Genetic diversity analysis of the natural regeneration loci of Liriodendron chinense in artificial mixed forests in the rocky desertification area of Western Hunan

Plant materials and DNA extraction

The site located in the Wuling Mountain area of southwest China, serving as the national long-term research base for comprehensive management of rocky desertification (29°3′4″N, 110°13′35″E, 467 m a.s.l.). We measured the diameter at breast height (DBH) of L. chinense with a diameter ≥ one cm in the plot and recorded their coordinates by using real-time kinematic technology (RTK). And we utilized ArcGIS (Esri, Redlands, CA, USA) software to create a population L. chinense distribution map based on individual distribution coordinates (Fig. 1).

We collected 318 one-year-old leaves from tagged L. chinense individuals, including a few L. chinense trees artificially planted in 1978 and their naturally regenerated individuals. Leaf tissues were stored at −80 °C for subsequent DNA extraction and were ground into powder using liquid nitrogen. Genomic DNA was extracted from L. chinense using the DP320 DNA Secure Plant Genomic DNA Extraction Kit (TIANGEN DNAsecure Plant Kit DP320-03; TIANGEN, Beijing, China). The quality of the DNA samples was determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis.

EST-SSR Primers, PCR Amplification, and microsatellite genotyping

The primers used in the experiment refer to the NCBI database expressed sequence tag-simple sequence repeat (EST-SSR) (http://www.ncbi.nlm.nih.gov/dbEST/index.html) primers and combined with the sequence information of SSR primers developed in published paper (n = 202) (Hirata et al., 2006). The target DNA fragments were amplified using the simple sequence repeat markers with tailed primer M13 (TP-M13-SSR) polymerase chain reaction (PCR) method (Supplement information).

The PCR amplification was performed using a thermal cycler (Bio-Rad, Hercules, CA, USA), and the PCR products were analyzed using an ABI 3730XL DNA analyzer (Applied Biosystems, Waltham, MA, USA). Sequencing was conducted using an ABI 3730 XL automated sequencer (Applied Biosystems) with LIZ-500 employed as the internal standard. Consequently, eleven pairs of highly polymorphic primers were selected for subsequent experimental research (Table 1).

Genetic diversity analysis

Peak profiles from capillary electrophoresis were read, verified, and genotyped using GeneMarker software version 2.6, generating a data matrix that served as the basis for subsequent analyses (Holland & Parson, 2011). We used GenALEx 6.5 software to analyze genetic diversity, the following data were obtained: number of alleles (N_A), effective alleles (N_E), Shannon index (I), observed heterozygosity (H_O), expected heterozygosity (H_E), genetic distance, and geographical genetic distance (D). And we used GenALex 6.5 software to calculate the fixation index (F), in-breeding coefficient within the population, genetic differentiation coefficient (F_ST), total inbreeding coefficient (F_IT) for the final identified loci in the L.chinense population, while calculating gene flow (Nm) between loci using the formula Nm = (1−F_ST)/4 F_ST, and testing whether the selected 11 pairs of SSR primers deviate from Hardy-Weinberg equilibrium (HWE) (Khan et al., 2021). Polymorphic information content (PIC) was calculated using CERVUS 3.07 software (Kalinowski, Taper & Marshall, 2010). Invalid alleles (F(null)), also known as zero loci, were detected using Micro-Checker software (Van Oosterhout et al., 2004).

