2.1 Study site

The study was performed in a mixed deciduous forest on a southwest-facing hillside on Jigong Mountain (32°06′44″N, 114°02′21″E, 226 m a.s.l.), central China. The area has a typical monsoon climate with a mean annual air temperature of 15.2°C, warmest monthly mean air temperature in July (33.6°C), and coldest monthly mean temperature in January (1.9°C). Mean annual rainfall of this region is approximately 1,063 mm, with 60% falling from May to September (China Meteorological Data Sharing Service System, http://data.cma.gov.cn). The soil is classified as a yellow-brown sandy-loam soil (Miao et al., 2019). The vegetation at the experimental site is dominated by Morus alba Linn., Liquidambar formosana Hance, Diospyros kaki Thunb, and Quercus acutissima Carruth (Chen, Cai, Zhang, Rao, & Fu, 2017).

2.2 Plot design

In July 2016, one 50 m × 50 m plot was established in a mixed deciduous forest with homogeneous and flat topography. The background values of soil organic C and total nitrogen (N) at 0–10 cm depth were 16.6 and 1.91 g kg⁻¹, respectively. Canopy height averaged 18.0 m, and the mean diameter at breast height (DBH) of the living trees (DBH ≥ 5 cm) was 27.8 cm. The mean stand density was 487.4 trees ha⁻¹. In this plot, tree species of M. alba, L. formosana, D. kaki, and Q. acutissima accounted for 5.2, 27.3, 16.4, and 43.4% in total aboveground biomass estimated by allometric equations, respectively. The SLA of undamaged leaves for M. alba, L. formosana, D. kaki, and Q. acutissima was 192.3, 174.6, 101.5, and 95.7 cm² g⁻¹, while the values for damaged leaves was 212.6, 186.6, 93.4, and 83.1 cm² g⁻¹, respectively (Table 1).

2.3 Litter collection

Six trees with DBH > 10 cm for each of the four species (M. alba, L. formosana, D. kaki, and Q. acutissima) were randomly selected in the plot. Three 1 m × 1 m litter traps were placed under the canopy of each selected tree from three different directions. Freshly fallen leaf litter of the four species was collected three times each month from October to November 2016. Leaf litter was air-dried at room temperature (~25°C) and pooled by species. Leaves of each species were sorted into two groups by the presence or absence of herbivory. Leaves with obvious damage by chewing insects were separated out as herbivore-damaged litter (hereafter, damaged), and leaves without any damage either by insects or physical breakage were classified as undamaged litter (hereafter, undamaged). Based on our previous observation, Lampronadata cristata and Ochrostigma albibasis, feeding on live leaves, are two main leaf-chewing insects in the deciduous forest in this region. The potential differences in morphology and element induced by different insect species are not analyzed in this study. Leaf with perforation area accounted for 30–40% of the total area based on visual judgment was selected as damaged leaf litter. Three randomly taken subsamples out of each group of different species were used for the determination of initial total C and N analysis using the dichromate oxidation method (Chen, Cao, Zhou, Li, & Fu, 2019; Nelson & Sommers, 1996) and micro-Kjeldahl technique (Bremner, 1996; Chen et al., 2019), respectively.

2.4 Litter decomposition

We used the litterbag method to study the decomposition process of damaged and undamaged litters for different species. Litter bags made out of nylon screen (1 mm mesh size, 20 cm × 25 cm) were filled with 6.0 ± 0.1 g of air-dried leaf litter. A total of 96 litterbags [4 species × 2 groups (damaged/undamaged) × 4 harvest × 3 replicates] were used for the decomposition experiments. Litterbags were prepared in the laboratory and transported in individual plastic bags to the field site in January 2017. After carefully removing loose forest floor litter, labeled litterbags were placed on the soil surface and held in place by nails. Litterbags were separated from each other by at least 40 cm and randomly distributed in the field. Three replicated litterbags for each group were retrieved in four-time steps after 90, 127, 250, and 400 days of decomposition in the field. After retrieval, the remaining litter was carefully cleaned from adhering mineral soil by gently brushing, and dried at 60°C until constant mass and weighed. Mass remaining rate (%) was calculated as the ratio of actual dry mass of litter in the bags to the initial dry mass. The oven-dried remaining litter of each collection was milled to powder for total C and N analysis by using the dichromate oxidation method (Nelson et al., 1996) and micro-Kjeldahl technique (Bremner, 1996), respectively.

