Abstract
Extreme precipitation events, which have intensified with global warming, will have a serious influence on society. Based on the latest generation of CMIP6 climate models and high-resolution grid observation data, the quantile mapping method and Taylor diagrams were used to correct and evaluate the simulation values. Six extreme precipitation indices for the near (2021–2040), middle (2051–2070), and far (2081–2100) periods of the 21st century were analyzed under shared socioeconomic pathway scenarios (SSP2-4.5 and SSP5-8.5). The results show that most of the extreme precipitation indices will increase over the Minjiang River Basin in the future, and both extreme precipitation and persistent drought will increase at the same time, which is more likely to cause extreme drought and flood disasters. For seasonal variation, the total amount and intensity of extreme precipitation will increase fastest in autumn, while the frequency of extreme precipitation will increase most in summer. Multimodel median changes show a decrease in the return period of RX1day (the maximum 1-day precipitation). By the end of the 21st century, under SSP5-8.5, the extreme events expected once every 100 years under the current climate are expected to occur approximately every 18.6 years.
HIGHLIGHTS
Quantile mapping and Taylor diagram were used to correct and evaluate the CMIP6 models.
The future changes of Minjiang River Basin extreme precipitation under SSP2-4.5 and SSP5-8.5 scenarios were revealed by using six ETCCDI indices.
Future seasonal variations in extreme precipitation were discussed.
GEV distribution indicates that the frequency of precipitation extremes will increase in the future.
Graphical Abstract
INTRODUCTION
According to the Intergovernmental Panel on Climate Change Sixth Assessment Report (AR6), the global surface temperature in 2011–2020 was 1.09 °C (0.95–1.20 °C) higher than that in 1850–1900 (IPCC 2021). Global warming is believed to be linked to the recent increase in extreme precipitation events due to increased atmospheric water vapor and warmer air (Ziegler et al. 2003; Christensen & Christensen 2004). With the continuous emission of greenhouse gases, the intensity and frequency of extreme precipitation events have increased, and this trend is very likely to continue in the future (Easterling et al. 2000; Allen & Ingram 2002; Meehl et al. 2005). It should be noted that precipitation extremes have posed ever-intensifying threats to natural ecosystems (Peñuelas et al. 2017), water resources (Giuliani et al. 2016), human life (Rosenberg et al. 2010), agricultural production (He et al. 2018), hydroelectric power generation (Curtis et al. 2017), transportation, and urbanization (Rosenzweig et al. 2001). Therefore, the mitigation and adaptation of climate change, particularly extreme climate events, have received increasing attention from governmental agencies, scientific communities, and the general public (Zhai et al. 2005; Alexander et al. 2006; Donat et al. 2016).
Located in the East Asian monsoon region, China is characterized by a climate with large natural variability, and the region is vulnerable to climate extremes due to its relatively low adaptive capacity, high population density and mobility (Qin 2012). Extreme precipitation events have been the main cause of natural disasters in China since 1990, with an average annual economic loss of 139 billion Yuan and 984 deaths per year (He et al. 2018). In recent years, several scholars have discussed the changes in extreme precipitation events within the year or the season on a nationwide scale (Fu et al. 2013; Jiang et al. 2015; Chi et al. 2016; Li et al. 2018). In river basins, Su et al. (2008) showed that extreme precipitation in the Yangtze River Basin has increased significantly in recent decades, causing severe flooding. Fischer et al. (2011) found that rainfall intensity increased along the coastline and in the far west of the Zhujiang River Basin from 1961 to 2007. Lv et al. (2018) also found that the warming and drying climate is one key reason for the significant decrease in runoff in the Yellow River Basin. However, most research has been carried out at large spatial scales, mostly on the scale of large river basins, and there are few assessments of small and medium-sized watersheds in southeast China. The Minjiang River Basin (MRB) is located in the subtropical climate zone along the southeast coast of China, with frequent droughts and floods. The spatial and temporal distribution of precipitation over the MRB is affected by monsoons, ocean currents, tropical cyclones, and other factors, making it a highly sensitive area to climate change. Although the MRB has a small area of only 60,992 km2, it has abundant precipitation and an average annual runoff of 1,980 m3s−1, ranking 7th in China. Due to the large intraannual and interannual variability of precipitation in this region, the water cycle is further accelerated due to climate warming, and the spatial and temporal distribution of precipitation may become more uneven in the future. Therefore, it is necessary to conduct a comprehensive study on the change characteristics of MRB extreme precipitation events.
