Quantitative Characteristics and Environmental Interpretation of Vegetation Restoration in Burned Areas of the Dry Valleys of Southwest China

1. Introduction

As an important part of the global terrestrial ecosystem, forests have important ecological functions such as preventing soil erosion, preventing wind and sand, and regulating climate (Lourens et al. 2017). There are many factors affecting forest ecosystems, among which fire is the most direct (Di et al. 2007). Fire can promote the secondary succession of vegetation communities while destroying the forest (Di et al. 2007), and the renewal and restoration of vegetation after forest fire disturbance is an important problem faced by environmental researchers and human society (Calvo et al. 2005).

After vegetation communities are disturbed by forest fire, the greatest contribution to community changes is made by the growth of herbs and shrubs (Lasaponara et al., 2009). Shrub-grass vegetation can efficiently absorb soil nutrients and grow rapidly to cover the surface (Lasaponara et al. 2009). Shrub-grass vegetation communities develop rapidly in burned areas, and then slowly succeed to woody plant communities. This causes the recovery rate of the normalized difference vegetation index (NDVI) in fire areas to be relatively slow (Lasaponara et al. 2009, Schoennagel et al. 2004, Ted 1971, Dalel et al. 1979). The effects of fire on different vegetation communities vary, and there are significant differences in their degree of recovery. The similarity of the vegetation community shrub layer is higher than that of the herb layer in burned areas (Xofis et al. 2021). When the forest fire disturbance years are closely separated, the similarity of the vegetation community shrub layer in the burned areas is much higher than that when the forest fire disturbance years are far apart (Xofis et al. 2021, Wang et al. 2004). The forest aboveground biomass is significantly affected by forest fire, which not only has profound impact on the biomass of different tree species communities but also has different effects on communities in different succession periods (Virginia et al. 2018). In previous research, after forest fire disturbance, the vegetation community in the burned areas was single in the early recovery stage relatively, and the vegetation community types differed little among plots (Jin and Wu 2021, Cai et al. 2012). The intensity level of forest fire disturbance is related to the characteristics of the understory plant community composition (Jens et al. 2019). The fine-scale heterogeneity of the forest structure after fire is an important driving factor of vegetation species diversity (Jens et al. 2019). Forest fires promote the development of different dominant vegetation through their positive effect on soil seed bank dynamics, which has great significance in maintaining vegetation species richness (Hideo et al. 2011).

Since the 1980s, the quantitative classification and ordination of vegetation has been a hot topic in ecological research, and has become an indispensable means of vegetation community ecology research (Zhang et al. 2000, Yu et al. 2005). In China, the research on vegetation quantitative classification and ordination can be summarized as the process from introduction, learning to trial, application and development (Zhang et al. 2000, Yu et al. 2005). Quantitative classification refers to the transformation of classification concepts from qualitative description to quantitative analysis using mathematical methods, which has great significance for promoting the development of biological taxonomy (Michal et al. 2016). Compared with the quantitative classification method (determining the disjunctive distribution of the community), the quantitative ordination method can reveal the continuous relationship of the community distribution (Wang et al. 2006). Two-way indicator species analysis (TWINSPAN), detrended correspondence analysis (DCA), and detrended canonical correspondence analysis (DCCA) are the main methods used in vegetation quantitative classification and ordination. These methods can objectively and quantitatively reveal the distribution pattern of vegetation communities and the relationship between vegetation communities and environmental factors (Chian et al. 2016, Rahman et al. 2017, Wang et al. 2006, Jiang et al. 2015).

Due to its unique dry valley climate conditions and human factors, the Anning River Basin has become an areas with concentrated fire occurrence (Bai et al. 2020). To date, research on the burned areas in the dry valleys of the Anning River Basin has focused on revealing the law of vegetation community restoration (Liu et al., 2022, Bai et al., 2020) and vegetation evaluation (Jiang et al. 2022), while the quantitative classification and ordination of vegetation communities in burned areas remain unclear. The purpose of this study was to explore the classification and distribution pattern of vegetation communities and their influencing factors in the early stage of natural recovery after forest fire disturbance in dry valleys. The results of this study have important guiding significance for the prevention of soil erosion and ecological civilization construction in dry valleys.

