Spatiotemporal effects of urban sprawl on habitat quality in the Pearl River Delta from 1990 to 2018

Since the implementation of the Chinese economic reforms. The habitat quality of coastal has gradually deteriorated with economic development, but the concept of "ecological construction" has slowed the negative trend. For quantitative analysis of the correlation between the Pearl River Delta urban expansion and changes in habitat quality under the influence of the policy, we first analyzed the habitat quality change based on the InVEST model and then measured the impact of construction land expansion on the habitat quality through habitat quality change index (HQCI) and contribution index (CI) indicators. Finally, the correlation between urbanization level and habitat quality was evaluated using geographically weighted regression (GWR) and the Self-organizing feature mapping neural network (SOFM). The results indicated that: (1) during the study period from 2000 to 2020, habitat quality declined due to urban sprawl, indicating a deterioration of ecological structure and function, and the decrease was most significant from 2000 to 2010. (2) The urbanization index had a negative effect on the habitat quality, but the negative effect have improved after 2000, reflecting the positive effect of policies such as "ecological civilization construction" (3) The implementation degree of ecological civilization varies greatly among cities in the study area: Shenzhen, Dongguan, Foshan, and Zhongshan have the best level of green development. These results reflect the positive role of policies in the prevention of damage to habitat quality caused by economic development and provide a reference for the formulation of sustainable urban development policies with spatial differences.


31
Habitat quality refers to the ability of an ecosystem to provide suitable living 32 conditions to sustain a species, which can reflect the level of biodiversity and ecological 33 services to a certain extent 1,2 . Urbanization is the main driving factor that puts 34 tremendous pressure on biodiversity conservation 3 . Since the implementation of the 35 reform and opening-up policy, China's urbanization rate has increased rapidly, from 36 17.9% in 1978 to 60.6% in 2019, and is expected to reach 65.5% by 2025. The rapid 37 landscape patterns, and habitat quality, as well as analyzing the impact of natural factors such as DEM, temperature, precipitation, and slope aspect, or human factors such as 63 degradation of habitat quality 3,22 . Land-use cover change and expansion of construction

150
The land-use information graph is a geospatial analysis model combining 151 attributes, processes, and spaces, which can reflect the spatial differences and temporal 152 changes in land-use types. In its function expression, let the state variables be = ( , ) (1) 156 where represents land-use characteristics.
(1) To realize the spatial description 157 of land attributes, when t is constant, the function relation of changing with is 158 constructed.
(2) The process description of land attributes can be realized, and when 159 is constant, the function relation of changing with can be constructed. The 160 combination of these two functions can form a conceptual model of the land-use 161 information graph and realize a composite study of land space, process, and attributes. 162

