Analyzing the impact of different forms of precipitation (rain and snow) on drought Open Access

With global climate change and temperature increase, patterns and effects of precipitation for Earth's ecosystems and socio-economy have become a focal point of attention among scholars worldwide (Khosravichenar et al. 2023; Kotz et al. 2024). Precipitation can have significant impacts on flood disasters (Dottori et al. 2023), urban waterlogging (Zhang et al. 2021), water resources management (Li, M. et al. 2022), and drought disasters (Ciais et al. 2005). Based on the study of Mishra & Singh (2010), drought means a sustained, extended deficiency in precipitation. Due to deficiency in precipitation, many extreme drought events have occurred across China. For example, Yao et al. (2018) investigated an extreme drought event across the Yangtze River Basin in the midsummer of 2022. Lyu et al. (2023) revealed an extreme drought event across southwestern China from autumn 2009 to spring 2010 (Li et al. 2019). Precipitation exhibits various forms and uneven spatiotemporal distributions in China (Aalıjahan et al. 2021; Xiao et al. 2024). Therefore, this study was necessary to investigate the relationship between precipitation and drought over China.

Precipitation exists in various forms, including rainfall, snowfall, sleet, and hail. Most precipitation forms in China are rainfall and snowfall. Many studies have analyzed the reasons for different forms of precipitation. When atmospheric temperature is relatively high, water vapor condenses into liquid droplets and rain falls to the ground under warm conditions (Aalijahan et al. 2022; Liu et al. 2024). When atmospheric temperature is low, water vapor condenses into ice crystals, forming snowflakes, and snow falls to the ground under cold conditions (Aalijahan et al. 2019; Zhao et al. 2023). Sleet forms under conditions of dramatic temperature changes in the atmosphere, causing raindrops and ice crystals to mix at the boundaries between different temperature layers (Nouri & Homaee 2021; Rezaei et al. 2023). Snowfall commonly occurs in regions with relatively low temperatures (Wei, R. et al. 2022). Ding et al. (2014) developed a new parameterization scheme to separate snow, sleet, and rain from precipitation. The parameterization scheme depended on statistical formulas and daily meteorological data to separate different forms of precipitation. Some precipitation products were developed by different institutions, including the reanalysis of data (Wu et al. 2023), multivariate fusion data, and spatial interpolation data. Reanalysis of data involves reanalyzing historical observational precipitation data using numerical models to generate a consistent time-series and spatial resolution of continuous precipitation data. The ERA5 dataset was typical reanalysis data. The ERA5 dataset was driven by the integrated forecasting system model and developed by the European Centre for Medium-Range Weather Forecasts. Fusion data combined multiple sources and types of satellites, ground-based measurement stations, and radar sensor observation data to provide more comprehensive and accurate information (Wei, L. et al. 2022). IMERG was typical fusion data and combined satellite and ground-based precipitation data by multiple algorithms. Spatial precipitation interpolation data were based on ground observation data, utilizing mathematical interpolation methods (including bilinear interpolation, inverse distance weighting interpolation, and kriging interpolation) to estimate and fill precipitation distributions in geographical space (Gong et al. 2014; Thrasher et al. 2022). The ground-based observation data were commonly used to estimate the performance of different precipitation products in many studies (Wei et al. 2021). However, reanalysis and fusion data might have more advantages than spatial interpolation data in periods and regions that lack ground observation data or high spatiotemporal resolution. More than 2,400 meteorological stations that covered China have been observed since the 1960s (Miao & Wang 2020). Therefore, spatiotemporal precipitation data of ground observation might be sufficient for China. Spatial precipitation interpolation data might have more reliability and availability than reanalysis and fusion data in this study.