Analysis of spatial genetic structure

The spatial genetic structure (SGS) of L. chinense across different diameter classes and for the entire population within the plot was analyzed using SPAGeDi software version 1.5 (Sofletea et al., 2020). Pairwise kinship coefficients (F_ij) for L. chinense individuals within the sample plot were calculated based on multiple loci, where F_ij represents the probability of identity between the genotypes of two random individuals, i and j, sampled from the population (Loiselle et al., 1995). Based on the number and spatial distribution of individuals in the population, the distance classes were defined such that the number of individual pairs was approximately equal across all classes. Finally, 15 distance classes were divided into 10, 16, 22, 28, 34, 40, 46, 52, 58, 64, 70, 77, 85, 96 and 180 m. We calculated the average F_ij value for each of the 15 distance classes, plot the regression of ln(rij) to obtain the regression slope b_F (where rij represents the spatial geographical distance between samples i and j), confirm the linear positive correlation between F_ij and ln(rij), and verify the significant deviation of genotype from random distribution through 10,000 permutations (Rousset, 2000). The strength of the SGS in L. chinense was assessed utilizing the Sp statistic, which represents the standard error of the regression of pairwise kinship coefficients on distance. The calculation formula for the Sp statistic is Sp = −b_F/(1−F1), where b_F is the linear regression slope of F_ij on the natural logarithm of distance classes, and F1 is the average kinship coefficient of all pairs within the first distance class (Rousset, 1997; Vekemans & Hardy, 2004). In addition, for the analysis, each distance class must include data from more than 30 individuals to ensure the statistical validity of the analysis when considering diameter classes (Doligez & Joly, 1997). The L. chinense population was divided into five diameter classes (Table 2). Since the number of individuals in Class V is less than 30, Classes IV and V were combined for analysis.

Parentage analysis

The probability of identity (PID) and the probability of identity among siblings were calculated using GenALEx software version 6.5 (Peakall & Smouse, 2006). Geographical distances between individuals were calculated based on their spatial coordinates. These data were then used to statistically estimate dispersal distances for both seed and pollen flow. Parentage analysis was performed using CERVUS software version 3.07, which employs a maximum likelihood approach. It determines the parent that best matches the offspring by estimating the logarithm of the likelihood ratio (LOD) between the offspring and the parents. The parent with the highest LOD value is determined to be the most likely parent, while the simulator assesses the significance of the most likely parent based on the calculated critical value (Marshall et al., 1998). In parentage analysis with unknown parental sexes, the optimal parental combination is identified as the one with the highest and most significant offspring-parent trio LOD score (Trio LOD score). The relationship between gene flow (dg) and pollen dispersal (dp), ds is calculated as: dg² = ds²+0.5dp². Distances and probabilities of seed and pollen dispersal within a 180-meter radius were compared. Dispersal pattern maps for pollen and seed flow were generated based on the spatial coordinates of offspring, male parents, and female parents.

L. chinense is a deciduous species belonging to the Liriodendron genus of the Magnolia family, which is monoecious, with pollen primarily dispersed by insects, and reproduces sexually through seeds. The fruit is a compound samara, with small, winged fruits that rely on wind and gravity for dispersal. Based on the methods of Dow and Ashley, and considering the dispersal characteristics of the L. chinense seeds (reliant on wind and gravity) and pollen (reliant on insect and wind dispersion), three hypotheses are proposed for parentage determination as followed: (1) when the trio LOD scores are significant, both parental candidates are identified. The candidate that is geographically closer to the offspring is inferred to be the maternal parent, and the more distant candidate is inferred to be the paternal parent; (2) if the trio LOD scores are not significant, but a significant pairwise LOD score is obtained with one candidate parent, that candidate is inferred to be the maternal parent of the offspring; (3) if neither the trio LOD scores nor any pairwise parental LOD scores are significant, no parental candidates for the offspring could be identified within the sampled population. For the offspring, the geographical distance between the male parent and female parent represents the distance of dp, while the geographical distance between the female parent and the offspring represents the distance of seed dispersal (ds).

Genetic diversity of Liriodendron chinense

A total of 128 alleles were amplified from 318 Liriodendron individuals, representing two populations, using 11 SSR primers. Primer LT089 yielded the fewest alleles (five), whereas primer LT131 yielded the maximum number (20). The Shannon diversity index (I) for the LT131 locus was the highest at 2.651, whereas the I value for the LT002 locus was the lowest at 1.387, resulting in an average I value of 1.940 (Table 3). The average values for the number of alleles (N_A), effective alleles (N_E), observed heterozygosity (H_O), expected heterozygosity (H_E), fixation coefficient (F), and null allele frequency were 11.636, 6.414, 0.673, 0.812, 0.171, and 0.0981, respectively.

