1. Introduction
Extreme precipitation events are of considerable interest because they can have major impacts on the environment, society, and economy (Sugiyama et al. 2010). According to the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC), the frequency, intensity, spatial extent, duration, and timing of weather and climate extremes are significantly influenced by climate change (Jiang et al. 2015). Severe and extreme precipitation events are more prone to occur with the warming climate (Min et al. 2011; Rajczak et al. 2013; Stott 2016), and thus the potential changes in extreme rainfall events with climate change have become increasingly important (Hallegatte et al. 2013).
Currently, climate models are the primary tools available for making projections of future mean climate and extreme events (O’Gorman and Schneider 2009; Sillmann et al. 2013; Madsen et al. 2014; Jiang et al. 2015; Bao et al. 2017). To be confident in the projected future change in extreme precipitation events, it is essential to understand the model’s capability in reproducing historical extreme rainfall events. This capability has been examined by numerous studies, and it was shown that large uncertainty still exists in the state-of-the-art general circulation models (GCMs; Stephens et al. 2010; Min et al. 2011; Sillmann et al. 2017). Xu et al. (2011) examined the performance of models from the World Climate Research Programme’s (WCRP) phase 3 of the Coupled Model Intercomparison Project (CMIP3) and showed that these models can reproduce the observed spatial distribution of extreme precipitation in the last half of the twentieth century, but the observed interannual variations in extreme precipitation events are not well simulated. By analyzing archives from phase 5 of the CMIP (CMIP5), Rosa and Collins (2013) showed that GCMs underestimate the incidence of heavy rainfall events but overestimate the persistence of heavy precipitation. Jiang et al. (2015) quantitatively assessed the CMIP5 models in simulating precipitation extremes and showed that the performance of the models is quite different between western and eastern China, with wet biases in the west and dry biases in the east. Kopparla et al. (2013) compared the simulations of daily extreme precipitation events between high-resolution CAM4 (~0.25°) and the same model at lower resolutions (~1° and 2°). They concluded that extreme precipitation is more accurately simulated at high resolution, but considerable biases remain in the high-resolution model.
Previous studies focusing on the model performance for extreme events generally evaluate the quantitative statistical features of extreme precipitation based on long-term model outputs. Since the rainfall intensity is usually underestimated in climate models (Dai et al. 2007; Li et al. 2015), the evaluation based on relative thresholds can hardly represent the extreme event in nature. The severity of the extreme events in the model simulation has not been fully revealed. Moreover, although these evaluations have provided a straightforward synthesis of model performance (Gleckler et al. 2008; Reichler and Kim 2008), details of the individual rainfall processes and the related synoptic circulation patterns in the model are masked by the statistical average and aggregation. A detailed investigation of the hourly evolution of extreme precipitation events could improve the understanding of how these events are produced in the model and whether they are valid for the right reason, which is important in understanding the reliability of climate predictions performed with the model (Trenberth et al. 2003). Therefore, this study aims to simulate a typical extreme precipitation event by a GCM and evaluate the model’s ability and limitations in reproducing the severity and process of the event.
In the midsummer of 2016, a particularly extreme rainfall event hit the Yangtze River valley (YRV) between 30 June and 6 July, with six national rain gauge stations observing a record-breaking daily rainfall amount over 250 mm. The extreme event was characterized by a long-lasting heavy rain belt with a widely covered area that led to a dramatic increase in flooding along the YRV (Bi et al. 2017). The affected areas experienced the country’s worst economic losses since 1998, making it a high-impact weather event of international significance. The long-lasting heavy rainfall was closely related to strong synoptic forcings, which provides an opportunity to examine the model’s performance in reproducing the hourly distribution of rainfall processes and its related synoptic circulations. In this study, the extreme precipitation over the YRV is simulated using a climate model. The model is integrated as a weather prediction model using the Transpose Atmospheric Model Intercomparison Project (T-AMIP)-type integration (Williams et al. 2013). This allows the examination of the precipitation process and related synoptic systems of the specific event with a relatively realistic initialization (Phillips et al. 2004). By analyzing the simulation of this extreme event, this study aims to discuss the possible causes of the model’s capabilities and deficiencies.
The remainder of the paper is organized as follows: Section 2 describes the model and the experiments and provides a brief description of the data and methods used in the study. The AMIP results are evaluated in section 3 to verify the model’s ability to simulate the climatology of precipitation over China. Section 4 shows the simulation results for the extreme precipitation event. The summary and discussion are provided in section 5.
2. Data, methods, model, and experiments
a. Data and methods
A dataset of hourly rain gauge records at 2240 stations covering mainland China during 2001–15 is used in this study. This dataset was obtained from the National Meteorological Information Center of the China Meteorological Administration and has undergone strict quality control. The quality control includes extreme value checks, internal consistency checks (whether the relationship between the different elements conforms to an objective law), and time consistency checks (whether the variation in the elements conforms to an objective law). All these stations have more than 60% of continuous hourly records during the 15 warm seasons (April–October). The locations of the 2240 stations are shown in Fig. 1, and there are 393 stations in the central region of the extreme event marked by the red rectangle (27°–33°N, 108°–120°E). For a better comparison between gauge observations and the model simulation, the hourly rain gauge records are interpolated to the model grid using the patch rendezvous method, which takes the least squares fit of the surrounding surface patches (Khoei and Gharehbaghi 2007) in the Earth System Modeling Framework (ESMF) function (http://www.earthsystemmodeling.org/esmf_releases/public/ESMF_5_2_0rp1/esmf_5_2_0rp1_regridding_status.html).
