Spatiotemporal Distribution of Drought and Humidity in China Based on the Pedj Drought Index (PDI)

: Drought is one of the most devastating natural disasters, especially in China. In drought assessment, the PDI has high robustness, is easier to obtain than indices such as SPEI and PDSI, and is more advantageous in regions with sparse stations. The present study employs the annual PDI with the precipitation and temperature data from 830 meteorological stations to systematically study the interannual variation characteristics of drought and humidity in China during 1970–2019. The results showed the following: (1) 26.6% of the total statistics from meteorological stations showed signiﬁcant ( p < 0.05) increases in annual PDI values throughout China during 1970–2019. (2) Air temperature plays a more remarkable role than precipitation in assessing the drying trend with PDI throughout China. (3) About 71% of stations experienced more drought events (PDI > 1) than humidity events (PDI < − 1), 14% of stations experienced more humidity events than drought events. (5) All stations experienced drought events (PDI > 1) with a frequency range from 6% to 32% and humidity events (PDI < − 1) with a frequency range from 8% to 36%. Most of the stations experienced extreme drought events (PDI > 3) with a frequency range from 2% to 10%, while only 177 stations experienced extreme humidity events (PDI < − 3) with a frequency range from 2% to 4%. (6) More than 67% of stations experienced drought conditions during both periods of 1972–1974 and 2000–2002 and even exceeded 80% in the three years 1972, 2000, and 2001. Both periods of 1976–1983 and 1985–1999 can be considered to be a humidity period throughout China. In conclusion, the PDI successfully expresses the interannual variation characteristics of drought and humidity events throughout China previously captured by other prominent, recommended drought indices.


Introduction
Drought is one of the most devastating natural disasters [1][2][3] and is directly associated with geographical location, altitude, and the distance from each big river system. The term drought is differently characterized according to research characteristics [4]. Drought is generally defined a significant shortage of natural freshwater supplies over a long period of time due to changes in precipitation and temperature patterns [5]. In recent years, the increasing human activities and climate warming have resulted in a change in the spatial distribution pattern of drought [6,7]. Several drought indices have already been adopted to evaluate various drought characteristics [8]. The more common drought indices are Palmer Drought Severity Index (PDSI) (Palmer, 1965), Standardized Precipitation Evapotranspiration Index (SPEI) [9], Standardized Precipitation Index (SPI) [10], Reclamation Drought Index (RDI) [11], Moisture Anomaly Index (Z-index) [12], and Precipitation Anomaly Index (RAI) [13]. Precipitation is the only indispensable input hydrometeorological variable in these indices. was recommended for evaluating drought characteristics due to its acceptable robustness and treatability for detecting drought [1,22]. Although readily available and popular, the sensitivity of the SPI in assessing drought is closely related to the soil moisture and groundwater. Meanwhile, the assessment ignores temperature, an important meteorological factor. These factors greatly reduce the SPI's ability to assess droughts in regions with extensive arid and semi-arid areas [4,26,27].
In contrast, the Pedj Drought Index (PDI) is sensitive to both precipitation and temperature, with high performance for assessing the onset of drought over extended periods (annual) [25]. In previous research, the PDI has been reported to identify drought events in a similar way to the well-known United Nations Environment Programme [28] Drought Index (AI) [25,29]. Another bright feature of the PDI is its outstanding ability to evaluate drought characteristics in regions with sparse meteorological stations and poorly recorded data [29]. In summary, the PDI has high robustness in drought assessment and is easier to obtain than SPEI, PDSI, etc., and has more advantages in drought assessment in regions with sparse stations. Therefore, the PDI is efficient and reliable in assessing drought characteristics. However, the adaptability of the PDI in spatiotemporal drought variability research throughout China is yet to be reported. The present article aims to fill this research gap.
The present article employs the annual PDI with the precipitation and temperature data from 830 meteorological stations to systematically study the interannual variation  (1) identify the drought and humidity severities of stations with statistically significant (p < 0.05) trends in the past 50 years in China; (2) analyze the impacts of temperature and precipitation changes on such significant trends; (3) calculate the frequency, longest duration, and spatial extent of both drought and humidity events throughout China.

