Evaluation of the WRF experiments (CP3 and GZ9) is mainly for the surface climate of T2m, Tmax, Tmin, and precipitation, as well as the diurnal variation of precipitation over the TP. In order to compare with the in-situ observations and the satellite precipitation products, the WRF simulation results were interpolated onto the stations when compared with in-situ observations as well as the grid cells of the GSMaP dataset when compared with the satellite precipitation products, using the inverse distance weight interpolation method. Due to the elevation difference between WRF grids and station location, lapse rate (LR) is used to bias correct the WRF simulated surface air temperature when evaluation the WRF outputs against in-situ observations. According to the spatiotemporal variability of LR proposed by Wang et al. (2018), the mean LRs over the western TP, northeastern TP, and southeastern TP during summer are −4.90, −4.53, and −4.03 K/km, respectively, which are consistently lower than the commonly used global mean LR (−6.5 K/km) and are used for bias correction in this study.
3.1 Summer mean surface air temperature and precipitation
Figure 2 shows the 11-year averaged (2009-2019) summer mean daily T2m, Tmax, and Tmin from the in-situ observations, and the differences between the WRF simulations and the observations. The observed T2m decreases from the southeastern TP to the northwestern TP, with the maximum T2m at about 22℃ over the eastern TP while the minimum T2m below 8℃ over the central TP. Both the WRF experiments can well reproduce the spatial pattern of T2m with the spatial correlation coefficients (SCCs) larger than 0.94, but underestimate the T2m over the regions south of 35°N. Compared to GZ9, CP3 clearly improve the surface air temperature simulation with lower biases. The simulated distributions of Tmax and Tmin also agree well with the observations, with the SCCs above 0.90 and the RMSEs below 2.7℃. Overall, CP3 tends to simulate higher surface air temperature than GZ9 over the TP, and thus shows more skillful performance in reproducing the spatial patterns of T2m, Tmax, and Tmin with reduced RMSEs and higher SCCs, especially over the southern TP. The daily temperature range (DTR) is higher over the regions north of 30°N according to the station observations, exceeding 13℃ (figure not shown), while the minimum DTR below 10°C is detected over the southeastern TP. All the WRF experiments can well simulate the spatial pattern of DTR with the SCCs higher than 0.75 and the RMSEs less than 1.5℃. However, obvious clod bias exists over the southern TP. Moreover, CP3 can reduce the RMSE by about 0.8°C over the southeastern TP compared with GZ9.
Figure 3 shows the 11-year averaged summer mean precipitation from the station observations, the GSMaP satellite dataset, and the differences between the WRF simulations and the observations. The observed precipitation decreases from southeast to northwest, with the maximum above 6mm/day located at the southeastern corner of TP and the minimum less than 1mm/day over the northeast TP. Both CP3 and GZ9 can well capture the spatial distribution of summer mean precipitation, with SCCs larger than 0.72 and RMSEs below 1.1 mm/day compared with the station observations. However, the WRF model clearly underestimates summer precipitation over most regions of TP, especially over the southern TP. When compared against the GSMaP precipitation dataset, WRF simulations underestimate summer precipitation over southeastern TP and overestimate it over northwestern TP. Both CP3 and GZ9 have similar SCCs above 0.7 and relatively large RMSEs greater than 2mm/day. Comparison with both observational datasets shows consistency over the southeastern TP where both WRF simulations demonstrate dry bias. However, in most regions over the northwestern TP without the station observations, results show that GZ9 has a larger wet bias than CP3 based on the comparison against the GSMaP dataset. Thus, it may be deduced that by increasing the horizontal resolution, the dry (wet) bias over the southeastern (northwestern) TP can be reduced. In general, both experiments can reproduce the distribution of summer mean precipitation over TP and show consistent dry bias over southeastern TP and wet bias over northwestern TP. With higher horizontal resolution than GZ9, CP3 tends to improve the precipitation simulation by producing more precipitation over southeastern TP and less precipitation over the northwestern TP.
