Differentiated Effects of Urbanization on Precipitation in South China

: In this paper, precipitation data from the Tropical Rainfall Measuring Mission (TRMM), together with atmospheric reanalysis data, are employed to identify warm-season precipitation (1998–2014) changes and their association with rapid urbanization in south China. Three urban clusters (Chenyu, Yangtze Delta, and Fujian Guangdong coast) are focused. The results reveal that, for the inland Chengyu urban cluster, a lack of precipitation trend is likely due to insigniﬁcant trends in convective available potential energy (CAPE) and total column water vapor (TCWV). They are likely resulted from a reduced local moisture recycling in urban areas, balanced by an increased evapotranspiration of rural areas, together with a stable advection of water vapor input. For the Yangtze River Delta urban cluster, a negative trend in precipitation is associated with a slightly decreased CAPE and an increased TCWV, but is very likely related to urbanization induced an increased planetary boundary layer (PBL) and reduced land surface evaporation. For the Fujian Guangdong coast urban cluster, a marked positive precipitation trend is well explained by positive trends in CAPE and TCWV. The increased precipitation likely beneﬁts from enhanced moisture recycling due to improved vegetation cover in rural areas, and enhanced advection moisture inputs due to urbanization along the coast. These results suggest urbanization effects on precipitation vary with regional conditions. In the coastal area, urbanization enhances sea breezes, which may beneﬁt precipitation if sea breezes go along with the prevailing moisture. In inland area, urbanization likely leads to a warmer-dryer climate if large-scale land cover keeps stationary.


Introduction
At present, 55% of the world's population resides in urban areas. The urban population is expected to reach 68% by 2050 (the Population Division of the UN DESA, 2018). Urbanization's effect on precipitation has been recognized for four decades [1][2][3]. Such effects are receiving increasing attention because they may exacerbate torrential rain, which threatens urban societies [4]. Previous studies suggest that urban-induced changes in precipitation processes most likely result from one or more of the following five mechanisms [5][6][7][8][9][10]: (1) atmospheric destabilization through influences of mesoscale circulation, (2) increased low-level convergence due to surface roughness, (3) modification of microphysical and dynamic processes by the addition of aerosols, (4) modification of low-level atmospheric moisture content, and (5) large structures create a bifurcation zone that steers storms around cities. These mechanisms potentially have either positive or negative effects on precipitation processes depending on the available moisture in the environment.
Numerous investigations of urbanization impact on precipitation are based on statistical analysis of rain gauge networks [11][12][13], ground-based radar [14][15][16], or model simulations [17][18][19][20][21]. Some authors have applied remote-sensing data (Tropical Rainfall Measuring Mission precipitation data, TRMM) to identify patterns in warm-season precipitation anomalies around cities [22][23][24][25][26][27][28][29][30][31]. Although useful, these studies are limited to As the urban effects on precipitation appear to be different between upwind and downwind to the urban area [8,42,63], the mean May-September 700 hPa horizontal wind during 1998-2014 (the study period) is analyzed, summarized in Figure 1e. The 700 hPa level is chosen as the representative level for the mean steering flow, following previous studies [64]. Southwesterly winds prevail in the South China during the warm-season, which is associated with the East Asian summer monsoon system. As the urban effects on precipitation appear to be different between upwind and downwind to the urban area [8,42,63], the mean May-September 700 hPa horizontal wind during 1998-2014 (the study period) is analyzed, summarized in Figure 1e. The 700 hPa level is chosen as the representative level for the mean steering flow, following previous Water 2021, 13,1386 4 of 15 studies [64]. Southwesterly winds prevail in the South China during the warm-season, which is associated with the East Asian summer monsoon system.
The urban effects are most pronounced during warm-season months [37,63,65]. Warmseason precipitation over China is significantly correlated with the convergence of the water vapor transport and deep convections [66][67][68]. During the warm season, the urbanizationinduced mesoscale circulation is more prominent and can significantly alter boundary layer processes. Thus, the present study focuses on the warm-season months (May-September) of 1998-2014, because the available TRMM rain data just cover a period of 1998-2014.

