Research on the Dynamics and Evolution of Regional Blue-green Space Driven by the Development of World-class Urban Agglomerations

： In recent years, the process of urbanization in China has accelerated, 11 and changes in the underlying surface have caused the difference in average 12 temperature between built-up areas and suburbs to increase, resulting in an 13 urban heat island effect, which has become an important environmental issue for 14 today's urban sustainable development. The Yangtze River Delta urban 15 agglomeration region is the fastest-growing region in China, with economically 16 developed and populous cities such as Shanghai, Nanjing and Suzhou. It has 17 become one of the six major urban agglomerations in the world, and its heat 18 island effect is particularly prominent. The single urban heat island phenomenon 19 gradually evolves into the urban agglomeration heat island phenomenon with 20 urbanization. However, the dynamic transfer process of key blue-green space 21 landscapes that can alleviate land surface temperature (LST) and regional 22 thermal environment (RTE) is still poorly understood, especially in the context 23 of urban agglomerations. With the approval of the State Council on the 24 development plan of the Huaihe River Ecological Economic Belt, the 25 construction of which has been officially upgraded to a national strategy. The 26 Eastern HaiJiang River and Lake Linkage Zone (EJRLLZ) emphasizes 27 areas 31 agglomeration, the rich water body and green space in the ERLLZ area are also 32 destroyed and affected. Therefore, we take this region as a case to further 33 quantify the impact of urbanization and urban agglomeration development on 34 the dynamics and evolution of blue-green space. 35 In this study, MODIS land surface temperature products and Globe land 36 cover products were used for analysis. With the help of Google cloud computing, 37 Markov model and spatial analysis, the seasonal and interannual variations of 38 land surface temperature and relative land surface temperature in the study area 39 from 2000 to 2020 were analyzed from the perspective of temporal and spatial 40 changes.This paper reveals that (1) there are significant differences in the 41 cooling effect of the gains and losses of ecological land, which provides evidence 42 for the value of the existing natural ecological system (especially forest land) to 43 climate adaptation because the newly constructed ecological land does not 44 provide the same cooling effect. (2) Land cover change is not only affected by 45 land cover patterns and processes, but also significantly affected by specific land 46 conversion processes. (3) From 2000 to 2020, the development land in the ERLLZ 47 increased significantly, while the arable land decreased significantly. The urban 48 cooling island was gradually isolated and dispersed, and the urban heat island 49 was interconnected and interacted to form a regional heat island. This study 50 deepens the understanding of the dynamics and evolution of blue-green space in 51 the context of urban agglomerations, and provides an important perspective for 52 the protection of existing natural ecosystems and climate adaptation planning.

Temperat ure thres hol d Extremely low temperature zone Q≤T-2.5s Low-temperature zone T-2.5s <Q≤T-1.5s S ub-l ow temperature zone T-1.5s <Q≤T-0.5s M edium temperature zone T-0.5s <Q≤T+0.5s S ub-hig h temperature zone T+0.5s <Q ≤T+1.5s High-temperature zone T+1.5s <Q ≤T+2.5s Extremely high temperature zone T+2.5s <Q region. In this study, acc ording to a previous study [20],the region with RLST 38 below 0 °C is defined as a low temperature zone, or we c all it a regional c o oling 39 island (RCI ), the region with RLST above 2 °C is defined as a high temperature 40 zone, or we c all it a regional heat island (RHI). 41

Dynamic detection of land cover 42
In this study, land use transfer matrix (LUTM) method (Markov model) was 43 used to detect the dynamic and evolution of land cover from 2000 to 2020. LUTM 44 method originates from the quantitative description of system state and state 45 transition in system analysis. In general, as shown in table 2, Tk represents land 46 cover changes for each period , and X1 and X2 are the beginning and end stages of 47 the period, respectively. Qnn represents the area of surfac e c over Sn in X1, and 48 converted to land c over Sn in X2. Then Qn+ and Q+n represent the total area of land 49 cover Sn in X1 and X2, respectively[21]. In addition, a differenc e value will be 50 c alc ulated to determine the general change of indic ators during this period . 51 Ta

