Evolutionary Characteristics of Daytime and Nocturnal Precipitation Heterogeneity in Gansu Province, Northwest China

Abstract

The diurnal variation in precipitation and its evolution are important foundations for understanding the regional impact of climate change and the parameterization of the model. Based on the daily precipitation data set of 23 national meteorological stations during 1970–2019, the spatial and temporal distribution characteristics of precipitation concentration degree (PCD) and precipitation concentration period (PCP) in Gansu province were evaluated on daytime and nocturnal scales. The results show the following: (1) Annual precipitation ranges from 69.1 ± 24.7 mm to 578.3 ± 96.6 mm, mainly (54.1 ± 2.6%) occurring at night, and the spatial distribution of the nocturnal precipitation rate is positively (r = 0.53, p < 0.01) correlated with annual precipitation; the wetting trend (12.7 mm/10 a, p < 0.01) in the past 50 years is obvious, and is mainly dominated by the frequency of precipitation (r = 0.58, p < 0.001), with both performing better during the day. (2) Most PCD is located between 0.55 and 0.75, showing a basic distribution pattern for daytime greater than nocturnal, higher values, and stronger interannual fluctuations in arid areas; the significant decreasing trend (p < 0.05) of PCD is very clear and highly consistent, especially in the high-altitude area, and the increase in precipitation in the dry season and the improvement in precipitation uniformity in the wet season play a key role. (3) PCP often fluctuates slightly around the 39th–41st pentad, but the general rule that daytime values are smaller than night values and the interannual variability is larger in arid areas also requires special attention; PCP has shown a relatively obvious advance trend in a few regions, but this is because the prominent and complex changes in the monthly precipitation distribution pattern have not been fully reflected. Along with continuous humidification, the decrease in PCD and the advance of PCP are likely to be the priority direction of precipitation evolution in the arid region of Northwest China, especially during the day. These findings provide a new perspective for understanding regional climate change.

1. Introduction

In March and July 2023, the IPCC Sixth Assessment Synthesis Report: Climate Change 2023 and the Blue Book on Climate Change in China (2023) were released, respectively. Whether in 2022 or 2011–2020, the global surface temperature (GST) was 1.1 °C higher than in 1850–1900, with higher increases on land (1.6 °C) than on the ocean (0.9 °C) [1,2]. Meanwhile, GST has increased faster since 1970 than in any other 50-year period over at least the last 2000 years [1]. As a sensitive area and a significant area of impact of global climate change, the level of warming in China was higher than the global average for the same period, reaching 0.16 °C/10 a (per decade) between 1901 and 2022 [2]. With global warming, not only is the water cycle undergoing many changes [3,4,5,6], but also droughts, hot extremes [7,8,9], heavy precipitation, floods [10,11,12], and various secondary disaster events [13,14,15] are occurring frequently around the world. For example, from the night of 29 July to the early morning of 2 August 2023, Beijing suffered its worst rainstorm in nearly 140 years (Beijing Meteorological Bureau); in July 2018, heavy rainstorms in western Japan killed more than 220 people, and so on.

It is clear that, in addition to a more intuitive response, the precipitation’s subsequent influence is more extensive and profound [16,17,18,19]. Therefore, the current spatial and temporal distribution pattern of precipitation should be necessarily different from that of the past. On a global scale, the average annual precipitation over land has increased, especially in the mid latitudes of the northern hemisphere [3,20]. In the last four decades, a faster increase in global land precipitation was observed, with large interannual variability and regional heterogeneity [20,21]. There is a general increase in the probability of daily rainfall exceeding 50 mm, and it is mainly caused by the general enhancement of rain intensity [22], particularly on sub-daily timescales [23,24]. China is one of the representative regions with the most obvious increase. During the period 1961–2022, the rate of increase in the national average annual precipitation was as high as 0.8%/10 a, and the regional and seasonal differences are striking [2]. Average annual precipitation in the Qinghai–Tibet region increased significantly (9.4 mm/10 a), while that in Southwest China decreased (−9.6 mm/10 a) [2,25]. Precipitation increases in autumn and winter, but decreases in summer [26], and snowfall on the Tibetan Plateau shows a similar change [27]. Regarding precipitation patterns, the frequency and intensity of rainstorm events or extreme daily rainfall have varied significantly and involve the vast majority of provinces [28,29,30]. In addition, even if annual precipitation totals changed little, the trend of increasing precipitation intensity and decreasing precipitation days may be obvious [31]. In Eastern China, the increasing trend for heavy rainfall is accompanied by the decreasing frequency of low and moderate rainfall [32]. On the other hand, the evolutionary characteristics of annual distribution and the diurnal variation of precipitation have also attracted extensive attention, but the existing understanding is insufficient. Based on a series of indicators represented by PCD, PCP, PCI (Precipitation Concentration Index), and CI (Concentration Index), the precipitation concentration and its changes in Northeast China [33], Inner Mongolia [34], the Huaihe River [35], Liaohe River [36], Weihe River [37] and the Three Gorges Reservoir area [38] have been effectively revealed. The above results confirm that the spatial distribution of precipitation concentration in China has complex meridional and zonal variability, and the correlation index generally shows a downward trend. It is well known that precipitation has a distinct diurnal variation process [39,40,41] due to the alternation of day and night, but a series of scientific issues related to it, such as whether its response to climate change is consistent in different periods and which period is more critical, have yet to be determined. Given this, we selected Gansu province, which has diverse climate types and is located deep in the interior of Northwest China, as the research object, to bridge the cognitive gap of precipitation concentration and its evolution on the diurnal scale in such regions, based on the data set of daily values of 23 national meteorological stations from 1970 to 2019. Benefiting from the novelty of the diurnal and nocturnal scales, this study attempts to provide substantial data support and a theoretical basis for the mechanism of precipitation change and the prevention of hydrological disasters.

