Effects of Land Use Changes Across Different Urbanization Periods on Summer Rainfall in the Pearl River Delta Core Area

The Pearl River Delta (PRD) is one of the three urban agglomerations in China that have experienced rapid development. For this study, a core area of the PRD was identified, comprising the highly urbanized areas of Guangzhou, Foshan, Zhongshan, Zhuhai, Shenzhen, and Dongguan Cities. The expansion of these urban areas was tracked across three time periods—the year population urbanization rate exceeded 70% (2000), 18 years before (1982), and 18 years after (2018). This study used the Weather Research and Forecasting (WRF) model to explore summer rainfall changes across different urbanization periods in the PRD core area. The results show that urban land expansion mainly occurred in the post urbanization period. Rainfall changes across different urbanization periods were roughly consistent with previously observed spatial and temporal changes accompanying urban expansion in the PRD core area. Extreme rainfall mainly increased in the post urbanization period, shifting rainstorm center towards the PRD core area. Further causal analysis revealed that land use changes affected rainfall by altering thermodynamics and water vapor transfer. The urban expansion changed the surface energy balance, resulting in increased surface heating and heat island effects. The heat island effects thickened the planetary boundary layer and increased vertical wind speeds, which initiated dry island effects, thereby causing more water vapor transportation to the atmosphere. Consequently, rainstorms and extreme rainfall events have become concentrated in urban areas.


Introduction
China is one of the fastest growing economies in the world, and its urbanization is developing rapidly. According to previous studies Chen et al. 2016), China entered the accelerated development stage of urbanization in the mid-1990s. Urban expansion, as one of the most prominent spatial characteristics of urbanization, changes surface topography, thermal conductivity, and hydraulic permeability (Junior et al. 2010;Li, Chen, et al. 2016;Wang et al. 2021). These changes affect ecosystem structure and function, resulting in ecological problems such as the urban heat island effect (Zhou et al. 2015;Rath et al. 2022;Tong et al. 2022), water pollution (Paerl et al. 2014;Li, Lu, et al. 2016), floods (Du et al. 2012;Zhou et al. 2013), and biodiversity losses (Ibáñez-Álamo et al. 2017;Cheng et al. 2019). The urbanization of the Pearl River Delta (PRD) has developed rapidly in recent decades, but it is spatially uneven. Highly urbanized areas include Guangzhou, Foshan, Zhongshan, Zhuhai, Shenzhen, and Dongguan Cities, which form an A-shaped core area around the Pearl River Estuary (Yang et al. 2020). With rapid urbanization, extreme rainfall and flood events are increasing. For example, on 11 May 2014, a rainstorm in Shenzhen City lasted for 22 h, while in the Longhua District of that city, the center of the rainstorm, the maximum rainfall was 310 mm in only 6 h, reaching the 180-year flood mark.
Rainfall is the direct cause of urban floods. Therefore, knowing how urbanization impacts rainfall is helpful for disaster management. Using observational data analysis and numerical simulations (Kishtawal et al. 2010;Shastri et al. 2015;Niyogi et al. 2017), past studies have shown the effects of heat islands (Oke 1982;Yang et al. 2016;Wu et al. 2019;Rath and Panda 2020), urban canopy roughness effect (Miao et al. 2009;Miao et al. 2011;, and energy-related aerosol emissions Ou et al. 2017;Misra et al. 2019) on rainfall formation. For instance, Souma et al. (2013) used a cloud-resolution model Fig. 1 Location of the study area, the Pearl River Delta core area, as well as meteorological stations in the core delta region and adjacent districts 1 3 to compare the effects of artificial land cover and artificial heat on a local heavy rainfall event. The results showed that both artificial heat and artificial land cover increase precipitation by increasing near-surface air temperature. Furthermore, rainfall formation has been correlated to urban topography, circulation, and climate characteristics Li et al. 2021).
Previous studies about the effects of urbanization on rainfall in the PRD mainly adopt the method of observation data analysis. Kaufmann et al. (2007) analyzed the effects of urbanization on rainfall in the PRD based on data from meteorological stations and found that urbanization leads to the decrease of total rainfall in the PRD, which may be caused by surface hydrological changes due to land use changes. Wu et al. (2019) found that urbanization increases extreme rainfall in the PRD, and showed that the increase of short-duration rainfall is more significant than that of longduration rainfall. Further analysis of extreme rainfall events showed that the increase of extreme rainfall is stronger in the areas where a heat island effect is prominent. Yan et al. (2020) observed that urbanization increases heavy rainfall and extreme rainfall in the PRD, and the increasing rainfall trend was more significant during the flood season. Although these studies offer qualitative conclusions regarding the effects of urbanization on rainfall by analyzing observed rainfall data, the quantitative results are challenging to isolate from the impacts of climate change. Furthermore, these conclusions do not involve the physical mechanism of rainfall. Therefore, the causes and physical mechanisms of rainfall (especially extreme rainfall) changes caused by urbanization remain unclear.
Some previous studies used numerical simulation to explore the impact of urbanization on rainfall in the PRD. For instance, Wang et al. (2012) designed two land use scenarios to explore the impact of urbanization on the regional climate of three large urban agglomerations in China using the Weather Research and Forecasting (WRF) model. 1 The results show that urbanization increases land surface temperature and the heat stress index, while it simultaneously decreases rainfall in urban areas. Zhang et al. (2019) used the WRF model to simulate the summer rainfall process under two urbanization scenarios and found that urbanization increases extreme rainfall and slightly reduces total rainfall. These studies included the entire PRD, however, and did not consider the effects of different urbanization periods on rainfall. Huang et al. (2021) found that rainfall mainly changes significantly in the PRD core area due to urbanization, but not in other areas based on the observation data. Thus, an analysis of the entire PRD will reduce the specificity of the effects on rainfall because of the uneven distribution of urbanization. We also know that urbanization can be divided into several different periods, and each period has its unique characteristics. Therefore, studying the effects of different urbanization periods on rainfall helps us to gain a deeper understanding about urban rainfall.
Thus, this study used the PRD core area as the study area and divided the urbanization process into pre urbanization and post urbanization according to urbanization rates. The WRF model was used to simulate the rainfall of three land use scenarios to analyze the effects on summer rainfall in different urbanization periods. Past studies found that anthropogenic heat (AH) also affects rainfall (Feng et al. 2014). Therefore, the default value of AH in the model was replaced by data from the PRD to improve the simulation accuracy. The objectives of the study were to: (1) obtain the spatial and temporal characteristics of urban expansion in the PRD core area during different urbanization periods; (2) analyze the effects of land use changes in different urbanization periods on summer rainfall in the PRD core area; and (3) explore the physical mechanisms of these effects.

