Urbanization Signicantly Reduces Frequency of Light Rain: an Example From Beijing City

An hour precipitation dataset of 42 automatic weather stations is developed and applied to analyze the temporal and spatial characteristics of light precipitation in urban areas of Beijing City during 2007-2017. The stations are classied into three groups, including 18 sites in central urban area (4th RR), 10 sites in peri-urban area (4th-5th RR) and 14 sites in suburban area (5th-6th RR). Light precipitation is dened as hourly rainfall of 0.1-0.3mm. Analysis shows that light precipitation occurred in urban area in the whole day, with the peak value in 0600 LST and minimum value in 1600LST; monthly variation of light precipitation frequency (LPF) was characterized by the highest value in summer and the lowest value in winter; remarkable differences are found for the various urbanized areas, with the annual and seasonal mean LPF being generally small in central urban area and gradually increasing toward suburban area; the hourly mean LPF during morning and nighttime is higher in summer than those in other seasons in each of the urban areas; relative humidity, aerosol and wind speed might have been the major inuential factors for the observed temporal and spatial pattern of light precipitation.

occur in recent decades (Ren et al. 2000;Fu et al., 2008;Gao et al., 2018). In addition to the increasing extremely intense precipitation events that lead to more damages and losses, the reduction of light rains is also remarkable across mainland China Ma et al., 2015;Ren et al. 2015).
Unlike heavy precipitation, light precipitation plays a more positive role in climate and environment. It can be absorbed directly by soils to maintain soil humidity and to alleviate drought (Trenberth, 2011). Due to the important effects on drought and agriculture, studies on light precipitation draw more attentions for many years over the world. The studies on long-term change of rain frequency have found that there are fewer rain days in recent years, but the daily precipitation intensity is on rise (Zhai et al. 2005;Liu et al. 2005;Pendergrass and Hartmann, 2014). The most important contribution to this reduction comes from a decrease in light-rain days (Gong et al. 2000;Liu et al. 2005;Fu et al. 2008). Decreased light precipitation frequency has also been reported in many other countries including Japan, India and so on (Fujibi, 2005;Goswami et al. 2006). In China, a decrease of light-rain days has been observed in many regions as well as in the country as a whole (Huang and Wen, 2013;Li and Su, 2014;Ren et al. 2015;Zhou et al., 2020). Change in temperature, aerosol concentrations, cloud, and relative humidity are examined to explain the possible reasons for light precipitation decrease, but no consensus has been reached (Huang and Wen, 2013;Wu et al., 2016. Wen, et al., 2016, though Ren et al. (2016) attributed the reduction to a combined in uence of urbanization, aerosols and measurement error.
The previous studies have been focused on regional scale. The detailed comparison among different regions or between urban and rural areas is lacking. Besides, the previous studies paid much attention to long-term trends of the light precipitation during recent decades, while its climatology in particular its diurnal variation characteristics in different areas was rarely analyzed (Huang and Wen, 2013;Wen et al., 2016;Zhou et al., 2020). One of the main reasons is the lack of high-density observation records. On the other hand, a large number of automatic weather stations (AWSs) have been set up in China till the past several years, based on the operational standard issued by China Meteorological Administration (CMA).
By 2017, a dense AWS network has nally been built in Beijing area (Yang et al , 2019. About twothirds of the stations attached to the network can provide hourly precipitation data.
The purpose of this paper is to examine the spatial and temporal variation of light precipitation in Beijing urban areas and the urban effects on light precipitation. Beijing urban area will be classi ed as three types for analyzing the different features of light precipitation in detail. The causation of the light precipitation variation will also be brie y discussed before the conclusions are drawn.