The spatial distribution of the rain gauge stations (black dots) used in this paper. The topography is shown by color shading (m). The central region of the extreme event (27°–33°N, 108°–120°E) is marked by the red rectangle. The Yangtze River and Yellow River are drawn as blue lines.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
In this study, the climatological precipitation amount is calculated by dividing the cumulative precipitation amount during the study period (April–October 2001–15) by the total nonmissing hours and then multiplying by 24 h for normalization to the daily amount (mm day−1). Precipitation frequency is defined as the percentage of the total hours with measurable precipitation (≥0.1 mm h−1) versus the total nonmissing hours. Precipitation intensity is obtained by dividing the cumulative precipitation amount by the number of rainy hours (the hours with precipitation ≥ 0.1 mm h−1).
To investigate the information on related cloud features during the extreme event, the brightness temperature (BT) from infrared channel 1 (IR1) of Fengyun-2G (FY-2G), the operational geostationary meteorological satellite of China, is used. The data cover a full disc image covering Earth’s surface from 60°N to 60°S in latitude and from 45° to 165°E in longitude with high spatial (5 km) and temporal (30 min) resolution (Liu et al. 2009; Guo et al. 2016), which is available online (http://satellite.nsmc.org.cn/PortalSite/Default.aspx). A BT calculation scheme developed by the Australian Bureau of Meteorology (Rikus 1997; Sun and Rikus 2004) is applied to calculate the model BT. The simulated narrowband BT corresponding to the IR1 channel of the FY-2G satellite is generated offline using a radiative transfer scheme, which makes the observed and simulated BTs quantitatively comparable.
The ERA-Interim reanalysis data with 6-h intervals and 0.75° horizontal resolution are used to analyze and track the location of the southwest vortex influencing the extreme precipitation event along the YRV. In addition, to obtain the hourly circulation differences between the model and observations, the ERA5 climate reanalysis dataset, which was developed through the Copernicus Climate Change Service (C3S), was also used in this study. ERA5 is the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalyses of the global climate and provides hourly analysis fields at a horizontal resolution of 31 km (http://climate.copernicus.eu/products/climate-reanalysis).
Two kinds of topography data are used in this study. One was derived from the Global 30 arc s elevation dataset (GTOPO30) with a horizontal resolution of 0.083° (Fig. 1). The other is the topography used in the model at T255 resolution (Fig. 10a).
b. Model description
The model used in this work is the atmospheric component of the Climate System Model of the Chinese Academy of Meteorological Sciences (CAMS-CSM). This model is a modified version of the ECHAM5 atmospheric model (Roechner et al. 2003) developed by the Max Planck Institute for Meteorology. It is a global primitive equation model and has 31 vertical levels with a top at 10 hPa. The model employs a spectral dynamical core and utilizes a semi-implicit leapfrog time-differencing scheme. A hybrid sigma-pressure system is used in the vertical direction. Several modifications have been made at CAMS to improve the performance of ECHAM5, especially over the East Asian region. A full land surface model, Common Land Model (CoLM; Dai et al. 2003), was implemented in the model. A two-step shape-preserving advection scheme proposed by Yu (1994) was ported into the dynamical core to replace the original flux-form semi-Lagrangian scheme. This advection scheme has been proven to significantly reduce the overestimation of precipitation over the steep edges of the southern Tibetan Plateau (Yu et al. 2015). A new radiation scheme developed by the Beijing Climate Center (Zhang et al. 2012) was implemented to improve the simulation of shortwave cloud radiative forcing over East Asia. For an additional detailed description of the ECHAM5 model, please refer to Roechner et al. (2003).
c. Experimental design
Two types of experiments were conducted in this work (see Table 1). The first type is an AMIP-type integration (Gates et al. 1999), which aims to examine the model’s performance in simulating the long-term mean state. The model was integrated at a horizontal resolution of T255 (approximately 0.47°) with observed monthly sea surface temperature (SST) and sea ice cover for the period 2000–15. The integration was started with the land states from a separate 20-yr prespinup run with a land model coupled to a standalone atmospheric model. The data from 2001 to 2015 were analyzed in this study.
Experimental design.
To evaluate the model’s ability to reproduce the extreme precipitation event, a T-AMIP-type integration (Williams et al. 2013) was also performed at the same resolution. T-AMIP experiments are designed to run weather forecasts using climate models, which enables a detailed evaluation of the processes involved in meteorological events. The T-AMIP experiments were performed using the CAMS-CSM model from 29 June to 5 July 2016 when the extreme rainfall event occurred along the YRV. The global climate model was started at 1200 UTC on each day and was integrated for five days, that is, seven global hindcasts were produced during the period of the extreme event. The model state variables (zonal and meridional wind, temperature, and humidity) were initialized from ERA-Interim reanalysis data from the ECMWF (Dee et al. 2011), while other nonstate variables that spin up quickly were initialized from zero. Daily SST data from NOAA OISST (Banzon et al. 2016) were used in all hindcasts. The land surface model was initialized from the final state of 29 June after a 3-yr continuous integration. The initial land condition was fixed for all seven hindcasts. The output of the first 12 h was discarded. The results of 13–36 h of each hindcast (from 0000 UTC to 0000 UTC of the next day) are referred to as day 1 of the hindcast in this work. In addition, the day-1 hindcasts are used in the following analysis to evaluate the model’s performance in reproducing the extreme event.
