Stress Memory in Cynodon dactylon (L.) Pers During Succession in Drawdown Zones: Implications for Vegetation Restoration and Sustainable Management

Abstract

Reservoir drawdown zones are repeatedly affected by water-level fluctuations and anthropogenic regulation, making vegetation recovery an important issue for ecological restoration and sustainable reservoir management. This study focused on Cynodon dactylon, a dominant herbaceous species in the drawdown zones of five reservoirs in the Jinsha River Basin, southwestern China. Drawing on the existing concept of stress memory, which emphasizes the retained effects of previous environmental stress exposure on subsequent plant responses, we established an integrated assessment framework based on species dominance, functional traits, landscape pattern indices, and the soil seed bank. This framework was used to evaluate variation in the stress memory of C. dactylon across different successional stages and inundation gradients. The results showed that the overall stress memory of C. dactylon increased with successional progression in both the upper and lower zones, indicating continuous adaptive accumulation under long-term hydrological disturbance. The memory reflected by individual component indicators also generally increased, although their accumulation patterns varied among indicators. These findings suggest that dominance, functional traits, landscape pattern, and the soil seed bank can jointly characterize the adaptive responses of C. dactylon during vegetation recovery. Overall, the stress memory framework provides a systematic approach for identifying stage-specific vegetation changes, evaluating restoration potential, and informing ecological restoration and sustainable management in reservoir drawdown zones.

1. Introduction

Reservoirs play important roles in flood control, water supply, and hydropower generation. However, their construction and operation can also disrupt river hydrological connectivity, alter natural hydrological regimes, and exert persistent and complex impacts on terrestrial ecosystems adjacent to reservoirs [1]. Reservoir drawdown zones are special areas formed by the periodic inundation and exposure of land caused by reservoir impoundment and flood discharge. These zones exhibit distinct characteristics in energy exchange, material cycling, and ecological pattern dynamics, and are of great significance to both reservoir engineering safety and aquatic ecological environments [2]. Long-term or high-intensity flooding disturbance may not only deteriorate soil physicochemical properties, intensify habitat fragmentation, and cause vegetation degradation, but also further weaken the stability and recovery capacity of drawdown zone ecosystems [3,4,5]. Therefore, understanding how vegetation in drawdown zones adapts to and recovers under persistent hydrological stress, and using this knowledge to support vegetation restoration and regional ecological governance, has become an important issue in reservoir ecological protection and sustainable management.

In recent years, increasing attention has been paid to the role of plant memory in ecology, particularly as the “memory function and evolution of plants under environmental stress” has been proposed as an important scientific issue [6]. Ecological memory is generally considered an important component of ecosystem resilience, and a richer ecological memory may help ecosystems maintain key structures and functions after disturbance and recover to a relatively stable state within a shorter period [7]. On this basis, the concept of ecological stress memory further emphasizes, at the individual plant level, that the retention and accumulation of previous stress experiences can influence subsequent responses to similar environmental changes [8,9]. Stress memory is conceptually related to stress tolerance but is not identical to it. While tolerance refers to the ability of plants to endure adverse conditions, stress memory highlights the retained effects of previous stress exposure on subsequent responses. These effects may be mediated by physiological acclimation, trait plasticity, gene-expression regulation, or epigenetic mechanisms, and they may be expressed ecologically through changes in population performance, functional traits, reproductive capacity, and spatial organization [10,11]. For reservoir drawdown zones, which are subject to the long-term combined effects of water-level fluctuations, alternating wetting and drying, and habitat instability, stress memory may serve not only as an important manifestation of how dominant plants adapt to persistent stress, but also as one of the intrinsic mechanisms underlying their continued survival, expansion, and successional advancement. From this perspective, understanding vegetation recovery in drawdown zones through the lens of stress memory may not only deepen our knowledge of plant adaptive strategies, but also provide a new analytical framework for restoration assessment and management practice.

