Next Article in Journal
Lightweight Design for Infrared Dim and Small Target Detection in Complex Environments
Next Article in Special Issue
Changes in Urban Green Spaces in the Pearl River Delta Urban Agglomeration: From the Perspectives of the Area, Spatial Configuration, and Quality
Previous Article in Journal
MPG-Net: A Semantic Segmentation Model for Extracting Aquaculture Ponds in Coastal Areas from Sentinel-2 MSI and Planet SuperDove Images
Previous Article in Special Issue
Fine-Scale Mangrove Species Classification Based on UAV Multispectral and Hyperspectral Remote Sensing Using Machine Learning
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 16
Issue 20
10.3390/rs16203762
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Evaluating the Multidimensional Stability of Regional Ecosystems Using the LandTrendr Algorithm

by
Lijuan Li
1,2,
Jiaqiang Du
1,2,*,
Jin Wu
3,
Zhilu Sheng
1,2,
Xiaoqian Zhu
1,2,
Zebang Song
1,2,
Guangqing Zhai
1,2 and
Fangfang Chong
1,2
1
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Science, Beijing 100012, China
2
State Environmental Protection Key Laboratory of Regional Eco-Process and Function Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China
3
Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2024, 16(20), 3762; https://doi.org/10.3390/rs16203762
Submission received: 14 July 2024 / Revised: 7 October 2024 / Accepted: 8 October 2024 / Published: 10 October 2024
(This article belongs to the Special Issue Remote Sensing Applications in Land Cover Changes and Associated Environmental Effects: Progress, Challenges, and Opportunities)
Browse Figures
Versions Notes

Abstract

:
Stability is a key characteristic for understanding ecosystem processes and evolution. However, research on the stability of complex ecosystems often faces limitations, such as reliance on single parameters and insufficient representation of continuous changes. This study developed a multidimensional stability assessment system for regional ecosystems based on disturbances. Focusing on the lower reaches of the Yellow River Basin (LR-YRB), we integrated the remote sensing ecological index (RSEI) with texture structural parameters, and applied the Landsat-based detection of trends in disturbance and recovery (LandTrendr) algorithm to analyze the continuous changes in disturbances and recovery from 1986 to 2021, facilitating the quantification and evaluation of resistance, resilience, and temporal stability. The results showed that 72.27% of the pixels experienced 1–9 disturbances, indicating the region’s sensitivity to external factors. The maximum disturbances primarily lasted 2–3 years, with resistance and resilience displaying inverse spatial patterns. Over the 35-year period, 61.01% of the pixels exhibited moderate temporal stability. Approximately 59.83% of the pixels recovered or improved upon returning to pre-disturbance conditions after maximum disturbances, suggesting a strong recovery capability. The correlation among stability dimensions was low and influenced by disturbance intensity, underscoring the necessity for a multidimensional assessment of regional ecosystem stability based on satellite remote sensing.

1. Introduction

Ecosystem stability refers to the capacity of a system to maintain and restore its initial state when subjected to external disturbances, without exceeding the system’s threshold. Stability is influenced by the ecosystem’s structure, function, evolutionary traits, and the intensity and characteristics of the disturbance. With the ongoing impacts of climate change, frequent extreme events, and the effects of large-scale, high-intensity human development and utilization activities, ecosystems are increasingly challenged in their ability to resist external disturbances and maintain normal functions and stability. When the disturbance exceeds the adaptive limits of an ecosystem, a previously dynamically balanced ecosystem may change its stable state [1], exhibiting irreversible damage or destruction [2,3], which impacts the sustainable development of regional ecosystems.
Ecosystem stability encompasses multiple dimensions [4,5], of which typical examples include resistance, resilience, and temporal stability. These dimensions are not completely independent, and the proposal of a multidimensional stability framework [5] has significant implications for standardizing the concept of stability and predicting regime shifts. However, existing research often only focuses on a single dimension of stability, which may hinder the overall understanding of system stability. Studies that have addressed various dimensions typically focused on plot scales or specific ecosystems, and research into the multidimensional stability of regional composite ecosystems is still in the exploratory phase.
Currently, stability assessment methods mainly include selecting key characteristics such as the number and area of indicator species, dominant species, or environmentally sensitive species of typical ecosystems to construct statistical parameters like the coefficient of variation, or mathematical analysis methods studying the uniqueness of model equilibrium solutions [6,7]; landscape ecology methods revealing the stable state of ecosystems through landscape indices reflecting landscape spatial heterogeneity and patterns [8,9]; comprehensive evaluation methods constructing a relatively complete index system and assessment framework using mathematical models [10,11]; and methods based on ideal benchmark and variation evaluation [12]. These methods can determine the current state of ecosystem stability and changes over a certain period, but for global-scale research, they are time-consuming and labor-intensive, and lack the ability to accurately detect continuous changes and the trajectories of various disturbances and recoveries in the system, making it difficult to quantify the multidimensional stability of ecosystems under disturbance.
Disturbance detection algorithms based on long time series remote sensing data are increasingly gaining attention due to the advantage of containing continuous temporal information, and they are uniquely applied in simultaneously detecting sudden, chronic, and persistent disturbances, with consistency across regions and time, playing an important role in the research of stability responses to disturbance and recovery processes. A series of algorithms, such as Landsat-based detection of trends in disturbance and recovery (LandTrendr) [6,7,13,14], continuous change detection and classification (CCDC) [15,16], vegetation change tracker (VCT) [17,18], breaks for additive season and trend (BFAST) [19,20], and dynamic time warping (DTW) [21], have been used to quantify the disturbed conditions of specific ecosystems. The LandTrendr algorithm has been widely used in forest disturbance and recovery detection, habitat monitoring, disturbance and restoration of marsh vegetation, and land use change detection [22,23,24,25]. Current applications of algorithms like LandTrendr for disturbance detection primarily rely on spectral reflectance bands, simple vegetation indices, or single-function indicators such as vegetation productivity or biomass [6,7]. These approaches focus on detecting disturbances and recovery within specific ecosystems, with limited expansion to regional composite ecosystems. Furthermore, research utilizing the LandTrendr algorithm to explore stability characteristics, such as ecosystem resistance and resilience, remains insufficient.
The structure of an ecosystem is the basis and foundation for its function, and changes in landscape structure will ultimately affect the provision of services [26]. Ecosystem quality reflects the ecological characteristics of the state and evolutionary laws of ecosystem elements, processes, and functions within a specific spatiotemporal range determined by natural resources and environmental conditions [27]. Coupling ecosystem structure and quality for the stability assessment research of complex ecosystems can more comprehensively grasp system characteristics, which is the current trend in regional ecosystem stability theoretical research and comprehensive application.
The lower reaches of the Yellow River basin (LR-YRB) are key areas for China’s modernization and an essential part of its ecological security. They are the front line for achieving the goals of the “Major National Strategy for the Yellow River” but are also typical ecologically fragile and climate-sensitive areas. Studying the multidimensional stability of ecosystems under disturbance is essential for understanding the historical evolution and development patterns of ecosystems in the LR-YRB. This can guide the selection of priority areas for ecological protection and restoration, as well as inform ecosystem management strategies to address future climate changes and human disturbances. Furthermore, this research is critical for promoting the high-quality development of ecosystems throughout the entire river basin.

2. Materials and Methods

2.1. Study Area

The Yellow River Basin is located between 96° to 119° east longitude and 32° to 42° north latitude, with the LR-YRB starting at Taohuayu in Henan Province. The LR-YRB flows through Henan and Shandong provinces, situated in a temperate monsoon climate zone with a semi-humid climate. This region encompasses a complex and dynamic ecosystem, including the Central China Loess Plateau Mixed Forests, Huang He Plain Mixed Forests, and Bohai Sea Saline Meadow ecoregions. The main land use/landscape types include farmland, construction land, forests, water bodies, and grasslands (Figure 1). Rapid socio-economic development and interruptions in the Yellow River’s flow have severely hindered the natural development of wetlands, leading to significant degradation, a sharp reduction in wetland area, and challenges in maintaining ecosystem stability, integrity, and biodiversity. Additionally, high sediment loads and seasonal climate variations often lead to floods and droughts in the LR-YRB, posing significant challenges to the ecosystems.