2. Materials and Methods

2.1. Study Area

Gansu province (92°13′–108°46′ E and 32°11′–42°57′ N, Figure 1), located at the intersection of the Qinghai-Tibet Plateau, the Mongolian Plateau and the Loess Plateau, is an important part of the inland region in northwest China. Although only a medium-sized province in terms of total area (4.26 × 10⁵ km²), it spans more than 1600 km from southeast to northwest. Therefore, the local climate types are particularly complex and diverse, covering humid (southwest), semi-humid, semiarid, and arid (northwest), and the latter two are dominant, accounting for about 75%. In addition to the complex and changeable landform, the terrain is also strongly undulating, with an altitude range of 587–5692 m above sea level. In general, the average annual temperature ranges from 0 to 15 °C, and the annual precipitation is between 40 and 750 mm (http://www.gansu.gov.cn (accessed on 10 September 2023)).

2.2. Data Acquisition

The daily data set of basic meteorological elements of the China National Surface Weather Station is obtained from the Resource and Environment Science and Data Center, Chinese Academy of Sciences (http://www.resdc.cn/data.aspx?DATAID=230 (accessed on 10 September 2023)). Taking into account practical conditions such as starting time and missing data, as well as basic requirements for time series consistency, 23 stations (Figure 1) with complete data and strict quality control were selected for the period 1970–2019. They are distributed between 1079 and 3471 m above sea level, and there are 2 (8.7%), 3 (13.0%), 8 (34.8%), and 10 (43.5%) stations above 3000 m, 2000–3000 m, 1500–2000 m, and below 1500 m, respectively. The daily precipitation data consists of nocturnal precipitation (2000–0800, Beijing Time), daytime precipitation (0800–2000), and daily precipitation (2000–2000), all of which use 0.1 mm as the lower limit to judge whether precipitation occurs (Chinese National Standard “Grade of Precipitation”, GB/T 28592-2012). The related results of precipitation concentration in the above three research periods are calculated by pentad precipitation (the first 5 pentads of each month include 5 days, and the remaining days are classified as the last pentad, so there are a total of 72 pentads in the whole year). For the statistical results of precipitation elements, the data of a single station are obtained first, and then the average value of 23 stations is obtained for the whole province.

2.3. Methods

The coefficient of variation (CV), which can be calculated from Equation (1), is a relative quantity that represents the size of the standard deviation ($\mathsf{\sigma }$) relative to the mean ($\overline{X}$) and can indicate the degree of dispersion of a set of variables. A higher CV value indicates a higher degree of dispersion, which means that the data are more volatile.

$$CV=\raisebox{1ex}{$\mathsf{\sigma }$}\!\left/ \!\raisebox{-1ex}{$\overline{X}$}\right.$$

(1)

The precipitation anomaly in percentages (PA), which is an extension of the concept of anomaly, is the percentage of the difference between the precipitation for a certain study period and the annual mean value accounting for the annual mean precipitation in the same period. The calculation is shown in Equation (2).

$$PA=\frac{P-\overline{P}}{\overline{P}}\times 100\%$$

(2)

where PA is the precipitation anomaly in percentages. P represents the precipitation of a certain study period, and $\overline{P}$ represents the annual mean precipitation in the same period. According to the Chinese National Standard “Grades of Meteorological Drought” implemented since 2018, this study classifies PA as moderate drought (−40% < PA ≤ −30%), slight drought (−30% < PA ≤ −15%), normal (−15% < PA < 15%), slight moist (15% ≤ PA < 30%), moderate moist (30% ≤ PA < 40%) and other levels.

