Mapping of Ecological Environment Based on Google Earth Engine Cloud Computing Platform and Landsat Long-Term Data: A Case Study of the Zhoushan Archipelago

: In recent years, with the rapid advancement of China’s urbanization, the contradiction between urban development and the ecological environment has become increasingly prominent, and the urban ecological system now faces severe challenges. In this study, we proposed an ecological index-based approach to monitor and evaluate the ecological environment using a


Introduction
The ecological environment refers to the sum of the various ecological factors that affect human production, life, and ecosystem development, and it is closely related to social sustainable development [1][2][3]. In recent years, human activities have had an increasingly strong impact on the global ecological environment, leading to a continuous decline in the ability of the ecosystem to recover and self-purify, as well as the destruction of the environment in which humans depend on for survival [4][5][6]. With the transformation of China's economy from high-speed development to high-quality development and urbanization in the new era, China is attaching increasing importance to the ecological environment's protection [7,8]. Therefore, rapid and accurate monitoring and evaluation of the ecological environment can provide a reliable scientific basis for regional environmental governance and sustainable development [9][10][11].
Before the rise of remote sensing technology, monitoring the ecological environment mainly relied on traditional methods such as fixed-point monitoring and on-site investigation [12]. These monitoring methods can obtain real-time data for a region with targeted accuracy, but they are expensive, inefficient, lack long-term observational data, and have difficulty obtaining data for a large area and a long-term series of ecological environment monitoring [13,14]. Compared to traditional methods, satellite-based remote sensing technology can provide large-area, long-term, and continuous observations, thereby quickly and efficiently obtaining the spectral characteristics of the spatially distributed land features; thus, it has become an important means of monitoring changes in land surface morphology and human activity impacts.
The first category is the single-index evaluation method, which evaluates a specific ecosystem, such as forest, urban, and arid areas, using a single remote sensing index, such as normalized difference vegetation index (NDVI) [31], vegetation cover fraction (PVF) [32], net primary productivity (NPP) [33], land surface temperature (LST) [34], normalized difference impervious surface index (NDISI) [35], normalized multi-band drought index (NMDI) [36], and normalized difference drought index (NDDI) [37,38]. Although this method is simple and easy to implement, the ecological environment is a comprehensive and dynamic system, and a single indicator cannot fully reflect its changes. Therefore, using a single remote sensing index to characterize the state of the ecological environment is a one-sided approach [39]. The second category is based on the Ecological Environment Condition Evaluation Technical Specification (HJ/T 192-2015) published by the Chinese Ministry of Environmental Protection. Using this method, the ecological environment evaluation indicator system is constructed by comprehensively considering the biological abundance index, vegetation coverage index, water network density index, land degradation index, and environmental quality index [40][41][42]. Most of the indicators used in this index system are derived from annual environmental statistical data regarding the study area, which have difficulty reflecting the spatial changes in the ecological environment [43,44]. The third category is the remote sensing-based ecological index (RSEI) proposed by Hanqiu Xu [45]. This index is based on four important indicators that reflect the natural ecological environment-greenness, wetness, heat, and dryness-which are used to construct an ecological environment evaluation model [46]. The model is based on remote sensing data and can objectively and quickly evaluate the urban ecological environment. It has been widely used in the evaluation of ecological environments in various regions [47][48][49]. For example, Xu [50] used an improved RSEI to investigate the land cover types and ecological conditions in the Xiongan New Area and predict the impact Remote Sens. 2023, 15,4072 3 of 30 of population growth on the average ecological conditions. Zhu [51] took the Wuhan Urban Development Zone as an example and proposed a locally adaptive RSEI (RSEILA) based on the remote sensing-based ecological index to evaluate the ecological environment status of the Wuhan Urban Development Zone from 2013 to 2019. This method restricts the geographic scope of the study area by setting a window, which makes the evaluation results more accurate. Jia [52] used the Google Earth Engine (GEE) platform to invert a modified remote sensing-based ecological index (MRSEI) to study the changes in the ecological environmental quality of the Qaidam Basin from 1986 to 2019. In addition, through the assistance of meteorology and socio-economic auxiliary data, the main factors affecting the changes in the ecological environment were explored. Xiong [53] constructed an RSEI based on the GEE platform to study the spatiotemporal changes in the ecological environment in the Erhai Lake area, which provided a valuable reference for the study of the interactions between human activities and ecosystem services in a watershed system. Firozjaei provided a new method based on the RSEI-impervious surface percentage (ISP) feature space to comprehensively evaluate the social and economic poverty levels in Budapest and the surrounding rural areas in Europe [54,55]. Musse combined remote sensing data and census data, referred to some indicators of the RSEI, and proposed a quantitative method to evaluate the urban environmental quality index (UEQI) in Cali, Colombia [56]. In summary, many studies have focused on the evaluation of the ecological environments of large-and medium-sized cities, while studies of island ecological environments are relatively scarce [57][58][59][60]. An island is a land area surrounded by water on all sides that is located above the water level during high tide periods [61,62]. Due to their unique characteristics, the ecological environments of islands are extremely sensitive to the dual impacts of high-intensity human activities and global change [63][64][65].
In this study, we aimed to map an ecological environment based on the Google Earth Engine cloud computing platform and Landsat long-term data and analyze the spatial variations and temporal evolution characteristics of the Zhoushan Archipelago's ecological environment from 1985 to 2020. This study can provide scientific evidence and technical support for the dynamic monitoring of the ecological environment and the formulation of regional sustainable development policies.
The paper is organized as follows: Section 1 gives an overview of research into the monitoring of the ecological environment using remote sensing technology; Section 2 describes the methodology used; Section 3 discusses the study area and data sources; Section 5 discusses three aspects: rationality, uncertainty, and prospect; and Section 6 is the conclusion, summarizing the results and analyzing the future work.

Methods
In this study, the RSEI proposed by Xu [44] was utilized to conduct remote sensing monitoring and evaluation of the ecological environment in the Zhoushan Archipelago using the GEE cloud-based platform. Firstly, the surface reflectance Landsat data for the period 1985-2020 were obtained using the GEE cloud-based platform, and the images for each year from May to October were selected and synthetized, clouds and shadows were removed, and vector cropping was conducted to obtain an image covering the entire study area. Then, the ecological indicators of greenness, humidity, heat, and dryness were calculated based on the NDVI, Wet, LST, and NDBSI. The results were standardized, and principal component analysis was conducted on each index to select the first principal component required to calculate the RSEI. A spatial distribution map of the RSEI with a resolution of 30 m for the Zhoushan Archipelago from 1985 to 2020 was then drawn to study the changes in the mean value of the RSEI and the conversion of the ecological environment grades. Finally, the Theil-Sen median trend analysis, Mann-Kendall test, and Hurst index were used to identify the spatiotemporal change characteristics of the ecological environment of the Zhoushan Archipelago. The time-series stability of the RSEI was tested using the coefficient of variation method. The technical route used in this study is illustrated in Figure 1.