3. Simulation of the climatic rainfall characteristics by CAMS-CSM
Since the performance of the CAMS-CSM model has not been fully evaluated, the 15-yr AMIP results are first validated to assess the model’s capability of reproducing the climatology of precipitation over East Asia. Figure 2 compares the distribution of the observed and simulated warm season (from April to October) mean precipitation amount, frequency, and intensity. The comparison reveals that the model adequately reproduces the general pattern and magnitude of the precipitation amount (Fig. 2b), despite the underestimation over the southeastern coastal region. The major precipitation zone along the YRV is reasonably simulated. As shown in Table 2, the simulated precipitation amount averaged over 27°–33°N, 108°–120°E is 4.60 mm day−1, which is very close to the observed value (4.76 mm day−1). However, there is an overall overestimation of the hourly precipitation frequency in the model results, especially over the Tibetan Plateau and other mountainous regions (Fig. 2d). Correspondingly, the hourly precipitation intensity is underestimated over most of the grids shown in Fig. 2f. This bias in the frequency–intensity structure is a common issue in most climate models and is closely related to the model resolution (Li et al. 2015) and the model physics (Zhang and Chen 2016). Over the YRV, the simulated frequency (intensity) is 48.7% (34.4%) higher (lower) than that of the observations, indicating a large portion of unrealistic weak precipitation. Nonetheless, the spatial distribution is comparable between the simulated and observed fields, and the pattern correlation for frequency and intensity is 0.60 and 0.77, respectively.
Spatial distribution of the 2001–15 warm season (April–October) mean precipitation (a),(b) amount (mm day−1), (c),(d) frequency (%), and (e),(f) intensity (mm h−1) from (left) station observations and (right) the CAMS-CSM AMIP simulation. The Yangtze River and Yellow River are drawn as dark gray lines. The central region of the extreme event (27°–33°N, 108°–120°E) is marked by the black rectangle in (a).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
Comparison results between the observations and AMIP simulation for the features of precipitation averaged over 27°–33°N, 108°–120°E.
Figure 3 shows the distribution of the 90th percentile values of hourly intensity, which have been widely used as thresholds of extreme precipitation. The patterns of the large-value zones over northern China, the YRV, and southern China are reproduced, although the values are significantly lower than the observed values. The simulated 90th percentile average value over the YRV (27°–33°N, 108°–120°E) is 3.07 mm h−1, which is 33.4% lower than the observed value (4.61 mm h−1). As the percentile level increases, the relative bias also increases from −41.0% (95th) to −49.9% (99th) (Table 2), indicating that the extreme rainfall events in the model are still much weaker than the observed events.
As in Fig. 2, but for the 90th percentile values of hourly precipitation (mm h−1) in the (a) gauge observations and (b) AMIP simulation.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
As a key metric of the climate model evaluation, the diurnal cycle of precipitation is sensitive to many relevant physical processes and is closely related to the mechanisms of precipitation. Over contiguous China, the diurnal cycle of precipitation has distinct modes of peak time and considerable regional features, which makes this region a versatile test bed for the evaluation of climate models (Yu et al. 2007). Compared with the observed diurnal peaks of the precipitation amount (Fig. 4a), the model can capture certain major features of the diurnal patterns (Fig. 4b). The afternoon peaks over northeastern China and southern inland China are reproduced, despite the bias of earlier occurrences. Focusing on the YRV, the model simulates the midnight peaks over the eastern edge of the Tibetan Plateau, the late-night peaks of approximately 105°–110°E, and the early morning peaks eastward. The simulation of the coherent eastward transition of diurnal phases, which is a key feature of precipitation along the YRV (Chen et al. 2010a), adds confidence to the validity of the model performance.
As in Fig. 3, but for the diurnal peaks of precipitation amount (local time).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
The above analyses of the AMIP results lead to the conclusion that the model can reasonably simulate the climatological distribution and diurnal variation in precipitation over the YRV and surrounding regions, although there is considerable bias in the hourly frequency–intensity structure.
4. Simulation of the extreme rainfall event via the T-AMIP experiments
The observed and simulated accumulated precipitation amount during the extreme event (from 30 June to 5 July 2016) are presented and compared in Fig. 5. There is a belt of heavy precipitation in the observations along the YRV and to the east of 108°E (Fig. 5a). In the model results (Fig. 5b), a similar precipitation belt exists over the middle and lower extent of the Yangtze River. To compare the locations of the bands of heavy precipitation, the 200-mm contour from the observations is depicted in Fig. 5b by a white dashed line. It is shown that the spatial pattern of the observed precipitation belt matches well with that simulated by the model. The white dashed line encloses 106 grid points in Fig. 5b (the horizontal resolution of the grid is approximately 0.47° × 0.47°). Correspondingly, there are 94 grid points exceeding 200 mm in the model results, and 73 of these grids are located within the white dashed line. The simulated precipitation amount averaged over the YRV (27°–33°N, 108°–120°E) is 132.3 mm, which is slightly lower than the observed value (153.9 mm). The highest total precipitation amount at all the national meteorological stations is as high as 537.2 mm and is recorded at the station located at approximately 31.6°N, 117.9°E. The simulated highest precipitation reaches 555.4 mm and is located at 30.7°N, 116.7°E, which is comparable to the observations. The above comparison shows that the model can reasonably reproduce the spatial distribution and magnitude of the 6-day accumulated precipitation amount for this extreme case via the T-AMIP experiment.