The PCD and the PCP are two standard statistical parameters proposed by Zhang and Qian [42] based on vector analysis, which can be defined to describe the pattern of the temporal distribution of annual precipitation. PCD can reflect the level of precipitation concentration in one year. The PCP, corresponding to the azimuth of the composite vector, is the center of gravity for when the annual precipitation is regarded as a whole, and can characterize the theoretical period when the maximum probability of peak precipitation occurs. The PCD and PCP can be defined as follows:

$${PCD}_i=\frac{\sqrt{R_{xi}^2+R_{yi}^2}}{R_i}$$

(3)

$${PCP}_i=\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{t}\mathrm{a}\mathrm{n}\left(\frac{R_{xi}}{R_{yi}}\right)$$

(4)

$$R_{xi}={\sum }_{j=1}^N{r}_{ij}\times \sin {\theta }_j$$

(5)

$$R_{yi}={\sum }_{j=1}^N{r}_{ij}\times \cos {\theta }_j$$

(6)

where i is the year (i = 1970, 1971, …, 2019), and R_i stands for annual precipitation in the ith year. j is the pentad (j = 1, 2, …, 72). θ_j represents the corresponding azimuth angle of the jth pentad (j = 1, 0° ≤ θ_j < 5°; j = 2, 5° ≤ θ_j < 10°; … ; j = 72, 355° ≤ θ_j < 360°. The center point of each interval is used for the calculation), while the year can be seen as 360°. r_ij denotes the precipitation of the jth pentad in the ith year. PCD_i is the degree of precipitation concentration in the ith year, and the range of values is from 0 to 1. If annual precipitation occurs in a specific pentad, the maximum value (1) of PCD_i can be obtained. If annual precipitation is evenly distributed, the figure can reach its minimum value (0). PCP_i represents the theoretical pentad with the highest probability of peak precipitation in the ith year. More detailed calculation methods and theories are enriched by References [36,38,42].

3. Results and Discussion

3.1. Basic Characteristics of Precipitation

According to the regional average observation results of 23 meteorological stations, the annual precipitation of Gansu province from 1970 to 2019 was about 343.4 ± 47.0 mm (Figure 2a). The minimum value and the maximum value appeared in 1972 (238.7 mm) and 2018 (458.1 mm), respectively, and the latter was 0.92 times higher than the former (Figure 2a). The overall interannual variability (CV = 0.11) was small, and PA was only in moderate drought, slight drought, slight moist and moderate moist grades for 1, 5, 5 and 1 year, respectively (Figure 2b). The average annual precipitation days (that is, precipitation frequency) of the entire province fluctuated between 53.4 d/year (1972) and 100.7 d/year (1988), with the main body being 82.0±10.3 d/year (Figure 2c). In the last 50 years, the evolutionary trend of precipitation intensity (4.2 ± 0.5 mm/d, the ratio of the amount of precipitation to the number of precipitation days) was not clear, but annual precipitation (r =0.39, p < 0.01; Figure 2a) and its precipitation frequency (r = 0.51, p < 0.001; Figure 2c) both showed an extremely significant increasing trend, with the rate of climate change tendency reaching 12.7 mm/10 a (Figure 2a) and 3.6 d/10 a (Figure 2c), respectively. This has been most prominent especially in the last decade or two (Figure 2a,c). Furthermore, annual precipitation was mainly dominated by precipitation frequency (r = 0.58, p < 0.01; Figure 2d), followed by precipitation intensity (r = 0.48, p < 0.01; Figure 2d), which is also supported by the results of partial correlation analysis. Therefore, annual precipitation in Gansu province has shown general characteristics of small interannual variability and a strong humidification trend over the past half-century, and the frequency of precipitation plays the most critical role in this process. These findings are consistent with existing research results [24]. However, the ability of the daily data set to reflect precipitation intensity and precipitation duration is very limited. Therefore, only with the popularization of hourly data with higher precision can the real evolutionary trend and the role of the above two factors be more scientifically revealed.

