. Moreover, our value of atmospheric heating might be slightly 230 overestimated, since our measurements were conducted during a relatively warm and sunny field season. This is reflected among others in the relatively large daily mean SWin that we measured during our field season (255 W m -2 ) compared to long term studies at similar latitudes (200 W m -2 , Eugster et al., 2000). Nevertheless, our study shows that large variations in radiation dynamics exist within alpine tundra depending on the vegetation composition.
The marginally lower LW* for the lichen plots (Fig. 6) is surprising, since it implies that the surface of lichens is 235 warmer than the surface of shrubs. The larger longwave radiation loss of the lichen plots is a result of a larger LWout, since LWin is similar for the paired lichen and shrub plots. Due to the dependence of LWout on the surface temperature following Stefan-Boltzmann's law (Oke, 2002), a larger LWout for the lichen plots suggests a higher surface temperature for the lichen plots, which seems counterintuitive considering the higher albedo of lichens. Moreover, time series of the LWout show that the difference in LWout between lichens and shrubs is made during daytime, while there is no difference during nighttime 240 (Appendix E). This points to additional processes that dominate over the effect of the albedo, showing an opposite effect.
Contrasting and counterintuitive results have also been found by previous studies. For example, Stoy et al. (2012) measured a higher surface temperature for the lichen species Cladonia rangiferina than for the moss species Sphagnum fuscum despite the higher albedo of C. rangiferina, while Gauslaa (1984) found a 20 °C higher thallus temperature of the dark-colored lichen species Bryocaulon divergens than the light-colored lichen species Alectoria ochroleuca. Parallel to our findings, Gersony et 245 al. (2016) measured with infrared thermography that the leaf temperature of B. nana is lower than for any other species in plots from a range of tundra types located in northern Alaska. A possible explanation could be that a difference in canopy morphology between lichens and shrubs leads to differences in the energy balance (i.e. evapotranspiration, see Sect. 4.2.3, Gersony et al., 2016), but this cannot be concluded from our measurements. Therefore, more research including the full energy balance and surface temperature of vegetation is needed to draw a solid conclusion on the dynamics between the albedo and 250 the surface temperature of lichens and other vegetation types.

Subsurface microclimate
The higher soil heat flux and soil temperature underneath the lichen plots during nearly the entire field season was not in line with our hypothesis. We thus infer that the higher albedo is not generating a cooler subsurface compared to shrubs, but that other differences between lichens and shrubs are more determinative. Our results are supported by the study of Mikola et al. 255 https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.
(2018), who also found lower soil temperatures below shrub plots than below lichen plots in the Siberian arctic tundra. We consider three reasons that might lead to a lower soil heat flux and soil temperature below shrubs compared to lichens: (1) differences in the subsurface between lichens and shrubs (i.e. litter layer); (2) shading of the surface underneath the taller shrubs with dense canopy; (3) differences in the energy balance of lichens and shrubs. In the following sections we will discuss how these three reasons might have affected our measurements. 260

Litter layer
A difference in subsurface between the lichen and shrub plots could lead to differences in the thermal properties (e.g. thermal conductivity, heat capacity) of the subsurface (Abu-Hamdeh and Reeder, 2000;Oke, 2002), which can in turn affect the soil heat flux and soil temperature. Since we did not find a difference in several mineral soil properties nor a difference in soil moisture between the lichen and the shrub plots (Fig. 3), it is unlikely that differences in the mineral soil initiated the higher 265 soil heat flux and soil temperature below lichens. However, we did find a large difference in litter layer thickness below lichens and shrubs. This organic litter layer contains among others dead leaves and roots and has a low thermal conductivity (Abu-Hamdeh and Reeder, 2000). Due to this low conductivity, the litter layer has an insulating effect on the mineral soil underneath and is likely to lead to a lower soil heat flux and soil temperature below shrubs (Beringer et al., 2001). Only some studies have addressed the insulating capacity of litter in the field (Beringer et al., 2001;Barrere et al., 2017). For example, Barrere et al. 270 (2017) measured a thermal conductivity of 1.36 W m -1 K -1 for an arctic soil and 0.19 W m -1 K -1 for a dry litter layer from shrubs in the Canadian Arctic and simulated that this litter layer decreased the summer soil temperature considerably. The insulating properties of litter might be of specific interest for our study, since the thermal conductivity is mainly depending on moisture availability (De Vries, 1963;Ochsner et al., 2001;Oke, 2002). Since our measurements were conducted during a relatively dry summer, the thermal conductivity of the litter might be even lower than during a normal summer and therefore the 275 insulating effect might have been amplified.
The insulating properties of litter can potentially also explain the smaller daily amplitude and delay in maximum and minimum for the soil heat flux and soil temperature below shrubs (Fig. 8). It appears that the soil below lichens gains and loses heat much easier than the soil below shrubs and that the soil temperature below lichens is more strongly coupled with the air temperature than the soil temperature below shrubs. Figure 5c also indicates a stronger relationship between soil-and air 280 temperature at the lichen plots than at the shrubs plots.

