The study area ( Figure 1 ) covered four major estuarine deltas in China (Liao River, Yellow River, Yangtze River and Pearl River).

The Liao River Estuary (LRE; 40° 40’–41° 25’ N, 121° 25’–123° 55’ E) is located in the southern Liaohe Plain and to the north of Liaodong Bay in the Bohai Sea. It is the largest warm-temperate coastal wetland in Asia and has an average annual temperature of 8.4°C and average annual precipitation of 623 mm (Li et al., 2021).

The Yellow River Estuary (YRE; 37° 59’–37° 91’ N, 119° 09’–119° 39’ E) is located at the junction of the Bohai Sea and Laizhou Bay. The wetland here is the best preserved, most typical and youngest wetland ecosystem in this warm temperate region. It has an average annual temperature of 12.1°C and average annual precipitation of 552 mm (Cui et al., 2009).

The Yangtze River Estuary (YTRE; 30° 42’–31° 56’ N, 121° 36’–122° 12’ E) is mainly located in Shanghai and Nantong, Jiangsu Province. It has an average annual temperature of 15.3°C and average annual precipitation of 1022 mm (Zhao et al., 2009).

The Pearl River Estuary (PRE; 21° 52’–22° 46’ N, 112° 58’–114° 3’ E) contains waters near Lingdingyang Bay, Huangmao Sea, Hengqin Island and Nanshui Island. The surrounding land, from east to west, features the cities of Hong Kong, Shenzhen, Dongguan, Zhongshan, Zhuhai and Macao. Since the 1980s, China began to implement a policy of reform and opening-up, giving priority to the development of coastal areas. Gradually, a coastal economic belt centred on the PRE was formed. It has an average annual temperature of 21.9°C and average precipitation of 1647 mm (Lu et al., 2020).

Landsat 5 surface reflectance datasets for 2000, 2005, 2010 and Landsat 8 surface reflectance datasets for 2013, 2015, 2018 and 2020 were obtained via the GEE platform. These datasets contained Landsat data with radiometric and atmospheric corrections. Pixel values in the visible, near-infrared and short-wave infrared bands are corrected surface reflectance, and pixel values in the thermal infrared band are corrected temperature values at the sensor. To avoid cloud cover degrading the data quality, we selected images with< 20% cloud cover in a year; except in the PRE (a subtropical monsoon climate zone with more cloud cover), for which the threshold was< 50%. Then, based on pixel quality assessment (QA) band data provided by GEE, the clouds and shadows in the original data were removed. Finally, the mean values of these data were used to synthesize good-quality surface reflectance data, which was prepared for the calculation of ecological parameters. The technical process is shown in Figure 2 and the specific methods are introduced in Section 2.3.

Land-use/cover data for 2000–2020 (Yang and Huang, 2021) were used to analyse the relationship between land cover and the RSEI. Land cover was divided into nine categories: cropland, forest, shrub land, grassland, water, snow and ice, barren land, impervious land, and wetland. The overall accuracy of the original data in China reached 79.31%. Based on the characteristics of the research area and Landsat and Google Earth images, we reclassified the data into forest, barren land, built-up land, grassland, cropland, natural waterbodies, reservoirs and ponds. There were some errors in the original data, which were manually revised using visual interpretation. In addition, estuarine wetland data (Mao et al., 2020) were downloaded from National Earth System Science Data Center (http://www.geodata.cn/data/datadetails.html?dataguid=267682210636266&docId=10). The overall accuracy of the original data for China reached 95.1%. In order to ensure the accuracy of these data, we also modified the wetland data by combining Landsat and Google Earth images with visual interpretation.

Figure 3 shows the RSEI values in the four estuaries from 2000 to 2020. To further explore the temporal and spatial variations in RSEI, it was divided into five grades: 0.0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8 and 0.8–1 (for the classification criteria, refer to Yao et al., 2022). If the RSEI grade of a certain pixel did not change, it was marked as 0; if it increased by 1, 2, 3 or 4 grades, it was marked as 1, 2, 3 or 4, respectively, and if it decreased by 1, 2, 3 or 4 grades, it was marked as −1, −2, −3 or −4. In our results ( Figure 4 ), there are no pixels that changed by 4 grades, so the results range from −3 to 3. Through Figures 3 , 4 , we can obtain the following results:

In summary, the regions with deterioration in eco-environmental quality in the LRE and YRE were mainly distributed in the coastal zone, while deterioration in the YTRE and PRE mainly occurred in urban expansion areas.