Study Area
The People's Republic of China is located in the east of the Asian continent and on the west coast of the Pacific Ocean, with a land area of 9.6 million km 2 , extending from latitude of about 3 • 51 N to 53 • 33 N and from longitude of about 73 • 33 E to 135 • 05 E. The proportion of the landforms in China is as follows: mountain area 33.3%, plateau area 26%, basin area 18.8%, plain area 12%, and hilly area 9.9%, which makes it complex and diverse.
China has a vast territory, spanning five climatic zones from south to north: tropical zone, subtropical zone, warm temperate zone, middle temperate zone, and cold temperate zone [18]. There are multiple climate types in China, e.g., temperate monsoon climate, subtropical monsoon climate, tropical monsoon climate, tropical rain forest climate, temperate continental climate, and plateau mountain climate. Since the East Asian monsoon climate zone covers the vast majority of continental China, which is wet in summer and dry in winter, the annual distribution of precipitation is obviously different and the interannual variation is very significant [3,30]. Meanwhile, there is a significant negative correlation between precipitation and temperature [22]. Due to differences in precipitation (data of 830 meteorological stations during 1970-2019), China is divided into three main regions: humid region, transition region (sub-humid and semi-arid region), and arid region (Figure 1), accounting for 47%, 29%, and 24% of the total area, respectively. Because it is deeply influenced by the East Asian monsoon, southeastern China (humid region) is dominated by a warm and humid climate and has a relatively wetter climate, with annual precipitation ranging from 800 to 2200 mm and temperature ranging from 12.5 to 22°C [18]. In the transition region, the precipitation (400 to 800 mm) is reduced compared with that in the humid region due to the weakening of the monsoon's influence, and the temperature is also reduced (5 to 16°C). In the northwest of China, the cold and arid climate dominates much of the region, with annual precipitation ranging from 50 to 400 mm and temperature ranging from −2 to 8°C. It should be emphasized that precipitation and temperature show a significant negative correlation with the distance from the ocean [22].

Data
Daily precipitation and air temperature data from 1970 to 2019 were acquired from the National Meteorological Science Data Center of China (http://data.cma.cn, accessed on 22 December 2021). The missing values in each station were filled with the mean values of the nearest five stations calculated with the Tyson Polygon Method.
In order to adequately analyze the research results, China was divided into three major regions: humid region (I), sub-humid and semi-arid region (II), and arid region (III), with annual precipitation of >800 mm, 400-800 mm, and <400 mm, respectively. The regional divisions were obtained based on the annual precipitation data from 830 meteorological stations with the Inverse Distance Weighting Interpolation Method, the data were interpolated into1 km gridded data (see Figure 1). There were 383, 261, and 186 meteorological stations in region I, II, and III, respectively (see Table 1).

Calculation of PDI
The PDI developed by Pedj [31] was selected to identify drought periods and severity with multiple spatial and temporal scales [25,29]. PDI is an indicator determined by both temperature and precipitation. Hence, two standardized anomaly indices (SAIs) need to

Data
Daily precipitation and air temperature data from 1970 to 2019 were acquired from the National Meteorological Science Data Center of China (http://data.cma.cn, accessed on 22 December 2021). The missing values in each station were filled with the mean values of the nearest five stations calculated with the Tyson Polygon Method.
In order to adequately analyze the research results, China was divided into three major regions: humid region (I), sub-humid and semi-arid region (II), and arid region (III), with annual precipitation of >800 mm, 400-800 mm, and <400 mm, respectively. The regional divisions were obtained based on the annual precipitation data from 830 meteorological stations with the Inverse Distance Weighting Interpolation Method, the data were interpolated into1 km gridded data (see Figure 1). There were 383, 261, and 186 meteorological stations in region I, II, and III, respectively (see Table 1).

