Spatiotemporal Dynamics of Drought–Flood Abrupt Alternations and Their Delayed Effects on Vegetation Growth in Heilongjiang River Basin

Abstract

Drought–flood abrupt alternations (DFAAs) have a greater impact on ecosystems and socioeconomic environments than lone droughts or floods. Despite the significant impact of DFAAs, research has paid little attention to their evolutionary characteristics, particularly in relation to vegetation growth in the Heilongjiang River Basin. Therefore, this study focuses on the Heilongjiang River Basin and employs the DFAA Index to identify and analyze abrupt alternation events from 1970 to 2019. It also examines the annual and interannual distributions of vegetation growth changes from 2000 to 2019, based on the Normalized Difference Vegetation Index. Lastly, it utilizes correlation analysis to investigate the responsive relationship between vegetation growth and DFAA events. The results indicate the following: (1) Within the Heilongjiang River Basin, the number of drought-to-flood events increased over time, whereas the number of flood-to-drought events decreased over time. The frequency of mutation was relatively high in the northern region, low in the eastern region, elevated in spring and summer, and reduced in winter. (2) The Normalized Difference Vegetation Index was lowest in January, highest in July, and approximately 0 during the winter. The vegetation coverage reached its peak during the summer. (3) Vegetation changes in response to DFAAs exhibited a significant time lag. Vegetation changes in spring–summer lagged behind DFAA events by 3–4 months, while in summer–autumn, the lag was approximately 3 months. These results are of great significance for the early warning and prevention of DFAAs in the Heilongjiang River Basin.

1. Introduction

Drought–flood abrupt alternation (DFAA) events are defined as “the phenomenon wherein drought and flood occur alternately over time or coexist within a spatial context”, and the alterations in frequency and intensity present a significant threat to ecosystems, social economy, and regional sustainable development [1,2]. Recently, the spatiotemporal patterns of DFAAs have become increasingly complex. This is, in part, due to the intensification of global climate change. Global warming alters the atmospheric circulation pattern, markedly influences the spatiotemporal distribution of precipitation and the frequency of extreme climate events, and complicates regional water resource regulation [3,4,5]. Human activities, including but not limited to land use change, urbanization, and vegetation degradation, further exacerbate DFAA risk, rendering their evolutionary characteristics more prominent, both locally and regionally [6]. Vegetation growth is an important factor in ecosystem stability and the hydrological cycle. Its spatiotemporal variation serves as a direct manifestation of climate change and human activities and exerts a feedback effect on DFAAs [7]. Nevertheless, the responses of various vegetation to drought–flood abrupt alternation events differ significantly. Certain vegetation types maintain their stability under DFAA conditions via long-term adaptation. However, other types (such as rice and cotton) suffer severe degradation during extreme climate events [8,9]. Therefore, it is crucial to investigate the response of vegetation growth to DFAA events.

In recent decades, this issue has attracted significant attention from experts and scholars. Research on crops like corn and rice has revealed that DFAA events augment the loss of nitrogen and phosphorus in the soil, curtail the absorption of these elements by crops, and consequently diminish crop yields [10,11,12]. Several studies have demonstrated that DFAA events can significantly affect vegetation growth by altering the availability of sunlight and soil nutrients, thereby disrupting photosynthesis processes and leading to a reduction in vegetation coverage and species diversity. These effects are particularly pronounced in areas with frequent flooding, where the synergistic and antagonistic impacts of drought and flood alternation exacerbate the stress on vegetation [13,14]. Although previous research has highlighted the widespread impact of DFAA events, their temporal distribution and spatial extent remain inadequately understood, particularly in terms of the abrupt transitions between drought and flood conditions and their lag effects on vegetation. Further research is needed to clarify the specific characteristics of DFAA events and their influence on vegetation dynamics, especially regarding their long-term effects on ecosystem recovery and biodiversity.

Recently, drought–flood abrupt alternation has emerged as a prominent topic within the broader field of global climate change research. The current research regarding DFAAs primarily relies on observational data and numerical simulation [15,16]. Index construction is the chief methodology in DFAA research, and indices such as DFAAI, LDFAAI, and SDFAAI have been widely employed in quantitative analysis [14,17,18]. The spatiotemporal evolution analysis predicated by these indices, in conjunction with climate models, can facilitate deeper exploration into the mechanisms that underlie DFAAs. Multi-scale data fusion techniques, combined with both satellite remote sensing and ground observation, have also been extensively utilized in regional research [19,20]. Recently, more attention has been devoted to the spatiotemporal distribution of DFAAs across diverse regions. Studies have revealed that global warming has augmented the frequency and magnitude of drought and flood events, with the Asian monsoon region and the Midwest region of North America labeled as high-risk regions. Studies have uncovered significant correlations between DFAAs and climate factors, including the El Nino–Southern Oscillation (ENSO), the Arctic Oscillation (AO), and the Indian Ocean dipole (IOD). These factors are closely related to the seasonal regulation of precipitation and evaporation [21,22]. By improving hydrological and climate models, the simulation accuracy and predictive capabilities for DFAAs have been enhanced. However, terrain, soil moisture, and environmental alterations cause substantial discrepancies in the local DFAA prediction capabilities of different models [23,24]. Although a plethora of methods have been employed to conduct quantitative analyses on DFAA events, there remains a scarcity of studies on DFAA spatiotemporal evolution in specific regions. Investigating the region-specific spatiotemporal distributions and changing trends of DFAAs helps bridge gaps in regional research and deepen the scientific understanding of their spatiotemporal characteristics.

The Heilongjiang River Basin is situated in the northeastern region of China and the Russian far East. It represents a crucial ecological and agricultural area, wherein its hydrological processes and ecosystems exhibit an acute sensitivity to global climate change. Recently, the frequency and intensity of DFAA events in the Heilongjiang River Basin have undergone significant changes due to climate warming and human activities [25]. Despite the importance of these events, there remains a critical gap in research on their spatiotemporal distribution and their impact on vegetation growth in the basin. This study aims to address this gap by analyzing the spatiotemporal patterns of DFAA events, integrating these patterns with vegetation growth characteristics, and examining the coupling mechanisms between DFAA events and vegetation responses. Thus, this study will contribute to a deeper understanding of the effects of climate change and human activities on the ecosystem and provide valuable insights for ecological protection and sustainable development in the Heilongjiang River Basin.