Shading by the shrub canopy
A second reason that we consider likely to cause the lower soil heat flux and soil temperature below shrubs compared to lichens, is the shading effect of the subsurface by the shrub canopy. Loranty et al. (2018) state that the amount of energy available for the soil heat flux depends among others on the thermal gradient between the ground surface and the underlying 285 soil. They advocate that the temperature of the ground surface, which includes only ground-cover vegetation such as lichens and mosses, is a better variable than the temperature of the land surface, which includes tall overlying vegetation canopies, https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.
since it is the ground-cover vegetation that is in direct contact with the underlying soil. It is plausible that the ground surface temperature in our lichen plots was higher than the ground surface temperature in our shrub plots due to shading of Betula nana on the ground cover vegetation in the shrub plots. Therefore the thermal gradient between the ground cover vegetation 290 and the soil was likely to be larger in our lichen plots and this led subsequently to a higher soil heat flux.
Multiple studies have addressed the shading effect of shrub canopies (Bewley et al., 2007;Juszak et al., 2014;Williams et al., 2014;Juszak et al., 2016). Juszak et al. (2016) measured an average growing season transmittance of only 0.36 below Betula nana. Moreover, multiple studies have also measured the impact of this shading on the soil microclimate below shrubs (Blok et al., 2010;Blok et al., 2011;Myers-Smith and Hik, 2013;Juszak et al., 2016). Blok et al. (2010) measured a 295 lower QG below Betula nana plots with a dense canopy compared to plots where the canopy was removed, despite a higher net radiation for the plots with a dense canopy. This low QG led to a decrease in active layer thickness below the plots with a dense canopy. Also, Myers-Smith and Hik (2013) found a 2 °C lower soil temperature below shrub plots compared to tundra plots due to shading of the shrub canopy. The above-mentioned studies show the possibility that shading by shrubs can cause a lower soil heat flux and soil temperature and this might therefore have led to a higher soil heat flux and soil temperature below 300 our lichen plots compared to our shrub plots.

Energy balance
A third reason for the unexpected results that we consider, is that a larger part of the net radiation of the shrub plots is used for evapotranspiration compared to the lichen plots and therefore a smaller fraction of net radiation is left to heat the soil. The net radiation that is available at the earth surface is usually partitioned over three components (Eq. 4) (Oke, 2002): in which QH is the energy that is used to heat up the atmosphere (sensible heat flux), QE is the energy that is used for evapotranspiration (latent heat flux), and QG is the energy that penetrates into the soil (soil heat flux). Since lichens do not have roots, they will not take up water actively from the soil and transpire back in the atmosphere. Therefore, the latent heat flux over a lichen canopy is solely dependent on evaporation and is relatively low. This is in contrast to shrub vegetation that does 310 take up and transpire water actively. The latent heat flux of shrubs is the sum of evaporation and transpiration and can thus be expected to be relatively high.
To verify that a larger part of the available energy is used for QE of our shrub plots compared to our lichen plots, we estimated QE of the plots measured in 2019 in a similar way as Eaton et al. (2001), using the formula of Priestley and Taylor (1972). Table 3 shows how the net radiation is partitioned over the three fluxes in our plots (see Appendix F for calculations 315 and additional results). These values imply that shrubs do use a larger part of the net radiation for evapotranspiration than lichens. The fraction QE/Q* of 0.55 for our lichen plots is close to the 0.49 found by both Eaton et al. (2001) and Boudreau and Rouse (1995) for a lichen-heath tundra. Moreover, the Bowen ratios fall within the range of low arctic upland tundra and low arctic shrub tundra reported by Eugster et al. (2000). Multiple other studies have found a higher QE/Q* for shrub tundra compared to shrub-free tundra heaths (McFadden et al., 1998;Eugster et al., 2000), even though their tundra heaths consisted 320 https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.  when both vegetation types are part of the understory vegetation of a mountain birch forest. Therefore, we infer that the 325 relatively high latent heat flux for our shrub plots is also a potential reason for their lower soil heat flux and soil temperature compared to our lichen plots. In addition, the lower QH/Q* for the shrub plots could be an explanation for the lower LWout and inferred lower surface temperature of the shrub plots compared to the lichen plots (see Sect. 4.1). However, measurements on the complete energy balance of lichens and shrubs are needed to confirm the partitioning of the net radiation over the three heat fluxes. 330