Global climate models (GCMs) are the main tools used by the scientific community to reproduce the current climate and project future changes in extreme precipitation events (IPCC 2007; Li et al. 2015). Currently, the underway CMIP6 consists of the largest number of participating models, the most complete scientific experiments designed, and the largest simulation data of the last few decades, which will support global climate research in the next 5–10 years (Eyring et al. 2016; Xin et al. 2020). CMIP6 developed a set of shared socioeconomic pathway scenarios (SSPs), which included future social and economic changes such as population, economic development, ecosystem, resources, institutions, and social factors (Chen et al. 2020). It also includes future mitigation, adaptation, and response efforts to climate change. Data from CMIP6 are still being released, but relevant research works have been carried out extensively (Akinsanola et al. 2020; Li et al. 2020; Fu et al. 2021). Due to the difference in the spatial and temporal distribution of extreme precipitation (Jiang et al. 2020; Zhu et al. 2020), the results of future projections for different river basins may be very different, so it is necessary to use the CMIP6 model specifically to discuss the future prediction of MRB. The research results will be helpful to understand the climate change and water resource distribution in this region and have important reference significance for ecological resource protection, flood and drought disaster management, etc.
The remainder of the paper is organized as follows: Section 2 introduces the study area and its climatic characteristics. Section 3 describes the data, models, and methods. Section 4 introduces the bias correction and future prediction results of CMIP6 models. Finally, conclusions and some further discussions are provided in Section 5.
STUDY AREA
DATA AND METHODS
Data
Observation data
The observed data in this paper are CN05.1 daily precipitation grid data provided by The National Meteorological Information Center from 1971 to 2010, with a spatial resolution of 0.25° × 0.25°. CN05.1 is a set of high-resolution grid data set established through quality control and interpolation based on the observation data of more than 2,400 Chinese surface meteorological stations, which has been widely used in climate change research (Dong et al. 2015; Zhu et al. 2020).
CMIP6 models
In this paper, daily precipitation data from GCMs were used to assess the impact of climate change in the MRB under historical and future scenarios. The GCMs outputs from CMIP6 were obtained from the Earth System Grid Federation (ESGF) via https://esgf-node.llnl.gov/search/cmip6/. According to the results of Zhu et al. (2021), six CMIP6 models with good precipitation simulation in China were selected, and SSP2-4.5 and SSP5-8.5 scenarios were used for future prediction. Table 1 gives a detailed description of these models, including the model number, model acronym, affiliated institution, and horizontal resolution. SSP2-4.5 is the updated scenario of RCP4.5 in CMIP5, representing the combination of moderate social vulnerability and moderate radiative forcing. SSP5-8.5, updated from the RCP8.5 scenario, is the only shared socioeconomic path that can achieve an anthropogenic radiative forcing of 8.5 W m−2 by 2100 (Eyring et al. 2016). Only the first realization, first initialization, and the first physics (r1i1p1f1) for each model were chosen, although all models provide certain ensemble members. The research period of the 21st century is divided into the near future (2021–2040), mid future (2051–2070), and far future (2081–2100). The historical reference period is 1991–2010, when extreme events increased significantly. As the multimodel ensemble (MME) can reduce the error of a single model (Sillmann et al. 2013), the MME method is employed in this work. Considering the different resolutions of the observed data and model, all model data were uniformly bilinear interpolated to the same latitude and longitude grid with the same resolution of observed data (0.25° × 0.25°) to facilitate the ensemble averaging of all models (Mishra et al. 2020).
