1. Introduction
Information on subdaily rainfall is important for multiple weather and climate applications, including validation of climate model performance. Therefore, many observational studies (Gray and Jacobson 1977; Dai 2001; Sorooshian et al. 2002; Nesbitt and Zipser 2003; Bowman et al. 2005; Hirose et al. 2008; Yang and Smith 2006; Yamamoto et al. 2008) and modeling studies (Collier and Bowman 2004; Dai and Trenberth 2004; Dai 2006; Sato et al. 2009; Noda et al. 2012) have investigated the global diurnal rainfall characteristics. However, although the global domain could allow comparisons between different regions around the globe, regional details, which are important for local communities, are difficult to achieve (e.g., Fig. S1 in the online supplemental material, to be discussed later). Additionally, the physical mechanisms that affect the diurnal rainfall variations are very complex and depend on regional characteristics (Romatschke et al. 2010; Chen et al. 2014, 2018). Therefore, instead of analyzing global domain, more and more recent studies have examined the characteristics and mechanisms of variations in the regional diurnal rainfall (Yuan et al. 2012; Satoh and Kitao 2013; Chen et al. 2014; Zhang et al. 2015; Huang and Wang 2017; Oh and Suh 2018; Huang et al. 2019; Riley Dellaripa et al. 2020). Thus, the present study was conducted with a regional perspective, particularly with a focus on the diurnal rainfall features over Southeast Asia (defined as the domain of Fig. 1a), where the diurnal rainfall characteristics and maintenance mechanisms among the land areas within this region were similar. Additionally, the related literatures were reviewed.
(a) Topography and geographic location of South China (brown-outlined box), Taiwan, and Luzon. (b) The 3-hourly rainfall averaged over the 2000–14 JJA seasons, and area-averaged over three selected land areas: South China (yellow bars), Taiwan (purple-outlined bars), and Luzon (red bars). (c) Infrared (IR) cloud image (extracted from http://www.ncdc.noaa.gov/gridsat/) of a CAR case, taken at 1700 LT (i.e., 17 h) 28 Jul 2015. (d) As in (b), but for the variation in 3-hourly rainfall that occurred on 28 Jul 2015.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Topography and geographic location of South China (brown-outlined box), Taiwan, and Luzon. (b) The 3-hourly rainfall averaged over the 2000–14 JJA seasons, and area-averaged over three selected land areas: South China (yellow bars), Taiwan (purple-outlined bars), and Luzon (red bars). (c) Infrared (IR) cloud image (extracted from http://www.ncdc.noaa.gov/gridsat/) of a CAR case, taken at 1700 LT (i.e., 17 h) 28 Jul 2015. (d) As in (b), but for the variation in 3-hourly rainfall that occurred on 28 Jul 2015.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Topography and geographic location of South China (brown-outlined box), Taiwan, and Luzon. (b) The 3-hourly rainfall averaged over the 2000–14 JJA seasons, and area-averaged over three selected land areas: South China (yellow bars), Taiwan (purple-outlined bars), and Luzon (red bars). (c) Infrared (IR) cloud image (extracted from http://www.ncdc.noaa.gov/gridsat/) of a CAR case, taken at 1700 LT (i.e., 17 h) 28 Jul 2015. (d) As in (b), but for the variation in 3-hourly rainfall that occurred on 28 Jul 2015.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
During summer [June–August (JJA)], rainfall formation over the land areas of Southeast Asia (including South China, Taiwan, and Luzon; marked in Fig. 1a) frequently consists of a distinct diurnal signal (Fig. 1b) (Yuan et al. 2012; Zhang et al. 2015; Huang et al. 2019). This diurnal rainfall signal is mainly associated with a local convective afternoon rainfall (CAR) event (Fig. 1c), which is a weather feature primarily induced by the diurnal cycle of solar radiation. In general, CAR events tend to generate high local diurnal rainfall in the afternoon (Fig. 1d) (Huang et al. 2019). Climatologically, CAR events are among the most frequently observed summer rain patterns over South China, Taiwan, and Luzon (Wang and Chen 2008; Chen et al. 2018). On an average, more than 50% of the total summer rainy days over Taiwan can be considered as CAR events (Wang and Chen 2008). Additionally, it is common for extreme CAR events, which featured a short life cycle but intense rainfall, to cause local flooding over Taiwan (Chiang et al. 2019). Similar disasters also occur frequently over South China (Zhong 2020) and Luzon (Riley Dellaripa et al. 2020).
Because of its importance and effects on daily life, understanding how the characteristics of CAR activities over Southeast Asia have changed has attracted significant attention (Chen et al. 2014; Huang and Wang 2014; Chen et al. 2018; Huang et al. 2019). Using observational data, studies have noted that the diurnal rainfall variation and related CAR characteristics over Southeast Asia over the past several decades consist of long-term trend signals (Huang and Wang 2014; Huang and Chen 2015). For example, by examining the CAR characteristics over South China and Taiwan in May and June during the 1982–2012 period, Huang and Chen (2015) noted that CAR events have become more frequent and intense. Such an increase in CAR frequency is suggested to be attributed to the intensification of daytime local wind convergence and thermal instability, while the increase in CAR intensity is suggested to be maintained by the intensification of moisture supply over Southeast Asia (Huang and Chen 2015).
In addition to observational data, model simulation data were also widely adopted by earlier studies to help clarify the causes of local weather changes (Satoh and Kitao 2013; Riley Dellaripa et al. 2020). Since 1995, an international Coupled Model Intercomparison Project (CMIP) has been organized to help better understand past, present, and future climate changes (Meehl 1995). Subsequently, many studies have evaluated the performance of global climate models (GCMs), mainly phase 3 (CMIP3) or phase 5 (CMIP5), in depicting the diurnal rainfall variation over various regions (Yang and Slingo 2001; Collier and Bowman 2004; Sato et al. 2009; Kim et al. 2019). It was noted that most GCMs in CMIP3 or CMIP5 have diurnal rainfall maximum peaks too early (i.e., occurring earlier than observed) over Southeast Asia (Dai 2006; Yuan et al. 2013). However, Huang and Wang (2017) examined 18 CMIP5 models and reported that one model—the Centro Euro-Mediterraneo sui Cambiamenti Climatici Climate Model (CMCC-CM), which has the finest spatial resolution—can capture diurnal rainfall variation over East Asia similar to the observation. Similarly, by dynamical downscaling using a lower-resolution model to drive a higher-resolution model, other studies found that this approach can improve the simulation of CAR activities over South China, Taiwan, and Luzon (Huang et al. 2016a,b). Therefore, high-resolution model simulation is assumed to better estimate the diurnal rainfall variation and the related CAR activities (e.g., frequency, intensity, and amount) over Southeast Asia.
As a result of this assumption, few studies have investigated the future changes in regional diurnal rainfall variation using the low-resolution GCMs directly (Huang and Wang 2017). In contrast, multiple studies have used the high-resolution regional model driven by the boundary conditions obtained from low-resolution GCMs (including CMIP3 and CMIP5) to project the future regional diurnal rainfall characteristics (Huang et al. 2016a,b; Oh and Suh 2018; Bowden et al. 2020). Among the documented studies that adopted the dynamical downscaling approach, Huang et al. (2016a,b) showed that the CAR events likely become fewer but more intense over South China, Taiwan, and Luzon at the end of the twenty-first century. However, the projected changes shown by Huang et al. (2016a,b) were based only on few (2–4) high-resolution model simulations.
Currently, the CMIP is in its sixth phase (CMIP6) (Eyring et al. 2016). Relative to CMIP3 and CMIP5, the majority of the CMIP6 models exhibit fine spatial resolution and parameterization improvements. Therefore, it is expected that CMIP6 models will perform better than CMIP3 or CMIP5 in depicting many global and regional features. Indeed, this expectation has been supported by recent studies (Gusain et al. 2020; Wu et al. 2019; Fu et al. 2020; Xin et al. 2020). For example, Xin et al. (2020) found that the CMIP6 models are more skillful than the CMIP5 models in the climatological rainfall pattern over eastern China. However, relative to the number of studies that examine CMIP3 and CMIP5 models’ diurnal rainfall ability (Covey et al. 2016; Huang and Wang 2017; Kamworapan and Surussavadee 2019), the performance of newly released CMIP6 models in depicting diurnal rainfall features and the related CAR activities over Southeast Asia has been less studied and deserves research attention.