However, in reservoir drawdown zones, existing studies have mainly focused on vegetation composition, community structure, functional traits, and recovery patterns, whereas systematic quantification remains limited regarding how dominant plants retain and accumulate adaptive responses through stress memory during succession [12,13]. In particular, it remains unclear how stress memory can be linked with observable ecological indicators, such as species dominance, functional traits, landscape pattern, and the soil seed bank, and how these indicators can be used to evaluate vegetation recovery and restoration potential. Cynodon dactylon is a widely distributed and highly adaptable perennial herbaceous species in reservoir drawdown zones of southwestern China. In the investigated drawdown zones, it was consistently present and maintained high dominance across successional stages, reflecting its strong survival, regrowth, and expansion capacity under repeated inundation and exposure. Therefore, C. dactylon provides a suitable focal species for examining vegetation recovery and adaptive accumulation during drawdown-zone succession. Accordingly, this study focused on C. dactylon in the drawdown zones of five reservoirs in the Jinsha River Basin, southwestern China. The specific objectives were to: (1) characterize the variation in dominance, functional traits, landscape pattern indices, and soil seed bank density of C. dactylon across successional stages and inundation gradients; (2) develop an integrated multi-indicator framework for quantifying stress memory based on measurable ecological attributes; and (3) evaluate the variation in overall and component-level stress memory during succession and discuss its implications for vegetation restoration and sustainable management in reservoir drawdown zones.

2. Materials and Methods

2.1. Study Area Description

The Jinsha River Basin (90°23′ E–104°37′ E, 24°28′ N–35°46′ N), located in the upper reaches of the Yangtze River, originates from the Tanggula Mountains in Qinghai Province and extends across the Yunnan–Guizhou Plateau and the western margin of the Sichuan Basin. The Basin covers an area of 475,000 km², accounting for approximately 26.3% of the total area of the Yangtze River Basin. Its main stem is 3464 km long and flows through Qinghai, Tibet, Yunnan, Sichuan, and Guizhou Provinces [14]. The basin is characterized by higher elevation in the northwest and lower elevation in the southeast. The dominant soil type is torrid red soil, with mountain red soil, yellow-brown, and brown soil also occurring in the Basin [15]. Most of the middle and lower reaches of the Jinsha River belong to a subtropical monsoon climate zone. In this region, the river gradient gradually decreases and the river channel becomes broader. The mean annual temperature remained above 10 °C, and the mean annual precipitation ranged between 600 and 800 mm [16]. To minimize the effects of topographic variation and human disturbance, while ensuring the representativeness of reservoir drawdown zones, we selected near-natural drawdown zones with relatively consistent topographic conditions and land-use histories as fixed observation sites. These sites were located in the Liyuan Reservoir (27°20′ N, 100°13′ E), Guanyinyan Reservoir (26°37′ N, 101°18′ E), Ludila Reservoir (26°13′ N, 100°29′ E), Longkaikou Reservoir (26°32′ N, 100°17′ E), and Ertan Reservoir (26°49′ N, 101°49′ E), as shown in Figure 1. These sites are all located in the hot-dry valley region of the middle and lower reaches of the Jinsha River, characterized by high temperatures and strong evaporation, and thus share relatively similar climatic backgrounds and habitat conditions. The terrain of the drawdown zones is generally gentle, with slopes mostly less than 35°, while steeper areas (>35°) have sparse vegetation. The dominant soil in the study area is torrid red soil, with minor occurrences of neutral brown soil, and no significant differences in substrate conditions were observed between the upper and lower zones. On this basis, and with reference to previous studies [17,18], vegetation cover was used as the criterion for stage classification, and the five sites were accordingly assigned, in ascending order of restoration degree, to the first through fifth stages of the successional sequence in reservoir drawdown zones. The sites are all located along reservoir banks far from towns and villages, where direct human disturbance is relatively limited. Vegetation is dominated by herbaceous communities, with only a small number of trees and shrubs occurring sporadically in the upper part of the drawdown zone. Owing to differences in reservoir operation patterns, the five reservoirs exhibit distinct gradients of interannual water-level fluctuation, providing a suitable natural setting for examining the stress memory of Cynodon dactylon under hydrological regulation. Therefore, these sites may serve as representative cases of reservoir drawdown zones in the dry-hot valley region of southwestern China and provide useful references for vegetation restoration in similar riparian habitats.

