Local Variation in Ground Arthropod Diversity Rises as Distance to Residential Areas Decreases in a Mature Evergreen Forest

Abstract

Ground-dwelling arthropods interact with vertebrates, plants, detritus, and microbes as important players in forest ecosystems. Human disturbance threatens the diversity of forest arthropods, with varied impacts on different taxa. However, we understand little of the impact of human disturbance on overwintering ground-dwelling arthropod diversity in mature subtropical evergreen forests. In order to test how ground-dwelling arthropod diversity varies by the distance to residential areas, we set 108 pitfall traps along four 100 m transects beginning near residential areas along the edges of a mature subtropical evergreen forest in Central China. We collected 30,616 arthropods, representing 96 morphospecies. The results show that the Shannon, Simpson, and Pielou’s evenness indices, as well as the effective number of species at α = 1 and 2, decrease when the pitfall traps are within 60 m of the residential areas. Moreover, the coefficients of variation in these three indices are higher at the sites closer to the residential areas by 11.54–17.72%. Such high variations in these widely used diversity and evenness indices indicate that estimation bias in arthropod diversity is more likely to occur at sites closer to residential areas. We suggest that different aspects of community composition should be studied to assess the effects of human disturbance on ground-dwelling arthropod diversity.

1. Introduction

Forests conserve a large number of arthropod species [1,2,3,4], which structure the ecological networks that contain mutualistic and antagonistic interactions within forest ecosystems [3,5,6,7,8]. Arthropods are important components of forest ecosystems, with some taxa, such as spiders (Araneae), ants (Hymenoptera: Formicidae), and dung beetles (Coleoptera: Scarabaeoidea), being indicators for forest health and integrity [3,9,10,11]. Understanding how human disturbance, such as deforestation, land use change, and residential activity, affects forest arthropod diversity is crucial for maintaining the ecological functions and ecosystem services of forests [12,13,14,15,16]. The negative effects of human disturbance on arthropod communities can be huge [2,13,17,18]. For example, deforestation can result in the loss of over 6000 arthropod species in a tropical forest [2] and over 700 in a Sahelian forest [19]. But there are findings indicating positive effects [20,21,22] or no effect [23] of human disturbance on arthropod diversity. In some cases, arthropod richness in forests increases during human disturbance due to the introduction of invasive species [22,24,25] and the growth of generalist species [21,26], while some endemic and specialist species may be less common and distribute less widely [27,28]. Some studies show that different taxa of arthropods may respond differently to human disturbance [29,30,31,32,33]. For example, the diversity of spiders decreases as canopy openness increases given the composition shift from generalists to specialists in areas with intact and closed canopies [32,34]. Meanwhile, invasive ants [25] and dung beetles [29,30] are common in fragmented forests, where open canopies, wood debris, and human wastes promote the population growth of these species. Here, populations of some endemic and specialist ants decline as the level of human disturbance increases [26,31]. Within certain orders of arthropods, such as Araneae [35], Diptera [4], and Hymenoptera [36], the variation in diversity at the different levels of human disturbance is high [37].

We have limited understanding whether a high variation in responses to human disturbance is present across the taxa of arthropods, especially among overwintering arthropods [38,39] and those in subtropical forests [18,23]. With the growing human populations in subtropical regions [40,41], deforestation for urbanization increases and is projected to continue increasing [42,43,44]. Land use and other human activities have also become more frequent as the results of increasing human populations in these regions [45,46]. Despite the efforts in afforestation and restoration to achieve a net increase in forested area in Central China, deforestation and fragmentation occur more frequently in mature subtropical evergreen forests to accommodate human populations within this highly populated region [47,48]. These urban forests provide ecosystem functions and services for millions of humans from several mega cities, including Wuhan, Changsha, and Hefei, and from hundreds of surrounding townships [49,50]. But they are more vulnerable to urbanization and industrialization than those forests in Southern and Western China in terms of maintaining ecological functions and ecosystem services [48,51]. Under increasing human disturbance, it is important to understand how human disturbance affects arthropod diversity in mature subtropical evergreen forests in Central China.

