The middle and lower YZR crosses 7 provinces in China, including Hubei Province, Hunan Province, Jiangxi Province, Anhui Province, Jiangsu Province, Zhejiang Province, and Shanghai City. The average annual temperature in this area varies from 14 ℃ to 18 ℃ and annual precipitation spans 1000 mm to 1500 mm. This region is a substantial production base for grains, vegetable oil, and cotton in China, and it is the most water-rich region in China. Famous freshwater lakes in the region include Poyang Lake, Dongting Lake, Taihu Lake, and Chaohu Lake. The lake and marsh areas are rich in aquatic biological resources. This area is also an essential industrial base in China, especially for the steel, machinery, electricity, textile, and chemistry industries.

We obtained land use and land cover data (for years 1980, 1990, 2000, 2010, 2015, and 2018), railroad data and road data, gross domestic product (GDP) data (2000, 2010, 2015), population density data (2000, 2010, 2015), normalized difference vegetation index (NDVI) data (2000, 2010, 2015), and digital elevation data (DEM) from the Resource and Environment Data Cloud Platform (http://www.resdc.cn/). The Land use data (Table 1) is derived from Landsat 8 and was interpreted visually, with a resolution of 1 kilometer. The overall evaluation accuracy of the first level of land use is greater than 93% (Ning et al., 2018). We obtained the land use and land cover data of future scenarios (A1B, A2, B1, B2) (Table 2) from the Geographical Simulation and Optimization Systems (GeoSOS) (Liu et al., 2017). This product used the Future Land Use Model to simulate global land use and land cover data based on future climate factors (emission scenarios of IPCC) (Nakicenovic et al., 2000), social-economic factors, and environmental factors. A total of 15 spatial driving factors were derived from the original dataset, including land use, human activities, climatic factors, soil information, and ecological factors. The overall accuracy of the simulated land use patterns is 0.75 for all land use types, Cohen’s Kappa coefficient is 0.67, and the FoM value is 19.62%. The future land use data includes six land use types: arable land, woodland, grassland, waterbody, built-up land, and unused land (Table 1).

The A1B scenario describes a future world of rapid economic growth, low population growth, and the rapid introduction of new and more efficient technologies, with a balanced emphasis on all energy sources. The A2 scenario describes a world in which each country is self-reliant, and the global population continuously increases. Economic development is primarily regionally oriented, and economic growth is more fragmented and technological change is slower than in the other scenarios. The B1 scenario describes a convergent world with the same global population that peaks in mid-century and declines thereafter, with rapid changes in economic structures toward a service and information economy, with reductions in material intensity and the introduction of clean and resource-efficient technologies. The emphasis is on global solutions to economic, social, and environmental sustainability, including improved equity, but without additional climate initiatives. The B2 scenario describes a world which emphasizes local solutions to economic, social, and environmental sustainability issues. It is a world with a continuously increasing global population at a rate lower than A2, intermediate levels of economic development, and with less rapid, but more diverse technological changes than in the B1 and A1B scenarios. While this scenario is oriented toward environmental protection and social equity, it focuses on the local and regional levels (Gaffin et al., 2002).

2.3.1 Land use change analysis

The land use change analysis assesses the quantitative change in each land use type and spatial distribution of land use conversion in the study area over the years. We used the “reclassify” function in ArcGIS 10.7 to reclass the land use and land cover data into six types and used the ‘intersect’ function in ArcGIS 10.7 to detect the conversion of land use types for different years.

2.3.2 The InVEST habitat quality model

Habitat quality in the InVEST model refers to the ability of the ecosystem to provide conditions appropriate for the persistence of individuals and populations. It is considered a continuous variable in the Model, ranging from low to medium to high, based on the resources available for survival, reproduction, and population persistence, respectively (Sharp et al., 2018). The habitat quality score ranges from 0 to 1. A higher habitat quality in the region will get a higher score. We divided the habitat quality score into 5 classes according to a previous study ([0, 0.2); [0.2, 0.4); [0.4, 0.6); [0.6, 0.8); [0.8, 1]).

The InVEST Habitat Quality model combines information on LULC and threats to biodiversity and produces habitat quality maps. There are four critical factors in this model: 1) each threat’s relative impact; 2) the sensitivity of each habitat type to each threat; 3) the distance between the habitats and the sources of the threats; and 4) the degree to which the land is legally protected. To assess the habitat quality accurately, we need to first select the threat factors and their impacts on habitat, and then determine the sensitivity of the habitats to the threats.