Data and Methodology
This section introduces the study area, the data used by the WRF model, and the study methodology.

Study Area
This study focused on the middle section of the PRD, referred to as its "core." It covers an area of about 11,692 km 2 , including Shenzhen, Zhongshan, and Dongguan Cities, as well as parts of Guangzhou, Zhuhai, and Foshan Cities (Fig. 1). Having subtropical monsoon climate, the area is warm and humid year-round. The average annual temperature is 21-23 °C, with average annual rainfall exceeding 1500 mm.
As one of three major urban agglomerations in China, the PRD has a high urbanization level, but the urbanization development is spatially uneven. By 2012, the core area's urbanization rate reached more than 85%, comprising at least 75% of the population of the PRD, while in other areas of the PRD, the urbanization rates did not exceed 50% (Guangdong Provincial Statistics Bureau 2013). Figure 2 shows that population aggregation, economic development, and urban expansion mainly occurred in the core area of the PRD. Therefore, using the core area as the study area can focus on the effects of land use changes in the PRD on rainfall in highly-urbanized areas.

Data
Observational data are diurnal scale data, including data on rainfall, temperature, and relative humidity.  This dataset is a national scale and multi-period land use/ land cover thematic database of China, which was constructed by visual interpretation based on Landsat remote sensing images.

Methodology
This section introduces the configuration of the WRF model and the calculation of error evaluation index and urban expansion index.