The Taylor diagrams are presented to evaluate the two WRF experiments in simulating the spatial distributions of summer temperature and precipitation (Figure 4) over the TP for each year (2009-2019). GZ9 and CP3 have similar performances in simulation the distributions of T2m, Tmax, and Tmin with the SCCs larger than 0.95. CP3 slightly improved performance with higher SCC and closer distance to observations (REF point) than GZ9. For the precipitation, the two WRF experiments also show comparable performances when compared with the station observations. However, CP3 outperforms GZ9 with higher SCCs (> 0.7) when the GSMaP precipitation dataset is used as a reference. In general, both experiments exhibit comparable ability in reproducing the spatial pattern of summer mean surface air temperature and precipitation during 2009-2019, with reasonable cold bias and dry bias, especially over the regions south of 35°N.
3.2 Daily surface temperature and precipitation
The 11-year averaged (2009-2019) daily variations of the regional mean (over the TP) T2m, Tmax, Tmin, and DTR from the in-situ observations and WRF experiments are shown in Figure 5. The observed T2m ranges from 12℃ to 16℃ throughout the summertime, with the maximum T2m in early and middle July. All the WRF experiments can well capture the daily variation of T2m with the temporal correlation coefficients (TCCs) higher than 0.95 and the RMSEs less than 1.1℃. CP3 outperforms GZ9 by reducing the cold bias. The Tmax ranges from 18℃ to 24℃ based on the observation, and CP3 and GZ9 also well simulate the daily variation with the cold bias of about 1.0℃ for CP3 and 2.0℃ for GZ9. For Tmin, CP3 can also reduce the RMSE by about 0.6℃ compared to GZ9. The observed DTR varies between 9℃ and 15℃, with the minimum DTR occurring in early July. Both WRF experiments reproduce the daily DTR variation with the TCCs larger than 0.91 and the RMSEs less than 1.2℃, and colder bias occurs in late July and early August. CP3 tends to simulate the DTR closer to observation. Both WRF experiments show comparable performance in reproducing the daily variation of surface air temperature and DTR, but underestimation is obviously reduced in CP3, especially for Tmax with the most significant reduction at about 1℃.
Regarding the daily variation of regional mean precipitation over the TP (Figure 6), both CP3 and GZ9 can reproduce the daily variation with the TCCs all above 0.9 and the RMSEs below 0.75 mm/day when compared with the station observations. Compared with in-situ observation, both experiments can better simulate the precipitation with negligible bias in early and middle June but growing dry bias since early July. While the two experiments obviously overestimate the daily precipitation when selecting the GSMap as the reference, with wetter bias in GZ9 has and a larger RMSE of about 0.92 mm/day than CP3. Whether station observation or GSMaP dataset is selected as reference, both WRF experiments consistently have TCCs above 0.9 and RMSEs below 1mm/day.
The spatial distributions of TCCs and RMSEs of T2m, Tmax, and Tmin at each observation station are shown in Figure 7 and Figure 8, respectively. The spatial patterns of TCCs of T2m and Tmax are quite similar in both CP3 and GZ9, with high TCCs above 0.9 are located over the northeastern TP and decrease from north to south. Compared with GZ9, CP3 exhibits higher TCCs of T2m and Tmax, especially over the southern TP. The WRF model has relatively lower performance in simulating variation of Tmin than that of T2m, with the TCCs ranging from 0.5 to 0.8. GZ9 shows higher TCCs for Tmin compared with CP3. For RMSEs, both experiments show large RMSEs (above 3.0℃) for Tmax over the southern TP and for Tmin over the northern TP, while CP3 demonstrates smaller RMSE for T2m compared to GZ9. CP3 can also reduce RMSE for Tmax and Tmin over the southern TP, especially for Tmax with the maximum reduction at about 1.6℃. Therefore, it can be concluded that CP3 improves the simulation of T2m and Tmax with higher TCCs and lower RMSEs, especially over the southern TP.
The spatial distribution of TCC and RMSE of the simulated daily precipitation for WRF experiments against station observation is presented in Figure 9. The two experiments show quite similar spatial patterns of TCC and RMSE. High TCCs exist over eastern TP and low RMSEs are located over central and northern TP. CP3 has slightly increased the TCCs by about 0.1 and reduced the RMSEs by about 0.5 mm/day over the southeastern TP, which is also validated by the comparison against GSMaP (figure not shown).
In general, both CP3 and GZ9 show comparable performance in reproducing the daily variation of surface air temperature, and CP3 obviously improves the performance by reducing the cold biases. Meanwhile, CP3 improves the ability to reproduce T2m and Tmax with higher TCCs and lower RMSEs compared to GZ9, especially over the southern TP. For precipitation, both experiments can well reproduce the spatiotemporal characteristics, while CP3 shows better skills with slightly increased TCC and reduced RMSE over the southeastern TP.
3.3 Diurnal cycle of precipitation
To evaluate the performance of WRF in simulating the diurnal variation of summer precipitation over the whole TP, the GSMaP satellite precipitation dataset is selected as the reference. The precipitation frequency (PF) is defined as the percentage of all hours during the period which had measurable precipitation (> 0.1 mm/h), the precipitation intensity (PI) is defined as the average precipitation rate over all the precipitating hours, and the precipitation amount (PA) is defined as the accumulated precipitation amount over a given time period.
The 11-year averaged (2009-2019) occurrence time of maximum PA and PF in a day from the observation and WRF experiments is presented in Figure 10. The observed maximum PA and PF mostly occurs after 18:00 Local Standard Time (LST) over the TP, with the latest occurrence time of maximum PA after 22:00 LST located over the southeastern TP. All the WRF experiments can capture the spatial pattern of the occurrence time of maximum PA, with about 3 hours in advance over the eastern TP. However, over the central and western TP, both experiments simulate a postponed occurrence time of maximum PA of about 2 hours.
Figure 11 shows the 11-year averaged regional mean diurnal cycle of PA and PI over the whole TP and four subregions (TP-NW, TP-SW, TP-NE, and TP-SE). It can be found that the observed PA ranges from 0.05 to 0.2 mm/hour over the TP, with two peaks occurring at 17:00 LST and 22:00 LST. Both CP3 and GZ9 can reproduce the bimodal structure of the diurnal variation of PA over the TP, but they all have delayed the peak at 22:00 LST by 3 hours. GZ9 can successfully capture the afternoon precipitation peak while CP3 slightly simulates the peak by 1 hour earlier. Compared with GZ9, CP3 shows reduced wet bias, especially from late afternoon to night. The diurnal variations of PF are similar to those of PA (figure not shown) with obvious nocturnal precipitation. Both GZ9 and CP3 can capture the diurnal variation of PI with consistent overestimation, which may contribute to the overestimation of PA. Both experiments can well capture the characteristics of the diurnal cycle of PA over TP-NW but with stronger amplitudes, where the RB is about 117.1% in CP3 and 148.8% in GZ9 respectively. The WRF experiments show the most skillful performance in simulating the diurnal variation over TP-NE with the most reduced wet bias. Over TP-SE and TP-SW, the observed diurnal variations of PA are quite similar, with peaks occurring in the afternoon and night. Both experiments can generally reproduce the diurnal variation with obvious dry bias at nighttime.
In general, WRF experiments tend to simulate much stronger afternoon precipitation over the eastern TP, which is ahead of observation, while simulating stronger nighttime precipitation over the western TP, which is behind the observation. CP3 can reduce wet bias over the northern TP, while it shows little superiority in reducing the dry bias at nighttime over the southern TP. Again, as a satellite-derived product, GSMaP needs to be calibrated with station observations. However, there are only fewer station sites located over the central and western TP. Currently, the uncertainty in this satellite-derived dataset is hard to be avoided. With more observations and reduced uncertainty in precipitation datasets, a better understanding can be got.