Based on previous research (Liu et al., 2019; Xu et al., 2019; Wang et al., 2020; Zhang et al., 2020), we chose built-up land, arable land, main roads, and railroads as the threat factors for the habitats (Table 3, Table 4). At the same time, we selected arable land, woodland, grassland, water, and unused land as the natural habitats for various creatures. Due to the data limitations, we did not add the protected land when analyzing the data.

The following equations give the impact of threat r in raster y to habitat x:

where i_rxy is the impact of threat r in raster y to habitat x; d_xy is the linear distance between grid cells x and y, and d_{r max} is the maximum sufficient distance of threat r’s reach across space.

D_xj gives the total threat level in grid cell x with LULC or habitat type j.

where D_xj is the habitat degradation or total threat level in grid cell x with LULC or habitat type j; R is the number of threat factors; r represents the threat layer; Y_r indicates the set of grid cell on r’s raster map; W_r indicates the weight of each threat factor (value range from 0 to1); r_y indicates the effect of threat r that originates in the grid cell; i_rxy indicates the distance between the habitat and the threat source and the impact of the threat across space; β_x is the factor that mitigates the impact of threats on habitat by environmental policies (here, β_x =1); S_jr indicates the sensitivity of LULC type j to threat factor r. The weights of threats are normalized so that the sum across all the threat weights equals 1. By normalizing the weights such that they sum to 1, we can think of D_xj as the weighted average of all threat levels in grid cell x. Note that the map of D_xj will change as the set of weights that are used changes.

A grid cell’s degradation score is translated into a habitat quality value using a half-saturation function, where the user must determine the half-saturation value, and as a grid cell’s degradation score increases its habitat quality decreases.

where Q_x,j is the quality of habitat in parcel x that is in LULC type j; H_j indicates the habitat suitability of LULC type j; k is the half-saturation constant; and z = 2.5.

2.3.3 Correlation analyses

We used the “Slope” and “Aspect” functions in ArcGIS 10.7 to extract the slope and slope direction as the topography of the study area. Due to the data constraints, we did not collect the NDVI data before 2000 or population data after 2015. We analyzed the relationships between GDP, population density, elevation, slope, slope direction, NDVI, and habitat quality from 2000 to 2005 in the study area by Pearson correlation analysis in SPSS 20.0. For all statistical tests, we selected P < 0.05 as the significance cut-off value.

All types of land other than cultivated land have increased from 1980 to 2018, and built-up land increased the most (Fig. 2, Fig. 3e). The cultivated land was steadily decreasing, while built-up land and waterbody area were steadily increasing. At the same time, other types of land fluctuated but increased overall.

The increasing occupation of arable land from 1980 to 2018 is conspicuous (Fig. 3b, c, e, Table l). The arable land decreased from 371.13×10³ km² in 1980 to 283.23×10³ km² in 2018, for a total change of 87.90×10³ km² (Fig. 2a). Most of the increased cultivated land was from woodland (55.77%, 38.04×10³ km²), which is mainly in the mountainous area of central China and southern Zhejiang Province (Fig. 3b, Table 5). About one-fourth (24.43%) of the increase in cultivated was from built-up land in urban areas of either the YZR Delta, Huaibei Plain, or Jianghan Plain, Dongting Lake Plain, and Poyang Lake Plain in the middle reach of the YZR (Fig. 3e, Table 5). Built-up land increased from 36.08×10³ km² in 1980 to 78.51×10³ km² in 2018, more than doubling (Fig. 2e). Most of the newly built-up area (77.17%) was converted from arable land, which is mainly in northern Anhui Province, northern Jiangsu Province, and Yangtze River Delta (Fig. 3e). The woodland increased from 421.59 ×10³ km² in 1980 to 428.58×10³ km² in 2010, with a slight decline of 1.89×10³ km² in 2015, and it then increased to 447.35×10³ km² in 2018 (Fig. 2b). Arable land and grassland contributed 53.95% and 24.56% of the increased woodland, at 38.03×10³ km² and 17.31×10³ km², respectively (Table 5). Most of these areas were distributed in the countryside or villages of central China and the south of the YZR Delta. The grassland increased from 35.99×10³ km² in 1980 to 36.89×10³ km² in 1990, then steadily decreased to 33.18×10³ km² in 2015 and increased once again to 42.15×10³ km² in 2018 (Fig. 2c). Waterbody areas contributing 38.14% of the new grassland are mainly distributed in central China, such as Dongting Lake and Poyang Lake. The waterbody increased from 48.53×10³ km² in 1980 to 63.99×10³ km² in 2018, or 15.45×10³ km² in total area increase (Fig. 2d). Most of the increased waterbody (64.03%) was converted from arable land at 19.24 ×10³ km². The unused land decreased from 2.38×10³ km² in 1980 to 1.91 × 10³ km² in 2015 and then increased to 3.24×10³ km² in 2018 (Fig. 2f). The waterbody and arable land contributed 44.48% and 24.98% of the newly increased unused land, respectively.