Model Configuration
The Weather Research and Forecasting model (WRF V4.2) 4 coupled single-layer urban canopy model (SLUCM) was used to investigate the impact of urbanization on summer rainfall characteristics in the PRD core area. The WRF model is a comprehensive, nonhydrostatic, mesoscale model developed by the U.S. National Center for Atmospheric Research (NCAR) and the U.S. National Oceanic and Atmospheric Administration (NOAA) in collaboration with an international array of universities and research institutes Li et al. 2017;Powers et al. 2017). First developed by Kusaka, SLUCM can describe the height, density, and shape of urban buildings in detail (Kusaka et al. 2001;Kusaka and Kimura 2004). Coupled with the WRF model, SLUCM can improve the simulation accuracy of urban land use change and predict the influence of urban geometric factors on surface energy and wind change (Kusaka et al. 2012). This approach has been widely used in mesoscale meteorological simulations (Kusaka et al. 2012;Li et al. 2013;Yang et al. 2015).
Two one-way-nested domains were configured (Fig. 1). The outer domain (d01) covered the South China Sea and the inner domain (d02) covered the PRD core area. The horizontal grids were 200 × 160 and 85 × 82 in numbers, with horizontal grid spacings of 9 km and 3 km, respectively. The vertical grids contained 33 sigma levels, with the upper boundary set at 50 hPa. Our simulations were carried out for each summer (May, June, July, and August) from 2018 to 2020 to eliminate irregularities of a certain year. First, the impacts of land-use changes on summer rainfall in 2018, 2019, and 2020 were analyzed separately, and then the impacts in these 3 years were averaged to obtain a nearly universal pattern. Our model was initiated at 00 UTC 15 May and terminated at 00 UTC 1 September of each year. The month of May for each simulation was used as a spin-up period and will not be considered in the following analyses. The initial and boundary conditions were provided by the Global Data Assimillation System (GDAS) with final analysis data from the U.S. National Center for Atmospheric Research (NCAR), using a 0.25° × 0.25° spatial resolution and a 6 h temporal resolution (Kleist et al. 2009). The physical option "sst_update" was turned on to update the lower boundary conditions periodically after extended periods of simulations.
Previous studies indicate that rainfall simulations are more sensitive to cumulus parameterization and microphysics than to other physical parameterizations. This study ran a set of sensitivity models focusing on three microphysics schemes: (1) the Lin et al. scheme (Lin et al. 1983); (2) the Thompson scheme (Thompson et al. 2006); and (3) the WRF single-moment 6 scheme (WSM6) (Lim 2006), and two cumulus parameterizations: (1) the Tiedtke scheme (Tiedtke 1989) and (2) the Kain-Fritsch (K-F) scheme (Kain and Fritsch 1993). These schemes have been used frequently for rainfall simulations in China Zhang et al. 2019;Wen et al. 2020). By comparing the simulation accuracy of six groups of summer rainfall in 2018, the Lin et al. scheme and K-F scheme combinations were selected.
Other physics schemes are shown in Table 1, which have been used to simulate the southern China regions in previous studies. The physics schemes used in this study mainly include the Lin et al. scheme (Lin et al. 1983), Kain-Fritsch scheme (Kain and Fritsch 1993), Yonsei University scheme (Hong et al. 2006), Rapid Radiative Transfer Model scheme (Iacono et al. 2008), Dudhia scheme (Chen and Dudhia 2001), and Noah Land Surface Model .
The spatial and temporal resolutions of the final output data are 6 h and 3 km × 3 km, respectively. The single-layer urban canopy model (SLUCM) was coupled to the WRF V 4.2 model to better simulate the thermodynamic properties of the urban land surface system and the interactions of urban surfaces and atmospheric processes. Three land use situations from the years 1982, 2000, and 2018 were selected to evaluate the impact of different urbanization periods. The land use categories were classified into three subcategories according to impervious rates (IR): low-intensity residential (IR ≤ 0.5), high-intensity residential (0.5 < IR ≤0.8), and commercial/industrial (IR > 0.8). In this study, all parameters of the SLUCM, except the daily anthropogenic heat and daily allocation coefficient, were taken as default values, with the daily anthropogenic heat replaced by the daily anthropogenic heat of the PRD in 2018. The anthropogenic heat daily allocation coefficient was replaced by the allocation coefficient of Guangzhou (Zhan and Dan 2014) (Fig. 3). The daily anthropogenic heat data in 2018 were calculated from the anthropogenic heat data in 2016 using the weighted average method (Wen et al. 2020) based on the statistical yearbook data of the PRD ( Table 2). The anthropogenic heat data from 2016 are grid data with 500 m × 500 m spatial resolution (Wang et al. 2020), which were obtained from the National Tibetan Plateau Data Center. 5 Three simulations schemes, S1982, S2000, and S2018, are show in Table 3. S2018 is the control scheme. In S1982 and S2000 only land use data were changed from S2018, so as to exclude the interference of other factors and explore the effects of land use changes on rainfall. Therefore, the results of S2000−S1982 and S2018−S2000 can be regarded as the effects of land use changes on rainfall in pre urbanization and post urbanization phases, respectively. The results of S2018−S1982 show the influence of land use changes , and 2018 (c) 5 https:// data. tpdc. ac. cn/ zh-hans/ data/ 637e9 9a1-be38-4a9f-8af9-a8600 e13d5 68/.
1 3 on summer rainfall in the core area of the PRD during the whole urbanization period.

