**4. Discussion**

Our results demonstrated that both the input and output of litter and litter carbon in the subalpine forest streams showed two peaks with critical periods, which was not consistent with our first hypothesis that the seasonal input and output of litter and litter carbon might have different dynamic patterns. Climate factors (temperature and precipitation) together with stream characteristics (sediment depth, length, and flow velocity) drive the seasonal sink-source pattern of litter and litter carbon in the subalpine forest streams. Meanwhile, the results showed that the maximum and minimum values of all the indices

mentioned above appeared in the later growing season and snowmelt periods, respectively, which was consistent with our second hypothesis that the maximum and minimum values of the indices mentioned above would appear in the litterfall and snowmelt periods, respectively. In particular, the litter and litter carbon input to the streams were higher than the output from the streams in the forest-stream ecosystem, which was consistent with our third hypothesis that the litter and litter carbon input to the streams would be higher than the outputs from the streams in forest-stream ecosystems, implying that subalpine forest streams play an important role in carbon sink in the subalpine forest region.

#### *4.1. Dynamics of Litter and Carbon Input in the Subalpine Forest Streams*

Worldwide, the annual litter input to forest streams varies from 3 to 1000 g m<sup>−</sup><sup>2</sup> y<sup>−</sup>1, and the highest and lowest values are observed in a coniferous forest in North Carolina and in a shrub/grass-covered riparian zone of Deep Cr, Idaho, respectively (Table 4) [20–22,47–50]. In this study, the annual litter input to the subalpine forest streams was 20.14 g m<sup>−</sup><sup>2</sup> and ranged from 2.47 to 103.13 g m<sup>−</sup><sup>2</sup> for our one-year investigation. Meanwhile, litter as the carrier of carbon biogeochemical linkages between forest and stream ecosystems [7], our results also showed the average litter carbon input to forest streams was 8.61 mg m<sup>−</sup><sup>2</sup> and ranged from 0.11 to 40.57 mg m<sup>−</sup>2. Annual litter and its carbon input to the forest streams are dominated by the litter production of mountain forest and riparian vegetation through which the streams flow. That is, the higher the litter production in neighboring forest and riparian vegetation, the higher the litter and more litter carbon input to the butted streams. Meanwhile, the streams are linked more closely with the forest ecosystems, and more litter can be inputted to the streams since longer and wider streams can receive more litter. For instance, Wenger et al. (1999) have found that the interception and conversion efficiencies of riparian zones for litter depend on the width of the riparian zones [51]. Previous investigations have also found that subalpine deciduous forests have higher litter production than those in the subalpine shrub and grass vegetation [15]. The subalpine forest region in this study consists mainly of dark coniferous tree species and deciduous coniferous and broadleaved tree species, and litter production varies greatly with forest types [52]. Higher litter production is observed in the Minjiang fir-dominated primary coniferous forest and *Larix masteriana*-dominated deciduous coniferous forest, and the lower value is observed in the rhododendron shrub forest [15]. Consequently, annual litter and litter carbon input to the subalpine forest streams also varied greatly with the investigated streams, resulting from the investigated streams having different length and width, different forest types and riparian vegetation types along the streams (Table 1). In this study, both litter and litter carbon input were significantly and positively correlated with the stream length (Table 2), indicating that the longer the stream, the more litter input to the streams, which was consistent with the litter input. However, the stream width correlated slightly and negatively with litter input owing to the wider streams flowing through the riparian zones where shrub and herb species produce less plant litter. These results sugges<sup>t</sup> that the litter input to the streams from mountain forests plays more important roles in maintaining the structure and function of the subalpine forest stream ecosystem, since plant litter is a pivotal component of stream food webs and ecosystem functioning [3].


**Table 4.** Litter Input to Forest Streams Reported Worldwide.

Litter and litter carbon input to the forest streams also showed two peaks with critical periods, and the maximum and minimum values were observed in the late growing season and seasonal snowmelt season, respectively. A reasonable explanation is that seasonal climate change controls the plant rhythm and in turn dominates the seasonal dynamics of litterfall in mountain forests and riparian vegetation. Two peaks of litterfall in the subalpine forest have been observed in the early and late growing seasons [15,53]. The litterfall peak that occurs in the early growing season contributes to the abnormal litter due to windstorms harming young leaves, while the maximum litterfall peak in the late growing season contributes mainly to the plant rhythm due to plant leaf senescence and falling in autumn [52]. Consequently, higher litter input to the forest streams was observed in the later growing season and early growing season. Correspondingly, lower litterfall was accompanied by lower litter input in the snowmelt season. Meanwhile, stream characteristics are often regulated by seasonal precipitation and then influence the seasonal dynamic pattern of litter input. Heavier rainfall can increase the velocity and width of stream flow, leading to more forest floor litter input to the forest stream. Johnson and Jones (2001) have revealed that the stream water level is a limiting factor for the interception and transformation of elements, and more litter can be inputted to the forest stream when the stream water level rises [54]. Briefly, the litter input to forest streams is affected by seasonal climate change, stream characteristics and plant growth rhythms.

