Spatiotemporal Variation in Precipitation during Rainy Season in Beibu Gulf, South China, from 1961 to 2016

: The spatiotemporal variation in precipitation is an important part of water cycle change, which is directly associated with the atmospheric environment and climate change. The high-resolution spatiotemporal change of precipitation is still unknown in many areas despite its importance. This study analyzed the spatiotemporal variation in precipitation in Beibu Gulf, South China, during the rainy season (from April to September) in the period of 1961–2016. The precipitation data were collected from 12 national standard rain-gauge observation stations. The spatiotemporal variation in precipitation was evaluated with incidence rate and contribution rate of precipitation. The tendency of variations was analyzed using the Mann–Kendall method. The precipitation in the rainy season contributed 80% to the total annual precipitation. In general, there was an exponential decreasing tendency between the precipitation incidence rate and increased precipitation durations. The corresponding contribution rate showed a downward trend after an initial increase. The precipitation incidence rate decreased with the rising precipitation grades, with a gradual increase in contribution rate. The precipitation incidence rate and contribution rate of 7–9 d durations showed the significant downward trends that passed the 95% level of significance test. The results provide a new understanding of precipitation change in the last five decades, which is valuable for predicting future climate change and extreme weather prevention and mitigation.


Introduction
The Intergovernmental Panel on Climate Change (IPCC) reports emphasize the global warming [1]. The intensified human activities made serious effects on the regional and global water circulation systems, leading to frequent disasters, such as floods, droughts, and storm surges [2]. Consequently, the human society and natural ecosystems have been terribly impacted [2][3][4][5]. Precipitation is an important part of the water cycle, and scholars have focused on the evolution characteristics and patterns of precipitation in changing environment, from different time scales, such as inter-annual, annual, monthly, rainy season, non-rainy season, and various space scales, for example, global, continent, country, and river basin [6][7][8][9]. Most existing studies have focused on the total amount of rainfall, mean value, extreme value, structural change, and others [10][11][12]. The response characteristics of precipitation to the environmental factors have also been explored. The above studies have laid a foundation for regional water resources planning and management. studied the precipitation in South China. However, most of these studies have focused on the variation trend of total annual precipitation [66], with teleconnection analysis [67]. Few studies have been performed to explore the specific variation of precipitation structure in the rainy season.
In order to fill the knowledge gap, this study was performed based on the daily precipitation data from meteorological observation stations in the Beibu Gulf, Guangxi, South China. The spatiotemporal variation characteristics of the precipitation structure in rainy season were analyzed, from the aspects of rainfall duration and precipitation grade. The incidence rate and contribution rate of precipitation were applied as the evaluation indices. The results will lay the foundation for predicting and guiding regional agricultural and industrial production, engineering construction, as well as disaster prevention and mitigation.

Study Area
The Beibu Gulf was in the southwest end of coastal area of China ( Figure 1) with subtropical monsoon climate. In the northwest part of the region, there is the Xijiang River, a tributary of the Pearl River. In the southeast part, there are Nanliujiang River, Qinjiang River, and Beilun River, and they all flow into the ocean. The region is rich in precipitation, with uneven distribution, and drought and waterlogging occur frequently. Beibu Gulf consists of six prefecture-level cities in Guangxi province: Nanning, Beihai, Qinzhou, Fangcheng, Yulin and Chongzuo. The total area of Beibu Gulf is approximately 72,800 km 2 , with more than 23,000,000 permanent residents by the end of 2018. Beibu Gulf is in the connection area of South China, Southwest China, and Association of Southeast Asian Nations (ASEAN) economic circle. It is the only coastal area in the west China, which is also the existing ocean channel between China and ASEAN countries. Beibu Gulf shares borders with these ASEAN countries, thus enjoying obvious location superiority and the important strategic position [68]. In the last four decades, under the influence of China's the Reform and Opening-Up Policy, rapid economic development has occurred in most parts of China's coastal areas, except for Beibu Gulf. In the 1980s, the regional economic development almost stopped due to the Sino-Vietnam War. In the 1990s, border trade in the region began to develop as Sino-Vietnam diplomatic relations gradually normalized. In 2010, China and ASEAN formally established the China-ASEAN Free Trade Area. Afterwards, the development of the Beibu Gulf has become a national strategy. This region has been regarded as an important international economic cooperation zone, opening to ASEAN for potential cooperation. In recent years, the Beibu Gulf is embracing the new development opportunities under the support of Chinese government's the Belt and Road Initiative [69].