Error Evaluation Index
In this study, mean deviation (MD), root mean square error (RMSE), correlation coefficient (CC), and relative error (RE) were chosen as the error evaluation indices, calculated as follows: where n is the simulation time, obs i represents the ith period observation for the station, and sim i represents the ith period simulation for the grid point. The bilinear interpolation method was used to interpolate the grid where the station is located, and then the simulation results obtained from

Urban Expansion Index
The expansion rate and intensity are important indices to indicate the characteristics of urban expansion (Kuang et al. 2014). Urban expansion rate (R ur ) refers to the annual growth rate of the urban area, which is used to represent the urban expansion trend of different regions in the same period. The urban expansion intensity index (I ur ) is used to standardize the average annual urban expansion rate of each region by area, which is used to compare the urban expansion rate of the same region in different periods. The calculation formulas are as follows: where ΔU ij is the urban expansion area of the ith region in the jth period, Δt ij is the time of the jth period, ULA ij is the urban area of the ith region at the beginning of the jth period, and TLA ij is the total land area of the ith region.

Results
This section first analyzes the characteristics of urban expansion in the core area of the PRD. Then the WRF model is verified. Finally, based on the simulation results of the WRF model, the characteristics and causes of rainfall changes in the PRD core area are explored.

Characteristics of Urban Expansion in the Core Area of the Pearl River Delta
The urbanization rate is the proportion of the urban population to the total population, which is a measure of urbanization. When the urbanization rate exceeds 70%, urbanization development has entered the post urbanization phase. According to the Guangdong Provincial Bureau of Statistics, the urbanization rate of the PRD reached 71.6% in 2000. This statistic helped determine the choice of land use scenarios of the years 1982, 2000, and 2018 for this study (Fig. 4).
Corresponding to the administrative areas, the PRD core area was divided into five regions: Guangzhou (GZ), Dongguan (DG), Shenzhen (SZ), Foshan (FS), and Zhongshan and Zhuhai (ZZ). As shown in Tables 4 and 5, in the pre urbanization phase, the urban expansion rate of SZ and DG was relatively high, and urban construction land was concentrated in GZ and DG, accounting for about 50% of the urban construction land in the PRD core area. In the post

Fig. 5
Changes of extreme rainfall intensity in the whole urbanization period from 1982 to 2018 in the Pearl River Delta core area (mm/ day). "0" is in the "−10−0" range. Grids with values not significant at the 95% confidence level (Student's t test) have been masked out urbanization phase, FS and ZZ experienced higher urban expansion, with urban expansion rates exceeding 6%, and the proportion of urban area in 2018 was over 30%, forming a true PRD core area. As shown in Table 6, urban expansion in each region primarily occurred during the post urbanization phase and the urban expansion rate in the post urbanization phase was higher than that in the pre urbanization phase.

Model Validation
To test the simulation accuracy of the model, statistical comparisons between the observations and model simulations of temperature at 2 m (T2) and relative humidity at 2 m (RH2) were verified using the MD, RMSE, and CC for S2018. The verification results of T2 and RH2 are shown in Table 7. The MD and RMSE of air temperature are less than 2 (except HY), and the CC is greater than 0.5, except in ZH. The MD and RMSE of RH2 range from 6.02−11.02 and 8.02−12.82, respectively. The CCs exceeded 0.6, indicating a significant correlation. In summary, the simulation results of T2 and RH2 are robust. Extreme daily rainfall (95th percentile of all rainy days), mean daily rainfall, and the number of rainfall days were selected to examine the simulation results of rainfall using the RE. These data indicate that the simulation results had acceptable simulation errors (Table 8). The accuracy of simulated extreme daily rainfall was even higher than other simulated rainfall characteristics. Deviations between the simulated results of the model and the observed values may be attributed to the simulation deviations of temperature, relative humidity, and the initial field conditions ).