Accumulated precipitation (mm) for the 6-day period (30 Jun–5 Jul 2016) from the (a) gauge observations and (b) T-AMIP simulations. The white dashed line in (b) is the 200-mm contour from the gauge observations. The Yangtze River and Yellow River are drawn as thick gray lines.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
In addition to the accumulated precipitation amount, the precipitation frequency and intensity at an hourly scale are also key metrics to evaluate the performance of the model. As shown in Fig. 6, the frequency in the T-AMIP experiments is also overestimated, but the overestimations are smaller than those in the AMIP-type simulation. Over the central region of the extreme event (27°–33°N, 108°–120°E), the 6-day mean frequency from the gauge observations is 36.32%, while it is 39.31% in the simulation. Correspondingly, the intensity is slightly underestimated, with the observed intensity at 2.61 mm h−1 and the model intensity at 2.02 mm h−1. To further quantitatively examine the model performance, 393 rain gauge stations and 312 model grids over the target region are selected. These stations and grids are grouped according to the mean hourly intensity in the 6-day period, and the percentages of the stations and grids in different intensity categories are shown in Fig. 7a. Approximately 49.9% (82.7%) of the stations (grids) have a mean intensity value less than 3 mm h−1, indicating an underestimation of the hourly intensity in the model. The simulated mean intensity averaged over the central region is 2.0 mm h−1, which is much lower than that of the observations (3.1 mm h−1). Figure 7b shows the results for the maximum hourly precipitation intensity at 5 mm h−1 bins. In the observations, 43.5% of the stations recorded a maximum intensity larger than 20 mm h−1. In contrast, precipitation heavier than 20 mm h−1 is reproduced only at 6.4% of the model grids. The strongest intensity is 62.4 mm h−1 in the observations and 30.3 mm h−1 in the simulation. Although the T-AMIP experiment simulates a reasonable pattern of accumulated precipitation amount, the model significantly underestimates the intensity of the extreme event.
As in Fig. 5, but for the 6-day mean precipitation frequency (%).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
Percentages of the model grids (gray dashed line with triangles) in the T-AMIP experiments and stations (solid black line with circles) in different ranges of (a) mean hourly precipitation intensity (1 mm h−1 bin) and (b) maximum hourly precipitation intensity (5 mm h−1 bin) during the 6-day period.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
To evaluate the hourly characteristics, the time–longitude diagram of precipitation averaged over 27°–33°N is shown in Fig. 8. A striking feature in the observations is the eastward movement of various precipitating systems during the extreme event (Fig. 8a). Compared with the observations, the model can partially reproduce the zonal propagation of precipitation (Fig. 8b). The BT observed by the FY-2G geostationary satellite and simulated by the model is compared in Fig. 9. Consistent with the observed gauged rainfall, the eastward movement is also obvious by the variation in the satellite BT (Fig. 9a). The model also simulates the eastward movement of convective systems during the extreme event, and the BT during the heavy rainfall periods is comparable with that observed by the satellite.
Hovmöller diagram of meridionally averaged (27°–33°N) precipitation (mm h−1) from the (a) observations and (b) T-AMIP simulations.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
As in Fig. 8, but for the IR1 BT (K) from the (a) FY-2G satellite observations and (b) T-AMIP simulations. The contours indicate BT values lower than 238 K, which specify convective areas apparent in both the observations and model.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
A typical period of good simulation is in the first two days (from 30 June to 1 July), during which the model presents a realistic phase and speed of the eastward movement of both rainfall centers (Fig. 8) and convective systems (Fig. 9). The dominant synoptic system in this period is a low vortex moving along the Yangtze River. The low vortex, which is generated over the Sichuan basin (27°–33°N, 103°–108°E) and is known as the southwest vortex, is an important synoptic-scale cyclone that greatly influences the rainfall over China (Kuo et al. 1988; Li et al. 2010; Li et al. 2017). The convection and rainfall centers prefer to trigger and develop over the southeast of the low-level vortex and move with the vortex. As shown in Fig. 10a, the southwest vortex formed over the Sichuan basin on 29 June, stayed in the basin for one day, and started moving out at approximately 1200 UTC 30 June. After 6 h (1800 UTC 30 June), the vortex climbed over the mountains to the east of the basin (approximately 110°E) and reached the plain (112°E) at 0000 UTC 1 July. Then, it moved eastward steadily along a track to the north of the YRV. The track in the model results (Fig. 8b) agrees well with the observed track, which is consistent with the realistic variation in the simulated precipitation. To examine the dynamic features of the vortex, the geopotential height anomalies (normalized by the 6-day mean) at the location of the low vortex are compared in Fig. 8c. The simulation generally captures the variation in the low vortex during the first two days. The low vortex strengthened moving into the basin and weakened moving out of the basin, which are both consistent in the reanalysis and the model. The model shows some biases when the low vortex climbs over the mountains to the east of the basin from 0600 to 1800 UTC 30 June.