2.2. Data

(1) Satellite Data: We utilized Landsat TM/ETM+/OLI data covering the study area from 1986 to 2021. Surface reflectance (SR) datasets were selected, and processing was conducted on the Google Earth Engine (GEE) cloud computing platform [28]. To reduce the impact of vegetation phenological differences on spectral recognition, images from the growing season (1 June to 30 September) were screened and applied. To minimize noise caused by phenology or solar elevation angle changes, median compositing of reflectance data was performed annually [29], and cloud and shadow masking was applied. Linear interpolation was used to fill in the null values in the composite images, creating an image dataset with minimal cloud cover. In addition, spectral corrections were made to reduce spectral differences between sensors [30].
(2) Verification event data: The data of impervious water surfaces were selected from the 1986–2022 dataset independently developed by our team [31]. Verification datasets for fire, flood, and drought events were sourced from GEE. Detailed information on the verification datasets can be found in Table S1.
(3) Other auxiliary Data: We chose the annual China Land Cover Dataset (CLCD) as the land use/landscape type data for this study [32], and the data of 2020 shown in Figure 1 as the basic data of classification statistics for this study. Yellow River Basin water body data [33] and JRC Yearly Water Classification History data [34] were taken into account when calculating RSEI year by year. In addition, basic geographic data and DEM were also collected.

2.3. Methods

We first constructed composite ecosystem metrics by integrating structural and quality parameters from remote sensing data. Next, we applied the LandTrendr algorithm to capture disturbance–recovery processes and extract relevant characteristic data. Using these results, we developed a multidimensional stability assessment framework focused on disturbances, calculating the stability for each pixel and time period in the LR-YRB. Finally, we explored the relationships between the different dimensions of ecosystem stability. The specific methods were as follows:

2.3.1. Stability Measurement Parameters

The remote-sensing-based ecological index (RSEI) was chosen to reflect the quality of the ecosystem. The RSEI uses principal component analysis to couple four categories of indicators: vegetation index, humidity index, surface temperature, and a bare soil and construction index (Table S2), representing the four important ecological elements of greenness, humidity, heat, and dryness, respectively [35], with the first principal component used to create a comprehensive ecological index. The recently proposed kernel normalized difference vegetation Index (kNDVI) is considered a robust alternative indicator of ecosystem productivity, capable of effectively addressing the saturation and mixed pixel issues encountered by traditional vegetation indices [36]. In this study, the kNDVI was used as the greenness indicator. To avoid interference in the humidity index from large water bodies, water body datasets were used to mask large permanent water bodies before calculating the RSEI.
We also utilized the kNDVI as input for gray-scale images, and selected angular second moment (ASM), entropy (ENT), and correlation (COR) as three representative and indicative texture feature parameters of ecosystem disturbances. Using a gray-level co-occurrence matrix (GLCM) [37], we calculated the texture index with a 5 × 5 pixel sliding window and synthesized the three parameters using the entropy weighting method, representing them as spatial structure parameters.
Finally, the above RSEI and texture features were averaged and weighted to serve as the comprehensive disturbance and stability metrics in this study.

2.3.2. Disturbance and Recovery Detection

The LandTrendr algorithm was selected for disturbance detection. The idea of this algorithm is to extract the surface spectral change trajectory from Landsat time series data and simplify complex change features into connected straight-line segments, thereby capturing disturbance characteristics and obtaining information on disturbance start and end years, disturbance magnitude, and recovery. This method only considers the trend component of time series data, enabling it to capture short-term changes and smooth long-term trends, eliminating noise without losing necessary details [38]. The algorithm can offer significant advantages in disturbance detection and provides a strong foundation for multidimensional stability assessments.
This algorithm can identify spectral trajectories that show a continuous increase or decrease and segment them into a series of breakpoints and fitted values. For each pixel, a disturbance breakpoint is defined as a decrease in the metric parameters, and a recovery breakpoint is defined as an increase in the metric parameters. The disturbance/recovery magnitude in the segmentation is defined as the calculated difference in the metric parameter value between the starting and ending breakpoints of the disturbance/recovery. The algorithm was implemented using the operator “ee.Algorithms.TemporalSegmentation.LandTrendr” in GEE, which organizes input parameters into a time series of the ImageCollection type. Based on existing literature [38] and the specific requirements of this study, the LandTrendr parameters utilized were as summarized in Table 1.

2.3.3. Multidimensional Stability Assessment

Referring to the general meaning of stability, ecosystem stability was defined as the ability of an ecosystem to maintain or recover to its initial state before disturbance, to maintain structural optimization and quality improvements.
(1) Multidimensional Stability Indicators
Based on the basic characteristics of ecosystem stability, ecosystem structure and quality parameters were obtained using dense time series remote sensing images, and key parameters of disturbance and recovery were obtained pixel by pixel based on the Landtrendr time series analysis method. A complete disturbance cycle was considered from the start of the disturbance to the end of the immediate recovery. Accordingly, multidimensional feature indices were designed to reflect the stable condition of the basin’s ecosystem (Figure 2). The specific multidimensional stability indicators were as shown in Table 2.
(2) Grading of Stability Indicators
Resistance was classified into five categories based on disturbance level thresholds (0.8, 0.6, 0.4, 0.2). Similarly, resilience and temporal stability were also divided into five categories, using approximately equal intervals based on the data distribution, to ensure consistent classification outcomes. The corresponding threshold values for each level were as shown in Table 3.

2.3.4. Assessment of Relationships among Various Stability Dimensions

The Spearman rank correlation coefficient is a statistical analysis indicator that reflects the magnitude of rank correlation using the ranks of sample values of two random variables instead of actual data. It reflects the association between the direction and strength of the change trend of two random variables. It does not require a linear relationship or normal distribution of data and is insensitive to outliers, making it suitable for quantifying the relationship between the various dimensions of ecosystem stability. We selected pairs of resistance, resilience, and temporal stability based on the maximum disturbance intensity, and conducted a Spearman rank correlation analysis between them.

2.3.5. Disturbance Result Validation Method

The validation method calculated the probability that the LandTrendr algorithm would detect disturbance nodes in areas where known disturbance events occurred. This probability was then compared to the average disturbance occurrence rate detected by the LandTrendr algorithm across the entire region and time period (calculated as the ratio of pixels identified as change nodes to the total number of pixels). If the disturbance detection probability in the validation areas was significantly higher than the regional average, this indicated that the method was effective in detecting disturbances. Additionally, time series fitting results were analyzed for validation points across different regions and landscape types to further verify the accuracy of the disturbance detection results.

