1. Introduction
In recent years, under the influences of human activities and climate change, the exhaustion of natural resources, environmental pollution, biodiversity decline, and the rising frequency of extreme weather events like droughts and floods have posed serious threats to the balance of ecosystems and the sustainable development of human society [
1,
2,
3,
4]. Based on diverse research perspectives including ecosystem services [
5], ecological vulnerability [
6], ecological security patterns [
7], and ecological carrying capacity [
8], the monitoring and assessment of changes in Ecological Environment Quality (EEQ) and its driving mechanisms have become essential for governments and ecological conservation organizations to formulating ecological conservation policies.
Xu H. Q. [
9] introduced the Remote Sensing Ecological Index (RSEI), which evaluates EEQ by integrating four indicators—wetness, greenness, heat, and dryness—through principal component analysis (PCA). Compared to methods using single indicators such as the normalized difference vegetation index (NDVI) [
10] or Leaf Area Index (LAI) [
11] for assessing EEQ, the RSEI contain more comprehensive information. The RSEI has been widely applied in evaluating EEQ in urban ecosystems [
12,
13], natural ecosystems [
14,
15], and key ecological conservation areas [
16] at different scales.
There are two limitations that can impact the assessment results of RSEI. First, the underlying data of RSEI rely on remote sensing images, where variations in sensors and acquisition times can lead to pseudofluctuations. Challenges like cloud cover, periodic stripes, bright lines, and spots make significant noise interference, substantially influencing the results of the driving analysis [
17,
18]. Second, ecosystems, especially urban ones, are influenced by the combined effects of natural, social, and economic factors [
19]. Natural and human factors are interwoven in ecosystem components, integrated in structure, coupled in processes, complementary in functions, and synergistic in services. Therefore, traditional correlation analysis and regression analysis have limitations in quantifying the contributions of driving factors to EEQ and their interactions.
The Jiaodong Peninsula, characterized as a complex coastal ecosystem, benefits from favorable climatic conditions and abundant natural resources, experiencing significant human activity. Over the past two decades, the region has become one of the fastest-growing urban agglomerations in northern China, associated with increasing vulnerability to coastal erosion, fragmented landscapes in ecological source areas, and declining biodiversity [
20,
21]. According to Huang et al., there was a significant decrease of 16.8% in the ecological resilience of the Shandong Peninsula region from 2000 to 2018 [
22]. Intense human activities [
23,
24] and extreme weather events [
25] have posed significant challenges to the ecological balance of the Jiaodong Peninsula. Using RSEI to assess the historical and current status of the EEQ in Jiaodong Peninsula, exploring the driving mechanisms of human activities and natural elements, is crucial for formulating targeted local ecological conservation policies and enhancing the resilience of the complex coastal ecosystem of the Jiaodong Peninsula to disturbances.
To overcome the issues associated with conducting ecological assessments based on remote sensing images, this research using the Random Sample Consensus Algorithm (RANSAC) proposed by Fischler and Bolles in 1981 to address the impact of remote sensing noise on correlation analysis. RANSAC exhibits strong robustness to noisy data, and has been widely applied in measurement engineering, image matching, and point cloud orientation [
26,
27]. GeoDetector is a statistical method for detecting spatial differentiation and revealing the driving forces behind it [
28], which can identify the interactions between driving factors and has been well applied in the analysis of EEQ driving factors [
29,
30].
The objectives of this study are (1) to evaluate the EEQ of the Jiaodong Peninsula from 2000 to 2020 using the RSEI model and analyze spatial clustering characteristics through LISA clustering; (2) to comprehensively analyze the relative contributions of human activities and natural factors to EEQ using RANSAC and GeoDetector, providing references for promoting the green and high-quality growth of the Jiaodong Peninsula urban agglomeration.
3. Results
3.1. Spatial and Temporal Evolution of RSEI
Figure 3 illustrates the distribution and proportional areas of RSEI levels across the Jiaodong Peninsula. In 2000, the average RSEI value was 0.60. By 2010, it decreased to 0.57, indicating a 5% decline. However, by 2020, the average RSEI value had risen to 0.66, representing a notable increase of 15.79%. Over the years 2000, 2010, and 2020, areas classified as “Good” and “Excellent” predominated, with proportions of 56.48%, 51.02%, and 67.17%. The proportion of areas classified as “Moderate” increased in 2010 but decreased below the 2000 level by 2020. Conversely, the proportions of “Worst” and “Poor” levels decreased annually, comprising 26.43%, 23.79%, and 19.31% of the total area.