2. Materials and Methods

2.1. Study Area

The Anning River (27°05′–29°02^′N, 101°38′–102°43′E) is a tributary of the Yalong River, which is a tributary of the Jinsha River (Figure 1). The main stream of Anning River flows through Mianning, Xichang, and Dechang in Liangshan Prefecture and Miyi County in Pan Zhihua City, with a total length of 320 km and a drainage area of 11150 km² (Xu 2004). The north–south climate change and vertical difference in the basin are large, the precipitation is concentrated from May to October, and the annual precipitation is more than 1000 mm (Xu 2004). The forests in this area are dominated by Pinus yunnanensis Franch. (Py)–Pinus massoniana Lamb. (Pm) mixed forests. Under the forest, the vegetation community mainly consists of shrubs and herbs such as Quercus guyavaefolia H. Leveille (Qg), Machilus pingii Cheng ex Yang (Mp), Duhaldea cappa (Buchanan-Hamilton ex D. Don) Pruski & Anderberg (Dc), Cymbopogon goeringii (Steud.) A. Camus Cymbopogon (Cg), and Paspalum paspaloides (Michx.) Scribn. (Pp). The results of on-site investigation showed that after the forest fire disturbed the study sample area (Table 1), all the aboveground parts of shrubs and grasses died and only trees trunk remained with burned branches, while the soil structure and underground parts of vegetation suffered less damage.

2.2. Methods

2.2.1. Quadrat Setting

In this study, the investigation quadrats were set up in the burned areas during the early natural recovery period after forest fire disturbance in the Anning River Basin (Figure 1 and Table 1) from 2020–2021. Ten plots (10 m × 10 m) were set for each burned area, then quadrats (5 m × 5 m) were placed in the center and four corners of each plot. A total of six burned areas and 360 quadrats, 60 shrub plots, and 300 grass quadrats were set. The type, quantity, plant height, and coverage of understory vegetation in the quadrats were recorded, and the four environmental factors of elevation, slope, slope aspect, and slope position were measured in each quadrat (Qiu et al. 2000).

2.2.2. Importance Value Calculation

The importance value of a vegetation species refers to a comprehensive index of the relative importance of certain vegetation in the community to which it belongs. The importance value calculation method selected in this study is as follows (Sun and Hu 2010):

Trees and Shrub	IV = (RD + RH + RC) / 3,	(1)
Herbs	IV = (RH + RC) / 2.	(2)

In formulas (1) and (2), IV is the importance value; RD is the relative density; RH is the relative height; and RC is the relative coverage.

2.2.3. TWINSPAN Quantitative Classification

TWINSPAN quantitative classification is a typical hierarchical classification method that has been widely used in vegetation community studies (Cho et al. 2015). In the present study, WinTWINS 2.3 software was used to perform TWINSPAN quantitative classification based on the importance value matrix and environmental factor matrix of vegetation species in typical quadrats (Zhao et al., 2019, Gholamhosein et al. 2018).

2.2.4. DCA and DCCA

DCA and DCCA are multivariate analysis techniques used to study the relationship between vegetation and the environment (Zhao et al. 2019, Jan et al. 2009). In the present study, DCA and DCCA of the quadrats and environment were performed using CANOCO 4.5 software, and then the sorting diagram was drawn using CanoDraw for Windows 4.5 (Zhao et al. 2019).

For the slope position, 1, 2, 3, 4, and 5 represented the bottom of the slope, downhill slope, middle slope, uphill slope, and top of the slope respectively. The larger the number, the higher the slope position. The original value of the slope direction could not directly represent the degree of sunshine exposure. Therefore, each slope direction was assigned a numerical value, where 1 refers to the north slope (317.15°–22.15°), 2 refers to the northeast slope (22.15°–67.15°) and northwest slope (292.15°–317.15°), 3 refers to the east slope (67.15°–112.15°) and west slope (247.15°–292.15°), 4 refers to the southeast slope (112.15°–157.15°) and southwest slope (202.15°–247.15°), and 5 refers to the south slope (157.15°–202.15°). The higher number of slope direction, the hotter it is (Qiu and Zhang 2000).

3. Results

3.1. Classification of the Vegetation Community

Based on the vegetation community classification methods and principles in China, the community types were named according to the survey results of dominant species. Forty typical vegetation quadrats and 86 vegetation species were quantitatively classified using TWINSPAN. The fourth classification level was used as the classification result, which could be divided into the following seven community types (Figure 2 and Table 2).