The spatial distribution of each land use type in the future scenarios is the same as in 2018 (Fig. 4a-h), although the proportions of each land type are different. Since land use conversion in the future scenarios are similar to the types of land use conversion from 1980 to 2018, this article does not analyze them further. The grassland and unused land will decline in the future for all scenarios, especially scenarios A1B (-60.52×10³ km² for grassland, -1.76 ×10³ km² for unused lands) (Fig. 4a, b) and B2 (-59.91×10³ km² for grassland, -1.55×10³ km² for unused land) (Fig. 4g, h). In the A1B and B1 scenarios, a considerable amount of arable land near the middle of YZR, in the Han River, Xiang River, Gan River, and southern Jianghuai Plain, would be converted to woodland. Cropland in the B1 scenario (-125.66×10³ km²) (Fig. 4e, f) will be reduced more than in the A1B scenario (-104.02×10³ km²). Correspondingly, the woodland in the A1B scenario (166.3×10³ km²) will be higher than in the B1 scenario (153.46×10³ km²) (Fig. 4e, f). The cropland acreage will increase from 2050 to 2100 in the A2 (65.54×10³ km²) (Fig. 4c, d) and B2 scenarios (89.02×10³ km²) (Fig. 4g, h). There is also an increase in build-up land from 2050 to 2100 in the A2 scenario (22.55×10³ km²) and the B2 scenario (0.88×10³ km²). Correspondingly, the woodland in the A2 scenario (-47.34×10³ km²) and the B2 scenario (-28.44×10³ km²) would decrease by 2100.

Many factors influence habitat quality, including human activities and physical geography. The results of the relationship analysis (Table 6) show that slope (R = 0.502, P < 0.01, in 2015), elevation (R = 0.003, P < 0.05, in 2015), population density (R = -0.299, P < 0.01, in 2015), and NDVI (R = 0.366, P < 0.01, in 2015) in the study area are significantly correlated with habitat quality (Table 4). However, the GDP and slope direction in the study area did not relate to the habitat quality. The steeper and higher an area was, the lower the population density; and the more plants there were, the higher the quality of that habitat. The relationship coefficients between these factors and habitat quality have dropped from 2000 to 2015.

The places with higher habitat quality were in the mountainous areas, especially in Yandang Mountain in southern Zhejiang, Dabie Mountain, Huang Mountain in Anhui Province, Wudang Mountain in western Hubei, Wuling Mountain in western Hunan, Xuefeng Mountain and southern South Ling and southern Jiangxi Province. The regions with lower habitat quality were distributed in the urban built-up areas, including Shanghai and Hangzhou in the Yangtze River Delta, Nanjing and Hefei in the north of lower Yangtze River, and Nanchang, Wuhan, Changsha in the middle of the Yangtze River. The habitat quality of the countryside near metropolitan areas gradually deteriorated (Fig. 6). In the first two decades, areas with low habitat quality were distributed as dots (Fig. 6a, b). With the ongoing construction of transportation and other infrastructures, more and more construction land threatened the habitat. As a result, the distribution pattern of the areas with low habitat quality gradually developed into a network, especially in the Yangzte River Delta (Fig. 6c, d, e).

The habitat quality of each province decreased from 1980 to 2018, except for Zhejiang Province in 2018 (Fig. 7). The average values of habitat quality in Shanghai and Jiangsu were lower than those in other provinces. Jiangxi Province has the highest habitat quality among the provinces, followed by Hubei Province.