3. Results

3.1. Disturbance Detection

3.1.1. Disturbance Frequency

As depicted in Figure 3, between 1986 and 2021, 72.27% of the pixels in the LR-YRB experienced one to nine disturbances, while the remaining 27.73% of the pixels showed only stable or recovery phases. Among the disturbed pixels, more than half experienced one or two disturbances, and pixels disturbed more than five times constituted less than 3% of the total. The spatial distribution of disturbance frequencies shows that the urban area of Zhengzhou (site 1 in Figure 1) and its surroundings predominantly experienced a single disturbance. The southwestern large agricultural lands exhibited clusters of three to four disturbances. Undisturbed areas were primarily contiguous around central lakes, forests areas, and landward sides of the river deltas. The delta region showed a gradient of decreasing disturbance frequency from sea to land, with the seaside contributing more to the frequency of four or more disturbances.
The analysis of disturbance frequency proportions across different landscape types revealed a general trend, where the occurrence of three to six disturbances progressively decreased, while the proportions of zero, one, and two disturbances varied among types. Notably, impervious surfaces were the only type where 1-time disturbances outnumbered no disturbances, and these also had the smallest proportion of undisturbed areas compared to other types. Approximately 50% of pixels in forest, shrubland, and grassland remained undisturbed. Within these, forests and shrublands more frequently experienced 1-time disturbances compared to 2-time disturbances, whereas grasslands showed the opposite pattern. Cropland and impervious surfaces exhibited similar proportions of 1-time and 2-time disturbances, with both displaying diminishing proportions for disturbances occurring more than twice. Water bodies and bare lands showed a clear distinction in the proportion of 2-time disturbances, but the other disturbance frequencies were comparable. Overall, the variation in disturbance frequency distributions among landscape types was primarily evident for zero to two disturbances, with a more uniform pattern for disturbances occurring more than three times. Natural/near-natural vegetation types had a higher proportion of undisturbed pixels and a relatively consistent distribution of disturbance frequencies, whereas landscapes heavily influenced by human activities showed diverse distributions under low-frequency disturbances.

3.1.2. Maximum Disturbance Occurrence Year and Duration

As illustrated in Figure 4, in the LR-YRB, the pixels experiencing the maximum disturbances were primarily concentrated in the years 1986 and 2002, accounting for 34.75% and 24.50% of the total disturbed pixels, respectively. This was followed by the years 2003, 1999, and 2010, with maximum disturbances affecting 8.66%, 5.64%, and 4.33% of pixels, respectively, while other years had relatively fewer pixels with the maximum disturbance. The primary areas affected by the maximum disturbances were urban areas and their surroundings, as well as landscapes significantly impacted by human activities. The year 2002 was marked by sparse rainfall and severe drought conditions in the Yellow River Basin, making it the year with the second-highest disturbance after 1986. The lingering effects of the drought also resulted in a relatively high number of maximum disturbances in 2003. After a significant shift in 2002, the rate of change in the cumulative frequency distribution of maximum disturbances was slightly lower than in previous periods. This could be attributed to the frequent interruptions of the Yellow River’s flow from the 1970s to 1990s. However, starting in 1999, unified water management and control were implemented, and the completion of the Xiaolangdi Water Control Project in 2001 helped the Yellow River’s main channel achieve a continuous flow for 23 consecutive years, thereby gradually aiding the ecological recovery of the basin. Additionally, since 2000, China has enacted comprehensive natural forest protection policies. These measures have both directly and indirectly stabilized and progressively improved the ecological conditions in the LR-YRB, leading to a reduction in negative disturbances.
Regarding the duration of maximum disturbances in the LR-YRB, aside from disturbances lasting zero years, the most common duration was two years, accounting for 20.17% of the disturbances. Pixels with disturbances lasting three and four years represented 7.84% and 3.13%, respectively. Disturbances lasting for 35 years, mainly in the Zhengzhou urban area and the southwestern agricultural regions, accounted for 10.3% of the pixels, indicating that these areas underwent overall trend changes and exhibited characteristics of continuous disturbance. Disturbances lasting 15 years accounted for 4.45% of the cases, while all other durations constituted less than 3%. In summary, the durations of maximum disturbances were predominantly characterized by short-term changes driven by anthropogenic or natural factors, though long-term disturbance–recovery dynamics and interannual variability during growth seasons were also commonly observed.

3.2. Multidimensional Stability Analysis

3.2.1. Resistance

From the distribution of resistance under maximum disturbance (i.e., minimum resistance) (Figure 5), in the past 36 years, most areas in the LR-YRB had stronger (IV) or strong (V) resistance, with proportions of 47.99% and 40.62%, respectively. Pixels with weak (I) resistance were mainly concentrated near the water bodies of the Yellow River channel, lakes, and river deltas, accounting for about 5.47% of the resistance distribution area. Weaker (II) resistance was sporadically distributed throughout the study area, accounting for less than 0.8%. Ordinary (III) resistance accounted for 5.16% and was present in multiple areas of the study area, showing some clustering.
From the perspective of each landscape type, bare land and non-permanent water bodies had a weaker resistance, and the standard deviation reflected a larger magnitude of dispersion. The resistance of other landscape types was relatively strong, with no significant differences in the magnitude of resistance. Among them, forests and shrubs had the strongest resistance and were more concentrated in distribution, cropland and grassland had the next strongest resistance, with cropland overall stronger than grassland, but with a larger magnitude of dispersion.

3.2.2. Resilience

As shown in Figure 6, the overall resilience corresponding to the maximum magnitude of disturbance was weak (I), with a pixel proportion reaching 70.32% of the disturbed pixels. Weaker (II) resilience was scattered among them, accounting for about 16.7% of the disturbed pixels. Ordinary (III) resilience was distributed, with clustering in southwestern cultivated land, accounting for 9.24%. The resilience near the estuary and around water bodies was relatively strong, with strong (V) and stronger (IV) recovery areas accounting for a total of 3.75% of the disturbed area. Spatially, the graded distribution features were different overall from those of resistance.
The differences in resilience among various landscape types were obvious. Bare lands and water bodies had much higher resilience than other landscape types, and their magnitude of dispersion was also relatively large. The resilience of construction land was next, followed by the resilience of cultivated land. The resilience of shrubland, grassland, and forest, three types of vegetation landscapes, was smaller, with little difference among them. This indicates that, within a specific region where external environmental conditions are relatively consistent, simpler systems tend to recover more quickly after being damaged.

3.2.3. Temporal Stability

From Figure 7, over the entire time series, most pixels (61.01%) in the region had an ordinary (III) temporal stability, followed by weaker (II) stability (20.73%) and stronger (IV) stability (13.39%). Weak (I) stability (3.42%) was mainly distributed along the river and in parts of the river delta, while strong (V) stability (1.14%) was mainly distributed around lakes/reservoirs, aquaculture water bodies at the estuary, and surrounding coastal wetlands, with many also distributed in the central forest area.
Forests and shrubs had higher temporal stability, followed by grassland and farmland, bare lands and impervious land were next, and water bodies had the poorest temporal stability. From the standard deviation statistics, the stability values of water bodies and bare lands were more dispersed, while the standard deviation of farmland and grassland was smaller.

3.2.4. Regime Shift Rate

Regime shift rate of stability is the ratio of the disturbance magnitude to the corresponding recovery magnitude. This study calculated the regime shift rate corresponding to the maximum magnitude of disturbance and the subsequent magnitude of recovery. A regime shift rate >1 indicates that an ecosystem has recovered to a state better than before the disturbance, a regime shift rate <1 indicates that the ecosystem cannot recover to its pre-disturbance state, and a regime shift rate = 1 indicates that the ecosystem has recovered to the same state as before the disturbance. As shown in Figure 8, more than half (about 58.83%) of the pixels strongly recovered from the disturbance, about 40.17% of the pixels had a stronger magnitude of disturbance than recovery, and only 1% of the pixels had the same magnitude of disturbance and recovery under maximum disturbance.
From the average regime shift rate of pixels in each landscape type, about 74.43% of forests, 74.15% of shrubs, and 59.32% of grasslands had a regime shift rate <1. Within cropland and impervious land, the proportions of regime shift rate <1 were 55.88% and 64.07%, respectively, and the proportions of pixels with a regime shift rate = 1 were about 0.31% and 1.21%. Non-permanent water bodies and bare lands had a higher proportion of regime shift rate = 1 than other types, at about 20%, and the proportions of pixels with a regime shift rate <1 were 65% and 63.79%, respectively. Overall, more than half of the pixels recovered to or exceeded their pre-disturbance state after the maximum disturbance. This indicates a general improvement in ecosystem quality and stability within the LR-YRB.

