Change in summer daily precipitation and its relation with air temperature in Northwest China during 1957-2016

CaiXia Zhang , XunMing Wang, YongZhong Su, ZhiWen Han, ZhengCai Zhang, Ting Hua

Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, Gansu 730000, China

ABSTRACT On the basis of the summer daily-precipitation meteorological data collected from weather stations across Northwest China from 1957 to 2016, this study evaluated the trends in 12-daily precipitation indices in the summer season and their relations with air temperature. Precipitation-event intensity, which was averaged over the total study area, increased in recent decades although the total precipitation continuously decreased. In particular, intensity generally decreased in the northern and eastern parts and increased in the southern and western parts of the study area. None of the 12 precipitation indices was significantly correlated with temperature in Xinjiang; R95N (number of events with precipitation greater than the long-term 95th percentile), RX1day (greatest 1-day total precipitation), PI (simple daily intensity), and R10 (number of heavy-precipitation days) were significantly and positively correlated with temperature in Qinghai-Gansu. However, low correlation coefficients were observed. In the Loess Plateau, P (total precipitation), WS (maximum number of consecutive wet days),R95N, and WD (number of wet days) were significantly and negatively correlated with temperature, whereas Gini (gini concentration index) and DS (maximum number of consecutive dry days) were significantly and positively correlated with temperature. Results of the study suggested that climate shift was evident in terms of daily precipitation, and the study area faced new challenges involving precipitation-event intensity increasing in the southwestern part and unevenly dispersing in the northwest.

Keywords: summer daily precipitation; temperature; China

1 Introduction

The global average land-and-sea surface temperature has increased by 0.85 °C since the mid-19th century (IPCC, 2013). Climate warming is expected to intensify precipitation variability worldwide, i.e., low precipitation frequency, long dry periods, and large intensity individual precipitation events (Easterling et al., 2000; Min et al., 2011; IPCC, 2013). Similar to other world regions, China has experienced significant changes in precipitation events, including regime,interannual variability, and intensity of extreme events (Zhai et al., 2005; Wang et al., 2010; Ye,2014). Some of the most important changes are the decreasing number of precipitation days and increasing precipitation intensity (Zhai et al., 2005), that is, a high percentage of precipitation amount occurring in the span of a few days or rainstorm events. Increased precipitation extremes exert hydrological and environmental consequences, such as an increase in flood occurrence and soil erosion (Miller, 1994; Zhang et al.,2017). For example, high occurrences of floods and geological disasters in China are suggested to be linked with intensified precipitation events and changes in spatiotemporal precipitation concentration since the 1990s (Li et al., 2011). Extreme climatic events, such as unusual floods and droughts, can also exert a strong impact on society and ecosystems (Shi et al., 2007; Garbero and Muttarak, 2013; Ye et al.,2016).

Various indices have been used to quantify precipitation extremes (Sen Roy and Balling, 2004; Moberg and Jones, 2005; Alexander et al., 2006; Sillmann and Roeckner, 2008; Sharma and Babel, 2014). The Expert Team on Climate Change Detection and Indices suggested 12 percentile-, threshold-, or duration-based indices derived from daily-precipitation data (Sillmann and Roeckner, 2008; Zhang et al., 2011). Using these daily-precipitation indices, Knapp et al. (2015)and Ye et al. (2018) analyzed the importance of each index for distinguishing extremely wet from extremely dry years. Several studies have suggested that a warmer climate has possibly resulted in precipitation extremes, including a lower frequency of rain events, more days with heavy or very heavy precipitation, and extended drought spans (Easterling et al.,2000; Min et al., 2011; Ye, 2014; Fischer and Knutti,2016). Wang and Zhou (Diao et al., 2015) analyzed the trends in annual and seasonal mean precipitation,and the trends in extreme precipitation events in China during 1961-2001. The results indicated that the annual mean precipitation increased significantly in southwestern, northwestern, and eastern China and decreased significantly in central, northern, and northeastern China. However, the characteristics of different indices and their capacity to describe climate changes have not been studied.

Moreover, previous studies mainly compared the annual and seasonal trends in climate variables before and after climate shift (Shi et al., 2003, 2007) and analyzed the overall trend of heavy-precipitation events during the last decades (Zhang et al., 2012;Wan et al., 2014). Since 1987, the climate regime in the western part of northwestern China has shifted from warm-dry to warm-wet (Shi et al., 2003, 2007).Based on the important findings of these studies, the total time series was divided into two periods, i.e.,1957-1986 and 1987-2016.