On the sub-daily scale, the precipitation frequency (Figure 2c) is generally slightly higher during the daytime (55.9±6.6 d), about 1.05±0.09 times that at nighttime (53.5 ± 7.0 d). On the other hand, the proportion of nocturnal precipitation can usually reach 54.1 ± 2.6% (Figure 2a), and the interannual difference is much weaker (CV = 0.05). Even in the rare anomalous years (1996, 2015), the value is not lower than 47.3%. Under the influence of these years, the intensity of precipitation (3.5 ± 0.5 mm/12 h) at night was higher than that during the daytime (2.8 ± 0.3 mm/12 h). In terms of temporal distribution, both daytime (nocturnal) precipitation (Figure 2a) and daytime (nocturnal) precipitation frequency (Figure 2c) are generally consistent with the corresponding results mentioned above, with only the extreme value year, ranking sequence, and value size being slightly different. It should be noted that the PA (Figure 2b) of nocturnal precipitation and daytime precipitation is not in completely the same direction, and some years they are even opposite. For example, nocturnal precipitation (−11.4%) in 1996 was relatively low, but daytime precipitation (9.5%) was relatively high. Furthermore, the number of years in which daytime precipitation reached moderate moist increased by two times (Figure 2b). The evolutionary characteristics of significant increase also apply to nocturnal precipitation, daytime precipitation (Figure 2a), nocturnal precipitation frequency and daytime precipitation frequency (Figure 2c). The climatic tendency rates were 5.4 mm/10 a (r = 0.29, p < 0.05), 7.3 mm/10 a (r = 0.47, p < 0.001; Figure 2a), 2.0 d/10 a (r = 0.41, p < 0.001) and 2.5 d/10 a (r = 0.56, p < 0.001; Figure 2c), respectively. Although the percentage of nocturnal precipitation has shown a decreasing trend (r = −0.21, p > 0.05), it has not yet reached a significant level. It can be seen that daytime precipitation, although usually slightly lower than nocturnal precipitation, has played a greater contribution in the evolutionary process of increasingly abundant precipitation in Gansu province, benefiting from the substantial increase in the frequency of daytime precipitation. In terms of mechanism, this may be due to the greater temperature range during the day, especially in the afternoon, which is very conducive to the formation of local convective rain [39,40,41].

Among the regions, not only is the difference in the precipitation elements larger, but the variation characteristics are also more complex. First, the annual precipitation extremes were 69.1 ± 24.7 mm and 578.3 ± 96.6 mm, located in the northernmost Mazongshan and the highest elevation, Maqu, respectively. On the whole, Gansu province has a precipitation distribution pattern of abundant precipitation in the southeast and scarce precipitation in the northwest, and the proportion of annual precipitation (AP) belonging to arid (AP < 200 mm), semiarid (200 mm ≤ AP < 400 mm), and semi-humid (400 mm ≤ AP < 800 mm) areas was roughly 30: 22: 48 (

Table 1. Correlation coefficients of precipitation amount and precipitation frequency evolutionary trends during 1970–2019. From the top down, the significance index is less and less, and the evolutionary trend is gradually weakened. Results that did not pass the significance test are not listed. * p < 0.05, ** p < 0.01, *** p < 0.001.

Station	Elevation /m	Average Annual Precipitation/mm	Precipitation Amount			Precipitation Frequency
			Nocturnal	Daytime	Daily	Nocturnal	Daytime	Daily
Wushaoling	3045	397.5	0.50 ***	0.54 ***	0.59 ***	0.59 ***	0.56 ***	0.61 ***
Hezuo	2910	525.2	0.42 **	0.41 **	0.50 ***	0.42 **	0.57 ***	0.50 ***
Yongchang	1977	208.4	0.39 **	0.45 ***	0.50 ***	0.43 **	0.46 ***	0.46 ***
Maqu	3471	578.3	0.34 *	0.53 ***	0.49 ***	0.53 ***	0.62 ***	0.56 ***
Jiuquan	1477	91.4	0.29 *	0.31 *	0.35 *	0.43 **	0.46 ***	0.47 ***
Wuwei	1532	168.5	0.39 **		0.33 *	0.39 **	0.32 *	0.36 **
Huajialing	2451	460.7		0.41 **		0.33 *	0.53 ***	0.52 ***
Gaotai	1332	111.2		0.27 *		0.35 *	0.54 ***	0.51 ***
Minxian	2315	561.5		0.30 *		0.32 *	0.33 *	0.44 **
Zhangye	1461	129.8				0.38 **	0.41 **	0.43 **
Linxia	1917	500.1				0.32 *	0.45 ***	0.38 **
Yuzhong	1874	373.9				0.33 *	0.38 **	0.39 **
Minqin	1368	116.9				0.29 *	0.39 **	0.35 *
Xifeng	1421	538.5		0.28 *			0.41 **	0.34 *
Huanxian	1256	419.4		0.37 **	0.30 *		0.30 *
Kongtong	1347	494.4		0.31 *			0.30 *	0.28 *
Gaolan	1669	250.9					0.36 **	0.39 **
Lintao	1894	503.9					0.34 *	0.29 *
Jingtai	1631	189.4	0.33 *
Maiji	1085	512.4					0.28 *
Mazongshan	1770	69.1
Jingyuan	1398	227.4
Wudu	1079	468.3