Synthesis
It is not possible from our study to conclude if one or more of the proposed reasons lead to the lower soil heat flux and soil temperature below shrubs compared to lichens. Considering the delay in maximum and minimum soil heat flux and the lower daily amplitude of the soil heat flux and soil temperature below shrubs (Fig. 8), it is likely that the litter layer below the shrubs plays an important role, since the other two proposed reasons would not generate this effect. However, additional 335 measurements are needed to give a solid conclusion. Nevertheless, our study does show that the high albedo of lichens is not leading to lower soil temperatures below lichens than below shrubs during the growing season. Since recent studies have shown that differences in color, and therefore albedo, did not even cause a difference in soil temperature between lichen species (Nystuen et al., 2019;Van Zuijlen et al., 2020), it is unlikely that albedo is an important factor determining the difference in soil temperature between the two different vegetation types. Instead, the marked differences in canopy structure between 340 lichens and shrubs are a more essential factor.
Our study shows that a shift from lichens to shrubs decreases the summer soil temperature, while other studies showed that shrub expansion can lead to higher winter soil temperatures, since the shrub canopy is trapping snow that insulates the soil (Sturm et al., 2001a;Myers-Smith and Hik, 2013). As a result, a shift from lichen heaths to shrub vegetation leads to lower soil temperature fluctuations during the course of a year. The change in fluctuation will be even more distinct with a shift from 345 lichen heaths to shrub vegetation than with a shift from a general arctic tundra towards shrub vegetation. Reason for this is that lichen heaths occur mainly on areas with shallow or missing snow cover, which are characterized by low winter soil temperatures (Odland and Munkejord, 2008;Sundstøl and Odland, 2017). In addition, a shift towards shrubs might have https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.
important consequences for permafrost, soil microbial activity and carbon storage due to a changing soil temperature (Myers-Smith et al., 2011;Loranty et al., 2018). Therefore, these consequence might be more distinct when the initial vegetation stage 350 includes a large abundance of lichens, since the soil temperature change is larger.
Multiple studies have discussed the cooling capacity of lichens on the underlying soil, and have argued that this cooling capacity is a result of their high albedo and the insulating properties of lichens due to their low conductivity. However, most studies measured or modelled lower soil temperatures below lichens in relation to bare soil or disturbed lichens (Beringer et al., 2001;Gold et al., 2001;Porada et al., 2016;Nystuen et al., 2019;Van Zuijlen et al., 2020), but the comparison with 355 another vegetation type has rarely been made. Our study shows that the cooling capacity of lichens does not lead to a lower soil temperature compared to shrubs. In addition, Van Zuijlen et al. (2020) concluded that the difference in soil microclimate between lichen species is not driven by the color of lichen species, but by lichen mat morphology. Therefore, our study and the study of Van Zuijlen et al. (2020) imply that the insulating capacity of lichens is a much more important factor determining the soil temperature than the high albedo. 360 Although the high albedo of lichens does not have a cooling effect on the subsurface, it will have a cooling effect on the atmosphere. Since lichens might continue to decrease in abundance due to shrub expansion, it is important to estimate the impact of such a shift on regional and possibly global climate. We measured an average increase in atmospheric heating of 3.35 MJ per day during the growing season with every square meter of lichen that turns into shrub. This value is among others dependent on the incoming solar radiation and can therefore change with latitude and day of the year. Modelling studies should 365 use our measurements to estimate the impact of the loss of lichen cover on the climate over alpine and arctic areas.

Conclusion
Our study shows that lichens have a lower net radiation than shrubs during the growing season. In addition, we show that the soil underneath the lichens has a higher soil temperature and a higher soil heat flux than the soil below shrubs, especially during warm days. This implies that the relatively high albedo of lichens affects the radiation balance, but not the subsurface 370 microclimate. Potential reasons for this could be the thicker litter layer, shading by the canopy or more evapotranspiration in the shrub plots. We conclude that the decline of lichens due to shrub expansion will lead to atmospheric heating (i.e. higher latent + sensible heat flux), but has a cooling effect on the subsurface during the growing season. Future studies should focus on the quantification of the effect of lichen decline on the climate on a regional and possibly on the scale of the arctic.

Variable
Sensor

Appendix F
To test if the shrub plots used more energy for the latent heat flux than the lichen plots, we calculated the latent heat flux for 385 the plots of the field season of 2019 in a similar way as Eaton et al. (2001) using the formula of Priestley and Taylor (1972): in which α is an empirical constant (-), S is the slope of the saturation vapour-temperature curve (Pa K -1 ) depending on the air temperature and γ is the psychrometric constant (65 Pa K -1 ). We used the α for upland lichen-heath tundra (0.90) and shrub tundra (1.08) estimated in the Canadian Arctic for our estimation (Eaton et al. 2001). Subsequently, we calculated the sensible 390 heat flux with: https://doi.org/10.5194/bg-2020-407 Preprint. Discussion started: 4 November 2020 c Author(s) 2020. CC BY 4.0 License.

Data availability 395
The data will be made available on the open-access research repository (Figshare) of the University of South-Eastern Norway after acceptance.