Data
Monthly atmospheric datasets, including convective available potential energy (CAPE), total column water vapor (TCWV), are obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF, https://apps.ecmwf.int/datasets/data/interim-fullmoda/levtype=sfc/, accessed on 20 February 2019). The data have a resolution of 0.125 • *0.125 • (about 12.5 km), and cover a period of 1979-2019 [69]. Data of hourly surface rain rate (unit: mm·h-1) are obtained from the TRMM 3G68 (hereafter referred to as PR 3G68) dataset, provided by GSFC/NASA (Goddard Space Flight Center, National Aeronautics and Space Administration, ftp://trmmopen.gsfc.nasa.gov/pub, accessed on 20 February 2019). It is a gridded product that combines the TRMM satellite Precipitation Radar (PR), TRMM Microwave Imager (TMI) and TRMM combined instrument at 0. 5 • *0. 5 • (about 50 km) and hourly resolution,1998-2014. The rainfall rate also includes total rainfall from each instrument and convective rainfall and convective fraction. Because the data are gridded hourly and put in a daily file, there are a lot of hours with missing data due to satellite sampling, the percent of sampling error are higher for low rain rate than high rain rate [70][71][72]. The GIMMS 3g NDVI v1 dataset provided by ECOCAST (NASA Ames Ecological Forecasting Lab, https://ecocast.arc.nasa.gov/data/pub/gimms/, accessed on 20 February 2019), at 1/12 • × 1/12 • and 15-days resolution, 1981-2015.

Methods
Although the TRMM dataset is not sufficient to establish true climatological behavior, the 17-year period, overlapped with the rapid urbanization in South China, allows for the investigation of anomalies in precipitation associated with urban effects. The analysis is primarily conducted based on hourly precipitation rates (mm·h −1 ) in 0.5 • × 0.5 • cells.
For more detailed analysis, mean surface precipitation rates at each grid point are calculated for the months of May, June, July, August, and September. For a given grid cell, hourly surface precipitation rates are averaged for the five warm-season months. This average hourly rain rate is multiplied by (24 h × 30/31 days × 5 months) to obtain the warm-season precipitation of the grid cell for the year. This procedure is repeated for each of the 17 years, resulting in a 17-year time series of warm season precipitation for 1998-2014.
The possible relationship between precipitation trend and urbanization in South China is analyzed as follows. Firstly, the 17-year warm season precipitation is examined for trends. Secondly, the variability of atmospheric parameters is examined to explain the precipitation trends (if there are). Thirdly, possible connections between these mechanisms and the urbanization are examined.

Interannual Variation of Warm Season Precipitation in South China
No obvious pattern of any trend of the spatially average warm-season precipitation occurs over South China during 1998-2014, while the population density and urban built-up area, represented by the selected cities increase rapidly (Table 1). Nevertheless, the precipitation trends are observed in subregions. The maximum trend of warm-season precipitation (3-6 mm·mon −1 ·year −1 , Figure 2a) distributes in the downwind and surrounding areas of the Fujian Guangdong coast cluster (Figure 1e). This confirms previous research result: an increasing trend of precipitation existed at the Pearl River Delta [12,13]. Warm-season rainfall decreases over the Chengyu urbanized areas (

Factors
The Linear Fitting Equation result: an increasing trend of precipitation existed at the Pearl River Delta [12,13]. Warmseason rainfall decreases over the Chengyu urbanized areas (Figure 2a, −3 mm·mon −1 ·year −1 ) and increases slightly in its downwind areas (Figure 2a Table 2). Table 2. The trend of rain, CAPE, TCWV, skin-temperature, sensible heat flux, boundary layer height, evaporation, 1000 hPa specific humidity in warm season during 1998 to 2014.