Dwelling on the ground or near the land surface, ground-dwelling arthropods are useful indicators for human disturbance in forests because of their sensitivity to different levels of disturbance and efficiency of sampling [37,52,53]. On the one hand, the composition, including abundance [54] and richness [39], of different taxa of ground-dwelling arthropods varies between intact and disturbed forests. Many common taxa of ground-dwelling arthropods, such as spiders [30,32,34], ants [25,55], dung beetles [29,30], crickets (Orthoptera: Grylloidea) [56], flies (Diptera) [4,57], and isopods (Isopoda: Oniscidae) [58], are sensitive to human disturbance. On the other hand, the efficiency of the sampling of ground-dwelling arthropods is higher than that for flying arthropods [52,59,60]. Ground-dwelling arthropods can be collected using pitfall traps and identified according to morphological differences [52]. Pitfall trapping, first developed by Hertz [61] and Barber [62], is the most frequently used sampling technique for ground-dwelling arthropods [52,60]. The two common spatial layouts of pitfall traps are grid and transect [60]. The former evenly covers the sampling location with known separation, while the latter is suitable for testing the impact of abiotic gradients on arthropods [60,63]. By controlling for the edge effect and the effect of habitat heterogeneity on arthropod diversity [64,65], gridded pitfall traps set along transects can be used to measure how different levels of human disturbance affect ground-dwelling arthropod diversity [63,66].

Although some studies on the impact of human disturbance on ground-dwelling arthropods in subtropical forests identified the decline in diversity of some common arthropod taxa [4,25,29,30,34,55,56,57,58,67], little evidence of such impacts on the communities of ground-dwelling arthropods is available [67,68]. Under the rapid change in land use from forests to urbanized area in Central China [43,44,49,51], there is an urgent need to understand the impact of human activities on arthropods [7,22,28], especially among overwintering arthropods [38,39] within urban forests [8,18,26]. Overwintering arthropods are important players in maintaining the ecological stability and ecosystem functions of subtropical forests in winter when other decomposers become less active than in warmer seasons [38,39]. This study aims to assess how distance to residential areas affects overwintering ground-dwelling arthropod diversity, while controlling for the edge effect and the impact of habitat heterogeneity using gridded pitfall traps in a mature subtropical evergreen forest in Central China. The results will improve our understanding of the responses of forest arthropod communities to human disturbance and inform the sustainable management of subtropical evergreen forests to maintain biodiversity, functions, and ecosystem services.

2. Materials and Methods

2.1. Study Area

This study was conducted in a 32-hectare mature subtropical evergreen forest in Central China (30°13′50″–30°13′52″ N, 115°3′45″–115°3′47″ E) with a residential area surrounding it (Figure 1). Within the 1 km distance from this urban forest, there are over 35 thousand residents [69], and the population density ranges from 1050 to 1800 per hectare based on our family visits. The study area is located in the subtropical monsoon climate zone, with a mean annual temperature of 17 °C, and there is 1352 mm of annual precipitation [70]. The forest age is 62 years old, and the dominant tree species is Cinnamomum camphora (L.) J.Presl, with a mean canopy closure of 83% and a canopy height between 16 and 26 m [48]. Elevation ranges from 21 m to 90 m, the slope is between 3 and 18° [48], and the soil type is yellow-brown soil [69]. The soil was documented under the code of B12 in the Genetic Soil Classification of China (GSCC) [71]. It is a type of luvisol with a high organic carbon content, a pH between 5.5 and 7.0, and a cation exchange capacity (CEC) above 24 cmol/kg [72].

2.2. Pitfall Trapping

We prepared 108 pitfall traps using 200 mL transparent polyethylene plastic cups (top diameter 6.2 cm, bottom diameter 4.5 cm, and height 8.2 cm), each containing 100 mL mixture of brown sugar water and 70% ethanol. The 100 mL mixture includes 5 g brown sugar, 15 mL white vinegar, 80 mL water, and 5 mL 70% ethanol, which is effective for attracting and preserving ground-dwelling arthropods [73]. We inserted the pitfall traps into 8 cm wide and 10 cm deep holes on 27 October 2024 and collected overwintering ground-dwelling arthropods for 10 days until 6 November 2024. The daily temperature range during the collection period was 15–25 °C, and the mean daily temperature was 22 °C (±4 °C). The total precipitation during this period was 2 mm, the mean relative humidity was 81% (±6%), and the mean daily near surface wind speed was 2 m/s (±2 m/s) based on the meteorological data recorded by a weather station 1 km away from the study area [70]. The maximum length of consecutive sunny days in December 2024 and January 2025 was less than 5 days, and the study area was closed in February 2025 due to the Spring Festival, which prevented further sampling efforts on the arthropods over this winter.