Community I: The Vaccinium fragile Franch. (Vf)–Imperata cylindrica (Linn.) Beauv. (Ic) community included quadrats 20, 29, 30–32, 34–35, and 51–53. The community was mainly distributed on gentle, semi-sunny slopes. Except for community indicator species, the shrub layer also included Qg, Rhododendron simsii Planch. (Rs), Mp, and Cotoneaster hissaricus Pojark. (Ch). The herb layer also included Pp, Cg, Uraria lagopodoides (L.) Desv. ex DC. (Ul), Leontopodium leontopodioides (Willd.) Beauv. (Ll), Ageratina adenophora (Sprengel) R. M. King & H. Robinson (Aad), and Tripogon chinensis (Franch.) Hack. (Tc).

Community II: The Dc–Leptochloa chinensis (L.) Nees (Lc) community included quadrats 12–13, 16, 18, 37, 54, and 55. The community was mainly distributed on steep semi-shady slopes or shady slopes. Major species included Qg, Hypericum patulum Thunb. ex Murray (Hp), Mp, Elsholtzia rugulosa Hemsl. (Er), Pp, Cg, Ll, Artemisia argyi Levl. et Van (Aar), and Hedyotis herbacea L. (Hh).

Community III: The Mp–Monogramma trichoidea J. Sm. (Mt) community included quadrats 9, 14, 15, 17, 19, 21, 24–26, 36, 38, and 56–59. The community was distributed on steep, sunny slopes with large elevation drops. The main dominant species in the shrub layer was Mp, but the Qg and Vf were relatively few. In addition to the main dominant species of Mt, the herbaceous layer was mainly composed of Er, Pp, Cg, Polygonatum verticillatum (L.) All. (Pv), Aa, and Cyperus cyperoides (L.) Kuntze (Cc).

Community IV: The Qg–Potentilla leuconota var. brachyphyllaria (Pl) community included quadrats 4, 22–23, 27–28, and 60. The community was distributed on steep, semi-shady slopes or semi-sunny slopes. In addition to community indicator species or dominant species, the shrub layer also mainly included Vf, Hp, and Mp, while the herb layer also included Er, Mt, Commelina diffusa N. L. Burm. (Cd), Desmodium microphyllum (Thunb.) DC. (Dm), Ll, and Aa.

Community V: The Qg–Pp community included quadrats 6–8, 10–11, 33, 41, 42, and 45–47. The community was mainly distributed on steep, shady slopes at high elevations. The major species included Mp, Leptodermis potanini Batalin (Lp), Campylotropis hirtella (Franch.) Schindl. (Ch), Cc, Er, Saussurea hieracioides Hook. f. (Sh), Lc, Ll, and Cg.

Community VI: The Qg–Monochasma savatieri Franch. ex Maxim. (Ms) community included quadrats 1, 2, 5, 43, 44, and 48–50. The community was mainly distributed on gentle, semi-shady slopes or semi-sunny slopes at high elevations. The main species in the shrub layer included Vf, Ch, Mp, and Lespedeza davidii (Ld). The herb layer included the major species Scutellaria baicalensis Georgi (Sb), Polygonum paleaceum Wall. ex HK. f. (Ppa), Pv, Er, Mt, and Sh.

Community VII: The Qg–Ll community included only quadrat 3. The vegetation coverage of this community reached 90%, and it was distributed on a gentle sunny slope at a high elevation. The major species included Vf, Ld, Mp, Lp, Cc, Ms, Rabdosia adenantha (Diels) Hara (Ra), Erigeron speciosus (Lindl.) DC. (Es), Anaphalis sinica Hance (As), and Iris tectorum Maxim. (It).