The F_ST values of L. chinense showed significant differences (Table 4), with a range of 0.006 to 0.068, and the average genetic differentiation coefficient among loci was 0.031. These results indicate that the genetic differentiation of L. chinense is not significant within a 4.5 m range; although genetic variation exists among loci, it accounts for only a small proportion of the total variance, as the majority of the genetic variation was found within loci. Wright-Fisher model suggested that when Nm > 1, genetic drift can be avoided, thereby preventing genetic differentiation between loci (Ishida & Rosales, 2019). Based on the F_ST values, the range of Nm among L. chinense loci at different loci varied from 3.400 to 43.696, with an average of 14.972, indicating a high level of gene flow among loci. The fixation index (F_IS) for the selected loci of L. chinense is greater than 0, with an average fixation coefficient of 0.150.

Spatial genetic structure

The SGS intensity of L. chinense within each diameter class and the entire 2.5 hm² population was calculated (Table 5). The SGS intensities for diameter Classes I, II, III, IV, and V, as well as the entire population, were 0.0298, 0.0220, 0.0363, 0.0479, and 0.0262, respectively. Except for diameter Class II, the Sp values of Liriodendron tend to increase with the higher diameter class. Notably, the Sp value of the smaller diameter classes is significantly lower than that of the larger diameter classes; for example, the Sp values for diameter Classes IV and V are 1.61 times that of Class I and 2.18 times that of Class II, while the Sp value for Class III is 1.22 times that of Class I and 1.65 times that of Class II.

Spatial genetic structure analysis was conducted for each diameter class and the entire 2.5 hm² population (Figs. 2 and 3). As shown in Fig. 4, when the geographical distance between individuals of the entire population of L. chinense in the 2.5 hm² plot is less than 10 m, F_ij is 0.0813 (which is the average correlation coefficient for the first distance class). The coefficients of relatedness for full siblings, half siblings, and cousins are 0.232, 0.112, and 0.051, respectively. Thus, within the 10 m range, the relatedness among individuals in the L. chinense population is slightly more distant than that of half siblings. When the geographical distance between individuals is within 49 m, F_ij  > 0 and tends to gradually decrease. When the geographical distance exceeds 49 m, F_ij is less than zero. These results indicate that when the distance between individuals in the L. chinense population is less than 49 m, the genetic similarity among individuals is relatively high, whereas genetic similarity among individuals at other distance classes is lower.

As shown in Fig. 3A, when the geographic distance between individuals in the population of L. chinense at the I tree diameter class is less than 10 m, F_ij is 0.0838 (the average correlation coefficient for the first distance class). The coefficients of relatedness for full siblings, half-siblings, and cousins are 0.232, 0.112, and 0.051, respectively; thus, within 10 m, the relatedness among individuals in the I tree diameter class is slightly greater than that of half-siblings. When the geographic distance between individuals is within 49 m, F_ij > 0, showing a gradually decreasing trend. When the geographic distance between individuals is greater than 49 m, F_ij is less than zero for all cases. The results indicate that when the distance between individuals in the population of L. chinense at the I tree diameter class is less than 49 m, genetic similarity among individuals is relatively high, while genetic similarity at other distance classes is lower.

As shown in Fig. 3B, when the geographic distance between individuals in the population of L. chinense at the II tree diameter class is less than 10 m, F_ij is 0.0874 (the average correlation coefficient for the first distance class). Within 10 m, the coefficients of relatedness for full siblings, half-siblings, and cousins are 0.232, 0.112, and 0.051, respectively; thus, within 10 m, the relatedness among individuals in the II tree diameter class is slightly greater than that of half-siblings. When the geographic distance between individuals is within 29 m, F_ij > 0, demonstrating a gradually decreasing trend. When the geographic distance exceeds 29 m, F_ij is also less than zero. The results indicate that when the distance between individuals in the II tree diameter class is less than 29 m, genetic similarity among individuals is relatively high, while genetic similarity at other distance classes is lower.