Recently, Wu et al. (2019) examined the main progress of the Beijing Climate Center Climate System Model (BCC-CSM) from CMIP5 to CMIP6 and indicated that it shows significant improvements in the diurnal cycle of rainfall over many areas of the world. However, one can note from Fig. 20 of Wu et al. (2019) that BCC-CSM in CMIP6 has problems (i.e., peaks too early) in capturing the temporal phase of diurnal rainfall maximum over Southeast Asia (including South China, Taiwan, and Luzon). Therefore, one of the main objectives of this study is to understand whether the bias seen in the BCC-CSM for the presentation of diurnal rainfall over Southeast Asia is a common error observed in other CMIP6 models (to be discussed in section 3). The other objective of this study is to evaluate the performance of CMIP6 models by the simulation of present-day local CAR activities (to be discussed in section 4). Moreover, this study also aims to clarify whether the projected changes in CAR activities using the CMIP6 models are consistent with those observed by Huang et al. (2016a,b) using high-resolution regional models (to be discussed in section 5). Clarifying these issues is important because it not only can help us better understand the applicability of CMIP6 models to study regional rainfall features, but also can provide insights into the variations in the local rainfall patterns in response to global warming in the future.
2. Data and method
a. Information of data and analyzed periods
CMIP6 models used in this study.
Additionally, we used satellite precipitation estimations from the Integrated Multisatellite Retrievals for Global Precipitation Measurement (IMERG), version 6, final run product (hereinafter simply IMERG; https://gpm.nasa.gov/data/imerg) (Huffman et al. 2019) as the observational reference base to help evaluate models’ capability in simulating present-day rainfall variations. Huang et al. (2020) demonstrated that the IMERG data are more capable than the frequently used Tropical Rainfall Measuring Mission (TRMM) 3B42 data in representing the characteristics of CAR events over Taiwan. Since the IMERG data were available since 2000 and the CMIP6 model historical run simulations ended in 2014, the overlapping period of 2000–14 JJA seasons was used to validate the present-day precipitation simulations of the models. A similar approach of using the overlapping time period between observations and the model simulations has been frequently adopted by earlier studies to evaluate the performance of the model (e.g., Covey et al. 2016; Kamworapan and Surussavadee 2019; Xin et al. 2020). For evaluation, all datasets (including observation and simulation) examined in this study were regridded into a unified spatial resolution 0.7° × 0.7° (i.e., the finest horizontal resolution listed in Table 1), with a temporal resolution of 3 h. Hereinafter, all the analyses are presented in Taiwan local time (LT; UTC + 8 h).
b. Identification of CAR events
To identify CAR events, we used methods derived from earlier studies (Huang and Chen 2015; Huang et al. 2016a). In the first step, we identified rainy days when the daily accumulated rainfall > 0.1 mm day−1. In the second step, a CAR day was identified as a rainy day meeting three criteria: 1) rainfall accumulated during 1200–2200 LT > 80% of the daily rainfall, 2) rainfall accumulated during 0100–1100 LT < 10% of the daily rainfall, and 3) not affected by other nonlocal rainy patterns (e.g., fronts and typhoons). Other details with regard to the identification of CAR events were given in Huang and Chen (2015) and Huang et al. (2016a).
3. Present-day simulation of diurnal rainfall features
Empirical orthogonal function (EOF) analysis has been frequently adopted by earlier studies to illustrate the regional spatiotemporal characteristics of the observed diurnal rainfall variations over Southeast Asia (Huang and Wang 2017; Huang et al. 2020). Similarly, we applied EOF analysis on the 3-hourly rainfall anomalies (i.e., daily mean removed) extracted from IMERG and averaged during 2000–14 JJA seasons (Figs. 2a–c). Hannachi et al. (2007) provided a detailed review of the formulation and application of the EOF analysis. The features showed in Figs. 2a–c are as follows: 1) the first and second EOF modes explain approximately 72.3% and 25.1%, respectively, of the total diurnal variability of IMERG; 2) the spatiotemporal pattern of the first EOF mode infers that the observed diurnal rainfall was mainly characterized by an afternoon rainfall peak at 1700 LT over most land areas and an early morning peak at 0200 LT over most oceanic areas; and 3) the combination of the first two EOF modes hints at a westward propagation of summer diurnal rainfall over Luzon and nearby oceans (Riley Dellaripa et al. 2020), and a movement of maximum diurnal rainfall from the coastal regions to the inner land areas of South China (Chen et al. 2018). Notably, the combination of the first two EOF modes explains more than 97% of the total variability of the observed diurnal rainfall, and thus can be used as a reference to evaluate the performance of model simulations (Huang and Wang 2017).
Spatial patterns of the (a) first and (b) second principal mode of the EOF analysis applied on the 3-hourly rainfall extracted from IMERG and averaged during 2000–14 JJAs. In (a) and (b), the eigenvector has been normalized by the maximum value in the analyzed domain. (c) Plot of the associated temporal pattern of the first (red line) and second (blue line) EOF mode of IMERG. (d)–(f) As in (a)–(c), respectively, but in terms of the EOF results of ensemble mean of the 10 CMIP6 models (EN_CMIP6). The percentage indicates the extent of the total diurnal rainfall variability explained by the EOF modes.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Spatial patterns of the (a) first and (b) second principal mode of the EOF analysis applied on the 3-hourly rainfall extracted from IMERG and averaged during 2000–14 JJAs. In (a) and (b), the eigenvector has been normalized by the maximum value in the analyzed domain. (c) Plot of the associated temporal pattern of the first (red line) and second (blue line) EOF mode of IMERG. (d)–(f) As in (a)–(c), respectively, but in terms of the EOF results of ensemble mean of the 10 CMIP6 models (EN_CMIP6). The percentage indicates the extent of the total diurnal rainfall variability explained by the EOF modes.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Spatial patterns of the (a) first and (b) second principal mode of the EOF analysis applied on the 3-hourly rainfall extracted from IMERG and averaged during 2000–14 JJAs. In (a) and (b), the eigenvector has been normalized by the maximum value in the analyzed domain. (c) Plot of the associated temporal pattern of the first (red line) and second (blue line) EOF mode of IMERG. (d)–(f) As in (a)–(c), respectively, but in terms of the EOF results of ensemble mean of the 10 CMIP6 models (EN_CMIP6). The percentage indicates the extent of the total diurnal rainfall variability explained by the EOF modes.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
On the basis of the comparison of these observed features (Figs. 2a–c) with the simulated features estimated from EN_CMIP6 (i.e., ensemble of all 10 CMIP6 models; Figs. 2d–f), EN_CMIP6 demonstrated the capability of depicting the variability of the first (71.9%) and second (25.4%) EOF modes. The model-to-model variations in the explained variance of the first two EOF modes are provided in Table S1 in the supplemental material. Overall, all individual models showed more than 95% of the total diurnal rainfall variability was explained by the combination of the first two EOF modes, while their ensemble (i.e., EN_CMIP6) demonstrated the explained variability closest to IMERG. Spatially, the first EOF mode of EN_CMIP6 represents the land–sea contrast, with higher diurnal rainfall variability over the land areas than over the oceanic areas; this is similar to IMERG. However, the second EOF mode of EN_CMIP6 shows a spatial pattern that differs from IMERG, particularly over Luzon, where a maximum center is observed over the western coastal region for IMERG (Fig. 2b), but it is observed over the eastern coastal region for EN_CMIP6 (Fig. 2e). This spatial difference implies that EN_CMIP6 could not represent the westward propagation of diurnal rainfall features over Luzon correctly.