3.2. DCA Ordination of Vegetation Community Quadrats and Environmental Factors

In this study, the results of TWINSPAN classification and DCA ordination revealed the distribution patterns of various vegetation communities in the DCA ordination map (Figure 3). According to the correlation coefficient (r) of the environmental factor matrix of DCA ordination, the first axis of DCA reflected the change of altitude gradient (r=-0.49). The altitude decreased gradually along the direction of the first axis. The direction of the first axis was mainly associated with the temperature condition of the quadrats. The second axis of DCA reflected gradient changes of the slope (r=-0.23) and slope direction (r=0.13). The slope gradually decreased and the degree of sunshine gradually increased along the direction of the second axis. The second axis direction was mainly associated with changes in the hydrological conditions of the quadrats. According to the clustering conditions of 40 typical vegetation quadrats on the DCA ordination diagram, it could be divided into the following five ecotopes:

Ecotope A: This ecotope was dominated by community I, with a large elevation drop (1825–2115 m), a gentle slope (10–25°), and high vegetation coverage (40–85%). Py was the main tree species in this ecotope, and there were many kinds of shrub and grass vegetation. This ecotope was a savanna community.

Ecotope B: This ecotope mainly included community II and community III, with an elevation of 2090–2140 m, a slope of 30–50°, and vegetation coverage of 20–60%. In this ecotope, the trees included Py and Pm, and the shrubs mainly included Dc and Mp. This ecotope was a savanna community.

Ecotope C: This ecotope included communities II, III, IV, and V. It was mainly distributed between an elevation of 2100–2180 m, with a steep slope (30–40°) and 35–70% vegetation coverage. There were various kinds of shrubs and grasses, mainly including Qg and Pp. This ecotope was a savanna community.

Ecotope D: Community III was dominant in this ecotope, and communities I, V, VI, and VII were also included. The ecotope was mainly located between an elevation of 2090–2245 m, with a steep slope of 25–45° and high vegetation coverage that reached 90%. In addition to indicator species or dominant species, this ecotype contained other vegetation species, such as Er, Pp, and Cg. This ecotope was a shrub-grass community.

Ecotope E: This ecotope included community V and community VI. The quadrats were located at a high elevation (2240–2390 m), with a slope of 10–45° and vegetation coverage of 50–80%. The main vegetation of this area was Qg. This ecotope was a shrub-grass community.

3.3. DCCA Ordination of Vegetation Community and Environmental Factors

In this study, DCCA ordination was used to analyze the relationship between the vegetation community and environmental factors in the burned areas of the Anning River Basin (Figure 4). The arrow line in the figure represents each environmental factor, while the quadrant of the arrow represents the positive and negative relationships between the environmental factor and the DCCA ordination axis. The slope between the arrow line and the ordination axis represents the correlation between the environmental factor and the ordination axis. The vertical distance from the quadrat to the arrow line represents the impact of the environmental factor on the quadrat. The first ordination axis mainly reflects the change of elevation (r= −0.65) and slope position (r= −0.38). The elevation and slope position gradually decrease along the ordination axis. The second ordination axis mainly reflects the changes of slope (r= −0.46) and aspect (r= 0.48). The slope decreases with the ordination axis direction, and the aspect transitions from shady slope to sunny slope.

The comparison between the DCA ordination diagram (Figure 3) and the DCCA ordination diagram (Figure 4) showed that the quadrat distribution of the DCCA ordination diagram was more concentrated than that of the DCA ordination diagram, but the demarcations between quadrats were not clear. Therefore, DCCA ordination was not as intuitive as DCA ordination when classifying vegetation communities. The eigenvalues of the DCA ordination axis were significantly greater than those of the DCCA ordination axis (Table 3). The correlation coefficients of species–environmental factors of the DCCA ordination axis were significantly larger than those of the DCA ordination axis (Table 3). The conclusion indicated that DCCA ordination focused on the relationship between vegetation species and environmental factors.

4. Discussion

4.1. Vegetation Restoration Features of Burned Areas in the Dry Valleys

Based on the climax pattern hypothesis, seven vegetation communities in the study area have not reached the climax level, and the stability of the community structure and spatial distribution pattern is poor. The vegetation community competes fiercely for nutrient resources in the early stage of vegetation regeneration after forest fire disturbance. The unique arid valley climate and topographical features of the Anning River Basin led to the serious deterioration of environmental conditions (such as a dry climate and poor soil) in the early stage after forest fire disturbance. The secondary succession process of vegetation communities was slow, and only vegetation species with strong tolerance (such as Qg, Mt, and Pp) could grow and cover the surface rapidly. Therefore, many community types with Qg as an indicator species were formed in the Anning River Basin. The community distribution range is large, but the species diversity is relatively single and the stability is low (Chian et al. 2016, Michel and Claire 2013), resulting in a fragile ecosystem structure that is vulnerable to natural disasters such as debris flows. When conducting the artificial restoration of vegetation in burned areas, it is necessary to consider soil and water conservation and vegetation diversity to accelerate the process of vegetation restoration (Jin and Wu 2021). Species that are more tolerant and are suitable to grow in the area, such as Dc and Vf, should be sown to break the dominance of Qg. In addition, it is necessary to consider the influence of environmental factors such as altitude and slope to design a reasonable vegetation configuration so as to improve the species diversity and stability of the ecosystem, and achieve the best restoration effect.