To visualize the changes in habitat quality of this study area, we derived the change in habitat quality using the “raster calculator” function of ArcGIS 10.7. We subtracted the later period habitat quality from the former period habitat quality, classified the results into 10 classes of equal intervals (5 classes for negative, 5 classes for positive), and summarized the proportions of land in each class (Fig. 8, Fig. 9).

From 1980 to 2018, 61.93% of the study area experienced habitat degradation; while 38.07% of the study area experienced habitat improvement (Fig. 8, Fig. 9). From 1980 to 1990, the degraded habitat area accounted for 91.95% of the study area, which declined in value by less than -0.2. Only 6.65% of the study area experienced an increase in habitat quality, mainly in the lower reaches of the YZR, including Taihu Lake in Jiangsu Province, Yandang Mountains in southern Zhejiang Province, Huang Mountain and Chao Lake in Anhui Province, Hanjiang Plain in Hubei Province, and Poyang Lake Plain in Jiangxi Province (Fig. 8a, Fig. 9). From 1990 to 2000, the habitat quality of 96.48% of the study areas declined by less than -0.2, and only 3.05% of the areas had a rise in habitat quality (Fig. 8c, Fig. 9). The region with improved habitat quality was clustered in Yandang Mountain in Zhejiang, Huangshan in Anhui, and Hongze Lake in Jiangsu. The habitat quality of a few places in the central YZR also increased, but they were not clustered. At this time, the decline in habitat quality in the middle and lower reaches of the YZR became accelerated, and habitat quality values fell by more than 0.4 in some eastern coastal areas. From 2000 to 2010, 96.90% of the study area experienced a decline in habitat quality, but the decline was not less than -0.2, and only 2.21% of the places experienced habitat improvement (Fig. 8c, Fig. 9). From 2010 to 2015, the declining habitat quality value of 97.56% of the areas ranged from -0.4 to -0.2. Only 0.11% of the area, distributed in the coastal land in the northeast of Jiangsu and near the river basins of those provinces, experienced improved habitat quality (Fig. 8d, Fig. 9). From 2015 to 2018, the places which experienced habitat degradation accounted for 46.13% of the study area—and places where habitat quality increased were interspersed with degraded habitat areas (Fig. 8e, Fig. 9).

The habitat quality of the study area shows continuing decline at a variable rate (Fig. 10). At the beginning of the two decades, its decrease was relatively slower. It declined by -0.04% per year from 1980 to 1990 and by -0.01% per year from 1990 to 2000. However, as economic growth continued, ecological degradation became more rapid, at -0.10% per year from 2000 to 2010 and -0.14% per year from 2010 to 2015. The habitat quality decreased by -0.10% per year from 2015 to 2018. By 2100, the study area's habitat quality will improve under the A1B and B1 scenarios, but it will decrease under the A2 and B2 scenarios. In conclusion, the habitat quality of the study area would be the highest under the B1 scenario, and would be the lowest under the A2 scenario.

Under any of the scenarios, habitat quality degradation was greater in the lower or northern YZR than in the midstream or southern areas. The habitat quality in the north of YZR is worse than in the south of YZR. The habitat quality in the plain is worse than in the mountainous areas (Fig. 11). The area with habitat quality below 0.6 is distributed in the plains, including the Lianghuai Plain of northern Jiangsu Province and northern Anhui Province, the Dongting Lake Plain of the northeastern Hunan Province, the Jianghan Plain, the Han River of the east Hubei Province, and the Poyang Lake of the north Jiangxi Province. Habitat quality under the A2 scenario (Fig. 11c, d) is the worst compared to other scenarios, followed by the B2 scenario (Fig. 11g, h). Under the A2 scenario, the habitat quality score of most of the Yangtze River Delta region is below 0.2. A rapid decline occurred in the quality of the human environment in urban and peri-urban areas. Areas of low habitat quality are distributed in networks. The habitat quality score in the B1 scenario (Fig.11e, f) is the highest compared to other scenarios, followed by the A1B scenario (Fig. 11a, b). In the A1B and B1 scenarios, the habitat quality in a few regions will increase from 2050 to 2100, and the increase interval is [0.2, 0.6] (Fig. 12a, c). The areas with increased habitat quality in these two scenarios are mainly clustered in the rivers, the lakes and their surrounding areas, such as the Han River and Jing River in Hubei, the Dongting Lake, Xiang River, Zishui, Yuan River in Hunan Province, and the Poyang Lake and Gan River in Jiangxi Province. The habitat quality of other places will decrease at a low rate.