From 2000 to 2010, regions experiencing a strong improvement in RSEI were primarily concentrated in the eastern hilly areas of the Jiaodong Peninsula, while areas exhibiting serious degradation were mainly observed in Weifang and Qingdao. From 2010 to 2020, regions showing a strong improvement in RSEI were predominantly distributed in the eastern hilly areas of the Jiaodong Peninsula, with fewer areas demonstrating serious degradation, which were scattered.
Using the standard deviation ellipse and centroid migration analysis method, this study investigates the spatial distribution characteristics of different levels of RSEI across the Jiaodong Peninsula (
Figure 4). The major axis of the standard deviation ellipse indicates the primary direction of RSEI distribution across various levels, while the minor axis signifies the extent of RSEI distribution. The distribution direction of each level of RSEI is consistent, trending from southwest to northeast. For the Good and Excellent levels, the major axis elongates and the minor axis shortens, with the centroid shifting towards the northeast. For the Moderate level, the sizes of the major and minor axes of the standard deviation ellipse remain stable, with minimal changes in the centroid position. For the Worst and Poor levels, the major axis of the standard deviation ellipse shortens, the minor axis widens, and the centroid shifts towards the southwest.
At the urban scale, the rankings of average RSEI values in 2000 were as follows: Rizhao > Weifang > Qingdao > Yantai > Weihai. In 2010, the rankings were: Rizhao > Weifang > Qingdao > Yantai > Weihai. By 2020, the rankings changed to: Rizhao > Weihai > Yantai > Qingdao > Weifang.
Figure 5 illustrates the proportional distribution of RSEI levels across the five cities of the Jiaodong Peninsula urban agglomeration. The proportions of Good and Excellent areas increased to over 65% in each city, with Rizhao notably reaching 73.85% by 2020. The Moderate areas witnessed an increase in all cities by 2010 followed by a subsequent reduction by 2020. Except for Weifang, the proportions of the Worst and Poor areas reached their lowest levels by 2020.
3.2. Spatial Disparity Analysis of RSEI
The LISA cluster map of RSEI is shown in
Figure 6.
2000: H-H clustering areas are primarily located in the mountainous and hilly regions of Rizhao and Weifang cities in the western part of the Jiaodong Peninsula, as well as the Mount Lao district of Qingdao. L-L cluster areas are concentrated in the coastal urban areas and bay regions of each city, with significant L-L clustering also observed in the mountainous and hilly areas of the western part of the Jiaodong Peninsula, especially in Weihai city.
2010: H-H cluster areas remained prominent in the mountainous and hilly terrains of Rizhao and Weifang cities, as well as in the Mount Lao district of Qingdao, albeit with a noticeable decrease in spatial extent. L-L cluster areas continued to appear along the coastal urban zones of each city, although their overall coverage decreased. Notably, clustering within the urban precincts of Weifang and Qingdao expanded during this period, contrasting with the general trend. Other regions did not exhibit significant clustering tendencies.
2020: H-H cluster areas were predominantly distributed in hilly terrains, while the coverage of L-L cluster areas along the coastline notably diminished in Weihai and Yantai cities. L-L cluster areas primarily in the urban zones of Qingdao, Weifang, and the bays of Jiaozhou and Laizhou. The majority of remaining regions did not exhibit pronounced clustering patterns.
3.3. Single-Factor Correlation Analysis Based on RANSAC
In
Figure 7, the black and red lines represent the results of least squares (OLS) and RANSAC fitting, respectively. The gray points in the figure denote the inliers fitted by the RANSAC algorithm, while the blue points represent the corresponding outliers or noise points. The coefficient of determination (
) is used to assess the fitting performance of the algorithms. The
values of the RANSAC algorithm consistently exceed those of the OLS method, suggesting the superior capability of the RANSAC algorithm in capturing the influence of factors on the evolution of RSEI.
RANSAC analysis reveal a negative correlation between human activities and RSEI, while the association between natural factors and RSEI demonstrates interannual variations.
The negative correlation between LUCI and RSEI reached its lowest point in 2000 but increased annually thereafter. By 2020, it reached a peak with an R2 value of 0.56, the highest among all impact factors.
The for PD and RSEI was relatively low in 2000, reaching its peak in 2010 at 0.25. By 2020, the value for POP ranked second among all impact factors. The negative impact of population aggregation and urbanization on EEQ has become increasingly apparent.
The correlation between GDP and RSEI peaked in 2010. In contrast, the correlation between GDP and RSEI weakened in 2020 compared to PD.
In 2000, SPEI exhibited a positive correlation with RSEI, highlighting the importance of agricultural drought on EEQ. However, this correlation turned negative by 2010 and became inconspicuous by 2020.