Habitat quality
163

Habitat quality evaluation 164
We used InVEST-HQ to evaluate the habitat quality in the Pearl River Delta region. 165 Based on land-use types, this module calculated the habitat degradation degree and 166 habitat quality index by using threat factors, the sensitivity of different habitat types to 167 threat factors, and habitat suitability. Habitat degradation and habitat quality were 168 calculated using the following formulas: 169 is the habitat quality of grid x in land-use type j, is the habitat 172 suitability of land-use type j, is the habitat degradation degree of grid x in land-173 use type j, k is the half-satiety sum constant, r is the number of threat factors, and y is the relative sensitivity of threat sources. 、 , and are, respectively, the 175 interference intensity and weight of the grid where the threat factor r is located, and the 176 interference generated by the habitat.
、 are the anti-disturbance ability of 177 habitat type x and its relative sensitivity to various threat sources, respectively. 178 The value range of habitat degradation degree is [0,1], and the larger the value, the 179 more serious the habitat degradation. The value of habitat quality is between 0 and 1, 180 and the higher the value, the better the habitat quality. 181   The CI was used to analyze the causes of the changes in habitat quality, and the 199 following formula was used to quantitatively represent the contribution of land-use 200 conversion to habitat quality change. In this study, the total value of habitat quality loss 201 The research unit is a river basin, which has both natural and social attributes. It is 230 a relatively independent and complete system, which can connect and explain the 231 coupling phenomenon of society, economy, and nature 45 . The hydrological analysis 232 module in ArcGIS was used to divide the research area into 374 small basins. When 233 calculating the cumulative flow of the grid, 100,000 was used as the threshold value, 234 and basins less than 5 km 2 were combined with the adjacent basins.
The SOFM neural network was proposed by Kohonen, a Finnish scholar, and 237 constructed by simulating a "lateral inhibition" phenomenon in the human cerebral 238 cortex. It has been widely applied in classification research in geographic and land 239 system science 39,40 . The advantages of this method in classifying the coupling 240 relationship between urbanization and habitat quality are as follows :(1) it simulates 241 human brain neurons through unsupervised learning, which is objective and reliable. (2) 242 It maintains the data topology during self-learning, training, and simulation to obtain 243 reasonable partition results and identify the differences between different basins. (3) 244 For massive data, the SOFM network has a good clustering function while maintaining 245 its characteristics and uses the weight vector of the output node to represent the original 246 input. This method can compress the data while maintaining a high similarity between 247 the compression results and the original input data. 46 . We exported the data from 248 ArcGIS, and conducted cluster analysis on the four factors of NTL, POP, LUR and 249 habitat quality using SOFM. Finally, the analysis results are imported into ArcGIS for 250 display.
Therefore, before analyzing the spatiotemporal change of habitat quality and its degree 257 of coupling with urbanization level, the urban expansion in the Pearl River Delta region 258 from 1990 to 2018 was examined as reflected in Figure 3 and 4. 259    To reflect the spatial and temporal changes in habitat quality in the Pearl River 320 Delta region during the study period, an interannual map of habitat quality was drawn. Shenzhen, and Zhongshan, while there are many areas with increased habitat quality in 330 Zhaoqing, Jiangmen, and Huizhou, which preliminarily indicates that these areas attach 331 greater importance to ecological protection. 332 The above results can be further verified by analyzing Table 5. During the three 333 periods, the change in habitat quality was most prominent from 2000 to 2010. The 334 increased value of low-grade habitat quality was as high as 2893.47 2 , and the 335 decrease in higher and higher quality habitat quality was more than 500 2 , which 336 highlighted the negative impact of rapid economic development on ecology during the 337 decade. In the past 30 years, the area of low-grade habitat quality increased by 4911.07 338 2 , while the sum of the area reduced to IV and V habitat quality was approximately 339 1500 2 . Substantial degradation in habitat quality has become an urgent concern. 340 These results are consistent with the situation of urban sprawl discussed in Section 3.1.1, 341 which can preliminarily infer the correlation between urban sprawl and habitat quality.  Table 3. According to the HQCI, all land conversions to construction land will 347 lead to habitat quality degradation. The HQCI was negative, and the absolute value was 348 greater than 0.10, and the effect of grassland transfer on habitat quality was the most 349 obvious, with an HQCI value of −0.30. It can be observed from the CI value that the 350 conversion of cultivated land leads to the degradation of habitat quality the most, with 351 a CI value of −386.02, followed by woodland and water areas. The reason why the 352 HQCI value of these land transfers is smaller but the CI value is larger is that they cover 353 a larger area. However, grassland conversion had the greatest impact on habitat quality 354 per unit area, but the total loss to habitat quality was not obvious because of the small 355 area of grassland transfer.

Impact of socioeconomic factors on habitat quality 360
The changes in urban expansion and habitat quality reflect the important influence 361 of human activities on the ecological space. In this study, habitat quality was taken as 362 the dependent variable, and NTL, POP, and LUR were selected as independent variables 363 by referring to existing studies 20,35 . The OLS and GWR were used for the analysis, and 364 it was found that the explanatory power of the GWR model at four time points was 365 superior to that of the OLS model, and the Sigma and AICC values of the former were 366 lower. Therefore, the GWR model was selected to obtain a better fitting effect and 367 higher accuracy. 368 During the study period, there was a negative correlation between habitat quality 369 and the NTL, POP and LUR in most areas. With the passage of time, this effect first 370 intensified and then gradually improved (Figure 7,8,9). From 1990 to 2010, these three 371 urbanization factors were negatively correlated with habitat quality in more than half Second, the SOFM classification results show that these areas have a low level of 493 urbanization and habitat quality is not considerably higher than that of the Pearl River 494 Delta core areas, so they have high development potential. 495