Based on long-term precipitation records (1957-2016)over northwestern China, this study focused on daily precipitation and examined the effect of global climate shift on the spatial and temporal trends of 12 daily-precipitation indices over northwestern China.

2 Materials and methods

2.1 Description of study sites and data

The study area of northwestern China (33°N-48°N,73°E-116°E), which includes five northwestern provinces and the Loess Plateau, is a typical arid and semi-arid area (Figure 1) with limited water resources and fragile ecological conditions and is thus considered an area sensitive for climate change (Shi et al., 2007). Precipitation in the region is characterized by high spatial-temporal variations： (1) summer-season (June, July, and August) precipitation markedly varies annually, ranging from tens of millimeters per year in dry years to more than 1,000 mm per year in flood years; (2) precipitation is unevenly distributed during the season; and (3) spatial distributions of precipitation are heterogeneous. Rainfall amounts decreased in the eastern part and increased in the western part of northwestern China; heavy precipitation was dominant in the whole region, and the number of rainfall days played an important role in the rainfall amounts (Chen and Dai, 2009). For the convenience of presentation, the total study area was divided into three subregions, from east to west： the Loess Plateau,Qinghai-Gansu, and Xinjiang. The climate regime in the western part of northwestern China shifted from warm-dry to warm-wet in 1987; i.e., the average precipitation during 1987-2000 increased in northwestern China, especially in Xinjiang, compared with that in 1961-1986 (Shi et al., 2003, 2007). Heavy-precipitation events tended to occur severely and frequently since 1980 in Xinjiang (Zhang et al., 2012). Total precipitation showed a different trend across the Loess Plateau in northwestern China, decreasing at most sites during 1957-2009 (Wan et al., 2014). However,more of the precipitation tended to occur in several extreme events, resulting in an increasing trend in erosive precipitation; these trends were evident in areas of sediment sources on the Loess Plateau (Wan et al., 2014) .

2.2 Data

Station-based meteorological data were quality-controlled (Zhai et al., 2005) and obtained from the China Meteorological Administration Data Sharing Service System (http：//cdc.nmic.cn/home.do). The quality control for daily precipitation included these components： (1) negative values were set to missing data; and(2) outliers in daily-precipitation data were defined as those higher than an upper limit of 500 mm(http：//etccdi.pacificclimate.org/software.shtml), with outliers also set to missing values. This study focused on a 60-year time series spanning from 1957 to 2016,due to the lack of daily observational data prior to 1957 at several meteorological stations in northwestern China (Ye et al., 2013). A total of 175 observatories with precipitation data missing a maximum of five years were included in the study (Figure 1).

2.3 Methods

Trends at individual stations were evaluated with the nonparametric Mann-Kendall (M-K) test, which reliably identifies monotonic linear and nonlinear trends in abnormal datasets containing outliers and commonly has been used to assess trends in meteorological parameters (Mann, 1945). In recent decades,the M-K test has been used to analyze long-term precipitation trends (Ye et al., 2010; Sayemuzzaman and Jha, 2014). The ordinary Kriging interpolation method (López-Moreno et al., 2010) was applied to interpolate station-based data to a raster surface for the total study region.

To thoroughly characterize daily precipitation, we selected 12 indices (Table 1) to describe different aspects of the daily-precipitation regimes. These indices had been widely used in previous studies(Sen Roy and Balling, 2004; Moberg and Jones,2005; López-Moreno et al., 2010; Sharma and Babel,2014).

Figure 1 Summer-season precipitation (averaged during 1957-2016) across the study area and distribution of meteorological stations

Table 1 Acronyms and definitions of the 12 selected precipitation indices

3 Results

3.1 Trends in precipitation indices

When averaged for each subregion, precipitation indices (including PRCPtot, R95pT, RX1day, and RX5day) significantly increased in Xinjiang and Qinghai-Gansu and slightly decreased (insignificantly) in the Loess Plateau during the last 60 years(Table 2). In particular, an increased number of meteorological stations in Xinjiang showed significantly positive trends (rather than negative) for PRCPtot(18 stations with positive trend vs. 1 negative), R95pT(13 vs. 0), RX1day (9 vs. 1), and RX5day (11 vs. 0).The same spatial patterns were evident in Qinghai-Gansu for PRCPtot (14 vs. 1), R95pT (9 vs. 0), RX1day(9 vs. 0), and RX5day (9 vs. 0). In the Loess Plateau,most stations showed nonsignificant trends for the four indices (Figure 2). PRCPtot decreased at 87% of the stations, whereas only two stations showed statistically significant changes. Only one station showed significant increase in PRCPtot. By contrast, more stations displayed significantly positive trends for R95pT (3 vs. 2), RX1day (3 vs. 1), and RX5day (3 vs. 1).