AMP exhibits a positive correlation with RSEI, with its value in 2000 only slightly lower than that of SPEI. However, by 2010, the correlation between AMP and RSEI was relatively low. In 2020, the of AMP reached the highest level among the natural factors.
In 2000, there was a positive correlation between AMT and RSEI, although with a low level of correlation. In 2010 and 2020, AMT showed no significant correlation with RSEI.
3.4. Driving Force Analysis Based on GeoDetector
Table 7 displays the statistical results of GeoDetector’s factor detector.
2000: Natural factors have a greater impact on RSEI compared to human activities. SPEI ranks higher than AMP, with AMT following as the third most influential factor.
2010: Human activities exert a stronger influence on RSEI than natural factors. LUCI had the strongest influence on RSEI evolution. GDP and SPEI are ranked second and third in terms of their importance to RSEI.
2020: Human activity factors dominate the top three spots in the ranking of factor importance. The q-value of LUCI reaching 0.45, with POP’s importance surpassing that of GDP.
The interaction detector results exhibited double-factor and nonlinear enhancement among driving factors (
Figure 8).
2000: The SPEI∩GDP interaction, followed by AMP∩GDP, had the greatest impact on EEQ, with a q value reaching 0.358.
2010: The LUCI∩SPEI interaction, followed by LUCI∩AMT, had the greatest impact on EEQ, with a q value reaching 0.383.
2020: The LUCI∩GDP interaction, followed by LUCI∩AMT, had the greatest impact on EEQ, with a q value reaching 0.494.
The double-factor and nonlinear enhancement among human factors gradually intensified, reaching its peak in 2020, while, among nature factors, it decreased annually in 2000, 2010, and 2020.
4. Discussion
4.1. Impact of Natural Factors on EEQ
According to previous studies, the impacts of human activities and natural factors on ecosystems exhibit significant spatial variations [
50,
51]. In 2000, natural factors outweighed human activity factors in influencing EEQ; SPEI was the predominant determinant of EEQ in the Jiaodong Peninsula (
Table 7). The results of the correlation analysis showed that, as SPEI values rise, indicative of alleviation of agricultural drought conditions, RSEI will correspondingly increase (
Figure 8).
During 1999–2000, northern China experienced extreme drought disaster due to anomalous circulation over the Eurasian continent [
52,
53]. During this period, the SPEI values were at their lowest in the eastern Jiaodong Peninsula, indicating the most severe drought conditions. This inhibited vegetation growth and resulted in a significant decline in NDVI, which contributed the most to RSEI (
Figure 2 and
Table 3). In 2010 and 2020, no extreme drought events occurred, leading to a notable enhancement in EEQ within the eastern hilly regions (
Figure 3 and
Figure 4). It should be noticed that the importance of SPEI exhibited a gradual decline in 2010 and 2020 compared to 2000 (
Figure 8 and
Table 7), potentially attributed to advancements in agricultural management practices, including irrigation technology. Hossain et al. [
54] propose that, due to disparities in plant drought resilience and agricultural management practices like irrigation, mild drought may not necessarily elicit significant impacts on the ecological environment, indicating that positive human activities can counteract the negative impacts of natural factors. Hua et al. [
55] observed that, even in areas with the same land cover type, the eastern regions exhibit higher sensitivity of vegetation activity to drought compared to western regions due to differences in climate and hydrological characteristics. This explains the concentration of L-L cluster areas of RSEI in the eastern regions of the Jiaodong Peninsula in 2000 (
Figure 3 and
Figure 5).
Compare to SPEI, the importance of AMP and AMT is relatively minor (
Table 7), This could be attributed to moderate precipitation and temperature changes, which may not significantly affect enzyme activity within plant tissues. Vegetation exhibits significant decline only when conditions exceed their adaptive range [
56,
57]. Interaction detection results indicate significant enhancement effects for SPEI∩AMT and AMP∩AMT (
Figure 8). Prolonged drought leads to insufficient soil moisture replenishment and exacerbated transpiration due to high temperatures, both of which have detrimental effects on the ecological environment.
With global warming, studies suggest an increased frequency of extreme weather events in the Jiaodong Peninsula [
58]. Extreme droughts, floods, frosts, and other natural disasters can lead to reduced vegetation coverage, decreased crop yields, and other ecological issues [
59]. Governments should pay attention to enhancing climate event monitoring and warning capabilities, controlling greenhouse gas emissions, and building infrastructure to address climate change. These efforts are necessary to improve climate disaster management and respond promptly to potential ecological risks.