Index Definition
This study focused on the spatiotemporal variation characteristics of the precipitation structure in the rainy season. Based on previous studies and the World Meteorological Organization (WMO) recommendations [51,52,71,72], the comparison of different indices indicates the suitability of

Data
Daily precipitation data were collected from 12 national standard rain-gauge observation stations in the Beibu Gulf, including Longzhou (LZ), Tiandeng (TD), Chongzuo (CZ), Dongxing (DX), Mashan (MS), Nanning (NN), Qinzhou (QZ), Beihai (BH), Hengxian (HX), Bobai (BB), Yulin (YL), and Rongxian (RX) (Figure 1). All of the stations continuously collected daily precipitation data, covering the duration from 1961 to 2016. The precipitation data were obtained from the National Climate Center (NCC) of the China Meteorological Administration (CMA). There are two stations containing missing observation data. Both stations had less than 0.1% of missing values. The durations of continuous gaps were mainly one day. The missing data were replaced by the average values of neighbouring days. This gap filling method has little influence on the result [70]. The data were checked with strict quality control and three aspects inspection (reliability, consistency, representativeness) inspection, showing good integrity.

Index Definition
This study focused on the spatiotemporal variation characteristics of the precipitation structure in the rainy season. Based on previous studies and the World Meteorological Organization (WMO) recommendations [51,52,71,72], the comparison of different indices indicates the suitability of rainfall duration and precipitation grade for the current study. When the daily precipitation equals to or exceeds 0.1 mm, it means the effective precipitation, according to the relevant provisions of the meteorological department [72,73]. Similar to previous studies [51,52,71,72], the precipitation duration was defined as the number of consecutive days for the precipitation events (daily rainfall ≥ 0.1 mm) from the beginning to the end. The precipitation durations were divided into ten categories: 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d, 9 d, and ≥ 10 d. In addition, the precipitation events were classified according to the daily precipitation amount [74], e.g. light rain (0.1 mm ≤ daily rainfall < 10.0 mm), moderate rain (10.0 mm ≤ daily rainfall < 25.0 mm), heavy rain (25.0 mm ≤ daily rainfall < 50.0 mm), and torrential rain (daily rainfall ≥ 50.0 mm). The WMO and the World Climate Research Program (WCRP) have jointly established the Expert Team on Climate Change Detection and Indices (ETCCDI). ETCCDI defined 27 representative climate indices (including 16 temperature indices and 11 precipitation indices), among which the CWD (consecutive wet days) index was a description of such precipitation events. CWD has been widely used in the studies of climate events [51,52,71,72].
The variation characteristics of the precipitation in the rainy season of the Beibu Gulf were comprehensively evaluated with two indices defined in this study: the precipitation incidence rate and precipitation contribution rate [53,55].
The precipitation incidence rate (W) was defined as the ratio of occurrence frequency of a specific precipitation event to the total frequency of all precipitation events: where n is frequency of a specific precipitation event and N is the total frequency of all precipitation events. The precipitation contribution rate (M) was defined as the ratio of precipitation amount of a specific precipitation event to the total precipitation amount of all precipitation events: where d is precipitation amount in a specific precipitation event and D is the total precipitation amount of all precipitation events. The Mann-Kendall test (MK) was applied to evaluate the significance of the variation trends of the incidence rate and contribution rate of various precipitation events. Mann first proposed this method [75], and then modified and improved [76]. At present, it has been widely applied in the meteorological hydrology [77], which has also been recommended by the WMO [78].

Collection of Historical Precipitation Events
Combing with the index analyses, historical precipitation events causing heavy losses in the region were summarized (Table 1). * Direct economic losses were calculated based on the price level in the year of the disaster. "\" means no available record.

Statistics of Precipitation in Rainy Season
The average precipitation in the rainy season (from April to September) was about 1359.59 mm, accounting for 81.20% of the annual rainfall (about 1674.41 mm), according to the statistics of precipitation in the Beibu Gulf ( Figure 2). method [75], and then modified and improved [76]. At present, it has been widely applied in the meteorological hydrology [77], which has also been recommended by the WMO [78].

Collection of Historical Precipitation Events
Combing with the index analyses, historical precipitation events causing heavy losses in the region were summarized (Table 1). * Direct economic losses were calculated based on the price level in the year of the disaster. "\" means no available record.

Statistics of Precipitation in Rainy Season
The average precipitation in the rainy season (from April to September) was about 1359.59 mm, accounting for 81.20% of the annual rainfall (about 1674.41 mm), according to the statistics of precipitation in the Beibu Gulf ( Figure 2).    Figure 3 shows the incidence rates and contribution rates of precipitation events in different durations of rainy season in the Beibu Gulf. The precipitation incidence rate exponentially decreased (y = 49.5e −0.4x , R 2 = 0.9957), along with the increased precipitation duration. The incidence rates of precipitation for 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d, 9 d, and ≥10 d were 32.57%, 21.12%, 14.88%, 10.14%, 6.85%, 4.33%, 2.68%, 2.24%, 1.17%, and 4.02%, respectively. However, for different durations, the variation in contribution rates was different from the variation in the incidence rates; instead, it increased firstly and decreased later along with the increased duration. When the duration ≥ 10 d, there was a significant rising tendency. For the duration of 1-9 d, the contribution rate at 3 d was the highest (14.13%), followed by 4 d (13.94%), 2 d (11.95%), and 5 d (11.26%). When the duration ≥ 10 d, the precipitation incidence rate was only 4.02%, while the contribution rate was up to 16.93%. Figure 3 shows the incidence rates and contribution rates of precipitation events in different durations of rainy season in the Beibu Gulf. The precipitation incidence rate exponentially decreased (y = 49.5e −0.4x , R 2 = 0.9957), along with the increased precipitation duration. The incidence rates of precipitation for 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d, 9 d, and ≥10 d were 32.57%, 21.12%, 14.88%, 10.14%, 6.85%, 4.33%, 2.68%, 2.24%, 1.17%, and 4.02%, respectively. However, for different durations, the variation in contribution rates was different from the variation in the incidence rates; instead, it increased firstly and decreased later along with the increased duration. When the duration ≥ 10 d, there was a significant rising tendency. For the duration of 1-9 d, the contribution rate at 3 d was the highest (14.13%), followed by 4 d (13.94%), 2 d (11.95%), and 5 d (11.26%). When the duration ≥ 10 d, the precipitation incidence rate was only 4.02%, while the contribution rate was up to 16.93%.