Study Area, Data And Methods
Beijing Municipality, a high-density mega city covering about 16.4 thousand square kilometers, is located at the northern end of the North China Plain and in the south to the Yanshan Mountains. The at southeast area occupies about 38% of the total area of Beijing, and the elevations are roughly below 100 meters above sea level. The north and west of Beijing are mainly mountainous areas. Climatologically, Beijing is characterized by a typical temperate continental monsoon climate, which is hot and humid in summer and cold and dry in winter. The rainy days and precipitation amount in Beijing are highly concentrated in summer (Jun-Jul-Aug), contributing to at least two thirds of its annual total precipitation (Yang et al. 2019).
With accelerating urbanization, population in Beijing City experiences a dramatic growth and over half of the population settled in the urban areas. Along with the bewildering urban sprawl, a multi Ring-Road (RR) system of transportation (including 4th, 5th and 6th RRs, Fig. 1) has been developed in the urban regions (Yang et al. 2013). The zones inside different RRs actually represent different levels of urbanization with varied densities of city population and buildings.
The observed hourly rainfall amount data recorded by 56 stations inside the 6th Ring-Road from 2007 to 2017 are obtained from the Meteorological Information Center, Beijing Meteorological Bureau (MIC/BMB). The data have been preliminarily checked and quality-controlled by the MIC, and the possibly wrong records have been taken out by applying a regionally climatological extreme value method. The data were con rmed to be accurate in the previous studies. (Yang et al. 2011(Yang et al. , 2013. The stations with too many erroneous records have been ruled out from the dataset directly, and the methods of threshold values and manual identi cation are also applied in checking. Based on the observational records and the precipitation feature of Beijing, different seasons have been given different threshold values. Each hourly record above its seasonal threshold is checked and then is adjusted if proved wrong by comparing the records of nearby stations. Only those sites with missing values less than 3% of the total records for the 11 years are selected for use, in order to minimize the in uence of missing records on the analysis. Additional description of quality-control techniques was reported in details by Yang et al. (2011Yang et al. ( , 2013. After the quality control, 42 observational stations inside the 6th RR are used ( Fig. 1) eventually.
To gure out the urban effects on light precipitation among the distinct sites of the built-up areas, all the selected urban stations are geographically classi ed into three groups: those inside the 4th RR or central urban area (comprised of 18 stations), those between the 4th-5th RRs or peri-urban area (comprised of 10 stations) and those between the 5th-6th RRs or suburban area (comprised of 14 stations). The RR loops well represent the different boundaries of varied urbanized zones at present, and also the boundaries of urban-rural zones in different time periods, in Beijing (Mu et al., 2012). The densest and tallest buildings and almost all the central commercial areas concentrate within the 4th RR.
In the study, the hourly precipitation is de ned as rain record equal or more than 0.1 mm precipitation amount accumulated during an hour. In the previous studies, light precipitation is generally regarded as 0.1-10 mm (Zhou et al., 2020) based on daily precipitation data, but a clear de nition of light rain based on hourly precipitation is rare. Considering the feature of Beijing precipitation, light precipitation in this paper is de ned as hourly rain record with 0.1-0.3 mm. The light rain as de ned in this paper contains the slightest portion of the daily precipitation based light rain.
Due to the slight accumulated amount of the light precipitation, the statistic result of precipitation reveals that the feature of light precipitation amount is in accordance with its occurrence. This paper mainly analyzes the feature of light precipitation frequency (LPF) in order to unravel the general pattern of the category of hourly precipitation in Beijing urban and suburban areas. For each site, the annual/seasonal/monthly mean LPF is the total number of precipitation hours per year/season/month. Besides, in order to compare the results of different areas, the average values of LPF of those sites inside each of the zones are also be calculated.

Characteristics of total precipitation
An overview of the characteristics of precipitation in Beijing city is helpful for studying the light precipitation. In Beijing, most rainfall events occur in warm seasons, especially in July and August. We calculated the annual mean frequencies and amounts of precipitation of each site inside the 6th RR, as shown in Fig. 2a and2b.
The distributions of annual mean frequencies and amounts of precipitation are different. There are at least three high-value zones in the frequencies map, scattering in urban center and its northwestern and northeastern parts. The largest mean frequency of more than 370 hours per year is registers at Zi ZhuYuan (ZZY) station. In the southern areas farther from urban center, the mean frequencies are usually less than 290 hours per year. The distribution of annual mean rainfall amount ( Fig. 2b) is more consistent with the divergent expansion of built-up areas and transportation system. The annual mean amount of precipitation in the study region is 584mm, which is a bit higher than the average rainfall amount (561.5mm) over the whole Beijing Municipality. An obvious high-value center appears in the eastern part of central urban area, with Si Hui Qiao station surpassing 680mm. The northwestern parts witness the lowest mean precipitation amounts, generally less than 520mm, and the lowest center appears in Dao XiangHu (DXH). Our previous study of urban heat island (UHI) showed a negative UHI intensity at this station (Yang et al., 2013). The sparser buildings and negative UHII might be one of the reasons for the lowest amount of precipitation.  (Fig. 3a) and monthly (Fig. 3b) cycle of LPF over the study areas. Overall, light precipitation occurred in each hour of the whole day. For the study period, an early morning maximum (0600 LST) of 5.6 times is seen in the mean value of LPF. The secondary maximum value occurs at 1100 LST. The minimum of LPF appears during the afternoon (1600 LST) at a value of 3.7 times. The standard deviation (SD) of each hour in day and night has no signi cant difference. The maximum value of SD (1.21 times) is observed around 0600 LST, which is coincided with the time of peak mean frequency. The minimum value of SD (0.66 times) occurs around 2200, which is different from the valley time (1600 LST) of LPF. The diurnal variation of light precipitation amounts and their standard deviations are almost as same as frequency (not shown). The average amount of light precipitation in a day is 0.60mm, with the peak of 0.73mm and the valley of 0.50mm. Late morning at 1100 LST can be considered as a dividing point, with a high level of LPF dominant before 1100, especially from 0200 to 1100 LST, and a low level of LPF prevailing after 1100, particularly during the period of afternoon (1300-1700 LST).