Data Pre-Processing
In this study, the remote sensing data pre-processing was conducted using the GEE platform. Firstly, the surface reflectance data from Landsat 5 for the period 1985-2010 and Landsat 8 for 2015-2020 were selected, and images for each year from May to October were chosen. Then, the cloud and shadow related pixels were masked using the cloudmasking technique [66] to extract the cloud shadow and cloud cover fields from the quality assurance (QA) band and establish a masking function. This step was taken to remove the cloud-covered areas in each image and obtain a minimum cloud cover image dataset for the target year that covered the entire study area. The images were further overlaid using the median composite method, through which the median function in the GEE calculated the median of each pixel in the image stack to generate a new image.  [45] combined four parameters closely related to the natural ecological environment to construct a remote sensing-based ecological index, namely greenness, humidity, heat, and dryness. The meanings and methods of calculating of each ecological factor used in the index are described below.

Calculation of
(1) Greenness index Vegetation is an extremely important factor that reflects the quality of the ecological environment in a region. The greenness indicator was expressed by the normalized difference vegetation index (NDVI), which could represent the growth status of plants, the distribution of the vegetation density, and the coverage of vegetation [67]. The formula was defined as follows: where ρ NIR is the reflectance of the near-infrared band in remote sensing data, and ρ R is the reflectance of the infrared band in remote sensing data. In Landsat 8 data, they corresponded to the 5th and 4th bands, respectively. In Landsat 5 data, they corresponded to the 4th and 3rd bands, respectively.
(2) Humidity index The humidity can reflect the moisture content of vegetation and soil [68][69][70]. In this study, the humidity component in the tasseled cap transformation was used instead of the humidity index, being denoted as tasseled cap wet. The formulas used to calculate the humidity index based on Landsat TM/OLI images was as follows: (2) where Wet TM is the humidity result obtained using the Landsat 5 satellite data, and WET OLI is the humidity result obtained using Landsat 8 satellite data. ρ B , ρ G , ρ R , ρ NIR , ρ SWR1 , and ρ SWR2 are the reflectances of the blue, green, red, near-infrared, shortwave infrared 1, and shortwave infrared 2 bands, respectively.
(3) Dryness index Soil drying caused by developed land and bare soil can seriously harm the ecological environment in a region. In this study, the NDBSI, which reflected the degree of soil drying, was calculated by combining an index-based built-up index (IBI) and a normalized bare soil index (SI) [71]. The formula used was as follows: where IBI and SI are the index-based built-up index and normalized bare soil index, respectively. The specific formulas used to calculate these indices were as follows: where ρ B , ρ G , ρ R , ρ NIR , ρ SWR1 , and ρ SWR2 are the reflectances of the blue, green, red, nearinfrared, shortwave infrared 1, and shortwave infrared 2 bands, respectively.
(4) Heat index The LST is an important component of the Earth's energy budget and an important parameter representing the surface environment. In this study, the obtained LST was used to represent the heat index. Regarding the Landsat series of satellites, Landsat 5 has one thermal infrared band (band 6, 10.40-12.50 µm), and the TIRS sensor on the Landsat 8 has two thermal infrared bands (band 10, 10.60-11.19 µm; band 11, 11.50-12.51 µm). Band 10 is located in a lower atmospheric absorption region than band 11 and has a higher atmospheric transmission accuracy. Therefore, band 6 of Landsat 5 and band 10 of Landsat 8 were selected as the channels used to perform the LST inversion.
The statistical mono-window model (SWM) was used to invert the LST [72]. The calculation of the land surface emissivity adopted the vegetation cover method [73], and the fractional vegetation cover was calculated using the NDVI. The specific formulas used to calculate the SWM were as follows: where L 6 is the radiance-calibrated thermal infrared band reflectance, and T b is the temperature value at the sensor. DN is the gray value of the data pixel, and gain and bias are the band gain value and bias values, respectively, which can be obtained using the image header. K 1 and K 2 are calibration parameters, which can be obtained by referring to the user manual (for Landsat 5, K 1 = 607.76 W × m −2 × um −1 × sr −1 and K 2 = 1260.56 K; for Landsat 8, K 1 = 774.89 W × m −2 × um −1 × sr −1 and K 2 = 1321.08 K). ε is the land surface emissivity, which was calculated by applying a threshold to the NDVI according to Sobrino's method [74]. pv is the fractional vegetation cover, NDVI is the normalized difference vegetation index, and NDVI min is the minimum value of the NDVI, which represents the NDVI value of completely bare soil or areas with no vegetation cover. NDVI max is the maximum value of the NDVI, which represents the NDVI value of pure vegetation pixels.

Construction of RSEI
Using remote sensing technology to calculate the vegetation index, humidity component, LSY, and NDBSI, which represent the greenness, humidity, heat, and dryness, respectively, and are closely related to the surface ecological environment, the RSEI was constructed. The specific formula used to calculate the RSEI was as follows: where NDVI, Wet, LST, and NDBSI represent the normalized difference vegetation index, humidity component obtained via the tasseled cap transformation, the land surface temperature, and the normalized difference bare soil index, respectively. f is the RSEI as a function of these four indexes, and in this paper, principal component analysis (PCA) was used to integrate these four ecological indexes. PCA is a multidimensional data compression technique that uses orthogonally rotated coordinate axes and linear transformations to condense information derived from multiple variables into a few characteristic components [76]. Using this method to synthesize multiple indicators can avoid the bias caused by subjective factors in the weighting process and ensure the representativeness of the obtained principal components [77]. To avoid imbalance in weights caused by non-standardized units, the ecological factors were normalized to the range of (0-1) before performing PCA, and the formula used in normalization iwass as follows: where NI i is the result of normalized processing of the index, and I i , I min , and I max are the values of the ith pixel of the index, the minimum value, and the maximum value, respectively. Based on the results of the PCA, the first component, i.e., PC1, integrated the various ecological factors and contained information about the vast majority of the ecological indexes, meaning that it could be used to characterize the quality of the ecological environment of the ground corresponding to the pixel. To ensure that a higher PC1 value represented a better ecological condition, we could obtain the initial ecological index RSEI 0 [78] by subtracting PC1 from Equation (1). The formula is as follows: To facilitate measurement and comparison of the indices, RSEI 0 was normalized as follows: where RSEI f is the remote sensing-based ecological index constructed, and its value ranges from 0 to 1. The closer the RSEI value was to 1, the better the ecological condition; the closer the value was to 0, the worse the ecological condition [79]. The ecological environmental quality of the Zhoushan Archipelago was analyzed at five-year intervals from 1985 to 2020. The ArcGIS raster reclassification tool was used to classify the ecological environmental quality of the study area into five levels, creating a spatial distribution map of the ecological environmental quality in the Zhoushan Archipelago. The proportion of the areas occupied by each level of ecological environmental quality and the total area were then calculated. Finally, two RSEI grading maps at five-year intervals were overlain to generate the ecological environment grading transition matrix. Based on this matrix, the ecological environment grading transition Sankey diagram was generated.

