**3. Results**

#### *3.1. Temperature Characteristics and Snow Depth*

Temperature characteristics of litter layer fluctuated consistently in different forest gaps and elevations are shown in Table 1. For the non-growing and growing seasons, a higher mean temperature was observed in forest gaps when compared to that under the closed canopy in all the three elevations. Furthermore, the highest mean temperatures were shown in the large gap in the 3000 m, while it was found in the middle or small gaps in the 3300 m and 3600 m plots. Freeze–thaw cycles' frequency was lower in large and middle gaps at the 3600 m and 3300 m sites, while a higher cycle frequency was found at the 3000 m site. During the non-growing season (November to April), the snow depth on the forest soil was consistently ranked in the following two orders: large gap > middle gap > small gap > closed canopy, 3600 m > 3300 m > 3000 m. At the OF and DF of the first two decomposing years, the accumulated snow was deeper than for other stages.


**Table 1.** Seasonal environmental factors of in the study sites.

Notes: OF, onset of freezing stage; DF, deep freezing stage; TS, thawing stage; EGS, early snow-free stage; MGS, middle snow-free stage; LGS, late snow-free stage; NGS, non-growing season; GS, growing season.

#### *3.2. Litter Mass Loss and Decomposition Constant k*

After the four decomposition years, the *A. faxoniana* litter mass loss reached 50~53%, and it was significantly affected by gap size, elevation, and decomposition time, as well as their interactions (Table 2). Specifically, the litter in the forest gaps exhibited significantly higher mass loss at the 3300 m and 3600 m sites than that under the closed canopy, especially in the late decomposition period (Figure 1b,c). However, litter at the 3000 m site showed significantly higher mass loss under the closed canopy when compared to those in the forest gaps (Figure 1a). Furthermore, there were higher litter mass loss rates during the non-growing season than during the growing season (Figure 2a,b). Although there were no significant differences, the mass loss was ranked in the order: large gap > middle gap > small gap in the non-growing season, but this showed a opposite trend in the growing season. For the four years, the mass loss was significantly lower at the 3000 m site than that at the 3300 m site. Besides, the decomposition k exhibited a gradually increasing trend from the large gap, middle gap, small gap to closed canopy plots at the 3000 m site, while that at 3300 m and 3600 m sites showed a totally opposite decreasing trend (Figure 1d).

**Table 2.** Results (F-values) of linear mixed effects models testing the effects of elevation (3000, 3300, 3600 m), gap size (large, middle, small), decomposition time (sampled times) and their interactions on mass loss and carbon content and release in in decomposing *A. faxoniana* litter. \*\* *p* < 0.001.


**Figure 1.** Mass loss and decomposition constant (*k*) of *A. faxoniana* litter in forest gaps (large gap, middle gap, small gap, closed canopy) along an elevation gradient (3000, 3300, 3600 m) at different decomposition stages in a subalpine forest of southwestern China. (**a**) Mass loss at 3000 m; (**b**) Mass loss at 3300 m; (**c**) Mass loss at 3600 m; (**d**) Decomposition constant k. Mean ± SE, n = 3. OF, onset of freezing stage; DF, deep freezing stage; TS, thawing stage; EGS, early snow-free stage; MGS, middle snow-free stage; LGS, late snow-free stage; NGS, non-growing season; GS, growing season. Asterisk above lines indicates significant differences among forest gaps (*p* < 0.05), lowercase letters above columns indicate significant differences among forest gaps (*p* < 0.05), capital letters above columns indicatesignificantdifferencesamongelevations(*p* <0.05).

**Figure 2.** Effects of forest size and elevation on mass loss and carbon release in decomposing *A. faxoniana* litter. (**a**) Mass loss on forest gap scale; (**b**) Mass loss on elevation scale; (**c**) Carbon release on forest gap scale; (**d**) Carbon release on elevation scale. Mean ± SE, n = 3. NGS, the non-growing seasons in the four decomposing years; GS, the growing seasons in the four decomposing years. Lowercase letters indicate significant differences among forest gaps or elevations (*p* < 0.05).