Statistical Analysis of Incidence Rate and Contribution Rate of Precipitation Durations
The incidence rates of most precipitation were distributed in the durations of 1-4 d, and the sum of incidence rate reached up to 78.71%. As for the contribution rates, the sum of precipitation for durations of 2-5 d was up to 51.28%, followed by ≥10 d (16.93%).
The probability of flood disasters caused by precipitation was analyzed. The precipitation contribution rate of duration of ≥10 d was up to 16.93%, when the time for forming runoff from rainfall was considered, the disaster probability caused by precipitation of 2-4 d obviously exceeded that of the precipitation of ≥10 d.  Figure 4 shows the incidence rates and contribution rates of different precipitation duration in rainy season at 12 weather stations in the Beibu Gulf. The precipitation incidence rate of each station decreased, along with the increased precipitation duration. In terms of the incidence rates at the durations of 1 d, 2 d, and ≥10 d, there were relatively remarkable differences between regions. Little difference was observed at duration of 3-9 d, indicating no regional difference. For the 12 stations, the incidence rates of precipitation for 3-9 d ranged 13-16%, 9-12%, 6-8%, 3-6%, 2-4%, 1-3%, 1-2%, respectively. Further comparison showed that the precipitation incidence rates for 1 d and 2 d showed the basically consistent spatial variation trend (Appendix A). These results show that the precipitation incidence rates were high in Beihai (in the southern); Nanning, Chongzuo and Longzhou (in the middle and western region). The incidence rate of 1 d was over 34%, where Beihai was up to 40.71%; the precipitation incidence rate of 2 d was 23-24%. Nevertheless, the incidence rates were relatively low in Mashan (in the northern region), Dongxing (in the southern region), and Bobai, Yulin, and Rongxian (in the eastern region). The incidence rates of 1 d and 2 d were 28-30% and 18-20%, respectively. The precipitation incidence rate for ≥10 d had opposite spatial variation trend with those of 1 d and 2 d (Appendix A). The incidence rates at Dongxing, Bobai, Yulin, and Rongxian stations (in the eastern area) were relatively high (5-7%), while the incidence rates at Beihai, Longzhou, Chongzuo, and Nanning stations (in the middle and western region) were relatively low (1.5-2.5%). The incidence rates of most precipitation were distributed in the durations of 1-4 d, and the sum of incidence rate reached up to 78.71%. As for the contribution rates, the sum of precipitation for durations of 2-5 d was up to 51.28%, followed by ≥10 d (16.93%).

Spatial Differences of Precipitation Incidence Rates and Contribution Rates
The probability of flood disasters caused by precipitation was analyzed. The precipitation contribution rate of duration of ≥10 d was up to 16.93%, when the time for forming runoff from rainfall was considered, the disaster probability caused by precipitation of 2-4 d obviously exceeded that of the precipitation of ≥10 d. Figure 4 shows the incidence rates and contribution rates of different precipitation duration in rainy season at 12 weather stations in the Beibu Gulf. The precipitation incidence rate of each station decreased, along with the increased precipitation duration. In terms of the incidence rates at the durations of 1 d, 2 d, and ≥10 d, there were relatively remarkable differences between regions. Little difference was observed at duration of 3-9 d, indicating no regional difference. For the 12 stations, the incidence rates of precipitation for 3-9 d ranged 13-16%, 9-12%, 6-8%, 3-6%, 2-4%, 1-3%, 1-2%, respectively. Further comparison showed that the precipitation incidence rates for 1 d and 2 d showed the basically consistent spatial variation trend (Appendix A). These results show that the precipitation incidence rates were high in Beihai (in the southern); Nanning, Chongzuo and Longzhou (in the middle and western region). The incidence rate of 1 d was over 34%, where Beihai was up to 40.71%; the precipitation incidence rate of 2 d was 23-24%. Nevertheless, the incidence rates were relatively low in Mashan (in the northern region), Dongxing (in the southern region), and Bobai, Yulin, and Rongxian (in the eastern region). The incidence rates of 1 d and 2 d were 28-30% and 18-20%, respectively. The precipitation incidence rate for ≥10 d had opposite spatial variation trend with those of 1 d and 2 d (Appendix A). The incidence rates at Dongxing, Bobai, Yulin, and Rongxian stations (in the eastern area) were relatively high (5-7%), while the incidence rates at Beihai, Longzhou, Chongzuo, and Nanning stations (in the middle and western region) were relatively low (1.5-2.5%).