Analysis of Ecological Environment Change Trend
Trend analysis compared the same indicators or ratios recorded in different time periods, directly observed their changes and magnitudes, examined their development trends, and predicted their future development prospects [80]. In the time-series trend analysis, the Mann-Kendall (M-K) test was combined with the Theil-Sen median trend analysis method to analyze the change trend of the ecological environment in the study area from 1985 to 2020. The Mann-Kendall test is a non-parametric statistical test method used to determine the significance of trends. It does not require the sample to follow a certain distribution and is not affected by a few outliers [81]. The computational formula used to perform the Mann-Kendall test was as follows: Set {RSEI i }, i = 1985, 1990,. . . , 2015, 2020.
The Z statistic was defined as follows: where RSEI i and RSEI j are the RSEI values of pixel i in year i and year j, respectively; n is the length of the time series; and sgn is the sign function. The value range of the Z statistic is the real set R. At the given significance level α, when |Z| > µ 1− ∂ 2 , there was a significant change in the research sequence at level α [82]. In this paper, α was set at 0.05 to determine the significance of the RSEI time-series trend at a confidence level of 0.05.
Theil-Sen median trend analysis is a robust non-parametric method used to calculate trends that can reduce the influence of data outliers [83]. The Theil-Sen median trend analysis method calculates the median of the slope of n(n − 1)/2 combinations of data. The computational formula used was defined as follows: when S RSEI > 0, the RSEI time series exhibited an increasing trend, and when S RSEI < 0, the RSEI time series exhibited a decreasing trend.

Analysis of Ecological Environmental Change Sustainability
Sustainability analysis predicts future trends based on existing changes. In this study, the sustainability of the ecological environment time-series change was analyzed based on the Hurst index. The Hurst index, which is also known as the index dependency or index long-term dependence, can quantify the relative trend of a time series. Rescaled range (R/S) analysis was first proposed by Hurst (1951) for the analysis of hydrological data for the Nile and has subsequently been developed. Among the many algorithms used to estimating the Hurst index, the most famous example is the R/S method used by Mandelbrot and Wallis based on Hurst's hydraulic research results [84]. The R/S method is used to analyze long-term time-series correlations and has wide applications in hydrology, economics, climatology, geology, and geochemistry [85]. Its basic principle is as follows [86].
For a time series {RSEI(t)}, t = 1, 2, . . . , n, the mean series was defined as follows: The formula for cumulative deviation was The formula for the range sequence was The formula for the standard deviation sequence was By taking the ratio of R(τ) to S(τ), we obtained The Hurst exponent, which was denoted as H, ranged from 0 to 1 and characterized the existence of the Hurst phenomenon in a time series. The value of H was used to determine whether the RSEI sequence was completely random or exhibited persistence based on the following three scenarios: (1) When H > 0.5, the process exhibits sustainable properties, that is, the future trend is consistent with the past trend, and the stronger the persistence is, the closer H is to 1. (2) When H = 0.5, the RSEI time series is a random sequence, without long-term correlation. (3) When H < 0.5, the time series exhibits non-sustainability, and the stronger the non-sustainability, the closer H is to 0 [87].

Temporal Stability Analysis of the Ecological Environment
The coefficient of variation is the ratio of the standard deviation to the mean of the original data, reflecting the absolute degree of dispersion of the data. A larger value indicates a more dispersed data distribution and greater inter-annual fluctuations. Conversely, a smaller value indicates a more concentrated data distribution, smaller inter-annual fluctuations, and a greater temporal stability in the time series. By calculating the coefficient of variation in the RSEI in the Zhoushan Archipelago, we determined the degree of dispersion of the RSEI values in the study area and the stability of the long-term RSEI time series. The calculation formula used was as follows [88]: where CV is the coefficient of variation, n is the number of years, i is the year index, and RSEI i is the remote sensing-based ecological index of the ith year.

Study Area
The Zhoushan Archipelago is located on the coast of the East China Sea and bordered by Hangzhou Bay to the west, Shanghai to the north, and Ningbo to the south ( Figure 2). It serves as a significant geographic gateway to the Yangtze River Basin and the Yangtze River Delta [89]. Geographically, the Zhoushan Archipelago is characterized by a low hilly landscape, with approximately 37.4% of the terrain consisting of hills, and a complex distribution of water systems, including underdeveloped surface water systems [65]. It has a subtropical monsoon humid climate, having a moderate climate throughout the year, distinct seasons, warm winters, and cool summers. The average annual temperature is 16 • C. The vegetation in the archipelago is dominated by evergreen broad-leaved forests in the Chinese subtropical region, which primarily consist of mixed forests of broad-leaved trees and shrubs [90]. Additionally, the archipelago is home to abundant biological resources and plays an important role as a habitat for island-dwelling birds.
Zhoushan Island is the largest in the Zhoushan Archipelago and the third largest in the country. It is home to 74% of the total population of Zhoushan City and serves as the administrative and economic center of the urban area located in the Zhoushan Archipelago [92]. Due to the unique geographic location of the islands, the fragile ecological environment in the Zhoushan Archipelago is susceptible to frequent interactions between land, sea, atmosphere, human activities, natural disasters, intensified urbanization, and global climate change [93].

Data Sources
In this study, Landsat series satellite images were used. The Landsat series is one of the longest-operating optical remote sensing satellite systems and has the advantages of a long time series, high spatial resolution, and high data availability [94,95]. Since the 1970s, a total of nine Landsat satellites have been launched, all of which have successfully acquired a substantial amount of Earth observation data, except for Landsat 6, which failed to launch [96,97]. The latest Landsat satellite, known as Landsat 9, was launched in September 2021 [98,99]. It will continue to provide long-term data products, achieve the goal of continuously monitoring land surface landscape changes in the Landsat plan, and continue the long-term Landsat observation record [100,101]. Currently, the Landsat series has become an important information source for the long-term monitoring of land use and land cover changes and the ecological environments of Earth [102][103][104]. Zhoushan City was the first prefecture-level city to be established in island form in China, covering a total of 2085 islands, representing 20% of China's islands [91]. Moreover Zhoushan Island is the largest in the Zhoushan Archipelago and the third largest in the country. It is home to 74% of the total population of Zhoushan City and serves as the administrative and economic center of the urban area located in the Zhoushan Archipelago [92]. Due to the unique geographic location of the islands, the fragile ecological environment in the Zhoushan Archipelago is susceptible to frequent interactions between land, sea, atmosphere, human activities, natural disasters, intensified urbanization, and global climate change [93].