Table 2 Trends in 12 precipitation indices during 1957-2016 in three subregions in Northwest China.The trends were estimated as the slopes of regional M-K trend tests

Figure 2 Spatial distribution of the trends of different precipitation indices describing the magnitude of precipitation events in the summer season

When averaged for each subregion, WS, PI, C95,R10, R95N, and WD significantly increased, whereas DS significantly decreased in Xinjiang (Figure 3,Table 2). These precipitation indices showed generally similar trends in Qinghai-Gansu, except WS and DS were not statistically significant. In the Loess Plateau, WD is the only index that displayed a significant trend (negative). Notably, PI slightly increased despite a decrease in the total precipitation in the area.The Gini concentration index showed no significant trend in all three subregions (Table 2). Moreover, WS did not change at most stations in all three subregions during the last 60 years (Figure 2). Only one station showed a significant positive trend in Xinjiang. WS significantly increased at four stations and decreased at one station in Qinghai-Gansu. In Xinjiang, more stations showed significantly positive trends than significantly negative trends for PI (7 stations with a positive trend vs. 0 with negative), C95 (16 vs. 1), R95N(17 vs. 1), R10 (7 vs. 1), and WD (17 vs. 1), whereas considerably fewer stations showed significantly positive trends for Gini (3 vs. 11) and DS (1 vs. 8). The same spatial patterns were evident in Qinghai-Gansu for PI (10 stations with positive trend vs. 2 with negative), C95 (12 vs. 1), R95N (12 vs. 2), R10 (12 vs. 1),WD (9 vs. 2), Gini (5 vs. 8), and DS (2 vs. 7). In the Loess Plateau, most stations did not show a significant trends for PI, C95, R95N, and R10. In general,more or a similar number of stations displayed significant positive trends for PI (5 stations with a positive trend vs. 0 negative), C95 (3 vs. 1), R95N (2 vs. 2),and R10 (1 vs. 1). By contrast, fewer stations showed significant positive trends for Gini (2 vs. 14) and DS(0 vs. 10). Compared with DS, a majority of the stations (68%) displayed negative trends in WD in the Loess Plateau; and results for 12 stations were significant.

Figure 3 Trends in daily-precipitation indices during the summer seasons of 1957-1976 and 1977-2014

3.2 Abrupt changes in precipitation indices

Despite their similar trends during the total period of 1957-2016 in Xinjiang and Qinghai-Gansu, precipitation indices showed markedly different abrupt changes in the two subregions. Most precipitation indices (except for PI) showed an abrupt change before and after regional climate shift in 1986 (Figure 3).Upward shifts were observed in PRCPtot, R95pT,RX1day, RX5day, WS, C95, R10, R95N, and WD,whereas downward shifts were found for Gini and WD. By contrast, abrupt changes occurred in different years in Qinghai-Gansu. Four indices (P, C95,R10, and R95N) displayed an abrupt change in 1978,one index (R95T) in 1993, one index (PI) in 1986,and one index (Gini index) in 1973 in Qinghai-Gansu(Table 3). Upward shifts were found in P, C95, R10,R95N, R95T, and PI, whereas downward shifts were observed in the Gini index. These index shifts were not as evident as those in Xinjiang. No abrupt change was detected for five indices (RX1day, RX5day, WS,DS, and WD) in Qinghai-Gansu (Table 3). In the Loess Plateau region, 9 of the 12 precipitation indices did not show any abrupt change; the other three indices (Gini, DS, and WD) displayed abrupt change in 1996. Gini and DS displayed upward shifts, whereas WD showed downward shifts.

Table 3 Abrupt changes in 12 precipitation indices during 1957-2016 in the three subregions in Northwest China

3.3 Relationship of precipitation indices with air temperature

None of the 12 precipitation indices was significantly correlated with temperature in Xinjiang (Table 4). R95T, RX1day, PI, and R10 were significantly and positively correlated with temperature in Qinghai-Gansu; however, the correlation coefficients were low (Table 4). In the Loess Plateau, P, WS, R95N,and WD were significantly and negatively correlated with temperature; and Gini and DS were significantly and positively correlated with temperature (Table 4).Moreover, the correlation coefficients were relatively higher for Gini, DS, and WD.