4.2. Impact of Human Activities on EEQ
With the development of the social economy, the impact of human activities on EEQ has gradually increased. In 2010, with favorable natural conditions, the EEQ of the Jiodong Peninsula reached 0.57 (
Figure 3). L-L cluster areas were primarily in the urban zones of cities and the coastal zone (
Figure 6). LUCI, followed by GDP, became the most significant influencing factor (
Table 7). Industrialization and urbanization led to excessive population concentration in cities, resulting in a reduction in per capita natural and social resources, creating immense pressure on the ecological environment of the Jiaodong Peninsula, and even brought about higher risks to life safety [
60].
With the increasing awareness of sustainable development and the growing need for a better urban living environment, Shandong Province and the five cities of Jiaodong Peninsula have formulated a series of targeted ecological protection policies over the past 20 years. By 2020, with favorable natural conditions, the EEQ of the Jiaodong Peninsula had increased to 0.67. However, in the urban zones of Weifang City and coastal ecological sources such as Jiaozhou Bay, the L-L cluster areas continued to increase (
Figure 6). The interaction effect of land use with other driving factors reached its peak in 2020 (
Figure 8).
To mitigate the negative effects of human activities on the complex coastal ecosystem of the Jiaodong Peninsula, we propose the following recommendations: (1) In urban zones where L-L cluster areas are mainly concentrated, human activities should be regulated and more green spaces should be planned within the cities to control the expansion of urban areas. (2) In hilly regions, human intervention should be reduced to maintain the integrity of natural landscapes and protect natural resources such as forests and farmland. Attention should be paid to natural restoration in important natural reserves or ecological sources like Lao Mountain, and artificial intervention methods such as cloud seeding can be employed if necessary. (3) In important coastal areas like Jiaozhou Bay and Laizhou Bay, strict regulations should be imposed on excessive development activities such as artificial land reclamation. More ecological enhancement projects such as increasing the area of wetland parks should be implemented. (4) Balancing urban and rural development and providing fair access to education and healthcare are effective in easing urban population pressure.
4.3. Limitations and Prospects
Compared to single-factor analysis, RSEI can provide richer information on the ecological environment. However, RSEI is only applicable to terrestrial areas, and indicators such as the distribution of water resources and environmental pollution cannot be incorporated into the assessment method, indicating a need to enrich the dimensions of RSEI. In future research, we should consider incorporating carbon emissions and environmental pollution factors into the assessment system to construct a more comprehensive ecological remote sensing assessment method.
This research using Landsat data with a spatial resolution of 30 m for the years 2000, 2010, and 2020. Therefore, we are unable to provide the nuanced changes over a twenty-year period of EEQ. Due to challenges in obtaining annual data on land cover and GDP, we were unable to investigate the temporal lagged effects of driving factors at a consistent temporal resolution. In the future, with improved data availability, we will consider extending the research to include historical data or future projections.
5. Conclusions
This study utilized RSEI to evaluate EEQ and employed standard deviation ellipses, centroid migration, and LISA clustering methods to assess the spatiotemporal distribution changes and spatial clustering effects of EEQ in the Jiaodong Peninsula for the years 2000, 2010, and 2020. To mitigate noise interference in remote sensing images, we applied the RANSAC algorithm, resulting in single-factor correlations between human and natural factors and RSEI. Additionally, GeoDetector was utilized to quantify the contributions of human and natural factors to EEQ and their interactions during the same years.
The results indicate that the Jiaodong Peninsula generally exhibits favorable EEQ, with average RSEI values of 0.60, 0.57, and 0.65 for 2000, 2010, and 2020, with Good and Excellent level areas comprising the largest proportion. For Good and Excellent RSEI levels, the centroid shifted towards the northeast. The cluster map of RSEI revealed that H-H cluster areas were primarily located in hilly regions, while L-L cluster areas were mainly found in urban and coastal areas around Laizhou Bay and Jiaozhou Bay.
The driving analysis results revealed that, in 2000, the significance of nature factors exceeded that of human factors. SPEI emerged as the predominant determinant of EEQ, with a q value reaching 0.222. However, in 2010 and 2020, the negative effects of human factors gradually intensified, with LUCI identified as the most influential driving factor, having q values of 0.191 and 0.450. The driving factors exhibited double-factor and nonlinear enhancement. The most significant interaction factors affecting the EEQ of the Jiaodong Peninsula in 2000, 2010, and 2020 were identified as SPEI∩GDP, LUCI∩SPEI, and LUCI∩GDP.
Governments and environmental protection agencies should implement measures such as improving climate disaster management, increasing urban green space planning, strengthening coastal ecological enhancement projects, and protecting critical ecological sources to enhance the resilience of the complex coastal ecosystem, promoting the sustainable development of the Jiaodong Peninsula.