Urban Clusters Chenyu
The Yangtze River Delta

Fujian Guangdong Coast
Rain trend

Association of Spatially Differentiated Trends in Precipitation and Two Atmospheric Variables Surrounding the Three Urban Clusters
Precipitation is a product of condensation of the atmospheric water vapor, in which the atmospheric instability and convection play a significant role. The CAPE is a good indicator for atmospheric instability and convection [30,51]. Previous studies have shown a tight relationship between precipitation and TCWV as well [68,73]. To examine the associated physical mechanisms leading to the precipitation trends observed in surrounding the three urban clusters, variations in CAPE and TCWV are first examined.
The spatially distributed CAPE and TCWV trends are shown in Figure 3 and Table 2 The overall northeast-southwest oriented zone of both negative CAPE and TCWV trends explain the slightly negative trend of warm season precipitation in these areas of South China

Possible Mechanisms
From Section 3.2, it seems that the overall the spatially differentiated warm season precipitation trends in South China are associated with the trends in CAPE and TCWV. Whether and how the urbanization in South China has contributed to these trends remain to be seen. If urbanization has an influence, it must realize through modifying the surface and low atmosphere conditions. Figure 4 and Table 2 show the changes in the surface and lower atmosphere during warm-season of 1998-2014 in South China. The urbanization effects on skin temperature are clear. The maximum trend value of warm-season skintemperature is 0.12 °C·year −1 (Figure 4a), occurs in Chengyu urbanized areas. The trend of the Yangtze River Delta urban cluster and the Fujian Guangdong coast urban cluster are around 0.03 °C·year −1 , which is close to the results of previous studies based on model simulation [20,21]. Our research, based on remote-sensing data, classify and discuss the possible mechanisms of urban effect on precipitation under different geographic and climatic conditions. . For this urban cluster, urban precipitation has decreased slightly, while the rural areas experienced a small increasing precipitation. During the same period, no significant change in large-scale advection of water vapor input has occurred to the region (Figure 4f). These can be explained by, in the marked urban areas, the decreased annual maximum Normalized Difference Vegetation Index (NDVI, Figure 5, −50-0 index·year −1 ) leads an enhanced skin-temperature, an increased sensible heat flux and a higher PBL,

Possible Mechanisms
From Section 3.2, it seems that the overall the spatially differentiated warm season precipitation trends in South China are associated with the trends in CAPE and TCWV. Whether and how the urbanization in South China has contributed to these trends remain to be seen. If urbanization has an influence, it must realize through modifying the surface and low atmosphere conditions. Figure 4 and Table 2 show the changes in the surface and lower atmosphere during warm-season of 1998-2014 in South China. The urbanization effects on skin temperature are clear. The maximum trend value of warm-season skintemperature is 0.12 • C·year −1 (Figure 4a), occurs in Chengyu urbanized areas. The trend of the Yangtze River Delta urban cluster and the Fujian Guangdong coast urban cluster are around 0.03 • C·year −1 , which is close to the results of previous studies based on model simulation [20,21]. Our research, based on remote-sensing data, classify and discuss the possible mechanisms of urban effect on precipitation under different geographic and climatic conditions.

