The changing precipitation storm properties under future climate change

Precipitation is one of the most important drivers of regional hydrological processes, yet it is also much more uncertain compared to the other climate variables such as temperature in the future projections (Chen et al. 2013; Wang & Chen 2014). The spatiotemporal distributions of precipitation have been altered due to changes of climate and land use (Trenberth 2011; Pathirana et al. 2014; Sohoulande Djebou & Singh 2016). Precipitation variations are of concern to society as the societal impacts of changes in precipitation have become more apparent. The 201.9 mm hourly rainfall in Zhengzhou, China 2021, the unusually heavy downpours in the central-western Europe, 2021, and the continuous 3 months of heavy rain in Pakistan, 2022 had profound impacts on both human society and the natural environment. Therefore, it is critical to investigate the changing pattern of precipitation in the future for flood and water resources management studies.

Precipitation variability has been examined by various methods. These methods could be broadly divided into two categories. The first one is using different precipitation indices including percentile-based indices, absolute indices representing maximum or minimum values within a certain period, threshold indices, duration indices, etc. (Jiang et al. 2015; Niu et al. 2021). Based on the definition of precipitation indices, they could be calculated for a station or a gird at a certain period and then used for the analysis of the spatiotemporal characteristics of precipitation. The second one is applying traditional intensity-duration-frequency (IDF) analysis or depth-duration-frequency (DDF) analysis (Agilan & Umamahesh 2016; Lima et al. 2016; Liuzzo et al. 2017). These methods focus more on total precipitation over a certain period rather than the concept of integrated precipitation event which has its start, development, and ending processes. The precipitation storm (referred as a storm here and after) properties are closely related to floods, soil erosion, agriculture, ecological habitat, etc. (Walther et al. 2002; Angel et al. 2005; Zhang et al. 2012) As a result, more analysis should be conducted on storm properties in the changing conditions.

The precipitation variations receive more attention in the southeastern coast of China as this region is undergoing rapidly changing social dynamics, pressure from an expanding population, and a greater risk of flooding (Gao et al. 2017; Liu et al. 2021). The flood season is characterized by two periods: Meiyu flood season (MFS), during which the continuous rainfall event is mainly caused by the planetary-scale Meiyu front (Ding et al. 2007) and Typhoon flood season (TFS) with heavy rainstorms caused by a typhoon (Tian et al. 2012; Gao et al. 2017). High intensity rainstorms during TFS may lead to severe flooding problems in this region. The intensity of rainfall during MFS is relatively small, but the large accumulation of rainfall during a long duration could also contribute to severe floods (Volonté et al. 2021). As the rainfall events during MFS and TFS are caused by different climatic systems, they may exhibit different trends or variations in the future. It is necessary to investigate the evolution of precipitation events separately during these two periods. Tian et al. (2012) investigated the trends of extreme precipitation in both MFS and TFS and indicated that the increase in total precipitation is mainly due to the increased precipitation during TFS. However, to our knowledge, few studies are conducted to explore the precipitation variability from the perspective of storm events for MFS and TFS, and fewer studies are designed to analyze the future trends of storm properties based on the climate model outputs.

To address the knowledge gaps mentioned above, we investigate the storm properties during two flood seasons based on station precipitation records, we also explore the future trends of storm properties based on Coupled Model Inter-comparison Project 6 (CMIP6) model outputs. Mean storm duration, inter-storm period, storm intensity, and within-storm patterns are determined for both MFS and TFS in the study region, the statistical matrices of storm properties are compared to illustrate different responses of storm events to changing conditions. The thorough analyses of precipitation variability during the two flood seasons are expected to provide detailed information on precipitation evolution to forecasters and local water managers.

Storm properties are analyzed and compared in MFS and TFS using hourly station data. In this study, the MFS ranges from April 16th to July 15th and the TFS ranges from 16 July to 15 October. In this case, the length of MFS and TFS are identical so that the storm properties are comparable. Two stations (ningxi and changtan) were chosen in CRB (Figure 1) with sufficient coverage for a long period (1960–2019) and high-quality data (less than 1% accumulated precipitation data). The Shangyang station is not involved due to the high percentage of missing data. The accumulated data are equally distributed to each hour for the accumulation period (generally 3–8 h) as the calculation of storm properties requires continuous hourly data. The reconstruction will not change the calculation of storm duration, inter-storm period, and average storm intensity based on the definition of these storm properties. It only impacts the determination of the within-storm pattern. However, the accumulated data accounts for less than 1% of the total precipitation data and the calculation of within-storm pattern is conducted on multiple months for a long period. As a result, the reconstruction will have little impact on the determination of storm properties. The future changes in storm properties are determined using hourly precipitation projections from CMIP6 climate models (Eyring et al. 2016).