2. Research Area and Data

2.1. Location of the Study Area

The Heilongjiang River Basin, located at coordinates (41°42′–55°56′ N, 107°31′–141°14′ E), lies in eastern Eurasia and is one of the ten largest rivers in the world. It encompasses an area of approximately 184.3 × 10⁴ km² and represents the largest transboundary basin in northeast Asia. Most of its eastern and western portions are situated in Russia, while a small western section lies in Mongolia, and the remainder is located in the Heilongjiang Province, Inner Mongolia, and the Jilin Province in China. In spring, it is often windy; in summer, it is humid; in autumn, the weather is erratic; and in winter, it is cold and dry. The average annual temperature ranges from −14.37 °C to 6.75 °C, demonstrating significant seasonal and regional variations. Much of the precipitation occurs between June and September, with the precipitation in July accounting for nearly 70% of the annual precipitation. The average annual precipitation varies significantly throughout the year and across different regions of the basin. For example, the annual precipitation in the western region ranges from 200 to 400 mm, while that in the eastern coastal region ranges from 500 to 800 mm. Moreover, the precipitation exhibits substantial interannual variability. For instance, in 2016, the precipitation amounted to 632.30 mm. In contrast, it was only 270.15 mm in 1999, representing a difference of approximately 2.3 times. The Heilongjiang River Basin is partitioned into eight zones, based on characteristics of the basin’s water system. These areas include the Heilongjiang River’s main stream area, the Jieya River Basin, Brea River Basin, the Amgon River Basin, the Shilka River Basin, the Erguna River Basin, the Songhua River Basin, and the Ussuri River Basin. Figure 1 provides a general overview of the research area.

2.2. Data Sources

2.2.1. Runoff Data

The runoff data used in this study were obtained from the daily runoff simulation run by the Community Water Model (CWatM), which is a large-scale hydrological and water resources model with a spatial resolution of 0.5°. The model can simulate global and regional hydrological processes within a daily timestep range of 30 arcmin to 30 arcsec. CWatM encompasses general surface water and groundwater hydrological processes, while also accounting for human activities—such as water use and reservoir regulation—by calculating water demand, water use, and backflow [26]. Reservoirs and lakes are also incorporated into the model framework. In this study, the runoff data spanning from 1970 to 2019 were selected for analysis.

In this study, runoff and evapotranspiration data from 2007 to 2012 were used for calibration, and data from 2012 to 2017 were used for validation. The model’s performance was evaluated using the Kling–Gupta Efficiency (KGE) metric, which provided a comprehensive assessment of the simulation accuracy [27,28].

2.2.2. Normalized Difference Vegetation Index (NDVI)

The Normalized Difference Vegetation Index (NDVI), initially proposed by Rouse, can accurately reflect the physical mechanisms of vegetation growth, which lie in the disparity between chlorophyll absorption in the red wavelength and the green scattering in the near-infrared wavelength. GIMMS NDVI 3g, SPOT NDVI, and MODIS NDVI (MOD13 and MYD13) represent the most prevalently utilized vegetation index datasets. The MODIS Normalized Difference Vegetation Index, provided by the National Aeronautics and Space Administration (NASA) (https://lpdaac.usgs.gov/products/mod13a3v061/, accessed on 13 October 2024), complements the NDVI product of NOAA’s Advanced Very High Resolution Radiometer (AVHRR) and ensures the continuity of historical timeseries applications. The MOD13A3v061 dataset from 2000 to 2019, with a spatial resolution of 1 km, was used in this study. Firstly, the vegetation coverage was estimated using the pixel binary model. Then, a series of extractions, mosaicking, projections, cropping, null value removals, and resampling processes were conducted. The monthly NDVI data were subsequently converted into the TIFF format. Lastly, the monthly NDVI raster data from 2000 to 2019 were extracted based on the Heilongjiang River Basin’s vector boundary.

2.2.3. Unified Data Resolution and Analysis Scale Selection

In this study, the original spatial resolution of the runoff data was 0.5°, which served as the main basis for analysis. The NDVI data, with an original resolution of 1 km, was primarily used to characterize vegetation conditions within the study area, rather than for a pixel-by-pixel correlation analysis. Therefore, all correlation analyses were conducted at the 0.5° resolution of the runoff data.

Regarding the resolution difference, although the NDVI data offer higher spatial precision, this study focused on hydrological–ecological processes at the basin scale (i.e., the 0.5° grid scale), making it both reasonable and necessary to adopt the runoff data’s 0.5° resolution as the unified analytical scale. Moreover, aggregating the high-resolution NDVI data to the 0.5° scale helped retain more of the original spatial heterogeneity of vegetation. Compared to directly using lower-resolution vegetation data, this approach improved the representativeness and spatial accuracy of vegetation indicators after aggregation.

3. Method

3.1. Community Water Model (CWatM)

Within a grid cell, “runoff concentration” refers to the process through which the runoff generated within the cell converges towards the river path. The runoff originating from each unit is directed to the corners of the corresponding unit. The concentration time (peak time) is determined using the land cover grade, slope, and runoff group (including surface runoff, confluence runoff, or base flow). Subsequently, the triangular weighting function is employed to compute the total runoff of the grid cells [26]. The function is as follows:

$$Q(t)={\displaystyle \sum _{land\mathrm{cov}er}}{\displaystyle \sum _{runoff}}{\displaystyle \sum _i^{\max }c(i)Q_{runoff}(t-i+1)}$$

(1)

where Q(t) is the total runoff of the grid cell with a timestep; runoff is the runoff component (surface, interflow, base flow); Q_runoff is the land cover runoff of runoff component; t is the time (1 d); and c(i) is a triangular function, which can be expressed as follows:

$$c(i)={\displaystyle \underset{i-1}{\overset{i}{\int }}\frac{2}{\mathrm{m}\mathrm{a}\mathrm{x}}}-\left|u-{\displaystyle \frac{\mathrm{m}\mathrm{a}\mathrm{x}}{2}}\right|\cdot {\displaystyle \frac{4}{\mathrm{m}\mathrm{a}{\mathrm{x}}^2}}du$$