As illustrated in Fig. 3C, when the geographic distance between individuals within the III tree diameter class population of L. chinense was less than 10 m, the F_ij was 0.1366, which represents the average correlation coefficient for the first distance class. Within 10 m, the coefficients of relatedness for full siblings, half-siblings, and cousins are 0.232, 0.112, and 0.051, respectively. Therefore, within 10 m, the relatedness among individuals in the III tree diameter class is slightly greater than that among full siblings. When the geographic distance between individuals is within 50 m, F_ij > 0, exhibiting an overall gradually decreasing trend. When the geographic distance exceeds 50 m, F_ij is less than zero for all cases. The results indicate that for individuals within the III tree diameter class, genetic similarity was relatively high at distances less than 50 m, whereas it was significantly lower in other distance classes. Due to the small number of individuals in the population of L. chinense at the V tree diameter class being less than 30 pairs, the V tree diameter class population was merged into the IV tree diameter class population for spatial genetic structure analysis.

As shown in Fig. 3D, when the geographic distance between individuals in the population of L. chinense at the IV & V tree diameter classes is less than 10 m, F_ij is 0.1130 (the average correlation coefficient for the first distance class). The coefficients of relatedness for full siblings, half-siblings, and cousins are 0.232, 0.112, and 0.051, respectively; thus, within 10 m, the relatedness among individuals in the IV & V diameter classes is slightly greater than that among full siblings. When the geographic distance between individuals is within 40 m, F_ij > 0, showing an overall gradually decreasing trend. When the geographic distance exceeds 40 m, F_ij is less than zero for all cases except in the range of 48 to 50 m. The results indicate that in the IV & V diameter class loci, genetic similarity among individuals is relatively high at distances of 40 m and within the range of 48 to 50 m, while genetic similarity at other distance classes is lower. A significant spatial genetic structure was detected for L. chinense in the I, II, III, IV, and V tree diameter classes, as well as the entire population, at distances of 49 m, 29 m, 50 m, 40 m, and 49 m, respectively. Furthermore, the shapes of the correlation coefficient curves were similar, indicating a general decreasing trend in the F_ij with increasing geographic distance.

Parentage analysis

According to parentage analysis (Table 6), both parents were identified for 91 of the 126 seedlings and Class I saplings, corresponding to an assignment rate of 72.2%. Uniparental assignment (exclusively maternal) was achieved for 15 individuals, whereas no parental matches were detected for the remaining 20 individuals. 68 individuals were identified as full- or half-siblings, representing 53.97% of all L. chinense individuals in diameter class I. For instance, half-siblings M170, M176, and M244 shared the same maternal parent (M174) but had different paternal parents. Conversely, half-siblings M40, M54, and M75 shared the same paternal parent (M208) but had different maternal parents. Based on geographical coordinates and parentage analysis results, seed and pollen dispersal distances for L. chinense were estimated. As indicated in Table 6, seed dispersal distance within diameter Class I ranged from 1.3 to 115.1 m, whereas pollen dispersal distance ranged from 3.7 to 112.9 m. The mean dispersal distance was 29.8 m for seeds and 35.6 m for pollen.

The results of the parentage analysis presented in Table 7 indicate that among the 124 young trees in the II diameter class, a total of 84 individuals found both parents, accounting for 67.742% of the total number of individuals. Additionally, 14 individuals were found to have only one parent, while 26 individuals did not have either parent identified. Among these, there are 89 full siblings or half-siblings, which represent 71.77% of all L. chinense in the II diameter class. For instance, the full siblings M91 and M139 both originated from parents M97 and M112, while half-siblings with the same female parent but different male parents, M82, M87, and M210, all came from the female parent M99. Half-siblings with the same male parent but different female parents, M34, M79, and M82, all came from the male parent M122.

Based on the geographical coordinates and the results of the parentage analysis, the dispersal distances of L. chinense seeds and pollen can be calculated. The table shows that in the II diameter class, the dispersal distance of L. chinense seeds ranges from 5.6 to 102.0 m, while the dispersal distance of pollen ranges from 2.2 to 91.5 m. The average dispersal distance for seeds is 36.1 m, while for pollen, it is 27.5 m.