2.2. Field Survey and Selection of Stress Memory Indicators

According to the hydrological rhythm of the study area, the reservoir riparian zone is usually exposed from May to September. In July, after 1–2 months of recovery, vegetation typically reaches a vigorous growth stage. Therefore, field investigations were conducted in July 2024 at the five study sites. In this study, a DJI Phantom 4 Multispectral unmanned aerial vehicle (UAV), equipped with a five-band multispectral camera array (red, green, blue, red-edge, and near-infrared) and a downward irradiance correction sensor, was used to conduct low-altitude photogrammetry over each site. Under adequate sunlight conditions, flight routes were planned according to the topographic characteristics of the study area, with a forward overlap of 85%, a side overlap of 75%, a flight speed of 2 m/s, and a flight altitude of 20 m. During data processing, ten ground control points (GCPs) that had been manually deployed in the field were used to mathematically model and correct the geometric distortion of the UAV remote sensing images. The coordinates of the GCPs were measured using a Qianxun Xingju SR1 intelligent network RTK receiver. At the same time, species reconnaissance and quadrat surveys were conducted for the plant communities in the drawdown zones. Referring to previous studies that classified inundation gradients according to inundation duration [19], the drawdown zones of the five reservoirs in this study were divided into two inundation gradients based on flooding duration: the lower-zone (severely inundated, inundation duration ≥ 5 months) and the upper-zone (lightly inundated, inundation duration < 5 months). Along each inundation gradient, five quadrats were randomly established for sampling, and each quadrat measured 1 m × 1 m. For each quadrat, plant species name, individual number, height, and species coverage were recorded.

The selection and quantification of stress memory indicators are crucial for the practical application of the concept of stress memory [20]. Therefore, appropriate ecological memory indicators need to be screened for application to the stress memory of C. dactylon [21]. An ideal indicator system should include key variables that influence ecosystem structure and the functioning of C. dactylon, while also remaining feasible for routine measurement. In addition, when exploring the role of stress memory in the succession and recovery of drawdown zones, the selection of indicators should consider not only the intrinsic implications of stress memory, but also the influence of each indicator on successional and recovery processes. Accordingly, indicators were selected from four aspects—species dominance [22], functional traits [23], landscape pattern indices [24], and the soil seed bank [25]—to characterize the stress memory of C. dactylon.

2.3. Measurement of Functional Traits and the Soil Seed Bank

During the field survey, three complete individuals of the target dominant species were selected from each small quadrat at every study site. These individuals were required to have favorable light conditions, mature growth status, and similar plant size, and to adequately represent the population characteristics of the dominant species within the quadrat. Field measurements were conducted using a measuring tape and a digital caliper with an accuracy of 0.01 mm. Plant height, main stem length and width, and main root length and width were recorded. Subsequently, the roots, stems, and leaves of the target plants were weighed using an electronic balance with an accuracy of 0.001 g. Leaf area was measured using a YMJ-A handheld leaf area meter, and chlorophyll content was determined using a TYS-A handheld chlorophyll meter. Mean values were calculated for each functional trait across sampled individuals to ensure data reliability and facilitate subsequent analyses of the potential relationships between inundation gradients and the functional traits of herbaceous plants.

Meanwhile, five sampling points were randomly selected within each inundation gradient at each study site, and 1 m × 1 m quadrats were established. The total number of quadrats across all sampling points was 2 × 5 × 5 = 50. Following the approach of using a large number of small quadrats [26], five soil subsamples were collected from each vegetation survey quadrat using a plum blossom sampling method [27], with each subsample measuring 10 cm × 10 cm × 10 cm. The five subsamples were then mixed into one composite sample, and five composite soil samples were obtained from each inundation gradient. In total, 2 × 5 × 5 = 50 soil seed bank samples were collected and brought back to the laboratory for germination experiments. After being naturally air-dried at room temperature, the soil samples were passed through a 4-mesh sieve to remove weed residues and other debris. Disposable plastic containers were used as germination trays. Under sufficient sunlight, water was applied daily as needed to keep the soil surface moist and to promote full seed germination. Germinated seedlings were counted every 3 days, and the recorded C. dactylon seedlings were removed after counting. The experiment was considered complete when no further germination was observed for two consecutive weeks. The germination experiment began on 14 July 2024 and ended on 16 September 2024, lasting 64 days.

2.4. Data Processing and Analysis

Based on the quadrat survey results, species composition, plant individual number, frequency, coverage, and the functional traits of dominant species were summarized separately for the upper and lower zones of each study site. Plant importance value and dominance were then calculated according to the following equations [28,29].

$$I_v={\displaystyle \frac{RC+RF+RD}{3}}$$

(1)

$$Y=P_i\times f_{\mathrm{i}}$$

(2)

In the formulas: RD (relative density) = (number of individuals of a given species/total number of individuals) × 100%; RF (relative frequency) = (number of quadrats in which the species occurs/sum of occurrences across all species) × 100%; RC (relative coverage) = (total cover of the species/total cover of all species) × 100%. (P_i = N_i/N), where (N_i) is the importance value of species (i), (N) is the sum of importance values for all species within the quadrat, (f_i) is the frequency of occurrence of species (i). To identify the sensitive traits of C. dactylon under drawdown zone habitat conditions, redundancy analysis (RDA) was conducted using the vegan package in R to examine the relationships between the functional traits of C. dactylon and environmental factors associated with drought and inundation. Landscape pattern indices were calculated using Fragstats 4.2. The detailed procedures for extracting landscape pattern indices and vegetation coverage for each study site are provided in the Supplementary Materials.