3.3. Analysis of Stability Dimensions

A Spearman rank correlation analysis was conducted for the three stability dimensions for all years inferred from the remote sensing data (Figure 9), with the resistance values represented in logarithmic form. Overall, all pairwise correlations of the stability dimensions showed clustering intervals in the scatterplot kernel density maps, the distribution of temporal stability was relatively dispersed when the resistance axis interval was greater than 3, there was a high-density clustering area of temporal stability between 0 and 3 on the resistance axis, the temporal stability clustered around 0 when resilience was greater than 0.5, and the distribution of temporal stability values was more dispersed as the resilience approached 0, forming two high-density clustering areas of temporal stability in the 0~0.5 interval of resilience. The scatterplot between resistance and resilience split into two layers similar in shape around a resilience value of 0.75. From the overall correlation between stability components, temporal stability and resistance were positively correlated, with a correlation coefficient of about 0.458. Temporal stability and resilience were negatively correlated, with a correlation coefficient of about −0.447. The correlation between resistance and resilience was not obvious, with a correlation coefficient of only −0.175. This indicated that each stability component had its own characteristics, and the correlation between components was not strong, nor were they completely independent, suggesting that no single aspect of stability was sufficient to reflect the overall stability of the system, highlighting the importance of studying the dimensions of regional ecosystem stability for a proper assessment of system stability.
To further explore whether the relationship between stability components was affected by the size of the disturbance, the correlation was calculated separately for each of the five levels of disturbance. The correlation between resistance and resilience was weakly negative in all segments, with correlation coefficients roughly between 0 and −0.2. The relationship between temporal stability and resistance, as well as between temporal stability and resilience, was generally consistent with the overall state in terms of sign under different disturbance levels, with some differences in values, and the correlation in the fourth level was opposite in sign to the overall state. The line graphs of temporal stability with resistance and temporal stability with resilience by stage show symmetry. In summary, the correlation between stability components by level grouping was basically consistent with the overall correlation in terms of sign, but there were significant differences in the values of the correlation between the temporal stability and the other two components. This suggests that different magnitudes of disturbance may have affected the nature and magnitude of the stability dimensions.

4. Discussion

4.1. Advantages and Uncertainties of the Assessment Framework

This study presents a framework for assessing the multidimensional stability of complex ecosystems based on disturbances, without distinguishing specific disturbance types, and considering multiple disturbances over multiple years. Firstly, we aimed to determine comprehensive descriptive indicators that reflect both the integrity of the ecosystem and have spatiotemporal representativeness. Studies have shown that under conditions of lower biomass density, estimating the resilience of vegetation ecosystems using only optical satellite vegetation indices involves considerable uncertainty [41], and spectral data alone cannot detect low-frequency disturbances [42]. Disturbances can change the state and trajectory of a system, leading to spatiotemporal heterogeneity, and the magnitude of recovery after a disturbance is highly related to the quality of the regional ecological environment. Integrating pixel spatial neighborhood features, considering spatial neighborhood relationships and the relationship between landscape structure/pattern and processes, has been proven to improve the accuracy of remote sensing disturbance detection [43,44,45,46] and enhance the detection accuracy of recovery endpoints. In this study, we used the RSEI and ecosystem texture parameters based on long time series remote sensing data as the basic parameters for disturbance and stability assessment, integrating spatial structure conditions into time series detection, enabling simultaneous monitoring and rapid assessment of different types of ecosystems, and, to a certain extent, establishing a response relationship with regional-scale landscape patterns and ecological process changes. In terms of specific indicators, the four parameters of the RSEI index in this study contributed an average of 75.86% to the first principal component, demonstrating its suitability for application in the LR-YRB. Moreover, the texture parameters were weighted and combined with the RSEI with the same weight coefficient, allowing the retention of structural and quality characteristics without subjective bias.
In natural ecosystems, the relationships between different stability dimensions can vary due to differences in intrinsic population growth rates and responses to environmental changes. These dimensions are often regulated by distinct ecological processes and mechanisms [47]. In composite ecosystems, these relationships are more complex, with stability varying across different disturbances and landscape or community types. Therefore, conducting multidimensional assessments of stability in regional composite ecosystems is crucial for gaining a comprehensive understanding of their dynamics. The LandTrendr [38] algorithm is capable of detecting disturbances caused by both single and mixed factors. It effectively captures both gradual and abrupt events, maintaining consistency across regions and time scales. Additionally, LandTrendr can identify changes in both vegetation and non-vegetation features, making it well-suited for detecting multiple disturbances over extended periods in composite ecosystems.
However, some uncertainties remain regarding the multidimensional stability assessment of this study. To create a scalable classification standard applicable across different spatial and temporal scales, as well as varying levels of disturbances, we employed approximately equal-interval classifications. This approach may have led to insufficient differentiation of stability dimensions. While we relied on relative comparisons between landscape types to mitigate this limitation, further exploration of classification thresholds and the sensitivity of category divisions is necessary. Developing clearer and more distinct classifications would enhance the visualization of spatial patterns and improve the comparability of different stability dimensions.

4.2. Rationality of Stability Assessment Results

4.2.1. Multidimensional Stability of the Ecosystem under Maximum Disturbance

Under maximum disturbance conditions, ecosystems in the LR-YRB displayed a characteristic pattern of “high resistance–low resilience” across various regions and landscape types. Forests showed higher resistance, due to their species richness, biodiversity, and complex structural composition, making them significantly more resistant than simpler ecosystems like grasslands. Grasslands, while structurally simpler, displayed resilience comparable to forests, as their smaller area and proximity to forested regions helped balance between resistance and resilience [48]. Bare land, being structurally simple, showed weaker resistance but stronger resilience. Coastal wetlands exhibited lower resistance, primarily due to frequent human activity and river dynamics, though the vegetation in these areas demonstrated high resilience. Farmland, heavily influenced by human activity, presented variable resistance and resilience depending on the crop type and community structure. A study by Shi [49] on the middle reaches of the Yellow River Basin similarly concluded that forests are the most resistant vegetation type, followed by farmland, with grasslands being the least resistant. The findings from this study are consistent with those results.

4.2.2. Analysis of the Relationships between Stability Dimensions

Pairwise correlation analysis of the time series for stability dimensions revealed a positive correlation between temporal stability and resistance. This suggests that ecosystems with higher resistance can maintain their primary structure and function for extended periods in the face of external disturbances. However, temporal stability was negatively correlated with resilience. Ecosystems with high resilience tended to respond quickly to disturbances, but this often came at the cost of reduced ecosystem diversity and complexity. For instance, large-scale farmland may rapidly recover to a “new steady state” [50] dominated by fast-growing species, but this new state typically features simpler functions and structures, diminishing the system’s capacity to handle long-term disturbances and resulting in lower temporal stability. Furthermore, when a system recovers to a higher-level steady state, increased variability can lead to reduced temporal stability.
Previous studies supported the complexity of these relationships. For example, Huang [7] found no significant correlation between resistance and resilience in vegetation studies in the karst regions of southwest China. Resistance and variability (temporal stability) showed a weak negative correlation, while resilience and variability were uncorrelated. Chen [51], in a global vegetation study, demonstrated that resistance was strongly correlated with temporal stability, while resilience remained independent of the other two dimensions. Xu [52] indicated that ecosystems with high structural resistance often exhibit lower recovery rates, while for functional stability, there is no clear relationship between initial resistance and post-disturbance resilience. This highlights the complexity of stability relationships, suggesting that conclusions from regional studies may not be easily generalizable across different ecosystems. This variation is likely due to the spatial heterogeneity of disturbance impacts, elevation dependence [53,54], and the distinct ecological processes regulating stability dimensions. As a result, ecosystem management strategies should be tailored to specific objectives, to ensure the long-term maintenance of ecosystem functions.
The weak correlations between stability dimensions in this study indicate a low redundancy among them, highlighting the need for multidimensional stability assessments. This underscores the importance of using a multidimensional approach to effectively assess the stability of composite ecosystems.