The results of the parentage analysis presented in Table 8 indicate that among the 28 mature trees in the III diameter class, a total of 24 individuals found both parents, accounting for 85.714% of the total number of individuals. Additionally, two individuals were found to have only one parent, while two individuals did not have either parent identified. Among these, there are 21 full siblings or half-siblings, which represent 75% of all L. chinense in the III diameter class. For example, the full siblings M229 and M281 both originated from parents M279 and M282, while half-siblings with the same female parent but different male parents, M120 and M284, both came from the female parent M89. Half-siblings with the same male parent but different female parents, M112 and M114, both came from the male parent M268. Based on the geographical coordinates and the results of the parentage analysis, the dispersal distances of L. chinense seeds and pollen can be calculated. The table shows that in the III diameter class, the dispersal distance of L. chinense seeds ranges from 4.6 to 93.3 m, while the dispersal distance of pollen ranges from 4.9 to 104.6 m. The average dispersal distance for seeds is 28.9 m, while for pollen, it is 36.9 m.

The results of the parentage analysis presented in Table 9 indicate that among the 38 mature trees in the IV diameter class, only two individuals found both parents, accounting for 5.263% of the total number of individuals. Additionally, 10 individuals were found to have only one parent, while 26 individuals did not have either parent identified. Among these, there are 12 full siblings or half-siblings, representing 31.58% of all L. chinense in the IV diameter class. For instance, the full siblings M122 and M295 both originated from parents M272 and M310; half-siblings with the same female parent but different male parents, M89 and M98, come from female parent M272, but no half-siblings with the same male parent but different female parents were found. Based on the geographical coordinates and the results of the parentage analysis, the dispersal distances of L. chinense seeds and pollen can be calculated. The table shows that in the IV diameter class, the dispersal distance of L. chinense seeds ranges from 23.4 to 90.1 m, with an average dispersal distance of 79.4 m, while the average dispersal distance for pollen is 73.5 m.

The parentage analysis results indicate that a total of 201 L. chinense individuals were found in the 2.5 hm² plot, which accounts for 63.21% of the total number of individuals. Additionally, 41 individuals were found to have only one parent, comprising 12.89% of the total, while 74 individuals did not have either parent identified. Among the 74 L. chinense individuals without identified parents, 26 have a DBH greater than 30 cm. During sampling at the experimental site, some trees with a DBH > 40 cm were felled. Consequently, the parents of these 26 individuals may have been among the felled trees or may have been located outside the plot. There are 190 full siblings or half-siblings, representing 59.75% of all L. chinense in the 2.5 hm² plot. For instance, half-siblings with the same female parent but different male parents, M120 and M284, both came from female parent M89; whereas half-siblings with the same male parent but different female parent, M112 and M114, both came from male parent M268.

For the 91 L. chinense individuals in diameter Class I that found both parents and the 15 individuals that found only one parent, a diagram was created to observe the dispersal distances and directions of seed flow and pollen flow, as shown in Fig. 4A. The trajectory from the male parent to the female parent represents pollen flow, whereas the trajectory from the maternal parent to the offspring represents seed flow. In diameter Class I, the distribution of L. chinense offspring is concentrated in three locations. The western offspring distribution forms two central point’s radiating outward, while the eastern offspring distribution is more dispersed. Both parents are primarily concentrated in the western area, with the main direction of seed flow dispersing from the southwest towards the north and northeast, with occasional dispersal in other directions. The main direction of pollen flow disperses from the southwest to the northeast and from the northeast to the southwest.

For the 84 L. chinense individuals in diameter Class II that found both parents and the 14 individuals that found only one parent, a diagram was created to observe the dispersal distances and directions of seed flow and pollen flow, as illustrated in Fig. 4B. The trajectory from the male parent to the female parent represents pollen flow, while the trajectory from the female parent to the filial generation represents seed flow. In diameter Class II, the distribution of L. chinense offspring is concentrated in the western area, forming a central point from which the western offspring radiate outward. In contrast, the distributions of eastern and northern offspring are more dispersed. Parental trees were predominantly located to the north and east of their offspring. The predominant direction of seed dispersal was from the southeast to the northwest, although occasional dispersal occurred in other directions (e.g., northeast). Pollen flow was primarily oriented along two axes: from the southeast to the northwest and from the northeast to the southwest.