Spearman rank correlation analysis was used to examine the monotonic relationship between successional stage and overall stress memory, with stages I–V coded as ordinal values from 1 to 5. Given the small number of stage-level observations, exact two-sided p-values were reported.

2.5. Characterization of Stress Memory

Stress memory was operationally quantified as the similarity in adaptive ecological attributes between each successional stage and the reference stage, based on dominance, functional traits, landscape pattern indices, and soil seed bank density. Specifically, the framework combined four indicator groups: species dominance, functional traits, landscape pattern indices, and soil seed bank density, representing population dominance, individual functional performance, spatial occupation, and reproductive potential, respectively. For each successional stage and inundation gradient, these indicators were compared with the reference stage using the Bray–Curtis similarity index, and the weighted similarity values were summed to obtain the overall stress memory value. The indicator weights used in the calculation are provided in Section 2.6.

Following this framework, the stress memory of C. dactylon at different restoration stages was assessed based on the amount of “residual resources” [30] retained relative to the reference ecosystem. In this study, “residual resources” were interpreted as the degree of resource overlap or similarity between the selected successional stage and the reference stage. Stress memory between a selected successional stage and the preceding stage was defined as retrospective memory, whereas stress memory between the selected stage and the subsequent stage was defined as prospective memory. Accordingly, stress memory can be expressed by the following equation [31]

$$\mathrm{E}\mathrm{M}={{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\omega }_{\mathrm{i}}{\mathrm{R}}_{\mathrm{i}}$$

(3)

where EM denotes ecological stress memory; ${\mathrm{R}}_{\mathrm{i}}$ represents the degree of resource overlap between the ith stress memory indicator systems; ω is the weight assigned to ${\mathrm{R}}_{\mathrm{i}}$; and n is the number of stress memory indicators. We used the Bray–Curtis similarity index [32] to quantify stress memory because it measures the compositional similarity between a reference stage and a target stage across multiple ecological indicators. Conceptually, higher similarity indicates that the current stage retains more of the “residual resources” or adaptive effects accumulated from previous stress experiences, which aligns with the definition of stress memory.

$$\mathrm{S}=1-{\displaystyle \frac{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{m}}} |{\mathrm{y}}_{\mathrm{i}1}-{\mathrm{y}}_{\mathrm{i}2}|}{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{m}}} {\mathrm{y}}_{\mathrm{i}1}+{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{m}}} {\mathrm{y}}_{\mathrm{i}2}}}$$

(4)

where S is the Bray–Curtis similarity index between systems, y_i1 is the measured value of the ith indicator in Ecosystem 1, and y_i2 is the measured value of the ith indicator in Ecosystem 2. Accordingly, Equation (1) can be rewritten as follows:

$$\mathrm{E}\mathrm{M}={{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\omega }_{\mathrm{i}}{\mathrm{S}}_{\mathrm{i}}={\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}} {\omega }_{\mathrm{i}}(1-{\displaystyle \frac{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}} |{\mathrm{y}}_{\mathrm{i}1}-{\mathrm{y}}_{\mathrm{i}2}|}{{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}} {\mathrm{y}}_{\mathrm{i}1}+{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}} {\mathrm{y}}_{\mathrm{i}2}}})$$

(5)

This study selected the Ertan Reservoir as the reference baseline for calculating the overall stress memory of C. dactylon, as it had the highest vegetation cover among the five sites and represented the most advanced stage of successional development. Using this reservoir as a reference provides a consistent and ecologically meaningful baseline for evaluating stress memory across other stages.

2.6. Expert Scoring Method for Stress Memory Indicators

To quantitatively evaluate the importance of community assessment indicators in reservoir drawdown zones, the expert scoring method was adopted to determine the weights of the selected indicators [33]. The evaluation system included six indicators: plant height (PH), number of branch stems (NBS), dominance, percentage of landscape (PLAND), largest patch index (LPI), and the seed bank density. First, five experts in ecology and botany were invited to score each indicator according to its contribution to community structure and ecological function on a scale of 1 to 5, thereby forming an expert scoring matrix, A = [a_ij], where i denotes the indicator and j denotes the expert. The detailed expert scoring matrix is shown in

Table 1. Expert scoring matrix.