Characteristics of Rainfall Changes in the Core Area of the Pearl River Delta
The characteristics of rainfall changes are analyzed from the following three aspects: rainfall intensity, rainfall frequency, and migration of the rainstorm center. Fig. 6 Changes of extreme rainfall intensity in the pre urbanization (a) and post urbanization (b) periods in the Pearl River Delta core area (mm/ day). "0" is in the "−10−0" range. Grids with values not significant at the 95% confidence level (Student's t test) have been masked out

Rainfall Intensity
Because it had the highest accuracy in simulations, the extreme rainfall intensity (95th percentile of the daily rainfall series) was selected as the indicator with which to study the changes of rainfall intensity caused by urbanization. From 1982 to 2018, extreme rainfall intensity significantly increased in most of the PRD core area, accounting for 81% of the total core area, with an average increase rate of 16.6 mm/day (Fig. 5). Figure 6 shows the spatial changes of extreme rainfall intensity in the pre urbanization (1982−2000) and post urbanization (2000−2018) periods. During the pre urbanization period, extreme rainfall intensity increased significantly in Guangzhou and Dongguan, accounting for 56% of the total core area, with an average increase of 7.7 mm/day. During the post urbanization period, extreme rainfall intensity increased significantly in Shenzhen, Zhuhai, Zhongshan, and Foshan, accounting for 56% of the total core area, with an average increase of 8.9 mm/day. In different urbanization periods, the proportion of the area with significant increase in extreme rainfall intensity was the same, but the location of the area varied greatly, and the post urbanization rates of increase were larger than the pre urbanization rates. This increasing trend may be related to the spatial and temporal characteristics of urban expansion in the PRD core area.

Rainfall Frequency
To examine the changes of rainfall frequency caused by urbanization, rainfall events were divided into five categories according to the criteria of the China Meteorological Administration: light rainfall (0.1−9.9 mm/day), moderate rainfall (10−24.9 mm/day), heavy rainfall (24−49.9 mm/ day), rainstorm (50−99.9 mm/day), and extreme rainfall (above 100 mm/day). The changes in the five classes of rainfall events in the PRD core area during the whole urbanization period are shown in Fig. 7. The frequency of extreme rainfall, rainstorm, and heavy rainfall increased by an average of 2.07, 2.38, and 1.74 days, respectively, with growth rates of 40.4%, 31.8%, and 15.4%. The frequency of moderate rainfall and light rainfall decreased by an average of 0.4 and 3.76 days, respectively, decreasing by 2.5% and 4.8%. Similarly, rainfall frequency changes were separately assessed according to the pre urbanization and post urbanization periods, as shown in Fig. 8. The frequency of extreme rainfall increased in the pre urbanization and post urbanization categories by an average of 0.35 day and 1.72 days, respectively, with growth rates of 6.8% and 33.6%. The rainstorm frequency also increased in the pre urbanization and post urbanization periods by an average of 0.5 day and 1.88 days, respectively, with growth rates of 6.7% and 25.1%. The frequency of heavy rainfall and moderate rainfall decreased in the pre urbanization period and increased in the post urbanization period, showing a great uncertainty. Both pre urbanization and post urbanization light rainfall frequencies decreased.

Migration of the Rainstorm Center
As shown in the previous subsections, land use changes have the most significant impact on rainstorm (24 h rainfall above 50 mm) intensity and frequency. Therefore, the spatial distributions of rainstorms in 1982, 2000, and 2018 are outlined  Fig. 9, based on the frequency of rainstorms in order to analyze the influence of land use changes on the migration of rainstorm center under constant climate conditions. Figure 9 shows that the rainstorm center in 1982 was located to the northeast of the PRD core area. Rainstorm frequency in the core area averaged 8.16 days, and no rainstorm center was formed in the core area. In 2000, the rainstorm frequency in the core area increased by 0.83 day on average, and the rainstorm center located in the northeast showed some expansion toward the PRD core area. In 2018, a rainstorm center was formed in the PRD core area, and the rainstorm frequency significantly increased in the core area, with an average frequency of 12.61 days. These results indicate that with the land use changes due to urbanization, the rainstorm center gradually migrated to the PRD core area under the same climate conditions.