For the 24 L. chinense individuals in diameter Class III that found both parents and the two individuals that found only one parent, a diagram was created to observe the dispersal distances and directions of seed flow and pollen flow, as shown in Fig. 4C. The trajectory from the male parent to the female parent represents pollen flow, while the trajectory from the female parent to the filial generation represents seed flow. In diameter Class III, the distribution of L. chinense offspring is concentrated in the northwest and southeast areas, with the northwest offspring forming a central point from which they radiate outward, while the southeast offspring distribution is more dispersed. Both parents are primarily located in the northwest and southeast positions. Seed dispersal occurred primarily along two axes: from the northwest to the southeast and from the southwest to the northeast, with occasional dispersal in other directions. Pollen flow exhibited a similar pattern, with primary directions from the northwest to the southeast and from the southwest to the northeast; occasional dispersal was also observed in other directions, including the northwest.

For the two L. chinense individuals in diameter Class IV that found both parents and the 10 individuals that found only one parent, a diagram was created to observe the dispersal distances and directions of seed flow and pollen flow, as shown in Fig. 4D. The trajectory from the male parent to the female parent represents pollen flow, while the trajectory from the female parent to the filial generation represents seed flow. In diameter Class IV, the distribution of L. chinense offspring is scattered in the northeastern area of the plot, forming a central distribution point. The main direction of seed flow dispersal is from south to north. The primary direction of pollen flow dispersal is from the northeast to the southwest.

Based on geographical coordinates and the results of the parentage analysis, the dispersal distances of L. chinense seeds and pollen were estimated. In the 2.5 hm² Western Hunan L. chinense plot, the seed dispersal distance ranges from 1.3 to 115.1 m, while the pollen dispersal distance ranges from 2.2 to 112.9 m. As shown in Table 10, the average dispersal distance for seeds is 45.0 m, while for pollen, it is 47.0 m. This pattern is primarily influenced by small-scale spatial constraints. Intense resource competition within the forest stand limits dispersal, resulting in most seeds being deposited near the maternal parent tree.

Distance intervals of 10 m were used to categorize the distances and to statistically analyze the dispersal frequencies of L. chinense seeds and pollen within each distance class. The results indicate (Fig. 5) that both seed and pollen dispersal frequencies generally decrease with increasing distance from the individuals. The maximum dispersal distance for L. chinense seeds is 115.1 m (offspring M169, female parent M181), with an average distance of 45.0 m; 78.5% of the seeds are dispersed within 60 m. For L. chinense pollen, the maximum dispersal distance is 112.9 m (offspring M160, female parent M167, male parent M203), with an average dispersal distance of 47.0 m, where 68.2% of the pollen is dispersed within 50 m.

Genetic diversity of Liriodendron chinense

Genetic diversity is a key determinant of population adaptability and evolutionary potential. Habitat fragmentation often reduces within-population genetic diversity and increases genetic differentiation among populations (Young, Boyle & Brown, 1996; Booy et al., 2000). This study selected 11 primers that are highly polymorphic, specific, and stable for genetic diversity analysis. L. chinense populations in Western Hunan exhibited relatively high genetic diversity (H_E = 0.812, I = 1.940) compared to congeneric species like L. tulipifera (H_E = 0.67) (Huang et al., 2018). And the expected heterozygosity (H_E) of Liriodendron in Qingping population is like that of some endangered plants (Thomas et al., 2021), which indicated that the Qingping Liriodendron population was still in “endangered” habitat, and it was urgent to carry out Liriodendron resource protection. This discrepancy may be attributed to variations in the number of microsatellite markers employed, the geographical sampling range, and the number of individuals sampled (Gerber et al., 2000).