Possible Mechanisms for Stationary Warm Season Precipitation in the Chengyu Urban Cluster
Over and upwind of Chengyu urbanized areas, there are significant trends of skin-temperature (0.12 • C·year −1 , Figure 4a . For this urban cluster, urban precipitation has decreased slightly, while the rural areas experienced a small increasing precipitation. During the same period, no significant change in large-scale advection of water vapor input has occurred to the region (Figure 4f). These can be explained by, in the marked urban areas, the decreased annual maximum Normalized Difference Vegetation Index (NDVI, Figure 5, −50-0 index·year −1 ) leads an enhanced skin-temperature, an increased sensible heat flux and a higher PBL, and hence the water vapor to be mixed more evenly in a larger thickness in the lower atmosphere. The decreased evaporation associated with a higher PBL leads to a decreased 1000 hPa specific humidity, which reduces latent heat release with air masses upward motion. Reduced local moisture recycling in the urban cluster, contributes to reducing CAPE. For the rural areas downwind from this urban cluster, local moisture recycling increases (including evaporation and TCWV) likely associated with improved vegetation cover in the rural China in recent decades [29,54,56], an increased trend of annual maximum NDVI in this rural area (Figure 5, 50-100 ndvi·year −1 ). Thus, the overall insignificant precipitation trend over this urban cluster and its downwind area is likely supported by little change in large-scale circulation (Figure 4f), and the offset between reduced urban moisture recycling and increased rural evapotranspiration. Figure 4b), and a PBL height (2-4 m·year −1 , Figure 4c). They together explain a slightly decreased CAPE. An increased TCWV is supported by an increased 1000 hPa specific humidity (2 × 10 −5 kg·kg −1 , Figure 4e), but not by a decreased evaporation (around −1 × 10 −7 Kg·m −2 ·s·year −1 , Figure 4d).
The negative trend in precipitation, is not explained well by a slightly decreased CAPE and increased TCWV. Nevertheless, it is consistent with an increased PBL, reduced land surface evaporation. The reduced local moisture recycling is associated with urbanization in this area, reduced annual maximum NDVI ( Figure 5, −50-0 ndvi·year −1 ), this is consistent with the result of the cited literature [57]. In addition, the urbanization has led to a slightly increased land-sea 1000-925 hPa average temperature difference (−0.01-0.01 °C·year −1 over sea and 0.01-0.02 °C·year −1 over land, Figure 6), which likely has enhanced sea breezes in the daytime. However, the sea breezes may go against synoptic prevailing moisture direction. Their impact on precipitation is not clear. Thus, very likely the observed negative precipitation trend in the Yangtze River Delta is mostly likely associated with urbanization induced an increased PBL, and a reduced land surface evaporation.   (e) (f)
The negative trend in precipitation, is not explained well by a slightly decreased CAPE and increased TCWV. Nevertheless, it is consistent with an increased PBL, reduced land surface evaporation. The reduced local moisture recycling is associated with urbanization in this area, reduced annual maximum NDVI ( Figure 5, −50-0 ndvi·year −1 ), this is consistent with the result of the cited literature [57]. In addition, the urbanization has led to a slightly increased land-sea 1000-925 hPa average temperature difference (−0.01-0.01 • C·year −1 over sea and 0.01-0.02 • C·year −1 over land, Figure 6), which likely has enhanced sea breezes in the daytime. However, the sea breezes may go against synoptic prevailing moisture direction. Their impact on precipitation is not clear. Thus, very likely the observed negative precipitation trend in the Yangtze River Delta is mostly likely associated with urbanization induced an increased PBL, and a reduced land surface evaporation.
zation, but more likely associated with an improved annual maximum NDVI ( Figure 5, 0-50 ndvi·year −1 ), vegetation cover improved in this region [53,55,57]. Local positive feedback provides a favorable moisture environment. Nevertheless, the urbanization leads to an increased land-sea 1000-925 hPa average temperature gradient (0.01-0.02 °C·year −1 over sea and 0.03-0.04 °C·year −1 over land, Figure 6) [74], enhancing sea breezes. The direction of sea breezes in this region, partly along with the prevailing wind (Figure 4f), likely increases advective moisture input to the area.
. Thus, the increased precipitation in this urban cluster is likely benefitted from enhanced moisture recycling due to an improved vegetation cover in the rural areas, and enhanced adjective moisture inputs due to urbanization along the coast.

Discussion
Based on the above discussion, we conceptualize the possible mechanisms for different precipitation trends at the three urban clusters (Figure 7).
In the Chenyu cluster, the urbanization has led to a reduced moisture recycling and an elevated PBL. Both tend to reduce CAPE and TCWV. These effects, however, are offset by increasing moisture recycling in the mountainous rural areas where vegetation cover The reduced PBL and increased local moisture inputs are unlikely related to urbanization, but more likely associated with an improved annual maximum NDVI ( Figure 5, 0-50 ndvi·year −1 ), vegetation cover improved in this region [53,55,57]. Local positive feedback provides a favorable moisture environment. Nevertheless, the urbanization leads to an increased land-sea 1000-925 hPa average temperature gradient (0.01-0.02 • C·year −1 over sea and 0.03-0.04 • C·year −1 over land, Figure 6) [74], enhancing sea breezes. The direction of sea breezes in this region, partly along with the prevailing wind (Figure 4f), likely increases advective moisture input to the area.
Thus, the increased precipitation in this urban cluster is likely benefitted from enhanced moisture recycling due to an improved vegetation cover in the rural areas, and enhanced adjective moisture inputs due to urbanization along the coast.