(2)

The hydrographic station was located at the outlet of the Heilongjiang River Basin. The calibration period encompassed 37 years of observational data, with the results provided in Figure 2. With the revised Kling–Gupta efficiency (KGE) serving as the objective function, a KGE value closer to 1 indicates a better fit [29]. The formula is as follows:

$$\mathrm{K}\mathrm{G}{\mathrm{E}}^{\prime }=1-\sqrt{{\left(r-1\right)}^2+{\left(\beta -1\right)}^2+{\left(\gamma -1\right)}^2}$$

(3)

where $r$ is the correlation coefficient between simulated and observed values; $\beta$is the bias ratio; and $\gamma$ is the variability ratio.

3.2. Drought–Flood Abrupt Alteration Index (DFAAI)

In this study, the drought–flood abrupt alteration index (DFAAI) was used to quantitatively describe DFAA events. The DFAAI is defined as follows:

$$\mathrm{D}\mathrm{F}\mathrm{A}\mathrm{A}\mathrm{I}=(N{Q}_i-N{Q}_{i-1})\cdot (\left|N{Q}_i\right|+\left|N{Q}_{i-1}\right|)\cdot {\alpha }^{-\left|N{Q}_i+N{Q}_{i-1}\right|} (i=2,4,3,\dots ,n)$$

(4)

$$N{Q}_i={\displaystyle \frac{Q_i-\overline{Q}}{\sigma }}$$

(5)

where $N{Q}_i$ and $N{Q}_{i-1}$ represent the normalized monthly runoff for month i and month i−1; $Q_i$ is the average monthly runoff; $\overline{Q}$ and $\sigma$ are the mean and standard error of $Q_i$; $\alpha$ is the monthly weight coefficient (a value of 3.2 is more appropriate); i is the month number; and n is the total number of months in the study period [18,30]. The item ($N{Q}_i$ − $N{Q}_{i-1}$) represents the intensity of the DFAAs; the item ($\left|N{Q}_i\right|$ + $\left|N{Q}_{i-1}\right|$) denotes the magnitude of the droughts and floods; and ${\alpha }^{-\left|N{Q}_i + N{Q}_{i-1}\right|}$ is the weight coefficient (which may decrease the weight of droughts or floods in two consecutive months and increase the weight of the DFAAs). DFAAI values greater than 1 (>1) or less than −1 (<−1) are defined as DTF or FTD, respectively. The greater the absolute value of the DFAAI, the stronger the DFAA.

3.3. Wavelet Analysis

For time series x(t), the Continuous Wavelet Transform (CWT) is defined as follows:

$$W_x(s,\tau )={\int }_{-\infty }^{\infty }x(t){\psi }^{*}({\displaystyle \frac{t-\tau }{s}})dt$$

(6)

where Wx(s,τ) is the CWT coefficient on the time scale (τ,s); s is a scale parameter (inversely proportional to frequency); τ is the time translation parameter; $\psi \left(t\right)$ is the parent wavelet function; and ${\psi }^{*}\left(t\right)$ is the complex conjugate.

For time series x(t) and y(t), the Cross Wavelet Transform (XWT) is defined as follows:

$$W_{xy}(s,\tau )=W_x(s,\tau )\cdot W_y^{*}(s,\tau )$$

(7)

where Wx(s,τ) and Wy(s,τ) are the wavelet transforms of x(t) and y(t); Wy∗(s,τ) is the complex conjugate of the y(t) wavelet transform; and Wxy(s,τ) is the cross wavelet coefficient of x(t) and y(t) on the time scale (τ,s). Cross Wavelet Transform (XWT) is concerned with the joint energy distribution of two sequences in time and scale, but it cannot provide a normalized correlation measure [31,32].

Wavelet Coherence (WTC) is a standardized measure of correlation that provides the local linear correlation between two sequences in time and scale, compensating for XWT’s inability to distinguish between strong energies and correlations. It is an effective method for studying the scale correlation between two geophysical variables and is used to analyze the coherence of two time series in the time-frequency domain, thereby examining their covariance. The wavelet coherence coefficient is equivalent to the local correlation coefficient in the time-frequency domain and is defined as follows:

$$R_{\mathrm{xy}}^2(s,\tau )={\displaystyle \frac{{\left|S(s^{-1}{W}_{xy}(s,\tau ))\right|}^2}{S(s^{-1}{\left|W_x(s,\tau )\right|}^2)\cdot S(s^{-1}{\left|W_y(s,\tau )\right|}^2)}}$$

(8)

where s is the scale; τ is the time; Wx(s,τ) and Wy(s,τ) are the wavelet transform of time series x and y; Wxy(s,τ) is the Cross Wavelet Transform; s is the smoothing operator used to smooth the local wavelet coefficients in time and scale; and ${\left|W_x(s,\tau )\right|}^2$ and ${\left|W_y(s,\tau )\right|}^2$ is the local wavelet energy of two time series [33,34].

3.4. Pearson’s Correlation Analysis

For the variables X = {x₁, x₂, …, x_n} and Y = {y₁, y₂, …, y_n}, Pearson’s correlation coefficient r is calculated as follows:

$$r={\displaystyle \frac{{\sum }_{i=1}^n(x_i-\overline{x})(y_i-\overline{y})}{\sqrt{{\sum }_{i=1}^n(x_i-\overline{x}{)}^2{\sum }_{i=1}^n(y_i-\overline{y}{)}^2}}}$$

(9)

where x_i and y_i are the ith observed values of the two variables; $\overline{x}={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{x}_i$ is the mean of x; $\overline{y}={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{y}_i$ is the mean of y; numerator ${\sum }_{i=1}^n(x_i-\overline{x})(y_i-\overline{y})$ is covariance of two variables; denominator $\sqrt{{\sum }_{i=1}^n(x_i-\overline{x}{)}^2{\sum }_{i=1}^n(y_i-\overline{y}{)}^2}$ is the product of the standard deviations of two variables; Pearson’s correlation coefficient r ranges from [−1,1]; r = 1 indicates a perfect positive correlation (the two variables have a perfect linear positive relationship); r = −1 indicates a perfect negative correlation (a perfectly linear inverse relationship between the two variables); and r = 0 means no linear correlation (there is no linear relationship between the two variables, but other nonlinear relationships may exist) [5].