The track of the low vortex (red line) in the (a) ERA-Interim reanalysis and (b) CAMS-CSM T-AMIP simulation. The location of the low vortex is marked by dots and circles at 6-h intervals, and the time is labeled at 12-h intervals. The topography is shown by color shading (m). The Yangtze River and Yellow River are drawn as blue lines. (c) The comparison of 6-h series of geopotential anomalies (normalized by the 6-day mean) at the location of the low vortex at 850 hPa in ERA-Interim (blue line with open circles) and CAMS-CSM T-AMIP simulations (red line with triangles).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
In contrast to the good simulation in the first two days, the variation in the modeled precipitation on 4 July differs significantly from the measured results (Fig. 8). In Fig. 8b, there is an artificial precipitation center from 1500 to 2100 UTC 4 July at approximately 116°E, which corresponds to a steady strong convection center over the area (Fig. 9b). The distribution of the difference between the simulated and observed precipitation amount during these hours is presented in Fig. 11. Large positive values are located to the south of the mountains, and the highest overestimation reaches 112.5 mm (at 29.7°N, 115.8°E). Accompanying the artificial precipitation, in the model results, there is a strong southwesterly wind and considerable convergence in the lower troposphere over the upwind slope of the mountains at 32°N, 116°E (Fig. 11). The 850-hPa divergence averaged over the dashed box shown in Fig. 11 (29°–32°N, 114°–118°E) exceeds −5.3 × 10−5 s−1, while the value is 4.7 × 10−6 s−1 in the ERA-Interim reanalysis data, indicating a weak divergence. Therefore, the large overestimation of precipitation over the upwind slope may be ascribed to the unrealistic convergence in the lower troposphere.
The bias of precipitation (difference between the T-AMIP simulation and observations) accumulated from 1500 to 2100 UTC 4 Jul. The positive (negative) bias is shown by blue solid (red dashed) contours. The positive (negative) contours start from 30 mm (−30 mm) with intervals of 30 mm, and the zero line is omitted. The vectors are for the simulated 850-hPa wind averaged from 1500 to 2100 UTC 4 Jul. The topography used in the model is shown by color shading (m). The box marked by the black dashed lines encloses the region of 29°–32°N, 114°–118°E.
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
To further examine the factors responsible for the precipitation bias over the dashed box in Fig. 11, the variation in the bias of the 850-hPa divergence and the precipitation averaged over this region is presented in Fig. 12a. The difference in the divergence between the simulation and reanalysis is shown as a red line in Fig. 12a, and the blue line represents the precipitation bias. The mean precipitation bias during the whole event is −0.29 mm h−1, showing a weak underestimation over the topography. The largest overestimation reaches 6.1 mm h−1 and occurs at 1900 UTC 4 July. The strongest convergence bias is one hour earlier than the peak of the overestimation of the precipitation. As shown in Fig. 12a, there is a significant negative correlation between the biases of divergence and precipitation. The correlation coefficient at lag 1 (hour) between the red and blue lines is −0.74. The red line in Fig. 12b shows the difference in the total column water vapor (TCWV) averaged over 29°–32°N, 114°–118°E between the simulation and reanalysis. The precipitation bias is positively correlated with the bias of the TCWV (coefficient of 0.74 at a 1-h lag), indicating the significant modulation of the TCWV on the bias of precipitation. Thus, both low-level convergence and TCWV contribute to the inadequacy in precipitation over this mountainous region. Because of the inadequacies in the representation of the effect of the local topography, the model produces an unrealistic convergence in the lower troposphere, which increases the TCWV and leads to an overestimation of precipitation.
(a) Hourly series of the difference in 850-hPa divergence (precipitation) between the simulation and ERA5 (observations) averaged over the dashed box in Fig. 11 (29°–32°N, 114°–118°E). The blue line represents precipitation (left y axis; mm) and the red line represents divergence (right y axis; s−1); (b) as in (a), but the red line represents the difference in TCWV (kg m−2).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
5. Summary and discussion
After examining the basic performance of CAMS-CSM in simulating the climatic features of rainfall in China, an extreme precipitation event that occurred along the YRV during the midsummer of 2016 is simulated by T-AMIP experiments using the climate model in the weather forecast mode. Two typical precipitation periods and the related synoptic circulation patterns are evaluated to examine the model’s capability of reproducing the severity and process of the extreme rainfall event. The main conclusions are as follows:
The T-AMIP experiments successfully simulate the spatial pattern and heavy centers of the extreme rainfall event as validated by gauge observations along the YRV from 30 June to 5 July 2016. The accumulated rainfall amount exceeds 300 mm in the model, which is comparable with the observations during the event. The model significantly underestimates the maximum hourly intensity in the extreme event, which is quite similar to its climatic biases.
The model performance in representing the rainfall processes depends on the scale of the synoptic systems. The model realistically reproduces the zonal propagation of heavy rainfall from 30 June to 1 July, which is closely related to the adequately simulated eastward movement of the low-level southwest vortex. For another heavy rainfall period on 4 July, the model produces an artificial heavy rainfall center around the high terrain at approximately 116°E, which may be caused by the unrealistic low-level convergence and precipitable water.
The comparison of the simulation of the two typical heavy rainfall periods shows that the model’s capability largely depends on the spatial scale of the related synoptic systems and the related topography forcing. To illustrate the influences of these factors, the low-level circulation during the two periods is further compared in Fig. 13. As shown in Fig. 10a, the successful simulation of the rainfall process from 30 June to 1 July is closely related to the reasonable reproduction of the southwest vortex. At 1800 UTC 30 June, when the southwest vortex (located over 27°–32°N, 106°–110°E) moves out of the basin, the size of the vortex is approximately 350 km (Fig. 13a), which is approximately the size of the Sichuan basin where it initializes. Such a large-scale circulation pattern is well resolved in the model with T255 resolution (Fig. 13b), which contributes to the successful simulation of the heavy rainfall process. Nevertheless, for the second heavy rainfall period, the dominant low-level system is the smaller vortex (located over 30°–32°N, 115°–117°E) generated in the shear line along the southwesterly low-level jet. The size of the small vortex at 1800 UTC 3 July (the hour when the strongest convergence bias occurs) is much smaller (Fig. 13c). The model reproduces the shear line along the low-level jet, but hardly resolves the integrity of the small vortex over the shear line (Fig. 13d), which in turn fails to reproduce the eastward propagation of the heavy rainfall center along with the small vortex in the observations. The low vortex in the model probably dissipates too quickly after the model initialization. The moisture, which should have been converted to rainfall along the track of the vortex, condenses and precipitates out when the air is uplifted after encountering the topography (Fig. 12). The artificial persistent heavy rain in the model appears over the upwind slope.