53
By c alc ulating the RLST changes of each land cover c onversion in different 54 periods, the influence of land c over dynamics on RTE mode and evolution is 55 evaluated [22]. The equation is as follows: 56 where T_DIFF represents the RLST difference of each land c over c onversion 58 in each period, s is land c over transition type , x1 represents the beginning stage 59 of the c yc le, x2 represents the end stage. Therefore, the positive value of T_DIFF 60 means that the RLST of land c over conversion type inc reases during this period, 61 and the negative value of T _DIFF means that the RLST decreases. 62  Subsequently, the isolated heat 'islands' were gradually c onnected to form a 72 regional heat island, with regional cities ex panding from 2010 to 2020, espec ially 73

Results
Yangzhou and Taizhou, whic h belong to the Yangtze River Delta urban 74 agglomeration. The water body and arable land patches around the EJRLLZ are 75 obviously distributed. There are many small patc hes of water in the northe ast of 76 Gaoyou Lake in the whole region, whic h gradually fragmented during 2000 -2020. 77 The fragmentation in 2020 is the most serious, which is obviously c aused by the 78 rapid development of urbanization between c ities and the expansion of urban 79 agglomeration h eat island effec t. In addition, it is obvious that human ac tivities 80 are mainly affec ted by river and terrain distribution. Spec ific ally (Tables 3-4 inc reased rapidly, mostly from arable land. However, the EJRLLZ has 99 experienc ed another wave of urban expansion and agglomeration in T2, mainly 100 contributed by arable land (7.46%), which means that the region fac es rapid 101 urban expansion and agglomeration from 2010 to 2020. In general, development 102 land and water increased rapidly from 2000 to 2020, and the overlong arable land 103 dec reased (Table 5). During this period, arable land c ontributed most to urban 104 expansion (56.92%), followed b y water (1.92%) and grassland (0.44%). Over the 105 years, arable land suffered a huge rec ession, most of which bec ame grassland 106 (9.83%), followed by water (6.07% cooling island effect is the strongest in winter, followed by autumn, summer heat 138 island effect is the strongest. 139 The results of Fig.4-11 show that the regional coolingisland effect (RCI) is 140 gradually isolated and RHI is signific antly enhanced exc ept Hongze Lake and 141 Gaoyou Lake, espec ially in the urban and urban districts of the EJRLLZ. Sinc e 142 2000, several isolated urban heat islands have gradually merged, which may be 143 due to the integration of Chuzhou, Yangzhou, Taizhou and Nanjing metropolitan 144 area, resulting in the increasing land coverage of development land. From 2000 145 to 2020, the RHI around the Yangtze River estuary c ontinued to expand, but 146 some RHI in the north of the EJRLLZ dec reased, espec ially in the low RLST area. 147 These mitigation trends in rec ent years may be c aused by the so -c alled ec ologic al 148 red line projec t and greenway network c onstruc tion implemented by loc al 149 governments in the EJRLLZ. 150 Spring (March-May): Seasonal variation of urban land surfac e temperature 151 spatial pattern in the EJRLLZ in spring is shown in Fig.4-5. Three years (2000,152 2010, 2020) cooling island intensity in the spatial variation range is roughly the 153 same, are c onc entrated in Hongze Lake, Gaoyou Lake . In terms of the overall and 154 loc al spatial pattern changes, the urban area as a whole shows the heat island 155 effect, which is not very obvious in the region, and the c ooling island effect is the 156 main advantage. The spatial pattern of heat island in differe nt years is quite 157 different in different seasons. The cooling island effec t in Yangzhou and Taizhou  158 in 2000 is significantly higher than that in 2010 and 2020. The heat island effec t 159 and the range of heat island in Chuzhou gradually inc reased from 2000 to 2020. 