4. Results

4.1. Spatiotemporal Analysis of DFAAs

4.1.1. Spatial Distribution of DFAAs

To quantify the spatial distribution of DFAA events in the Heilongjiang River Basin, we tallied the amounts of drought-to-flood (DTF) and flood-to-drought (FTD) events in each sub-basin from 1970 to 2019. The proportion of these events was computed to illustrate the distribution frequency of DFAAs in the region. In Figure 3, HeilongjiangR illustrates the overall pattern of the region to highlight the general trend; arrows pointing to the individual sub-basins were not included in the figure to avoid confusion with the detailed analysis of the sub-basins. Figure 3 demonstrates that over the past 50 years, the occurrences of DFAA events in the Heilongjiang River main stream area (HLJRMS), Jieya River Basin (JieyaR), Brea River Basin (BreaR), Amgon River Basin (AmgonR), Shilka River Basin (ShilkaR), Erguna River Basin (ErgunaR), Songhua River Basin (SHR), and Ussuri River Basin (UssuriR) measured 14.15%, 10.72%, 7.02%, 5.79%, 14.31%, 14.74%, 18.92%, and 14.36%, respectively. Of these, the greatest frequency of DTF events was observed in the Songhua River Basin (SHR), while the lowest was in the AmgonR. The SHR also had the most FTD events, while the Brea River Basin (BreaR) had the least.

Differences were observed in the frequencies of the DTF and FTD events in the different sub-basins. Generally, the DTF events exhibited an upward trend, increasing at a rate of 3.51% per decade. In the AmgonR, however, the frequency of DTF events decreased by 1.17% per decade, making it the only sub-basin with a declining trend. The remaining seven sub-basins demonstrated an upward trend, with HLJRMS, JieyaR, BreaR, ShilkaR, ErgunaR, SHR, and UssuriR increasing at rates of 7.33%/decade, 2.14%/decade, 1%/decade, 2.67%/decade, 1.8%/decade, 8.48%/decade, and 1.29%/decade, respectively. FTD events demonstrated a downward trend, with a decreasing rate of 4.52% per decade. The frequency of FTD events in ShilkaR (1.9%/decade) and UssuriR (0.91%/decade) increased slightly. However, the other six sub-basins exhibited a downward tendency. The decreasing rates of HLJRMS, JieyaR, BreaR, AmgonR, ErgunaR, and SHR were 8.46%/decade, 9.4%/decade, 13.33%/decade, 13.33%/decade, 1.9%/decade, and 2.78%/decade, respectively.

The DTF events exhibited an overall upward trend from 1970 to 2019, whereas the FTD events demonstrated a downward trend. This reflects the spatiotemporal heterogeneity of DFAAs. Spatially, differences existed in both the frequency and trend of DFAA events. The frequency of alternations was relatively higher in the northern regions (i.e., the ErgunaR and the ShilkaR) and relatively lower in the eastern regions (i.e., JieyaR and the BreaR). This discrepancy is likely related to the climatic conditions, topographic features, and hydrological environments of each sub-basin.

4.1.2. Time Distribution of DFAA

To understand the trends of DFAAEs’ overextended time scales and identify the diverse characteristics of seasonal droughts and floods, we analyzed the occurrences of DFAA events in the Heilongjiang River Basin over the course of 5 decades (1970–1979, 1980–1989, 1990–1999, 2000–2009, and 2010–2019) and across four different seasons (spring, summer, autumn, and winter).

According to the seasonal distribution shown in Figure 4, clear seasonal differences were observed in the spatial distribution characteristics of DFAAs and their associated events (FTD and DTF) in the Heilongjiang River Basin. Spring and summer were the peak periods for DFAA events, which were characterized by broad spatial distributions and extensive influence ranges. In contrast, the frequencies of such events were relatively low in autumn and winter and were accompanied by a reduction in spatial coverage. The spatial coverage of FTD events in spring, summer, autumn, and winter were 52.88%, 48.77%, 40.94%, and 0.05%, respectively. The spatial coverage of DTF events were 39.29%, 26.28%, 20.58%, and 0.16%. In spring, the regions with high DFAAI values were concentrated in the middle and southern parts of the basin, particularly in the SHR and the HLJRMS. This indicates that DFAAs occurred more frequently there. FTD events were predominantly distributed in the southern region of the SHR and the northern region of the ErgunaR, whereas DTF events were more concentrated in the central region of the SHR and the ShilkaR. During summer, northern ErgunaR and JieyaR had high DFAAI values. FTD events occurred more frequently in the northern and eastern basins covering the ErgunaR and UssuriR. Meanwhile, the DTF events were concentrated in the southern part of the SHR and the UssuriR. In autumn, the high-value area of DFAAI was primarily located in the southern SHR and the UssuriR. However, the high-frequency area of the FTD events shrank to the middle region of the river basin. The DTF events remained concentrated in certain sections of the SHR and the UssuriR, yet their frequencies decreased significantly, compared to those in spring and summer. In winter, the spatial distribution range of DFAAI was the narrowest, and the high-value area was confined to the southern part of the SHR. The FTD events were scattered across a few areas within the basin, whereas the DTF events rarely occurred.

Generally, the spatial coverage of DFAA events was broadest in spring and summer. In autumn, the events were mainly concentrated in the southern region. In winter, their frequency reached the lowest point accompanied by a significant reduction in spatial range. Moreover, FTD events were more prevalent in the northern and central basins, whereas DTF events were usually observed in the southern and eastern basins. These seasonal distribution differences might be caused by the seasonal variations in precipitation, snowmelt, the topographic features of the watershed, and human activities.