Population genetic structure

Generally, compared to herbaceous plants and shrubs, trees have longer pollen dispersal distances and greater longevity, resulting in insignificant genetic structure (Ghazoul, 2005). The presence of a significant SGS in trees is typically associated with specific biological conditions, such as elevated rates of self-fertilization or inbreeding, which can be influenced by factors including greater seed mass and higher population density (Nyquist & Baker, 1991). The results of the spatial genetic structure study of L. chinense in the 2.5 hm² Western Hunan area indicate that the spatial genetic structure for all individuals, as well as for diameter Classes I, II, III, and IV & V, are 0.0262, 0.0298, 0.0220, 0.0363, and 0.0479, respectively. These values are consistent with the SGS reported for other species with similar life-history traits and suggest a relatively strong SGS. The study found that the spatial genetic structure strength of hybrid plants (average Sp = 0.0372) (Vekemans & Hardy, 2004), trees (average Sp = 0.0102), insect-pollinated plants (average Sp = 0.0171), wind-dispersed seed plants (average Sp = 0.0120), and gravity-dispersed seed plants (Sp range = 0.0041–0.2632) is significantly higher in comparison. Our results also align with previous findings for insect-pollinated tree species, such as Vouacapoua americana, Sextonia rubra and Eperua grandiflora (Rousset, 1997). Compared to other endangered tree species, L. chinense exhibits intermediate Sp values (Zhang & Ma, 2008).

The individual distribution of L. chinense loci predominantly displays a clustered pattern. Consequently, as the distance class increases, the frequency of individuals in the two loci generally decreases. There is a significant genetic structure present within the L. chinense population (within a range of 49 m).

Spatial autocorrelation analysis reveals that the linear regression slope b_F is negative, indicating that with increasing distance, the pairwise relatedness F_ij decreases. A significant spatial genetic structure (F_ij > 0) is detected overall within the 49 m range. Beyond this distance, F_ij sharply declines to show significant negative or non-significant correlation, which aligns with the isolation by distance (IBD) pattern (Wiens & Colella, 2024).

Different diameter class loci of L. chinense exhibit significant genetic structure. The SGS intensity for diameter Class I is Sp =0.0298 (with F_ij = 0.0838 at the first distance class of 10 m), for diameter Class II it is Sp =0.0220 (F_ij is 0.0874 within the first distance level, i.e., 10 m) SGS strength Sp = 0.0363 for diameter Class III (F_ij 0.1366 within 10 m at the first distance class), SGS strength Sp = 0.0479 for diameter Class IV & V (F_ij 0.1130 within 10 m at the first distance class), The first distance level of all diameter classes was greater than that of Cousins (F_ij = 0.051), and even the F_ij of diameter Classes III and IV & V were greater than that of half sib individuals (F_ij =0.112), indicating that pollen transmission could be effectively carried out within the population, but long-distance gene ex-change was hindered. Gene exchange is mainly between close individuals with higher relative distance.

The F_ij of all diameter classes decreased linearly, indicating that the coefficient of affinity between individuals in a population decreased with the increase of distance. Relative to pollen, seeds exhibited a shorter dispersal distance. Seed flow had a greater influence at closer distances, whereas pollen flow had a stronger effect at longer distances (Smouse & Sork, 2004). However, in the second diameter class, the effect of pollen flow was greater in the short distance, while the effect of seed flow was greater in the long distance (Cain, Milligan & Strand, 2000). The original habitat likely consisted primarily of deciduous broad-leaved tree species with low population density. Lower community and population densities may effectively reduce the intensity of SGS within a population, thereby facilitating greater dispersal of both seeds and pollen.

Parentage analysis

The gene flow constituted by seed dispersal and pollen dispersal is the most funda-mental and primary ecological process within a population. Many plants can disperse their pollen over greater distances, with some reaching several hundred meters and others even extending to tens of kilometers (Luna et al., 2001; Schulke & Waser, 2001). This may be due to the wider visibility in the tree layer for pollinators, making distant flowers easier to detect, which facilitates further pollen dispersal.

In our results, the seed and pollen dispersal distances of 2.5 hm² in Qingping rocky desertification area of Western Hunan Province were relatively short, with 68.2% pollen dispersal distances mainly within 50 m and 78.5% seed dispersal distances mainly within 60 m. It is much smaller than the plant pollen distribution range (hundreds to thousands of meters) summarized by Wang et al. (2010). This may be due to the high stand density of the mixed forest, and the fact that it grows in the rocky desertification area with dense stone forest, rich species and strong habitat heterogeneity, resulting in the pollen being confined to a small area.