Table 4 Relationship of precipitation indices with air temperature

When averaged for each subregion, warming trends were observed in all three subregions in the total time series of 1957-2016 (Figure 4). Abrupt changes were detected for mean air temperature in Xinjiang and Qinghai-Gansu in 1993 and in the Loess Plateau in 1996 (Figure 4). In all subregions, temperature was markedly higher in the recent two decades,contributing the most to warming trends over the 60 years. The abrupt change in temperature occurred later than that of precipitation indices in Xinjiang. The same pattern was observed in Qinghai-Gansu, but the abrupt change of R95 T occurred in the same year as that of temperature. In the Loess Plateau, abrupt changes in the three precipitation indices (Gini, DS,and WD) and those in temperature occurred simultaneously (Table 4). In particular, abrupt changes in all precipitation indices did not concurrently occur with those of the temperature at the majority of stations in Xinjiang and Qinghai-Gansu (Figure 5). Stations with abrupt changes in precipitation that occurred earlier than abrupt changes in temperature accounted for the largest proportion of the total stations in the two subregions. In the Loess Plateau, the number of stations with abrupt changes in precipitation was similar to that of stations with abrupt changes in temperature,which is relatively high for Gini, DS, and WD. For Gini, change was observed simultaneously at 35 stations and at different times at 27 stations. By contrast,WD changed at 34 stations simultaneously and at different times at 28 stations. For DS, abrupt changes in precipitation and temperature occurred simultaneously,and these stations accounted for the largest proportion.

Figure 4 Abrupt changes occurred in all three subregions during the total time series of 1957-2016

4 Discussion

Various indices have been applied to quantify precipitation extremes (Sen Roy and Balling, 2004;Moberg and Jones, 2005; Sharma and Babel, 2014);therefore, an understanding of the spatial and temporal characteristics of different indices is essential.

Similar to findings of previous studies (Shi et al.,2003, 2007), a warm and wet trend was observed in the western and middle parts of northwestern China(Xinjiang and Qinghai-Gansu). Differing from the previously reported absence of trends as observed during 1961-2000 (Shi et al., 2003, 2007), the long-term observation of the present study suggested a warm and dry trend in the eastern part of northwestern China (the Loess Plateau). The analysis of this study showed that abrupt changes occurred in 1986 for most precipitation indices in Xinjiang, which agrees with the 1987 climate shift observed in previous studies(Shi et al., 2003, 2007). Seven precipitation indices presented abrupt changes in Qinghai-Gansu, four of which occurred in 1978 (Table 3). That point in time was close to the global climate shift that occurred in 1976 (Hartmann and Wendler, 2005; IPCC, 2013; Ye,2014). In Xinjiang and Qinghai-Gansu, abrupt changes in precipitation indices occurred earlier than those in temperature (in 1993) and slightly later than the global climate shift in 1976. Most precipitation indices did not significantly correlate with precipitation during the study period of 1957-2016. Therefore, increased precipitation and its extremes were not caused by regional warming but were inherent to the climate change worldwide (Feng et al., 2006; Zhang et al.,2012). In the Loess Plateau, shifts in precipitation indices were strongly correlated with those in temperature, thus suggesting that regional warming may cause changes in precipitation.

In the total study area, the western and middle parts (Xinjiang and Qianghai-Gansu) presented lower mean summer precipitation than those in the east(Loess Plateau). The spatial redistribution of precipitation resulted in a trend of "increasingly wet dry re-gions and increasingly dry wet regions" in northwestern China, which differs from the global trend of "dry regions drying out further, whereas wet regions become wetter as the climate warms" (Chou et al.,2009). The opposite precipitation pattern observed in this study was consistent with that in a recent study in which aridity changes over land did not always follow a simple intensification of existing patterns(Greve et al., 2014).

Figure 5 Abrupt changes occurred in mean temperature(a) Xinjiang, (b) Qinghai-Gansu, (c) Loess Plateau

5 Conclusion

In summary, trends in daily-precipitation indices in the summer season and their relations with air temperature differ substantially across space. Precipitation intensity generally decreased in the northern and eastern arts and increased in the southern and western parts of Northwest China. Precipitation indices were uncorrelated with temperature in Xinjiang. In the Loess Plateau, six precipitation indices were significantly correlated with temperature. The results of the study suggested that climate shift in terms of daily precipitation was evident, and the study area faced new challenges where precipitation-event intensity increased in the southwest and was more unevenly dispersed in the northwest. Climate warming may have resulted in significant changes in daily precipitation in the Loess Plateau.

Acknowledgments:

This research was supported by the National Natural Science Foundation of China (Nos. 41101006 and 31570467) and the Key Frontier Program of the Chinese Academy of Sciences (Grant No. QYZDJSSW-DQC043).