Calculation of PDI
The PDI developed by Pedj [31] was selected to identify drought periods and severity with multiple spatial and temporal scales [25,29]. PDI is an indicator determined by both temperature and precipitation. Hence, two standardized anomaly indices (SAIs) need to be calculated with priority, i.e., mean temperature and precipitation, respectively. The SAI is calculated as follows: where x is a particular year record, and x and s are the mean and the standard deviation of all year records over the time period, respectively. Then PDI is calculated as where SAI T and SAI p are the SAIs of mean temperature and precipitation, respectively, on the corresponding time scale. In the present study, the magnitudes of drought and humidity were defined as nine types, and each standard of the PDI range is shown in Table 2.

Statistical Analysis Method
Sen's slope estimator [32] was used to calculate the trends of PDI, SAI T , and SAI P time series. The Spearman rank correlation test [33] was used to detect the pairwise correlation between PDI, SAI T , and SAI P . The Spearman's rank correlation coefficient evaluates the linear relationship between two arrays, the coefficient is between −1 and +1, the negative coefficient indicates decreasing trend and vice versa. Since the test is nonparametric, the two sets of variables do not need to be normally distributed [29]. Additionally, the nonparametric Mann-Kendall (MK) [34][35][36] test was applied for data significance testing.

Results
Trend analysis identified 221 meteorological stations with significant (p < 0.05) increases in annual PDI values studied throughout China during 1970-2019, accounting 26.6% of total statistics stations. Regions I, II, and III had 121, 63, and 37 stations with significant increases in annual PDI, accounting for 31.6%, 24.1%, and 19.9%, respectively, of the number of statistical stations in their respective regions. Only 57 stations showed significant (p < 0.05) decreases, accounting 6.9% of total statistics stations. Regions I, II, and III had 13, 17, and 27 stations with significant decreases in annual PDI, accounting for 3.4%, 6.5%, and 14.5%, respectively, of the number of statistical stations in their respective regions (see Table 3). Among the 221 stations with increased annual PDI trends, 216 (accounting for 97.7%) showed significant increases in annual SAI T , and 5 showed significant decreases in annual SAI T . However, 204 (accounting for 92.3%, 204 out of 221) did not show significant trends in annual SAI p , 14 showed significant increases in annual SAI P , and 3 showed significant decreases in annual SAI P (see Figure 2a-c). Among the 57 stations with decreased annual PDI trends, 31 (accounting for 54.4%, 31 out of 57) showed significant decreases in annual SAI T , 14 showed significant increases in annual SAI T , and 12 did not show a significant trend in annual SAI T . Of these 57 stations, 48 (accounting for 84.2%) showed a significant increase trend in annual SAI P , while 9 did not show a significant trend in annual SAI P (see Figure 2a-c). The top 5% of stations with significant changes in annual PDI in each region (see Figure 2a) were selected for extreme stations' analysis (see Figures 3 and 4). The top 5% of region I stations (6 out of 121) with significant increases in annual PDI had values of 0.0482, 0.0438, 0.0415, 0.0396, 0.0389, and 0.0386/year, respectively. The top 5% of region II stations (3 out of 63) with significant increases in annual PDI had values of 0.0671, 0.0437, and 0.0430/year, respectively. The top 5% of region III stations (2 out of 37) with significant increases in annual PDI had values of 0.0471 and 0.0462/year, respectively. Extreme stations with significant increases in annual PDI of region I had more drought events after 2000 (see Figure 3a-f). Extreme stations of region II had more drought events since the middle 1990s (see Figure 3g-i). The one extreme station of region III had more drought events after 2000 (see Figure 3j), the other one had more drought events since the 1980s (see Figure 3k). The top 5% of region I, region II, and region III stations (1 out of 13, 1 out of 17, 1 out of 27) with significant decreases in annual PDI had values of 0.430, 0.471, and 0.462/year, respectively. Extreme stations of region I, II, and III with significant decreases in annual PDI had more humid years since the early 1990s (see Figure 4a), the middle 1980s (see Figure 4b), and the early 1980s (see Figure 4c), respectively.  The top 5% of stations with significant changes in annual PDI in each region (see  Figure 3a-f). Extreme stations of region II had more drought events since the middle 1990s (see Figure 3g-i). The one extreme station of region III had more drought events after 2000 (see Figure 3j), the other one had more drought events since the 1980s (see Figure 3k). The top 5% of region I, region II, and region III stations (1 out of 13, 1 out of 17, 1 out of 27) with significant decreases in annual PDI had values of 0.430, 0.471, and 0.462/year, respectively. Extreme stations of region I, II, and III with significant decreases in annual PDI had more humid years since the early 1990s (see Figure 4a), the middle 1980s (see Figure 4b), and the early 1980s (see Figure 4c), respectively.    Annual PDI of 703 (85%, 703 out of 830) stations showed statistically significant correlations with annual SAIT (see Figure 5a). Only five of them showed negative correlations, all of which were in region I. The rest of them showed positive correlations, region I, II, and III had 318 (83%, 318 out of 383), 222 (85%, 222 out of 261), and 158 (85%, 158 out of 186) stations, respectively (see Figure 5a). Slightly dissimilarly, annual PDI showed statistically significant negative correlations with annual SAIP at all stations (see Figure 5b). There were 651 (78%, 651 out of 830) stations that showed positive relationships between annual SAIT and SAIP throughout China, but only at about 29% (188 out of 651) of them were significant (p < 0.05) (see Figure 5c). There are 179 (22%, 179 out of 830) stations that Annual PDI of 703 (85%, 703 out of 830) stations showed statistically significant correlations with annual SAI T (see Figure 5a). Only five of them showed negative correlations, all of which were in region I. The rest of them showed positive correlations, region I, II, and III had 318 (83%, 318 out of 383), 222 (85%, 222 out of 261), and 158 (85%, 158 out of 186) stations, respectively (see Figure 5a). Slightly dissimilarly, annual PDI showed statistically significant negative correlations with annual SAI P at all stations (see Figure 5b). There were 651 (78%, 651 out of 830) stations that showed positive relationships between annual SAI T and SAI P throughout China, but only at about 29% (188 out of 651) of them were significant (p < 0.05) (see Figure 5c). There are 179 (22%, 179 out of 830) stations that showed negative relationships between annual SAI T and SAI P , but only at about 4% (8 out of 179) of them were significant (p < 0.05) (see Figure 5c).