4.2.3. Analysis of Regime Shifts and Their Significance for Ecosystem Management

After experiencing maximum disturbances, more than half of the pixels exhibited a regime shift rate greater than 1. Combined with the resistance and resilience results, this suggests that, although the recovery may be slow, the degree of recovery was relatively high, ultimately reaching or exceeding the pre-disturbance steady state. This recovery was likely driven by both natural and human factors. In natural ecosystems, when disturbances do not exceed a critical threshold, self-regulation and natural recovery help maintain system stability. Human interventions, such as reservoir regulation and irrigation management, have mitigated the impacts of extreme events like droughts and floods. Additionally, ecological projects, such as the Grain for Green Program and the Natural Forest Protection Project, have enhanced ecosystem quality, promoting a higher regime shift rate. Overall, the ecosystem in the LR-YRB has shown gradual improvement and significant recovery potential, demonstrating a strong long-term resilience. Therefore, future management efforts should prioritize long-term restorative strategies, focusing on natural recovery processes and optimizing ecological restoration measures. Implementing differentiated strategies will promote the health and sustainable development of regional ecosystems.

4.3. Reasonability of Disturbance Detection

Based on the validation results of typical disturbance events, the average occurrence rate of disturbance/recovery events over the past 35 years was 9.46%. Analyzing independent disturbance events in typical years revealed that the probability of detecting disturbances in areas of impervious surface changes, floods, fires, and droughts was 36.53%, 17.27%, 45.87%, and 39.62%, respectively. The detection probabilities in areas with typical disturbance events were significantly higher than for average rate of change detection events, indicating that the time series disturbance detection results in this study were spatially reasonable. The inability to detect some pixels as disturbances in typical disturbance areas may have been due to mismatches between disturbance event levels. Some events’ intensity and impact ranges may not have been classified as core disturbances in long-term changes, or they may have occurred outside the growing season, resulting in an interannual lag in their effects. However, a more accurate validation would require ground-based observations and the use of new tools to ensure the alignment between the multi-temporal and spatial-scale datasets involved in the validation process [55].
We also chose four verification points (Figure 1) for different areas and landscape types, to intuitively assess the feasibility of the research results in this paper. Site 1 represented the urban area of Zhengzhou, Site 2 represented cropland, Site 3 represented forest, and Site 4 represented coastal farmland area. We obtained a sequence chart of the RSEI and composite stability parameter index (i.e., stability parameters combining RSEI and texture structure), corresponding to a complete disturbance–recovery cycle at five time points at the verification points (the color range from blue to red indicates a gradual deterioration in ecological conditions), and plotted a LandTrendr disturbance and recovery detection time pattern chart based on the composite parameters (Figure 10). Overall, the LandTrendr algorithm could generally detect different disturbance patterns and stage trends, but its applicability was not strong when the amplitude of the ecosystem descriptor parameter curve was large and the interval of changes was short, especially when a single disturbance–recovery event was shorter than two years, being prone to missed detection. Moreover, looking at the key change points of the RSEI and composite parameters at the verification points, the detected disturbance change trend was reflected in both the composite parameters and RSEI, with an obvious deterioration after the disturbance, the recovery end better than the disturbance end/recovery onset, and the disturbance characteristics reflected in both the structure and quality indicators.

4.4. Other Limitations and Prospects

Additional limitations were related to cloud cover and climate conditions, which often limit the availability of complete time series data, and may affect disturbance detection accuracy. Future studies could incorporate multi-source remote sensing data and appropriate interpolation methods to address this issue. In addition, the stability indicators used in this study were relative measures of disturbance. Incorporating reference benchmarks, such as historical or optimal background conditions observed on the ground, could provide an absolute baseline for stability assessments, enhancing the comparability across different periods and improving predictions of future stability trends. Additionally, this study treated all disturbances uniformly, without differentiating between types. Future research should specify the nature, patterns, and combinations of disturbances to improve the understanding of the impacts of multiple disturbance types.

5. Conclusions

This study conducted a long-term multidimensional stability assessment of the LR-YRB ecosystem, focusing on resistance, resilience, temporal stability, and state transitions in response to disturbances. A composite ecosystem stability parameter was developed by integrating ecosystem quality and spatial texture characteristics using Landsat imagery, while the LandTrendr algorithm was applied for continuous disturbance and recovery detection. The results indicated that between 1986 and 2021, 72.27% of the pixels in the LR-YRB experienced one to nine disturbances, with one to two disturbances being the most common. A significant portion of pixels in near-natural vegetation landscapes remained undisturbed. The most substantial disturbances occurred in 1986 and 2002, primarily as short-term events lasting 2 to 3 years, while 10.3% of pixels subjected to the maximum disturbance exhibited gradual, long-term changes over 35 years.
The spatial patterns of resistance and resilience were generally inverse, with vegetation exhibiting “high resistance–low resilience” and bare land showing weaker resistance but higher resilience. Approximately 59.83% of the pixels recovered to, or exceeded, their pre-disturbance state, indicating an overall improvement in ecosystem stability and significant recovery potential following disturbances. The correlations between the multidimensional stability components in the LR-YRB were weak. Resistance and resilience showed a slight negative correlation (R = −0.175), while resistance and temporal stability, and resilience and temporal stability, were positively (R = 0.458) and negatively (R = −0.447) correlated, respectively. The effectiveness of these stability dimensions was also influenced by varying disturbance intensities. These findings highlight the need for multidimensional assessments of regional composite ecosystem stability using satellite remote sensing. This approach lays the groundwork for integrating multidimensional stability into early warning systems and recovery management, offering a holistic perspective to support decision-making in ecological restoration and sustainable development.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/rs16203762/s1, Table S1. The disturbance validation datasets used in this study. Table S2. Construction of RSEI indicators. References [56,57,58,59,60,61,62,63,64] are cited in Supplementary Materials.

