Soil respiration variation along an altitudinal gradient in Italian Alps: Disentangling forest structure and temperature effects

To understand the main determinants of soil respiration (SR), we investigated the changes of soil respiration and soil physicochemical properties, including soil carbon (C) and nitrogen (N), root C and N, litter C and N, soil bulk densities and soil pH at five forest sites, along an elevation/temperature gradient (404 to 2101 m a.s.l) in Northern Italy, where confounding factors such as aspect and soil parent material are minimized, but an ample variation in forest structure and composition is present. Our result indicated that SR rates increased with temperature in all sites, and about 55% - 76% of SR was explained by temperature. Annual cumulative SR, ranging between 0.65 and 1.40 kg C m-2 yr-1, declined along the elevation gradient, while temperature sensitivity (Q10) of SR increased with elevation. However, a high SR rate (1.27 kg C m-2 yr-1) and low Q10 were recorded in the old conifer forest stand at 1731 m a.s.l., characterized by a complex structure and high productivity, introducing nonlinearity in the relations with elevation and temperature. Reference SR at the temperature of 10°C (SRref) was not related to elevation. A significant linear negative relationship was found for bulk density with elevation. On the contrary, soil C, soil N, root C, root N, pH and litter mass were better fitted by nonlinear relations with elevation. However, it was not possible to confirm a significant correlation of SR with these parameters once the effect of temperature has been removed (SRref). These results show how the main factor affecting SR in forest ecosystems along this Alpine elevation gradient is temperature, but its regulating role can be strongly influenced by site biological characteristics, particularly vegetation type and structure. This study also confirms that high elevation sites are rich in C stored in the soil and also more sensitive to climate change, being prone to high carbon losses as CO2. Conversely, forest ecosystems with a complex structure, with high SRref and moderate Q10, can be more resilient.

relationship between soil organic matter (SOM) and elevation and reported that global soil 66 organic C stock at high elevation is more sensitive to climate change and is predicted to decrease 67 in a warming climate [14,23,29,30,31,32]. However, several researchers have reported opposite 68 trends and found lower SOM and higher SR at a higher elevation [30,33,34]. This variability 69 may be partially due to confounding factors affecting SR other than temperature. Besides 70 elevation, mountain landscapes are, in fact, characterized by substantial changes of other site 71 parameters such as slope and aspect, which can affect microclimatic conditions and, therefore 72 soil C dynamics [35]. Furthermore, due to the heterogeneity of geological substrates, soils of 73 mountain regions are highly diverse over short spatial scales and this can generate marked 74 contrasts in soil biogeochemical functions [36]. Different results have also been found when SR 75 is related to soil organic carbon [26,37]. 76 Besides, there is evidence that diverse plant biome types can influence SR rate differently. 77 Therefore, the various plant communities can affect differently, microclimate, soil and litter 78 composition, and root distribution, therefore affecting soil respiration rate [18,26,38,39,40]. 79 However, within the same plant biome, there is a high spatial heterogeneity of SR. Some authors 80 found a possible linkage between the topography, plant community structure (e.g., forest type 81 and speed of regeneration), and SR within the same forest ecosystems [18,38,39,40]. Further, 82 forest management can also play a crucial role in SR [41]. For instance, tree removal can directly 83 influence soil respiration due to the removal itself (i.e., reduction of plant biomass) but even Currently, the temperature dependency of SR and SOC decomposition is a major interest 87 regarding global climate change and the role of terrestrial ecosystems in regulating Earth´s 88 climate [43,44]. Therefore, there is a need to better understand the interactions between 89 temperature and soil CO2 efflux. The general goal of this study is to disentangle the possible 90 multi-effects on SR of soil properties, temperature, SOM, and vegetation structure (tree height in 91 particular) along a plant biome-elevation gradient. In particular, the existing differences in 92 vegetation structure allowed us to investigate the extent to which these biological variables and 93 the induced variation in microclimatology can alter the relation between elevation and SR. 94 Specifically, i), we tested the hypothesis that SR and SOM accumulation change linearly with 95 elevation. We also hypothesized that the Q10 value increases linearly with elevation as well. 96 Furthermore, ii) we analyzed which are the main factors affecting SR other than temperature. To 97 better isolate the effect of temperature on SR, the study was conducted along an 98 altitudinal/temperature gradient in Italian Alps, in conditions where confounding factors like 99 slope, aspect, and soil parent material are minimized. The differences in vegetation structure 100 allowed us to investigate to which extent these biological variables and the induced variation in 101 microclimatology, can alter the relation between elevation and SR.

103
Study areas 104 Five experimental sites were established between the top of the Rittner Horn mount and the city 105 of Bolzano, Italy, on the southern side of the Alps (Fig 1a). The overall elevation gradient 106 between the highest and the lowest site is 1700 m and the elevation separation between each site 107 is approximately 420 ± 60 m. All sites are characterized by a soil evolved upon a glacial till laid 108 on a porphyric bedrock and an SSE slope orientation. Annual precipitation ranges between 800 109 and 1000 mm.