As depicted in Figure 5, during the period from 1970 to 1979, DFAAs were predominantly concentrated in the northern and central regions of the Heilongjiang River Basin, particularly in the ErgunaR, ShilkaR, and SHR. In these regions, the DFAA frequency was relatively high. As time progressed, these events gradually expanded towards the eastern part of the basin, with regions like the UssuriR and AmgonR also exhibiting a relatively higher mutation frequency. The general trend indicated that the spatial distribution range of DFAA events gradually expanded, while the mutation frequency in the southern region significantly increased. The temporal evolution of the high-value area showed that, in the 1970s, the high-frequency area was concentrated in the northern region and gradually expanded to cover the entire basin.

The spatial coverage of FTD events increased from 90.67% (1970–1979) to 93.69% (2010–2019). The highest frequencies in the periods of 1970–1979, 1980–1989, 1990–1999, 2000–2009, and 2010–2019 were 20, 17, 22, 20, and 23, respectively. During the period from 1970 to 1979, the FTD events were predominantly concentrated in the central and northern regions of the SHR and the ErgunaR, forming a relatively concentrated high-frequency area. From 1980 to 1999, the spatial range of the FTD events expanded significantly, with the high-frequency area gradually encompassing the ShilkaR and a part of the HLJRMS. From 2000 to 2019, the FTD events exhibited an overall downward trend, with high-frequency areas diminishing but remaining concentrated in the northern and central basins.

The spatial coverage of DTF events declined from 55.7% (1970–1979) to 40.62% (2010–2019). The highest frequencies registered during 1970–1979, 1980–1989, 1990–1999, 2000–2009, and 2010–2019 were 27, 29, 27, 27, and 33, respectively. During the period from 1970 to 1979, DTF events occurred on a relatively limited scale and were concentrated in the southern and central regions of the basin. Over time, the area with high values gradually expanded and extended towards the north, encompassing regions like the ErgunaR and the JieyaR. From 2000 to 2019, the spatial distribution of the DTF events became more expansive, and the high-frequency areas increased substantially (specifically within the SHR and the UssuriR).

The spatial distribution of DFAAs expanded gradually. Once concentrated in the northern region, they eventually covered the entire basin. The DTF events showed a more pronounced increase in the southern and eastern regions, whereas the FTD events experienced a gradual decline in the high-frequency northern areas. The DFAA events within the Heilongjiang River Basin exhibited significant temporal variations, with these changes likely due to climate warming in the basin, alterations in the spatiotemporal distribution of precipitation, and human activities.

4.2. Analysis of NDVI Evolution Law

4.2.1. Annual Variation

A significant positive correlation existed between the NDVI, ecosystem parameters, biomass, and vegetation growth within the Heilongjiang River Basin. Specifically, a high NDVI value corresponded to greater vegetation coverage and better growth conditions. The multi-year average NDVI data from January to December (2000 to 2019) revealed both the annual NDVI variations and the temporal dynamic characteristics of vegetation growth in the basin. As shown in Figure 6, the vegetation coverage within the basin exhibited an initial increase and then a decrease over the course of a year. The coverage was greatest in July and lowest in January. This pattern was primarily influenced by climate shifts and moderate precipitation. During winter, the NDVI values were relatively low, approaching 0 in most areas. This was especially true in the northern regions. These results reflected sparse vegetation and/or low activity levels, which align with the characteristics of cold weather, snow cover, and vegetation dormancy. In spring, the NDVI values gradually rose, and vegetation recovered in the southern and central regions. In summer, the NDVI values reached their annual peak (close to 1 in most areas). Extensive vegetation covered the SHR, the UssuriR, and the southern plain areas. Meanwhile, NDVI increased in the ErgunaR and the HLJRMS in the north. In autumn, the NDVI values gradually declined, vegetation activity waned, and some areas entered a state of dormancy or decay. NDVI’s spatial distribution exhibited significant differences between the northern and southern regions. Generally, the NDVI values in the southern areas (the SHR and the UssuriR) were higher throughout the year, likely due to the warm, humid climate conditions. In contrast, in the northern regions (the ErgunaR), the vegetation growth and the growing season were constrained by the high altitudes and cold climate. Along the Heilongjiang River’s main stream area and its main tributaries, the NDVI values were relatively higher in spring and summer, and the vegetation thrived under the water supply. In summer, the areas with high NDVI values were concentrated in the southern and plain regions. These areas generally shifted to low values in winter, indicating that seasonal variations significantly influenced vegetation activity. Overall, the distribution of NDVI within the Heilongjiang River Basin exhibited significant seasonal and spatial differences. Vegetation growth was active, with extensive coverage in spring and summer. Moreover, the NDVI values in the southern regions were significantly higher than those in the northern regions, and in winter, vegetation growth was largely stagnated. This was likely due to the climatic, topographic, and hydrological conditions of the basin.

4.2.2. Interannual Variation

This study used monthly NDVI data to calculate a multi-year average NDVI dataset to examine the spatial patterns of vegetation growth within the Heilongjiang River Basin. As indicated by the legend, a darker color represents a higher NDVI value, signifying greater vegetation coverage and better growth conditions. Conversely, lighter colors signify lower vegetation vitality, which could correspond to bare land, construction land, or lakes. As illustrated in Figure 7, the multi-year average of NDVI from 2000 to 2019 reveals generally high NDVI values across the entire Heilongjiang River Basin, particularly in the central and southern parts of the basin. In these areas, the NDVI values are close to 1, suggesting high vegetation cover density and lush growth. In contrast, the NDVI values in certain cold, high-altitude areas in the northern and western parts of the basin were relatively low. In some regions, they were close to 0, indicating sparse vegetation coverage, possibly due to the environmentally restrictive terrain, lakes, and vegetation.