Indicator	Expert 1	Expert 2	Expert 3	Expert 4	Expert 5
PH	2	2	1	2	2
NBS	2	1	2	1	2
Dominance	4	5	4	4	4
PLAND	4	4	5	4	4
LPI	1	1	2	1	1
Seed bank density	5	5	5	5	5

Note: The indicators evaluatedt include plant height (PH); number of branching stems (NBS); Dominance; percentage of landscape (PLAND); largest patch index (LPI) and seed bank density.

.

An equal-weight sensitivity analysis was additionally performed to assess whether the overall stress memory pattern was sensitive to the expert-derived weights, with detailed comparisons provided in the Supplementary Materials.

The expert scores for each indicator were averaged to obtain the composite score of that indicator:

$$\overline{a_i}={\displaystyle \frac{1}{n}}{\displaystyle \sum _{j=1}^n}{a}_{ij}$$

(6)

The composite scores of all indicators were then normalized to calculate the weight of each indicator, ${ \omega }_i$

$${\omega }_i={\displaystyle \frac{\overline{a_i}}{{\sum }_{k=1}^m\overline{a_k}}}$$

(7)

The final weights are shown in

Table 2. Weights of the component indicators.

Indicator	Mean Score	Actual Weight	Assigned Weight	Ecological Significance
PH	1.8	0.1	0.1	Reflects the vertical structure of the community
NBS	1.6	0.09	0.1	Reflects horizontal expansion capacity
Dominance	4.2	0.23	0.2	Reflects species dominance status
PLAND	4.2	0.23	0.2	Reflects community coverage status
LPI	1.2	0.07	0.1	Reflects the proportion of the largest patch
Seed bank density	5	0.28	0.3	Reflects community regeneration capacity

Note: The indicators evaluatedt include plant height (PH); number of branching stems (NBS); Dominance; percentage of landscape (PLAND); largest patch index (LPI) and seed bank density.

:

3. Results

3.1. Multidimensional Evolutionary Characteristics of Cynodon dactylon During the Restoration Process

3.1.1. Analysis of the Dominance of Vegetation in the Study Area

Across the drawdown zones of five typical reservoirs in the Jinsha River, field surveys recorded a total of 32 vascular plant species belonging to 29 genera and 15 families. Annual herbs were predominant (21 species, 65.63%), followed by perennial herbs (10 species, 31.25%), and one shrub species (3.13%). A complete species list with records for each plot is provided in the Supplementary Materials.

In ecological studies, a species is generally identified as a dominant species when its dominance value exceeds 0.01, and as an absolute dominant species when its dominance value exceeds 0.1, indicating a leading position within the community [34]. The dominance values of the main dominant species across different successional stages are shown in Figure 2a. As shown in Figure 2a, the dominance value of Cynodon dactylon was greater than 0.1 in all five successional stages.

The dominance values of C. dactylon in the upper and lower zones across different successional stages are shown in Figure 2b. Except for the lower zone of Stage I, where no exposed vegetation was present during the initial stage of succession and the dominance value was therefore 0, the dominance value of C. dactylon exceeded 0.1 in both the upper and lower zones of all other stages. This indicates that C. dactylon maintained an absolute dominant position across different successional stages and spatial locations. Overall, except for Stage III, the dominance of C. dactylon was higher in the upper zone than in the lower zone, revealing a clear pattern of spatial differentiation. Among all stages, the highest dominance value in the lower zone occurred in Stage III, reaching 0.266, whereas the dominance value in the upper zone reached 0.351 in Stage V, which was the maximum value observed across all stages. In general, C. dactylon exhibited high dominance across different successional stages and in both upper and lower positions, while also showing a certain degree of spatial variation.

3.1.2. Functional Trait Analysis of Cynodon dactylon

The Redundancy Analysis (RDA) ordination results for plant functional traits and environmental factors are shown in Figure 3. As shown in Figure 3a, the first two ordination axes jointly explained 70.4% of the variation in the functional traits of C. dactylon, with RDA-1 and RDA-2 explaining 42.1% and 28.3%, respectively. Among the environmental variables, inundation duration and inundation depth had relatively long arrows, indicating that they had strong explanatory power for the differentiation of functional traits in C. dactylon. Among the functional trait variables, plant height and branch number showed markedly longer arrows than the other traits, suggesting greater variation along the ordination axes and a stronger response to environmental gradients.