Data Sources
In this study, Landsat series satellite images were used. The Landsat series is one of the longest-operating optical remote sensing satellite systems and has the advantages of a long time series, high spatial resolution, and high data availability [94,95]. Since the 1970s, a total of nine Landsat satellites have been launched, all of which have successfully acquired a substantial amount of Earth observation data, except for Landsat 6, which failed to launch [96,97]. The latest Landsat satellite, known as Landsat 9, was launched in September 2021 [98,99]. It will continue to provide long-term data products, achieve the goal of continuously monitoring land surface landscape changes in the Landsat plan, and continue the long-term Landsat observation record [100,101]. Currently, the Landsat series has become an important information source for the long-term monitoring of land use and land cover changes and the ecological environments of Earth [102][103][104].
In this study, surface reflectance (SR) data from Landsat 5 thematic mapper (TM) sensors for the years 1985,1990,1995,2000,2005, and 2010 and Landsat 8 operational land imager/thermal infrared sensor (OLI/TIRS) for 2015 and 2020 were selected through the GEE cloud-based platform. Landsat images acquired during the growing sensor (from May to October) were selected, and we proposed annual composite images using median values to ensure the seasonal consistency of the temporal analysis. The cloud and shadow mask algorithm on the GEE cloud-based platform was then applied to remove clouds and shadows from the obtained images, achieving the goal of eliminating anomalous values and improving the image quality. The data used in this study are described in Table 1.

RSEI Calculation for the Zhoushan Archipelago
Based on satellite remote sensing data used to characterize ecological environmental factors, such as greenness, humidity, temperature, and dryness, the RSEI was calculated using PCA. The results of the PCA for the study area from 1985 to 2020 are presented in Table 2, with the analyzed indicator values of the first to fourth principal components denoted as PC1, PC2, PC3, and PC4, respectively.  1985,1990,1995,2000,2005,2010,2015, and 2020, respectively. (2) Compared to the other components, the first principal component concentrated over 65% of the characteristic information of each indicator and could integrate the information of each indicator well, representing the ecological environmental characteristics of the region. Therefore, it was used to construct the RSEI and characterize the ecological environmental status of the region.
Based on the calculation results of the remote sensing-based ecological index for each year of the study period, a spatial distribution map of the RSEI with a 30-meter resolution for the Zhoushan Archipelago from 1985 to 2020 was generated ( Figure 3).

2020
Eigenvalue 0.0735 0.0085 0.0012 0 Percent eigenvalue 88.25% 10.25% 1.50% 0.00% As can be seen in Table 2, (1) the contribution rates of the four indicators to the first principal component exceeded 65% in all years, with values of 67.95%, 66.54%, 66.61%, 74.10%, 74.14%, 77.71%, 89.94%, and 88.25% recorded in 1985,1990,1995,2000,2005,2010,2015, and 2020, respectively. (2) Compared to the other components, the first principal component concentrated over 65% of the characteristic information of each indicator and could integrate the information of each indicator well, representing the ecological environmental characteristics of the region. Therefore, it was used to construct the RSEI and characterize the ecological environmental status of the region.
Based on the calculation results of the remote sensing-based ecological index for each year of the study period, a spatial distribution map of the RSEI with a 30-meter resolution for the Zhoushan Archipelago from 1985 to 2020 was generated ( Figure 3).  As can be seen in Figure 3, the range of the remote sensing-based ecological index values from 1985 to 2020 in the Zhoushan Archipelago was 0-1. The areas with relatively high RSEI values were mainly located in the interior regions of the major islands in the archipelago, while the areas with relatively low RSEI values were mainly distributed in the eastern part of Zhoushan Island, the southwestern part of Jintang Island, the northern part of Daishan Island, the southern part of Qushan Island, the northern and northwestern parts of Zhujiajian, and the northwestern and eastern parts of Liuheng Island. To quantitatively analyze the changes in the RSEI during the study period, the mean and standard deviation of the RSEI and the four component indexes for each year were calculated, and a line graph recording the changes in the remote sensing-based ecological index in the Zhoushan Archipelago from 1985 to 2020 was created. The results are presented in Table  3 and Figure 4.  As can be seen in Figure 3, the range of the remote sensing-based ecological index values from 1985 to 2020 in the Zhoushan Archipelago was 0-1. The areas with relatively high RSEI values were mainly located in the interior regions of the major islands in the archipelago, while the areas with relatively low RSEI values were mainly distributed in the eastern part of Zhoushan Island, the southwestern part of Jintang Island, the northern part of Daishan Island, the southern part of Qushan Island, the northern and northwestern parts of Zhujiajian, and the northwestern and eastern parts of Liuheng Island. To quantitatively analyze the changes in the RSEI during the study period, the mean and standard deviation of the RSEI and the four component indexes for each year were calculated, and a line graph recording the changes in the remote sensing-based ecological index in the Zhoushan Archipelago from 1985 to 2020 was created. The results are presented in Table 3 and Figure 4.  As can be seen in Table 3 and Figure 4, the overall ecological environment of the Zhoushan Archipelago was good from 1985 to 2020, but the mean RSEI exhibited a decreasing trend over time. The mean RSEI decreased from 0.7719 in 1985 to 0.7566 in 2000, and it further decreased to 0.5817 in 2020. This declining trend indicates that the ecological environment of the Zhoushan Archipelago has deteriorated by approximately 25% in the past 35 years. The decline was relatively slow from 1985 to 2000, though it was more severe from 2000 to 2020. During the study period, the standard deviation of the RSEI was relatively small, indicating that the data concentration was high, and the research results were more reliable.