List of CMIP6 models used in this study
Model number . | Model Acronym . | Institute/Country . | Horizon resolution (lon × lat) . | Reference . |
---|---|---|---|---|
1 | BCC-CSM2-MR | Beijing Climate Center, China Meteorological Administration (BCC)/China | 1.125° × 1.125° | Wu et al. (2019) |
2 | EC-Earth3-Veg | EC-EARTH consortium | 0.7° × 0.7° | Döscher et al. (2022) |
3 | GFDL-ESM4 | National Oceanic and Atmospheric Administration (NOAA), Geophysical Fluid Dynamics Laboratory (GFDL)/USA | 1.0° × 1.25° | Dunne et al. (2020) |
4 | IPSL-CM6A-LR | L'Institut Pierre-Simon Laplace (IPSL)/France | 1.26° × 2.5° | Boucher et al. (2020) |
5 | MRI-ESM2-0 | Meteorological Research Institute (MRI)/Japan | 1.125° × 1.125° | Yukimoto et al. (2019) |
6 | NorESM2-MM | Norwegian Climate Centre (NCC)/Norway | 0.94° × 1.25° | Seland et al. (2020) |
Model number . | Model Acronym . | Institute/Country . | Horizon resolution (lon × lat) . | Reference . |
---|---|---|---|---|
1 | BCC-CSM2-MR | Beijing Climate Center, China Meteorological Administration (BCC)/China | 1.125° × 1.125° | Wu et al. (2019) |
2 | EC-Earth3-Veg | EC-EARTH consortium | 0.7° × 0.7° | Döscher et al. (2022) |
3 | GFDL-ESM4 | National Oceanic and Atmospheric Administration (NOAA), Geophysical Fluid Dynamics Laboratory (GFDL)/USA | 1.0° × 1.25° | Dunne et al. (2020) |
4 | IPSL-CM6A-LR | L'Institut Pierre-Simon Laplace (IPSL)/France | 1.26° × 2.5° | Boucher et al. (2020) |
5 | MRI-ESM2-0 | Meteorological Research Institute (MRI)/Japan | 1.125° × 1.125° | Yukimoto et al. (2019) |
6 | NorESM2-MM | Norwegian Climate Centre (NCC)/Norway | 0.94° × 1.25° | Seland et al. (2020) |
Extreme precipitation event indices
Due to the limitation of global climate long-series data and the lack of a unified definition of extreme climate event indicators in various countries, the research and development of extreme climate events is hindered to some extent (Alexander et al. 2006). To change this situation, the World Meteorological Organization (WMO) and the World Climate Research Programme (WCRP) established the Expert Group on Climate Change Detection and Indicators (ETCCDI, http://etccdi.pacific.climate.org/indices.shtml) and proposed 27 extreme climate indices for quantifying extreme temperature and precipitation events, which have been widely used in global extreme weather and climate research (Yin & Sun 2018). This study adopted the extreme climate index promoted by ETCCDI and selected six extreme precipitation indices, as shown in Table 2. These were total precipitation on wet days (Prcptot), consecutive dry days (CDD), heavy precipitation days (R10mm), very heavy precipitation days (R20mm), maximum 1-day precipitation (RX1day), and simple precipitation intensity index (SDII).
Extreme precipitation event indices considered in this study
Indicator name . | Abbreviation . | Definitions . | Units . |
---|---|---|---|
Total precipitation | Prcptot | Total precipitation in wet days with precipitation ≥ 1 mm | mm |
Consecutive dry days | CDD | Maximum number of consecutive days with precipitation <1 mm | d |
Heavy precipitation days | R10mm | Days with precipitation ≥ 10 mm | d |
Very heavy precipitation days | R20mm | Days with precipitation ≥ 20 mm | d |
Maximum 1-day precipitation | RX1day | The maximum 1-day value for period | mm |
Simple precipitation intensity index | SDII | Total wet days precipitation divided by the number of wet days | mm d−1 |
Indicator name . | Abbreviation . | Definitions . | Units . |
---|---|---|---|
Total precipitation | Prcptot | Total precipitation in wet days with precipitation ≥ 1 mm | mm |
Consecutive dry days | CDD | Maximum number of consecutive days with precipitation <1 mm | d |
Heavy precipitation days | R10mm | Days with precipitation ≥ 10 mm | d |
Very heavy precipitation days | R20mm | Days with precipitation ≥ 20 mm | d |
Maximum 1-day precipitation | RX1day | The maximum 1-day value for period | mm |
Simple precipitation intensity index | SDII | Total wet days precipitation divided by the number of wet days | mm d−1 |
Methods
Bias correction
The bias correction method is an important technology for reducing simulation biases (Feddersen et al. 1999). The quantile mapping method (QM), a probability distribution-based correction method, is used to correct the deviation of CMIP6 models in this paper. Previous studies have shown that this correction method can effectively reduce the deviation of model simulation, especially for the tail of the probability density distribution (Cannon et al. 2015; Tang et al. 2022).
This transfer function will remain valid in the future, and the bias-corrected simulation can be calculated in the future period. Here, 1971–1990 is selected as the original training period, and the historical reference period (1991–2010) is corrected as the test period. Then, the modeling is applied to the daily precipitation data of the next three periods (2021–2040, 2051–2070, and 2081–2100).
Taylor diagram
To evaluate the overall skill in reproducing the spatial pattern of the present-day climate indices, the Taylor Diagram is used (Taylor 2001) in this paper. It statistically summarizes how well the historical model outputs match the observed data by means of their corresponding spatial correlation coefficient (CC), centered root mean square error (RMSE), and the ratio of spatial standard deviation (RSD). There is a mathematical transformation relationship among the three variables, so the difference in the simulation ability of different models can be compared more intuitively by putting them on the same graph. A perfect simulation would be one with an RMSE equal to 0 and both the CC and RSD close to 1 (Li et al. 2015).