). Second, the average annual precipitation frequency ranged from 33.3 ± 7.7 d/year to 135.9 ± 26.6 d/year, and the relationship with the corresponding annual precipitation can be accurately characterized by the linear equation (r = 0.92, p < 0.01) or the power function (r = 0.96, p < 0.01). Similarly, the spatial pattern of precipitation intensity (2.1 ± 0.7 mm/d~5.8 ± 1.1 mm/d) was also strengthened linearly with increasing annual precipitation (r = 0.87, p < 0.001). Third, the variation amplitude of the nocturnal precipitation rate expanded sharply, with the minimum value dropping to 40.2% and the maximum value as high as 62.1%. At the same time, the index showed a distribution law of the more abundant the annual precipitation, the greater the value (r = 0.53, p < 0.01). Finally, the evolutionary results of precipitation amount and precipitation frequency in the three statistical periods (nocturnal, daytime, and daily) jointly revealed that at least one of them had reached a significant increase level (p < 0.05) in most areas (87%), and only three stations (13%) did not show a clear trend (Table 1). Among these, nearly 70% of the weather stations with three or more indexes increased significantly, especially Wushaoling, Hezuo, Yongchang, and Maqu (all relatively high elevations), where almost all parameters reached extremely significant increases (p < 0.01), and are the most representative (Table 1). Overall, the evolutionary trend of the precipitation frequency is much ahead of the amount of precipitation, and both are more powerful during the day.

Therefore, against the climate background where precipitation elements such as annual precipitation (Table 1), precipitation frequency, precipitation intensity, and nocturnal precipitation percentage are generally decreasing from southeast to northwest, the basic trend of increasing humidification in Gansu province is very clear (Figure 2a), and it is mainly caused by the rise in precipitation frequency (especially in the daytime) and the more prominent performance in high-altitude areas (Figure 2c,d, Table 1). Why do these high-altitude zones generally have a more prominent performance? This must be the result of the interaction of faster warming rates, larger daily temperature ranges, more abundant water vapor sources, and other factors [12,18,22]. For example, the special topography is conducive to blocking the increasing abundance of external water vapor in the atmospheric circulation, and favorable hydrological conditions can make full use of the local internal circulation of water vapor. Once sufficient water vapor reaches the plummeting temperature, it is natural for water vapor to condense into precipitation, especially in the early morning and afternoon [39,40,41]. With global warming, precipitation events should form more easily at higher elevations, and the increase in precipitation amounts is also likely to be significantly ahead of that at lower elevations. There is no doubt that these preliminary inferences and mechanisms of action still need to be revealed and confirmed by further research, because climate change mainly affects large-scale weather systems, and local-scale factors may sometimes balance or overcome the effects.

3.2. Spatiotemporal Variation of PCD

As shown in Figure 3, the annual mean value of PCD is usually the highest on the daytime scale (0.59~0.75), followed by the daily scale (0.56~0.71), and the lowest on the nocturnal scale (0.53~0.69). To accurately describe the difference between the values, PCD is divided into the following five grades by the method of equal division: highly uniform (0 ≤ PCD ≤ 0.2), mildly uniform (0.2 < PCD ≤ 0.4), medium (0.4 < PCD < 0.6), mildly concentrated (0.6 ≤ PCD < 0.8) and highly concentrated (0.8 ≤ PCD ≤ 1.0). It is obvious that most of these are at the level of mildly concentrated, and only some of the results of Xifeng, Maiji, Wudu and other stations on the southeastern edge are medium (Figure 3a). However, it should not be ignored that the interannual volatility of PCD is greatly enhanced (Figure 3a), with CV generally exceeding 0.15 and even approaching 0.30 (on the nocturnal scale or in arid areas). This implies that although PCD in Gansu province is usually weak, it may still reach a highly concentrated level in a few years, to which attention especially needs to be paid in those regions with relatively poor precipitation. As is known, the single peak precipitation distribution which is centered in summer is the basic climatic characteristic of Gansu province, so most of the regions are expected to be at a mildly concentrated or medium level [43]. In the extremely arid region of the northwest, the occasional highly concentrated PCD must be due to the scarcity of annual precipitation, almost all of which occurs during the rainy season.