The 108 pitfall traps were set along four 100 m transects from the edge near residential areas to the inner forest (Figure 1). At the start, midpoint, and end of each transect, three grids of 9 traps were deployed by 3 rows and 3 columns with a 5 m distance between each pair of adjacent traps (Figure 1). Each transect started 5 m away from the nearest residential area, so the distances from residential areas of the three grids are 5–15 m, 50–60 m, and 95–105 m, respectively. To control for the edge effect and the impact of habitat heterogeneity [66], the four transects were set in parallel to forest edges (the distances of 5, 10, and 15 m from the edge) near four residential buildings (Figure 1). We checked that the canopy closure at each pitfall trap was above 80% and the slope of the transect was below 5°. We also recorded the number of pieces of human trash at size > 5 cm and the number of visible pieces of plastic trash within a circle of diameter 5 m from each pitfall trap.

2.3. Diversity Indices

At the lab, we washed out the residue of mixture and dust from samples using 70% ethanol, and preserve the samples in 108 centrifuge tubes, each containing the arthropods from an individual pitfall trap. We then identified each collected arthropod using stereo microscopes (SZ 650 B2L 0.7–4.5× Zoom, Chongqing Optec Instrument Co., Ltd. (Chongqing, China)) based on the taxonomic and morphological characteristics of the samples and those of the identified morphospecies [74]. We sorted the individuals of the same morphospecies into a labeled centrifuge tube preserved in 70% ethanol. A data sheet for the morphospecies composition of the samples from each pitfall trap was prepared for rank abundance assessment and diversity analysis.

Based on the composition data, we first assessed the rank abundance of the morphospecies by fitting null, preemption, Zipf, and lognormal models using the R 4.4.3 package vegan [75]. The model with the lowest Akaike information criterion (AIC) fits the data best. Rank abundance is useful for assessing the representation of samples for a target community [76]. In order to assess the spatial dependence of the samples collected from the 108 pitfall traps, we calculated the Bray–Curtis dissimilarity (Equation (1); BC) of the arthropods between each pair of pitfall traps to evaluate the dissimilarity of species composition [77].

$${BC}_{ij}=1-{\displaystyle \frac{2C_{ij}}{S_i+S_j}},$$

(1)

where S_i and S_j are the numbers of species at the ith and jth pitfall traps, C_ij is the number of species occurring at both the pitfall traps, and BC_ij is the Bray–Curtis dissimilarity between the two pitfall traps. A value of 0 indicates that two pitfall traps share all the species, and a value of 1 indicates that the two pitfall traps do not share any species [77]. Bray–Curtis dissimilarity is useful for assessing if there could be a problem of pseudoreplication in the pitfall traps [78,79].

We then calculated the abundance (total number of individuals; Equation (2); N); morphospecies richness (total number of morphospecies; S); and four indices, namely Shannon’s diversity index (Equation (3); H) [80], Simpson’s diversity index (Equation (4); D) [81], Margalef’s richness index (Equation (5); d) [82], and Pielou’s evenness index (Equation (6); J) [83], of each pitfall trap.

$$N={\sum }_{i=1}^S{n}_i,$$

(2)

where n_i is the number of the ith of S morphospecies collected in the given pitfall trap.

$$H=-{\sum }_{i=1}^S{\displaystyle \frac{n_i}{N}}\mathrm{l}\mathrm{n}\left({\displaystyle \frac{n_i}{N}}\right),$$

(3)

$$D=1-{\sum }_{i=1}^S{{\displaystyle \frac{n_i}{N}}}^2,$$

(4)

$$d={\displaystyle \frac{S-1}{\ln \left(N\right)}},$$

(5)

$$J={\displaystyle \frac{H}{\ln \left(S\right)}},$$

(6)

We also estimated the effective number of species (ENS) [84] at α = 1 and 2 based on inverting Renyi entropy using the R package divo [85]. The effective number of species evaluates diversity as the number of species in an equivalent community with equally abundant species [86]. These six diversity and evenness indices have been widely used to measure and assess the diversity of biological communities [87,88,89], including arthropods [2,21].