Characteristics of light precipitation
It is notable from Fig. 3b that monthly variation is in accordance with the climatological feature of precipitation over Beijing. The average monthly value of LPF is 9.1 times. In summer, LPF owes high values of 15.5 times, while the low occurrence of 2.6 times is in winter. The monthly curve shows the peak value of LPF in July (17.1 times), accounting for 36.9% of all. Additionally, values in July and September are higher than those in the rest months, the sum of which accounts for about 67.0% of all light precipitation in whole year. It is interesting to note that the monthly mean value of LPF in August is not the largest during summer period, with a mean value 13.7 times, about 2 times lower than that in July or September. The minimum value appears in January with only 1.4 times. Standard deviations of each month are closely related to monthly mean value of LPF. SD is higher in summer but lower in winter. The convective heavy precipitation mainly happens in summer, and the rainfall has a larger temporal and spatial variability and higher SD as compared to that in winter.
The spatial distribution of annual mean LPF is shown in Fig. 4. The regional average annual mean LPF is 109 times. The maximum LPF center (134.2 times) appears at northeastern part of the urban areas. Other two large LPF centers are both located in the western part. It is obvious that all high LFP centers are in suburban area. The station with minimum LPF (83 times) is Yu Yuan Tan (YYT) in the northwest of the urban center inside the 4 th RR. This station is a small park, surrounded by dense buildings and several expressways.
In general, the distribution of LPF seems to be closely related to the urbanization. The more urbanized a site, the lower value the LPF. However, there are a few exceptions. An obvious relatively higher center occurs at the central area (Fig. 4), with three stations registering the annual mean LPF of 124.4, 123.8 and 122.9 times respectively. The site named Tian An Men (TAM) is next to the Tian An Men Square, which is probably the most spacious urban square in the world. The rest two stations are both in parks, surrounded by more grass, trees and water. The common feature of the three sites is that their surroundings are open and spare, being similar with the sites relatively far from urban center. Figure 5 shows the spatial distribution of seasonal mean LPF. There is a signi cant difference among seasons. The mean LPF is much larger in summer than any other seasons. The average value in summer is 46.3 times, accounting for 42.5% of the whole year. During the period, the seasonal mean LPF shows a considerable reduction toward suburban area, with 40 times isotherm covering most sites in central urban area. Spring and autumn register the moderate mean LPFs, with the average values of 21.7 times and 33.2 times respectively. However, there is an apparent difference between them. Although the LPFs of sites away from urban center are relatively higher in spring and summer, they are not the highest ones. In spring, the maximum value of LPF (29.1 times) occurs in Yun Gang (YG), which is in the southwestern part of the suburban area, while the largest value (51.1 hours) in summer is recorded by Lou Zi Zhuang (LZZ) in the northeastern part of the suburban area. The lowest center of spring and autumn is also different. The minimum value of site in spring is JG near the south 5 th RR, while the lowest value in autumn is at Si Hui Qiao (SHQ), western part of the central area. Besides, the central urban area also undergoes an opposite change between spring and autumn. There is a small center with high values in spring, while it does not appear in autumn. The seasonal mean LPF has the lowest value in winter, with most areas less than 8 times. Interestingly, sites with winter mean LPF more than 10 times are all in central urban area, which shows a completely opposite pattern to other seasons.
It is easily understandable that the annual and seasonal mean LPF is the highest in the suburban area and the lowest in the central urban area among the above three areas. Suburban (central urban) area is characterized by a lower (higher) condensation level during nighttime and a larger (smaller) relative humidity all the hours of a day (Yang et al. 2019), leading to the highest (lowest) LPF due to the shorter (longer) distance for raindrops to travel and less (larger) evaporation. The higher concentration of aerosols in central urban area may also contribute to the lower LPF as compared to the suburban area.
The high mean LPF in winter in central urban area might have been related to the stronger UHI intensity which leads to stronger convection. Figure 6 shows the annual mean diurnal LPF variation in the three urban areas. The standard deviation of each area is also given in the gure. It is obvious that there is a common diurnal variation pattern among the three areas, which is similar to the diurnal cycle of the whole area. However, there are signi cant differences among the urban and suburban areas.

