*2.2. Experimental Design*

From 2005 to 2008, pure cypress plantations with similar aspects, slopes, and altitudes were selected for managemen<sup>t</sup> experiments, including strip filling and ecological thinning practices. In the strip filling practice, after forming a series of 4–10 m bandwidths through artificial logging, native broadleaved tree species, such as *Alnus cremastogyne* and *Camptotheca acuminata*, were replanted for striping mixing. In comparison, the ecological thinning practice involved randomly cutting down part of the cypress trees with a removal of 10%–25% of the initial basal area. Meanwhile, the logging residues under both practices were removed from the experimental areas. Previous studies have found that there is a better ecological effect (e.g., high plant diversity and soil nutrient content) in the pattern of strip filling with a bandwidth of 6 m and in the pattern of ecological thinning with a thinning intensity of 20%–25% [33,34]. Accordingly, in the present study, we intended to further study the characteristics of litterfall under these two patterns.

In August 2019, five replicate plots (20 × 20 m in size and 50 m apart) were established in plantations with strip filling, ecological thinning and no artificial interference (pure forest), respectively. The stand characteristics of the three sampling plantations in 2019 are listed in Table S1. The meteorological data (wind speed, precipitation, temperature and relative humidity) from the Yanting Agro-ecological Station of Purple Soil, Chinese Academy of Sciences, from 2019 to 2020, are shown in Figure S1.

## *2.3. Litter Collection*

Six litter traps (1 × 1 m in size and 50 cm above the ground) made of nylon materials with a mesh size of 1 mm were distributed randomly in each plot for the three plantations. Specifically, in the strip filling practice, six litter traps were placed equally within the strip cutting area and an adjacent uncut area. Litterfall was collected approximately monthly from September 2019 to August 2020. These monthly subsamples were pooled to create composite three-month samples in approximately the early dry season (January– March), late dry season (April–June), early wet season (July–September) and late wet season (October–December), according to the tropical forest division [15] and previous climatic data for the region. All litter materials were classified as leaf litter (including broad leaf and needle leaf litter), twig litter (including bark), reproductive organ litter (including flower, fruit and seed litter) and miscellaneous litter (unidentified components, such as indistinguishable detritus, dead animals and insect excrement). All litter materials were oven-dried separately at 65 ◦C, and the total litterfall was calculated as the sum of the dry litter mass for all litter components.

## *2.4. Statistical Analyses*

Repeated-measures analysis of variance (ANOVA) was employed to test for effects of sampling time and plantation on litterfall production. One-way ANOVA with Tukey's honestly significant difference (HSD) test was used to test for the effect of plantation or season on litterfall production and the proportion of each litter component relative to total litterfall. Meanwhile, to calculate the relative influence of each factor, we employed redundancy analysis (RDA), using the "vegan" package, and hierarchical partitioning analysis using the "rdacca.hp" package in R [35]. Considering the consistency of the climate in each plantation and little within-year variation in vegetation factors, we analyzed the relative effects of vegetation factors on annual litterfall production and climate factors on monthly litterfall production. All statistical analyses were performed using R 4.0.5 (R Core Team 2021).

The allometric approach was used to analyze annual litterfall production, which was expressed as Y = β Xα or logY = α log X + log β, where X and Y represent the total litterfall production and its different components (leaf, twig and reproductive organ litterfall) of each planation, respectively; α is the regression slope (scaling exponent); and β is the regression intercept. When the 95% confidence interval contained 1, α indicated that the scaling relationship between Y and X was isometric; otherwise, the relationship was interpreted as allometric [36].

After we log10-transformed the litterfall production data, a model II regression method (i.e., reduced major axis regression, RMA) was applied to estimate the regression parameters and test whether the slopes were significantly different from 1, as well as different among the plantations [37]. When the slopes for the different plantations did not differ significantly, a common slope was first provided, and then the Wald test was performed to test whether there were intercept or coaxial drifts between plantations. The model II regression analysis was performed in SMATR Version 2.0.