The 850-hPa wind (vectors) from 1800 UTC (a),(b) 30 Jun and (c),(d) 3 Jul in the (left) ERA-Interim reanalysis and (right) CAMS-CSM T-AMIP simulation. The topography is shown by color shading (m).
Citation: Journal of Climate 31, 16; 10.1175/JCLI-D-17-0801.1
Previous studies using climate models to simulate and project extreme rainfall events mainly take all events together and investigate the overall features of these events. By examining a model’s performance in simulating the extreme rainfall event along the YRV in the midsummer of 2016, this study demonstrates that a climate model is capable of reproducing the severity of the heavy rainfall event. However, the model performance greatly depends on the model resolution, which determines the model’s ability to produce the synoptic systems that dominate the events. Therefore, before analyzing the features or trends of the extreme events based on the simulation or projection by climate models, it is necessary to classify the events based on the spatial scale of the dominated systems. The model could realistically reproduce the features, and probably the impact, of the events that are dominated by the synoptic systems resolved by the model. However, for those events dominated by systems that are too small to be resolved by the model, the simulation or projection would present large uncertainties. This also implies that for the predictions of extremes of rainfall in the East Asian region, it is better to use observations or hindcasts to calculate the rainfall contributed by unresolved systems of the kind such as in the second period of the event shown in this work and use that to statistically adjust regional climate projections.
The simulation of the East Asian summer monsoon (EASM) mean state and variability are challenging issues for the climate-modeling community, and the underlying causes are still not clear (Wang et al. 2005; Zhou et al. 2009; Chen et al. 2010b; Zhou et al. 2017). The East Asian mei-yu–baiu–changma front along the YRV, a major rain-bearing system in the subtropics and midlatitudes causing heavy rainfall events frequently (Huang et al. 2007; Ding et al. 2015), provides an ideal test bed to evaluate model performance and investigate the possible causes for the model biases. By using the T-AMIP experiment, this study emphasizes the importance of evaluation of rainfall processes to attribute model biases. These results could assist in understanding the strengths and weaknesses of climate models in simulating heavy rainfall events.
The current study mainly focuses on the representation of the magnitude of the rainfall amount and the influence of circulation systems during an extreme event, and the model biases are mainly attributed to the resolution. The initial condition of soil moisture, which could also lead to heavy precipitation (Rasmussen and Houze 2012; Kumar et al. 2014), is not considered in this work and still needs more investigation. The evolution of the rainfall systems and the relationship with physical parameterizations also require future work. Moreover, only a representative extreme case is explored in this study, and the roles of model resolution, the representation of synoptic-scale features, and topographic forcing will be further investigated using other cases to corroborate the results.
Acknowledgments
The authors gratefully acknowledge the anonymous editors, whose valuable comments contributed a great deal to improving the quality of this paper. This work is supported by National Natural Science Foundation of China (91637210, 41675075, and 91737306). The contributors (KF and SM) were supported by the United Kingdom–China Research & Innovation Partnership Fund through the Met Office Climate Science for Service Partnership (CSSP) China as part of the Newton Fund.
REFERENCES
Banzon, V., T. M. Smith, T. M. Chin, C. Liu, and W. Hankins, 2016: A long-term record of blended satellite and in situ sea-surface temperature for climate monitoring, modeling and environmental studies. Earth Syst. Sci. Data, 8, 165–176, https://doi.org/10.5194/essd-8-165-2016.
Bao, J., S. C. Sherwood, L. V. Alexander, and J. P. Evans, 2017: Future increases in extreme precipitation exceed observed scaling rates. Nat. Climate Change, 7, 128–132, https://doi.org/10.1038/nclimate3201.
Bi, B., X. Zhang, and K. Dai, 2017: Characteristics of 2016 severe convective weather and extreme rainfalls under the background of super El Niño (in Chinese). Chin. Sci. Bull., 62, 928–937, https://doi.org/10.1360/N972016-01136.
Chen, H., R. Yu, J. Li, W. Yuan, and T. Zhou, 2010a: Why nocturnal long-duration rainfall presents an eastward-delayed diurnal phase of rainfall down the Yangtze River valley. J. Climate, 23, 905–917, https://doi.org/10.1175/2009JCLI3187.1.
Chen, H., T. Zhou, R. B. Neale, X. Wu, and G. J. Zhang, 2010b: Performance of the new NCAR CAM3.5 in East Asian summer monsoon simulations: Sensitivity to modifications of the convection scheme. J. Climate, 23, 3657–3675, https://doi.org/10.1175/2010JCLI3022.1.
Dai, A., X. Lin, and K.-L. Hsu, 2007: The frequency, intensity, and diurnal cycle of precipitation in surface and satellite observations over low- and mid-latitudes. Climate Dyn., 29, 727–744, https://doi.org/10.1007/s00382-007-0260-y.