Temporally, the first EOF mode of EN_CMIP6 depicts the maximum diurnal rainfall at 1400 LT and minimum diurnal rainfall at 2300 LT (Fig. 2f); these are approximately 3 h earlier than the occurrence timing of maximum/minimum diurnal rainfall in the first EOF mode of IMERG (Fig. 2c). Similarly, the maximum and minimum values of the second EOF mode of EN_CMIP6 (Fig. 2f) peak approximately 3 h earlier than the second EOF mode of IMERG (Fig. 2c). Notably, Huang and Wang (2017) reported similar bias in time shifting during the examination of the EOF analysis of diurnal rainfall variation over Southeast Asia from the ensemble mean of 18 CMIP5 models (hereinafter EN_CMIP5). However, when compared with the EOF results of EN_CMIP5 [see Fig. 2a of Huang and Wang (2017); not shown], the EOF results of EN_CMIP6 in this study (Fig. 2d), revealed more details of the spatial characteristics. For example, EN_CMIP6 can demonstrate the day–night contrast for the land–ocean diurnal rainfall maximum over Southeast Asia for the first EOF mode, while EN_CMIP5 cannot. This difference might be partially attributed to the original spatial resolution of EN_CMIP6 (approximately 1.26° in longitude, 1.19° in latitude; estimated from the average of 10 CMIP6 models) used in this study being finer than that of EN_CMIP5 (approximately 2.21° in longitude, 1.94° in latitude; estimated from the average of 18 CMIP5 models) used by Huang and Wang (2017).
Huang and Wang (2017) indicated that despite the bias seen in EN_CMIP5, there is one model, CMCC-CM, with a spatial resolution of 0.75° × 0.75° capable of depicting the accurate occurrence timing of maximum diurnal rainfall over the land areas of Southeast Asia. Therefore, in addition to IMERG and EN_CMIP6, we constructed a phase diagram for individual CMIP6 models (Fig. 3a) to justify their capabilities to depict the occurrence timing of the maximum diurnal rainfall. Overall, the maximum diurnal rainfall for IMERG and EN_CMIP6 occurs at 1700 and 1400 LT, respectively, over most land areas (Fig. 3a). These features are consistent with our earlier discussions (Fig. 2). Notably, a similar bias that the maximum diurnal rainfall occurred earlier in EN_CMIP6 than in IMERG is observed in not only Southeast Asia, but in multiple regions of the globe (Fig. S1 in the online supplemental material). However, one should also note that although IMERG is capable of depicting the summer diurnal rainfall phase over Southeast Asia (Huang et al. 2020), it has limitations in representing the diurnal rainfall phase correctly in some regions of the world (O and Kirstetter 2018). Therefore, the phase differences between EN_CMIP6 and IMERG (Fig. S1c) might not be similar to the phase differences between EN_CMIP6 and rain gauge observations in other regions of the world.
(a) Phase diagram depicting the timing of the occurrence of diurnal rainfall maximum averaged during 2000–14 JJAs, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models (numbered 1–10). The hatched areas in (a) indicate that the phase difference between model and OBS exceeds 6 h. (b) Plot of the phase differences [from (a)] estimated from model minus OBS area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for each of the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Phase diagram depicting the timing of the occurrence of diurnal rainfall maximum averaged during 2000–14 JJAs, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models (numbered 1–10). The hatched areas in (a) indicate that the phase difference between model and OBS exceeds 6 h. (b) Plot of the phase differences [from (a)] estimated from model minus OBS area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for each of the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Phase diagram depicting the timing of the occurrence of diurnal rainfall maximum averaged during 2000–14 JJAs, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models (numbered 1–10). The hatched areas in (a) indicate that the phase difference between model and OBS exceeds 6 h. (b) Plot of the phase differences [from (a)] estimated from model minus OBS area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for each of the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
In addition, a global analysis of the difference between individual models and IMERG shows that EC-Earth3 and EC-Earth3-Veg outperform other models in capturing the diurnal rainfall phase over East Asia (Fig. S2 in the online supplemental material). When focusing on the regional feature (Fig. 3a), only EC-Earth3 and EC-Earth3-Veg could successfully represent the maximum diurnal rainfall occurring at 1700 LT over South China, Taiwan, and Luzon. Consequently, no phase differences are observed between these two models and IMERG for the three focused land areas (Fig. 3b). According to the previous studies (Kusunoki and Arakawa 2015; Covey et al. 2016; Kim et al. 2019; Xin et al. 2020), the errors of models in depicting the rainfall variations might be partially caused by the differences in the spatial resolution or other factors (e.g., the convection scheme) used in the models. Since EC-Earth3 and EC-Earth3-Veg have the finest spatial resolution among the 10 models (Table 1), we believe that an increase in spatial resolution adds values to increase its ability to simulate the diurnal rainfall features. More discussion of this suggestion will be given later in section 6.
Conversely, as shown in Fig. 3, BCC-CSM2-MR has the largest bias among the examined models. The error of BCC-CSM2-MR (Fig. 3) is consistent with that shown by Wu et al. (2019), supporting the robustness of our discussions for BCC-CSM2-MR. However, we would like to point out that BCC-CSM2-MR does not have the lowest spatial resolution among the 10 models (Table 1). Therefore, other factors (e.g., the convection scheme) apart from the spatial resolution might contribute to the comparatively higher errors in BCC-CSM2-MR than in other models for the simulation of diurnal rainfall over Southeast Asia. This suggestion, however, requires additional verification in the future (e.g., performing sensitivity testing of convective parameterizations).
4. Present-day simulation of CAR activities
It is hypothesized in this study that the model with comparatively more accurate representation of the diurnal rainfall phase performs better in capturing the occurrence frequency of CAR events. To verify this hypothesis, we examined the ability of the model to represent the climatological distribution of CAR frequency averaged over the 2000–14 JJA seasons (Fig. 4a). The details of identification of CAR events have been presented in section 2. As shown in Fig. 4a, a higher CAR frequency (units: days per JJA) is revealed over the land areas rather than over the oceanic areas. This is consistent with the fact that the diurnal rainfall formation observed by IMERG tends to peak in the afternoon over most land areas, but not over the oceanic areas (Fig. 3a). Quantitatively, by comparing EN_CMIP6 and IMERG shown in Fig. 4a, the former is found to be less frequent than the latter over most land areas. Such a problem in underestimating the CAR frequency over Southeast Asia can be observed in most studied models, expect that EC-Earth3 and EC-Earth3-Veg seem to have CAR frequency higher than those estimated by IMERG over Taiwan and Luzon.
(a) Occurrence of convective afternoon rainfall (CAR) frequency averaged over the 2000–14 JJA seasons, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models. (b) The mean ratios (MR; =model/OBS) [from (a)] of CAR frequency area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Occurrence of convective afternoon rainfall (CAR) frequency averaged over the 2000–14 JJA seasons, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models. (b) The mean ratios (MR; =model/OBS) [from (a)] of CAR frequency area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Occurrence of convective afternoon rainfall (CAR) frequency averaged over the 2000–14 JJA seasons, estimated from IMERG (i.e., OBS), EN_CMIP6, and each of the 10 models. (b) The mean ratios (MR; =model/OBS) [from (a)] of CAR frequency area-averaged over South China (yellow bars), Taiwan (purple bars), and Luzon (red bars) for the 10 models.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
A better illustration of the bias exhibited by CMIP6 models during the underestimation of the occurrence of CAR events over South China, Taiwan, and Luzon is shown in Fig. 4b. Based on the values of the mean ratio between the simulated and observed CAR frequency, we further noted that overall EC-Earth3 and EC-Earth3-Veg rank as the top two models (while ACCESS-CM2, BCC-CSM2-MR, and MIROC6 rank as the bottom three models) for representing the CAR frequency over the land areas of Southeast Asia. These features roughly correspond to their ability to depict the diurnal rainfall evolutions (Fig. 3), suggesting that better skill in depicting the diurnal rainfall evolutions might lead to better skill in capturing the CAR frequency. Despite the magnitude differences (Fig. 4), most examined models are capable of qualitatively representing higher CAR frequency over land than over ocean; this feature is similar to what is observed by IMERG.