During the period from 2000 to 2009, the overall NDVI values within the basin were lower compared to those of the subsequent decade. Specifically, the area with low NDVI values in the northern region was relatively larger, suggesting that vegetation growth was somewhat restricted. This situation might be associated with the climatic conditions, regional economic development activities, or land use changes that occurred in the late 20th and early 21st centuries. During the period from 2010 to 2019, the NDVI values increased significantly. The high-value areas in the middle of the basin and in the plain areas expanded further, specifically into the Songnen Plain and Sanjiang Plain regions. Meanwhile, the area with low NDVI values decreased, indicating that vegetation growth was enhanced. This improvement could stem from the ecological restoration policies implemented by the state (i.e., returning farmland to forest, wetland protection, and the amelioration of climate conditions). Comparing the NDVI distribution from 2000 to 2009 with that from 2010 to 2019, the average NDVI increase in the basin during the 2010–2019 period was 0.54%. Moreover, the area of vegetation growth improvement accounted for 17.29% of the total basin area, whereas the area of degradation accounted for only 3.05%. The area of regions with NDVI values > 0.8 increased, while the area of low-value regions with NDVI values < 0.3 decreased significantly.

In summary, the vegetation growth within the Heilongjiang River Basin exhibited an overall upward trend over the past 20 years. During the period from 2010 to 2019, the NDVI values experienced a significant increase, accompanied by an overall shrinkage in low-value areas. This alteration reflects both the amelioration of climate conditions (i.e., the extension of the vegetation growing season brought about by climate warming) and the positive impacts of national ecological protection measures (i.e., policies like the conversion of farmland back to forest and wetland protection) on vegetation restoration. Nevertheless, the high altitude, cold climate, and complex terrain in the northern region resulted in relatively slight improvements in vegetation growth.

4.3. Response of NDVI to DFAAE

To explore the specific DFAAI characteristics influencing the NDVI, we analyzed the time-frequency relationship between the DFAAI and NDVI using Cross Wavelet Transform and Wavelet Coherence analysis. The univariate wavelet spectra demonstrate the periodic characteristics of the DFAAI and NDVI. The Wavelet Coherence analysis revealed the correlation between the two time series within the frequency domain. Cross wavelet analysis isolated the common wave pattern shared by the two.

As seen in Figure 8, the DFAAI and NDVI in the Heilongjiang River Basin exhibited a significant correlation in spring and summer, particularly within the 8–16-year cycle. In the DFAAI figure, the regions with high energy values were concentrated in the 8–16-year range, specifically at the positions ranging from 40 to 80 on the time axis. This demonstrates strong cyclical fluctuations and contrasts with the NDVI figure, where the high energy values were concentrated within the 8–12-year cycle range. This suggests that the vegetation cover index experiences significant volatility during this cycle. Wavelet Coherence analysis revealed that the time-frequency coherence between the DFAAI and NDVI was primarily concentrated within the mesoscale period range of 8–12 years, specifically at the positions ranging from 40 to 80. Moreover, the coherence and common fluctuations of the two were the most pronounced during this range. During this period, the arrows predominantly pointed toward the lower right, indicating that the DFAAI slightly preceded NDVI, implying that DFAAs might have a lag effect on vegetation growth. The significant coherent regions in summer and autumn remained concentrated within the mesoscale period range of 8–12 years, further validating the delayed response relationship between drought–flood events and vegetation growth. Nevertheless, in local areas, the arrows pointed towards the upper right, which could indicate the recovery characteristics of vegetation in response to drought–flood events. The coherence at different periodic scales exhibited a certain phase difference, suggesting that the interaction between the two is complex and nonlinear. Additionally, the DFAAI also exhibited a significant correlation with the NDVI within the 30–40 period. This could be due to extreme climate events like summer and autumn floods. Overall, this correlation implies that drought and flood events induced by climate change will likely significantly impact vegetation growth within the Heilongjiang River Basin, particularly in the medium- and long-term cycles. Consequently, the sensitivity and response mechanism of vegetation to climate change warrant further attention and research.

In the spring–summer seasons (Figure 9), the red and blue regions in the figure were intertwined. They displayed different color intensities, and the overall color tone was relatively light, suggesting that the correlation in most regions had not fully manifested. After a 1–2-month lag, the dark colored area gradually expanded, the distribution of red and blue became more distinct, and the area of significant correlation broadened considerably. The dark blue area expanded and gradually became dominant. After a 2-month lag, the spatial proportion of the blue area was 6.49% greater than that of the red area, signifying that the adverse impacts of drought or flood had accumulated to a peak, potentially leading to a large-scale decline and the destruction of vegetation. Particularly, in flood-prone regions, the lag effect was more pronounced, indicating that the flood damage to vegetation had further accumulated. Meanwhile, the red area gradually expanded, and the dark red area became more concentrated, showing that the promoting effect of appropriate precipitation on vegetation was emerging. After a 3-month lag, the expansion and intensification of the deep red area indicated a maximum positive effect, implying that the recovery response of vegetation to water conditions in some areas required time to activate. The dark blue area remained significant, and the negative effects of droughts and floods persisted in some areas. After a 4-month lag, the correlation between the drought–flood mutation index and the vegetation cover index weakened to the point of being invalid. This could indicate that the lag effect had already passed, or that the vegetation change during this period was primarily driven by other factors, such as temperature and light. Overall, the significant correlation with a 3–4-month lag during the spring–summer period was the strongest and most extensive, suggesting that the lag effect of drought–flood fluctuations on vegetation had reached its peak.

In the summer–autumn season (Figure 10), the correlation during the period was rather weak, and the colors of most regions were relatively light, suggesting that the overall correlation distribution was not pronounced. After a 1–2-month lag, the red area gradually expanded, signifying that the lagged effect of summer precipitation exerted a clear, positive effect on vegetation growth. This could be due to the precipitation accumulation improving the soil moisture or the positive response of vegetation to precipitation as it enters the vigorous growth stage. The blue areas on the diagram became darker, particularly in areas affected by extreme precipitation (i.e., flooding). This indicates more substantial damage to vegetation or cumulative damage over an extended period by extreme drought and flood conditions. At this juncture, the overall correlation distribution was significantly strengthened, and the lag effect reached its peak. Following a lag of 3 months, the blue region further expanded, occupying 1.21% more area than the red region. The negative correlation became dominant, while the positive correlation weakened, suggesting that the cumulative effect of droughts and flooding on vegetation destruction was approaching its peak, particularly in ecologically sensitive areas. Nevertheless, the persistent effects of precipitation continued to have a positive influence on vegetation in certain regions. After a 4-month lag, the overall color gradually lightened, and the distribution of red and blue became more uniform. Theoretically, the restoration mechanism of vegetation could gradually take effect, diminishing the negative impacts of droughts and floods while reducing the positive effects of precipitation. Consequently, the lag effect may dissipate, and the correlation could return to a lower level.