From the perspective of spatial variation, it is easy to recognize that PCD increases gradually with decreasing annual precipitation (Figure 3a,b). Among these values, the valley values (less than 0.6) are usually distributed in the southeastern fringe area represented by Xifeng, Maiji, and Wudu. With the continuous advance to the northwest interior, the PCD in Wushaoling, Yongchang, Minqin, and other stations increased significantly, until the peak of about 0.75 appeared at Mazongshan (Figure 3a). Meanwhile, there was a significant negative correlation (p < 0.01) between PCD and its corresponding precipitation, and it was applicable on the daytime (r = −0.53), nocturnal (r = −0.61) and daily (r = −0.64) scales (Figure 3b). Regarding the evolutionary characteristics, not only is the trend clearer (except for Wudu and Maiji), but the consistency is also very strong (

Table 2. PCD evolution results at 10 representative weather stations from 1970 to 2019. ** p < 0.01, *** p < 0.001.

Station	Elevation/m	Nocturnal	Daytime	Daily
Wushaoling	3045	y = −0.0056x + 11.925 R2 = 0.45 ***	y = −0.0038x + 8.312 R2 = 0.35 ***	y = −0.0046x + 9.786 R2 = 0.43 ***
Huajialing	2451	y = −0.0045x + 9.568 R2 = 0.34 ***	y = −0.0043x + 9.199 R2 = 0.28 ***	y = −0.0042x + 8.954 R2 = 0.35 ***
Hezuo	2910	y = −0.0045x + 9.605 R2 = 0.30 ***	y = −0.0027x + 6.031 R2 = 0.29 ***	y = −0.0038x + 8.249 R2 = 0.35 ***
Maqu	3471	y = −0.0034x + 7.387 R2 = 0.29 ***	y = −0.0033x + 7.183 R2 = 0.32 ***	y = −0.0033x + 7.227 R2 = 0.34 ***
Gaotai	1332	y = −0.0053x + 11.192 R2 = 0.20 ***	y = −0.0048x + 10.186 R2 = 0.22 ***	y = −0.0052x + 10.932 R2 = 0.31 ***
Minxian	2315	y = −0.0023x + 5.170 R2 = 0.15 **	y = −0.0023x + 5.239 R2 = 0.28 ***	y = −0.0022x + 4.989 R2 = 0.24 ***
Yongchang	1977	y = −0.0032x + 7.115 R2 = 0.18 **	y = −0.0034x + 7.449 R2 = 0.22 ***	y = −0.0032x + 7.117 R2 = 0.22 ***
Mazongshan	1770	y = −0.0053x + 11.200 R2 = 0.17 **	y = −0.0042x + 9.092 R2 = 0.23 ***	y = −0.0045x + 9.703 R2 = 0.25 ***
Jiuquan	1477	y = −0.0065x + 13.658 R2 = 0.28 ***	y = −0.0040x + 8.667 R2 = 0.19 **	y = −0.0050x + 10.656 R2 = 0.25 ***
Zhangye	1461	y = −0.0029x + 6.480 R2 = 0.14 **	y = −0.0034x + 7.488 R2 = 0.23 ***	y = −0.0033x + 7.163 R2 = 0.22 ***

). In the past 50 years, nearly 85% of PCD decreased significantly (p < 0.05), and more than 60% of PCD reached the extremely significant (p < 0.01) level, regardless of the daily or sub-daily scale (Table 2). This change was most obvious (p < 0.001) in high-altitude areas such as Maqu, Hezuo, and Huajialing (Table 2). Taking Wushaoling as an example, the decreasing PCD rates on nocturnal, daytime, and daily scales were 0.06/10 a, 0.04/10 a, and 0.05/10 a, respectively (Table 2). Combined with the research results of obvious humidification in Section 3.1, it is not difficult to find that the representative regions of the two overlap greatly, indicating that these high-altitude regions have the most sensitive response to climate change and often play an indicative and leading role. Furthermore, the decreasing trend on the daytime scale was generally stronger than that on the nocturnal scale (Table 2). For China, the trend of decreasing precipitation concentration is widespread [44,45]. Consequently, although the evolution of PCD in Gansu province is also more prominent in the high-altitude region, its spatial and temporal variation is almost opposite to that of annual precipitation, with an overall pattern of northwest higher than southeast and decreasing with humidification.