There were many studies on analyzing drought, generally including investigation of drought characteristics, analyzing multi-type drought propagation, and developing new drought assessment methods. For example, Pandya & Gontia (2023) investigated severity, duration, and frequency characteristics of meteorological drought across the Gujarat State of India. They also estimated the evolution, severity, and trends of drought in mainland China. Guo et al. (2020) analyzed the triggering mechanism from meteorological drought to hydrological drought in the Wei River Basin. Zhang et al. (2019) developed a new drought index (standardized moisture anomaly snow index) to estimate global drought characteristics. Due to the varying physical characteristics of different forms of precipitation, rain, snow, and sleet might impact the regional hydrological cycles (Liu et al. 2022). Precipitation is a critical driving factor of drought. However, many studies related to drought ignored the impact of different forms of precipitation on drought characteristics and propagation. In high-latitude and plateau regions of China, precipitation often occurs in the forms of rainfall and snowfall. Therefore, this study was necessary to investigate the impact of different forms of precipitation on drought characteristics and propagation in different climate sub-regions of China.

The objectives of this study were (1) to investigate the precipitation amounts and trends of different forms in various climatic sub-regions of China, (2) to reveal the spatiotemporal differences of drought event characteristics identified based on different forms of precipitation, and (3) to investigate and analyze the impact of different forms of precipitation on drought propagation. This study could provide insights into the mechanisms of drought propagation and water resource allocation planning in high-altitude regions.

Considering the temperature, precipitation, moisture, and vegetation conditions, China is divided into seven climate sub-regions (Weng et al. 2023). The climate sub-regions are the temperate and warm-temperate deserts of northwest China (TWDN), the temperate grassland of Inner Mongolia (TGIM), the temperate humid and sub-humid northeast China (THSN), the warm-temperate humid and sub-humid north China (WHSN), the subtropical humid central and south China (SHCS), the Qinghai–Tibetan Plateau (QTP), and the tropic humid south China (THSC).

The gridded dataset of different forms of precipitation (snow, sleet, and rain) was provided by Su & Zhao (2022) (https://data.tpdc.ac.cn/en/data/). The parameterization scheme for separating different forms of precipitation was applied to develop this dataset (Ding et al. 2014). The observational data of daily meteorological data (including air temperature, precipitation, RH, and surface pressure) and elevation data were employed to develop the dataset of the different forms of the precipitation (Su et al. 2022). The gridded dataset of the different forms of the precipitation included rainfall, sleet, and snowfall data, and the spatiotemporal resolution was 0.25° × 0.25° and daily, respectively. The temporal and spatial coverage range was 1961–2016 and China, respectively. The gridded dataset of different forms of precipitation was developed based on observational data, and the dataset might have more reliability than reanalysis and simulation of datasets (Dehaghani et al. 2023). In this study, the different forms of precipitation were used to investigate meteorological drought.

A daily 0.25° × 0.25° terrestrial hydrological dataset (VIC-CN05.1) provided the runoff data, which was developed by Miao & Wang (2020). The daily 0.25° × 0.25° meteorological dataset, soil parameters, snow elevation bands, and digital elevation model were regarded as inputs for this hydrological dataset. The variable infiltration capacity (VIC) model was used to simulate the hydrological process for this hydrological dataset (Keller et al. 2023). This hydrological dataset provided many hydrological process variables, including total runoff, soil moisture, terrestrial water storage change, and evapotranspiration. The daily 0.25° × 0.25° terrestrial hydrological dataset (VIC-CN05.1) indicated that the simulated total runoff showed great agreement in the Yangtze, Southeast, and Pearl River basins, with a positive relative error ≤7% and a Nash–Sutcliffe efficiency coefficient ≥0.87. The spatiotemporal resolution of the hydrological dataset was 0.25° × 0.25° and daily, respectively. The temporal and spatial coverage range was 1961–2015 and China, respectively. The total runoff of the hydrological dataset was used to investigate hydrological drought.

In this study, the standardized precipitation index (SPI) and the standardized runoff index (SRI) were employed to describe meteorological drought and hydrological drought (McKee et al. 1993; Ding et al. 2021a). The monthly time-series of precipitation and runoff fit the distribution function (Yao et al. 2018). The distribution functions of gamma and log-normal were used to fit the SPI and the SRI, respectively (Ding et al. 2021b). The detailed calculation process of the SPI and the SRI can be found in Wu, J. et al. (2018). Based on different forms of precipitation, precipitation, rainfall, and snowfall were used to calculate the SPI. When the monthly time-series of precipitation, rainfall, and snowfall were used to calculate the SPI, it was defined as SPIp, SPIr, and SPIs, respectively. Both the SPI and the SRI were calculated for one- to 18-month time-series in this study.