Dai, Y., and Coauthors, 2003: The Common Land Model. Bull. Amer. Meteor. Soc., 84, 1013–1024, https://doi.org/10.1175/BAMS-84-8-1013.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Ding, Y., Y. Liu, Y. Song, and J. Zhang, 2015: From MONEX to the global monsoon: A review of monsoon system research. Adv. Atmos. Sci., 32, 10–31, https://doi.org/10.1007/s00376-014-0008-7.
Gates, W. L., and Coauthors, 1999: An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I). Bull. Amer. Meteor. Soc., 80, 29–56, https://doi.org/10.1175/1520-0477(1999)080<0029:AOOTRO>2.0.CO;2.
Gleckler, P. J., K. E. Taylor, and C. Doutriaux, 2008: Performance metrics for climate models. J. Geophys. Res., 113, D06104, https://doi.org/10.1029/2007JD008972.
Guo, Q., F. Chen, B. Chen, X. Feng, C. Yang, X. Wang, and Z. Zhang, 2016: Internal‐blackbody calibration (IBBC) approach and its operational application in FY‐2 meteorological satellites. Quart. J. Roy. Meteor. Soc., 142, 3082–3096, https://doi.org/10.1002/qj.2890.
Hallegatte, S., C. Green, R. J. Nicholls, and J. Corfee-Morlot, 2013: Future flood losses in major coastal cities. Nat. Climate Change, 3, 802–806, https://doi.org/10.1038/nclimate1979.
Huang, R., J. Chen, and G. Huang, 2007: Characteristics and variations of the East Asian monsoon system and its impacts on climate disasters in China. Adv. Atmos. Sci., 24, 993–1023, https://doi.org/10.1007/s00376-007-0993-x.
Jiang, Z., W. Li, J. Xu, and L. Li, 2015: Extreme precipitation indices over China in CMIP5 models. Part I: Model evaluation. J. Climate, 28, 8603–8619, https://doi.org/10.1175/JCLI-D-15-0099.1.
Khoei, A. R., and S. A. Gharehbaghi, 2007: The superconvergence patch recovery technique and data transfer operators in 3D plasticity problems. Finite Elem. Anal. Des., 43, 630–648, https://doi.org/10.1016/j.finel.2007.01.002.
Kopparla, P., E. M. Fischer, C. Hannay, and R. Knutti, 2013: Improved simulation of extreme precipitation in a high-resolution atmosphere model. Geophys. Res. Lett., 40, 5803–5808, https://doi.org/10.1002/2013GL057866.
Kumar, A., R. A. Houze Jr., K. L. Rasmussen, and C. Peters-Lidard, 2014: Simulation of a flash flooding storm at the steep edge of the Himalayas. J. Hydrometeor., 15, 212–228, https://doi.org/10.1175/JHM-D-12-0155.1.
Kuo, Y.-H., L. Cheng, and J.-W. Bao, 1988: Numerical simulation of the 1981 Sichuan flood. Part I: Evolution of a mesoscale southwest vortex. Mon. Wea. Rev., 116, 2481–2504, https://doi.org/10.1175/1520-0493(1988)116<2481:NSOTSF>2.0.CO;2.
Li, J., R. Yu, W. Yuan, H. Chen, W. Sun, and Y. Zhang, 2015: Precipitation over East Asia simulated by NCAR CAM5 at different horizontal resolutions. J. Adv. Model. Earth Syst., 7, 774–790, https://doi.org/10.1002/2014MS000414.
Li, L., R. Zhang, and M. Wen, 2017: Genesis of southwest vortices and its relation to Tibetan Plateau vortices. Quart. J. Roy. Meteor. Soc., 143, 2556–2566, https://doi.org/10.1002/qj.3106.
Li, Y., D. Li, S. Yang, C. Liu, A. Zhong, and Y. Li, 2010: Characteristics of the precipitation over the eastern edge of the Tibetan Plateau. Meteor. Atmos. Phys., 106, 49–56, https://doi.org/10.1007/s00703-009-0048-1.
Liu, Y., J. Xia, C.-X. Shi, and Y. Hong, 2009: An improved cloud classification algorithm for China’s FY-2C multi-channel images using artificial neural network. Sensors, 9, 5558–5579, https://doi.org/10.3390/s90705558.
Madsen, H., D. Lawrence, M. Lang, M. Martinkova, and T. R. Kjeldsen, 2014: Review of trend analysis and climate change projections of extreme precipitation and floods in Europe. J. Hydrol., 519, 3634–3650, https://doi.org/10.1016/j.jhydrol.2014.11.003.
Min, S.-K., X. Zhang, F. W. Zwiers, and G. C. Hegerl, 2011: Human contribution to more-intense precipitation extremes. Nature, 470, 378–381, https://doi.org/10.1038/nature09763; Corrigendum, 498, 526, https://doi.org/10.1038/nature09763.
O’Gorman, P. A., and T. Schneider, 2009: The physical basis for increases in precipitation extremes in simulations of 21st-century climate change. Proc. Natl. Acad. Sci. USA, 106, 14 773–14 777, https://doi.org/10.1073/pnas.0907610106.
Phillips, T. J., and Coauthors, 2004: Evaluating parameterizations in general circulation models: Climate simulation meets weather prediction. Bull. Amer. Meteor. Soc., 85, 1903–1915, https://doi.org/10.1175/BAMS-85-12-1903.
Rajczak, J., P. Pall, and C. Schär, 2013: Projections of extreme precipitation events in regional climate simulations for Europe and the Alpine region. J. Geophys. Res. Atmos., 118, 3610–3626, https://doi.org/10.1002/jgrd.50297.