2.5 Statistical analysis

Annual decomposition constant (k, yr⁻¹) was calculated with a single exponential model by the equation (Olson, 1963):

Where: M_t is the mass remaining at time t; M_o is the initial mass value; k is the exponential decay coefficient; t is time in days.

Total loss of C and N (%) from leaf litter in each litterbag during decomposition process was calculated as:

Where: M₀ and M_f are the initial and final dry weight, and C₀ and C_f are the initial and final C concentration of litter, and N₀ and N_f are the initial and final N concentration of litters.

All data were first tested for normality and homogeneity of variance, then two-way ANOVA was conducted to assess the effects of species and herbivory on initial quality, decomposition constant, and total C and N loss across the decomposition process. In addition, we used One-way ANOVA to analyze the differences in initial quality, mass loss rate, decomposition constant, and total C and N loss between undamaged and damaged litter. Multiple comparisons of Tukey tests were performed to examine the differences of decomposition constant, and total C and N loss among different species. To test the possible effect of initial litter quality on the decomposition process, linear regression was fitted between the decomposition constant and initial C and N content and C:N ratio. In order to assess the response sensitivity of decomposition constant to litter quality, we compared the difference of slopes illustrating the dependence of decomposition constant on litter quality between herbivory-damaged and -undamaged leaf litter using an analysis of covariance (ANCOVA). All statistical analyses were performed by SPSS 19.0 (SPSS Inc. Chicago, Illinois) for Microsoft Windows.

3.1 Variability of initial litter quality

Initial litter chemistries varied with both species and herbivory (Figure 2). Initial litter C concentrations for D. kaki (47.3%) and Q. acutissima (47.8%) were significantly higher than that for M. alba (43.7%) and L. formosana (45.2%; p < .01, Figure 2a). Herbivore-damaged leaf litter, on average, had higher C concentration (46.9%) than undamaged leaf litter (45.1%) (p < .01, Figure 2a), with significant differences for M. alba and L. formosana (p < .01, Figure 2a). However, litter initial C content showed no differences between damaged and undamaged litters for both D. kaki and Q. acutissima (p > .01, Figure 2a). Initial N concentration of leaf litter was significantly affected by species, herbivory, and their interaction (all p < .01, Figure 2b). For damaged and undamaged leaves, M. alba had a substantially higher N content (2.05%) than the other three species (p < .01, Figure 2b). Litter N content was 0.65% higher in damaged litter than undamaged litter for M. alba and 0.14% lower in damaged litter than undamaged litter for Q. acutissima (both p < .05, Figure 2b). Initial C:N ratio of leaf litter differed among species, and the magnitude depended on the presence and absence of herbivory (Table 1). M. alba showed a substantially lower C:N ratio (21.8%) than the other three species (p < .01, Figure 2c). The C:N ratio of the damaged litter was 5.78% lower than the undamaged litter for M. alba, and 5.35% higher than the undamaged litter for Q. acutissima (both p < .05, Figure 2c). For L. formosana and Q. acutissima, neither initial N content nor C:N ratio showed differences between damaged and undamaged leaf litters.

3.2 Litter mass remaining

Across our 400 days of incubation, litter mass remaining showed significant differences among species. Irrespective of the herbivory effect, Q. acutissima showed the continuously highest values of mass remaining (94.5–45.4%) and M. alba had the lowest values (55.2–16.6%, Figure 3) across the study period. The average mass remaining of damaged leaf litter (78.8–32.9%) was higher than that of undamaged leaf litter (76.3–29.3%) throughout the studied period. Herbivory effect on litter mass remaining of different species varied with incubation time. Herbivore damage increased litter mass remaining of D. kaki by 3.5% (p < .10) and Q. acutissima by 12.2% (p < .05) at the end of incubation, respectively (Figure 3c,d). For M. alba, herbivore damage showed positive effects on litter mass remaining at 90 days incubation (Figure 3a). D. kaki had significantly higher mass remaining in damaged litter than undamaged litter after 127 and 250 days incubation (Figure 3c). Across the 400-day experiment, litter mass remaining of L. formosana in damaged litter showed no difference with that in undamaged litter.