118
The site is characterized by an unvenaged distribution of tree diameters, approaching the 119 structure of old-growth forest stands (45,46]. Site C was located near the location of Riggermoos were applied between elevation and environmental variables (soil C and N, root C and N, litter C 216 and N, soil bulk density, soil pH). The linearity changes of these variables with elevation were 217 detected based on the lowest AIC and the highest R 2 . The association of Q10, soil C and soil N 218 with environmental variables were determined using Spearman's Correlation Test. All statistical 219 analyses were performed using R version 3.6.0 ([59], www.r-project.org).

222
Environmental factors variability along the altitudinal gradient 223 A significant difference in soil C stock was found only between site E (3891 ± 2756 g C m -2 ), at 224 the lowest altitude, where the C stock was smaller in comparison to sites A, B, and C (Fig 2a).

225
No significant differences were found for soil N stock in the different sites (Fig 2e). Root 226 biomass and root C and N stocks in site A were significantly higher than other sites (Fig 2b,  and litter mass in site B was significantly higher than site E (Fig 2h). However, the accumulated 228 C and N in the litter were not significantly different along the altitudinal gradient (Fig 2c, g).

229
Significant differences were found between pH values in the different sites: the lowest value of  Based on the AIC and R 2 , the linear relationship with elevation appeared in model selection only 241 for soil bulk density i.e., soil bulk density resulted to be linearly related to elevation (Table 2).

242
On the contrary, soil C, soil N, fine root mass, and root C, root N, pH and litter mass data were 243 fitted better with nonlinear relations with elevation (Table 2; more detail about equations can be 244 found in Table S1; supplementary material). Furthermore, a significant negative correlation was 245 found between soil C and soil N with soil pH, mean dominant tree height, and bulk density 246 (Table 3).  Both logistic and Q10 models confirmed that soil respiration rates increased with temperature in 275 all sites (Fig 3), and the seasonal pattern of SR was similar to that of air chamber temperature 276 (S1 Fig).   (Table 4). The Q10 value recorded in site A (highest elevation) was significantly different from the others 289 (Table 4). A significant linear relationship was identified between Q10 and elevation (Table 2).

290
However, the trend of Q10 against temperature was better described by a nonlinear relation 291 (Table 2). Furthermore, a significant negative correlation was found between Q10 and mean 292 dominant tree height ( The cumulative SR in site A was significantly lower than in the other sites (Fig 4). The results of 302 the statistical analysis also confirmed a nonlinear relationship between cumulative SR and 303 elevation (Table 2).  in the present study, all the sites are characterized by similar slopes and by the same south or 337 south-east facing; therefore, we tend to exclude an influence of these factors on soil C 338 accumulation in the examined sites. According to recent studies, SOM is not consistently related 339 to variation in climatic conditions along elevation gradients; it is also strongly affected by 340 productivity or by vegetation type/composition [30,33,66,67]. Our data confirms a significant 341 positive correlation between mean dominant tree height and soil C (Table 3). Therefore, the 342 higher amount of soil C found in the present study at intermediate altitude could be explained by 343 higher site productivity in sites B and C in particular, which is also suggested by the mean 344 dominant tree height. 345 Soil pH and bulk density are considered two of the main variables influencing other soil 346 properties, soil microbial activity and soil respiration [68,69]. Generally, at high elevation, the 347 higher precipitation and lower evapotranspiration rate decrease soil pH by increasing the 348 leaching of basic cations [65,70,71,72,73]. This is confirmed by the strong relationship between 16 349 elevation and soil pH found in the present study (Table 2, Fig 2). In addition, soil bulk density 350 was significantly diminished by increasing elevation (Fig 2; Table 2). One of the main factors 351 affecting soil bulk density is SOM content [74]. Therefore, the lowest values of soil bulk density 352 at high elevation could be explained by the high amount of soil C, as confirmed by the negative 353 association found between soil C and soil bulk density ( At a global scale, SR has been related to soil C, litter production and pH, and negatively 377 correlated with soil bulk density; therefore, a high value of soil C and litter accumulation could 378 lead to an increase in soil respiration [20,66,81,82,83]. In the present study, the highest amount 17 379 of soil C and dry litter weight were also observed in site B. However, we could not find a 380 significant correlation between SR and soil C and litter dry weight (Fig 5; Table 5).

381
An increase in SR has also been observed as a consequence of increasing soil pH between 4 and 382 7, because of the positive effect of pH on soil microbial activity within this range 383 [1,68,84,85,86]. In contrast, SR and bulk density are generally negatively correlated, as a low SR 384 indicates increasing rates of SOM accumulation and therefore a decrease in bulk density [82].

385
Furthermore, SR declines with increasing bulk density due to the lower soil porosity and oxygen 386 availability for microbial activity in compacted soils [18,59,87]. However, our analysis did not 387 confirm a significant correlation of SR with pH and bulk density (