Rank Division of Ecological Environment
To better analyze the changes in the ecological environment of the Zhoushan Archipelago, it was necessary to rank the RSEI.  (Table 4 and Figure 5).  As can be seen in Table 3 and Figure 4, the overall ecological environment of the Zhoushan Archipelago was good from 1985 to 2020, but the mean RSEI exhibited a decreasing trend over time. The mean RSEI decreased from 0.7719 in 1985 to 0.7566 in 2000, and it further decreased to 0.5817 in 2020. This declining trend indicates that the ecological environment of the Zhoushan Archipelago has deteriorated by approximately 25% in the past 35 years. The decline was relatively slow from 1985 to 2000, though it was more severe from 2000 to 2020. During the study period, the standard deviation of the RSEI was relatively small, indicating that the data concentration was high, and the research results were more reliable.  (Table 4 and Figure 5).  Table 4 and Figure 5 show the area in of the ecological environment levels in the Zhoushan Archipelago from 1985 to 2020. The statistical results show that the proportion of areas with excellent ecological environments decreased by 38.83%, while the proportion of areas with poor ecological environments increased by 8.53%. Therefore, the ecological environment of the Zhoushan Archipelago exhibited an overall decreasing trend. The specific analysis is as follows: (1) From 1985 to 2020, the ecological environment of the Zhoushan Archipelago was mainly at the excellent level, having an average of 49.80% per year, but the proportion of the excellent areas decreased over time. Indeed, the proportion of the excellent ecological environment was highest in 1985, accounting for 67.46% of the total area, and it was lowest in 2020, accounting for 28.63% of the total area. (2) From 1985 to 2020, the proportion of areas with good ecological environments in the Zhoushan Archipelago was second only to those with excellent ecological environments. Except for 1985, it exceeded one-fifth of the total area, having an average of 24.21%. Indeed, the proportion of areas with good ecological environments was highest in 2010, accounting for 33.73%, while it was lowest in 1985, accounting for 16.46%. (3) From 1985 to 2020, the proportion of areas with a general ecological environment in the Zhoushan Archipelago ranked third, having an average of 12.17%. Indeed, the proportion of areas with general ecological environments was highest in 2020, accounting for 16.68%, while it was lowest in 1990, accounting for 7.25%. (4) From 1985 to 2020, the proportions of areas with poor and relatively poor ecological environments in the Zhoushan Archipelago were relatively small, having averages of 4.60% and 9.22%, respectively. Except for 2015 and 2020, the sum of the areas with these two levels was less than 16% in all years, but their overall proportions exhibited an increasing trend. Indeed, the sum of the areas with poor and relatively poor ecological environments was highest in 2020, accounting for 27.67% of the total area. The sum of the areas with poor and relatively poor ecological environments was smallest in 1985 and 2000, accounting for 7.64% and 7.62% of the total area, respectively.   Table 4 and Figure 5 show the area in of the ecological environment levels in the Zhoushan Archipelago from 1985 to 2020. The statistical results show that the proportion of areas with excellent ecological environments decreased by 38.83%, while the proportion of areas with poor ecological environments increased by 8.53%. Therefore, the ecological environment of the Zhoushan Archipelago exhibited an overall decreasing trend. The specific analysis is as follows: (1) From 1985 to 2020, the ecological environment of the Zhoushan Archipelago was mainly at the excellent level, having an average of 49.80% per

Level of Conversion of Ecological Environment
In order to further analyze the spatiotemporal changes in the ecological environment of the Zhoushan Archipelago, the classified RSEI results for each study period were overlaid, and a Sankey diagram ( Figure 6) and an ecological environment level transition matrix ( Figure 7) were created to illustrate the results.  The Sankey diagram clearly describes the direction and volume of the changes in the ecological environment levels, which quantitatively expresses the transition relationship  The Sankey diagram clearly describes the direction and volume of the changes in the ecological environment levels, which quantitatively expresses the transition relationship The Sankey diagram clearly describes the direction and volume of the changes in the ecological environment levels, which quantitatively expresses the transition relationship between the ecological environment levels during different periods. It can be concluded based on Figure 6 that the characteristics of the ecological environment level conversions every 5 years can be summarized as follows: the transition between every two adjacent levels was relatively drastic. In particular, the transitions between excellent and good, good and general, general and relatively poor, and relatively poor and poor ecological environments were significant transitions.
The ecological environment level transfer matrix visually reflects the specific areas in which ecological environment level changes occurred, and it has great significance for quantifying the transitions between different ecological environment levels in each period. In this study, the ecological environment level transfer matrix was created every 5 years from 1985 to 2020 (Figure 7). It can be concluded based on Figure 7 that there were significant changes in the ecological environment levels in the Zhoushan Archipelago from 1985 to 2020. (1) During the study period, the conversion from the excellent to the good level was dominant, with the largest period of conversion occurring from 2010 to 2015, during which time 131.3087 km 2 were converted from the excellent to the good level. The main category converted to the excellent level was the good level, with a maximum conversion area of 186.8389 km 2 (2005-2010). (2) The main category converted from the good level was the excellent level, followed by the conversion from the good to the general levels. The maximum area converted from the good to the general level was 33.8996 km 2 (2015-2020), and the maximum area converted from the general to the good level was 79.0286 km 2 (2010-2015). (3) The main areas converted from the general level was converted to the good and relatively poor levels. The maximum area converted from the relatively poor to the general level was 83.2385 km 2 (2010-2015), and the maximum area converted from the general to the relatively poor level was 37.2699 km 2 (2015-2020). (4) Except for the conversion between the general and relatively poor levels, the conversion from and to of the relatively poor level mainly occurred with regard to the poor levels. The maximum area converted from the poor to the relatively poor level was 97.3449 km 2 (2010-2015), and the maximum area converted from the relatively poor to the poor level was 36.3026 km 2 (2015-2020). (5) Aside from the conversion of the very poor level, the amount of area converted from the poor level to other levels was relatively small, with the minimum conversion amount occurring during the period 1995-2000, during which time the area converted to the good level was only 0.0047 km 2 . The conversion to the poor level mainly occurred from the relatively poor and general levels, and the maximum area converted from the general to the poor level was 11.6758 km 2 (2015-2020).