160 It is worth noting that the heat island effec t of Chuzhou in spring is stronger 161 than other c ities in the whole four seasons. The heat island effec t of Huai'an, 162 Taizhou and Yangzhou is more and more concentrated in the urban area. 163 Summer (June-August): The main reason for the highest temperature season 164 in a year is the maximum solar radiation absorbed by the surfac e and sunshine 165 hours provide good conditions for the increase of LST. Seasonal variation of 166 urban land surface temperature spatial pattern in EJRLLZ in summer is shown in 167 Fig.6-7. In terms of the spatial variation range of cooling island intensity (the 168 maximum and minimum values of c ooling island intensity), the spatial variation 169 range of c ooling island intensity in the three years (2000,2010,2020) is roughly 170 the same, which is concentrated in Hongze Lake and Gaoyou Lake. In terms of 171 the change of the overall and local spatial pattern of the cooling island, the c ity 172 as a whole presents the heat island effec t. The spatial pattern of the heat island The c ooling island effect of Chuzhou in 2020 was signific antly higher than that in 182 2010 and 2000). 183 Autumn (September-November): LST began to decrease, which was mainly 184 affected by the reduc tion of solar radiation, the shortening of sunshine time, and 185 the reduc tion of vegetation coverage. Therefore, LST began to decrease again. 186 Seasonal variation of urban land surface temperature spatial pattern in EJRLLZ 187 in autumn is shown in Fig.8-9. The spatial variation range of c ooling island 188 intensity in three winter years (2000,2010,2020) acc ounted for a large area in the 189 region. The regional c ooling island effec t in 2000~2020 tends to be the area of 190 five c ity boundaries year by year, mainly distributed in the lakes with large 191 cooling island effect in the region: Hongze Lake and Gaoyou Lake. 192 Winter (Dec ember -February): As the temperature drops further, it is wi nter 193 wheat overwintering period, c rop growth is slow, so the land cover type presents 194 contiguous low value area. The interannual variation of the spatial pattern of 195 land surfac e temperature in the EJRLLZ of winter is shown in Fig.10-11. The 196 spatial variation range of c ooling island intensity (the maximum and minimum 197 of c ooling island intensity) in the three winter years (2000,2010,2020) is very 198 similar, which is larger than that in autumn. In terms of the seasonal changes 199 of the overall and loc al spatial patterns, in general, the winter region from 2000 200 to 2020 showed the c ooling island effec t as a whole, but compared with autumn, 201 the c ooling island phenomenon was more common. The spatial pattern of cooling 202 island in different years is quite different in the region (the intensity of cooling 203 island in Yancheng in 2010 is signific antly greater than that in 2000 and 2020. In 204 2000, the intensity of heat island in Chuzhou City was signific antly lower than 205 that in 2010 and 2020). In additi on, the range of regional c ooling islands in the 206 three years was relatively c onc entrated in 2000 and 2010, and gradually 207 dispersed into fragmentation distribution in 2020. 208 It c an be seen that the LST distribution in different months is c losely related 209 to the seasonal changes of solar radiation, sunshine time, LUCC type and 210 vegetation c overage in this period. Therefore, the LST value and RLST value in 211 the study area are in the order of summer > autumn > spring > winter, exc ept that 212 Chuzhou has the strongest heat island effec t in spring. In general, the average 213 LST in EJRLLZ has a strong spatial variability, and the seasonal variation of land 214 surfac e temperature is mainly determined by c limate factors, LUCC c overage 215 and spatial pattern changes. Overall, the s patial pattern of heat island in 216 different seasons is quite different among different c ities in regional c ities. 