Cause Analysis of Rainfall Changes Due to Urbanization
The data analysis showed that land use changes have a significant impact on the intensity and frequency of rainfall as well as the migration of the rainstorm center. The causes of the influence of land use changes on rainstorms were explored from the two aspects of thermodynamics and water vapor conditions. Figure 10 shows the influence of land use changes on thermodynamic variables during the whole urbanization period. In Fig. 10a, the sensible heat flux (SHF) mainly showed an upward trend in the whole simulated area, with the most significant increase (more than 30 W/m 2 ) in the core area of the PRD. Figure 10b indicates that latent heat flux (LHF) trended downward with land use changes, especially in the core area of the PRD, where a reduction of 30 W/m 2 can be observed. Figure 10c and d show that the level of free convection (LFC) 6 and lifting condensation level (LCL) 7 increased significantly in the core area of the PRD, with the increases above 100 m and 60 m respectively. Figure 10e shows that the planetary boundary layer height (PBLH) 8 also exhibited a significant increase (above 80 m) in the PRD core area. Figure 11 presents the influence of land use changes on water vapor variables during the whole urbanization period. Figure 11a and c show that the relative humidity (RH) and water vapor mixing ratio (WVMR) at 2 m in the simulated area decreased with land use changes, especially in the core area of the PRD (average decreases of 4.6% and 1.5 g/kg respectively), where a dry island effect is obvious. Urban surfaces are composed of reinforced concrete, replacing the natural vegetation surface, subsequently reducing the water vapor and the relative humidity on the ground. Figure 11b indicates that the temperature at 2 m (T2) increased with the change of land use. The whole core area of the PRD showed an obvious heat island effect, with an average temperature increase of 0.59 ℃, thus enhancing vertical wind speed (Fig. 11e), and resulting in more water vapor being transported upward. Figure 11d shows that water vapor mixing ratio at 850 hPa increased significantly in the core area of the PRD, with an average increase of 0.15 g/kg.