Figure 3b–e further illustrate the variation in plant height (PH) and number of branching stems (NBS) across different successional stages and in the upper and lower zones. Overall, both traits showed an increasing trend with successional progression. PH exhibited a relatively continuous increase in both the upper and lower zones, with a more pronounced rise after Stage III. NBS generally increased in the upper zone, whereas some fluctuations were observed in the lower zone, although its overall level remained higher than that in the initial successional stage. The specific values of the other functional traits are provided in Supplementary Table S5. Taken together, the RDA ordination results and stage-specific variation indicate that PH and NBS were relatively sensitive in reflecting the responses of C. dactylon to environmental gradients and successional differences, and can therefore be used as important functional trait indicators in the subsequent analysis of stress memory.

3.1.3. Analysis of Landscape Pattern Indices of the Study Sites

The landscape pattern indices of C. dactylon across different successional stages are shown in Figure 4. Among these indices, the percentage of landscape (PLAND) and the largest patch index (LPI) exhibited relatively clear stage-specific variation. Overall, both the coverage (PLAND) and the largest patch index (LPI) of C. dactylon in the upper and lower zones generally increased with successional progression, indicating a gradual increase in the proportion and dominance of C. dactylon patches within the landscape.

In terms of specific changes, PLAND and LPI in the upper zone reached their maximum values in Stage IV, at 52.67 and 50.51, respectively. Although both indices declined slightly in Stage V, they remained generally higher than those in the earlier stages. In the lower zone, PLAND and LPI also showed an overall increasing trend, although their values were generally lower than those in the upper zone. By contrast, patch density (PD) and mean patch area (AREA) fluctuated more markedly across different stages and between the upper and lower zones. Specifically, PD showed stage-dependent variation overall, whereas AREA increased substantially in the upper zone during the later stages and fluctuated more strongly in the lower zone. In general, both PLAND and LPI of C. dactylon exhibited an overall increasing pattern with successional progression.

3.1.4. Analysis of Soil Seed Density and Its Distribution Characteristics

As shown in Figure 5, the seed density of C. dactylon exhibited clear spatial differences between the upper and lower zones. Specifically, the mean seed density in the upper-zone plots across different successional stages ranged from 20 to 360 seeds/m², whereas that in the lower-zone plots was comparatively lower, ranging from 0 to 83 seeds/m². Overall, seed density was consistently higher in the upper-zone plots than in the lower-zone plots, indicating a greater accumulation of C. dactylon seeds in the upper zone.

In terms of stage-specific variation, the seed density of C. dactylon generally increased with successional progression in both the upper and lower zones. In the upper-zone plots, seed density increased gradually from 20 seeds/m² in Stage I to 360 seeds/m² in Stage V. In the lower-zone plots, it increased from 0 seeds/m² in Stage I to 83 seeds/m² in Stage V, also showing a clear increasing trend. Meanwhile, substantial differences between the upper and lower zones persisted across all stages, indicating pronounced spatial differentiation in seed density distribution. Combined with the variation observed at different spatial positions, the results suggest that seed density was positively associated overall with the water-level gradient, that is, the seed density of C. dactylon increased with increasing water-level gradient. Correspondingly, seed density remained relatively low in the lower zone, where flooding disturbance was stronger.

3.2. Stress Memory Across Different Successional Stages

3.2.1. Variation in Individual Component Indicators and Overall Stress Memory

Using the Ertan site as the reference endpoint of succession, the variation in the individual component indicators and the overall stress memory of C. dactylon across different successional stages and in the upper and lower zones was calculated (Figure 6 and Figure 7). Overall, the total stress memory of C. dactylon increased with successional progression in both the upper and lower zones. In the upper zone, overall stress memory gradually increased from 0.234 in Stage I to 1.000 in Stage V. In the lower zone, it increased from 0 in Stage I to 1.000 in Stage V. By comparison, the upper zone exhibited a relatively higher level of stress memory in the early stages of succession, whereas the lower zone showed a gradual accumulation after Stage II. These results indicate that overall stress memory increased continuously in both spatial positions, although the onset and pace of accumulation differed between the upper and lower zones.

The stress memory reflected by the individual component indicators also generally increased with successional progression, although their accumulation patterns were not entirely consistent. Among them, the stress memory associated with seed bank density and plant height exhibited relatively continuous increasing trends. Dominance, LPI, and coverage (PLAND) also showed overall increases, although some fluctuations occurred at specific stages. The stress memory of branch number likewise increased overall, but its stage-specific fluctuations were relatively more pronounced. Taken together, both the individual component indicators and the overall stress memory increased steadily, indicating that the stress memory of different ecological components of C. dactylon gradually accumulated and strengthened as succession progressed. Spearman rank correlation analysis further showed that overall stress memory was positively correlated with successional stage in both the lower and upper zones (ρ = 1.000, p = 0.0167).