Generalized extreme value distribution
Return periods such as the ‘T-year (100-year) event’ describe an event that has, on average, a 1-in-T (1% for the 100-year event) chance of occurring or being exceeded in any given year at a given location (Slater et al. 2021). It is commonly used for a wide range of applications in engineering planning, for instance, in the decision of hydrologic design, water management structure, dams, and bridges (Li et al. 2018). To estimate the return value, the annual maximum (AM) value method is used to select sample sequences from daily data at every grid point to be fitted by the generalized extreme value (GEV) distribution. According to classical extreme value theory, the approximate distribution of the maximum value in each period should belong to one of the three probability distributions of Gumbel, Frechet, and Weibull, while the GEV distribution is a unified description of these three forms, which can avoid the disadvantage of adopting a single distribution. GEV distribution was found to be suitable as a fit to the tails of the distribution for precipitation (Wang et al. 2013).
RESULTS
Model evaluation and biases correction
Bias in the simulation of extreme precipitation event indices from the observations after bias correction during the reference period: (a) Prcptot, (b) R10mm, (c) R20mm, (d) Rx1day, (e) SDII, and (f) CDD. The ‘ + ’ indicates that all the simulations simulated the negative/positive bias.
Bias in the simulation of extreme precipitation event indices from the observations after bias correction during the reference period: (a) Prcptot, (b) R10mm, (c) R20mm, (d) Rx1day, (e) SDII, and (f) CDD. The ‘ + ’ indicates that all the simulations simulated the negative/positive bias.
Taylor diagrams showing the performance of models in simulating climatological fields over Fujian province in extreme temperature events for six indices: (a) Hollow dots indicate no bias correction, while solid dots indicate bias correction, (b) (numbers 1–6 correspond to the model number in Table 1).
Taylor diagrams showing the performance of models in simulating climatological fields over Fujian province in extreme temperature events for six indices: (a) Hollow dots indicate no bias correction, while solid dots indicate bias correction, (b) (numbers 1–6 correspond to the model number in Table 1).
Spatial variation in precipitation extremes
Prcptot and CDD
Spatial distribution of variation of annual Prcptot in the near future (the left column), mid future (the middle column), and far future (the right column) relative to historical reference period in MRB under two scenarios (Black solid dots indicate the statistical significance of precipitation difference that exceeds the 95% confidence level).
Spatial distribution of variation of annual Prcptot in the near future (the left column), mid future (the middle column), and far future (the right column) relative to historical reference period in MRB under two scenarios (Black solid dots indicate the statistical significance of precipitation difference that exceeds the 95% confidence level).
R10mm and R20mm
Spatial distribution of variation of annual Prcptot in the near future (blue), mid future (green), and far future (red) relative to historical reference period in MRB under SSP5-8.5 scenario: (a) Prcptot, (b) CDD, (c) R10mm, (d) R20mm, (e) Rx1day, and (f) SDII. Bars represent the multimodel mean, the upper and lower limits represent the maximum and minimum values of multiple models. Please refer to the online version of this paper to see this figure in colour: https://dx.doi.org/10.2166/wcc.2022.145.
Spatial distribution of variation of annual Prcptot in the near future (blue), mid future (green), and far future (red) relative to historical reference period in MRB under SSP5-8.5 scenario: (a) Prcptot, (b) CDD, (c) R10mm, (d) R20mm, (e) Rx1day, and (f) SDII. Bars represent the multimodel mean, the upper and lower limits represent the maximum and minimum values of multiple models. Please refer to the online version of this paper to see this figure in colour: https://dx.doi.org/10.2166/wcc.2022.145.
Rx1day and SDII
Season projection
In this section, the ensemble mean of the projected changes under the SSP5-8.5 scenario are discussed due to the relevance of strong seasonality over this region. It defines spring as March to May, summer as June to August, autumn as September to November, and winter as December to February. Prcptot will increase in all seasons in the future, with the most significant increase in autumn (approximately 70–140%). The CDD will increase significantly in autumn and winter but decrease slightly in summer. This indicates that the variation between seasons increases, especially in the dry season, with the increase in Prcptot, drought increases greatly at the same time, and the possibility of extreme drought and flood events increases greatly. In the future, R10mm and R20mm days will increase in most seasons, with the largest increase in summer, indicating an increase in the frequency of extreme precipitation in summer. At the same time, it can also be seen that in near future spring and middle future autumn, R10mm and R20mm will decrease, suggesting that seasonal differences exist in a specific period of time. In the 21st century, precipitation intensity will increase significantly in all seasons, with the largest increase in autumn. Rx1day will increase by an average of 10–90% in each season, while SDII will increase by approximately 5–40%. Overall, these changes suggest that the total amount and intensity of extreme precipitation will increase fastest in autumn, while the frequency of extreme precipitation will increase most in summer. The MRB may experience a longer dry spell duration in the future, but extreme precipitation may be more frequent and more intensified when it occurs. Similarly, Supari et al. (2020) found that it will also occur in other parts of Asia, suggesting that extreme drought-flood events require wider attention in the future.