In addition to the CAR frequency, the contribution of CAR amount to the total rainfall amount (hereinafter CAR amount percentage) is another variable that has been frequently examined in previous studies (Wang and Chen 2008; Huang et al. 2016b, 2020). To understand the ability of the model to represent this variable, we constructed the climatological distribution of CAR amount percentage averaged during 2000–14 JJAs (Fig. 5a). Overall, Fig. 5a indicates that 1) when compared with IMERG, EN_CMIP6 underestimates the CAR amount percentage over the land areas, and 2) among the 10 models, EC-Earth3 and EC-Earth3-Veg rank highest (i.e., exhibit the best performance), while ACCESS-CM2, BCC-CSM2-MR, and MIROC6 rank lowest, for quantitatively representing the CAR amount percentage over South China, Taiwan, and Luzon (see also Fig. 5b). Therefore, one can conclude from Figs. 3–5 that EC-Earth3 and EC-Earth3-Veg are the best two models in representing diurnal rainfall evolution as well as the CAR characteristics (including frequency and amount) over the land areas of Southeast Asia.
As in Fig. 4, but for contribution of CAR amount to total summer rainfall amount [i.e., CAR amount percentage = (CAR amount/total summer rainfall amount) × 100%].
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 4, but for contribution of CAR amount to total summer rainfall amount [i.e., CAR amount percentage = (CAR amount/total summer rainfall amount) × 100%].
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 4, but for contribution of CAR amount to total summer rainfall amount [i.e., CAR amount percentage = (CAR amount/total summer rainfall amount) × 100%].
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
One might have a question about the models’ ability to simulate CAR intensity (=CAR amount/CAR frequency). To clarify this issue, we constructed the spatial distribution of climatological CAR intensity averaged during 2000–14 JJAs in Fig. 6a for both observation and models. Notably, since the CAR frequency over the oceanic areas in the CMIP6 models is too low (Fig. 4a) to reasonably calculate the CAR intensity, we demonstrated CAR intensity only over the land areas (Fig. 6a). The associated mean ratio between simulated and observed CAR intensity over the three selected land areas is shown in Fig. 6b. Overall, most models tend to underestimate the CAR intensity; this is similar to the results shown in Figs. 4 and 5. However, from Fig. 6, EC-Earth3 and EC-Earth3-Veg cannot be identified as the best models for quantitatively depicting the CAR intensity. Moreover, based on the comparison of Fig. 6 with Figs. 4 and 5, we noted that features exhibited in the CAR amount percentage (Fig. 5) are similar to those in CAR frequency (Fig. 4), but differ from those in CAR intensity (Fig. 6). This implies that the applicability of these models to simulate CAR amount over Southeast Asia was mostly determined by its ability to simulate the CAR frequency in comparison with its ability to simulate the CAR intensity. It will be interesting to examine whether future changes in CAR amounts would be mainly dominated by future changes in CAR frequency and less dominated by CAR intensity; this is presented in the next section.
As in Fig. 5, but for CAR intensity (millimeters per CAR day).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 5, but for CAR intensity (millimeters per CAR day).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 5, but for CAR intensity (millimeters per CAR day).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
5. Projected future changes of CAR activities
To investigate the uncertainty of future changes, we compared the percentage of projected changes for the selected variables derived from two different ensembles: EN_CMIP6 and EN_ECearth3 (representing the ensemble mean of EC-Earth3 and EC-Earth3-Veg, i.e., the top two models identified from Figs. 3–5). Figures 7a and 7b show the projected changes in CAR frequency estimated from EN_CMIP6 and EN_ECearth3, respectively. Here, only changes over land areas are presented and discussed. Overall, both EN_CMIP6 and EN_ECearth3 projects that CAR frequency over Southeast Asia will be suppressed in the future relative to the present. In particular, over the coastal area of South China and most of Taiwan, the decrease in CAR frequency estimated from EN_ECearth3 can exceed the present CAR frequency by 20%. To verify whether the projected changes in EN_CMIP6 are mainly dominated by EN_ECearth3, we further constructed boxplots using the projected changes estimated from each of the 10 CMIP6 models in the three study areas (Fig. 7c). The red and blue marks in Fig. 7c represent the area-averaged values estimated from EN_CMIP6 and EN_ECearth3, respectively. Notably, despite the magnitude differences, more than 6 of the 10 models project that the CAR frequency over South China, Taiwan, and Luzon will reduce in the future (Fig. 7c and Table 2). This suggests that the changes presented in Fig. 7a are a common feature in the CMIP6 models.
Projected changes {%; =[(future − present)/present] × 100%} of CAR frequency estimated from simulations over two selected eras: future (2071–2100 JJAs) and present (1985–2014 JJAs) for (a) EN_CMIP6 and (b) the ensemble mean of EC-Earth3 and EC-Earth3-Veg (hereinafter EN_ECearth3). In (a) and (b), the areas with changes that are significant at the 90% confidence interval are hatched. (c) Boxplot illustrating the range of projected changes estimated by 10 CMIP6 models for CAR frequency area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ). In (c), red and blue marks represent the values of EN_CMIP6 and EN_ECearth3, respectively. Also shown are the corresponding results for the projected changes in the CAR (d)–(f) amount and (g)–(i) intensity.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Projected changes {%; =[(future − present)/present] × 100%} of CAR frequency estimated from simulations over two selected eras: future (2071–2100 JJAs) and present (1985–2014 JJAs) for (a) EN_CMIP6 and (b) the ensemble mean of EC-Earth3 and EC-Earth3-Veg (hereinafter EN_ECearth3). In (a) and (b), the areas with changes that are significant at the 90% confidence interval are hatched. (c) Boxplot illustrating the range of projected changes estimated by 10 CMIP6 models for CAR frequency area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ). In (c), red and blue marks represent the values of EN_CMIP6 and EN_ECearth3, respectively. Also shown are the corresponding results for the projected changes in the CAR (d)–(f) amount and (g)–(i) intensity.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Projected changes {%; =[(future − present)/present] × 100%} of CAR frequency estimated from simulations over two selected eras: future (2071–2100 JJAs) and present (1985–2014 JJAs) for (a) EN_CMIP6 and (b) the ensemble mean of EC-Earth3 and EC-Earth3-Veg (hereinafter EN_ECearth3). In (a) and (b), the areas with changes that are significant at the 90% confidence interval are hatched. (c) Boxplot illustrating the range of projected changes estimated by 10 CMIP6 models for CAR frequency area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ). In (c), red and blue marks represent the values of EN_CMIP6 and EN_ECearth3, respectively. Also shown are the corresponding results for the projected changes in the CAR (d)–(f) amount and (g)–(i) intensity.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Number of models showing similar future projections for CAR activities.
Corresponding to Figs. 7a–c, Figs. 7d–f show the percentage of projected changes in CAR amount estimated from the CMIP6 models. Overall, both EN_CMIP6 (Fig. 7d) and EN_ECearth3 (Fig. 7e) project that the CAR amount would reduce in the future over most of South China, Taiwan, and Luzon. Among the 10 models, more than 7 models show similar projection of decrease in CAR amount in the future (Fig. 7f and Table 2). Visually, the projected changes in Figs. 7d–f seem to follow the trends of projected changes observed in Figs. 7a–c. However, the CAR amount was calculated based on the CAR frequency and CAR intensity (CAR amount = CAR frequency × CAR intensity). Thus, the next question with regard to the projected changes in CAR intensity needs to be answered. It is noted that EN_CMIP6 (Fig. 7g), EN_ECearth3 (Fig. 7h), and more than 8 of the 10 models (Fig. 7i and Table 2) project that the CAR intensity would increase over most of South China, Taiwan, and Luzon. These features of CAR intensity (Figs. 7g–i) are opposites to those of CAR frequency (Figs. 7a–c) and CAR amount (Figs. 7d–f), implying that the projected changes in CAR amount are mainly dominated by the projected changes in CAR frequency rather than the projected changes in CAR intensity.