496
Based on existing research, the following innovations were made in this study: not 497 only quantifying the harm of urban sprawl on habitat quality, but also identifying the 498 regional differences in three types of socioeconomic factors, and reflecting the 499 importance of policy in urban ecological protection through comparison between 500 different time periods. First, in order to examine the impact of policies on habitat quality 501 in different periods, HQCI and CI index were used to quantitatively analyze the specific 502 impact of land transfer on habitat quality dating back to the time when the Pearl River 503 Delta was just established as a "coastal economic development zone." Then, 504 considering the construction of ecological civilization as the starting point, the GWR 505 analysis results reflect the regions that pay attention to economic development but ignore the protection of ecological resources during the research period. Finally, based on the clustering results of the SOFM neural network, the differences in green 508 development in different regions over the past 30 years were discussed. Consistent with 509 previous studies, the results of this study also show that urban expansion and human 510 activities have a serious negative impact on habitat quality. The difference is that some 511 studies do not consider the spatial and temporal differences of various influencing 512 factors and the impact of ecological-protection-related policies in different regions 24,35 . 513 In this study, from the perspective of policy changes, the spatial heterogeneity of social 514 and economic factors is included to provide a reference for the social and economic 515 development process of coastal urban agglomerations and the relationship between 516 urbanization and the ecosystem. 517 Although this study has effectively supplemented and expanded the sustainable 518 development of urban ecology in the Pearl River Delta region, it still has some 519 limitations. First, the assessment of habitat quality is a complex task. Although the 520 InVEST -HQ model has been applied by many scholars to calculate the habitat quality 521 index, it needs to be improved in terms of pertinence and reliability because it is based 522 on land-use type. In the future, multi-source data will need to be considered to reflect 523 habitat quality. Second, because of the limitation of data and considering that both 524 population and night light are social and economic factors with stable growth in the 525 short term, the population density in 2018 and the night light grid layer in 1990 in this 526 study were obtained by linear fitting of the data of the most recent year, so as to maintain original data will be sought to avoid these errors. Finally, the spatial scale effect plays 529 an important role in studies related to geography and ecology. To reflect the natural and 530 social attributes at the same time, this study adopts the watershed as the research unit 531 when analyzing the correlation by GWR and classifying by SOFM to solve the problem 532 of inconsistent resolution of original data. We will then consider changing the size of 533 the research unit to explore the impact of urban development on habitat quality at 534 different spatial scales. 535

536
Rapid urban expansion and high-intensity human activities have greatly affected 537 habitat quality in the Pearl River Delta. Based on the analysis of the spatiotemporal 538 evolution characteristics of habitat quality, the GWR model was used to explore the 539 impact of urbanization on habitat quality, and the SOFM neural network was used to 540 cluster each river basin into four zones according to the green development status. The 541 habitat quality index was calculated based on the InVEST-HQ model, and urbanization 542 indexes included NTL, POP, and LUR. 543 The main results are as follows :(1) The period of urban expansion was the fastest 544 from 2000 to 2010, which coincided with the period of decreasing habitat quality, and 545 the area of urban expansion was mainly concentrated in the center of the Pearl River the types of land transferred to construction land, arable land accounts for the largest area, causing irreversible harm to development of the agricultural level. (2) The area of 549 low-grade habitat quality increased by 4911.07 2 during the study period, and the 550 sum of the reduction in areas of IV and V habitat quality was about 1500 2 . The 551 conversion of grassland to construction land per unit area had the most obvious effect 552 on habitat quality, while the conversion of cultivated land caused the greatest total loss 553 of habitat quality. Considerable degradation of habitat quality has become a matter of 554 urgent concern. (3) There were considerable negative correlations between habitat 555 quality and NTL, POP, and LUR in most areas during the study period. Before 2000, 556 this negative impact worsened but gradually improved from to 2000-2018, which is 557 closely related to a large number of policies related to "ecological civilization 558 construction" since the 21st century. (4) Different cities in the Pearl River Delta have 559 great differences in the importance they attach to the construction of ecological 560 civilization and green development. The level of green development in Shenzhen, 561 Foshan, and Zhongshan was the highest, while the levels of urbanization and habitat 562 quality in most areas of other cities were relatively low. 563 564