Rasmussen, K. L., and R. A. Houze Jr., 2012: A flash-flooding storm at the steep edge of high terrain: Disaster in the Himalayas. Bull. Amer. Meteor. Soc., 93, 1713–1724, https://doi.org/10.1175/BAMS-D-11-00236.1.
Reichler, T., and J. Kim, 2008: How well do coupled models simulate today’s climate? Bull. Amer. Meteor. Soc., 89, 303–312, https://doi.org/10.1175/BAMS-89-3-303.
Rikus, L., 1997: Application of a scheme for validating clouds in an operational global NWP model. Mon. Wea. Rev., 125, 1615–1637, https://doi.org/10.1175/1520-0493(1997)125<1615:AOASFV>2.0.CO;2.
Roeckner, E., and Coauthors, 2003: The atmospheric general circulation model ECHAM5. Part 1: Model description. Max Planck Institute for Meteorology Rep. 349, 127 pp.
Rosa, D., and W. D. Collins, 2013: A case study of subdaily simulated and observed continental convective precipitation: CMIP5 and multiscale global climate models comparison. Geophys. Res. Lett., 40, 5999–6003, https://doi.org/10.1002/2013GL057987.
Sillmann, J., V. V. Kharin, F. W. Zwiers, X. Zhang, and D. Bronaugh, 2013: Climate extremes indices in the CMIP5 multimodel ensemble: Part 2. Future climate projections. J. Geophys. Res. Atmos., 118, 2473–2493, https://doi.org/10.1002/jgrd.50188.
Sillmann, J., C. W. Stjern, G. Myhre, and P. M. Forster, 2017: Slow and fast responses of mean and extreme precipitation to different forcing in CMIP5 simulations. Geophys. Res. Lett., 44, 6383–6390, https://doi.org/10.1002/2017GL073229.
Stephens, G. L., and Coauthors, 2010: Dreary state of precipitation in global models. J. Geophys. Res., 115, D24211, https://doi.org/10.1029/2010JD014532.
Stott, P., 2016: How climate change affects extreme weather events. Science, 352, 1517–1518, https://doi.org/10.1126/science.aaf7271.
Sugiyama, M., H. Shiogama, and S. Emori, 2010: Precipitation extreme changes exceeding moisture content increases in MIROC and IPCC climate models. Proc. Natl. Acad. Sci. USA, 107, 571–575, https://doi.org/10.1073/pnas.0903186107.
Sun, Z., and L. Rikus, 2004: Validating model clouds and their optical properties using geostationary satellite imagery. Mon. Wea. Rev., 132, 2006–2020, https://doi.org/10.1175/1520-0493(2004)132<2006:VMCATO>2.0.CO;2.
Trenberth, K. E., A. Dai, R. M. Rasmussen, and D. B. Parsons, 2003: The changing character of precipitation. Bull. Amer. Meteor. Soc., 84, 1205–1217, https://doi.org/10.1175/BAMS-84-9-1205.
Wang, B., Q. Ding, X. Fu, I.-S. Kang, K. Jin, J. Shukla, and F. Doblas-Reyes, 2005: Fundamental challenge in simulation and prediction of summer monsoon rainfall. Geophys. Res. Lett., 32, L15711, https://doi.org/10.1029/2005GL022734.
Williams, K. D., and Coauthors, 2013: The Transpose-AMIP II experiment and its application to the understanding of Southern Ocean cloud biases in climate models. J. Climate, 26, 3258–3274, https://doi.org/10.1175/JCLI-D-12-00429.1.
Xu, C., Y. Luo, and Y. Xu, 2011: Projected changes of precipitation extremes in river basins over China. Quat. Int., 244, 149–158, https://doi.org/10.1016/j.quaint.2011.01.002.
Yu, R., 1994: A two-step shape-preserving advection scheme. Adv. Atmos. Sci., 11, 479–490, https://doi.org/10.1007/BF02658169.
Yu, R., T. Zhou, A. Xiong, Y. Zhu, and J. Li, 2007: Diurnal variations of summer precipitation over contiguous China. Geophys. Res. Lett., 34, L01704, https://doi.org/10.1029/2006GL028129.
Yu, R., J. Li, Y. Zhang, and H. Chen, 2015: Improvement of rainfall simulation on the steep edge of the Tibetan Plateau by using a finite-difference transport scheme in CAM5. Climate Dyn., 45, 2937–2948, https://doi.org/10.1007/s00382-015-2515-3.
Zhang, H., and Coauthors, 2012: Simulation of direct radiative forcing of aerosols and their effects on East Asian climate using an interactive AGCM-aerosol coupled system. Climate Dyn., 38, 1675–1693, https://doi.org/10.1007/s00382-011-1131-0.
Zhang, Y., and H. Chen, 2016: Comparing CAM5 and superparameterized CAM5 simulations of summer precipitation characteristics over continental East Asia: Mean state, frequency–intensity relationship, diurnal cycle, and influencing factors. J. Climate, 29, 1067–1089, https://doi.org/10.1175/JCLI-D-15-0342.1.
Zhou, T., B. Wu, and B. Wang, 2009: How well do atmospheric general circulation models capture the leading modes of the interannual variability of the Asian–Australian monsoon? J. Climate, 22, 1159–1173, https://doi.org/10.1175/2008JCLI2245.1.
Zhou, T., and Coauthors, 2017: A robustness analysis of CMIP5 models over the East Asia-western North Pacific domain. Engineering, 3, 773–778, https://doi.org/10.1016/J.ENG.2017.05.018.