The maximum monthly storm duration appears in June (Figure 3) during which the precipitation is mainly dominated by the Meiyu system. The relatively longer storm duration is mainly a product of the quasi-stationary front. There are no storm events during this period which has a total amount of precipitation over 250 mm. However, it is worth noting that large storm events with higher average storm intensity also exits during MFS (26 storm event with total precipitation larger than 100 mm, Table 2) due to the quasi-stationary and slow-moving linear mesoscale convective systems (Wang et al. 2014). Storms in July and August are characterized by shorter duration and higher intensity as the storms during the hot summer are mainly produced by the local convective systems and typhoons. So, the total precipitation during TFS is larger than that of MFS in the study area while the average account of storm events and the average storm duration are all smaller. In other words, the larger precipitation amount during TFS is mainly due to higher intensity storm events, which could be further proved by the account of storm events of which the total precipitation is larger than 250 mm in Table 2. Mass curves of three levels from TFS storm events show a broader range than that of MFS, indicating that the within-storm pattern is more uncertain and unsteady during TFS under the impact of tropical cyclones with huge power and abundant moisture.

In TFS, the maximum increase of TP appears in SSP585 due to a higher average storm intensity, TP exhibits less increase in the other scenarios except for SSP245 for the future1 period. The decrease of TP during this period for SSP245 is accompanied by more moderate storm events with shorter duration. TP in SSP126 also presents an obvious increase during both future periods. However, the increases are caused by changes in different storm properties. During the future1 period, it is mainly due to more high frequency – long duration precipitation events, while during the future2 period, it is mainly induced by storms with higher average intensity. As a whole, the future storm events are characterized by higher average storm intensity and will increase the flood risk in CRB, especially during the SSP585 scenario in the future2 period due to the longer storm duration.

The mass curves determined from historic simulation match well with those determined from future simulations for most emission scenarios at both MFS and TFS (Figure 5), which means that the within storm pattern is relatively stable under future climate change. Some exceptions also exist. For example, the future (both future1 and future2) mass curve range is narrower than the historic mass curve indicating a more stable within storm pattern in the future. This change in the within storm pattern is mainly caused by the changes in 10 percentile curves, which means more intense precipitation at the early age of a storm. Similar changes can be seen in the MFS future1 period for SSP585 and TFS future (both future1 and future2) period for SSP370. On contrast, the mass curve range becomes wider in future2 period than in the historic and future1 period in TFS for SSP585. This is mainly caused by the changes in 90% curves which mean less intense precipitation at the late part of a storm. The within storm pattern becomes more uncertain in TFS for SSP585.

In this study, we examine the precipitation characteristics in terms of storm properties in two rainy seasons at CRB. We also investigate the evolution of these storm properties by comparing the statistical matrices from historical and future simulations of selected CMIP6 GCM outputs. Our results highlight the changing patterns of storm duration, inter-storm period, average storm intensity, and within storm pattern in both MFS and TFS, which confirm and extend findings from previous studies (Jiang et al. 2016; Guan et al. 2020; Chen et al. 2022; Wei et al. 2022).

Precipitation storms in TFS exhibit shorter duration and higher average storm intensity than those of MFS. In other words, the larger total precipitation in TFS is induced by the higher average intensity of storms under the impacts of the local convective heat transport effects during the hot summers and tropical cyclones. However, this does not mean the risk of floods lies mainly during the TFS. The accumulative precipitation in MFS could also pose a great threat to local reservoirs, especially during June, during which the long-term mean storm duration is the largest of the year.

Future projections of storm properties show uncertainties among different emission scenarios and different seasons. However, the total precipitation and the average storm intensity exhibit a relatively consistently increasing trend while the inter-storm period presents a decreasing trend, especially during MFS. This is probably due to the ability of current GCMs in simulating large-scale frontal systems during MFS rather than local impacts such as local convective heat transport and random impacts on typhoon routine (Vitart et al. 1997; Camargo et al. 2005; Jiang et al. 2013). The uncertainties mainly lie in the projected changes of storm duration, which is not consistent either in magnitude or direction. However, changes in storm durations are not significant compared to changes in the average storm intensity and total precipitation. The higher intensity is the main factor that contributes to the increase in total precipitation during both MFS and TFS. The increase in average storm intensity is accompanied by the increase in storm duration during TFS for SSP585, which may pose the greatest challenges to reservoir operation strategies for mitigating the over level floods. The uncertainty arises not only from the emission scenarios, but also from internal variability and various climatic models (Hawkins & Sutton 2011; Rupp et al. 2013; Deser et al. 2014). The uncertainty from internal variability is not conducted as the long-term mean storm properties over 30 years which should not be very sensitive to internal variability (Rupp et al. 2013). For model uncertainty, there are many precipitation outputs from CMIP6 climate models at monthly and daily temporal scales. However, precipitation projections at an hourly scale are relatively few. Currently, only four hourly precipitation products are available, which may be not enough for a comprehensive evaluation of the uncertainties arising from models. As a result, this study focuses on the uncertainty from different emission scenarios and the model uncertainty could be investigated in future research when more hourly precipitation products are available.

This work was supported by the Fundamental Research Funds for the Central Universities (Grant No. 2019B05014); the Belt and Road Special Foundation of the State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering (Grant No. 2020490211; Grant No. 2020490208). We wish to thank the Program for Climate Model Diagnosis & Intercomparison (PCMDI) Earth System Model Evaluation Project for provide the data used in this paper.

All relevant data are available from https://github.com/pengjiang-hhu/HYDROLOGY-D-22-00142

The authors declare there is no conflict.