*2.2. Experimental Design*

Based on previous investigations [7,38], the input and output of litter and litter carbon in 15 representative permanent streams in the forest catchment were investigated. Three sampling sites for collecting litter were set up in the upper, middle, and end of every stream, and the streams' characters (length, width, sediment depth, and flow velocity) were also measured at every sampling time. The length was measured from the source to the estuary using a flexible rule along the banks of the streams. Meanwhile, the width was measured horizontally from one bank to another at every sampling site, the depth was measured vertically using a rule, and the flow velocity was measured using a flowmeter (Martin Marten Z30, Current-meter, Barcelona, Spain) every 30 min. Every index mentioned above was measured three times, then the mean was taken at every sampling time for every sampling site. These streams are in a typical subalpine forest catchment at elevations of 3600~3700 m, and the total area of the investigated forest catchment was 4.3 km2. At every site, a button thermometer (iButton DS1923-F5, Maxim/Dallas Semiconductor, Sunnyvale, CA, USA) was set to record the temperature every 2 h, and the precipitation was measured using a rainfall monitor (ZXCAWS600, Zxweather, Beijing, China) for real-time monitoring.

#### *2.3. Monitoring the Input and Output of Litter and Litter Carbon*

In order to collect litter, according to the stream length, a quadratic litter collector (0.8 m × 0.8 m) was randomly installed at the source, middle, and end of the stream (when the stream width < 0.8 m, one litter collector was positioned; when the stream width > 0.8 m and <1.6 m, two litter collectors were positioned; the stream widths are shown in Table 1), and each was installed 0.5 m above the water or ground surface (Figure 2). To avoid litter decay in the litter collectors caused by rainfall, the litter samples were collected every 15 days, but the litter was collected only once in the cold winter since litterfall in winter was rare. All of the collected litter samples were put into precleaned polyethylene bags and transported to the lab. The samples were dried to a constant weight and stored at 65 ◦C for less than one week until analysis.


**Table 1.** Basic characteristics of 15 representative subalpine forest streams in the investigated subalpine forest catchment.

**Figure 2.** Litter and litter carbon input and output monitoring system in the investigated streams of the Bipenggou Valley, located in the upper reaches of the Yangtze River.

To make sure the amount of litter collected was accurate, a litter collector (0.8 m × 0.8 m) was installed at the start of every stream (the width of all stream sources were less than 0.8 m). Two litter interception dams with different mesh specifications were set at the outlet of each stream, and the total amount of litter intercepted by the two interception dams was the output amount of litter (Figure 2). The interception dam with 3 cm × 3 cm apertures was in front, which mainly collected litter with a larger diameter, and the 100- mesh interception dam was installed behind, which was mainly used to collect litter of a smaller diameter. The cycles of measuring the litter output were consistent with the litter input measurement. During the sampling time, all litter collected from the interceptor was quickly brought back to the laboratory, dried to a constant weight at 65 ◦C, then weighed and recorded as the litter output.

We divided one year into five different periods, i.e., the snowmelt season (SMS: April to May), early growing season (EGS: May to June), growing season (GS: July to August), later growing season (LGS: September to October), and seasonal snow cover (SSC: November to April the following year) based on phenological changes, seasonal precipitation, and temperature dynamics [40]. Litter was collected nine times in the growing seasons and four times in the non-growing seasons. Specifically, litter in the collectors was collected during the LGS at approximately 15-day intervals. The quantity of inputted and outputted litter and litter carbon at each period was calculated as the cumulant values of this stream, and the 15 streams were treated as 15 repeats during data analysis.

The temperatures in the in-stream and riparian zones varied considerably with the ecosystem type during the two years of the experiment [7]. The temperature fluctuated sharply with critical periods, with the average daily temperatures ranging from 0 ◦C to 10.7 ◦C. However, the in-stream and riparian zone temperatures were almost always above 0 ◦C throughout the investigation period. Similarly, the water varied substantially between the stream and riparian zones for a comparable period, and varied significantly within the stream or riparian zone among different periods (Table 1).

#### *2.4. Analytical Methods and Calculations*

The concentration of litter carbon was determined using the potassium dichromate oxidation–external heating method [45]. Following this, 0.01 g of the dried litter that had been sieved through a 0.15-mm sifter was placed in the bottom of a 100-mL Erlenmeyer flask. The required amount of H2SO4 (5 mL) and 10 mL of potassium dichromate solution were added. After attaching a reflux condenser, the mixture was boiled at 220–230 ◦C for 15 min on an electric stove. After cooling and rinsing the condenser with water, 3 to 5 drops of N-phenylanthranilic acid were added. The titration was performed with a 0.2 M solution of ferrous sulfate salt at room temperature. With the addition of one drop, the color shifted from violet to bright green:

$$\mathbf{C\_S = (C\_C \times M\_L)/S\_S}$$

where CS is the litter carbon stock; CC is the carbon concentration, g kg −1; ML is the litter stock, g; and SS is the area of the stream, m2.

Linear mixed effect models were used to analyze the relationships of the input and output of litter and litter carbon with climate and stream characteristics in the subalpine forest stream among different sampling periods. First, we tested the normality of residuals, homoscedasticity of errors, and independence of errors to determine whether our data met the assumptions of the analyses. Second, the sampling period was treated as a fixed effect, and then we conducted a repeated measures analysis of variance (ANOVA) to examine the variability of the different variables (litter input, litter carbon input, litter output, and litter carbon output) at different critical periods. Third, to better illustrate the correlations of the input and output of litter and litter carbon with the explanatory variables, these variables were treated as fixed factors and the stream was included as a random factor. We used linear and quadratic models to fit the relationships of four indices of litter with the changes in the various explanatory variables. The relationships between the ratios of input to output of litter and litter carbon vs. stream characteristics were also tested by the linear mixed effect models. All analyses were conducted in R using the LME4 package [46].