3.3 Decomposition constant

The decomposition rate, expressed as decomposition constant, showed substantial differences among four species, with the highest value for M. alba, and the lowest value for Q. acutissima (p < .01, Figure 4a). Furthermore, herbivory effects on the decomposition constant depended on species. Damaged leaf litter showed higher decomposition constant for M. alba and L. formosana with greater SLA, whereas D. kaki and Q. acutissima had a lower decomposition constant for damaged leaf litter compared with undamaged leaf litter (all p < .05, Figure 4a).

3.4 Total C and N loss

Similar to the patterns of decomposition constant, total C and N loss displayed significant differences among species, with the greatest values for M. alba and the lowest values for Q. acutissima (p < .01, Figure 4b,c). Herbivory damage increased total N loss by 5.0 and 12.4% for M. alba and L. formosana across our 400 days decomposition, respectively (p < .05, Figure 4b). By contrast, however, total N loss of damaged litter was 7.2 and 15.5% lower than undamaged litter for D. kaki and Q. acutissima, respectively (p < .05, Figure 4b). Compared with the undamaged litters, greater total C loss was found for damaged ones of L. formosana, whereas lower values were displayed for damaged ones of D. kaki and Q. acutissima (p < .05, Figure 4c).

3.5 Decomposition constant versus initial litter quantity

According to the linear regression analyses, initial C content of leaf litter were negatively correlated with decomposition constant for both undamaged (R² = .27, p < .05) and damaged leaf litter (R² = .34, p < .05, Figure 5a). However, initial N content had positive effects on the decomposition constant of undamaged (R² = .72, p < .05) and damaged leaf litter (R² = .66, p < .05; Figure 5b). In addition, significant negative effects were found between initial C:N ratio and decomposition constant for both undamaged and damaged leaf litter (Figure 5c). Compared with the undamaged litter, herbivory-damaged leaf showed lower slope of the relationship between decomposition constant and initial litter N content (ANCOVAs: F = 4.905, p = .039, Figure 5b).

4.1 Herbivory effect on litter quality

It has previously been shown that insect herbivory can lead to substantial changes in leaf physicochemical properties (Ohgushi, 2005). For example, in a semiarid woodland, insect herbivory substantially enhanced N concentration and reduced C:N ratio of pinyon pines (Chapman et al., 2003). Similarly, in African savannas, outbreaks of mopane worm showed a positive effect on leaf N content of Colospbosperum mopane (de Swardt et al., 2018). In contrast, however, a decreased foliar N content following herbivory attack was found for Quercus rubra in North Carolina (Frost & Hunter, 2008) and for grass in riparian meadows (Sirotnak & Huntly, 2000). Therefore, the response of leaf litter quality to insect herbivory varied with tree species and ecosystem types. In agreement with previous studies, we also found that herbivory effect on leaf litter quality differed among tree species with different SLA. Increased leaf N content was found for M. alba (high SLA), but decreased leaf N content was found for Q. acutissima (relative low SLA) after herbivory attack (Figure 2). According to the study by Chapman et al. (2003), the insect-feeding leaf generally drops prior to normal abscission, thus the increased litter quality might be caused by the incomplete nutrient resorption (Zvereva & Kozlov, 2014). Notably, all leaf litters used in this study were collected at the end of the growing season. It means that the incomplete nutrient reabsorption may persist even though the leaf does not defoliate, probably due to the insect damage on the vein. Compared to low-SLA leaves, high-SLA leaves always show thinner thickness and finer vein, and thus more vulnerable to herbivory chewing. Therefore, lower probability of nutrient reabsorption could be expected for high-SLA leaves. Another possible explanation for the reduced leaf N content and an increased C:N ratio for Q. acutissima may be the increased proportion of recalcitrant content during the insect-feeding process (Uriarte, 2000). More nutrients may be allotted to synthesize structural matters (cellulose/lignin) for Q. acutissima due to its greater leaf thickness. Previous study has shown that insect herbivory effect on litter quality varied between evergreen coniferous trees and deciduous broadleaf ones (Chapman et al., 2006). Our finding corroborates that litter quality responses to herbivory vary even among different deciduous broadleaf trees probably due to the variation in leaf traits.