To reveal the internal mechanism of PCD reduction, the four most representative stations (Table 2) which have similar natural conditions, such as altitude, were selected to further analyze the annual distribution characteristics of precipitation. As can be seen from Figure 4a, Gansu province has distinct dry and wet seasons and the distribution of monthly precipitation within the year is unimodal. Among these, the dry season (October to April) lasts for more than half a year, but the total precipitation is less than 20% of the annual precipitation, and most months usually account for only about 1% (Figure 4a). For the wet season (May to September), the percentage of monthly precipitation generally does not fall below 11%, while the maximum monthly precipitation occurs mainly in July, followed by August, accounting for about 21% of the annual precipitation, and sometimes even exceeding 26% (Figure 4a). Through the decadal comparison of the last 50 years, it can be found that the main changes in the annual distribution of precipitation are as follows: On the one hand, precipitation in the dry season has increased dramatically, with the proportion jumping from about 5% in the 1970s to 19% in the 2010s (Figure 4a). On the other hand, the unevenness of precipitation in the wet season tended to weaken, with the proportion of precipitation in July and August decreasing from more than 25% to less than 20%, and being basically stable or slightly increased in other months (Figure 4a). Similar results are found on the nocturnal scale (Figure 4b) and the daytime scale (Figure 4c), but the maximum monthly precipitation is scattered between June and September for the former and is basically stable in July for the latter. All in all, the increase at both ends combined with the flattening of the middle part resulted in a significant reduction in the CV (from 1.3 in the early stage to 0.9 in the late stage) of the monthly precipitation percentage. The increase in precipitation in the dry season and the improvement in precipitation uniformity in the wet season are the key factors for the general decrease in PCD in Gansu province.

3.3. PCP Distribution Pattern

As can be seen in Figure 5a, the annual average values of PCP on the three statistical scales fluctuated slightly between 193° and 205° (39th–41st pentad), indicating that mid–late July is the most important precipitation period in Gansu province. On the other hand, the nocturnal PCP (195–207°, 40th–42nd pentad) is generally slightly late, while the daytime PCP (191–204 °, 39th–41st pentad) is slightly earlier (Figure 5a). Meanwhile, the interannual variation in PCP is generally weak, with most CV values less than 0.10, and there is a significant (p < 0.001) negative correlation with the corresponding precipitation on the nocturnal (r = −0.62), daytime (r = −0.83) and daily (r = −0.72) scales. As a result, PCP in arid areas tends to show a wider range of variations, with an advance to mid-June or a delay to mid-August being normal (Table 1, Figure 5a). In terms of evolutionary characteristics, there was almost no clear trend in PCP, and only a few stations such as Huajialing showed a relatively obvious advance trend on the daytime (r = −0.28, p < 0.05), daily (r = −0.25, p < 0.10), and nocturnal (r = −0.18, p > 0.10) scales (Figure 5b). It can be seen that mid-to-late July is usually the center of gravity of annual precipitation in Gansu province, but the greater variation in PCP in arid areas and the faintly visible advance trend [44] also deserve attention.

As mentioned in Section 3.2, the decrease in PCD is caused by the increase in precipitation in the dry season and the improvement in uniformity of precipitation in the wet season. So what are the implications of these shifts in PCP? To establish this, the percentage change in monthly precipitation in Huajialing Station was analyzed for each month in turn. First, the wetting trend from November to February is usually the most significant, but the perennial disadvantage of a monthly precipitation percentage of less than 2% severely weakens its role in the process of evolution of the annual distribution of precipitation (

Table 3. Distribution and evolution of the monthly precipitation percentage at Huajialing Weather Station from 1970 to 2019. * p < 0.05., ** p < 0.01., *** p < 0.001.