2.4. Analysis of Variance

We conducted two analyses to confirm if arthropod diversity was affected by human disturbance and if the variations in arthropod diversity indices were different at different levels of human disturbance. First, we checked the normality using the Shapiro–Wilk test and homogeneity of variance using Levene’s test of abundance, morphospecies richness, and the six diversity indices. We then conducted two-way analysis of variance (ANOVA) to test if abundance, morphospecies richness, and each of the six diversity indices of the 108 pitfall traps vary by the distances to the forest edge (5, 10, and 15 m) and the distances to the residential areas (5, 10, 15, 50, 55, 60, 95, 100, and 105 m). Heatmaps were produced to visualize mean abundance, morphospecies richness, and each of the six diversity indices by the distances to the forest edge and to the residential areas using the R package ggplot2 [90]. Multiple comparison over the estimated marginal means using the R package emmeans was conducted if significant effects were detected from two-way ANOVA (i.e., p < 0.05) [91], while the difference in Shannon index was confirmed based on the modified t-test [92]. Second, the coefficient of variation (CV; Equation (7)) of abundance, morphospecies richness, and each of the six diversity indices within each grid, namely the 9 pitfall traps set by 3 rows and 3 columns, was calculated. One-way ANOVA was then applied to the CV to analyze if CV varies by the distances from the grid centers to the residential areas, i.e., 5–15, 50–60, and 95–105 m. The distance to the forest edge was not considered here because no edge effect was detected from first analysis. Multiple comparison was conducted if the mean of the CV varies by the distance to the residential areas using the R package emmeans [91].

$$CV={\displaystyle \frac{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{d}\mathrm{e}\mathrm{v}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}}\times 100\%,$$

(7)

3. Results

3.1. Community Composition

In total, 30,616 arthropod individuals of five classes (i.e., Chilopoda, Arachnida, Malacostraca, Collembola, and Insecta), 16 orders, and 96 morphospecies were collected and identified (

Table 1. The morphospecies composition of the 12 grids by the five classes of the phylum Arthropoda. The abundance (N) and morphospecies richness (S) of the arthropods collected at each grid are listed. A, B, C, and D are the four transects set in the study area, and the grids numbered 1, 2, and 3 are 5–15 m, 50–60 m, and 95–105 m away from the residential areas, respectively.

	Chilopoda		Arachnida		Malacostraca		Collembola		Insecta		Arthropoda
Grid	N	S	N	S	N	S	N	S	N	S	N	S
A1	9	2	3	2	1	1	23	1	1988	26	2024	32
A2	10	2	13	10	6	1	7	1	3927	30	3963	44
A3	30	2	1	1	5	1	9	1	2541	25	2586	30
B1	14	1	5	3	1	1	20	1	1452	26	1492	32
B2	179	2	12	9	3	1	31	1	2124	26	2349	39
B3	395	1	22	8	3	1	103	1	1518	40	2041	51
C1	14	2	6	5	1	1	15	1	1484	34	1520	43
C2	13	2	7	5	6	1	8	1	1867	24	1901	33
C3	5	2	9	3	0	0	25	1	1142	27	1181	33
D1	51	2	18	4	2	1	18	1	5298	20	5387	30
D2	107	1	8	5	9	1	58	1	3840	23	4022	31
D3	148	1	5	5	2	1	135	1	1860	21	2150	29
Total	975	3	109	22	39	1	452	1	29,041	69	30,616	96

;

Table A1. A summary of the morphospecies richness, abundance, and relative abundance (%) of the 30,616 arthropods of the 16 orders collected from the 108 pitfall traps.