Comparison among different zones
The rst feature is that, the greater the urbanization, the lower the average LPF. The regional average annual mean LPF is 4.33, 4.35, 4.84 times in central urban area, peri-urban area and suburban area respectively. Although the annual mean LPF difference between central urban and peri-urban areas, the annual mean LPF in nighttime during 2200-0500 LST in central urban area is smaller than that in the periurban area. Daytime (0600-1800 LST) can be divided into two stages of morning (0600-1100 LST) and afternoon (1200-1800 LST). In the morning, the annual mean LPF in the central urban area is larger than that in the peri-urban area except for 0700 and 0900 LST. However, the afternoon stage shows an opposite pattern. The value of daytime LPF in the central urban area is more likely to be lower except for 1300 and 1600 LST, while the annual mean LPF in daytime in the peri-urban area is larger.
Early morning from 0200 to 0600 LST is characterized by a lower condensation level in all areas and a stronger UHII in central urban area, and this would be the main reason for the highest LPF in particular in central urban area. Similarly, the hourly LPF remain relatively low from 1300 to 1700 LST, because of the largely heightened condensation level, a weaker UHII and the lower relative humidity or large vapor pressure de cit, compared to the suburban area, during this time of the day. Occurrence time 1500 1600 1600 Figure 7 shows the monthly mean LPF and standard deviation in the three urban areas. The regional average annual mean values of LPF are 8.69, 8.82 and 9.79 times in the central urban area, peri-urban area and suburban area respectively ( Table 2). The maximum values in different zones all appear in June, and all of the minimum values are in January. It is also noted that, in wet seasons, the seasonal transitions are more abrupt and variable than those in dry seasons. The period from June to September is peak stage of light precipitation amount. The period from November to March is also noteworthy, because the monthly mean LPF in 4 th RR during this period is greater than during the other periods of the year. It also has a lower variability. Due to the rare occurrence of LPF, the disparities among the three urban areas and the standard deviations remain small during the cold season.  Figure 8 presents the hour-month plots of the LPF averaged for the different urban areas for the period 2007-2017. The detailed comparison generally con rms that the hourly mean LPF during morning and nighttime is higher in summer than those in other seasons in each urban zone. Nevertheless, higher daytime LPF in August is less obvious within the above areas. Once again, considerable differences exist among the different urban areas. In the suburban area, the hourly mean LPF can reach as high as 1.04 times during the early morning stage in July, approximately 0.95 times higher than that in the afternoon stage of December. By contrast, the maximum LPF in the central urban area is 0.91 times, which happens around 2100 LST in July, and the relatively larger LPF periods among summer months in this area are also different. For example, the large LPF period in July is in early morning (0500-0800 LST) and late evening (2100-2200 LST), while the peak stage in September is from 0200 to 0600 LPF, 2 hours earlier than that in July, and the values at nighttime are not so notable. Between the 4th-5th RRs, high LPF values in summer is relatively stable, prevailing in early morning (0200-0700 LST), and even a few of secondary high values are registered in the evening (1800-2100 LPF) in June.