#### *4.2. Seasonal Dynamics of Litter and Carbon Output from Subalpine Forest Streams*

The annual litter output from the investigated subalpine forest streams varied from 0.02 to 22.30 g m<sup>−</sup>2, and the average value was 0.56 g m<sup>−</sup><sup>2</sup> for the investigated streams during this one-year investigation. Correspondingly, the annual litter carbon output varied from 0.01 to 1.51 mg m<sup>−</sup>2, and the average value was 0.16 mg m<sup>−</sup>2. In theory, the litter and litter carbon output from the streams depend mainly on three pathways. Firstly, the stocks of litter in the stream ecosystem determine the output from the streams. In other words, the higher the reserves of litter in the streams, the higher the litter and litter outputs from the streams, implying that the litter and litter input to the streams manipulate the litter outputs from the streams [3,15,30]. Secondly, the decomposition and deposition of plant litter in the stream largely regulate the output of litter. For instance, Yue et al. (2016) have found that plant litter incubated in the subalpine forest streams decomposes completely after two years [7]. Our sediment data at the bottom of the streams (unpublished) showed that more than 95% of the litter was locked in the stream floor as sediments: the more litter that is stored in the form of sediment, the less the output of litter from the forest streams, leading to a lower carbon output. The significantly negative relationships found between litter and litter carbon output and sediment depth in this study were consistent with this (Table 3). Thirdly, litter loss in deeper streams was generally significantly higher for a specific litter species [7,55]. Given these cases, stream characteristics, particularly the sediment depth

and flow velocity, also play key roles in regulating litter carbon output [30]. For example, Bilby et al. (2003) also found that as stream width increased from 5 m to 15 m, the water velocity also increased; as the litter carbon captured by falling wood dropped by 80%, more carbon was outputted from the streams in the study of the Coastal Pacific Northwest of the United States [56]. The positive relationship between litter output and water level and flow velocity in this study was consistent with their results. Overall, forest streams are key bonds of biological matter linkage between mountain forest ecosystems and downstream.

In aquatic habitats, contrary to our expectation, the litter and its carbon output from the forest streams also presented similar seasonal patterns to those of the input. The results can be explained by the following reasons. On the one hand, the factors influencing litter carbon input to streams also modulate litter carbon output from the streams. For example, the peak and nadir values of litter carbon output are also observed in the later growing season and snow melting season, respectively. On the other hand, stream characteristics also play key roles in manipulating the output of litter and litter carbon from the streams, and these characteristics are always regulated by seasonal precipitation and plant rhythm. In this study, the highest values of litter carbon output from the forest streams were observed in the later growing season, owing to the higher flow velocity and litterfall peak in this period, while the lower flow rate, together with less litter production in the snowcovered period, might contribute more to the occurrence of the lowest values during the snowmelt season. Seasonal precipitation, together with seasonal melting of snow cover, dominates the periodic fluctuations in the depth, velocity, and width of forest streams and determines the seasonal dynamics of litter and litter carbon output from the streams. The forest streams act as the source of litter and litter carbon in the butted river, and the output dynamics are complexly regulated by the sink of litter and litter carbon in the streams and the stream characteristics, as affected by seasonal precipitation.

#### *4.3. Litter Carbon Budget in the Subalpine Forest Stream*

The average ratio of litter input to output in the subalpine forest catchment was 188.17, and the average ratio of input to output of litter carbon was 270.01, indicating that forest streams act as the sink of litter carbon in the subalpine forest region. Owing to this huge difference between the input and output, the variables that had a significant effect on litter and litter caron input could explain the significant variability of these ratios in the sampled streams to a grea<sup>t</sup> extent. Possible reasons for this wide gap between litter carbon input and output include the following. First, although much litter was inputted to the forest streams, more than 95% of the litter input was deposited at the bottom of the streams in the form of sediment, as mentioned above, resulting in very little carbon being exported from the streams in the form of litter. Second, the effects of abrasion caused by sediment transport in the streams could be another driver of the higher litter loss, which was also regulated by seasonal precipitation (rainfall and snowfall). This process could accelerate litter decomposition rates in aquatic ecosystems [55], leading to less litter carbon being output in the form of litter. Generally, the decomposition and deposition of plant litter in the stream largely regulate the gap between the input and output of litter carbon. For all of the reasons above, forest streams are indispensable carbon sinks in the subalpine forest stream ecosystem. However, the processes and mechanisms of the stream litter budget need to be fully investigated in the future.

Although the carbon output from the streams was smaller than the carbon input, the dissolved carbon output from the streams was a noteworthy part of the forest ecosystem [36,57]. The litter carbon output from the forest streams in this study was much less than that in the investigation of east Finland (0.62–0.94 g m<sup>−</sup><sup>2</sup> a<sup>−</sup>1) [58], which might be mainly due to the streams in our research having a shorter length and less anti-interference ability, or since much of the carbon stored in the litter was retained in the more complex ecosystems of rivers and lakes rather than in small streams, resulting in a smaller amount of carbon in the forest catchments [59]. Aquatic litter carbon (stream and riparian zone) could connect forest ecosystems with river ecosystems [16,60]. Therefore, investigations of

the litter carbon budget in the forest streams can provide baseline data for further studies on the interconnections between individual aquatic and terrestrial ecosystems within a forest stream ecosystem.