Spatial Differences of Precipitation Incidence Rates and Contribution Rates
incidence and contribution rates, especially the precipitation contribution rates. At the same time, the dry spell was longer than other regions, especially in the years with less total precipitation. In a two-day flood (29-30 August 1988), Nanning had a total precipitation of 158.1 mm, with the consequence of 400,000 people affected, eight death, more than 2000 houses damaged, 37,300 hectares of crops affected, and the direct economic loss of 130 million RMB Yuan [79]. In another two-day flood (20-21 July 1994), Beihai received 245.1 mm precipitation, causing six people dead, 325 livestock killed, 624 houses damaged, and the direct economic loss of about 350 million RMB Yuan [79]. In three seasons (spring, summer and autumn) during the period of 1961-2000, Nanning experienced the longest dry spell, followed by Chongzuo and Beihai [82].  Figure 5 show the incidence rate and contribution rate of each precipitation grade during rainy season in the Beibu Gulf. With the increased precipitation grade, the incidence rate decreased while the contribution rate of precipitation increased. In terms of precipitation incidence rate, the light rain dominated, accounting for 64.03%, while the contributions of moderate rain, heavy rain, and torrential rain were only 18.55%, 10.76%, and 6.66%, respectively. In terms of the precipitation contribution rate, the light rain only accounted for 13.22%, and the contribution rates of moderate rain, heavy rain and torrential rain were 23.19%, 27.53% and 36.06%, respectively. Therefore, the At all the 12 stations, the precipitation contribution rates increased first and decreased later along with the increase in precipitation duration (Figure 4b). When the duration was ≥10 d, the precipitation contribution rate increased significantly again. Relative to the incidence rates of precipitation, the duration of each station varied largely in terms of the precipitation contribution rate. The contribution rates of all durations were compared, and they were different between 1 d, 2 d, and ≥10 d. The contribution rates for the 3-9 d durations were similar, which were 12-17%, 11-16%, 9-13%, 7-11%, 5-8%, 4-7%, and 3-6%, respectively. This result was basically consistent with the precipitation incidence rates in different durations. Further analysis showed that, for 1 d and 2 d, a similar variation tendency was observed in the precipitation incidence rate and spatial contribution rate. The contribution rate in Beihai (in the southern) and Longzhou, Chongzuo and Nanning (in middle and western) (the contribution rates for 1 d and 2 d were 9-11% and 15-17%) were markedly higher than those in Qinzhou, Dongxing, Bobai, Yulin, and Rongxian (the contribution rates of 1 d and 2 d durations were 4-6% and 8-10%). The precipitation contribution rate for the duration of ≥10 d showed opposite spatial variation tendency with those of 1 d and 2 d (Appendix B). The contribution rates of Qinzhou, Dongxing, Bobai, and Rongxian (in the eastern) were relatively higher (for example, more than 23% and the contribution rate of Dongxing was up to 29.45%), while they were relatively lower in Beihai and Longzhou, Chongzuo, and Nanning (in middle and western, 9-10%).

Statistic Analysis for Incidence Rate and Contribution Rate of Different Precipitation Grades
Longzhou, Chongzuo, and Nanning (in middle and western) and Beihai (in southern region) were more prone to short-term flood disaster, according to above spatial differences in precipitation incidence and contribution rates, especially the precipitation contribution rates. At the same time, the dry spell was longer than other regions, especially in the years with less total precipitation. In a two-day flood (29-30 August 1988), Nanning had a total precipitation of 158.1 mm, with the consequence of 400,000 people affected, eight death, more than 2000 houses damaged, 37,300 hectares of crops affected, and the direct economic loss of 130 million RMB Yuan [79]. In another two-day flood (20-21 July 1994), Beihai received 245.1 mm precipitation, causing six people dead, 325 livestock killed, 624 houses damaged, and the direct economic loss of about 350 million RMB Yuan [79]. In three seasons (spring, summer and autumn) during the period of 1961-2000, Nanning experienced the longest dry spell, followed by Chongzuo and Beihai [82]. Figure 5 show the incidence rate and contribution rate of each precipitation grade during rainy season in the Beibu Gulf. With the increased precipitation grade, the incidence rate decreased while the contribution rate of precipitation increased. In terms of precipitation incidence rate, the light rain dominated, accounting for 64.03%, while the contributions of moderate rain, heavy rain, and torrential rain were only 18.55%, 10.76%, and 6.66%, respectively. In terms of the precipitation contribution rate, the light rain only accounted for 13.22%, and the contribution rates of moderate rain, heavy rain and torrential rain were 23.19%, 27.53% and 36.06%, respectively. Therefore, the dominated precipitation was light rain, and the sum of incidence rates of moderate, heavy, and torrential rain was less than 36%, but the contribution rate was up to about 87%. Precipitation in Beibu Gulf during rainy season was dominated by weak precipitation events, but the main contribution was dependent on strong precipitation events, indicating the large possibility of flood disaster and aggravated situation of regional flood.