Discussion
Although the amount of light precipitation accounts for only 2.5% of total precipitation in Beijing urban area, the of light precipitation frequency (LPF) accounts for 33.3% of the total rainy hours per year. As one of the major types of hourly precipitation events, the in-depth understanding of temporal and spatial characteristics of light precipitation, and the urban effect on light precipitation process, is scarce and insu cient. Table 3 and the above analysis, shows that the light precipitation seems to be less in both of frequency and amount in more urbanized areas. This indicates that urbanization has changed the pattern of precipitation. The similar conclusion has been found in previous studies, which were mainly focused on heavy or extreme rainfall events (Chakravari and Archibold, 1993;Rorbert et al., 2003). Combined with the previous observational studies (Yang et al., 2013(Yang et al., , 2019(Yang et al., , 2020, the possible in uence of urbanization on LPF in Beijing urban areas could be discussed as follows. Studies show that relative humidity of low-layer atmosphere was closely related to light precipitation (Ren et al., 2016;Zhou et al., 2020).  found that, in urban area of Beijing, especially in urban center, the relative humidity is apparently lower than the sites outside building-up area, forming remarkable urban dryness island effect. The feature of relative humidity distribution inside the 6th RR of Beijing city is well in accordance with the spatial distribution of LPF shown in Fig. 5.
The mechanism of relative humidity in uence could be summarized as follows. Firstly, the low relative humidity and large saturation vapor de cit in the boundary layer would facilitate the evaporation of some tiny droplets, well before they fall down to the ground. Secondly, the surroundings of rain gauge in urban area differs from that in rural regions. Lower relative humidity and higher surface air temperature promote evaporation of the in-gauge water, which may affect the collection and reading of light precipitation.
Thirdly, due to urban heat island effect, the condensation-layer heights would rise up in urban area, leading to a higher cloud-base height (Fig. 9). The higher cloud-base height could extend the dropping distance of droplets and would make them more likely to vaporize before reaching on ground (Ren et al., 2016).
The relation between aerosol and precipitation including light rain has been discussed for long time (Williams, 2002;Wu et al., 2017). Although the view that aerosol can change precipitation has been widely accepted, whether the in uence is positive or negative is still in debate. As the condensation nucleus of the cloud, if the aerosol maintains a high concentration level in regions with insu cient moisture, it will be di cult for cloud droplets to become rainy droplets through collisioncoalescence process, and the rain will fail to be formed (Rosenfeld 2000;Stevens and Feingold, 2009). The situation would more likely to exist in urban areas, where more pollutants have been emitted by human activity, and this would result in a further reduction of light precipitation in addition to the mechanism mentioned above (Ren et al. 2016). The sketch of the combined in uence of aerosol effects and UHI is shown in Fig. 9.
Wind speed is also likely to affect the veracity of the light precipitation records. The higher wind speed may increase the possibility of record error because the catch rate tends to decrease (Ren et al., 2016;Zheng et al., 2017). Our previous studies (Yang et al., 2020) have reported a lower wind speed in urban center, and the difference between urban and rural area is more signi cant in spring and winter. In this paper, the monthly variation of LPF among different urban areas shows that the LPF from November to April in central urban area are a little higher compared to the record in the other urban areas. This might have been related to the stronger urban stilling effect in the urban center, which leads to a higher catch rate of gauge and a larger frequency and amount of light precipitation. It is possible that a few stations within the 4th RR also record signi cantly larger annual mean LPF due to the weakened surface wind speed caused by the surrounding buildings, and a few others observe the smaller LPF probably due to the relatively stronger wind speed in an opener ground.
Although the spatial difference of light precipitation is detectable partly because of the varied wind surrounding, the general conclusion that annual and seasonal mean light precipitation frequency, except for winter, will decrease in more urbanized areas is not changed. Urbanization has exerted a signi cant in uence on the temporal and spatial pattern of light precipitation in Beijing city.
The ndings reported in this article would be helpful for deepening our understanding of urban precipitation structure and the effect of urbanization on local climate, and for the planning and management of urban ecosystem and environment. They will also provide a clue to the unsolved question on causes of the long-term decrease of light precipitation observed in many regions of the world over the last decades.

Conclusions
This article applied an 11-year ground-based hourly precipitation dataset to analyze the characteristics of light precipitation among different urbanized areas of Beijing city. Overall, the analysis reveals that the spatial and temporal pattern of light precipitation is highly related to the urbanization. The main conclusions are as follows: (1) The annual mean values of LPF in central urban areas or nearby areas tend to be small, and those in the suburban area are signi cantly larger.
(2) The seasonal mean LPF are larger in summer than any other seasons. Summer also witnesses the most obvious feature of small values in urban center, while an opposite pattern is seen for winter.
(3) The light precipitation occurs in urban area in the whole day, with the peak value in 0600 LST and minimum value in 1600LST; hourly mean LPF during morning and nighttime is higher in summer than those in other seasons in each urban area.
(4) Lower relative humidity, higher aerosol concentration and smaller wind speed might have been the major in uential factors for the observed temporal and spatial pattern of light precipitation in Beijing urban areas.

Declarations
Con ict of Interest