3.2.2. Stage Differences Between Prospective Memory and Retrospective Memory

The calculated results of prospective memory and retrospective memory are presented in

Table 3. Stress memory of Cynodon dactylon at different successional stages.

Inundation Gradient	Stage	V	IV	III	II	I
Lower-zone	V	1
	IV	0.889	1
	III	0.805	0.914	1
	II	0.67	0.772	0.855	1
	I	0	0	0	0	1
	Retrospective memory	0.889	0.914	0.855	0
	Prospective memory		0.889	0.914	0.855	0
Upper-zone	V	1
	IV	0.873	1
	III	0.82	0.946	1
	II	0.69	0.1	0.862	1
	I	0.379	0.464	0.504	0.615	1
	Retrospective memory	0.873	0.946	0.862	0.615
	Prospective memory		0.873	0.946	0.86	0.615

. Overall, the increase in stress memory from Stage I to Stage II was greater than that between the other adjacent stages, indicating that the main accumulation of stress memory in C. dactylon occurred during the early stage of succession. In both the upper and lower zones, prospective memory was higher than retrospective memory in Stage II and Stage III, suggesting that these two stages were in a process of positive succession. In contrast, at Stage IV, prospective memory was lower than retrospective memory, indicating that the later stage of succession exhibited different characteristics. It should be noted that Stage I and Stage V represent the initial and final stages of the successional sequence, respectively. Because no corresponding reference stage exists before Stage I or after Stage V, these two stages were not included in the calculation of retrospective memory or prospective memory.

4. Discussion

4.1. The Role of Stress Memory in the Dynamics of Cynodon Dactylon in Reservoir Drawdown Zones

Ecosystems often exhibit different adaptive characteristics across successional stages, and their transitions under persistent disturbance may depend largely on the balance between accumulated adaptation and subsequent environmental responses [35,36]. For plants in reservoir drawdown zones, retrospective memory reflects the retention and continuation of previous stress experiences, whereas prospective memory represents adaptive preparation for subsequent environmental changes. Together, these two forms of memory could contribute to plant growth strategies and population persistence, and may be consistent with stage-specific dynamics.

The stress memory of C. dactylon showed progressive accumulation along the recovery gradient. This suggests that under the long-term background of water-level fluctuations and repeated cycles of inundation and exposure, C. dactylon does not merely passively endure environmental variation but gradually develops substantial adaptive accumulation through repeated exposure to stress. Stress memory is not merely an accompanying feature of succession, but may represent an intrinsic factor contributing to the sustained adaptation, expansion, and progression of C. dactylon under long-term hydrological fluctuations. As a perennial herb, the growth, reproduction, and expansion of C. dactylon are continuously influenced by water-level fluctuations, changes in soil conditions, and habitat heterogeneity [37]. In this process, retrospective memory formed through previous environmental experience helps preserve adaptive responses to periodic inundation, short-term drought, and substrate variation, whereas prospective memory enables the species to respond in advance to potential subsequent changes based on current environmental signals [38,39]. This dual regulation, reflecting responses to both past experiences and potential future changes, may contribute to the survival, expansion, and recovery of C. dactylon under highly disturbed conditions [11,12].

Further analysis suggested that prospective memory exceeded retrospective memory in Stages II and III, whereas retrospective memory became more important in later stages. This suggests that C. dactylon prioritizes adaptive preparation for future environmental change during the early and middle stages, thereby promoting population expansion and community reconstruction, while shifting toward a stage-specific adjustment strategy in later stages. These results suggest that the recovery dynamics of C. dactylon may not follow a simple linear progression but could involve structural adjustments and reorganization of adaptive strategies within an overall recovery trend. Together, these patterns indicate that retrospective and prospective memory play complementary roles in the adaptive process of C. dactylon, help explain the stage-specific dynamics of vegetation recovery in reservoir drawdown zones, and provide insight into the intrinsic mechanisms underlying this recovery process.