Change Trend of Ecological Environment
By combining Theil-Sen median trend analysis and the Mann-Kendall test, we effectively analyzed the change trend and spatial distribution characteristics of the ecological environment in the Zhoushan Archipelago from 1985 to 2020. Since there were basically no regions with SRSEI values strictly equal to 0 in this study, the regions with SRSEI values of −0.0005 to 0.0005 were defined as regions with stable and unchanged ecological environments, regions with SRSEI values of greater than 0.0005 were defined as regions with an improved ecological environment, and regions with SRSEI values of less than −0.0005 were defined as regions with a degraded ecological environment, thus dividing the ecological environmental changes in the study area into three different categories. In the significance analysis, after setting the significance level at 0.05, the significance of the RSEI time-series trend change was assessed. The results of the significance test were divided into significant changes (Z > 1.96 or Z < −1.96) and slight changes (−1.96 ≤ Z ≤ 1.96). The results of the Theil-Sen median trend analysis and the Mann-Kendall test were superimposed to obtain the RSEI change trend results at the pixel scale from 1985 to 2020. The results were divided into five types of change (Table 5). As can be seen in Table 5, the area with an improved ecological environment accounted for 12.60% of the total area, the stable and unchanged ecological environment areas accounted for 3.05% of the total area, and the areas with a degraded ecological environment accounted for 84.35% of the total area. Therefore, most of the areas in the Zhoushan Archipelago exhibited a trend of ecological environmental degradation from 1985 to 2020. Indeed, 50.25% of the study area had a slightly degraded ecological environment, while 34.10% of the area had a seriously degraded ecological environment. The proportion of the areas with significant improvement in the ecological environment was the smallest, accounting for only 0.84% of the total area.
As can be seen in Figure 8, the area of ecological environment degradation in the Zhoushan Archipelago far exceeded the area of ecological environment improvement from 1985 to 2020. The areas with significantly degraded ecological environments were mainly located in the southern part of Zhoushan Island, the southwestern part of Jintang Island, the northwestern area of Zhujiajian Island, the northern area of Taohua Island, the central part of Liuheng Island, and the western parts of Qushan Island, Shengsi Island, and Yangshan Island. The areas with slightly degraded ecological environments were mainly located in the central part of Zhoushan Island, the eastern parts of Jintang Island and Zhujiajian Island, and the southern parts of Dachangtu Island and Yushan Island. The areas with stable ecological environments were scattered throughout the archipelago. The areas with slightly improved ecological environments were mainly located in the northeastern and eastern parts of Zhoushan Island, the southern parts of Taohua Island and Zhujiajian Island, the southwestern area of Liuheng Island, and the northern area of Qushan Island. The areas with significantly improved ecological environments were scattered within the inner areas of the slightly improved areas.  As can be seen in Table 5, the area with an improved ecological environment accounted for 12.60% of the total area, the stable and unchanged ecological environment areas accounted for 3.05% of the total area, and the areas with a degraded ecological environment accounted for 84.35% of the total area. Therefore, most of the areas in the Zhoushan Archipelago exhibited a trend of ecological environmental degradation from 1985 to 2020. Indeed, 50.25% of the study area had a slightly degraded ecological environment, while 34.10% of the area had a seriously degraded ecological environment. The proportion of the areas with significant improvement in the ecological environment was the smallest, accounting for only 0.84% of the total area.
As can be seen in Figure 8, the area of ecological environment degradation in the Zhoushan Archipelago far exceeded the area of ecological environment improvement from 1985 to 2020. The areas with significantly degraded ecological environments were mainly located in the southern part of Zhoushan Island, the southwestern part of Jintang Island, the northwestern area of Zhujiajian Island, the northern area of Taohua Island, the central part of Liuheng Island, and the western parts of Qushan Island, Shengsi Island, and Yangshan Island. The areas with slightly degraded ecological environments were mainly located in the central part of Zhoushan Island, the eastern parts of Jintang Island and Zhujiajian Island, and the southern parts of Dachangtu Island and Yushan Island. The areas with stable ecological environments were scattered throughout the archipelago. The areas with slightly improved ecological environments were mainly located in the northeastern and eastern parts of Zhoushan Island, the southern parts of Taohua Island and Zhujiajian Island, the southwestern area of Liuheng Island, and the northern area of Qushan Island. The areas with significantly improved ecological environments were scattered within the inner areas of the slightly improved areas.

Sustainability Analysis of Ecological Environment
The Hurst exponent was used to explore the sustainability of the change trends of the ecological environment of the Zhoushan Archipelago. The results show that the average Hurst index of the RSEI is 0.65, with the areas in which the Hurst index is less than 0.5 accounting for 13.39% of the total area and the areas in which it is greater than 0.5 accounting for 86.61% of the total area. These results indicate that the ecological environmental changes in the Zhoushan Archipelago have a strong positive persistence, meaning it will have a strong tendency to continue its original change trend in the future.
The Theil-Sen median trend analysis quantitatively reflects the change trend of the RSEI during a given period, while the Hurst exponent qualitatively predicts whether the change trend is sustainable. The sustainability of the RSEI changes in the Zhoushan Archipelago is obtained by superimposing the results of the RSEI change trend on the Hurst exponent results. In this paper, the results are categorized into four levels ( Table 6 and Figure 9).

Sustainability Analysis of Ecological Environment
The Hurst exponent was used to explore the sustainability of the change trends of the ecological environment of the Zhoushan Archipelago. The results show that the average Hurst index of the RSEI is 0.65, with the areas in which the Hurst index is less than 0.5 accounting for 13.39% of the total area and the areas in which it is greater than 0.5 accounting for 86.61% of the total area. These results indicate that the ecological environmental changes in the Zhoushan Archipelago have a strong positive persistence, meaning it will have a strong tendency to continue its original change trend in the future.
The Theil-Sen median trend analysis quantitatively reflects the change trend of the RSEI during a given period, while the Hurst exponent qualitatively predicts whether the change trend is sustainable. The sustainability of the RSEI changes in the Zhoushan Archipelago is obtained by superimposing the results of the RSEI change trend on the Hurst exponent results. In this paper, the results are categorized into four levels ( Table 6 and Figure 9).   As can be seen in Table 6 and Figure 9, in most areas of the Zhoushan Archipelago, the ecological environment exhibited a trend of degraded sustainably, having an area proportion of 73.40%. The regions in which the ecological environment exhibited a trend of sustainably stability accounted for 2.65% of the total area, and these regions were scattered in the central part of Zhoushan Island and the western part of Taohua Island. The regions in which the ecological environment exhibited a trend of improved sustainably accounted for 10.56% of the total area and were mainly located in the eastern and northeastern parts of Zhoushan Island; the northern, northwestern, and southern parts of Zhujiajian Island; the southern part of Taohua Island; the southwestern part of Liuheng Island; the northern part of Yangshan Island; and the southern and northern parts of Qushan Island. The area proportion in which the trend of the change was unsustainable was 13.39%, and these areas were scattered within the central areas of the islands, such as the central parts of Zhoushan island. In conclusion, the protection of the ecosystem in the Zhoushan Archipelago should be given due attention. In particular, the regions with a sustainable degradation trend require special attention.

Time-Series Stability Analysis of Ecological Environmental Evolution
To determine the time-series stability of the ecological environmental changes in the Zhoushan Archipelago recorded in this study, the coefficient of variation in the RSEI from 1985 to 2020 was calculated. The coefficient of variation values for each pixel were then counted and used to create a cumulative frequency graph ( Figure 10).
As can be seen in Table 6 and Figure 9, in most areas of the Zhoushan Archipelago, the ecological environment exhibited a trend of degraded sustainably, having an area proportion of 73.40%. The regions in which the ecological environment exhibited a trend of sustainably stability accounted for 2.65% of the total area, and these regions were scattered in the central part of Zhoushan Island and the western part of Taohua Island. The regions in which the ecological environment exhibited a trend of improved sustainably accounted for 10.56% of the total area and were mainly located in the eastern and northeastern parts of Zhoushan Island; the northern, northwestern, and southern parts of Zhujiajian Island; the southern part of Taohua Island; the southwestern part of Liuheng Island; the northern part of Yangshan Island; and the southern and northern parts of Qushan Island. The area proportion in which the trend of the change was unsustainable was 13.39%, and these areas were scattered within the central areas of the islands, such as the central parts of Zhoushan island. In conclusion, the protection of the ecosystem in the Zhoushan Archipelago should be given due attention. In particular, the regions with a sustainable degradation trend require special attention.