The changes in CAR activities using CMIP6 models presented above are consistent with the earlier findings by Huang et al. (2016a,b) using other models, contributing to the robustness of our suggestions pertaining to Southeast Asia. What would be the cause for these projected changes? Studying the maintenance mechanisms of CAR activities over Taiwan over the past several decades, earlier observational studies (Huang and Chen 2015; Huang et al. 2019) have found that the changes in CAR frequency and CAR intensity are mainly dominated by different atmospheric factors. In general, changes in the afternoon sea-breeze convergence and daily thermal instability are two of the major causes for explaining the changes in CAR frequency, while the changes in CAR intensity are mainly determined by the changes in the local atmospheric humidity (Huang and Chen 2015; Huang et al. 2019). Based on these concepts, we hypothesized that the projected changes in CAR activities observed in CMIP6 models might also be a response to the projected changes in the associated maintenance mechanisms. This hypothesis is clarified in the next section.
6. Discussion
a. Mechanisms responsible for the projected changes in CAR activities
Figure 8 shows the horizontal distribution of the simulation (Fig. 8a) and projected changes (Fig. 8b) for afternoon wind convergence, extracted from EN_CMIP6 at 1400 LT. Overall, the magnitude of the projected changes (Fig. 8b) is one order smaller than the present-day simulation (Fig. 8a), with a spatial distribution approximately contrasting to the present-day simulation. This implies that the wind convergence over land in the afternoon is projected to be suppressed in the future; EN_ECearth3 (blue mark in Fig. 8c) and most of the examined models (boxplot in Fig. 8c) yield a similar projection. According to Huang and Chen (2015), this decrease in the local wind convergence can be unfavorable for initializing local CAR formation. The causes of suppression of local wind convergence and sea-breeze circulation might be attributed to the decrease in the diurnal surface temperature range over land areas in a warm climate (Stone and Weaver 2003). This kind of change in surface temperature implies a weakening of daytime land–sea thermal contrast, which can further weaken the daytime sea-breeze circulation and related surface convergence (Huang et al. 2016b). Conversely, weakening of the daytime surface convergence because of decrease in CAR frequency is also possible.
(a) Diurnal anomalies (i.e., daily mean removed) of surface wind convergence [denoted as (−∇ · Vs)] at 1400 LT estimated from EN_ECearth3 for the present-day simulation averaged during 1985–2014 JJAs. (b) As in (a), but for the differences of (−∇ · Vs) at 1400 LT between the future (2071–2100 JJAs) and the present (1985–2014 JJAs). Areas with changes that are significant at the 90% confidence interval are hatched. (c) As in Fig 7c, but showing the range of changes of (b) estimated by 10 CMIP6 models. (d)–(f) As in (a)–(c), but for the result of vertical thermal instability estimated from the differences between daily mean of atmospheric temperature at 600 hPa and the surface, following Huang et al. (2016a).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Diurnal anomalies (i.e., daily mean removed) of surface wind convergence [denoted as (−∇ · Vs)] at 1400 LT estimated from EN_ECearth3 for the present-day simulation averaged during 1985–2014 JJAs. (b) As in (a), but for the differences of (−∇ · Vs) at 1400 LT between the future (2071–2100 JJAs) and the present (1985–2014 JJAs). Areas with changes that are significant at the 90% confidence interval are hatched. (c) As in Fig 7c, but showing the range of changes of (b) estimated by 10 CMIP6 models. (d)–(f) As in (a)–(c), but for the result of vertical thermal instability estimated from the differences between daily mean of atmospheric temperature at 600 hPa and the surface, following Huang et al. (2016a).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Diurnal anomalies (i.e., daily mean removed) of surface wind convergence [denoted as (−∇ · Vs)] at 1400 LT estimated from EN_ECearth3 for the present-day simulation averaged during 1985–2014 JJAs. (b) As in (a), but for the differences of (−∇ · Vs) at 1400 LT between the future (2071–2100 JJAs) and the present (1985–2014 JJAs). Areas with changes that are significant at the 90% confidence interval are hatched. (c) As in Fig 7c, but showing the range of changes of (b) estimated by 10 CMIP6 models. (d)–(f) As in (a)–(c), but for the result of vertical thermal instability estimated from the differences between daily mean of atmospheric temperature at 600 hPa and the surface, following Huang et al. (2016a).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
In addition, Figs. 8d–f show that the low-level thermal instability over the land areas of Southeast Asia is projected to be suppressed in the future; this feature is revealed in all 10 models. A possible cause for the weakening of low-level thermal instability might be that mountains at high elevation are found to be warming faster than the low plain areas under a warming climate (Huang et al. 2016b). The associated vertical temperature change implies a decrease in the thermal instability, which is known to be unfavorable for initiating local CAR formation (Huang et al. 2019). When the features in Fig. 8 were considered together, the projected changes in low-level wind convergence and thermal instability could explain the projection of decrease in CAR frequency in the future.
With regard to the projected changes in the maintenance mechanism of CAR intensity, Figs. 9a and 9b indicate that the vertically integrated specific humidity is projected to increase over Southeast Asia in the future than in the present. This trend was observed in all the studied models (Fig. 9c), suggesting that the projected increase in CAR intensity might be attributed to the increase in humidity in the future. Other studies have also suggested that the convective intensity increases in the presence of more specific humidity or water vapor (Kendon et al. 2010; Chou et al. 2012).
As in Fig. 8, but for the vertically integrated humidity (denoted as vintq) between the surface and 300 hPa.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 8, but for the vertically integrated humidity (denoted as vintq) between the surface and 300 hPa.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
As in Fig. 8, but for the vertically integrated humidity (denoted as vintq) between the surface and 300 hPa.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Notably, the maintenance mechanisms examined above can only qualitatively explain the projected changes in CAR activities. For example, Fig. 7c indicates that the decrease in CAR frequency is less evident over Luzon than the other two regions. This is inconsistent with the regional differences observed in Figs. 8c and 8f. In addition, Fig. 7i shows that the uncertainty of projected changes in CAR intensity is more evident over Taiwan than the other two regions; however, this is not observed in the projected changes in humidity (Fig. 9c). A possible explanation for these differences may be that not all factors affecting future changes in CAR activities are being examined in this study, leaving scope for further research.
b. Comparison of EC-Earth models from CMIP5 with CMIP6
As reported in sections 3 and 4, EC-Earth3 and EC-Earth3-Veg exhibit the best performance in depicting diurnal rainfall evolution, CAR frequency, and CAR amount percentage over Southeast Asia. We suggest that an increase in the spatial resolution and an improvement in parameterization might be the primary reasons for supplementing this finding. To support this suggestion, we examined the performance of the earlier version of EC-Earth3 and EC-Earth3-Veg from CMIP5 (i.e., EC-Earth) in simulating the diurnal rainfall evolution (Fig. 10a), CAR frequency (Fig. 10b), and CAR amount percentage (Fig. 10c) during 2000–09 JJAs. Additionally, we compared the performances of EC-Earth with those of EC-Earth3 and EC-Earth3-Veg in Figs. 10d–f. The analysis period was set to 2000–09 JJAs because observational data of IMERG (i.e., the reference base) are available since 2000 and historical simulation data of EC-Earth end in 2009.