Within a lag of 4 months, the correlation increased gradually, and the dark-colored area expanded continuously, indicating a significant lag effect from droughts and floods on vegetation. In the spring–summer period, both the positive and negative correlations reached their peaks during a lag of 3–4 months, suggesting that this specific time window was when droughts and floods significantly disrupted vegetation. In contrast, in the summer–autumn period, the correlation reached its peak during the 3-month lag. Moreover, the lag effect was more pronounced in the summer–autumn period than in the spring–summer period, and the spatial range, as well as the intensity of both the positive and negative correlations, were more conspicuous. The distinct color distribution reflects the differences in the response mechanisms to droughts and floods in different regions within the basin. For instance, dark red areas might correspond to wet environments like forests, whereas dark blue areas could correspond to more vulnerable grasslands or cultivated lands. The delayed response of DFAAs to vegetation presents obvious temporal and spatial heterogeneity. This suggests that ecological management should focus on the cumulative effects over time and develop tailored recovery strategies based on regional characteristics.

5. Discussion

5.1. Direct Effects of DFAAE on Vegetation Growth

Moderate flood events can briefly augment the regional soil water and nutrient supply, particularly in wetland ecosystems or for plants tolerant to waterlogging. When floodwaters recede, floods can introduce a substantial amount of sediment and nutrients into the soil, thereby facilitating the rapid recovery and expansion of vegetation [1,35]. Nevertheless, persistent flooding can result in a reduction in soil oxygen, root decay, and the significant degradation of water-intolerant vegetation [36]. Drought events decrease vegetation cover by reducing soil moisture and restricting plant transpiration and photosynthesis. Prolonged drought can completely kill vegetation and even alter affected vegetation types [37]. Our study demonstrates that there might be a positive or negative correlation between DFAA events and vegetation growth. This implies that the interaction between the two could have synergistic or antagonistic effects. Other studies have corroborated this finding, positing that in certain ecosystems, moderate droughts and floods might promote vegetation diversity and maintain ecosystem balance. For example, during dry periods, the number of non-drought-tolerant plants decreases, thereby providing more space for drought-tolerant plants to thrive. Meanwhile, during flood periods, water is supplied for other types of vegetation [1,38,39,40,41,42]. Thus, certain vegetation types can demonstrate their adaptability in long-term DFAAs, allowing vegetation growth to fluctuate periodically yet remain generally stable [43]. However, the combination of droughts and floods can also pose challenges for vegetation in terms of adaptation. Certain types of vegetation might not have time to recover after a drought, and subsequent flooding could further damage the soil and roots. There could also be significant competition among different types of vegetation in response to DFAAs. For example, drought-tolerant plants can be damaged during flood events, while water-tolerant plants may be degraded during drought events. Such competition can result in fluctuations in vegetation growth and serious regional degradation [44].

The impacts of DFAAs on vegetation growth varied significantly across different regions and ecosystems. The positive effect was more pronounced in humid regions, while the negative effect prevailed in more arid regions [45]. Meanwhile, climate warming likely increased the frequency and intensity of DFAAs, thereby introducing greater uncertainty into the dynamic changes of vegetation growth. Increasing research on the ability of vegetation to adapt to DFAAs is crucial, particularly in areas that are prone to frequent extreme events. In areas that experience severe vegetation degradation, adaptive management measures should be adopted in accordance with the characteristics of DFAAs. For instance, drought-tolerant or water-tolerant plants should be planted to enhance the resilience of the ecosystems.

5.2. Feedback Mechanism of Vegetation Growth Change to DFAAs

Based on the WTC and XWT figures presented in Section 4.3 of this study, most of the arrows within the significant region of the 8–16-year period were positioned either in the upper right or lower right. This suggests that the change in DFAA plays a significant role in influencing vegetation growth over a specific period of time. Nevertheless, in the lower period of 4–8 years, the arrows were positioned in the upper left or lower left, indicating that vegetation growth might exert a certain feedback effect on DFAAs. During droughts, reduced NDVI may result in soil drying and a decrease in evaporation, which, in turn, indirectly impacts the performance of DFAAs. Changes in vegetation growth can impact regional hydrological conditions by affecting soil moisture and local climate regulation, which in turn can provide feedback to DFAA. Similar situations have also been observed in previous studies. Specifically, an increase in vegetation cover can enhance the efficiency of the local water cycle via evapotranspiration and precipitation recycling. The leaf interception and root water absorption functions of vegetation help stabilize soil moisture and reduce the intensity of droughts and floods [46,47]. Furthermore, the hydrological regulatory effects of vegetation (i.e., rainwater retention, soil stabilization, and enhanced surface infiltration) are capable of significantly reducing surface runoff and soil erosion. This decreases the intensity of flooding [48]. A reduction in vegetation cover weakens evapotranspiration and diminishes the efficiency of local precipitation recycling, elevating the risk of drought. The reduction in surface vegetation can result in the deterioration of soil structure and a decrease in permeability, facilitating the formation of surface runoff and flooding. During the dry season, the water-holding capacity of the soil also weakens, further aggravating the risk of flooding [49].

Vegetation restoration measures, including the reversion of farmland back to forest and wetland restoration, can significantly improve water retention and regional climate regulation at the ecosystem level and reduce the frequency and intensity of DFAAs. For example, China’s Three-North Shelterbelt Project, by means of large-scale vegetation restoration, strengthened the stability of the regional water cycle and mitigated the amplifying effect of climate change on DFAAs [50]. An increase in vegetation diversity can also enhance ecosystems’ abilities to endure extreme climate events. Under DFFAs, highly diverse ecosystems are capable of achieving functional maintenance through the species complementarity effect [51].