Changes in the return period
The return period of Rx1day in different periods in the 21st century under 20-years (a), 100-years (b) return period in history under SSP2-4.5 and SSP5-8.5 scenarios. The upper and lower limits of the box indicate the 75th and 25th percentile values, respectively.
The return period of Rx1day in different periods in the 21st century under 20-years (a), 100-years (b) return period in history under SSP2-4.5 and SSP5-8.5 scenarios. The upper and lower limits of the box indicate the 75th and 25th percentile values, respectively.
CONCLUSIONS AND DISCUSSION
Using six climate models in the CMIP6 archive and ETCCDI indices, this study reveals the future changes in extreme precipitation over the MRB, as well as the return period under the SSP2-4.5 and SSP5-8.5 scenarios. The primary conclusions are as follows:
The CMIP6 climate model results are better than those of the original model after QM bias correction. On the Taylor diagram, the CC of most extreme precipitation indices increased from 0–0.6 to 0.5–0.9, the RMSE ranged from 0 to 0.5, and the RSD ranged from 0.7 to 1.5.
From the perspective of the annual average, most ETCCDI indices will grow in the future. Prcptot will increase in all periods in the future, and CDD will also increase in most cases. R10mm and R20mm have complex spatial changes at different times, and they will increase significantly during 2081–2100 under the SSP5-8.5 scenario. Rx1day and SDII, which represent precipitation intensity, will increase in the future throughout the 21st century. Overall, extreme precipitation and persistent drought will increase simultaneously over the MBR in the 21st century, and extreme climate and instability will increase, which is more likely to induce extreme drought and flood events. For seasonal variation, Prcptot will increase in all seasons in the future, with the most significant increase in autumn. The CDD will increase significantly in autumn and winter but decrease slightly in summer. R10mm and R20mm days will increase in most seasons, with the largest increase in summer. Rx1day will increase by 10–90% in each season, while SDII will increase by approximately 5–40%. These changes suggest that the MRB may experience a longer dry spell duration in the future, but extreme precipitation may be more frequent and more intensified when it occurs.
By GEV distribution analysis, the frequency of extreme events will increase in the future. The events that are expected to occur every 20 years in the current climate are expected to occur approximately every 11–12 years (near future), 7–10 years (middle future), and 4–5 years (far future). Extreme events expected to occur once every 100 years under the current climate are expected to occur approximately every 50–55 years (near future), 36–42 years (middle future), and 18–24 years (far future).
Admittedly, the CMIP6 models’ ability to simulate extreme precipitation indices is lower than their ability to simulate extreme temperature indices (Zhu et al. 2020). Compared with CMIP5, which was very persistent and widespread, the drying biases in CMIP6 in South China were significantly reduced but still existed. The system bias of GCMs and their ability to simulate precipitation extremes is closely related to the models’ ability to simulate atmospheric circulation, such as the Northwest Pacific subtropical high (NPSH) and southwesterly winds (Jiang et al. 2015). At the same time, the simulation ability of the model to the monsoon region is relatively poor, and the intermodel uncertainties of the projected precipitation extremes are still large in those regions where the East Asian summer monsoon (EASM) prevails (Xu et al. 2015). The MBR is located in the subtropical monsoon climate zone along the southeast coast of China, which is inevitably affected by model uncertainty. Future studies should also include more targeted analyses to better understand the physical processes underlying both the model biases and the changes in the regional distribution of extremes over the MRB. Moreover, the simulation bias between GCMs and observations is also attributed to the coarse resolution of GCMs. The NEX-GDDP-CMIP6 downscaling projection dataset provided by NASA is a good option (Thrasher et al. 2022). In addition, dynamic downscaling and regional climate models (RCMs) have better resolution and thus better capture extreme climate features at the regional scale (Dong et al. 2020). We will also adopt more CMIP6 models and apply different set schemes and preferred ordering to achieve a reasonable set average of the model results to reduce the uncertainty in future work.
DATA AVAILABILITY STATEMENT
All relevant data are available from an online repository or repositories: CN05.1: https://ccrc.iap.ac.cn/resource; CMIP6: https://esgf-node.llnl.gov/search/cmip6/.
CONFLICT OF INTEREST
The authors declare there is no conflict.