Order	Richness	Abundance	Relative Abundance (%)
Scolopendromorpha	2	974	3.181
Scutigeromorpha	1	1	0.003
Araneae	14	71	0.232
Sarcoptiformes	8	86	0.281
Isopoda	1	39	0.127
Entomobryomorpha	1	452	1.476
Dermaptera	3	38	0.124
Orthoptera	2	40	0.131
Blattodea	2	58	0.189
Thysanoptera	1	3	0.010
Hemiptera	7	1280	4.181
Hymenoptera	20	25,008	81.683
Coleoptera	11	107	0.349
Diptera	20	2055	6.712
Siphonaptera	1	399	1.303
Mecoptera	2	5	0.016

). The number of Insecta morphospecies collected from each grid of pitfall traps accounts for over 71.8% of the morphospecies richness from the same grid. The most abundant order is Hymenoptera, accounting for 81.7% individuals of the samples, followed by Diptera (6.7% of the samples), both of which are the most diverse order with the highest morphospecies richness of 20 morphospecies. There are thirty-three singletons (i.e., morphospecies collected from only one pitfall trap) and seven doubletons (i.e., morphospecies collected from only two pitfall traps), where Araneae (seven singletons and three doubletons) and Diptera (seven singletons and one doubleton) are the two orders with most singletons and doubletons. The abundance of each morphospecies varies from one to nine thousand six hundred and forty-eight individuals, with a mean of three hundred and nineteen and a median of five individuals. For 59 morphospecies, fewer than 10 individuals were trapped by the 108 pitfall traps. The numbers of arthropods collected by the 108 pitfall traps vary from 31 to 1202, with a mean of 283 and a median of 204 individuals. And the numbers of arthropod morphospecies of the 108 pitfall traps range from 5 to 35 with both a mean and a median of 13 morphospecies.

The rank abundance curves of these morphospecies are shown in Figure 2, with the best fitted model as the log-normal model (AIC = 9465.41). The samples represent the ground-dwelling arthropod community in the study area, with three morphospecies of ants being the three most abundant taxa [76]. Regarding dissimilarity in morphospecies composition within and among the grids, Bray–Curtis dissimilarity between each pair of the 108 pitfall traps ranges from 0.100 to 0.882, with only 10 pairs below 0.333, and 2880 out of 5778 pairs ≥ 0.500 (Figure 3). There is no difference in Bray–Curtis dissimilarity among the 12 grids (one-way ANOVA; p = 0.36).

3.2. Variation in Diversity Indices

Heatmaps of abundance, morphospecies richness, and each of the six diversity indices by the distances to the residential areas and the forest edge are shown in Figure 4. Based on two-way ANOVA, only the Shannon index (p = 0.044), the Simpson index (p = 0.021), and Pielou’s evenness index (p = 0.010) of the 108 pitfall traps are significantly different by the distances to the residential areas. No difference is detected in abundance (p = 0.409), morphospecies richness (p = 0.990), or Margalef’s richness index (p = 0.929) by the distances to the residential areas. Additionally, the edge effect or the interaction between the edge effect and the distance to the residential areas is not significant for abundance, morphospecies richness, or any of the six diversity indices of the 108 pitfall traps (each p > 0.05). Multiple comparison confirms that the pitfall traps 100 m away from the residential areas show significantly higher diversity (H and D) and evenness (J) than those 5 m, 15 m, 50 m, and 60 m away (

Table 2. The estimated marginal means of abundance (N), morphospecies richness (S), the Shannon index (H), the Simpson index (D), Margalef’s richness index (d), Pielou’s evenness index (J), and the effect numbers of species at α = 1 and 2 (ENS1 and ENS2) of the 108 pitfall traps by the distances from the residential areas. Under the column, the marginal means are superscripted with letters to indicate the statistical differences by the three distances from the residential areas. If two means share a same letter, this indicates that the two means are not different (p > 0.05), and vice versa (p < 0.05).