Statistic Analysis for Incidence Rate and Contribution Rate of Different Precipitation Grades
Water 2020, 12, x FOR PEER REVIEW 8 of 19 dominated precipitation was light rain, and the sum of incidence rates of moderate, heavy, and torrential rain was less than 36%, but the contribution rate was up to about 87%. Precipitation in Beibu Gulf during rainy season was dominated by weak precipitation events, but the main contribution was dependent on strong precipitation events, indicating the large possibility of flood disaster and aggravated situation of regional flood.  Figure 6 show the incidence rate and the contribution rate of different precipitation grades at 12 stations during rainy season. All of the stations met the characteristics that the precipitation incidence rate decreased with the increased precipitation grade, and the light rain was the dominated precipitation. The incidence rates of light rain at Dongxing, Qinzhou and Beihai stations (in the southern coast area) were extremely low, whereas the incidence rates of torrential rain were relatively high. The incidence rate of precipitation at other stations were similar, with the fluctuations around the average value. The letters along the y-axis denote precipitation grades. L is Light rain, M is Moderate rain, H is Heavy rain, and T is Torrential rain.

Spatial Differences of Incidence Rate and Contribution Rate of Different Precipitation Grades
The 12 stations showed that the contribution rates of light rain were the lowest (generally below 17%), according to the contribution rates of various precipitations grades (Figure 6b). The contribution rates of light precipitation ranged 7%-10% at Dongxing, Qinzhou, Beihai stations (in southern coastal region), being the lowest value in this region. The contribution rates of moderate rain, heavy rain, and torrential rain at Dongxing, Qinzhou, Beihai, Mashan, Bobai, and Hengxian stations showed an increasing tendency, especially at Dongxing, Qinzhou, Beihai stations, where all of the contribution rates of rainstorm were above 50%, being significantly higher than those at other stations (Figure 7). The region in the southern coast would be attacked by the strong convection  Figure 6 show the incidence rate and the contribution rate of different precipitation grades at 12 stations during rainy season. All of the stations met the characteristics that the precipitation incidence rate decreased with the increased precipitation grade, and the light rain was the dominated precipitation. The incidence rates of light rain at Dongxing, Qinzhou and Beihai stations (in the southern coast area) were extremely low, whereas the incidence rates of torrential rain were relatively high. The incidence rate of precipitation at other stations were similar, with the fluctuations around the average value. dominated precipitation was light rain, and the sum of incidence rates of moderate, heavy, and torrential rain was less than 36%, but the contribution rate was up to about 87%. Precipitation in Beibu Gulf during rainy season was dominated by weak precipitation events, but the main contribution was dependent on strong precipitation events, indicating the large possibility of flood disaster and aggravated situation of regional flood.  Figure 6 show the incidence rate and the contribution rate of different precipitation grades at 12 stations during rainy season. All of the stations met the characteristics that the precipitation incidence rate decreased with the increased precipitation grade, and the light rain was the dominated precipitation. The incidence rates of light rain at Dongxing, Qinzhou and Beihai stations (in the southern coast area) were extremely low, whereas the incidence rates of torrential rain were relatively high. The incidence rate of precipitation at other stations were similar, with the fluctuations around the average value. The letters along the y-axis denote precipitation grades. L is Light rain, M is Moderate rain, H is Heavy rain, and T is Torrential rain.

Spatial Differences of Incidence Rate and Contribution Rate of Different Precipitation Grades
The 12 stations showed that the contribution rates of light rain were the lowest (generally below 17%), according to the contribution rates of various precipitations grades (Figure 6b). The contribution rates of light precipitation ranged 7%-10% at Dongxing, Qinzhou, Beihai stations (in southern coastal region), being the lowest value in this region. The contribution rates of moderate rain, heavy rain, and torrential rain at Dongxing, Qinzhou, Beihai, Mashan, Bobai, and Hengxian stations showed an increasing tendency, especially at Dongxing, Qinzhou, Beihai stations, where all of the contribution rates of rainstorm were above 50%, being significantly higher than those at other stations (Figure 7). The region in the southern coast would be attacked by the strong convection weather systems, such as typhoon from the ocean, thus leading to frequent intense precipitation The letters along the y-axis denote precipitation grades. L is Light rain, M is Moderate rain, H is Heavy rain, and T is Torrential rain.
The 12 stations showed that the contribution rates of light rain were the lowest (generally below 17%), according to the contribution rates of various precipitations grades (Figure 6b). The contribution rates of light precipitation ranged 7%-10% at Dongxing, Qinzhou, Beihai stations (in southern coastal region), being the lowest value in this region. The contribution rates of moderate rain, heavy rain, and torrential rain at Dongxing, Qinzhou, Beihai, Mashan, Bobai, and Hengxian stations showed an increasing tendency, especially at Dongxing, Qinzhou, Beihai stations, where all of the contribution rates of rainstorm were above 50%, being significantly higher than those at other stations (Figure 7). The region in the southern coast would be attacked by the strong convection weather systems, such as typhoon from the ocean, thus leading to frequent intense precipitation [83,84]. At the other six stations, the contribution rates of moderate rain and heavy rain were similar, and they all showed the characteristics that the contribution rates of moderate or heavy rain slightly exceeded that of torrential rain.
Water 2020, 12, x FOR PEER REVIEW 9 of 19 [83,84]. At the other six stations, the contribution rates of moderate rain and heavy rain were similar, and they all showed the characteristics that the contribution rates of moderate or heavy rain slightly exceeded that of torrential rain.  Figure 8 shows the variation tendencies of precipitation incidence rate ( Figure 8a) and contribution rate (Figure 8b) of different durations in the rainy season at all stations. The variation tendency was evaluated using Mann-Kendall test and only the variation trends that passed the 95% level of significance test (MK > 1.96 or MK < −1.96, Figure 8) were included.