Time-Series Stability Analysis of Ecological Environmental Evolution
To determine the time-series stability of the ecological environmental changes in the Zhoushan Archipelago recorded in this study, the coefficient of variation in the RSEI from 1985 to 2020 was calculated. The coefficient of variation values for each pixel were then counted and used to create a cumulative frequency graph ( Figure 10). As can be seen in Figure 10, the coefficient of variation was mainly concentrated between 0 and 0.4, having an average value of 0.1627 and a standard deviation of 0.1467. The pixels with a coefficient of variation of 0.04 had the largest number of occurrences, having a frequency of nearly 350,000. The second largest frequency was for pixels with a coefficient of variation of 0.06, which had a frequency of greater than 250,000. This result indicates that the RSEI values in the Zhoushan Archipelago were relatively concentrated between 1985 and 2020, with small inter-annual fluctuations and a relatively stable time series.
According to Jenks [105], there are natural breakpoints and turning points in any data series, which are statistically significant and can be used to divide the research object into groups with similar properties. Therefore, natural breakpoints are good boundaries for classification. The natural breaks method is a method of classifying data based on natural breakpoints. It determines the best arrangement of values in a group by iteratively As can be seen in Figure 10, the coefficient of variation was mainly concentrated between 0 and 0.4, having an average value of 0.1627 and a standard deviation of 0.1467. The pixels with a coefficient of variation of 0.04 had the largest number of occurrences, having a frequency of nearly 350,000. The second largest frequency was for pixels with a coefficient of variation of 0.06, which had a frequency of greater than 250,000. This result indicates that the RSEI values in the Zhoushan Archipelago were relatively concentrated between 1985 and 2020, with small inter-annual fluctuations and a relatively stable time series.
According to Jenks [105], there are natural breakpoints and turning points in any data series, which are statistically significant and can be used to divide the research object into groups with similar properties. Therefore, natural breakpoints are good boundaries for classification. The natural breaks method is a method of classifying data based on natural breakpoints. It determines the best arrangement of values in a group by iteratively comparing the sum of squares of the differences between each grouping and the mean of the elements in the grouping to the observed values. Based on the calculated best arrangement, the cutoff points present in an ordered distribution are determined, and the elements are divided into multiple categories by setting boundaries at the cutoff points. This classification method can minimize the sum of the squared differences between groups, maximize the variance between groups, and minimize the variance within groups, achieving the goal of classification.
To explore the time-series stability of the ecological environmental changes in the study area, the coefficient of variation results were classified into five categories using the natural breaks method: very stable (0 < CV ≤ 0.10), relatively stable (0.10 < CV ≤ 0.21), slightly variable (0.21 < CV ≤ 0.33), moderately variable (0.33 < CV ≤ 0.48), and highly variable (0.48 < CV ≤ 1.0). The spatial distribution of the coefficient of variation values is shown in Table 7 and Figure 11.  As can be seen in Table 7 and Figure 11, there are spatial differences in the time-series stability of the ecological environment changes. The areas with moderately variable and highly variable coefficients of variation account for a small proportion of the total area, i.e., 11.22% and 4.52%, respectively. These areas are mainly located in the northeastern part of Zhoushan Island, the northern part of Daishan Island, the southern and northern areas of Yangshan Island, the edges of Liuheng Island, the western area of Zhujiajian Island, and the northwestern, northeastern, and southwestern parts of Jintang Island. These coastal areas are relatively flat and greatly affected by natural disasters and human activities, leading to large fluctuations in the RSEI values. The areas that are very stable and relatively stable account for a large proportion of the total area, i.e., 49.40% and 19.25%, respectively. These areas are mainly located in the central areas of the major islands, such as the central areas of Zhoushan Island, Daishan Island, Liuheng Island, Taohua Island, Jintang Island, Xiushan Island, Qushan Island, and Changtu Island, as well as the small islands that have not been developed and utilized by humans. These areas have a high As can be seen in Table 7 and Figure 11, there are spatial differences in the time-series stability of the ecological environment changes. The areas with moderately variable and highly variable coefficients of variation account for a small proportion of the total area, i.e., 11.22% and 4.52%, respectively. These areas are mainly located in the northeastern part of Zhoushan Island, the northern part of Daishan Island, the southern and northern areas of Yangshan Island, the edges of Liuheng Island, the western area of Zhujiajian Island, and the northwestern, northeastern, and southwestern parts of Jintang Island. These coastal areas are relatively flat and greatly affected by natural disasters and human activities, leading to large fluctuations in the RSEI values. The areas that are very stable and relatively stable account for a large proportion of the total area, i.e., 49.40% and 19.25%, respectively. These areas are mainly located in the central areas of the major islands, such as the central areas of Zhoushan Island, Daishan Island, Liuheng Island, Taohua Island, Jintang Island, Xiushan Island, Qushan Island, and Changtu Island, as well as the small islands that have not been developed and utilized by humans. These areas have a high vegetation cover and are less affected by natural phenomena and human activities, maintaining good ecological conditions, thus resulting in small overall fluctuations and a stable time-series of RSEI values over time. The areas classified as slightly variable accounted for 15.61% of the total area and were mainly distributed in the transition zone between the relatively stable and moderately variable regions.

Discussion
In this study, long-term time-series Landsat satellite remote sensing data provided by the GEE platform were utilized to efficiently monitor and assess the ecological environment of the Zhoushan Archipelago, and the spatiotemporal variations were comprehensively investigated. Although the experimental results exhibit some degree of effectiveness, uncertainties also exist in the findings.

Rationality of Mapping of Ecological Environment of Islands Using Remote Sensing Images
Remote sensing is a technique used to monitor the Earth's surface from a distance, and the acquired remote sensing images can be applied through interpretation, analysis, and visualization [106][107][108]. Compared to other conventional methods, this method has several significant advantages. Firstly, remote sensing technology provides abundant information with a short information acquisition cycle, meaning that it is a rich data source for long-term ecological environment monitoring. Based on the Landsat satellite remote sensing data derived from 1985 to 2020 and integrated into the GEE cloud-based platform, an RSEI model was constructed to monitor and evaluate the ecological environment of the Zhoushan Archipelago. This model accurately detected the characteristics of the ecological environmental changes, and the results provided technical support for the formulation of regional ecological environment monitoring and development planning strategies. Secondly, remote sensing technology can achieve large-scale, long-term, and continuous observation at fixed points, making it more efficient than other traditional methods [109,110]. Based on the remote sensing technology, powerful online visualization calculation ability, and cloud storage characteristics of the GEE cloud platform [111][112][113][114], in this study, the quality of the regional ecological environment was objectively and rapidly evaluated by obtaining indicators closely related to the ecological environmental quality, such as greenness, humidity, dryness, and heat. Thirdly, remote sensing technology is characterized by comprehensiveness and can form a widely distributed detection network while observing the temporal, spatial, and spectral characteristics of objects, thereby obtaining true geographic information about objects within the regional space [115]. Compared to social statistical yearbook data, it can reflect the geographic distribution characteristics and relationships between things. For example, the ecological environment grading map created in this study based on remote sensing data reflects the spatial distribution characteristics of the different ecological environment levels in the Zhoushan Archipelago and lays the foundations for subsequent analysis of the spatiotemporal evolution of the ecological environment.

