**4. Discussion**

After four years of decomposition, 50~53 and 58~64% of mass and carbon were lost (Figures 1 and 3), which were similar to other pine tree species in other temperate forests [44,45], and more than 50% of the losses were observed in the 3600 m sites during the non-growing seasons (Figure 2), indicating rather high mass loss and carbon release in decomposition. *A. faxoniana* litter was present at higher elevations of the subalpine forest subjected to low temperatures in the non-growing seasons. Forest gaps formed by natural stem breakage were commonly distributed in high-latitude and high-altitude ecosystems, which would modify the temperature, precipitation, snow coverage and radiation conditions within the gap areas when compared to the closed canopies [10,46]. Our results partially support the first hypothesis that forest gaps increased the mass loss and release in the decomposing *A. faxoniana* leaf litter, but only in the 3300 m and 3600 m sites. For the decomposition constant (*k*), we found that higher decomposition rates were in the gaps when compared to the closed canopy in both the 3300 and 3600 m sites. A previous study has demonstrated that there was more accumulated snow coverage within gaps, helping microbes maintain high activities through the isolated protection in winter, as well as the more intense precipitation and strong radiation attributed to the opening canopy increasing the leaching loss and photodegradation in the growing season, thereby promoting the litter mass loss [8,47], which was consist with our results. However, the gapstimulated effects were not observed in the 3000 m site, exhibiting a higher decomposition rate under the closed canopy. According to the environmental factors presented in Table 1, we recorded relatively small differences in the snow depth among different gaps and the closed canopy; besides, there was a lower frequency of freeze–thaw cycles under closed canopy than in gaps in the non-growing seasons, which was the opposite to the other two elevations. Therefore, a likely explanation was that the lower variables of environmental conditions among the gaps and closed canopy compared to the two high elevations resulted in different effects of gaps on the decomposition rate. Additionally, the Spearman correlation analysis showed that both mass loss and carbon release were significantly negatively related to the freeze–thaw cycles (Table 3), implying that the effects of gaps on litter decomposition were more correlated to the freeze–thaw cycles in winter. It has been reported that soil freezing would modify carbon release from plant litter in the subalpine forests of this region [9,48], and our study addressed that freeze–thaw cycle variations were a key driver changing the effects of gaps on litter decomposition at different elevations of high-altitude forest ecosystems, which also supported our third hypothesis.

In this study, the effects of forest gaps were not always in accordance with the gap size gradient. Previous studies have revealed that nitrogen mineralization in forest gaps increased in the first decomposition year and then decreased later [49,50]; thus, the soil microorganisms contributing to the degradation of carbon-associated compounds at the early decomposition stage would benefit from increasing the nitrogen availability [46,51]. However, carbon release did not consistently present a general pattern along with the gap size gradient (Figure 2c). In general, although deep accumulated snowpack on the forest floor within gaps decreased the contribution of soil freezing to the physical breakdown

of litter plant-derived materials [52,53], higher biological activities were still maintained under the snow cover [8,54]. Moreover, intense hydrological leaching during the snowmelt period also contributes to the labile carbon releasing from newly shed litter [55,56]. These parts of carbon, in turn, provide available carbon and nutrients for microbial utilization and thus affect the decomposition processes [50,57]. When compared with the lower elevation sites, the effect of forest gaps was greater in the high elevation site (3600 m) for both carbon content and release from *A. faxoniana* litter. Previous studies revealed that cold regions are proven to be more sensitive to environmental disturbances, and the relatively lower temperatures in the high elevation site are at much higher risk of promoting soil freezing, which would physically favor the decomposition of the *A. faxoniana* litter [9]. Meanwhile, snow coverage in forest gaps has also been documented to accelerate carbonassociated compound (both labile and recalcitrant substrate) degradation [58]. Therefore, the stimulative effects of forest gaps on mass loss and carbon release were more responsive to higher elevation in subalpine forests.

The stimulative effects of forest gap on litter decomposition were more remarkable in the non-growing seasons when compared to the growing seasons. Carbon content decreased over time at the three elevations as decomposition proceeded, particularly for the non-growing seasons (Figure 2). Moreover, a rather higher mass loss and carbon release in *A. faxoniana* litter was observed in subalpine forests subjected to low temperatures in the non-growing seasons. These results sugges<sup>t</sup> that carbon release greatly varies between non-growing and growing seasons in the subalpine forest. When carbon release in the nongrowing seasons throughout the four decomposition years was combined, we observed that the large gaps significantly enhanced the carbon release at the 3600 m site, suggesting that deep snow coverage at the high-elevation forest promoted carbon release from the *A. faxoniana* litter [59]. Forest gaps only exhibited significant effects at the early litter decomposition stage (the first two decomposition years), but such a gap influence was dissimilar, with the increasing carbon release in the first decomposition year and decreasing in the second year. Previous studies indicated that the difference between forest gaps was greater at the late stage of decomposition [23], which is attributed to the disturbances of the original environment resulting from the formation of forest gaps. However, in this study, the gap influence on carbon release disappeared by the end of four decomposition years, which supported our second hypothesis. Another likely explanation was that 56, 50 and 49% of the initial carbon was lost in the first two decomposition years at all the three sites, and the remaining carbon compounds were primarily recalcitrant components with a resistant structure, which decay quite slowly and are not able to be affected by environmental alteration easily [45].