(a) Phase diagram depicting the timing of diurnal rainfall maximum, (b) CAR frequency, and (c) CAR amount percentage, estimated from EC-Earth averaged during 2000–09 JJAs. In (a), the hatched areas indicate that the phase difference between EC-Earth and IMERG exceeds 6 h. (d) Phase differences [model − IMERG (i.e., OBS)] between the indicated models and the OBS area-averaged over SC, TW, and LZ. (e),(f) As in (d), but for the ratios (model/OBS) of CAR frequency and amount percentage, respectively.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Phase diagram depicting the timing of diurnal rainfall maximum, (b) CAR frequency, and (c) CAR amount percentage, estimated from EC-Earth averaged during 2000–09 JJAs. In (a), the hatched areas indicate that the phase difference between EC-Earth and IMERG exceeds 6 h. (d) Phase differences [model − IMERG (i.e., OBS)] between the indicated models and the OBS area-averaged over SC, TW, and LZ. (e),(f) As in (d), but for the ratios (model/OBS) of CAR frequency and amount percentage, respectively.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
(a) Phase diagram depicting the timing of diurnal rainfall maximum, (b) CAR frequency, and (c) CAR amount percentage, estimated from EC-Earth averaged during 2000–09 JJAs. In (a), the hatched areas indicate that the phase difference between EC-Earth and IMERG exceeds 6 h. (d) Phase differences [model − IMERG (i.e., OBS)] between the indicated models and the OBS area-averaged over SC, TW, and LZ. (e),(f) As in (d), but for the ratios (model/OBS) of CAR frequency and amount percentage, respectively.
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Figure 10a shows that in EC-Earth, the diurnal rainfall maximum over South China, Taiwan, and Luzon occurred during 0800–1400 LT. By comparing models (including EC-Earth, EC-Earth3, and EC-Earth3-Veg) and IMERG, we found that EC-Earth depicts the maximum diurnal rainfall earlier than IMERG, while the occurrence timing of the maximum diurnal rainfall simulated by EC-Earth3 and EC-Earth3-Veg is similar to IMERG. Corresponding to the performance in the diurnal rainfall phase, EC-Earth3 and EC-Earth3-Veg also show less error than EC-Earth in presenting the CAR frequency (Figs. 10b,e) and CAR amount percentage (Figs. 10c,f) in the present day. Different from EC-Earth3 and EC-Earth3-Veg with spatial resolution of 0.7° × 0.7°, EC-Earth has a spatial resolution of 1.125° × 1.125°. This difference implies that an increase in the spatial resolution allows EC-Earth3 and EC-Earth3-Veg to better simulate diurnal rainfall evolution, CAR frequency, and CAR amount percentage over Southeast Asia. Notably, the EC-Earth model from CMIP5 to CMIP6 not only has increased the horizontal and vertical resolution, but also has substantially changed the representation of aerosols and introduced a sophisticated description of aerosol indirect effects (Wyser et al. 2020). Therefore, improvement in parameterization will also assist EC-Earth3 and EC-Earth3-Veg to better simulate regional diurnal rainfall features.
Figure 11 indicates that EC-Earth3 and EC-Earth3-Veg perform better than EC-Earth in terms of quantitatively estimating CAR intensity over South China and Taiwan, but not Luzon. This suggests that an increase in the spatial resolution or an improvement in parameterization of EC-Earth3 and EC-Earth3-Veg does not guarantee a better simulation of CAR intensity over the land areas of Southeast Asia. Additionally, this also partially explains why EC-Earth3 and EC-Earth3-Veg, having the finest spatial resolution among the 10 models (Table 1), do not exhibit the best performance in the simulation of CAR intensity (Fig. 6). Further studies are suggested in the future to improve the ability of EC-Earth3 and EC-Earth3-Veg in simulating the CAR intensity over Luzon.
Similar to Fig. 10, but for (a) the present-day simulation of CAR intensity estimated from EC-Earth averaged during 2000–09 JJAs, and (b) the related ratio (i.e., model/OBS) between selected models (including EC-Earth, EC-Earth3, and EC-Earth3-Veg) and OBS (i.e., IMERG) area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Similar to Fig. 10, but for (a) the present-day simulation of CAR intensity estimated from EC-Earth averaged during 2000–09 JJAs, and (b) the related ratio (i.e., model/OBS) between selected models (including EC-Earth, EC-Earth3, and EC-Earth3-Veg) and OBS (i.e., IMERG) area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
Similar to Fig. 10, but for (a) the present-day simulation of CAR intensity estimated from EC-Earth averaged during 2000–09 JJAs, and (b) the related ratio (i.e., model/OBS) between selected models (including EC-Earth, EC-Earth3, and EC-Earth3-Veg) and OBS (i.e., IMERG) area-averaged over South China (SC), Taiwan (TW), and Luzon (LZ).
Citation: Journal of Climate 34, 12; 10.1175/JCLI-D-20-0788.1
7. Conclusions
This study assesses the present-day simulation and future projection of summer CAR activities (including frequency, amount, and intensity) over Southeast Asia from 10 CMIP6 models. For the present-day simulation, most models (except EC-Earth3 and EC-Earth3-Veg) exhibit a similar bias in depicting the timing of diurnal rainfall maximum earlier over the land areas while using IMERG as a reference. Among the 10 models, EC-Earth3 and EC-Earth3-Veg demonstrate the best skill in depicting the diurnal rainfall evolution, CAR frequency, and CAR amount percentage. This might be attributed to the fact that EC-Earth3 and EC-Earth3-Veg have the finest spatial resolution among the examined models. Analyses also show that relative to EC-Earth (the CMIP5 version), an increase in spatial resolution and an improvement in parameterization help EC-Earth3 and EC-Earth3-Veg (the CMIP6 version) to better simulate the local diurnal rainfall evolution, CAR frequency, and CAR amount percentage. As for the simulation of CAR intensity, EC-Earth3 and EC-Earth3-Veg do not always outperform their earlier version or other CMIP6 models. This information should be considered while using these models to study local weather changes over Southeast Asia.
In addition, by examining the differences between the future (2071–2100, SSP858) and the present (1985–2014, historical), we found that the frequency of CAR events will decrease, but its intensity will increase over Southeast Asia. The projected changes in CAR frequency and CAR intensity are the major and minor dominant factors, respectively, affecting future changes in CAR amount, which is projected to suppress in the future. Results also show that the projected decrease in local wind convergence and thermal instability are two possible mechanisms leading to the projected decrease in CAR frequency, while the projected increase in moisture may contribute to the projected increase in CAR intensity. However, these maintenance mechanisms can only qualitatively but not quantitatively explain the projected changes in CAR activities. This implies that there might be other factors contributing to future changes in CAR activities not being considered in this study; this deserves further research attention.
Note that, to the best knowledge of the authors, this is the first study to assess the simulation and future projection of CAR activities over Southeast Asia using CMIP6 models. Our findings highlight that CMIP6 can provide valuable information for projecting local weather changes over Southeast Asia. As there are many other model simulations from the CMIP6-Endorsed Model Intercomparison Project (e.g., the Global Monsoons Model Intercomparison Project, High-Resolution Model Intercomparison Project, etc.) (Eyring et al. 2016), further studies are planned in the future to assess the ability of these models to simulate the local weather/climate changes over Southeast Asia.
Acknowledgments
Model data provided by CMIP (phases 5 and 6) and IMERG data provided by NASA are highly appreciated. This study was supported by the Ministry of Science and Technology of Taiwan under MOST 106-2628-M-003-001-MY4, MOST 108-2625-M-003-002, and MOST 109-2625-M-003-002. Author L. Deng was supported by the National Natural Science Foundation of China (Grant 41875071), “Yangfan” Talent Project of Guangdong Province (Grant 000001005) and Doctoral Fund of Guangdong Ocean University (Grant R17002).
REFERENCES
Bowden, J. H., and Coauthors, 2020: High-resolution dynamically downscaled rainfall and temperature projections for ecological life zones within Puerto Rico and for the U.S. Virgin Islands. Int. J. Climatol., 41, 1305–1327, https://doi.org/10.1002/joc.6810.