Uncertainties of Mapping of Ecological Environment of Islands Using Remote Sensing Images
Remote sensing technology can quickly monitor the ecological environment status of a region, but some factors affect the accuracy of ecological environment assessment. From the perspective of the inherent defects of remote sensing technology, the spatial resolution determines the size of the pixels covering the Earth's surface in satellite images, and the RSEI represents the ecological environmental status of the ground covered by the pixel [116]. However, the mixed pixels in the remote sensing images create certain differences between the acquired information and the true values, which may cause some uncertainties regarding the calculation of the RSEI [117,118]. In some areas of the Zhoushan Archipelago, the vegetation cover is high, and the NDVI value may be saturated, which may lead to the inaccurate calculation of the greenness index. In the future, the leaf area index (LAI) will be used instead of the NDVI to invert the greenness index of the remote sensingbased ecological index [119][120][121]. As developed land is often covered by impermeable surfaces, such as building roofs, paved roads, and parking lots [122][123][124], the normalized difference impervious surface index (NDISI) can enhance the characteristics of the soil and developed land that cause dry land surfaces compared to the NDBSI. Therefore, the NDISI can be used instead of the NDBSI to invert the dryness index and make the results more accurate.

Prospects of Remote Sensing Technology for Ecological Environmental Monitoring
Remote sensing data have become important sources of information used to study surface conditions and monitor ecological environments. In recent years, with the development of remote sensing technology to include multiple platforms, sensors, and angles; higher spatial and spectral resolutions; and quantitative remote sensing in practical applications, remote sensing technology is playing an increasingly important role in ecological environmental monitoring. On one hand, the combination of satellite remote sensing data and measured data can improve the accuracy of ecological environmental remote sensing monitoring. On the other hand, the fusion of space-time technology has the ability to enhance image quality, obtain temporal remote sensing data that can be used to describe the ecological conditions in different periods, and reveal the evolutionary characteristics of the ecological environment at the fine scale [125][126][127][128]. Moreover, new techniques, such as cloud computing, big data, and artificial intelligence, have injected new vitality into remote sensing image processing and provide innovative technical means of solving the large-scale automatic processing of and intelligent information extraction from satellite data, which can fully tap the value of massive remote sensing data and promote the rapid development of remote sensing technology [129][130][131].

Conclusions
Based on the GEE cloud-based platform and Landsat satellite remote sensing data, in this study, the NDVI, Wet, LST, and NDBSI were selected to characterize the closely related greenness, humidity, heat, and dryness of the ecological environment, respectively. The RSEI was constructed using PCA, and RSEI distribution maps with 30-meter spatial resolutions for eight time periods from 1985 to 2020 in the Zhoushan Archipelago were created. The trends and patterns of the spatiotemporal evolution of the ecological environment in the Zhoushan Archipelago were explored using the Theil-Sen median trend analysis, Mann-Kendall test, Hurst index, and coefficient of variation method. The spatial pattern, temporal trend, and temporal stability of ecological environmental evolution in the Zhoushan Archipelago from 1985 to 2020 were clarified. The main conclusions are as follows: (1) The average RSEI values for the eight time periods (at five-year intervals) from 1985 to 2020 were 0.7719, 0.7532, 0.7657, 0.7566, 0.7293, 0.6682, 0.6250, and 0.5817. Except for 2020, the RSEI values for all years were within the range 0.6-0.8. During the entire study period, the average RSEI value in the Zhoushan Archipelago decreased from 0.7719 to 0.5817, indicating an overall declining trend regarding the ecological environment in the study area. During the study period, the largest area of transition from general to good grade was 79.0286 km 2 , occurring from 2010 to 2015. (3) The ecological environment in the Zhoushan Archipelago exhibited co-existing degradation and partial improvement, with the areas with a degraded ecological environment accounting for 84.35% of the total area and the areas with an improved ecological environment accounting for 12.61% of the total area. The proportion of the areas with a significantly improved ecological environment was the smallest, accounting for only 0.84% of the total area. The proportion of the areas with a heavily degraded ecological environment was significant, accounting for 34.10% of the total area. These areas were mainly distributed in the southern part of Zhoushan Island, the southwestern part of Jintang Island, the northwestern area of Zhujiajian Island, the northern area of Taohua Island, the central part of Liuheng Island, and the western parts of Qushan Island, Shengsi Island, and Yangshan Island. The decline in the ecological environment in these areas was related to urbanization and the development of tourism resources in the Zhoushan Archipelago. (4) The results of the Hurst exponent analysis indicate that the change trend of the ecological environment in most regions of the Zhoushan Archipelago is sustainable. The proportion of the areas with degraded sustainably was 73.40%, and these areas were extensively distributed on the major islands in the Zhoushan Archipelago. The proportion of the areas with stable sustainably was 2.65%, and these areas were scattered in the central part of Zhoushan Island and the western part of Taohua Island. The proportion of the areas with improved sustainably was 10.56%, and these areas were mainly located in the eastern and northeastern parts of Zhoushan Island; the northern, northwestern, and southern parts of Zhujiajian Island; the southern part of Taohuajian Island; the southwestern part of Liuheng Island; and the northern part of Yangshan Island. The proportion of the unsustainable areas was 13.39%, and these areas were scattered within the central areas of the islands. In conclusion, the ecological environment in the areas with degraded sustainably requires particular attention. (5) The coefficient of variation in the RSEI sequence from 1985 to 2020 was mainly concentrated within the range 0-0.4, and its average value was 0.1627, indicating that the overall variations in the RSEI value in the Zhoushan Archipelago during the study period were relatively small, having a stable temporal trend. The regions with a high coefficient of variation were mainly concentrated in the northeastern part of Zhoushan Island, the northern part of Daishan Island, the edges of Liuheng Island, the western side of Zhujiajian Island, and the northwestern and southwestern parts of Jintang Island, indicating that the RSEI values in these areas fluctuated considerably over time.
The contributions of this study are that it provides a feasible method for ecological environmental monitoring in island areas, as well as data and technical support for the monitoring and evaluating of the regional ecological environment. However, this study only provides a simple description of the trend of the ecological environment in the Zhoushan Archipelago from 1985 to 2020, without offering specific analysis and discussion of the reasons and driving factors causing the changes. In the future, we will consider the relationship between the ecological environment and land use to better reveal the underlying causes of the ecological environment's evolution. In addition, using data with longer time series and more dense time intervals to more accurately explore characteristics and trends will represent a follow-up work.