Integrating Remote-Sensing and Assimilation Data to Improve Air Temperature on Hot Weather in East China

: Land-surface characteristics (LSCs) and land-soil moisture conditions can modulate energy partition at the land surface, impact near-surface atmosphere conditions, and further affect land– atmosphere interactions. This study investigates the effect of land-surface-characteristic parameters (LSCPs) including albedo, leaf-area index (LAI), and soil moisture (SM) on hot weather by in East China using the numerical model. Simulations using the Weather Research and Forecasting (WRF) Model were conducted for a hot weather event with a high spatial resolution of 1 km in domain 3 by using ERA-Interim forcing ﬁelds on 20 July 2017 until 16:00 UTC on 25 July 2017. The satellite-based albedo and LAI, and assimilation-based soil-moisture data of high temporal–spatial resolution, which are more accurate to match ﬁne weather forecasts and high-resolution simulations, were used to update the default LSCPs. A control simulation with the default LSCPs (WRF_CTL), a main sensitivity simulation with the updated LSCP albedo, LAI and SM (WRF_CHAR), and a series of other sensitivity simulations with one or two updated LSCPs were performed. Results show that WRF_CTL could reproduce the spatial distribution of hot weather, but overestimated air temperature (Ta) and maximal air temperature (Tamax) with a warming bias of 1.05 and 1.32 ◦ C, respectively. However, the WRF_CHAR simulation reduced the warming bias, and improved the simulated Ta and Tamax with reducing relative biases of 33.08% and 29.24%, respectively. Compared to the WRF_CTL, WRF_CHAR presented a negative sensible heat-ﬂux difference, positive latent heat ﬂux, and net radiation difference of the area average. LSCPs modulated the partition of available land-surface energy and then changed the air temperature. On the basis of statistical-correlation analysis, the soil moisture of the top 10 cm is the main factor to improve warming bias on hot weather in East China.


Introduction
With global warming, the intensity and frequency of hot weather are generally rising [1]. China often suffers from hot weather, East China (EC), where the summer mean air temperature (Ta) is over 27.4 • C [2,3]. Since the 1990s, the intensity of hot weather has significantly increased in EC [2,3], affecting human health and life, energy supply and demand, water resources, and agricultural production, and causing economic losses [4]. Many studies showed that the increase in high temperature is closely related to land-surface heterogeneity [5]. The EC region has experienced rapid urbanization with economic and social development, which resulted in diverse and complex land-surface characteristics (LSCs) [6,7]. Studies found that some hot weather is sensitive to near-surface conditions and LSCs [4,8,9]. Therefore, the impact of LSC is becoming increasingly important for understanding hot weather in EC [10].
In recent years, due to global warming, high-temperature and heatwave weather occurs frequently during summer in China, and a wide range of record-breaking high temperatures appeared one after another [40], and 2017 was the warmest year on record without an El Niño affecting it; the hot weather showed obvious features of high intensity, long duration, and large area [40]. In late July 2017, the averaged maximal air temperature was more than 4 °C than that of the historical period, the number of high-temperature days was 3 to 6 days more than that of the historical period, and some stations broke the station's historical records [40]. In 21-25 July 2017, 500 hPa, the Asian middle and high latitudes had a trough and a ridge. The ridge was located in northern Siberia, and the trough was located in the northern part of Japan. The center of the subtropical high was located northward, and the southernmost part of China was controlled by the subtropical ridge ( Figure 2). Strong downdrafts and anticyclonic circulations stabilized the atmosphere and eventually led to hot weather [40].  In recent years, due to global warming, high-temperature and heatwave weather occurs frequently during summer in China, and a wide range of record-breaking high temperatures appeared one after another [40], and 2017 was the warmest year on record without an El Niño affecting it; the hot weather showed obvious features of high intensity, long duration, and large area [40]. In late July 2017, the averaged maximal air temperature was more than 4 • C than that of the historical period, the number of high-temperature days was 3 to 6 days more than that of the historical period, and some stations broke the station's historical records [40]. In 21-25 July 2017, 500 hPa, the Asian middle and high latitudes had a trough and a ridge. The ridge was located in northern Siberia, and the trough was located in the northern part of Japan. The center of the subtropical high was located northward, and the southernmost part of China was controlled by the subtropical ridge ( Figure 2). Strong downdrafts and anticyclonic circulations stabilized the atmosphere and eventually led to hot weather [40]. city, and most of Anhui province, which covered the most area of EC. Topographical height is relatively flat, and most of the area of topography height was lower than 100 m in the studied area ( Figure 2). Crop is the main land-use type except sea, followed forest, shrub, water (lakes and rivers), urban, and grass covering small areas ( Figure 2). All simulations used the ERA-Interim (ERA) 6 h boundary conditions. Simulations were initialized at 00:00 UTC (0800 LST) on 20 July 2017 until 16:00 UTC 25 July 2017 (0000 LST), and the last 5 days of LST were analyzed. We first performed two simulations. First, the control simulation (WRF_CTL) in which the default LSCP albedo and LAI were used and the SM data initialized by forcing fields ERA (data details described in Section 2.2). Second, the sensitivity simulation in which the LSCP albedo, LAI, and four layers of SM for initialization were updated on the basis of MODIS albedo and LAI and CLDAS SM (WRF_CHAR; data details described in Section 2.2). In addition, a series of supplied sensitivity simulations were performed, which included the ALB simulation (updated the albedo on the basis of the MODIS albedo), LAI simulation (updated the LAI on the basis of MODIS LAI), ALL simulation (updated the albedo and LAI on the basis of MODIS albedo and LAI), and CLDAS simulation (updated the SM on the basis of CLDAS SM). For LSC data processing, if the MODIS albedo was missing, the nearest years' valid value was used. Then, the albedo, LAI, and CLDAS SM datasets were reprojected onto the WRF Lambert projection to coincide with