Discussion
Based on the above discussion, we conceptualize the possible mechanisms for different precipitation trends at the three urban clusters (Figure 7).
In the Chenyu cluster, the urbanization has led to a reduced moisture recycling and an elevated PBL. Both tend to reduce CAPE and TCWV. These effects, however, are offset by increasing moisture recycling in the mountainous rural areas where vegetation cover have greatly improved in recent decades [29,54,56]. Thus, overall, no marked warm season precipitation trend is observed. Should there be no vegetation cover improvement in rural area, the urbanization in this inland region, would likely become warmer and dryer.
In the Yangtze River Delta cluster, urbanization has diffused into the whole area, leading to a reducing local moisture recycling, and an elevated PBL. Both contributes to a lower CAPE. Southerly anomaly of moisture input may slightly strengthen large scale moisture inputs to the region for precipitation. This enhancement may have contributed to surface specific humidity and TCWV, but this has not translated to an increased precipitation, likely limited by a slightly negative CAPE trend. vide the explanation. First, the Fujian Guangdong cluster is surrounded by a large area of mountainous landscapes where vegetation cover has greatly improved in recent decades [53,55,57], enhancing local moisture recycling, while this has not occurred the flat areas surrounding Yangtze River Delta [57]. Second, the urbanization enhanced sea breezes, tend to go along with the prevailing moisture direction for the Fujian Guangdong cluster, while against the prevailing moisture direction for the Yangtze River Delta cluster.  In the Fujian Guangdong coast cluster, urbanization enhanced land-sea thermal contrast has likely induced stronger sea breezes. Meanwhile, local moisture recycling is enhanced resulting from improved vegetation cover in the rural mountainous regions [53,55,57]. Both processes have resulted in positive trends in TCWV and CAPE. They together explain the positive warm season precipitation trend.
Both along the coast, the Fujian Guangdong cluster area has seen an increase precipitation, while Yangtze River Delta has a slightly decrease trend. Two differences may provide the explanation. First, the Fujian Guangdong cluster is surrounded by a large area of mountainous landscapes where vegetation cover has greatly improved in recent decades [53,55,57], enhancing local moisture recycling, while this has not occurred the flat areas surrounding Yangtze River Delta [57]. Second, the urbanization enhanced sea breezes, tend to go along with the prevailing moisture direction for the Fujian Guangdong cluster, while against the prevailing moisture direction for the Yangtze River Delta cluster.

Conclusions
Using a 17-year warm season analysis of hourly rainfall rates from the TRMM 3G68 PR, we have examined the possible connections between precipitation trends and urbanization in the region of South China. Key findings are summarized as follows: (1) There was no marked warm-season precipitation trend in Chengyu urban cluster. A downward trend of precipitation occurred over the Yangtze River Delta urban cluster. The Fujian Guangdong coast urban cluster experienced an obvious precipitation increase.
(2) For Chengyu urban cluster, a lack of precipitation trend is likely due to insignificant trends in CAPE and TCWV. These can be explained by a reduced local moisture recycling in the urban area, balanced by an increased evapotranspiration in the rural areas, together with a stable large-scale advection of water vapor input to the region.
(3) The negative trend in precipitation for the Yangtze River Delta cluster, is not explained well by a slightly decreased CAPE and increased TCWV, but seems to be associated with an increased PBL and reduced land surface evaporation. They both can be induced by rapid urbanization in this area.
(4) For the Fujian Guangdong coast urban cluster, a marked positive precipitation trend is explained by positive trends in CAPE and TCWV. The increased precipitation in this urban cluster likely benefits from enhanced moisture recycling due to improved vegetation cover in the rural areas, and enhanced adjective moisture inputs due to urbanization along the coast.
(5) In additional to urbanization, large-scale vegetation cover improvement in south China seems to have an impact on precipitation via increasing local moisture recycling. Without this vegetation cover change, an inland urbanization (e.g., Chenyu) likely makes the area warmer and dryer. In the coastal area, urbanization enhances sea breezes, which may benefit precipitation if sea breezes go along with the prevailing moisture. Therefore, in order to reduce the risk of making the area warmer and dryer, it is necessary to improve the large-scale vegetation cover in the process of urbanization.