Bowman, K. P., J. C. Collier, G. R. North, Q. Wu, E. Ha, and J. Hardin, 2005: Diurnal cycle of tropical precipitation in Tropical Rainfall Measuring Mission (TRMM) satellite and ocean buoy rain gauge data. J. Geophys. Res., 110, D21104, https://doi.org/10.1029/2005JD005763.
Cao, J., and Coauthors, 2018: The NUIST Earth System Model (NESM) version 3: Description and preliminary evaluation. Geosci. Model Dev., 11, 2975–2993, https://doi.org/10.5194/gmd-11-2975-2018.
Chen, G., T. Iwasaki, H. Qin, and W. Sha, 2014: Evaluation of the warm-season diurnal variability over East Asia in recent reanalyses JRA-55, ERA-Interim, NCEP CFSR, and NASA MERRA. J. Climate, 27, 5517–5537, https://doi.org/10.1175/JCLI-D-14-00005.1.
Chen, G., R. Lan, W. Zeng, H. Pan, and W. Li, 2018: Diurnal variations of rainfall in surface and satellite observations at the monsoon coast (South China). J. Climate, 31, 1703–1724, https://doi.org/10.1175/JCLI-D-17-0373.1.
Chiang, C. C., L. H. Hsu, T. J. Lee, and Y. C. Yu, 2019: Meteorological analysis of heavy rain events in Taoyuan and Taipei on July 2, 2019 (in Chinese). National Science and Technology Center for Disaster Reduction Newsletter, 16 pp., https://www.ncdr.nat.gov.tw/Files/NewsLetter/New/20190813100513.pdf.
Chou, C., C. A. Chen, P. H. Tan, and T. K. Chen, 2012: Mechanisms for global warming impacts on precipitation frequency and intensity. J. Climate, 25, 3291–3306, https://doi.org/10.1175/JCLI-D-11-00239.1.
Collier, J. C., and K. P. Bowman, 2004: Diurnal cycle of tropical precipitation in a general circulation model. J. Geophys. Res., 109, D17105, https://doi.org/10.1029/2004JD004818.
Covey, C., P. J. Gleckler, C. Doutriaux, D. N. Williams, A. Dai, J. Fasullo, K. Trenberth, and A. Berg, 2016: Metrics for the diurnal cycle of precipitation: Toward routine benchmarks for climate models. J. Climate, 29, 4461–4471, https://doi.org/10.1175/JCLI-D-15-0664.1.
Dai, A., 2001: Global precipitation and thunderstorm frequencies. Part II: Diurnal variations. J. Climate, 14, 1112–1128, https://doi.org/10.1175/1520-0442(2001)014<1112:GPATFP>2.0.CO;2.
Dai, A., 2006: Precipitation characteristics in eighteen coupled climate models. J. Climate, 19, 4605–4630, https://doi.org/10.1175/JCLI3884.1.
Dai, A., and K. E. Trenberth, 2004: The diurnal cycle and its depiction in the Community Climate System Model. J. Climate, 17, 930–951, https://doi.org/10.1175/1520-0442(2004)017<0930:TDCAID>2.0.CO;2.
Dix, M., and Coauthors, 2019: CSIRO-ARCCSS ACCESS-CM2 model output prepared for CMIP6 CMIP historical, version 20191213. Earth System Grid Federation, accessed 13 December 2019, https://doi.org/10.22033/ESGF/CMIP6.4271.
EC-Earth, 2019a: EC-Earth-Consortium EC-Earth3 model output prepared for CMIP6 CMIP historical, version 20191213. Earth System Grid Federation, accessed 13 December 2019, https://doi.org/10.22033/ESGF/CMIP6.4700.
EC-Earth, 2019b: EC-Earth-Consortium EC-Earth3-Veg model output prepared for CMIP6 CMIP historical, version 20191213. Earth System Grid Federation, accessed 13 December 2019,https://doi.org/10.22033/ESGF/CMIP6.4706.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fu, Y., Z. Lin, and D. Guo, 2020: Improvement of the simulation of the summer East Asian westerly jet from CMIP5 to CMIP6. Atmos. Oceanic Sci. Lett., 13, 550–558, https://doi.org/10.1080/16742834.2020.1746175.
Gray, W. M., and R. W. Jacobson Jr., 1977: Diurnal variation of deep cumulus convection. Mon. Wea. Rev., 105, 1171–1188, https://doi.org/10.1175/1520-0493(1977)105<1171:DVODCC>2.0.CO;2.
Gusain, A., S. Ghosh, and S. Karmakar, 2020: Added value of CMIP6 over CMIP5 models in simulating Indian summer monsoon rainfall. Atmos. Res., 232, 104680, https://doi.org/10.1016/j.atmosres.2019.104680.
Hannachi, A., I. T. Jolliffe, and D. B. Stephenson, 2007: Empirical orthogonal functions and related techniques in atmospheric science: A review. Int. J. Climatol., 27, 1119–1152, https://doi.org/10.1002/joc.1499.
Hirose, M., R. Oki, S. Shimizu, M. Kachi, and T. Higashiuwatoko, 2008: Fine-scale diurnal rainfall statistics refined from eight years of TRMM PR data. J. Appl. Meteor. Climatol., 47, 544–561, https://doi.org/10.1175/2007JAMC1559.1.
Huang, W. R., and S. Y. Wang, 2014: Impact of land-sea breezes at different scales on the diurnal rainfall in Taiwan. Climate Dyn., 43, 1951–1963, https://doi.org/10.1007/s00382-013-2018-z.
Huang, W. R., and K. C. Chen, 2015: Trends in pre-summer frontal and diurnal rainfall activities during 1982–2012 over Taiwan and southeast China: Characteristics and possible causes. Int. J. Climatol., 35, 2608–2619, https://doi.org/10.1002/joc.4159.
Huang, W. R., and S. Y. Wang, 2017: Future changes in propagating and non-propagating diurnal rainfall over East Asia. Climate Dyn., 49, 375–389, https://doi.org/10.1007/s00382-016-3348-4.
Huang, W. R., Y. H. Chang, H. H. Hsu, C. T. Cheng, and C. Y. Tu, 2016a: Summer convective afternoon rainfall simulation and projection using WRF driven by global climate model. Part II: Over South China and Luzon. Terr. Atmos. Oceanic Sci., 27, 673–685, https://doi.org/10.3319/TAO.2016.05.02.02.
Huang, W. R., Y. H. Chang, H. H. Hsu, C. T. Cheng, and C. Y. Tu, 2016b: Dynamical downscaling simulation and future projection of summer rainfall in Taiwan: Contributions from different types of rain events. J. Geophys. Res. Atmos., 121, 13 973–13 988, https://doi.org/10.1002/2016JD025643.
Huang, W. R., Y. H. Chang, and P. H. Huang, 2019: Relationship between the interannual variations of summer convective afternoon rainfall activity in Taiwan and SSTA(Niño3.4) during 1961–2012: Characteristics and mechanisms. Sci. Rep., 9, 9378, https://doi.org/10.1038/s41598-019-45901-w.
Huang, W. R., P. Y. Liu, Y. H. Chang, and C. Y. Liu, 2020: Evaluation and application of satellite precipitation products in studying the summer precipitation variations over Taiwan. Remote Sens., 12, 347, https://doi.org/10.3390/rs12030347.
Huffman, G. J., D. T. Bolvin, E. J. Nelkin, E. F. Stocker, and J. Tan, 2019: V06 IMERG release notes. GPM Project Doc., 6 pp., https://pmm.nasa.gov/sites/default/files/document_files/IMERG_V06_release_notes_190503.pdf.
Kamworapan, S., and C. Surussavadee, 2019: Evaluation of CMIP5 global climate models for simulating climatological temperature and precipitation for Southeast Asia. Adv. Meteor., 2019, 1067365, https://doi.org/10.1155/2019/1067365.
Kendon, E., D. Rowell, and R. Jones, 2010: Mechanisms and reliability of future projected changes in daily precipitation. Climate Dyn., 35, 489–509, https://doi.org/10.1007/s00382-009-0639-z.