Distance (m)	N	S	H	D	d	J	ENS1	ENS2
5	336 a	13 a	1.285 a	0.579 a	2.169 a	0.512 a	3.907 a	2.800 a
10	185 a	12 a	1.466 ab	0.611 ab	2.338 a	0.596 ab	5.022 ab	3.450 ab
15	349 a	13 a	1.344 a	0.590 a	2.268 a	0.524 a	4.222 ab	2.903 a
50	321 a	13 a	1.314 a	0.575 a	2.221 a	0.526 a	4.056 ab	2.954 a
55	320 a	13 a	1.407 ab	0.619 ab	2.291 a	0.544 ab	4.374 ab	3.010 ab
60	378 a	13 a	1.251 a	0.579 a	2.050 a	0.500 a	3.733 a	2.719 a
95	244 a	13 a	1.536 a	0.691 ab	2.230 a	0.620 ab	4.932 ab	3.558 ab
100	206 a	14 a	1.782 b	0.770 b	2.505 a	0.709 a	6.053 b	4.502 b
105	213 a	12 a	1.553 ab	0.696 ab	2.124 a	0.645 ab	4.964 ab	3.681 ab

). In addition, the distance to the residential areas is negatively correlated with the number of pieces of human trash (r = −0.577; p < 0.001; Figure A1A) and with the number of pieces of plastic trash (r = −0.559; p < 0.001; Figure A1B) within the circle of diameter of 5 m from each pitfall trap.

One-way ANOVA confirms that the CVs of the Shannon index (p = 0.032), the Simpson index (p = 0.017), and Pielou’s evenness index (p = 0.049) are different by the distances from the grids to the residential areas, while no difference in the CVs of abundance (p = 0.608), morphospecies richness (p = 0.595), and Margalef’s richness index (p = 0.899) is found among these grids. Multiple comparison shows that the CVs of the Shannon index (p = 0.029 and 0.038), the Simpson index (p = 0.018 and 0.019), and Pielou’s evenness index (p = 0.044 and 0.040) are significantly lower at the grids 95–105 m away from the residential areas than those 5–15 m and 50–60 m away, respectively. The mean differences in the CVs of the Shannon, the Simpson, and Pielou’s evenness indices between the grids 5–15 m and 95–105 m away from the residential areas are 14.23%, 17.72%, and 11.54%, respectively. Those between the grids 50–60 m and 95–105 m away from the residential areas are 13.34%, 17.62%, and 14.28%, respectively. Meanwhile, multiple comparison shows that the CVs of the Shannon index (p = 0.982), the Simpson index (p = 0.999), and Pielou’s evenness index (p = 0.867) of the grids 5–15 m away from the residential areas are not different from those of the grids 50–60 m away from the residential areas (Figure 5).

4. Discussion

From the 30,616 individuals collected at the 108 pitfall traps, we identify 96 morphospecies belonging to overwintering ground-dwelling arthropod communities in the 23-hectare mature subtropical evergreen forest (Figure 1). The 96 morphospecies follow a log-normal distribution according to the rank abundance curve (Figure 2). It can be expected that more morphospecies can be identified if the sampling efforts improve, e.g., setting more pitfall traps and extending the collection period, given that 33 singletons are present in our samples [93]. However, despite the unfavorable weather conditions in the subsequent two months and the temporary closure of the study area due to the Spring Festival, the sampling efficiency of the 108 pitfall traps is optimized to answer if the diversity of ground-dwelling arthropods decreases at the sites closer to the residential areas for the design of this experiment [60,73] and the collection period [39,55]. The samples collected from these pitfall traps (Table 1) indicate the high sampling efficiency of the design [66]. We collected the ground-dwelling arthropods in late autumn when the spatial variation in arthropod diversity within evergreen forests is lower comparing to that in spring and summer [38,94], making them good indicators for the impact of human disturbance on different taxa [39]. Although some studies on ground beetle diversity imply that the species composition of beetles collected by pitfall traps set within 20 m is similar, and replications by pitfall traps could be at the risk of pseudoreplication [95], the Bray–Curtis dissimilarity of morphospecies composition between nearly half of the pairs of pitfall traps is over 0.5, indicating a low risk of pseudoreplication in this study [78,79].