217 Bec ause there are a large number of water bodies (lakes and rivers) and green 218 spac e (arable land) in the region, there are relatively fixed cooling island areas in 219 any season in the region, and these c ooling island areas are directly 220 corresponding to the area where the water body is loc ated, mainly the land 221 surfac e temperature of the water body is far lower than the land surfac e   The increase in forest land, grassland, water and wetlands is then referred to as 251 ec ologic al land benefits, inc luding DE-GR、BA-GR、DE-WA、BA-WA、DE-FO and 252 BA-FO (Fig.12). It c an be seen that in the proc ess of urbanization and urban 253 agglomeration, the loss of ec ologic al land generally c ontributes to the increase of 254 temperature, while the inc rease of ecologic al land usually reduc es the 255 temperature. In addition, the transition from bare land to ec ologic al land 256 signific antly reduc ed RLST. For the c onversion between woodland, grassland, 257 water and wetland, the general model is that the land coverage transferred to 258 woodland, water and wetland reduc es RLST, while the land c overage transferred 259 to grassland usually inc reases RLST. 260 The results in Tables 8-10  Combined with the results of Fig.12 and Table 8-10, it c an be clearly seen 268 that in general, espec ially in summer, the land cover type transferred to the blue 269 system (water and wetland) c an reduc e the temperature more than the green 270 system (arable land, forest land and grassland). The RLST values of land c over 271 types transferred to grassland in T1 autumn and T2 summer were negative, and 272 the RLST values of conversion from c ropland and woodland to grassland were 273 mostly negative. This means that although conversion from DE and BA to 274 grassland c an reduc e temperature, grassland has a lower c ooling effec t than 275 water and wetlands. In addition, in T1 and T2, the RLST variation of BA and DE's 276 ec o-land income is generally less than that of BA and DE's loss. These results 277 show that compared with the cooling effect brought by ec ologic al land, the loss 278 of ec ologic al land, espec ially the type of forest land coverage, will signific antly 279 inc rease the regional temperature. The differenc e of RLST between ecological 280 land loss and ecologic al land income is of great signific anc e for ecologic al land 281 protec tion. 282 The results in Tables  conversion. Development land expansion or urbanization increases regional 293 temperature and leads to RHI. All the land c over types transferred to the blue 294 system will reduc e the temperature, among which GR -WE (-139 °C) and BA-WA 295 (-1.4 °C) have larger negative RLST. It c an be seen that the transformation to 296 water and wetland usually reduc es the temperature, and vic e versa, whic h means 297 that water will be the best choic e for the regional c limate adaptation.

326
It is widely believed that urbanization (and urban agglomerations) 327 signific antly reduc es UCI effec ts and increases RTE [5,[12][13][14]26]. In partic ular, 328 the impact of the model on the UCI effec t, such as the study of Weng[4] and 329 Cao[11], has proposed that LST is related t o some dominant land c over and land 330 use types within a c ertain temperature range. Similarly, this study also found 331 that the LST of water and wetland was signific antly higher than that of other 332 land coverage types, and the average RLST was mainly between 3 °C and 8 °C. 333 However, the results of this study show that the blue -green space is not only 334 dominated by WA and WE land c over types (as well as urbanization), but also 335 signific antly affec ted by spec ific land c onversion processes (e.g. AR -WA and 336 DE-WE), as well as the differenc e in the cooling effect of ecologic al land losses 337 and benefits (Table 8-10, Fig.12). These findings provide new evidenc e for 338 explaining the rapid urbanization mechanism of UCI. In addition, c ities are 339 generally isolated and c onstrained by administrative boundaries at the initial 340 stages of urbanization, partic ularly in the context of China. Regional Cooling 341 Island (RCI) is therefore not isolated. In the proc ess of c ommon development of 342 regional c ities (Fig.2), the deterioration of RTE mak es these c onnected RCIs 343 gradually isolated ( Fig.6 and 7). In addition, climate change and anthropogenic 344 heat emissions are the mechanisms for RCI fragmentation and weakening. 345 From the beginning of the 21st c entury, the government of the EJRLLZ has 346 also implemented projec ts such as returning farmland to forests, ec ologic al red 347 lines and greenway networks. As shown in Fig.4-11, the RCI intensity in the 348 northern part of the EJRLLZ began to rise slightly (2010 -2020). 