4.2. TWINSPAN Quantity Classification and DCA Ordination Features

TWINSPAN quantitative classification is based on indicator species. Generally, five main dominant species are selected as important indicator species to objectively classify vegetation communities using TWINSPAN (Cho et al. 2015, Jan et al. 2009). DCA ordination was proposed by Hill and Gauch after further research on the basis of correspondence analysis. This method effectively eliminates the bow effect of correspondence analysis, and is widely used in vegetation community analysis as a result (Zhang 1994, Li 2002). For different research sites, scales, and objects, the ecological interpretation of the DCA ordination axis will yield different research results (Zhang 1994). In this study, TWINSPAN quantitative classification was combined with the DCA ordination method, and the DCA ordination diagram was reclassified on the basis of TWINSPAN quantitative classification. The discontinuous and continuous relationships between vegetation classification and distribution were checked out each other(Zhao et al. 2019). These method clearly and accurately revealed the quantitative classification and distribution pattern of vegetation communities in the study area, which has great guiding significance for the efficient implementation of artificial restoration projects in the burned areas. Due to the low amount of quadrat data, only the classification results of fourth-level TWINSPAN quantitative classification were used in this study, so the classification of vegetation community types was not sufficiently detailed. In subsequent studies, the amount of quadrat data should be increased to achieve a more detailed classification of the vegetation community in the study area.

4.3. DCA Ordination and DCCA Ordination Features

The applicable objects of DCA ordination and DCCA ordination are different (Zhang 1994, Jennifer et al. 2006). Due to the clear boundaries between communities, DCA ordination is suitable for classifying vegetation communities and exploring the relationships between communities. The DCCA ordination is based on the DCA ordination, and then performs a multiple linear regression (environmental constraint) between the quadrat environment matrix value and the quadrat ordination value, which not only reflects the similarity of species composition among quadrats but also reflects the similarity of the environmental factors influencing each quadrat (Bai et al. 2020). DCCA ordination can display the three parameters of quadrat, species, and environmental factors on the coordinate plane of the sorting axis simultaneously. The interaction of the three parameters leads to the blurring of the boundaries between communities in the ordination diagram (Zhang 2004). Therefore, DCCA ordination is mainly suitable for revealing the relationships between vegetation species and environmental factors (Zhang 2004). Only a small number of environmental factors were investigated in this study, the results of the vegetation community boundary research had a certain amount of chance, and the environmental factors had a low interpretation rate for community distribution variation. Additional environmental factors, such as soil physicochemical properties and climate factors, should be further explored and comprehensively analyzed in subsequent research.

5. Conclusions

In this study, the spatial sequence method instead of the time successional sequence method was used to classify and analyze vegetation in the early natural recovery period after forest fire disturbance in the dry valleys of the Anning River Basin. The following conclusions were drawn.

(1)

After forest fire disturbance in the Anning River Basin, the TWINSPAN classification method was used to classify the vegetation community in the early stage of vegetation regeneration and succession into seven community types. The community types were dominated by savanna communities. The Qg community was widely distributed in the region.

(2)

In the DCA ordination diagram, the burned areas were divided into five ecotopes based on the distribution of quadrats. Elevation was significantly correlated with the classification and distribution pattern of community types in this area. With the decrease of elevation, the dominance of the main dominant species of Qg decreased continuously, and the community type changed from a shrub-grass community to a savanna community.

(3)

In the DCCA ordination diagram, the first ordination axis mainly reflected the changes in elevation and slope position. The elevation and slope position gradually decreased along the first ordination axis. The second ordination axis mainly reflected the changes of slope and aspect. The slope direction decreased along the second ordination axis direction, and the aspect transitioned from shady slopes to sunny slopes.