In this study, the use of multiple diversity and evenness indices, e.g., the Shannon and the Simpson diversity indices, effective numbers of species, and Pielou’s evenness index, in addition to abundance and morphospecies richness (Figure 4) is essential for analyzing the composition data based on pitfall traps with disturbance regimes [13,32,88,89]. Following the previous studies reporting a decline in diversity of many arthropod taxa, including spiders, ants, and flies, during human disturbance in subtropical forests [4,25,30,34,55,56,57,58,66], we also find that ground-dwelling arthropod diversity is lower at the sites closer to the residential areas than those 100 m away (Table 2). Higher levels of disturbance at the sites closer to the residential areas, such as light, noise, and solid waste, may cause the distribution shift of endemic species and specialist species from these sites to less-affected habitats [21,26]. Meanwhile, the populations of invasive species and generalist species which benefit from reduced competition and open niches as a consequences of human disturbance grow fast and become dominant species at disturbed sites [22,25,27,30,66]. These two processes of composition change are likely the causes of the decline in the Shannon, the Simpson, the effective number of species, and Pielou’s evenness indices in the study area [89]. Other than the impact of human disturbance, no edge effect or interaction between the edge effect and the impact of human disturbance is identified in this study. The spatial distribution of ground-dwelling arthropods within the forest is less affected by the edge effect than the soil profile [17], vegetation heterogeneity [38], canopy openness [34], forest age [65], and the stand characteristics [64], which are of less concern in this study.

In addition to the decline in the Shannon, the Simpson, and Pielou’s indices identified at the sites 5 m, 15 m, 50 m, and 60 m away from the residential areas (Table 2), the increase in the coefficient of variation in these three indices at the grids 5–15 m and 50–60 m away from the residential areas (Figure 5) implies that the impact of human disturbance on ground-dwelling arthropod diversity may be more complex than previously identified [7,8,18,22,26,28]. The higher variation in diversity indices of the grids closer to the residential areas than those 95–105 m away from the residential areas indicates that it is less likely to detect the impact of human disturbance at disturbed sites and more likely to reach a biased estimation of arthropod diversity at these sites. A larger fluctuation in population dynamics and higher variation in the species distribution of different taxa of ground-dwelling arthropods sometimes can be detected at disturbed habitats and fragments of forests [11,31], and sometimes cannot [7,18]. To reduce the chance of making a type II error (i.e., unable to reject the false null hypothesis), more sampling efforts, such as setting more pitfall traps per grid and repeated sampling, are needed for studies on the impact of human disturbance on ground-dwelling arthropod diversity [63,66]. We also suggest that the variation in diversity indices of sites at different levels of human disturbance should be considered in addition to the diversity indices themselves. Moreover, the temporal variation in arthropod diversity at disturbed and intact sites should be further investigated and incorporated for a comprehensive understanding of the impact of human disturbance on ground-dwelling arthropods [19,23].

Although the identification of morphospecies is sufficient for conducting the analysis of the impact of human disturbance on arthropod abundance, diversity, and evenness [2,21], if possible, molecular techniques in taxonomy should be used for species identification [4,96]. The DNA barcoding technique based on the cytochrome c oxidase subunit I (COI) gene can be used to verify the morphological identification of species [97]. It is also helpful to build a COI database of the arthropods collected from the study area for arthropod metabarcoding and environmental DNA analysis [4]. However, the numbers of individuals of the 59 morphospecies (accounting for 61.5% of the total morphospecies richness) collected from the pitfall traps are fewer than 10, which made it difficult to conduct a polymerase chain reaction (PCR) for these morphospecies [4]. We therefore will keep the samples in 70% ethanol and wait for the next sampling season to collect more individuals for these morphospecies before building a COI database for the arthropods. Under the increasing rate of urbanization in Central China [44,49,51], we need urgently and comprehensively understand how the arthropod community in the forests of this region responds to human disturbance [8,18,26]. Future studies on the spatiotemporal variation in multiple aspects of arthropod diversity, such as species composition, species turnover, and functional diversity, within different human disturbance regimes using optimized sampling methods and advanced techniques in taxonomy are needed to guide the sustainable management of subtropical evergreen forests.