Author Contributions

Conceptualization, J.D.; methodology, L.L. and J.W.; software, L.L. and J.W.; validation, Z.S. (Zebang Song), G.Z. and F.C.; formal analysis, J.D.; investigation, L.L. and X.Z.; resources, J.D.; data curation, L.L. and J.W.; writing—original draft preparation, L.L.; writing—review and editing, Z.S. (Zhilu Sheng) and X.Z.; visualization, L.L. and J.W.; supervision, J.D.; funding acquisition, J.D. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the Fundamental Research Funds for the Central Public-interest Scientific Institution, Grant No. 2024YSKY-37 and No. 2022YSKY-17, and the National Natural Science Foundation of China, Grant No. 41001055.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Scheffer, M.; Carpenter, S.; Foley, J.A.; Folke, C.; Walker, B. Catastrophic shifts in ecosystems. Nature 2001, 413, 591–596. [Google Scholar] [CrossRef] [PubMed]
  2. Donohue, I.; Hillebrand, H.; Montoya, J.M.; Petchey, O.L.; Pimm, S.L.; Fowler, M.S.; Healy, K.; Jackson, A.L.; Lurgi, M.; McClean, D.; et al. Navigating the complexity of ecological stability. Ecol. Lett. 2016, 19, 1172–1185. [Google Scholar] [CrossRef] [PubMed]
  3. Li, W.; Li, X.; Zhao, Y.; Zheng, S.; Bai, Y. Ecosystem structure, functioning and stability under climate change and grazing in grasslands: Current status and future prospects. Curr. Opin. Environ. Sustain. 2018, 33, 124–135. [Google Scholar] [CrossRef]
  4. Francesco, P.; Andreu, R. Effects of multiple stressors on the dimensionality of ecological stability. Ecol. Lett. 2021, 24, 1594–1606. [Google Scholar]
  5. Donohue, I.; Petchey, O.L.; Montoya, J.M.; Jackson, A.L.; McNally, L.; Viana, M.; Healy, K.; Lurgi, M.; O’Connor, N.E.; Emmerson, M.C. On the dimensionality of ecological stability. Ecol. Lett. 2013, 16, 421–429. [Google Scholar] [CrossRef]
  6. Kang, W.; Liu, S.; Chen, X.; Feng, K.; Guo, Z.; Wang, T. Evaluation of ecosystem stability against climate changes via satellite data in the eastern sandy area of northern China. J. Environ. Manag. 2022, 308, 114596. [Google Scholar] [CrossRef]
  7. Huang, Z.; Liu, X.; Yang, Q.; Meng, Y.; Zhu, L.; Zou, X. Quantifying the spatiotemporal characteristics of multi-dimensional karst ecosystem stability with Landsat time series in southwest China. Int. J. Appl. Earth Obs. Geoinf. 2021, 104, 102575. [Google Scholar] [CrossRef]
  8. Zhang, Y.; Liu, X.; Yang, Q.; Liu, Z.; Li, Y. Extracting Frequent Sequential Patterns of Forest Landscape Dynamics in Fenhe River Basin, Northern China, from Landsat Time Series to Evaluate Landscape Stability. Remote Sens. 2021, 13, 3963. [Google Scholar] [CrossRef]
  9. Eva, M.; Monika, I.; Štefan, K. The evaluation of anthropogenic impact on the ecological stability of landscape. J. Environ. Biol. 2015, 36, 1. [Google Scholar]
  10. Li, H.; Li, L.; Su, F.; Wang, T.; Gao, P. Ecological stability evaluation of tidal flat in coastal estuary: A case study of Liaohe estuary wetland, China. Ecol. Indic. 2021, 130, 108032. [Google Scholar] [CrossRef]
  11. Zhang, R.; Zhang, X.; Yang, J.; Yuan, H. Wetland ecosystem stability evaluation by using Analytical Hierarchy Process (AHP) approach in Yinchuan Plain, China. Math. Comput. Model. 2013, 57, 366–374. [Google Scholar] [CrossRef]
  12. Li, L.; Li, G.; Du, J.; Wu, J.; Cui, L.; Chen, Y. Effects of tidal flat reclamation on the stability of coastal wetland ecosystem services: A case study in Jiangsu Coast, China. Ecol. Indic. 2022, 145, 109697. [Google Scholar] [CrossRef]
  13. Tang, D.; Fan, H.; Yang, K.; Zhang, Y. Mapping forest disturbance across the China–Laos border using annual Landsat time series. Int. J. Remote Sens. 2019, 40, 2895–2915. [Google Scholar] [CrossRef]
  14. Yang, Y.; Erskine, P.D.; Lechner, A.M.; Mulligan, D.; Zhang, S.; Wang, Z. Detecting the dynamics of vegetation disturbance and recovery in surface mining area via Landsat imagery and LandTrendr algorithm. J. Clean. Prod. 2018, 178, 353–362. [Google Scholar] [CrossRef]
  15. Awty-Carroll, K.; Bunting, P.; Hardy, A.; Bell, G. Using continuous change detection and classification of Landsat data to investigate long-term mangrove dynamics in the Sundarbans region. Remote Sens. 2019, 11, 2833. [Google Scholar] [CrossRef]
  16. Zhu, Z.; Woodcock, C.E. Continuous change detection and classification of land cover using all available Landsat data. Remote Sens. Environ. 2014, 144, 152–171. [Google Scholar] [CrossRef]
  17. Li, M.; Huang, C.; Zhu, Z.; Wen, W.; Xu, D.; Liu, A. Use of remote sensing coupled with a vegetation change tracker model to assess rates of forest change and fragmentation in Mississippi, USA. Int. J. Remote Sens. 2009, 30, 6559–6574. [Google Scholar] [CrossRef]
  18. Huang, C.; Goward, S.N.; Masek, J.G.; Thomas, N.; Zhu, Z.; Vogelmann, J.E. An automated approach for reconstructing recent forest disturbance history using dense Landsat time series stacks. Remote Sens. Environ. 2010, 114, 183–198. [Google Scholar] [CrossRef]
  19. Waller, E.K.; Villarreal, M.L.; Poitras, T.B.; Nauman, T.W.; Duniway, M.C. Landsat time series analysis of fractional plant cover changes on abandoned energy development sites. Int. J. Appl. Earth Obs. Geoinf. 2018, 73, 407–419. [Google Scholar] [CrossRef]
  20. Verbesselt, J.; Hyndman, R.; Zeileis, A.; Culvenor, D. Phenological change detection while accounting for abrupt and gradual trends in satellite image time series. Remote Sens. Environ. 2010, 114, 2970–2980. [Google Scholar] [CrossRef]
  21. Ding, H.; Yuan, Z.; Yin, J.; Shi, X.; Shi, M. Evaluating ecosystem stability based on the dynamic time warping algorithm: A case study in the Minjiang river Basin, China. Ecol. Indic. 2023, 154, 110501. [Google Scholar] [CrossRef]
  22. Griffiths, P.; Kuemmerle, T.; Kennedy, R.E.; Abrudan, I.V.; Knorn, J.; Hostert, P. Using annual time-series of Landsat images to assess the effects of forest restitution in post-socialist Romania. Remote Sens. Environ. 2012, 118, 199–214. [Google Scholar] [CrossRef]
  23. Kennedy, R.E.; Yang, Z.; Braaten, J.; Copass, C.; Antonova, N.; Jordan, C.; Nelson, P. Attribution of disturbance change agent from Landsat time-series in support of habitat monitoring in the Puget Sound region, USA. Remote Sens. Environ. 2015, 166, 271–285. [Google Scholar] [CrossRef]
  24. Fu, B.; Lan, F.; Xie, S.; Liu, M.; He, H.; Li, Y.; Liu, L.; Huang, L.; Fan, D.; Gao, E.; et al. Spatio-temporal coupling coordination analysis between marsh vegetation and hydrology change from 1985 to 2019 using LandTrendr algorithm and Google Earth Engine. Ecol. Indic. 2022, 137, 108763. [Google Scholar] [CrossRef]
  25. Shen, Q.; Gao, G.; Duan, Y.; Chen, L. Long-term continuous changes of vegetation cover in desert oasis of a hyper-arid endorheic basin with LandTrendr algorithm. Ecol. Indic. 2024, 166, 112418. [Google Scholar] [CrossRef]
  26. Fu, B.; Chen, L.; Wang, J.; Meng, Q.; Zhao, W. Land use structure and ecological processes. Quat. Sci. 2003, 23, 247–256. [Google Scholar]
  27. Yu, G.; Wang, Y.; Yang, M. Discussion on the ecological theory and technological approaches of ecosystem quality improvement and stability enhancement. Chin. J. Appl. Ecol. 2023, 34, 1–10. [Google Scholar] [CrossRef]