4.2. Indicator Value of the Stress Memory Framework and Its Implications for Ecological Restoration

Research on stress memory not only provides a new perspective for understanding how plants adapt to high-frequency environmental fluctuations but also offers a new analytical framework for ecological restoration and management in reservoir drawdown zones [40,41]. Compared with approaches that assess restoration status solely on the basis of vegetation cover, dominance, or a single functional trait, this study incorporated multiple dimensions—dominance, functional traits, landscape pattern, and the soil seed bank—into a unified stress memory assessment framework, thereby enabling a more systematic evaluation of the adaptive variation of C. dactylon across different successional stages. The results showed that the memory reflected by all component indicators generally increased, although the modes of increase were not entirely consistent. Specifically, the memory intensity associated with the soil seed bank and plant height increased progressively with succession, showing relatively stable accumulation patterns. By contrast, dominance, coverage, branch number, and the largest patch index fluctuated to some extent across stages, but still exhibited overall increasing trends. These results indicate that different indicators vary in their sensitivity and stability in response to successional progression. Therefore, a single indicator is insufficient to fully capture the restoration status of vegetation in reservoir drawdown zones, whereas a multi-indicator integrated framework is more effective for identifying key ecological response indicators and their stage-specific variation. Although this study focused on C. dactylon, the proposed framework may be transferable to other dominant wetland or riparian species experiencing recurrent hydrological stress, such as Typha spp., Schoenoplectus spp., and other rhizomatous or clonal plants. Its transferability lies in the integration of population performance, functional traits, spatial organization, and reproductive potential. However, the specific indicators and weights should be adjusted according to species-specific life-history strategies, reproductive modes, and habitat preferences. Therefore, the framework should be regarded as a flexible assessment approach rather than a fixed indicator system limited to C. dactylon.

From the perspective of restoration practice, this framework has clear application potential. First, in vegetation restoration and species selection, C. dactylon should be prioritized as a core species, with locally abundant species such as Xanthium strumarium and Alternanthera sessilis considered as supplementary species to enhance community diversity and stability. Restoration strategies should take into account inundation duration, slope, and topography: gently sloping and lightly inundated areas are most suitable for planting, while heavily inundated or steep zones may require additional soil or hydrological management. Second, the differences among memory components across stages suggest that restoration assessment in drawdown zones should not rely on a single indicator. Instead, multiple aspects, including functional traits, spatial pattern, and reproductive potential, should be considered comprehensively according to the restoration stage and management objectives. Third, the stage-specific adjustment signals observed in the later stage of succession indicate that vegetation recovery in drawdown zones does not always proceed steadily in a single direction. Accordingly, long-term monitoring should be strengthened in management practice, with particular attention paid to community structural reorganization and changes in recovery rhythm, so as to avoid simply interpreting short-term fluctuations as restoration stagnation or failure.

Nevertheless, the formation mechanisms of stress memory in the highly disturbed habitat of reservoir drawdown zones still require further investigation. Previous studies have shown that epigenetic modification, gene expression regulation, and protein interaction networks may all participate in the formation of plant stress memory [42,43]. However, under drawdown zone conditions, how these mechanisms interact with long-term hydrological fluctuations and further influence species establishment, competition, and distribution patterns still requires additional verification [44,45]. Future research should therefore strengthen the integration of molecular biology, hydroecology, and restoration ecology in order to reveal, across multiple scales, the formation mechanisms of stress memory and its ecological consequences. Overall, this study indicates that stress memory not only provides an important perspective for understanding the long-term adaptation of dominant plants in reservoir drawdown zones, but may also represent one of the intrinsic mechanisms supporting their continued succession, while offering valuable scientific support for identifying restoration potential, screening key indicators, and informing ecological restoration practice.

5. Conclusions

The stress memory of C. dactylon increased with successional progression in both the upper and lower zones, exhibiting an overall trend of continuous accumulation. This indicates that C. dactylon developed substantial adaptive accumulation under long-term hydrological fluctuations. The memory reflected by all component indicators also generally increased, although their patterns of change differed, with plant height and seed bank density showing relatively greater stability. Notably, seed density was positively associated with the water-level gradient, increasing with higher positions along the gradient, while remaining relatively low in the lower zone, where flooding disturbance was stronger. Prospective memory was higher than retrospective memory in Stage II and Stage III, whereas the later stage of succession suggested possible stage-specific adjustment. This indicates that the recovery dynamics of Cynodon dactylon do not follow a simple linear progression, rather that they involve a certain reorganization of adaptive strategies within an overall recovery trend. Overall, the concept of stress memory and its integrated assessment framework not only help identify key ecological response indicators during vegetation recovery in reservoir drawdown zones, but also provide a scientific reference for judging vegetation succession, identifying restoration potential, and informing ecological restoration and sustainable management.