*2.1. Experimental Site*

This study was conducted at the Long-term Research Station of Alpine Forest Ecosystems, Sichuan, China (31◦14 N, 102◦53 E, 3579–3582 m). The experimental sites are located in a primeval alpine forest on the eastern Tibetan Plateau and have been extensively described elsewhere [18]. The mean annual temperature and precipitation are 2.7 ◦C and 850 mm, respectively. Snow cover develops in late October and melts in April of the following year with a maximum snow depth of approximately 50 cm and a soil freezing time of approximately 120 days [19]. This alpine forest is dominated by coniferous fir (*Abies faxoniana* Rehd. et Wils.), cypress (*Sabina saltuaria* Rehder & E.H.Wilson), deciduous larch (*Larix mastersiana* Rehd. et Wils.) and birch (*Betula albosinensis* Burk.). Canopy gaps induced by natural tree fall and other climatic extremes cover 13–23% of the forest landscape. Soils are classified as Cambisols (WRB soil taxonomy) [20] and the litter depth (Oi/Oe) is 7 cm with a C stock of 1.6 t ha−<sup>1</sup> in autumn (the peak of litter fall). The contents of C, nitrogen (N), phosphorus (P) and humic substances in the organic soils (Oa/A) are 16%, 0.58%, 0.17% and 6.1%, respectively [21].

## *2.2. Experimental Design*

We used a 2 (snow depth) × 4 (litter species) factorial design in this study. Litter decay at different snow depths was facilitated by incubating litter samples in the canopy gap (deep snow plot) and under the canopy (shallow snow plot). The alternative to this strategy was to perform repeated manual shoveling to reduce snow cover, but this strategy increases the risk of losing needle litter (fir, cypress and larch) from litterbags, which must have sufficiently large mesh sizes to permit access by soil fauna [22]. To decrease the uncertainty among litter species, four dominant tree species (fir, cypress, larch and birch) with different initial qualities (Table 1) were studied. Specifically, three sites (as replicates, 500–2000 m apart) with similar topographies and canopy covers that receive similar snowfall and litter fall were set up in this alpine forest in 2009. Each site had two pairs of deep/shallow snow plots (*n* = 6), and each plot had four subplots (3 × 3 m in size and 3–4 m apart) for incubating the four foliar litters (a total of 24 subplots).

**Table 1.** Initial chemical compounds in the four foliar litters.


The values are means with standard deviations shown in parentheses (*n* = 3). The values in the same columns with different superscript letters are significantly (*p* < 0.05) different among litter species. C: carbon, N: nitrogen, P: phosphorus, WSS: water-soluble substances, OSS: organic-soluble substances, ASS: acid-soluble substances, and AUR: acid-unhydrolyzable residues.

To assess the effect of variation in winter snow timing on litter decomposition, samplings were scheduled at the ends of the snow formation, snow coverage and snowmelt stages during winter from 2012 to 2014. From 2014 to 2016, we collected samples only once each winter (at the end of the snowmelt stage). The beginning and end of winter were defined as having hourly temperatures with a 50% probability (continuous over 12 h per day) of freezing (below 0 ◦C; Table 2) [23]. The winter was further divided into snow formation, snow coverage and snowmelt stages based on the snow dynamics and changes in temperature according to our previous meteorological monitoring in this alpine forest (Figure 1) [18] and the experiments that were conducted in other cold biomes [24]. We also performed sampling at the end of the growing season (beginning of the snow formation stage) of each year.

**Table 2.** Sampling comparison of the timing and length of winters during the four years of decomposition.


a Hourly temperature with 50% probability (continual 12 h per day) of freezing (below 0 ◦C) is defined as the beginning and end of winter in this study. This definition is not conservative because solar radiation is very strong in this alpine forest, which increases the temperature recorded by the data logger. b Sampling dates are scheduled approximately in late April and October within one week of intervals referring to the timing of the 2012/2013 winter.