  28. Tamiminia, H.; Salehi, B.; Mahdianpari, M.; Quackenbush, L.; Adeli, S.; Brisco, B. Google Earth Engine for geo-big data applications: A meta-analysis and systematic review. ISPRS J. Photogramm. Remote Sens. 2020, 164, 152–170. [Google Scholar] [CrossRef]
  29. Simonetti, D.; Simonetti, E.; Szantoi, Z.; Lupi, A.; Eva, H.D. First Results From the Phenology-Based Synthesis Classifier Using Landsat 8 Imagery. IEEE Geosci. Remote Sens. Lett. 2015, 12, 1496–1500. [Google Scholar] [CrossRef]
  30. Roy, D.P.; Kovalskyy, V.; Zhang, H.; Vermote, E.F.; Yan, L.; Kumar, S.; Egorov, A. Characterization of Landsat-7 to Landsat-8 reflective wavelength and normalized difference vegetation index continuity. Remote Sens. Environ. 2016, 185, 57–70. [Google Scholar] [CrossRef]
  31. Zhang, J.; Du, J.; Fang, S.; Sheng, Z.; Zhang, Y.; Sun, B.; Mao, J.; Li, L. Dynamic changes, spatiotemporal differences, and ecological effects of impervious surfaces in the Yellow river basin, 1986–2020. Remote Sens. 2023, 15, 268. [Google Scholar] [CrossRef]
  32. Yang, J.; Huang, X. The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019. Earth Syst. Sci. Data 2021, 13, 3907–3925. [Google Scholar] [CrossRef]
  33. Zhang, Y.; Du, J.; Guo, L.; Fang, S.; Zhang, J.; Sun, B.; Mao, J.; Sheng, Z.; Li, L. Long-term detection and spatiotemporal variation analysis of open-surface water bodies in the Yellow River Basin from 1986 to 2020. Sci. Total Environ. 2022, 845, 157152. [Google Scholar] [CrossRef]
  34. Pekel, J.F.; Cottam, A.; Gorelick, N.; Belward, A. High-resolution mapping of global surface water and its long-term changes. Nature 2016, 540, 418–422. [Google Scholar] [CrossRef]
  35. Xu, H.; Wang, M.; Shi, T.; Guan, H.; Fang, C.; Lin, Z. Prediction of ecological effects of potential population and impervious surface increases using a remote sensing based ecological index (RSEI). Ecol. Indic. 2018, 93, 730–740. [Google Scholar] [CrossRef]
  36. Camps-Valls, G.; Campos-Taberner, M.; Moreno-Martínez, Á.; Walther, S.; Duveiller, G.; Cescatti, A.; Mahecha, M.D.; Muñoz-Marí, J.; García-Haro, F.J.; Guanter, L. A unified vegetation index for quantifying the terrestrial biosphere. Sci. Adv. 2021, 7, eabc7447. [Google Scholar] [CrossRef]
  37. Haralick, R.M.; Shanmugam, K.; Dinstein, I.H. Textural features for image classification. IEEE Trans. Syst. Man Cybern. 1973, SMC-3, 610–621. [Google Scholar] [CrossRef]
  38. Kennedy, R.E.; Yang, Z.; Cohen, W.B. Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr—Temporal segmentation algorithms. Remote Sens. Environ. 2010, 114, 2897–2910. [Google Scholar] [CrossRef]
  39. Kéfi, S.; Domínguez-García, V.; Donohue, I.; Fontaine, C.; Thébault, E.; Dakos, V. Advancing our understanding of ecological stability. Ecol. Lett. 2019, 22, 1349–1356. [Google Scholar] [CrossRef]
  40. White, H.J.; Gaul, W.; Sadykova, D.; León-Sánchez, L.; Caplat, P.; Emmerson, M.C.; Yearsley, J.M. Quantifying large-scale ecosystem stability with remote sensing data. Remote Sens. Ecol. Conserv. 2020, 6, 354–365. [Google Scholar] [CrossRef]
  41. Smith, T.; Boers, N. Reliability of vegetation resilience estimates depends on biomass density. Nat. Ecol. Evol. 2023, 7, 1799–1808. [Google Scholar] [CrossRef]
  42. Liu, M.; Liu, X.; Wu, L.; Tang, Y.; Li, Y.; Zhang, Y.; Ye, L.; Zhang, B. Establishing forest resilience indicators in the hilly red soil region of southern China from vegetation greenness and landscape metrics using dense Landsat time series. Ecol. Indic. 2021, 121, 106985. [Google Scholar] [CrossRef]
  43. Hermosilla, T.; Wulder, M.A.; White, J.C.; Coops, N.C.; Hobart, G.W. Regional detection, characterization, and attribution of annual forest change from 1984 to 2012 using Landsat-derived time-series metrics. Remote Sens. Environ. 2015, 170, 121–132. [Google Scholar] [CrossRef]
  44. Hamunyela, E.; Verbesselt, J.; Herold, M. Using spatial context to improve early detection of deforestation from Landsat time series. Remote Sens. Environ. 2016, 172, 126–138. [Google Scholar] [CrossRef]
  45. Zhu, Z. Change detection using landsat time series: A review of frequencies, preprocessing, algorithms, and applications. ISPRS J. Photogramm. Remote Sens. 2017, 130, 370–384. [Google Scholar] [CrossRef]
  46. Meng, Y.; Liu, X.; Ding, C.; Xu, B.; Zhu, L. Analysis of ecological resilience to evaluate the inherent maintenance capacity of a forest ecosystem using a dense Landsat time series. Ecol. Inform. 2020, 57, 101064. [Google Scholar] [CrossRef]
  47. Hillebrand, H.; Langenheder, S.; Lebret, K.; Lindström, E.; Östman, Ö.; Striebel, M. Decomposing multiple dimensions of stability in global change experiments. Ecol. Lett. 2017, 21, 21–30. [Google Scholar] [CrossRef]
  48. Li, X.; Piao, S.; Wang, K.; Wang, X.; Wang, T.; Ciais, P.; Chen, A.; Lian, X.; Peng, S.; Peñuelas, J. Temporal trade-off between gymnosperm resistance and resilience increases forest sensitivity to extreme drought. Nat. Ecol. Evol. 2020, 4, 1075–1083. [Google Scholar] [CrossRef]
  49. Shi, X.; Chen, F.; Ding, H.; Shi, M.; Li, Y. Assessing Vegetation Ecosystem Resistance to Drought in the Middle Reaches of the Yellow River Basin, China. Int. J. Environ. Res. Public Health 2022, 19, 4180. [Google Scholar] [CrossRef]
  50. Scheffer, M.; Carpenter, S.R. Catastrophic regime shifts in ecosystems: Linking theory to observation. Trends Ecol. Evol. 2003, 18, 648–656. [Google Scholar] [CrossRef]
  51. Chen, J.; Chi, Y.; Zhou, W.; Wang, Y.; Zhuang, J.; Zhao, N.; Ding, J.; Song, J.; Zhou, L. Quantifying the dimensionalities and drivers of ecosystem stability at global scale. J. Geophys. Res. Biogeosci. 2021, 126, e2020JG006041. [Google Scholar] [CrossRef]
  52. Xu, Z.; Ren, H.; Li, M.H.; van Ruijven, J.; Han, X.; Wan, S.; Li, H.; Yu, Q.; Jiang, Y.; Jiang, L. Environmental changes drive the temporal stability of semi-arid natural grasslands through altering species asynchrony. J. Ecol. 2015, 103, 1308–1316. [Google Scholar] [CrossRef]
  53. De Keersmaecker, W.; Lhermitte, S.; Tits, L.; Honnay, O.; Somers, B.; Coppin, P. A model quantifying global vegetation resistance and resilience to short-term climate anomalies and their relationship with vegetation cover. Glob. Ecol. Biogeogr. 2015, 24, 539–548. [Google Scholar] [CrossRef]
  54. Geng, S.; Shi, P.; Song, M.; Zong, N.; Zu, J.; Zhu, W. Diversity of vegetation composition enhances ecosystem stability along elevational gradients in the Taihang Mountains, China. Ecol. Indic. 2019, 104, 594–603. [Google Scholar] [CrossRef]
  55. Valerio, F.; Godinho, S.; Marques, A.T.; Crispim-Mendes, T.; Pita, R.; Silva, J.P. GEE_xtract: High-quality remote sensing data preparation and extraction for multiple spatio-temporal ecological scaling. Ecol. Inform. 2024, 80, 102502. [Google Scholar] [CrossRef]
  56. Padilla, M.; Wheeler, J.; Tansey, K. ESA Climate Change Initiative–Fire_cci D4. 1.1 Product Validation Report (PVR); Universidad de Alcala: Madrid, Spain, 2018. [Google Scholar]
  57. Tellman, B.; Sullivan, J.A.; Kuhn, C.; Kettner, A.J.; Doyle, C.S.; Brakenridge, G.R.; Erickson, T.A.; Slayback, D.A. Satellite imaging reveals increased proportion of population exposed to floods. Nature 2021, 596, 80–86. [Google Scholar] [CrossRef] [PubMed]