Data
In the WRF model, default LAI and albedo data are derived from MODIS and AVHRR, respectively, which are better, and can show spatial distribution and temporal change characteristics of 12 months comparing traditional parameter tables depending on vegetation type [13,20,41,42]. However, the default LAI and albedo with coarser spatiotemporal features and without interannual variation do not display realistic conditions. In this study, the albedo for updating the lower boundary condition of WRF was derived from MODIS Collection 6 land products based on Terra and Aqua, with a 1 day temporal resolution and about 1 km horizontal resolution [43,44]. The LAI for updating the lower boundary condition of WRF was also derived from MODIS Collection 6 land products based on Terra and Aqua, with a 4 day temporal resolution and about 500 m horizontal resolution [43,44].
In addition, the SM is an important initial condition for the RCM. Initial SM data are obtained from forcing data, and those are often reanalysis data of global circulation models [24,31]. For example, in the WRF model, initial SM fields can be supplied from GFS, FNL, NCEP, ECMWF ERA-Interim, or JRA55 and other similar general atmospheric circulation models or general atmosphere-ocean circulation models, which, however, have coarse resolution [24,31]. Inadequate resolution is the main problem with this type of soil-moisture initialization [31]. For example, ERA-Interim (ERA) reanalysis data have a resolution of 0.75 • × 0.75 • , which is higher than that of similar reanalysis data, that is to say, one grid covers about 90 × 90 km in the middle latitude. In our study, the smallest domain, Dh-3, had a 1 km resolution, which means that one value in the ERA data represented a value of 90 × 90 grids, thus lacking realistic representation and spatial variations [31]. The China Meteorological Administration (CMA) Land Data Assimilation System (CLDAS) SM fields were used to update the initial SM in this work, with a high spatial resolution of 0.0625 • × 0.0625 • , and a high temporal resolution of 1 day. CLDAS SM and the station observation SM matched very well, had a region average correlation coefficient of 0.89, root-mean-square error of 0.02 m 3 m −3 , and a bias of 0.01 m 3 m −3 , and CLDAS SM was better than similar assimilation SM data of other agencies (http: //data.cma.cn/dataService/cdcindex/datacode/NAFP_CLDAS2.0_RT.html, accessed on Remote Sens. 2021, 13, 3409 5 of 16 11 February 2020). Therefore, satellite-based MODIS albedo data and LAI data, and assimilation-based CLDAS SM data were relatively reliable to update the LSCPs in the WRF model in EC (Table 1). Automatic hourly surface weather observation stations of 2 m air temperature datasets, which are produced and can be downloaded on the Climatic Data Center, the National Meteorological Information Center, CMA (http://data.cma.cn/site/index.html, accessed on 11 February 2020), were used to assess the simulated air temperature (Ta) and maximal air temperature (Tamax).

Model Setup
The WRF model, which was developed by NCAR/NCEP, the Forecast Systems Laboratory, and other university scientists, is widely used in weather forecast or prediction and long-term research [20,45,46]. The WRF model is also widely used to research extreme weather, especially during heatwaves or hot weather [9,[47][48][49][50].
In this study, we used WRF model version 3.8.1 to research hot weather by using physical schemes and parameterizations as follows: WSM6 microphysics scheme, RRTM longwave radiation scheme, Dudhia shortwave radiation scheme, MM5 Monin-Obukhov surface-layer scheme, Bougeault and Lacarrere TKE boundary-layer scheme, and the unified Noah land-surface model. Three nested domains with 25, 5, and 1 km horizontal resolution were used in the simulation, respectively ( Figure 2). The innermost domain-3 covered most of the area of the Yangtze River Delta, including Jiangsu province, Shanghai city, and most of Anhui province, which covered the most area of EC. Topographical height is relatively flat, and most of the area of topography height was lower than 100 m in the studied area ( Figure 2). Crop is the main land-use type except sea, followed forest, shrub, water (lakes and rivers), urban, and grass covering small areas ( Figure 2). All simulations used the ERA-Interim (ERA) 6 h boundary conditions. Simulations were initialized at 00:00 UTC (0800 LST) on 20 July 2017 until 16:00 UTC 25 July 2017 (0000 LST), and the last 5 days of LST were analyzed.
We first performed two simulations. First, the control simulation (WRF_CTL) in which the default LSCP albedo and LAI were used and the SM data initialized by forcing fields ERA (data details described in Section 2.2). Second, the sensitivity simulation in which the LSCP albedo, LAI, and four layers of SM for initialization were updated on the basis of MODIS albedo and LAI and CLDAS SM (WRF_CHAR; data details described in Section 2.2). In addition, a series of supplied sensitivity simulations were performed, which included the ALB simulation (updated the albedo on the basis of the MODIS albedo), LAI simulation (updated the LAI on the basis of MODIS LAI), ALL simulation (updated the albedo and LAI on the basis of MODIS albedo and LAI), and CLDAS simulation (updated the SM on the basis of CLDAS SM). For LSC data processing, if the MODIS albedo was missing, the nearest years' valid value was used. Then, the albedo, LAI, and CLDAS SM datasets were reprojected onto the WRF Lambert projection to coincide with the WRF model. All data were interpolated to 6 h intervals LSCPs to match with our WRF model's setups inputs by using a simple linear method.

LSCP Comparison
For the LSC, the WRF default albedo showed very coarse spatial distribution characteristics. It mainly displayed values of 0.14-0.16 in most research areas, and there were some variations south and east of Jiangsu province, and north of Anhui province (Figure 3(a1)). The MODIS albedo had more accurate spatial distribution characteristics and wide-ranging variations of 0.06-0.24, and the albedo was higher in the northwestern study area than that in the southeastern area ( Figure 3(a2,a3)). The area-averaged MODIS albedo was higher by 0.005 (3.40%) than the default albedo. The LAI showed a similar spatial distribution difference with that of the land-surface albedo. Besides accuracy spatial distribution characteristics (Figure 3(b1,b2)), the MODIS LAI was higher by 1.5-3.0 m 2 m −2 in the southwestern study area and lower by −0.5 to −1.5 m 2 m −2 in the northeastern study area (Figure 3(b3)). The area average MODIS LAI was 2.79 m 2 m −2 , which was higher by 11.15% than the default LAI of 2.51 m 2 m −2 .
The For the LSC, the WRF default albedo showed very coarse spatial distribution char teristics. It mainly displayed values of 0.14-0.16 in most research areas, and there we some variations south and east of Jiangsu province, and north of Anhui province (Figu 3(a1)). The MODIS albedo had more accurate spatial distribution characteristics and wid ranging variations of 0.06-0.24, and the albedo was higher in the northwestern study ar than that in the southeastern area ( Figure 3(a2,a3)). The area-averaged MODIS albedo w higher by 0.005 (3.40%) than the default albedo. The LAI showed a similar spatial dist bution difference with that of the land-surface albedo. Besides accuracy spatial distrib tion characteristics (Figure 3(b1,b2)), the MODIS LAI was higher by 1.5-3.0 m 2 m −2 in southwestern study area and lower by −0.5 to −1.5 m 2 m −2 in the northeastern study ar (Figure 3(b3)). The area average MODIS LAI was 2.79 m 2 m −2 , which was higher by 11.1 than the default LAI of 2.51 m 2 m −2 .
The  (Figure 3(c1,c2,d1,d2)). T figures of SM differences showed that there were negative differences of −0.01 to −0 m 3 m −3 in the western middle study area, and positive differences of 0.01-0.15 m 3 m −3 four soil layers (Figure 3(c3,d3)); the area-averaged differences were 0.032, 0.043, 0.0 and 0.053 m 3 m −3 , respectively, and the relative differences were 12.36%, 15.87%, 15.77 and 18.09%, respectively in the four soil layers, which gradually increased with the chan in soil depth (SM figures of 60 and 100 cm are not shown).
Generally, it is remarkable that the MODIS albedo and LAI could accurately descr the spatial distribution characteristics comparing the default climatology albedo and LA which are more suitable for high-resolution simulations. Area-averaged CLDAS SM d were higher by 12.36% to 18.09% than those of the ERA SM. te Sens. 2021, 13, x FOR PEER REVIEW 7 of Figure 3. Spatial-distribution characteristics and relative differences over research period in EC. (a1,a2,a3) Albedo and relative difference between WRF and MODIS data; (b1,b2,b3) LAI and relative difference between WRF and MODIS data; (c1,c2,c3) 10 cm soil water and relative difference between WRF and CLDAS data (unit: m 3 m −3 ) and (d1,d2,d3) 30 cm soil water and relative difference between WRF and CLDAS data (unit: m 3 m −3 ).

Model Validation
To validate the performance of the developed modeling system, the simulation r sults of the WRF_CTL were compared with the observed Ta from the AWS network. F ures 4a and 5a show the spatial distribution of the observed Ta during the hot weather EC, particularly mean Ta and Tamax. The large-area higher-observation Ta appear south of Jiangsu province and the Shanghai, and the highest 5 day mean observation reached 36.60 °C (Figure 4a). The WRF_CTL simulation could effectively capture the sp tial distribution of Ta with a spatial correlation coefficient (SCC) of 0.77 compared w the observed Ta ( Figure 4b). However, the WRF_CTL simulation systematically overes mated the Ta with a warming bias of 1.05 °C in most of the study area (Figure 4b,c).  . Spatial-distribution characteristics and relative differences over research period in EC. (a1,a2,a3) Albedo and relative difference between WRF and MODIS data; (b1,b2,b3) LAI and relative difference between WRF and MODIS data; (c1,c2,c3) 10 cm soil water and relative difference between WRF and CLDAS data (unit: m 3 m −3 ) and (d1,d2,d3) 30 cm soil water and relative difference between WRF and CLDAS data (unit: m 3 m −3 ).
Generally, it is remarkable that the MODIS albedo and LAI could accurately describe the spatial distribution characteristics comparing the default climatology albedo and LAI, which are more suitable for high-resolution simulations. Area-averaged CLDAS SM data were higher by 12.36% to 18.09% than those of the ERA SM.

Model Validation
To validate the performance of the developed modeling system, the simulation results of the WRF_CTL were compared with the observed Ta from the AWS network. Figures 4a and 5a show the spatial distribution of the observed Ta during the hot weather in EC, particularly mean Ta and Tamax. The large-area higher-observation Ta appeared south of Jiangsu province and the Shanghai, and the highest 5 day mean observation Ta reached 36.60 • C (Figure 4a). The WRF_CTL simulation could effectively capture the spatial distribution of Ta with a spatial correlation coefficient (SCC) of 0.77 compared with the observed Ta (Figure 4b). However, the WRF_CTL simulation systematically overestimated the Ta with a warming bias of 1.05 • C in most of the study area (Figure 4b,c).
The spatial distribution of observation Tamax had a similar pattern with that of observation Ta. The daily highest hourly observation Tamax of 41.10 • C appeared south of Jiangsu, and the 5 day mean observation Tamax is about 39.86 • C (Figure 5a). The WRF_CTL simulation could capture the spatial distribution of Tamax, and the SCC between observed Tamax and WRF_CTL Tamax is 0.67 (Figure 5b). The CTL simulation also overestimated the Tamax with a warming bias of 1.32 • C (Figure 5b,c). Comparing the simulated Ta by WRF_CTL, the SCC between simulated and observed Tamax was lower, but the warming bias was higher. Table 2 compares the statistical values of the observed and simulated 5-day mean Ta and Tamax in the study region. south of Jiangsu province and the Shanghai, and the highest 5 day mean observation reached 36.60 °C (Figure 4a). The WRF_CTL simulation could effectively capture the sp tial distribution of Ta with a spatial correlation coefficient (SCC) of 0.77 compared w the observed Ta (Figure 4b). However, the WRF_CTL simulation systematically overes mated the Ta with a warming bias of 1.05 °C in most of the study area (Figure 4b,c).  simulated Ta by WRF_CTL, the SCC between simulated and observed Tamax was low but the warming bias was higher. Table 2 compares the statistical values of the observ and simulated 5-day mean Ta and Tamax in the study region.

Impact of LSCPs on Air Temperature
To investigate the impact of LSC on hot weather, results from control simulati WRF_CTL and sensitivity simulation WRF_CHAR were compared and analyzed. Figu  6 shows the simulated averaged Ta and the difference in the modeling period WRF_CTL and WRF_CHAR simulations. WRF_CTL and WRF_CHAR presented simi Ta patterns, and the higher temperature regions mainly appeared to be cities, such Nanjing, Wuxi, Zhenjiang, and Shanghai. However, WRF_CHAR obtained a higher than WRF_CTL did, especially east of Jiangsu, with a minus difference of 2 °C . In most the southern region of Jiangsu, a cold bias of 0.2-1.0 °C was found, but north of Jiangsu slight warming bias of 0.2-0.4 °C appeared. Generally, a 5 day area-averaged Ta WRF_CHAR showed a slight cold bias of 0.38 °C compared with that in WRF_CTL.
The surface temperature (Ts) had a similar result with Ta from WRF_CTL a WRF_CHAR (figures not shown), but a cold bias of more than 3 °C appeared in the e side of Jiangsu, and a 5 day area-averaged Ts displayed a cold bias of 0.72 °C .

Impact of LSCPs on Air Temperature
To investigate the impact of LSC on hot weather, results from control simulation WRF_CTL and sensitivity simulation WRF_CHAR were compared and analyzed. Figure 6 shows the simulated averaged Ta and the difference in the modeling period by WRF_CTL and WRF_CHAR simulations. WRF_CTL and WRF_CHAR presented similar Ta patterns, and the higher temperature regions mainly appeared to be cities, such as Nanjing, Wuxi, Zhenjiang, and Shanghai. However, WRF_CHAR obtained a higher Ta than WRF_CTL did, especially east of Jiangsu, with a minus difference of 2 • C. In most of the southern region of Jiangsu, a cold bias of 0.2-1.0 • C was found, but north of Jiangsu, a slight warming bias of 0.2-0.4 • C appeared. Generally, a 5 day area-averaged Ta by WRF_CHAR showed a slight cold bias of 0.38 • C compared with that in WRF_CTL. the southern region of Jiangsu, a cold bias of 0.2-1.0 °C was found, but north of Jiangsu slight warming bias of 0.2-0.4 °C appeared. Generally, a 5 day area-averaged Ta WRF_CHAR showed a slight cold bias of 0.38 °C compared with that in WRF_CTL.
The surface temperature (Ts) had a similar result with Ta from WRF_CTL a WRF_CHAR (figures not shown), but a cold bias of more than 3 °C appeared in the e side of Jiangsu, and a 5 day area-averaged Ts displayed a cold bias of 0.72 °C . The surface temperature (Ts) had a similar result with Ta from WRF_CTL and WRF_CHAR (figures not shown), but a cold bias of more than 3 • C appeared in the east side of Jiangsu, and a 5 day area-averaged Ts displayed a cold bias of 0.72 • C. Figure 7a presents the hourly variations of area-averaged observed and simulated Ta and their differences in EC. Tamax was observed to occur at approximately 16:00 local standard time (LST) by observation, WRF_CTL and WRF_CHAR. WRF_CTL obtained a higher Ta than the observation did at almost the entire diurnal time period, the maximal warming bias was approximately 1.9 • C, and the averaged warming bias was 1.07 • C (Figure 7b). The Ta from WRF_CHAR was lower than that from WRF_CTL, with a maximal warming bias of 1.5 • C and an averaged warming bias of 0.71 • C, which was closer to the observed value and significantly reduced the warming bias. Statistical results showed that the WRF_CHAR slightly improved the simulated Ta with a value of 33.08%. Figure 8 presents the spatial distribution of mean Tamax for WRF_CTL, WRF_CHAR, and their difference, which presented a similar result with that in Figure 6, and WRF_CHAR generally obtained a lower Tamax than WRF_CTL did, with an area-averaged difference of approximately 0.45 • C. Daily variations of area-averaged observed and simulated Tamax obviously showed that WRF_CHAR obtained a lower Tamax than WRF_CTL did, though still higher than that of the observation. Comparing the observations, WRF_CHAR reduced warming bias from 1.31 to 0.92 • C. That is to say, CHAR improved the simulated Tamax of 29.24% (Figure 9).

Impact of LSCPs on Surface Energy Balance
LSCs modulate the partition of total available energy at the land surface between sensible heat fluxes (SHs) and latent heat fluxes (LHs), determine temperature and moisture at the interface between soil or vegetation and the atmosphere, and then impact near-surface atmospheric conditions [14,24], further resulting in changes in air temperature [24]. Figure 10 displays the 5 day averaged surface energy-flux differences between WRF_CHAR and WRF_CTL in EC. As is shown in Figure 10a, large areas had a negative SH difference, which was very similar with the Ta differences (Figure 6c), with an area-averaged SH difference of −9.25 Wm −2 , especially east of Jiangsu, with a minimal SH difference of about −30 Wm −2 . LH differences presented the opposite pattern when comparing SH differences, which mainly showed positive values in most areas with an area average of 13.35 Wm −2 , and the maximal SH difference was more than 30 Wm −2 . Generally, net radiation flux (Rn) presented a positive difference with an area average of 3.57 Wm −2 , the positive Rn was mainly distributed in the southeastern region of the research area, and the negative difference appeared northwest of the research area. Soil heat flux (GH) showed a slight positive difference with an area average of 0.35 Wm −2 . In the south of the research area, GH differences of 1-3 Wm −2 appeared, but in the north of the research area, there was a −1 to −3 Wm −2 GH difference. Figure 7a presents the hourly variations of area-averaged observed and simulated Ta and their differences in EC. Tamax was observed to occur at approximately 16:00 local standard time (LST) by observation, WRF_CTL and WRF_CHAR. WRF_CTL obtained a higher Ta than the observation did at almost the entire diurnal time period, the maximal warming bias was approximately 1.9 °C , and the averaged warming bias was 1.07 °C (Figure 7b). The Ta from WRF_CHAR was lower than that from WRF_CTL, with a maximal warming bias of 1.5 °C and an averaged warming bias of 0.71 °C , which was closer to the observed value and significantly reduced the warming bias. Statistical results showed that the WRF_CHAR slightly improved the simulated Ta with a value of 33.08%.  Figure 8 presents the spatial distribution of mean Tamax for WRF_CTL, WRF_CHAR, and their difference, which presented a similar result with that in Figure 6, and WRF_CHAR generally obtained a lower Tamax than WRF_CTL did, with an areaaveraged difference of approximately 0.45 °C . Daily variations of area-averaged observed and simulated Tamax obviously showed that WRF_CHAR obtained a lower Tamax than WRF_CTL did, though still higher than that of the observation. Comparing the observ tions, WRF_CHAR reduced warming bias from 1.31 to 0.92 °C . That is to say, CHAR i proved the simulated Tamax of 29.24% (Figure 9).

Impact of LSCPs on Surface Energy Balance
LSCs modulate the partition of total available energy at the land surface between sensible heat fluxes (SHs) and latent heat fluxes (LHs), determine temperature and moisture at the interface between soil or vegetation and the atmosphere, and then impact nearsurface atmospheric conditions [14,24], further resulting in changes in air temperature [24]. Figure 10 displays the 5 day averaged surface energy-flux differences between WRF_CHAR and WRF_CTL in EC. As is shown in Figure 10a, large areas had a negative SH difference, which was very similar with the Ta differences (Figure 6c), with an areaaveraged SH difference of −9.25 Wm −2 , especially east of Jiangsu, with a minimal SH difference of about −30 Wm −2 . LH differences presented the opposite pattern when comparing SH differences, which mainly showed positive values in most areas with an area average of 13.35 Wm −2 , and the maximal SH difference was more than 30 Wm −2 . Generally, net radiation flux (Rn) presented a positive difference with an area average of 3.57 Wm −2 , the positive Rn was mainly distributed in the southeastern region of the research area, and the negative difference appeared northwest of the research area. Soil heat flux (GH) showed a slight positive difference with an area average of 0.35 Wm −2 . In the south of the research area, GH differences of 1-3 Wm −2 appeared, but in the north of the research area, there was a −1 to −3 Wm −2 GH difference. The temporal variation of area-averaged surface energy-flux differences showed minimal SH differences, and maximal LH, Rn, and GH differences appearing in the day time, especially occurring at approximately 16:00 local standard time (LST), which corre- The temporal variation of area-averaged surface energy-flux differences showed minimal SH differences, and maximal LH, Rn, and GH differences appearing in the day time, especially occurring at approximately 16:00 local standard time (LST), which corresponds to Ta variation (Figure 7a). Obviously, the change in LSCP albedo, LAI, and soil moisture affected energy allocation. Albedo, LAI, and soil moisture affected the energy balance through shortwave radiation, canopy resistance, and surface evaporation, respectively. On the basis of Monin-Obukhov similarity theory, the surface energy flux established the relationship between the surface layer profiles of temperature, humidity, and wind speed through atmospheric turbulence in the atmosphere boundary layer, from which air temperature is changed.

Impact Comparison of Different LSCPs on Air Temperature
In this research, two LSCP types (albedo and LAI) and one type of land soil variation (SM) were integrated into the WRF model to research the impact of LSC on hot weather in EC. Sensitivity simulation WRF_CHAR, in which the LSCPs were updated, improved Ta and Tamax. However, it is still a problem that different LSCPs have a different effect or contribution on higher temperature, and which LSCP is the main affecting factor needs to be established. Figure 11 shows the scatter diagram of spatial-grid cell-based albedo, LAI, and SM of 10 cm differences, and the corresponding Ta differences of simulations from WRF_CHAR and WRF_CTL. Albedo and Ta differences had positive correlation with a value of 0.02 (Figure 11a). LAI and Ta differences had negative correlation with a value of −0.37 (Figure 11b). Moreover, SM10 and Ta differences obtained negative correlation of −0.68, which was the largest absolute value of the correlation coefficient (Figure 11c). wind speed through atmospheric turbulence in the atmosphere boundary layer, from which air temperature is changed.

Impact Comparison of Different LSCPs on Air Temperature
In this research, two LSCP types (albedo and LAI) and one type of land soil variation (SM) were integrated into the WRF model to research the impact of LSC on hot weather in EC. Sensitivity simulation WRF_CHAR, in which the LSCPs were updated, improved Ta and Tamax. However, it is still a problem that different LSCPs have a different effect or contribution on higher temperature, and which LSCP is the main affecting factor needs to be established. Figure 11 shows the scatter diagram of spatial-grid cell-based albedo, LAI, and SM of 10 cm differences, and the corresponding Ta differences of simulations from WRF_CHAR and WRF_CTL. Albedo and Ta differences had positive correlation with a value of 0.02 (Figure 11a). LAI and Ta differences had negative correlation with a value of −0.37 (Figure 11b). Moreover, SM10 and Ta differences obtained negative correlation of −0.68, which was the largest absolute value of the correlation coefficient (Figure 11c). Figure 11. Density scatter diagram of (a) albedo difference, (b) LAI difference, and (c) 10 cm soil moisture difference, and corresponding air-temperature difference.
The spatial distribution of the passed confidence-test correlation coefficient between Ta, albedo, LAI, and SM10 differences further supports the results of Figure 11 ( Figure  12). The correlation coefficient between Ta and SM10 differences presented patchy largearea negative values, and the other correlation coefficients mainly presented point spatial distribution. Figure 12. Spatial distribution of correlation coefficient (r) between (a) air-temperature and albedo difference, (b) LAI difference, and (c) 10 cm soil water difference over the research period in EC. Figure 11. Density scatter diagram of (a) albedo difference, (b) LAI difference, and (c) 10 cm soil moisture difference, and corresponding air-temperature difference.
The spatial distribution of the passed confidence-test correlation coefficient between Ta, albedo, LAI, and SM10 differences further supports the results of Figure 11 (Figure 12). The correlation coefficient between Ta and SM10 differences presented patchy large-area negative values, and the other correlation coefficients mainly presented point spatial distribution.
To further confirm the result, excluding WRF_CTL and WRF_CHAR, four more numerical simulations were performed: WRF_ALB (just updated the albedo), WRF_LAI (just updated the LAI), WRF_ALL (updated the albedo and LAI), and WRF_CLDAS (just updated the four layers' SM). The hourly variation of area-averaged observed Ta and simulated Ta from six numerical simulations and their differences are displayed in Figure 13. Comparing the five other simulations, area-averaged Ta from WRF_CLDAS were closer to those from WRF_CHAR, which together integrated the albedo, LAI and SM, especially in the daytime, which means that SM had the maximal contribution to improving the warming Ta bias. The albedo and LAI just resulted in a slight change in area-averaged Ta, and the effect of albedo and LAI on Ta and Tamax is mainly reflected in spatial distribution characteristics (figures not shown).
The spatial distribution of the passed confidence-test correlation coefficient betwee Ta, albedo, LAI, and SM10 differences further supports the results of Figure 11 (Figur 12). The correlation coefficient between Ta and SM10 differences presented patchy large area negative values, and the other correlation coefficients mainly presented point spatia distribution. To further confirm the result, excluding WRF_CTL and WRF_CHAR, four more nu merical simulations were performed: WRF_ALB (just updated the albedo), WRF_LAI (jus

Conclusions
In this study, to investigate the impact of LSCP albedo and LAI and SM on hot We integrated remote-sensing LSCP albedo and LAI and assimilation SM into the WRF model, and air temperature was improved. However, there are still remaining biases between the simulated and observed air temperature. This is probably due to uncertainties in the WRF model, atmospheric forcing data, and LSCP albedo, LAI, and assimilation SM data. In the past few years, numerical forecasting models have been improved, but some processes are still poorly simulated [13]. Additionally, atmospheric forcing data from different general atmospheric-circulation models or general atmosphere-ocean circulation models still have differences, which can also induce inaccurate simulations [51]. Satellitebased LSCP albedo and LAI and assimilation SM data are better than before, but different remote-sensing albedo and LAI and assimilation SM still have differences and can cause air temperature bias [52,53].

Conclusions
In this study, to investigate the impact of LSCP albedo and LAI and SM on hot weather that occurred in EC on 21-25 July 2017, a control simulation with the WRF model default LSCPs, and a series of sensitivity simulations with the satellite-and assimilation-based data were performed. The simulated Ta and Tamax were compared with the station observations. Then, we compared and analyzed the results of the control, and performed a series of sensitivity simulations. The conclusions can be summarized as follows: 1.
MODIS albedo and LAI, and CLDAS SM can better describe spatial distribution characteristics, and match fine weather forecasts and high-resolution simulations than the default climatology albedo and LAI can. The area-averaged accuracy of CLDAS SM data was higher by 12.36% to 18.09% than that of the ERA SM.

2.
The control simulation WRF_CTL could capture the spatial distribution of Ta and Tamax, but overestimated the Ta and Tamax with a warming bias of 1.05 and 1.32 • C, respectively. Sensitivity simulation WRF_CHAR improved the simulated Ta and Tamax by 33.08% and 29.24%, respectively.

3.
Comparing the WRF_CTL simulation, the WRF_CHAR simulation presented an area-averaged SH difference of −9.25 Wm −2 , LH difference of −9.25 Wm −2 , and Rn difference of 3.57 Wm −2 . The updated albedo, LAI, and SM changed the partition of surface energy and then resulted in a change of Ta and Tamax.

4.
Soil moisture is the main factor to improve warming bias in hot weather in EC based on the scatter diagram and the spatial correlation coefficient of albedo, LAI, and SM10 and Ta.