We used run theory to investigate the drought events for each grid cell (Zhang, Q., Miao, C. et al. 2023). Run theory can identify meteorological and hydrological drought events through pooling and excluding methods. By investigating the duration and severity of drought events in each grid, we could reveal the characteristics of drought (Pokhrel et al. 2021). When the drought index value was less than −1, the drought was defined as moderate drought based on the classification of drought levels (Kalisa et al. 2020). The threshold of the SPI and the SRI was defined as −1 in this study (Guo et al. 2020).

We employed the maximum Pearson correlation coefficient (MPCC) method to investigate the drought propagation degree and time from meteorological drought to hydrological drought (Xu et al. 2018). MPCC is a method used to compute the maximum correlation coefficient between two sets of random variables. The method aims to identify the strongest linear correlation between two time-series of random variables. Specifically, the MPCC method achieves the following steps: (1) We computed the Pearson correlation coefficients for different lags (time scales) between the SPI and the SRI. (2) The MPCC method was used to calculate the drought propagation time from meteorological drought to hydrological drought. (3) The maximum correlation coefficient was defined as the drought propagation degree from meteorological drought to hydrological drought when we calculated the maximum lag. Moreover, the MPCC method has been used to investigate drought propagation in many studies (Jehanzaib et al. 2023).

The box plot in Figure 8 quantifies the duration and severity characteristics of drought events in different climatic sub-regions. From Figure 8(a), we found that the severity of rainfall meteorological drought was generally higher than precipitation meteorological drought. The main range of rainfall meteorological drought severity was between 5.0 and 6.5 in TWDN, TGIM, and THSN. The main range of meteorological drought severity in terms of precipitation was between 3.2 and 3.9 in TWDN, TGIM, and THSN. The main range of meteorological drought severity in terms of rainfall was between 4.5 and 5.0 in WHSN. The main range of meteorological drought severity in terms of precipitation was between 3.6 and 3.8 in WHSN. Meteorological drought in terms of rainfall and precipitation showed a similar severity range, between 3.6 and 4.3. Meteorological drought in terms of rainfall (main range: 6.4–7.3) showed evidently higher severity than precipitation (main range: 3.1–3.6) in QTP. From Figure 8(b), we also found that the duration of rainfall meteorological drought was generally higher than that of precipitation meteorological drought in TWDN, TGIM, THSN, and QTP. Meteorological drought in terms of precipitation showed a similar duration (main range: 2.4–2.8 months) in different climate sub-regions. Meteorological drought in terms of rainfall presented a similar duration (main range: 3.2–4.6 months) in TWDN, TGIM, and THSN. Meanwhile, meteorological drought in terms of rainfall also presented a similar duration (main range: 2.4–3.0 months) in WHSN, SHCS, and THSC. Meteorological drought in terms of rainfall evidently showed a high duration (main range: 5.0–6.8 months) in QTP. From Figure 8(c) and 8(d), we found that hydrological drought exhibited relatively high and similar severity (main range: 4.0–7.0) and duration (main range: 2.5–4.2 months) in THSN and QTP. Hydrological drought exhibited relatively low and similar severity (main range: 3.0–4.1) and duration (main range: 2.1–3.0 months) in TWDN and TGIM. Hydrological drought in WHSN, SHCS, and THSC showed similar severity (main range: 3.5–5.1) and duration (main range: 2.5–3.5 months). In general, meteorological drought in terms of rainfall showed statistically significant higher severity and duration than precipitation in TWDN, TGIM, THSN, and QTP. The severity and duration of meteorological drought in terms of rainfall were similar to those in terms of precipitation. Hydrological drought exhibited relatively high severity and duration in THSN and QTP.