  58. Beguería, S.; Vicente Serrano, S.M.; Reig-Gracia, F.; Latorre Garcés, B. SPEIbase v. 2.9 [Dataset]. 2023. Available online: https://digital.csic.es/handle/10261/332007 (accessed on 12 February 2024).
  59. Crist, E.P. A TM tasseled cap equivalent transformation for reflectance factor data. Remote Sens. Environ. 1985, 17, 301–306. [Google Scholar] [CrossRef]
  60. Baig, M.H.A.; Zhang, L.; Shuai, T.; Tong, Q. Derivation of a tasselled cap transformation based on Landsat 8 at-satellite reflectance. Remote Sens. Lett. 2014, 5, 423–431. [Google Scholar] [CrossRef]
  61. Artis, D.A.; Carnahan, W.H. Survey of emissivity variability in thermography of urban areas. Remote Sens. Environ. 1982, 12, 313–329. [Google Scholar] [CrossRef]
  62. Sekertekin, A.; Bonafoni, S. Land surface temperature retrieval from Landsat 5, 7, and 8 over rural areas: Assessment of different retrieval algorithms and emissivity models and toolbox implementation. Remote Sens. 2020, 12, 294. [Google Scholar] [CrossRef]
  63. Kumar, D.; Soni, A.; Kumar, M. Retrieval of land surface temperature from landsat-8 thermal infrared sensor data. J. Hum. Earth Future 2022, 3, 159–168. [Google Scholar] [CrossRef]
  64. XU, H.-Q. A new index-based built-up index (IBI) and its eco-environmental significance. Remote Sens. Technol. Appl. 2011, 22, 301–308. [Google Scholar]
Remotesensing 16 03762 g001
Figure 1. Study area and spatial distribution of land use/landscape types (2020).
Figure 1. Study area and spatial distribution of land use/landscape types (2020).
Remotesensing 16 03762 g001
Remotesensing 16 03762 g002
Figure 2. Schematic diagram of ecosystem stability parameters under disturbance.
Figure 2. Schematic diagram of ecosystem stability parameters under disturbance.
Remotesensing 16 03762 g002
Remotesensing 16 03762 g003
Figure 3. Distribution of disturbance frequency.
Figure 3. Distribution of disturbance frequency.
Remotesensing 16 03762 g003
Remotesensing 16 03762 g004
Figure 4. Maximum disturbance occurrence years and duration.
Figure 4. Maximum disturbance occurrence years and duration.
Remotesensing 16 03762 g004
Remotesensing 16 03762 g005
Figure 5. Distribution of resistance (avg—the mean value; std—the standard deviation. Figure 5, Figure 6 and Figure 7 have the same meaning).
Figure 5. Distribution of resistance (avg—the mean value; std—the standard deviation. Figure 5, Figure 6 and Figure 7 have the same meaning).
Remotesensing 16 03762 g005
Remotesensing 16 03762 g006
Figure 6. Distribution of resilience.
Figure 6. Distribution of resilience.
Remotesensing 16 03762 g006
Remotesensing 16 03762 g007
Figure 7. Distribution of temporal stability.
Figure 7. Distribution of temporal stability.
Remotesensing 16 03762 g007
Remotesensing 16 03762 g008
Figure 8. Spatial distribution of regime shift rate of stability.
Figure 8. Spatial distribution of regime shift rate of stability.
Remotesensing 16 03762 g008
Remotesensing 16 03762 g009
Figure 9. Pairwise scatterplot kernel density maps (a–c) and correlation coefficients (d) between stability dimensions (p < 0.01).
Figure 9. Pairwise scatterplot kernel density maps (a–c) and correlation coefficients (d) between stability dimensions (p < 0.01).
Remotesensing 16 03762 g009
Remotesensing 16 03762 g010
Figure 10. Disturbance assessment accuracy case: (a) disturbance indicator parameters (top: RSEI; bottom: composite stability parameter index; “+” represents selected points); (b) disturbance and recovery detection results based on LandTrendr (blue line: original values of composite parameters; orange line: LandTrendr fitted curve; yellow vertical line: start year of disturbance; purple vertical line: end year of recovery).
Figure 10. Disturbance assessment accuracy case: (a) disturbance indicator parameters (top: RSEI; bottom: composite stability parameter index; “+” represents selected points); (b) disturbance and recovery detection results based on LandTrendr (blue line: original values of composite parameters; orange line: LandTrendr fitted curve; yellow vertical line: start year of disturbance; purple vertical line: end year of recovery).
Remotesensing 16 03762 g010
Table 1. Parameter settings for LandTrendr.
Table 1. Parameter settings for LandTrendr.
ParameterValueParameterValue
Max Segments12recoveryThreshold0.5
Spike Threshold0.95pvalThreshold0.1
Vertex Count Overshoot3bestModelProportion0.75
Prevent One Year RecoveryFalseminObservationsNeeded6
Table 2. Main stability indicators of this study.
Table 2. Main stability indicators of this study.
Stability ComponentExpressionExplanation
Resistance (RD)RD = 1/MResistance reflects an ecosystem’s ability to absorb disturbances, with the core concept being the capacity to “resist disturbances and maintain its original state.” It is quantified as the inverse of an ecosystem’s deviation from its equilibrium state following a disturbance [39]. A higher resistance value indicates a more stable system.
Resilience (RR)RR = M’/ΔT’Resilience refers to an ecosystem’s ability to recover to a stable state after being disturbed, with a core concept of “damaged but returning to its original state”. It is quantified as the ratio of the magnitude of changes in ecosystem parameters to the recovery time once the disturbance has been eliminated [40]. A higher resilience indicates a more stable system.
Temporal Stability (TS)TS = 1/cv
= μ/σ		Evaluated using the reciprocal of the coefficient of variation (cv), where μ is the mean value of the characteristic parameter, and σ is the standard deviation. The greater the temporal stability, the more stable the system.
Regime shift rate (RS)RS = M/M’Described by the ratio of the magnitude of disturbance to the magnitude of recovery [21]. This index can determine the direction of a regime shift.
Table 3. Grading of stability indicators.
Table 3. Grading of stability indicators.
LevelWeak (I)Weaker (II)Ordinary (III)Stronger (IV)Strong (V)
D≤1.25(1.25, 1.67](1.67, 2.5](2.5, 5]>5
RR≤0.2(0.2, 0.4](0.4, 0.6](0.6, 0.8]>0.8
TS≤1(1, 2](2, 3](3, 4]>4
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Li, L.; Du, J.; Wu, J.; Sheng, Z.; Zhu, X.; Song, Z.; Zhai, G.; Chong, F. Evaluating the Multidimensional Stability of Regional Ecosystems Using the LandTrendr Algorithm. Remote Sens. 2024, 16, 3762. https://doi.org/10.3390/rs16203762

AMA Style

Li L, Du J, Wu J, Sheng Z, Zhu X, Song Z, Zhai G, Chong F. Evaluating the Multidimensional Stability of Regional Ecosystems Using the LandTrendr Algorithm. Remote Sensing. 2024; 16(20):3762. https://doi.org/10.3390/rs16203762

Chicago/Turabian Style

Li, Lijuan, Jiaqiang Du, Jin Wu, Zhilu Sheng, Xiaoqian Zhu, Zebang Song, Guangqing Zhai, and Fangfang Chong. 2024. "Evaluating the Multidimensional Stability of Regional Ecosystems Using the LandTrendr Algorithm" Remote Sensing 16, no. 20: 3762. https://doi.org/10.3390/rs16203762

APA Style

Li, L., Du, J., Wu, J., Sheng, Z., Zhu, X., Song, Z., Zhai, G., & Chong, F. (2024). Evaluating the Multidimensional Stability of Regional Ecosystems Using the LandTrendr Algorithm. Remote Sensing, 16(20), 3762. https://doi.org/10.3390/rs16203762

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop