香京julia种子在线播放

In this study, the three urban agglomerations located in coastal Eastern China, i.e., Beijing-Tianjin-Hebei region (BTH), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD), are selected as research areas. These three regions are the most mature and competitive urban agglomerations in China, with the largest population density and the highest level of national economic development, which is the inevitable outcome of China’s new industrialization processes (Fang 2015). More specifically, the three urban agglomerations have an area of 370 × 10³ km², with a population of 102.19 million and a GDP of CNY 30.82 trillion (Chou et al., 2021), and these suggest that the degree of exposure and vulnerability to climate disasters and potential economic losses in three urban agglomerations are much larger than that in other regions.

The regional model used for dynamic downscaling in this study is the WRF model version 3.9.1 (Skamarock and Klemp 2008), which is coupled with the Urban Canopy Model (UCM) (WRF/UCM; Kusaka et al., 2001; Kusaka and Kimura 2004). Two one-way nesting strategy has been configured for WRF, with 15 km spatial resolution for outer domain and 5 km resolution for inner domain. As shown in Figure 1, three urban agglomerations are all located in the innermost domain, with 78 × 90 grid cells for BTH region, 90 × 110 for YRD and 45 × 57 grid points for PRD respectively. The Lambert conformal map projection was adopted for the model’s horizontal coordinates. The vertical grid contained 37 terrain-following eta levels from the surface up to 50 hPa. The main physical schemes adopted in this study include the WRF Single-Moment 6-class microphysics scheme (Hong and Lim, 2006), the Kain-Fritsch cumulus scheme (Kain, 2004), the Rapid Radiative Transfer Model for GCM (RRTMG) longwave and shortwave radiation scheme (Iacono et al., 2008), the Yonsei University boundary layer scheme (Hong et al., 2006), and the Noah land surface model (Chen and Dudhia, 2001) which is coupled with the single-layer UCM.

Atmospheric initial and lateral boundary conditions for driving the WRF model were taken from the bias-corrected CESM CMIP5 datasets (with a horizontal resolution of 0.9° × 1.25° in latitudinal and longitudinal directions) at 6-h intervals (https://rda.ucar.edu/datasets/ds316.1/). Compared with other CMIP5 models, CESM can reasonably reproduce the global temperature and precipitation distribution (Knutti and Sedlacek 2013), especially in China (Chen and Frauenfeld, 2014). For the future projection, a low-medium scenario (i.e., RCP45) was selected, which is corresponding to moderate emissions with a range of technologies and strategies for reducing GHG emissions (Thomson et al., 2011).

Moreover, the daily mean surface air temperature (SAT) retrieved from the Earth System Grid data portal was used to estimate the timing of 1.5 and 2°C global warming in the CESM projection under RCP4.5 scenario. Relative to the preindustrial period, the timing for 1.5 and 2°C warming is expected in year 2030 and 2047 respectively, based on the CESM CMIP5 projection under RCP4.5 scenario. To investigate future changes in summer HWs over urban agglomerations of eastern China under 1.5 and 2.0°C global warming, we take 1996–2005 as reference time period, 2025–2034 as 1.5°C warming time period, and 2042–2051 as 2°C warming time period in this study. Then three sets of 10-year dynamical downscaling experiments have been conducted, with WRF driven by the CESM simulation and projection results, i.e., 1) a historical downscaling run during 1996–2005 (Hereafter “Exp HIS-run”), with WRF driven by CESM historical simulation during the same period, 2) a scenario run under 1.5°C warming (Hereafter “Exp 1.5-run”), with WRF driven by CESM future projection results during 2025–2034, and 3) a scenario run under 2°C warming (Hereafter “Exp 2.0-run”), with WRF driven by CESM future projection results during 2042–2051. For each experiment, the model was initiated from 12:00 UTC 31 May and run until 00:00 UTC 1 September each year, with the first 12 h of each run being considered as the spin-up time.

The daily mean temperature (T_mean) and maximum temperature (T_max) during 1996–2005, which is from the meteorological stations of the China Meteorological Administration (https://data.cma.cn), have been adopted for the evaluation and bias correction of WRF-based dynamical downscaling results. There are 21, 27, and 7 meteorological stations selected in BTH, YRD, and PRD, respectively. These stations include both urban stations and suburban stations in model domain d02 (Figures 1D–F).

The default land cover data (USGS 24-category) used in the WRF model are based on 1-km Advanced Very High Resolution Radiometer data obtained from 1992 to 1993, which cannot provide up-to-date information. In the present study, a 5-km resolution ChinaLC dataset from 1981 to 2010 was used to represent land cover in the model. Detailed information regarding the data sources and the classification approach can be found in Li et al. (2017) and Yang et al. (2017). When compared with other large-scale land cover datasets, the ChinaLC dataset shows an average overall accuracy of approximately 75%, which is much higher than its counterparts (Xiao et al., 2015; Li et al., 2017). As the resolution of the inner WRF domain was set to 5 km, which is the same as the spatial resolution of ChinaLC data, the land use data in the model can be replaced by the ChinaLC data in the WRF Preprocessing System.

The classification scheme of the ChinaLC dataset is developed based on the International Geosphere-Biosphere Programme (IGBP) land cover classification scheme, which is different from that of the USGS dataset in the WRF model. Before utilizing the ChinaLC dataset, the classification transformation was carried out based on the method used by Liang et al. (2019). A land cover case of the 2010s from the ChinaLC dataset was introduced into the WRF/UCM to represent the land cover over the urban agglomerations. The land cover classifications of the 2010s for the innermost model domains are shown in Figures 1D–F. We can see that the main types of land cover in the urban agglomerations present wide divergence in different latitudes, especially for the urban/built-up land (shaded areas in red). However, the urban/built-up land is not the dominant land cover type in urban agglomerations. In the historical scenario, the percentages of urbanized areas in BTH, YRD, and PRD are 4.89, 8.49, and 17.86%, respectively. Despite the varying percentages caused by the different area sizes of selected research regions, the climate characteristics of urban parts are usually very different from those of the suburbs in the regions. Therefore, the response of extreme cases in urban areas to global warming may have been underestimated in previous studies by using the regional average changes.

Generally, there are several HW definitions which are based on relative or absolute thresholds, but it is also noted that there is no single HW definition which could be universally accepted (You et al., 2017). Following the heatwave warning criteria used by the China Meteorological Administration (www.cma.gov.cn/en/WeatherWarnings/), we define extreme high temperature with a threshold of 35°C in this study, and the day with T_max ≥ 35°C is defined as a high-temperature day. An HW event is then identified when there are three and more consecutive high-temperature days. Furthermore, three widely used HW indices are adopted for analysis in this study, i.e., the frequency index, which is the number of HWs in the study period; the duration index, which is the sum of the number of high temperature days during the HWs; the intensity index, which refers to the cumulative sum of the difference between the daily maximum temperature and the 35°C threshold for all HWs in study period. In this study, we focus on changes of HWs in summer season, so all above three HW indices are calculated for summer season, i.e., June, July and August.

To reduce the model uncertainty of WRF/UCM in simulating the daily mean and maximum temperature, bias correction (BC) for the WRF/UCM-based downscaled T_max is needed. Based on the average difference between the observed and simulated T_mean and T_max during the period 1996–2005, a climatology based bias correction method is applied in this study, just as mentioned in Li et al. (2018).

The bias-corrected temperature for the model ( T MC ) was calculated as follows: T MC = T model + ( T T OBS -T T model ) (1) Where T m o d e l represents the simulated daily mean/maximum temperature at each grid cell of the model, and ( T T OBS -T T model ) stands for the mean systematic error of the temperature. T T OBS and T T model are the mean values from the observation of all stations and the simulation of model corresponding to the stations during 1996–2005, respectively.

Using the bias corrected temperature results, the future changes of summer HWs in three urban agglomerations have been estimated, and the flow chart of analysis in this study is shown in Figure 2 as following.

To examine the performance of the WRF/UCM modeling system, the frequency histograms of simulated results versus observed data for Tmean and Tmax are shown in Figure 3. All daily values during the entire analysis period (JJA of 1996–2005) from the meteorological stations were used. For BTH, YRD and PRD, the records are 18,900 (92 days × 10 years × 21 stations = 18,900), 24,300 (92 days × 10 years × 27stations = 18,900) and 6,300 (92 days × 10 years × 7 stations = 6,300), respectively. From the frequency of occurrence in different temperature sections, there are good temporal correlations between the bias-corrected model outputs and observations for T_mean and T_max. The correlation coefficients are mostly greater than 0.6 for all three urban agglomerations. The mean bias ranges from 1.04 to 1.38 for T_mean and from 1.22 to 1.7 for T_max. Furthermore, the spatial distributions of summer T_mean and T_max in observations and bias-corrected model simulations are demonstrated in Figure 4. The HIS-run well replicates the observations in JJA with a spatial correlation coefficient of more than 0.6, which indicates that the bias-corrected model outputs capture the climatological features acceptably. The urban heat island effect can be seen, as T_mean and T_max tend to be higher in the urban areas and lower in the surrounding areas.

In summary, the bias-corrected model outputs can, to a great extent, generate the regional-scale characteristics of the summer temperatures realistically at each selected urban agglomeration faithfully, which suggests the reliability of using WRF downscaling for projecting the HW indices’ future changes. We note that a few efforts recently attempted to apply the WRF dynamic downscaling method to revisit the regional impacts of global warming (Yamamoto et al., 2018; Ge et al., 2021).

Figure 5 presents the WRF-projected changes in summer T_mean and T_max over the three urban agglomerations under 1.5 and 2.0°C warming conditions relative to the historical run. Projected T_mean and T_max both increase under 1.5 and 2.0°C warming throughout the urban agglomerations. When the warming reaches 1.5°C under RCP4.5, the averaged increasing magnitude of T_max are 0.97°C with the inhomogeneous changes range between 0.92 and 1.01°C over BTH region. As for YRD region, the 10-year mean increasing magnitude is 0.88°C, with the range between 0.82 and 0.94°C. The increasing magnitude is smallest over PRD region, with mean magnitude of 0.85°C, maximum T_max increase is 0.90°C, and minimum T_max increase is 0.82°C. The increasing magnitudes of T_max are stronger than those of T_mean. It suggests that the number of extreme high temperature events will soar in urban agglomerations. Under the scenario of 2.0°C warming, the changes of summer temperatures will be greater (Figure 5B), and the largest increases of T_mean and T_max occur in YRD.

How the increase from 1.5 to 2.0°C warming impacts on summer temperatures is shown in Figure 5C. With this additional 0.5°C temperature rise, T_mean will increase by 0.29°C (0.19–0.38°C), 0.68°C (0.6–0.75°C) and 0.34°C (0.33–0.36°C) over BTH, YRD and PRD, respectively, whereas T_max can rise by 0.2°C (0.1–0.26°C), 0.59°C (0.5–0.69°C) and 0.36°C (0.33–0.4°C) for BTH, YRD and PRD, respectively. Obviously, the increases of T_mean and T_max in response to the additional 0.5°C warming are largest in YRD, which implies that by enforcing the agreement on limiting the temperature increase to 1.5°C, the YRD region will benefit more than the other urban agglomerations in terms of reducing extreme hot events. This result is consistent with the results of multi-mode projection on extreme maximum temperature by Yu et al. (2018). Note that the increases of T_mean and T_max in response to the additional 0.5°C warming in the other urban agglomerations except YRD region is slightly weaker than the additional rise. This may be because our study focuses on summertime when the warming trend will be less significant than winter in China under future global warming (Yang et al., 2021c). In addition, Ge et al. (2021) pointed out that the smaller benefits achieved from the additional 0.5°C warming limit for the WRF projection than CESM.

Figures 6–8 show the spatial distributions of projected future changes in the HWs’ frequency, duration, and intensity over the three urban agglomerations. The three indices of summer HWs are enhanced in most areas of the urban agglomerations against the backgrounds of 1.5 and 2.0°C warming. According to the variation diagrams of HW frequency in Figures 6A,D,G, the HW frequency is enhanced in almost entire BTH and YRD under the 1.5°C warming scenario. In BTH, the enhancement mainly occurs over the urban underlying surface and the underlying surface of dry cropland to the south of the urban areas, while the urban and cropland/woodland mosaic areas witness most of the HW frequency increases in YRD. As for PRD, the HW frequency decreases in the urban areas of Guangzhou, but increases significantly in the surrounding cities. Under the 2.0°C warming scenario, there is a similar distribution of HW frequency changes, but the range is more extensive, and the intensity is stronger (Figures 6B,E,H). With the additional temperature increase of 0.5°C from 1.5 to 2.0°C, the frequency of HWs in metropolitan areas (such as Beijing, Shanghai, Guangzhou) will decrease (Figures 6C,F,I). It should be pointed out that the frequency of HWs in our study is defined based on the HW events, so the reduction of frequency may not mean the decrease of high-temperature days, because the increase of HWs’ duration may reduce HWs’ frequency in the fixed summertime.

The variations of HW duration under different scenarios are illustrated in Figure 7. Compared with the historical period, when reaching the warming scenarios of 1.5 and 2.0°C, the urban agglomerations show a ubiquitous characteristic of prolonged HW duration at different latitudes from north to south, and the more southwards the urban agglomerations are located, the larger increases their HW duration will have. In those urban agglomerations, the regions with relatively larger HW duration growth can generally match the regions with higher HW frequency increase shown in Figure 6. As can be seen from Figures 7C,F,I, with the additional 0.5°C warming, the duration of HWs increases both in the urban areas and in the suburbs. Why does the frequency of HWs in urban areas decrease, while in the suburbs on the contrary (Figures 6C,F,I), and the duration of HWs in the suburbs increases more than that in urban areas? It may be because the current urban areas have higher frequency and longer duration of HWs than the suburbs due to the heat island effect. Therefore, under global warming, HWs’ frequency and duration increase within a limited extent in cities, while the suburbs have a greater growth space.

Different from the changes in the frequency and duration of HWs, whether from the historical period to the 1.5 and 2.0°C warming periods or from 1.5 to 2.0°C warming, large increases of HW intensity are mainly concentrated over the urban underlying surface (Figure 8). This implies that the urban population will face stronger HWs in the future.

Furthermore, the future changes in the area-average frequency, duration, and intensity of HWs in each urban agglomeration are summarized in Table 1. It can be seen from the table that, under any scenario, the more southward the urban agglomerations are, the higher frequency, longer duration, and greater intensity their HWs will have, since the HW indices are related to the geographical latitudes of the urban agglomerations. Under the 1.5°C warming scenario, the frequency, duration and intensity of HWs are increased respectively by 50.4, 63.6, and 116.7% in BTH, 84.7, 84.6, and 123.2% in YRD, and 76.3, 109.8, and 130.4% in PRD compared with the historical period. Among the three HWs indices, the intensity grows the most. The additional warming of 0.5°C from 1.5 to 2.0°C will result in 15.5, 44.7, and 75.5% increases in the frequency, duration, and intensity of HWs in YRD, which are the largest among those of the three urban agglomerations.

From the spatial distribution figures concerning the future changes in HWs’ frequency, duration, and intensity, we can see large spatial differences in each urban agglomeration, which mainly lie between the urban and non-urban underlying surfaces. Four major land cover types were selected to analyze their HW index values in each urban agglomeration in detail. Figure 9 shows the area-average HW frequency, duration, and intensity values of different land cover types in each urban agglomeration. Compared with other major land cover types, the urban areas have much higher frequency, duration, and intensity of HWs. The response of HW indices to the warming is more robust in the urban areas, especially under 1.5°C warming climate. In the scenario of 1.5°C warming, the HWs’ frequency, duration and intensity in urban areas of BTH are 3.73, 19.36, and 51.98 respectively. Compared to the respective values of HW indices over the dryland cropland and pasture, which are the biggest among those of non-urban land cover types in BTH, the above values in urban areas are 60, 68, and 144% higher. In YRD, the HW frequency, duration, and intensity in urban areas are 3.44, 22.96, and 41.49 respectively, and those values are 65, 83 and 140% higher than the respective greatest values for non-urban land cover, which all belong to the irrigated cropland and pasture in this case. The HW frequency, duration, and intensity in the urban areas of PRD are 4.71, 38.74, and 59.54 respectively, which are 74, 126, and 200% higher than those of the cropland and woodland mosaic, the largest among the respective values for non-urban land cover. In PRD, the differences in the values of HW indices between urban and non-urban areas are particularly significant. The foregoing results thus suggest a much graver future risk of HWs in urban areas than in non-urban areas.