Figure 5 shows the RSEIs of the four estuarine wetland regions from 2000 to 2020. The results show that:

To analyse spatial autocorrelation, we used 1 km × 1 km grid resampled data and analysed four typical estuaries using Moran’s I and LISA. An RSEI vs Moran’s I scatter diagram is shown in Figure 6 . The points are mainly distributed in quadrants 1 and 3, indicating that ecological environmental quality in the study area was positively correlated with its spatial distribution. The Moran’s I values of the four regions in 7 years are shown in Table 1 , which all show positive correlated and aggregation distribution characteristics. From 2000 to 2020, the Moran’s I values of the YRE and PRE remained stable, while those of the LRE and the YTRE showed downward trends. The areas with deteriorating eco-environmental quality in the YRE were distributed in patches (mainly in aquaculture areas), while the eco-environmental quality in the PRE was relatively stable, so the Moran’s I values in these two estuaries were relatively stable. The eco-environmental quality of the LRE increased significantly but decreased in some regions, so the spatial aggregation decreased. The eco-environmental quality of the YTRE decreased and the patches of change were scattered ( Figure 4C ), and the spatial aggregation decreased significantly.

To explore the spatial characteristics of the RSEI, we generated LISA clustering maps ( Figure 7 ). The RSEI showed a low-low aggregation in coastal zones and urban construction zones, and high-high aggregation in forests, grassland and other areas with relatively lush vegetation. High-low and low-high clustering were rare and sporadic in the four study areas, indicating that none had extreme environmental quality. We further analysed the LISA clustering changes and obtained the following results:

Comparing the distribution maps of RSEI and land cover ( Figure 8 ), it can be found that the RSEI was lower in barren, urban and built-up regions and higher in cropland, grasslands and forest.

All raster data were resampled to 1 km × 1 km, then the NDVI, WET and RSEI were projected into a three-dimensional space together ( Figure 10A ), as were LST, NDBSI and RSEI ( Figure 10B ). Figure 10 shows that the RSEI increased with NDVI and WET, so RSEI is positively related to greenness and humidity, which is consistent with our own observations and other studies (Xiong et al., 2021). The RSEI decreased as LST and NDBSI increased, indicating that RSEI is negatively correlated with heat and dryness, which is also consistent with the actual situation and the results of Yao et al. (2022).

RESI is developed under the conceptual framework of pressure-state-response (PSR), which is one of the most commonly used environmental conceptual frameworks. (Niemeijer and de Groot, 2008). Within the framework of PSR, eco-environmental problems can be expressed as three different but interrelated index types: anthropogenic pressure (P), eco-environmental state (S) and climate response (R). Among the four indicators comprising the RSEI, NDBSI is used to represent the pressure caused by human activities on an ecosystem, NDVI measures the ecological state of land before and after human-induced land surface changes, and LST and WET reflect the response of the ecosystem to the climate (Xu et al., 2019). Therefore, the RSEI has a clear physical meaning.

The RSEI is a combination of four indicators and may be difficult to use to quantify actual values on the ground. Hence, it is difficult to conduct a quantitative verification of the RSEI and only qualitative evaluations can be made. According to Figure 9 , the changes in RSEI values are relatively consistent with changes in land cover. Pixels converted from other types of land cover to built-up areas, for instance, may have a clearly defined boundary showing the extent of the change. However, the border of the resulting eco-environmental change may not perfectly follow the border of land cover changes. Land cover changes may also influence the ecological conditions of adjacent pixels, which could result in changes to hot and wet surface conditions (Xu et al., 2019).

Many previous studies (Murray et al., 2014; Yim et al., 2018; Chen et al., 2019; Ma et al., 2019) have investigated how natural and human factors affect the temporal and spatial dynamics of coastal wetlands. Studies have shown that the ecological quality of coastal wetlands is affected by complex natural and human dynamic factors. Natural factors include runoff, temperature, precipitation, waves, and tidal currents, which affect the area and type of coastal wetlands and indirectly affect the biological composition of coastal wetlands (Shi et al., 2020; Zang et al., 2020). Human factors include beach reclamation, port development and water conservancy projects (Wang et al., 2014; Jiang et al., 2015). The purpose of reclamation is to transform cropland and develop salt farms and aquaculture. The large amount of tidal flat reclamation causes a decrease in the tidal flats area and changes the coastline length, intensifying the effect of tidal water on the coastal environment. It also reduces the habitat of intertidal organisms, directly leading to ecological deterioration (Sun et al., 2017). Port construction will straighten the coastline so that its resistance to marine dynamic action is weakened (Liu et al., 2018). Port construction leads to the expansion of coastal towns and the destruction of coastal wetlands (Wu et al., 2022). Water conservancy projects are used to intercept sediment, which directly reduces the total amount of sediment entering the river section downstream of the reservoir. It also directly reduces incoming water and sand in the coastal basin, which have great impacts on the coast, especially plain coast (Liu et al., 2013). Dam construction damages the plant habitat of coastal wetlands (Newton et al., 2020).