Discussion
The annual PDI time series trend shows that a significant portion of the region has become drier over the past 50 years, with a few areas getting wetter. Meanwhile, the wetter region (region I) had a higher proportion of stations with significant increases in annual PDI, and the drier region (region III) had a higher proportion of stations with the significant decreases in annual PDI. This reveals that there is a tendency toward greater wetness in arid areas, while the characteristics of climate change in humid areas were the opposite throughout China. This is consistent with the drought analysis using SPEI [3]. It is generally believed that the increase in drought trend (annual PDI trend increases) is associated with a significant increase in SAIT [25], which is consistent with the conclusion of this paper. However, there is no significant correlation between this drought trend (annual PDI trend increases) and SAIP throughout China. An important reason is that the change of air temperature changes the regional water vapor flux [37,38], which partly suggests that air temperature played a more remarkable role than precipitation in assessing the drying trend. This is consistent with the global analysis of drought with SPEI [39]. Hence, in the context of global warming, air temperature should be given priority in order to assess the drying trend more objectively.
Severity, frequency, duration, and spatial distribution are the four most distinctive features of drought [25,40]. Most regions experienced more drought events than humidity events throughout China during 1970-2019, and the stations with high frequency of drought events were mostly concentrated in the central and eastern regions, while the stations with high frequency of humidity events were scattered. This confirms the results of previous studies by some scholars [3,[41][42][43]. From the results of the statistical analysis, the long-duration drought of the stations was mainly concentrated in two periods, the 1970s and after 2000, this is consistent with the conclusions in a recent report [44,45]. However, the long-duration humidity of stations mostly concentrated in the 1980s and 1990s, this is consistent with the global analysis of drought using SPEI [39]. Additionally, China experienced extensive droughts during two periods, i.e., 1972-1974 and 2000-2002, with a relatively wet period between them, which confirms the conclusions of some research [46][47][48]. For the period after 2000, the present study indicates that a large part of China has experienced drought that has lasted to this day, which has been shown in recent research [24,39,49].
In present research, PDI was adopted to study the variation characteristics of drought and humidity in China during 1970-2019, and was in good agreement with the results of