Variation Tendency of Incidence Rate and Contribution Rate of Different Precipitation Durations
For the durations of 8-9 d, all of the stations had the significant downward trend of precipitation incidence rate (Figure 8a). For the duration of 7 d, there were eight stations with the significant downward trend. For the durations of 10 d and 6 d, there were three stations (Nanning, Chongzuo, and Beihai) and two stations (Dongxing and Chongzuo) with the significant downward trend, respectively. For other durations, there was only one or no station with the significant change trend.
By contrast, the variation tendencies of precipitation contribution rate and incidence rate for ≥7 d (7-9 d and ≥10 d) durations were generally the same. For 6 d, there were two stations (Mashan and Hengxian) with the significant downward trend (Figure 8b). For other durations, there was only one or no station with the significant change trend.   Figure 8 shows the variation tendencies of precipitation incidence rate (Figure 8a) and contribution rate (Figure 8b) of different durations in the rainy season at all stations. The variation tendency was evaluated using Mann-Kendall test and only the variation trends that passed the 95% level of significance test (MK > 1.96 or MK < −1.96, Figure 8) were included.

Variation Tendency of Incidence Rate and Contribution Rate of Different Precipitation Durations
Water 2020, 12, x FOR PEER REVIEW 9 of 19 [83,84]. At the other six stations, the contribution rates of moderate rain and heavy rain were similar, and they all showed the characteristics that the contribution rates of moderate or heavy rain slightly exceeded that of torrential rain.  Figure 8 shows the variation tendencies of precipitation incidence rate ( Figure 8a) and contribution rate (Figure 8b) of different durations in the rainy season at all stations. The variation tendency was evaluated using Mann-Kendall test and only the variation trends that passed the 95% level of significance test (MK > 1.96 or MK < −1.96, Figure 8) were included.

Variation Tendency of Incidence Rate and Contribution Rate of Different Precipitation Durations
For the durations of 8-9 d, all of the stations had the significant downward trend of precipitation incidence rate (Figure 8a). For the duration of 7 d, there were eight stations with the significant downward trend. For the durations of 10 d and 6 d, there were three stations (Nanning, Chongzuo, and Beihai) and two stations (Dongxing and Chongzuo) with the significant downward trend, respectively. For other durations, there was only one or no station with the significant change trend.
By contrast, the variation tendencies of precipitation contribution rate and incidence rate for ≥7 d (7-9 d and ≥10 d) durations were generally the same. For 6 d, there were two stations (Mashan and Hengxian) with the significant downward trend (Figure 8b). For other durations, there was only one or no station with the significant change trend.  For the durations of 8-9 d, all of the stations had the significant downward trend of precipitation incidence rate (Figure 8a). For the duration of 7 d, there were eight stations with the significant downward trend. For the durations of 10 d and 6 d, there were three stations (Nanning, Chongzuo, and Beihai) and two stations (Dongxing and Chongzuo) with the significant downward trend, respectively. For other durations, there was only one or no station with the significant change trend.
By contrast, the variation tendencies of precipitation contribution rate and incidence rate for ≥7 d (7-9 d and ≥10 d) durations were generally the same. For 6 d, there were two stations (Mashan and Hengxian) with the significant downward trend (Figure 8b). For other durations, there was only one or no station with the significant change trend. Figure 9 shows the variation tendencies of precipitation incidence rate (Figure 9a) and contribution rate (Figure 9b) of different grades during rainy season at all stations. In terms of incidence rates of torrential rain and heavy rain, there were two stations (Dongxing and Qinzhou) and three stations (Hengxian, Bobai, and Yulin) showing the significant upward trend, respectively. For incidence rate of moderate rain and light rain, there was no station showing the significant variation tendency (e.g., failing to pass the 95% level of test, −1.96 < MK < 1.96, Figure 9).