Discussion
The impacts of land use changes on summer rainfall characteristics in the core area of the PRD across different urbanization periods are notable. During the whole urbanization period (1982−2018), extreme rainfall intensity in most parts of the core area increased significantly, the frequency of rainstorms and extreme rainfall increased significantly, and the rainstorm center migrated to the core area. These results are consistent with previous studies on the PRD. Chang et al. (2021) found that the center of annual days of heavy rainfall has moved towards the built-up area during the urbanization in Beijing. However, using the PRD core area instead of the whole PRD as the study area, the quantitative results and pattern of influence found in this study are stronger than in previous studies. Zhang et al. (2019) found that the frequency of extreme rainfall and rainstorms increased by an average of 4.19% and 3.04% in the whole PRD due to urbanization, respectively. By analyzing the changes of rainfall during different urbanization periods, it was found that with the development of urbanization, the impact of land use changes on rainfall have changed from the impact on rainfall intensity to the impact on rainfall intensity and frequency. In the pre urbanization period, the rainfall intensity increased significantly, while the rainfall frequency increased slightly. In the post urbanization period, the rainfall intensity and frequency all increased significantly. By comparing the spatial changes of rainfall intensity, it was found that the spatial changes in different urbanization periods were opposite. Rainfall intensity increased significantly in Guangzhou and Dongguan in the pre urbanization period, while in Shenzhen, Foshan, and Zhongshan intensity increased in the post urbanization period, and extreme rainfall intensity and frequency rates showed a greater increase in the post urbanization period than in the pre urbanization period. This pattern of change roughly aligns with the spatial and temporal characteristics of urban expansion in the PRD core area. Wang et al. (2015) showed a strong positive correlation between extreme rainfall events and urban spatial patterns in the PRD, with stronger correlations of extreme indicators at shorter temporal scales. The spatial and temporal expansion of the PRD core area is characterized by the rapid urban expansion of Shenzhen and Dongguan in the pre urbanization period, with construction land concentrated in Guangzhou and Dongguan. In the post urbanization phase, Foshan, Zhongshan, and Zhuhai showed the highest rate of urban expansion. The high level of land urbanization in each region comprises the core area of the PRD, with urban expansion occurring at higher rate during the post urbanization period than the pre urbanization period. What causes land use changes to significantly impact rainstorms can be explored by analyzing the changes of thermodynamic and water vapor variables between the two land Fig. 10 Changes in critical environmental indices during the whole urbanization period in the Pearl River Delta core area and surroundings. Grids with values not significant at the 95% confidence level (Student's t test) have been masked out. a Sensible heat flux (SHF), b latent heat flux (LHF), c level of free convection (LFC), d lifted condensation level (LCL) and e planetary boundary layer height (PBLH) ◂ 6 http:// gloss ary. amets oc. org/ wiki/ Level_ of_ free_ conve ction. 7 http:// gloss ary. amets oc. org/ wiki/ Lifti ng_ conde nsati on_ level. 8 http:// gloss ary. amets oc. org/ wiki/ Atmos pheric_ bound ary_ layer. use scenarios (S2018−S1982). The results show that sensible heat flux increases significantly with expanding urbanization, which influences the thermodynamic factors such as thickening planetary boundary layer height, increasing levels of free convection, and lifting condensation levels (Zhang et al. 2019). The latent heat flux decreases significantly, which leads to the decline of maximum convective available potential energy . These changes will ultimately affect cloud formation and form conducive rainfall conditions, thus increasing the intensity and frequency of urban rainfall . At the same time, the heat island effect increases the thickness of the planetary boundary layer and vertical wind speed during urban expansion. This, coupled with the dry island effect, results in more water vapor being transported upward, promoting the formation of rainstorms and contributing to extreme rainfall events in urban areas .
Affected by the monsoon climate, the PRD region receives abundant rainfall, which is generally concentrated in April to September. Therefore, the simulation period should be appropriately extended to consider rainfall changes more comprehensively. The simulation results of the WRF model match well with the observation data in this study, especially the temperature at 2 m and relative humidity. But there are still some differences overall. In a potential future study, default parameters of the WRF-SLUCM model can be replaced with the characteristics of the PRD, such as types of building geometry and material, in order to make the results align better with the real-world scenario. Additionally, the PRD is affected more severely by typhoons than other areas are, especially in the summer. Therefore, a higher-resolution model that considers the impact of typhoons can be used to explore the impact of land use changes on summer rainfall in the core area of the PRD. Existing studies have found that global warming, anthropogenic heat, and aerosols also have an impact on rainfall. Therefore, changing the parameters of the WRF model to explore the impacts of these factors on rainfall is also a fruitful potential future research direction.

Conclusion
This study chose the PRD core area as its study area, analyzed the spatial and temporal pattern of urban expansion in different urbanization periods and its influence on summer rainfall characteristics based on three temporal situations of land use (1982, 2000, and 2018) and the WRF numerical simulation experiment, and explored the causal variables that have an effect on summer rainfall due to urbanization. This urban expansion of the PRD core area mainly occurred in the post urbanization period Throughout the urbanization period, land use changes have enhanced extreme rainfall intensity, and shifted the rainstorm center toward the core area. The impact of land use changes on rainfall varies in different urbanization periods. Land use changes affect rainfall by creating changes in thermodynamics and water vapor transfer. Urban expansion changes the surface energy balance, which affects thermodynamic variables, and leads to better conditions for rainfall. At the same time, the heat island effect thickens the planetary boundary layer and increases vertical wind speeds, which initiates a dry island effect. Ultimately, more water vapor is transported upward, which promotes water droplet condensation, rainstorm formation, and extreme rainfall events in urban areas. Fig. 11 Changes of relative humidity (RH) (a), temperature at 2 m (T2) (b), water vapor mixing ratio (WVMR) at 2 m (c), WVMR at 850 hPa (d), and vertical wind speed at 850 hPa (e) in the whole urbanization period in the Pearl River Delta core area and surroundings. Grids with values not significant at the 95% confidence level (Student's t test) have been masked out ◂