Although the conversion of forests into farmland might boost food production in the short term, it will undermine regional hydrological regulation. For instance, farmland is less capable of withstanding flooding and less resilient in the face of drought compared to natural vegetation [52]. Furthermore, feedback conflicts might arise among different types of vegetation, due to water competition. As an example, deep-rooted trees could impact the recovery of herbaceous vegetation by over-consuming groundwater resources. In the context of farmland conversion projects in northern India, certain revegetation projects have decreased the risk of flooding while augmenting the risk of regional drought [53].

Through climate warming, the feedback mechanism through which vegetation growth changes impact DFAAs has intensified. High temperatures accelerate transpiration and exacerbate drought effects, while extreme precipitation events heighten soil erosion and flood risks. Meanwhile, accounting for changes in atmospheric circulation and surface albedo, the feedback of vegetation growth on droughts and floods could extend from the local level to regional or even global scales, amplifying its overall impact.

5.3. Uncertainty of CWatM Simulation

The uncertainty in CWatM simulations is a significant issue in the research and application of hydrological modeling. It primarily stems from the complexity and ambiguity associated with numerous factors. Firstly, the input of vague or faulty data directly influences the model’s results [54]. For instance, discrepancies in spatial resolution and accuracy among different data sources, including observation, reanalysis, and remote sensing data, could give rise to significantly different simulation results. Errors in the input data (i.e., precipitation and temperature) will directly impact the simulation of evapotranspiration, runoff, and water resource supply and demand. Moreover, data projections related to land use change, population growth, and economic activity frequently contain errors that can result in inaccurate estimates of irrigation demand and water use [55]. Secondly, uncertainty surrounding the model’s structure can also influence the simulation accuracy. During the simulation, CWatM simplifies several complex hydrological and management processes, such as the interactions between groundwater and surface water and the rapid response to extreme events. Such simplification could lead to certain regions or specific events being inaccurately represented [56]. Simultaneously, the parameters within the model, namely soil moisture and evapotranspiration coefficient, rely on empirical formulas or regional calibration and could result in substantial simulation errors in uncalibrated regions [57].

Apart from the uncertainties associated with the data and model structure, the uncertainty surrounding scenario assumptions also presents challenges to the reliability of the results [58]. Predictions of future climate conditions vary considerably among climate models like CMIP6 and emission scenarios like the RCP or SSP pathways. Socioeconomic scenarios including population growth, policy changes, and technological advances are equally challenging to accurately predict. This increases the uncertainty of future water supply and demand assessments [59,60,61]. Meanwhile, the uncertainties related to spatial and temporal scales also warrant special attention. Due to the complexity of topography, land use, and water resources distribution across different regions, low-resolution simulations will likely be unable to capture local characteristics. Furthermore, simulations conducted over long time scales, such as those involving interannual variations, may fail to accurately depict the short-term dynamics of extreme events [62].

In light of the aforementioned insecurities, the following measures can be adopted to mitigate the errors: (1) Integrate diverse meteorological data and land use data, utilize remote sensing data to correct crucial input parameters like precipitation and evapotranspiration, and conduct a comprehensive evaluation of the impact of data errors on model results [63]. Local parameter calibration should be conducted for different regions. For example, the evapotranspiration and discharge data of the Heilongjiang River Basin were utilized to calibrate the model in this study, with the aim of enhancing the accuracy of the simulation. (2) Multi-model ensemble analysis should be carried out jointly with other global hydrological models to diminish the uncertainty resulting from a single-model structure. (3) Simulate multiple climate and socioeconomic scenarios and quantify the uncertainty of projections by comparing the ranges of outcomes among different scenarios. (4) Optimize the model grid resolution and utilize high time resolution data to conduct finer-grained simulations of complex regions. These methods can effectively mitigate the errors induced by inconsistencies in CWatM simulations.

Notwithstanding certain factors, the regional and global applications of CWatM demonstrate that its integrated and versatile architecture can support multi-scenario and multi-scale analyses. With improvements in the input data quality, the optimization of model structure, and the advancements in uncertainty quantification methods, CWatM holds significant potential for widespread application in water resources management and climate change.

6. Conclusions

In this study, the spatiotemporal distribution of drought–flood abrupt alternation events within the Heilongjiang River Basin was analyzed. The dynamic characteristics of vegetation growth were investigated through the utilization of NDVI, and the response mechanism of vegetation growth to drought–flood abrupt alternation events was identified. The following conclusions were drawn:

(1)

From 1970 to 2019, the DTF events were most frequent in the SHR and least frequent in the AmgonR, with an upward trend (3.51%/decade). The FTD events decreased (4.52%/decade) over time, were more frequent in northern regions like the ErgunaR and the ShilkaR, and less frequent in eastern areas like the JieyaR and the BreaR. Temporally, events peaked in spring and summer, were concentrated in the south in autumn, and were least frequent in winter. High-frequency areas expanded from the north in the 1970s–1980s into the entire basin. From 2010 to 2019, the DTF events were concentrated in the south and east, while the FTD events showed a decline in the northern high-frequency areas.

(2)

Annual NDVI values were lowest in January, highest in July, and close to 0 in winter. Vegetation coverage was greatest in summer. Southern areas like SHR and the UssuriR had relatively higher NDVI values all year. Northern regions, like ErgunaR, were more climate restricted. From 2000 to 2019, vegetation growth markedly increased. The NDVI values were lower from 2000 to 2009, especially in the northern and central high altitudes, but rose significantly from 2010 to 2019, with a reduction in low-value areas and an expansion of high-value areas.

(3)

Wavelet Coherence analysis revealed that drought and flood events have a 1–4-month lag effect on vegetation. The positive and negative correlations were most significant in the spring–summer period with a 3–4-month lag. In the summer–autumn period, the peak lagged by 3 months.

In conclusion, significant spatiotemporal heterogeneity exists in both drought–flood abrupt alternation events and vegetation growth within the Heilongjiang River Basin. The delayed response of vegetation to DFAAs implies the necessity of strengthening ecological restoration strategies to address the impact of climate change on ecosystems.