Month	Mean Percentage	Nocturnal	Daytime	Daily
January	0.9%	y = 0.0002x − 0.352 R2 = 0.090 *	y = 0.0003x − 0.505 R2 = 0.155 **	y = 0.0002x − 0.417 R2 = 0.147 **
February	1.2%	y = 0.0004x − 0.716 R2 = 0.245 ***	y = 0.0005x − 0.899 R2 = 0.159 **	y = 0.0004x − 0.760 R2 = 0.282 ***
March	2.6%	y = 0.0005x − 1.041 R2 = 0.117 *	y = 0.0007x − 1.282 R2 = 0.184 **	y = 0.0006x − 1.110 R2 = 0.164 **
April	5.8%	y = 0.0008x − 1.548 R2 = 0.055	y = 0.0008x − 1.508 R2 = 0.080 *	y = 0.0007x − 1.422 R2 = 0.075 *
May	11.6%	y = 0.0007x − 1.336 R2 = 0.020	y = 0.0002x − 0.190 R2 = 0.001	y = 0.0005x − 0.879 R2 = 0.013
June	14.5%	y = −0.0003x + 0.755 R2 = 0.004	y = −0.00001x + 0.170 R2 = 0.000	y = −0.0001x + 0.350 R2 = 0.001
July	20.2%	y = −0.0002x + 0.529 R2 = 0.001	y = −0.0007x + 1.542 R2 = 0.012	y = −0.0004x + 1.065 R2 = 0.006
August	19.8%	y = −0.0022x + 4.530 R2 = 0.079 *	y = −0.0020x + 4.137 R2 = 0.076 *	y = −0.0020x + 4.156 R2 = 0.102 *
September	14.4%	y = −0.0013x + 2.656 R2 = 0.051	y = −0.0006x + 1.408 R2 = 0.017	y = −0.0010x + 2.165 R2 = 0.053
October	7.0%	y = 0.0007x − 1.378 R2 = 0.037	y = 0.0003x − 0.617 R2 = 0.016	y = 0.0005x − 0.983 R2 = 0.045
November	1.5%	y = 0.0005x − 1.011 R2 = 0.161 **	y = 0.0005x − 0.990 R2 = 0.172 **	y = 0.0005x − 0.990 R2 = 0.210 ***
December	0.5%	y = 0.00005x − 0.087 R2 = 0.014	y = 0.0001x − 0.266 R2 = 0.048	y = 0.0001x − 0.177 R2 = 0.045

). Secondly, the evolutionary trend from March to June weakened successively (Table 3). In March, the increasing trend reached significant levels on the daytime (r = 0.43, p < 0.01), daily (r = 0.40, p < 0.01), and nocturnal (r = 0.34, p < 0.05) scales. Only the daytime (r = 0.28, p < 0.05) and daily (r = 0.27, p < 0.05) scales showed significant increases in April, while trends in May and June were unclear. Finally, the situation for the remaining four months is more complicated (Table 3). The results of the above three time scales showed an unknown trend in July, but decreased significantly (r: −0.27~−0.32, p < 0.05) in August. Both September and October were weaker, but the former tended to decline and the latter was the opposite (Table 3). These findings indicate that the pattern of annual precipitation distribution is undergoing significant and complex changes and that a more prominent daytime performance and a greater increase in the early rainy season are likely to be general characteristics of Gansu province. The interaction of these changes should be key to the fact that the vast majority of PCP has not yet shown a significant trend. The PCP is the center of gravity after considering annual precipitation as a whole, and this composite effect represents the theoretical period with the greatest probability of peak precipitation (applicable to the unimodal distribution). In addition, individual changes in monthly precipitation are also likely to be masked by mutual cancellation. For example, if the same change occurs on both sides of the peak, there will be no change in the result of PCP. In conclusion, the evolution of the annual distribution of precipitation requires a more refined systematic analysis, and the early and late rainy seasons are likely to be the most prominent periods.

4. Conclusions

Based on daily data from 23 national meteorological stations from 1970 to 2019, the spatiotemporal heterogeneity of daytime and nocturnal precipitation concentration in Gansu province was preliminarily analyzed with PCD and PCP as the core indexes. The main conclusions of this study can be summarized as follows:

(1)

Annual precipitation in Gansu province varies greatly, but nocturnal precipitation is generally more dominant, and its proportion is generally positively correlated with the former. The trend in humidification is obvious, and is mainly caused by the increase in the frequency of precipitation; both of these have a stronger performance during the day.

(2)

The majority of PCD is located between 0.55 and 0.75, showing a basic distribution pattern for daytime greater than the nocturnal, higher values, and stronger interannual fluctuations in arid areas. The decreasing trend for PCD is very clear and highly consistent, especially in the high-altitude area, and the increase in precipitation in the dry season and the improvement in precipitation uniformity in the wet season play a key role.

(3)

PCP often fluctuates slightly around the 39th–41st pentad (mid-late July), but it also illustrates the general rule that the value of daytime is earlier than that of nighttime, and the interannual variability is larger in arid areas. PCP has shown a relatively obvious advance trend in a few regions, and this is because the prominent and complex changes in the monthly precipitation distribution pattern have not been fully reflected.

Along with continuous humidification, the decrease in PCD and the advance of PCP are likely to be the priority direction for precipitation evolution in the arid region of Northwest China, especially during the day. For the deeper dynamic mechanism, the relationship between precipitation concentration and the East Asian summer monsoon, the South Asian summer monsoon, the Arctic Oscillation, and the Western Pacific Subtropical High will be the focus of the next work. These findings not only provide a new perspective for understanding and coping with regional climate change, but also have practical implication in various operational decisions in areas such as water resource management and the prevention of extreme drought and flood events.