This study analysed the impacts of runoff, sediment discharge, and the CLUDD and SLUDD of built-up land on the ecological quality of four estuarine wetlands. Runoff and sediment discharge in the LRE increased significantly from 2000 to 2010 and remained stable from 2015 to 2020 ( Figure 11 ), which is consistent with the RSEI trend. This indicates that both runoff and sediment discharge improved the eco-environmental quality of the LRE. Although YRE runoff was relatively stable from 2003 to 2010, sediment discharge had a downward trend, which is one of the reasons for the deteriorating eco-environmental quality during the study period. Due to a lack of adequate sediment supply, the development of the continental body in the delta was inhibited; eventually, fewer new wetlands were formed, the flora became further away from the sea and the ecological quality decreased (Yang, 2005). The runoff and sediment discharge of the YTRE and PRE were much higher than those of the LRE and YRE. The runoff of the two rivers remained stable and the sediment discharge decreased and then became stable, and the impact on eco-environmental quality was not as significant as that of the LRE and YRE.

CLUDD represents the rate of land-use change, while the SLUDD of built-up land reflects the rate of built-up land change. These two indicators reflect the degree of human activities such as land reclamation and port construction. Both CLUDD and SLUDD were low in the LRE and were relatively stable from 2000 to 2020 ( Figure 12 ), indicating that activities such as reclamation and port construction had little impact on eco-environmental quality. Under natural conditions, such as the increases in runoff and sediment discharge, the eco-environmental quality improved during this period. CLUDD and SLUDD increased first and then decreased in the YRE from 2000 to 2020, which is consistent with the RSEI trend, indicating that human activities had a great impact on the eco-environmental quality of the YRE, which was most severe during 2000-2010. According to Figures 8B, F , 9B , reclamation activities in the YRE related to aquaculture increased significantly. Wang et al. (2021) also reported that land reclamation and coastal zone development (conversion to aquaculture ponds) were the main causes of wetland loss in the YRE. Both CLUDD and SLUDD in the YTRE and the PRE had significant downward trends from 2000 to 2020, which demonstrates relatively stable eco-environmental quality in these regions. It must be noted that urbanization is still the main driver of the retreat of YTRE and PRE wetlands (Wang et al., 2021), but its influence is gradually decreasing. Due to the restriction of human disturbance and the expansion of Spartina alterniflora saltmarshes, the eco-environmental quality of coastal wetlands in Changxing Island and Jiuduansha Island gradually improved (Yang et al., 2018). Governmental conservation and restoration projects increased the area of mangroves, which also contributed to improvements in the eco-environmental quality of the PRE (Ma et al., 2019).

Overall, the ecological and environmental quality of coastal wetlands in China has increased in the past decade. Wang et al. (2021) also showed that a significant reduction in human activities since 2012 led to an increase in coastal wetland area. It is worth noting that in 2012, “ecological civilization construction” was incorporated into the national strategy, and the State Council of China approved the National Marine Function Zone (2011–2020) to improve the management of reclamation projects and rationally control their scale (Liu et al., 2018). Reduced human disturbance and greater government protection have contributed to the restoration of coastal wetlands in China since 2012 (Wang et al., 2021).

In our study, the RSEI was obtained by batch processing of data on the GEE platform to achieve rapid assessment of wetland eco-environmental quality based on Landsat images acquired during 2000–2020. The results are objective and easy to analyse, which facilitates large-scale ecological quality assessment of wetlands.

However, there are some deficiencies in our research. Firstly, the ecological environment is complex, so using only four RSEI indicators can only approximate its true status. We suggest that future studies use more diverse spatial data, such as net primary productivity (NPP) and aerosols. Secondly, the RSEI is primarily used in land areas and is not suitable for large bodies of water (e.g., oceans). Through the tasseled cap transformation, the water components are primarily related to vegetation and soil water. If there were more water in the study area, and the calculated moisture component will not accurately reflect the moisture contents of vegetation and soil. The water in this case has to be masked off. Therefore, this study cannot consider the large-scale ecological status of aquatic environments. In future studies, we suggest that water quality inversion results based on remote sensing, marine NPP and RSEI data should be integrated to evaluate the eco-environmental quality of coastal wetlands.