The drought propagation from meteorological to hydrological might be impacted by snowfall in a region. Based on Section 3.3, we found that SPIr–SRI showed a higher propagation degree than SPIp–SRI in the snowfall area. Moreover, the propagation time of SPIs–SRI was lower than SPIr–SRI and SPIp–SRI across the snowfall area. Freezing–thawing of permafrost and alpine snowmelt might extend the propagation time from meteorological drought to hydrological drought across the snowfall areas (Wang et al. 2023; Yang, J. et al. 2019, Yang, P. et al. 2019). From Figure 9, we found that the propagation time of SPIs–SRI generally presented 3–4 months longer than SPIp–SRI in the snowfall areas. Therefore, winter snowfall might contribute to spring runoff in the snowfall areas (Wu, Y. et al. 2018). Then, snowfall in winter might relieve hydrological drought in spring across the snowfall areas. However, we also revealed that the propagation degree of SPIs–SRI was relatively weak, as shown in Figure 9(c). Freezing–thawing of permafrost and alpine might be a complex nonlinear hydrological process (Zhang, Q., Shen, Z. et al. 2023; Zhu et al. 2023). From Figure 9, we found that the propagation time and degree between SPIr and SRI and SPIp–SRI were similar in southern China. From Figures 3(a) and 5(a), we found that both rainfall and runoff showed a higher amount in southern China. Higher rainfall and runoff amount might weaken the impact of different forms of precipitation (rain and snow) on drought propagation from meteorological to hydrological. Traditional linear methods might not accurately analyze the drought propagation process in the snowfall areas. Chen & Sun (2022) found that temperature and winter snowfall would increase in the future across northern China. Then, the future snowmelt in spring and the drought propagation process might be more complex than historically in northern China.

The study aimed to investigate the impacts of different forms of precipitation on drought characteristics and propagation. However, the methods, data, and results of this study still have certain limitations. In terms of method, although the MPCC method remains a widely used approach for determining drought propagation times, it might have limitations when investigating nonlinear issues (Zhou et al. 2022). The nonlinear limitations of MPCC have been confirmed by many studies (Vicente-Serrano et al. 2013). However, the drought propagation research field still lacks reliable nonlinear methods to replace MPCC (Zhang et al. 2022). In terms of data methods, although some observation data were used in this study, it still lacks field experiments to reliably support the result. Analyzing the spatial differences of drought characteristics across different regions remains the primary method in the current drought research field (Yu et al. 2023; Zeng et al. 2023). The gridded dataset of different forms of precipitation (snow, sleet, and rain) provided data on different forms of precipitation in this study. However, the dataset covered the period from 1961 to 2016 and has not been updated to the most recent years. Although the results of this study were based on long-term observational data, the outdated data might impose certain limitations on the applicability of the findings. In terms of results, this study only revealed that the propagation time of SPIs–SRI was longer than that of SPIp–SRI; however, we were unable to explain the complex propagation mechanism of SPIs–SRI in snowfall regions. The complex propagation mechanisms in plateau snowfall regions might be an interesting research direction in the future.

In this study, we investigated and analyzed the impact of different forms of precipitation on drought event characteristics and drought propagation across different climate sub-regions of China. The SPI and the SRI were used to describe meteorological drought and hydrological drought, respectively. Run theory was applied to identify the drought events for each grid cell. MPCC was used to estimate the drought propagation time and degree. The MK test was employed to monitor the changed trend of snowfall, sleet, and rainfall. The main conclusions were as follows:

• (1) The rainfall showed higher annual amounts in southern China. The annual snowfall amounts were higher in QTP and THSN. Both rainfall and snowfall obviously showed an increasing trend in TWDN. Snowfall showed decreasing trends in QTP and the Lesser Khingan mountains.

• (2) Both SERB and PRB showed a higher runoff amount and an increasing trend. The runoff in HARB and CRB had a lower runoff amount. The runoff of SERB and HARB overall presented the largest decreasing and increasing trends, separately.

• (3) The meteorological drought events of rainfall obviously presented higher severity and duration than the precipitation in TWDN, TGIM, THSN, and QTP. The hydrological drought event exhibited relatively high severity and duration in THSN and QTP.

• (4) The propagation degree of SPIr–SRI was higher than SPIp–SRI in TGIM, THSN, and QTP. The propagation time of SPIs–SRI displayed was shorter than SPIr–SRI and SPIp–SRI across QTP, THSN, TGIM, and northern TWDN.

This study was funded by the study of the impact of climate change on runoff conditions in the upper reaches of the Luo River (Grant No. A2019002).

All relevant data are available from an online repository or repositories: https://data.tpdc.ac.cn/.

The authors declare there is no conflict.