Variation Tendency for Incidence Rate and Contribution Rate of Different Precipitation Grades
For the contribution rate of torrential rain, the Dongxing and Mashan stations showed the significant upward trend (Figure 9b). For the contribution rate of heavy rain, Yulin and Rongxian stations showed the significant upward trend. For the contribution rate of moderate rain, the Mashan and Qinzhou stations showed the significant downward trend. For the contribution of light rain, there was no station with significant variation tendency.
Water 2020, 12, x FOR PEER REVIEW 10 of 19 Figure 9 shows the variation tendencies of precipitation incidence rate ( Figure 9a) and contribution rate (Figure 9b) of different grades during rainy season at all stations. In terms of incidence rates of torrential rain and heavy rain, there were two stations (Dongxing and Qinzhou) and three stations (Hengxian, Bobai, and Yulin) showing the significant upward trend, respectively. For incidence rate of moderate rain and light rain, there was no station showing the significant variation tendency (e.g., failing to pass the 95% level of test, −1.96 < MK < 1.96, Figure 9).

Variation Tendency for Incidence Rate and Contribution Rate of Different Precipitation Grades
For the contribution rate of torrential rain, the Dongxing and Mashan stations showed the significant upward trend (Figure 9b). For the contribution rate of heavy rain, Yulin and Rongxian stations showed the significant upward trend. For the contribution rate of moderate rain, the Mashan and Qinzhou stations showed the significant downward trend. For the contribution of light rain, there was no station with significant variation tendency.

Discussion
The link between China and the ASEAN countries will be more frequent under China's the Belt and Road Initiative. The development of this area is promising and foreseeable due to the special geographical location of Beibu Gulf. The environmental research is necessary before large-scale infrastructure construction and economic development [85]. In this study, the response characteristics of precipitation to the environmental factors have been explored. Additionally, the structural variation in precipitation could effectively reflect the variation in water cycle. Thus, the spatiotemporal evolution characteristics of precipitation structure are helpful for analyzing the response of hydrological cycle to the changing environment, including global warming. This study has laid a foundation for regional water resources planning and management [86]. Additionally, the study can provide new information for predicting and guiding regional agricultural and industrial production, and disaster prevention and mitigation.

Spatiotemporal Variation in Precipitation
In this study, short-duration precipitation of 1-4 days dominated in the incidence rate and contribution rate of precipitation, and light rain dominated the incidence rate of precipitation of all precipitation grades. Despite the similarity with previous studies [53,54,87,88], there were some differences. In particular, the contribution rate of precipitation was quite different for the duration ≥ 10 d and torrential rain. For example, the contribution rate of precipitation for the duration ≥ 10 d in the upper reached the Yangtze River was only 10.6% [53], and that of torrential rain in most inland areas of Northwest China was less than 15% [89]. This might be due to the various conditions of water vapor supply in different regions, which was affected by the continuous actions of weather system conducive to precipitation [90,91].

Discussion
The link between China and the ASEAN countries will be more frequent under China's the Belt and Road Initiative. The development of this area is promising and foreseeable due to the special geographical location of Beibu Gulf. The environmental research is necessary before large-scale infrastructure construction and economic development [85]. In this study, the response characteristics of precipitation to the environmental factors have been explored. Additionally, the structural variation in precipitation could effectively reflect the variation in water cycle. Thus, the spatiotemporal evolution characteristics of precipitation structure are helpful for analyzing the response of hydrological cycle to the changing environment, including global warming. This study has laid a foundation for regional water resources planning and management [86]. Additionally, the study can provide new information for predicting and guiding regional agricultural and industrial production, and disaster prevention and mitigation.

Spatiotemporal Variation in Precipitation
In this study, short-duration precipitation of 1-4 days dominated in the incidence rate and contribution rate of precipitation, and light rain dominated the incidence rate of precipitation of all precipitation grades. Despite the similarity with previous studies [53,54,87,88], there were some differences. In particular, the contribution rate of precipitation was quite different for the duration ≥ 10 d and torrential rain. For example, the contribution rate of precipitation for the duration ≥ 10 d in the upper reached the Yangtze River was only 10.6% [53], and that of torrential rain in most inland areas of Northwest China was less than 15% [89]. This might be due to the various conditions of water vapor supply in different regions, which was affected by the continuous actions of weather system conducive to precipitation [90,91].
The variation tendencies of precipitation incidence and contribution rate in short (1-3 d) and medium (4-6 d) durations were more complicated, and the conclusions were various in different regions. However, the incidence and contribution rates of precipitation with long duration (≥7 d) were generally the same in different regions, both with significant downward trends [53,87,92]. In recent years, the incidence rate and contribution rate of light rain have decreased, while the incidence rate and contribution rate of torrential rain have increased in many areas in China, including Hebei, Gansu, Qinghai, Chongqing, Guizhou, and Guangdong [54,89,[93][94][95][96][97]. However, the variation trends of light rain were insignificant, while the incidence and contribution rates of torrential rain increased significantly only at two stations in our study area.