Discussion
The annual PDI time series trend shows that a significant portion of the region has become drier over the past 50 years, with a few areas getting wetter. Meanwhile, the wetter region (region I) had a higher proportion of stations with significant increases in annual PDI, and the drier region (region III) had a higher proportion of stations with the significant decreases in annual PDI. This reveals that there is a tendency toward greater wetness in arid areas, while the characteristics of climate change in humid areas were the opposite throughout China. This is consistent with the drought analysis using SPEI [3]. It is generally believed that the increase in drought trend (annual PDI trend increases) is associated with a significant increase in SAI T [25], which is consistent with the conclusion of this paper. However, there is no significant correlation between this drought trend (annual PDI trend increases) and SAI P throughout China. An important reason is that the change of air temperature changes the regional water vapor flux [37,38], which partly suggests that air temperature played a more remarkable role than precipitation in assessing the drying trend. This is consistent with the global analysis of drought with SPEI [39]. Hence, in the context of global warming, air temperature should be given priority in order to assess the drying trend more objectively.
Severity, frequency, duration, and spatial distribution are the four most distinctive features of drought [25,40]. Most regions experienced more drought events than humidity events throughout China during 1970-2019, and the stations with high frequency of drought events were mostly concentrated in the central and eastern regions, while the stations with high frequency of humidity events were scattered. This confirms the results of previous studies by some scholars [3,[41][42][43]. From the results of the statistical analysis, the long-duration drought of the stations was mainly concentrated in two periods, the 1970s and after 2000, this is consistent with the conclusions in a recent report [44,45]. However, the long-duration humidity of stations mostly concentrated in the 1980s and 1990s, this is consistent with the global analysis of drought using SPEI [39]. Additionally, China experienced extensive droughts during two periods, i.e., 1972-1974 and 2000-2002, with a relatively wet period between them, which confirms the conclusions of some research [46][47][48]. For the period after 2000, the present study indicates that a large part of China has experienced drought that has lasted to this day, which has been shown in recent research [24,39,49].
In present research, PDI was adopted to study the variation characteristics of drought and humidity in China during 1970-2019, and was in good agreement with the results of other drought indices [3,39,41,50]. Therefore, the present research shows that the PDI, like other drought indices, successfully expresses the interannual variation characteristics of drought and humidity events throughout China and provides certain reference significance for the research and early warning of drought and flood disasters in China.

Conclusions
The present article employs the annual PDI with the precipitation and temperature data from 830 meteorological stations to systematically study the interannual variation characteristics of drought and humidity in China during 1970-2019. The following major conclusions were drawn: (1) There were 26.6% of stations that showed significant increases in annual PDI values throughout China during 1970-2019, the proportion of stations with significant increases in region I (31.6%), II (24.1%), and III (19.9%) gradually decreased. Only 6.9% of stations showed significant decreases throughout China, the proportion of stations in region I (3.4%), II (6.5%), and III (14.5%) gradually increased. (2) Air temperature played a more remarkable role than precipitation in assessing the drying trend with PDI. Both precipitation and air temperature were important in assessing the wetting trend, but the former occupied a more prominent position. (3) Most stations (85%) showed significant positive correlations between annual PDI and SAI T . PDI and SAI P were significantly negatively correlated at all stations. SAI T and SAI P were positively correlated in some stations (23%) and negatively correlated in a few stations (1%). (4) Most stations (71%) experienced more drought events, and a few (14%) experienced more wet events or humidity events over the past 50 years. The frequency of drought events ranged from 6% to 32% and humidity events ranged from 8% to 36%. Most stations (95%) experienced extreme drought events, with a frequency range from 2% to 10%. A few (21%) experienced extreme humidity events, with a frequency range from 2% to 4%.