349 4.2 Impacts on regional climate adaptation planning 350 Compared with previous studies foc using on a single c ity in a single 351 period [15,16,27,28], this study uses the LUTM method to quantify the 352 multi-period changes at the regional sc ale. The research results reveal the 353 general rule of RLST dynamic s and evolution in the process of rapid 354 urbanization, and provide a sc ientific basis for the adaptation and mitigation of 355 rapid urbanization. 356 The cooling effec t of water and wetland found in this study is also 357 consistent with many previous studies [29][30][31][32][33][34]. However, the cooling effec t of 358 grassland needs further analysis. The results of this study (Table 8 -10, Fig.12-13) 359 showed that grassland had no cooling effect similar to that of water and wetland. 360 This result is different from previous research results, the previous results show 361 that grassland also has c ooling effect [35][36][37]. In fac t, Yu[22] has proposed that 362 the cooling effec t of grass vegetation is greatly affec ted by its loc al bac kground 363 c limate, whic h shows that rainfall, irrigation and wind speed conditions c an 364 signific antly affec t the cooling effec t of grass vegetation. Kang[38] also pointed 365 out that the expansion of irrigation agric ulture reduc ed the land surfac e 366 temperature and moistened the surfac e air, but promoted the comprehensive 367 measurement of temperature and humidity, thereby enhanc ing the intensity of 368 heat waves. In addition, some studies have found that grassland vegetation may 369 have a positive impact on the thermal environment, thereby impeding the 370 formation of 'cooling island', mainly due to lac k of irrigation and diffic ulty in 371 maintaining 'green state'. Santamouris [23,24] also c onc luded that the c ooling 372 effect of grassland is still unc ertain and needs further investigation. Therefore, 373 we believe that grassland is not a good choic e to adapt to and mitigate c limate 374 change, whether in EJRLLZ or in other c limate zones . 375 The research results (Tables 8 -10) also found that water and wetland usually 376 had better cooling effec t than woodland, whic h provi ded new evidenc e for 377 disc ussing the differenc e in c ooling effec t between water and 378 woodland [11,[39][40][41]. We suggest that the land coverage of water body should be 379 given priority to in the agglomeration area of the EJRLLZ to alleviate the RHI 380 effect. 381 In addition, in general, th e pattern-process-scale-effec t diagram is the basic 382 princ iple of landsc ape ec ology. It c an also explain the dynamics and evolution of 383 thermal environment. The impac t of land landsc ape pattern on urban cooling 384 island (UCI) effec t has attrac ted much attentio n[29-34], but there is still a lac k of 385 understanding of the quantitative impac t of land c over proc ess on RTE effect, 386 espec ially the land c over change proc ess in a spec ific period [15,16,27]. For 387 example, Sun and Chen [19] found that, the transformation from impervious land 388 to green land has obvious cooling effec t, but th e expansion from impervious land 389 to green land will lead to signific ant changes in the internal thermal effect of 390 green land. Yu et al.[42] found in Fuzhou (China) that from 2000 to 2013, the land 391 surfac e temperature increased with the inc rease of the proportion of 392 development land, and the proportion of green space decreased sharply. In 393 addition, this study also quantified RLST changes in BA and DE's eco -land use 394 returns less than BA and DE's loss du ring urbanization (2000-2020) and regional 395 urban development. This result shows the value and importance of the existing 396 natural ec ologic al system, bec ause the newly built 'ec ologic al' land does not 397 provide the same cooling effec t. Moreover, the existing grasslands still play an 398 important role in mitigating RTE due to the larger RLST changes c aused by the 399 conversion of grassland to development land than the conversion of 400 development land to grassland. Therefore, these strategies c an foc us on how to 401 improve the cooling performance of the current grassland through better 402 adaptive management and planning (i.e., establishing tree -shrub-grass 403 struc ture).