Kim, H., M. Lee, D. Cha, and W. M. Putman, 2019: Improved representation of the diurnal variation of warm season precipitation by an atmospheric general circulation model at a 10 km horizontal resolution. Climate Dyn., 53, 6523–6542, https://doi.org/10.1007/s00382-019-04943-6.
Kusunoki, S., and O. Arakawa, 2015: Are CMIP5 models better than CMIP3 models in simulating precipitation over East Asia? J. Climate, 28, 5601–5621, https://doi.org/10.1175/JCLI-D-14-00585.1.
Mauritsen, T., and Coauthors, 2019: Developments in the MPI-M Earth System Model version 1.2 (MPI-ESM1.2) and its response to increasing CO2. J. Adv. Model. Earth Syst., 11, 998–1038, https://doi.org/10.1029/2018MS001400.
Meehl, G. A., 1995: Global coupled general circulation models. Bull. Amer. Meteor. Soc., 76, 951–957, https://doi.org/10.1175/1520-0477-76.6.951.
Müller, W. A., and Coauthors, 2018: A higher-resolution version of the Max Planck Institute Earth System Model (MPI-ESM1.2-HR). J. Adv. Model. Earth Syst., 10, 1383–1413, https://doi.org/10.1029/2017MS001217.
Nesbitt, S. W., and E. J. Zipser, 2003: The diurnal cycle of rainfall and convective intensity according to three years of TRMM measurements. J. Climate, 16, 1456–1475, https://doi.org/10.1175/1520-0442-16.10.1456.
Noda, A. T., K. Oouchi, M. Satoh, and H. Tomita, 2012: Quantitative assessment of diurnal variation of tropical convection simulated by a global nonhydrostatic model without cumulus parameterization. J. Climate, 25, 5119–5134, https://doi.org/10.1175/JCLI-D-11-00295.1.
O, S., and P. E. Kirstetter, 2018: Evaluation of diurnal variation of GPM IMERG-derived summer precipitation over the contiguous US using MRMS data. Quart. J. Roy. Meteor. Soc., 144, 270–281, https://doi.org/10.1002/qj.3218.
Oh, S. G., and M. S. Suh, 2018: Changes in seasonal and diurnal precipitation types during summer over South Korea in the late twenty-first century (2081–2100) projected by the RegCM4.0 based on four RCP scenarios. Climate Dyn., 51, 3041–3060, https://doi.org/10.1007/s00382-017-4063-5.
Riley Dellaripa, E. M., E. D. Maloney, B. A. Toms, S. M. Saleeby, and S. C. van den Heever, 2020: Topographic effects on the Luzon diurnal cycle during the BSISO. J. Atmos. Sci., 77, 3–30, https://doi.org/10.1175/JAS-D-19-0046.1.
Romatschke, U., S. Medina, and R. A. Houze Jr., 2010: Regional, seasonal, and diurnal variations of extreme convection in the South Asian region. J. Climate, 23, 419–439, https://doi.org/10.1175/2009JCLI3140.1.
Sato, T., H. Miura, M. Satoh, Y. N. Takayabu, and Y. Wang, 2009: Diurnal cycle of precipitation in the tropics simulated in a global cloud-resolving model. J. Climate, 22, 4809–4826, https://doi.org/10.1175/2009JCLI2890.1.
Satoh, M., and Y. Kitao, 2013: Numerical examination of the diurnal variation of summer precipitation over southern China. SOLA, 9, 129–133, https://doi.org/10.2151/sola.2013-029.
Semmler, T., and Coauthors, 2018: AWI AWI-CM1.1MR model output prepared for CMIP6 CMIP, version 20191219. Earth System Grid Federation, accessed 19 December 2019, https://doi.org/10.22033/ESGF/CMIP6.359.
Sorooshian, S., X. Gao, K. Hsu, R. A. Maddox, Y. Hong, H. V. Gupta, and B. Imam, 2002: Diurnal variability of tropical rainfall retrieved from combined GOES and TRMM satellite information. J. Climate, 15, 983–1001, https://doi.org/10.1175/1520-0442(2002)015<0983:DVOTRR>2.0.CO;2.
Stone, D. A., and A. J. Weaver, 2003: Factors contributing to diurnal temperature range trends in twentieth and twenty-first century simulations of the CCCma coupled model. Climate Dyn., 20, 435–445, https://doi.org/10.1007/s00382-002-0288-y.
Tatebe, H., and Coauthors, 2019: Description and basic evaluation of simulated mean state, internal variability, and climate sensitivity in MIROC6. Geosci. Model Dev., 12, 2727–2765, https://doi.org/10.5194/gmd-12-2727-2019.
Von Storch, H., and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.
Wang, S. Y., and T. C. Chen, 2008: Measuring East Asian summer monsoon rainfall contributions by different weather systems over Taiwan. J. Appl. Meteor. Climatol., 47, 2068–2080, https://doi.org/10.1175/2007JAMC1821.1.
Wu, T., and Coauthors, 2019: The Beijing Climate Center Climate System Model (BCC-CSM): The main progress from CMIP5 to CMIP6. Geosci. Model Dev., 12, 1573–1600, https://doi.org/10.5194/gmd-12-1573-2019.
Wyser, K., T. van Noije, S. Yang, J. von Hardenberg, D. O’Donnell, and R. Döscher, 2020: On the increased climate sensitivity in the EC-Earth model from CMIP5 to CMIP6. Geosci. Model Dev., 13, 3465–3474, https://doi.org/10.5194/gmd-13-3465-2020.
Xin, X., T. Wu, J. Zhang, J. Yao, and Y. Fang, 2020: Comparison of CMIP6 and CMIP5 simulations of precipitation in China and the East Asian summer monsoon. Int. J. Climatol., 40, 6423–6440, https://doi.org/10.1002/joc.6590.
Yamamoto, M. K., F. A. Furuzawa, A. Higuchi, and K. Nakamura, 2008: Comparison of diurnal variations in precipitation systems observed by TRMM PR, TMI, and VIRS. J. Climate, 21, 4011–4028, https://doi.org/10.1175/2007JCLI2079.1.
Yang, G., and J. Slingo, 2001: The diurnal cycle in the tropics. Mon. Wea. Rev., 129, 784–801, https://doi.org/10.1175/1520-0493(2001)129<0784:TDCITT>2.0.CO;2.
Yang, S., and E. A. Smith, 2006: Mechanisms for diurnal variability of global tropical rainfall observed from TRMM. J. Climate, 19, 5190–5226, https://doi.org/10.1175/JCLI3883.1.
Yuan, W., R. Yu, M. Zhang, W. Lin, H. Chen, and J. Li, 2012: Regimes of diurnal variation of summer rainfall over subtropical East Asia. J. Climate, 25, 3307–3320, https://doi.org/10.1175/JCLI-D-11-00288.1.
Yuan, W., R. Yu, M. Zhang, W. Lin, J. Li, and Y. Fu, 2013: Diurnal cycle of summer precipitation over subtropical East Asia in CAM5. J. Climate, 26, 3159–3172, https://doi.org/10.1175/JCLI-D-12-00119.1.
Yukimoto, S., and Coauthors, 2019: The Meteorological Research Institute Earth System Model version 2.0, MRI-ESM2.0: Description and basic evaluation of the physical component. J. Meteor. Soc. Japan, 97, 931–965, https://doi.org/10.2151/jmsj.2019-051.
Zhang, X. X., X. Q. Bi, and X. H. Kong, 2015: Observed diurnal cycle of summer precipitation over South Asia and East Asia based on CMORPH and TRMM satellite data. Atmos. Oceanic Sci. Lett., 8, 201–207, https://doi.org/10.1080/16742834.2015.11447260.
Zhong, S. X., 2020: Diurnal variation of the duration and environment for heavy rainfall during the warm season in South China. Atmos. Sci. Lett., 21, e969, https://doi.org/10.1002/asl.969.