Implication of the Results
The precipitation contribution rate of durations 3 d was the highest in duration 1-9 d, followed by 4 d and 2 d, and the total contribution rate of duration 2-4 d reached up to 40%. Considering the large precipitation in the rainy season (1359.59 mm) and many flood and waterlogging disasters in the region, it is important to pay more attention to the disasters that are caused by 2-4 d precipitation.
In Dongxing station, located in the south coast, the annual average contribution rate of torrential rain was as high as 57.16%. Worse, the incidence rate and contribution rate of torrential rain are still rising. Therefore, it was crucial to focus on the disaster that is caused by torrential rain in the region.
In recent years, under the background of the Belt and Road Initiative, the trade between China with the European Union, Africa, Central Asia, and ASEAN has increased gradually. In particular, Beibu Gulf is an important region for the trade development between China and ASEAN countries, and it is predicted for the population growth and rapid development of large infrastructure projects (large seaports, transportation lines, and others) in the coming decade. These changes will result in some environmental problems, for example, the increase in carbon emission, extensive surface hardening, accelerated surface runoff, urban waterlogging, and floods [98][99][100]. Proper measures are urgently needed in order to mitigate the damages [101,102]. The losses could be minimized by science-base planning, such as improving the design standards for drainage of built-up areas and enhancing flood control of rivers, as well as increasing vegetation coverage [103].

Limitation and Future Research
Similar to many studies, there are some limitation of this study. Only data from 12 meteorological stations in the area were collected and analyzed in the current study due to the relative lower economic development in the Beibu Gulf [104]. In recent years, more meteorological observation stations have been built in the region, providing more data with high-quality. With the development of remote sensing technology, observation data with high-resolution and wide observation range could be acquired from satellite, such as Tropical Rainfall Measuring Mission (TRMM), Climate Hazards group InfraRed Precipitation with Station data (CHIRPS), and Global Precipitation Mission (GPM). In addition, studies have shown that these data may be problematic in many areas and correction is required before application [22][23][24]105,106]. In the future, studies using the TRMM, CHIRPS and GPM data can expand our understanding of hydrological cycle in the area. However, in situ observations are generally more accurate and still indispensable. This study was performed from the perspective of statistical analysis, and the further study on the atmospheric circulation and other dynamics will increase knowledge regarding the impact mechanism of the hydrological cycle in South China.

Conclusions
This study analyzed the daily precipitation data of the Beibu Gulf in the period of 1961-2016, and explored the spatiotemporal variation characteristics of the precipitation structure in rainy season (from April to September), from the aspects of precipitation duration and grade, by introducing the indices of the precipitation incidence rate and the contribution rate. The following conclusions have been drawn from this study: (1). The precipitation during rainy season accounted for more than 80% of the total annual precipitation. With the increase in precipitation durations, the precipitation incidence rates decreased exponentially and the precipitation contribution rates increased first and decreased later. For the duration of ≥ 10 d, the precipitation contribution rates showed a marked increase. The spatial differences for all the durations were almost the same for the precipitation incidence rate and the contribution rate of durations 3-9 d. However, the differences for durations 1-2 d and ≥ 10 d were quite distinct.
(2). With the increase in precipitation grades, the precipitation incidence rate decreased and the precipitation contribution rate increased gradually. For different precipitation grades, there were spatial differences of the precipitation incidence rates and the contribution rates for light rain and torrential rain at Beihai, Qinzhou, Dongxing stations (in the southern region). In these three stations, low precipitation incidence rates and contribution rates were observed for light rain, while they were larger for torrential rain. Especially, the contribution rates of torrential rain were all over 50% at these three stations.
(3). The precipitation incidence rate and contribution rate of 7-9 d durations at most (7 d) and all (8-9 d) stations showed a significant downward trend, while the variation trends of other durations at most stations were insignificant. In terms of grade precipitation, the variation trends of all grade precipitation at most stations were insignificant.
(1). The precipitation during rainy season accounted for more than 80% of the total annual precipitation. With the increase in precipitation durations, the precipitation incidence rates decreased exponentially and the precipitation contribution rates increased first and decreased later. For the duration of ≥ 10 d, the precipitation contribution rates showed a marked increase. The spatial differences for all the durations were almost the same for the precipitation incidence rate and the contribution rate of durations 3-9 d. However, the differences for durations 1-2 d and ≥ 10 d were quite distinct.
(2). With the increase in precipitation grades, the precipitation incidence rate decreased and the precipitation contribution rate increased gradually. For different precipitation grades, there were spatial differences of the precipitation incidence rates and the contribution rates for light rain and torrential rain at Beihai, Qinzhou, Dongxing stations (in the southern region). In these three stations, low precipitation incidence rates and contribution rates were observed for light rain, while they were larger for torrential rain. Especially, the contribution rates of torrential rain were all over 50% at these three stations.
(3). The precipitation incidence rate and contribution rate of 7-9 d durations at most (7 d) and all (8-9 d) stations showed a significant downward trend, while the variation trends of other durations at most stations were insignificant. In terms of grade precipitation, the variation trends of all grade precipitation at most stations were insignificant.
Water 2020, 12, x FOR PEER REVIEW 14 of 19