#### *3.3. Litter Carbon Content*

During the four decomposition years, the *A. faxoniana* litter content gradually decreased by 18~29%, and it was significantly affected by gap size, elevation, decomposition time as well as their interactions (Figure 3, Table 2). Specifically, obvious declines were observed in the first non-growing season and the second growing season during the first two years. When compared with the closed canopy, higher carbon contents were shown in forest gaps at the 3000 m and 3300 m sites. However, at the 3000 m site, the highest carbon content was measured under the closed canopy in the first decomposition year, and then exhibited the following order: large gap > closed canopy > middle/small gap. Additionally, the non-linear regression indicates that the carbon content significantly declined with decomposition processing, meanwhile the decomposed litter with less remaining mass had a higher carbon content at the 3300 and 3600 m sites than that at 3000 m site (Figure 4a,b).

**Figure 3.** Carbon content of *A. faxoniana* litter in forest gaps (large gap, middle gap, small gap, closed canopy) along an elevation gradient (3000, 3300, 3600 m) at different decomposition stages in a subalpine forest of southwestern China. Mean ± SE, n = 3. OF, onset of freezing stage; DF, deep freezing stage; TS, thawing stage; EGS, early snow-free stage; MGS, middle snow-free stage; LGS, late snow-free stage; NGS, non-growing season; GS, growing season. Lowercase letters indicate significant differences among forest gaps (*p* < 0.05), *n.s*. indicates no significant differences (*p* > 0.05).

**Figure 4.** Carbon content versus remaining mass in decomposing *A. faxoniana* litter. (**a**) Forest gap scale; (**b**) Elevation scale. R<sup>2</sup> and *p* values from linear regression are shown in each panel.

#### *3.4. Litter Carbon Release*

After the four decomposition years, 58~64% of carbon was released from the *A. faxoniana* litter; carbon release was not affected by the gap size (Figure 5), but within the time period, it was significantly affected by gap size, elevation, decomposition time as well as their interactions (Table 2). When compared to the 3000 m and 3300 m sites, the carbon release variation was greater at the 3600 m site. When compared with closed canopy, more carbon was released from decomposing litter in forest gaps at the 3300 m and 3600 m sites. However, less carbon was released from decomposing litter in the large/small gasp at the 3000 m site. At the end of decomposition, there were no significant differences among forest gaps and the closed canopy at all sites, and significant differences were only observed in the first two years. Under the closed canopy, less carbon was released from decomposing litter at the 3300 and 3600 m sites in the first year but more carbon was released at the 3000 and 3300 m sites in the second year. Carbon released in the first two years accounted for 81~86% of the total four-year release, while that of the combined non-growing seasons' release across the four years accounted for 45~63%. There were significant differences in litter carbon release in the combined non-growing seasons across the four years among the gap sizes at the 3600 m site, but there were no significant differences in the combined growing seasons at all the sites. Further, there was a higher litter carbon release during the non-growing season than during the growing season (Figure 2c,d). The effects of gap size on carbon release were not significant, but at 3300 m they were significant higher and lower in the non-growing season and growing season, respectively.

#### *3.5. Key Drivers of Litter Mass Loss and Carbon Release*

Based on the Spearman correlation analysis, litter mass loss and carbon release were both significantly negative correlated with freeze–thaw cycles in the non-growing season, and carbon release was significantly negative correlated with freeze–thaw cycles for the whole four decomposition years. Besides, the litter carbon release was also significantly negative correlated with the positive accumulated temperature and the mean temperature in the non-growing season (Table 3).

**Figure 5.** Carbon release percentage of *A. faxoniana* litter in forest gaps (large gap, middle gap, small gap, closed canopy) along an elevation gradient (3000, 3300, 3600 m) at different decomposition stages in a subalpine forest of southwestern China. Mean ± SE, n = 3. 1st NGS, non-growing season in the first decomposing year; 2nd NGS, non-growing season in the second decomposing year; 3rd NGS, non-growing season in the third decomposing year; 4th NGS, non-growing season in the fourth decomposing year; 1st GS, growing season in the first decomposing year; 2nd GS, growing season in the second decomposing year; 3rd GS, growing season in the third decomposing year; 4th GS, growing season in the fourth decomposing year; NGS, the non-growing seasons in the four decomposing years; GS, the growing seasons in the four decomposing years. Lowercase letters indicate significant differences among forest gaps (*p* < 0.05), *n.s*. indicates no significant differences (*p* > 0.05).


**Table 3.** Spearman correlation coefficient between mass loss, carbon release and environmental factors in the non-growing season, growing season and the 4 total decomposition years. Significant correlations are indicated in a bold font, with \* *p* < 0.05, \*\* *p* < 0.001.