4.2 Herbivory effect on litter mass remaining

Litter mass loss represents the sum of many biotic and abiotic processes, which is largely related to the numerous chemical compounds of litter. The finding that the response of litter mass remaining to herbivory varies with species at the end of the decomposition (Figure 3) could be explained by the contrasting initial litter quality. Herbivory damage enhanced leaf C:N ratio of Dk and Qa (Figure 2), thus showed positive impact on litter mass remaining, which is consistent with the reports that nitrogen-poor litter generally decomposes more slowly than nitrogen-rich litter (Berg et al., 1993). For each species, herbivory effects on litter mass remaining varied with decomposition stage (Figure 3), which may be attributed to the following aspect. First, differences in specific chemical constituent (e.g., C:N and micronutrient) between damaged and undamaged leaf can lead to the variation in mass loss of each stage. Second, differences in environmental factors, such as temperature and decomposers, during each decomposition stage may also influence the litter mass loss.

4.3 Herbivory effect on litter decomposition constant

Previous studies have shown that the influences of insect herbivores on litter decomposition varied with ecosystem types. For example, an increased litter decomposition has been found in grassland (Wang et al., 2018) and coniferous forest (Chapman et al., 2006), while insect feeding showed no effects on leaf litter decomposition in neotropical rain forest (Cárdenas et al., 2015) and Malaysian tropical forest (Kurokawa & Nakashizuka, 2008). These contrasting effects may relate to plant functional trait and initial quality. As an evidence, our findings showed that the impacts of herbivory damage on litter decomposition rate varies with species SLA, showing positive effects on M. alba and L. formosana, and negative effects on D. kaki and Q. acutissima (Figure 4a). The differential responses could be attributed to the changes in litter quality induced by herbivory damage. The decomposition constant was positively dependent on litter initial N content and negatively related to the initial C:N ratio (Figure 5). The increased leaf N content of damaged litter for M. alba and L. formosana can offer more nutrient sources for microbes activity. It has been reported that higher leaf N content generally elevated leaf palatability for the aboveground consumers (Schädler, Jung, Auge, & Brandl, 2003) as well as belowground decomposers (Talbot & Treseder, 2012). Therefore, leaf litters with high N content show greater decomposition constant. In addition to the reduced leaf N content for D. kaki and Q. acutissima, herbivory-induced secondary compounds such as polyphenols, alkaloids, and terpenes (Berenbaum, 1995; Chapman et al., 2006), which may also contribute to the depressed decomposition rate.

By comparing the slopes of the regression models between decomposition constant and initial quality, we assessed the differences in response sensitivity of decomposition constant to initial quality between herbivory-damaged and -undamaged leaf litters. The dependence of decomposition constant on initial N content for the damaged leaf litter showed a lower slope (Figure 5b), which indicates that herbivory damage decreased the sensitivity of decomposition rate to the changes in leaf N content.

4.4 Implication for forest management

In this study, we investigated the short-term effects of herbivory on leaf litter decomposition, which implies that herbivory foraging shows negative influences on the decomposition of Q. acutissima but positive effects on L. formosana (Figure 4). As two dominant tree species in this region, the potential fluctuation of proportion with forest natural succession may lead to uncertainty in forest nutrient cycling due to the different responses of decomposition rate to herbivory damage. Given the suppression of herbivory damage on the decomposition of Q. acutissima, two aspects should be considered for forest managers to maintain nutrient balance. First, both the intensity and frequency of insect outbreak should be properly managed in Q. acutissima dominated forest. Second, the seedling of Q. acutissima could be partly removed during forest tending to release resources for other species. Similar to the report by Chen and Frank (2020), our previous study conducted in this forest also demonstrated that soil CO₂ emission during herbivory outbreak was 36.3% higher than that without herbivory disturbance forest (Liu et al., 2017). Herbivory-mediated organic matter in particulate and dissolved forms contributes considerably to the overall throughfall input of organic substances into the forest soil (Michalzik & Stadler, 2005). To maintain the health and stability of forests, the influence of natural disturbance such as insect herbivores outbreak, have to be taken into account by forest management and policies.