1. Introduction
Global warming and related changes very likely caused by anthropogenic increases in greenhouse gases can have devastating impacts on human society and the natural environment (Karl and Easterling 1999; IPCC 2007; Peterson et al. 2008; Hu et al. 2012; IPCC 2012; Wang et al. 2012; IPCC 2013; WMO 2013). Particularly, East Asia is expected to be highly vulnerable to the future extreme impacts under global warming (IPCC 2012), and there are the remarkably concentrated industry and high population densities in East Asian countries, as mentioned in Ho et al. (2011).
Reanalyses are multidecadal, gridded datasets produced by assimilating long time series of observations with a fixed and modern numerical weather prediction (NWP) model and data assimilation system to estimate a wide range of atmospheric, sea state, and land surface variables which are spatially complete and physically consistent because the equations of motion and physical processes are used (Dee et al. 2014). Such datasets are used for a great variety of applications (e.g., climate variability, chemistry transport, parameterization studies; Rood and Bosilovich 2010).
Most reanalyses have been produced for the global area (e.g., Schubert et al. 1993; Kalnay et al. 1996; Gibson et al. 1997; Kistler et al. 2001; Kanamitsu et al. 2002; Uppala et al. 2005; Onogi et al. 2007; Bosilovich 2008; Saha et al. 2010; Dee et al. 2011; Ebita et al. 2011; Rienecker et al. 2011; Kobayashi et al. 2015). Reanalyses that include the first half of the twentieth century also have been developed through many projects (e.g., Compo et al. 2008, 2011; Hersbach et al. 2015). Additional details of each reanalysis (e.g., coordinating organization, period, assimilation method, resolution) are described in Yang and Kim (2017).
In addition, due to the limitation of the coarse resolution of global reanalyses in analyzing regional-scale phenomena (e.g., Chen et al. 2014), many meteorological organizations started or planned to produce regional reanalyses with higher resolution (e.g., Mesinger et al. 2006; Bromwich et al. 2010; MoES and NOAA 2010; Renshaw et al. 2013; Mahmood et al. 2014; Barker et al. 2015; Borsche et al. 2015).
Accordingly, the East Asia Regional Reanalysis (EARR) system was recently developed based on the Unified Model, which has been an operational NWP model in the Korea Meteorological Administration (KMA; Yang and Kim 2017). The EARR system was established for the 2-yr period of 2013–14 to prepare for the long-term production of the EARR. Since a reanalysis is an estimate of the atmospheric state (Mahmood et al. 2014; Borsche et al. 2015), verification is crucial. The quality of regional reanalyses can be assessed mainly through 1) accuracy of short-range reforecasts integrated from an initial condition that is reanalysis and 2) comparison of reanalysis with independent observations not assimilated to produce the reanalysis itself, and it is vital to compare accuracy against global reanalysis, as mentioned in Barker et al. (2013).
Dee et al. (2011) evaluated the performance of ERA-Interim (hereinafter ERA-I) reanalysis by comparing skills of daily forecasts of tropical temperatures and winds obtained with reanalyses because the performance of the reforecasts can be affected by not only the quality of the model used but also the accuracy, completeness, and physical coherence of the reanalyzed fields. Especially, precipitation fields obtained from the North American Regional Reanalysis (NARR; Mesinger et al. 2006) and European regional reanalyses, which are not analyzed variables but forecast variables produced by forecast model dynamics and physics parameterizations, were examined against other reanalyses (Bukovsky and Karoly 2007; Dahlgren et al. 2016; Jermey and Renshaw 2016). Because of the higher resolution and better representation of orography in regional reanalysis, precipitation phenomena can be better captured in regional reanalyses than global reanalyses.
Although Yang and Kim (2017) investigated the characteristics and uncertainties of the EARR against ERA-I and observation data, they used dependent observations used in assimilation for verification and evaluated the reanalysis variables except for reforecast variables. Verification of reforecast fields against independent observation which is not used in producing reanalysis makes the verification results more reliable. Therefore, in this study, the quality of the EARR is evaluated by assessing the performance of precipitation reforecast. The reforecasts of EARR are compared with the reforecasts of ERA-I reanalysis and the operational forecasts of the KMA (OPER).
In section 2, the EARR system used to produce the EARR reforecast is described in comparison with the KMA operational system and how to monitor the process for producing reanalysis is introduced. Furthermore, verification methods are explained in section 2. The verification results in terms of equitable threat score (ETS), frequency bias index (FBI), probability of detection (POD), false alarm ratio (FAR), and comparison of spatial fields are presented and discussed in section 3. In addition, section 3 presents case studies and evaluation of forecast fields from the same NWP model. Finally, section 4 presents the summary and conclusions.
2. Methods
a. East Asia Regional Reanalysis system
1) Model and data assimilation system
The EARR system was developed on the Unified Model (UM; Davies et al. 2005) at the KMA, which has been an operational NWP model at the KMA since 2011 (Yang and Kim 2017). The UM used to produce the EARR is also used to produce reforecast fields of the EARR as the NWP model in this study. With respect to model resolution, the horizontal resolution is 12 km, and 70 vertical levels up to 80 km from the surface are used. The EARR domain is shown in Fig. 1. The Met Office four-dimensional variational data assimilation (4DVAR) scheme (Courtier et al. 1994; Rawlins et al. 2007) is used.
The domain of EARR. A blue box denotes the verification area.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
The lateral boundary conditions (LBCs) are provided by the KMA operational global UM with 25-km horizontal resolution (N512). Because the global UM produces analysis fields every 6 h and the LBCs are updated every 3 h, the analysis and 3-h forecast fields from the global UM are used as the LBCs (Fig. 2). More specific and detailed information on the model and observations in the EARR system can be found in Yang and Kim (2017).
A schematic diagram of the cycling of the forecast model and data assimilation system for EARR.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
In this study, the KMA operational forecast (hereinafter OPER) and ERA-I reforecast fields are introduced to compare with the EARR reforecast fields. The OPER is simply forecasts with 12-km horizontal resolution from the operational regional analyses produced by the KMA. The OPER is not able to benefit from additional datasets which are not available at the time when operational regional analyses are made, the latest updates with model, and the reprocessing, as mentioned in Smith et al. (2014), whereas the EARR is able to do so.
In addition, when the EARR system was established, there were several updates and improvements with model, compared to that for the OPER. As mentioned in Shin et al. (2015), the specific differences between the EARR and OPER are as follows:
For the OPER, soil moisture information was updated once a day, but for the EARR it is updated four times a day (0000, 0600, 1200, 1800 UTC). This information is provided by the global UM surface analysis process.
The typhoon bogus process is improved in the EARR system, compared to the OPER system. The typhoon bogus process being used at the KMA is to strongly reflect real-time information on typhoons into the model by inserting synthetic typhoon data simulated by observed data. Detailed improvements are as follows. When the intensity of the typhoon is weak, the synthetic typhoon algorithm is not applied. In contrast, for a typhoon with strong intensity, more data on the synthetic typhoon are inserted by alleviating quality control standards.
The thinning distance for satellite observations is modified from 80 to 125 km, which is determined by conducting sensitivity tests.
The climate value of eight kinds of aerosol used in the global UM is newly applied to longwave radiation parameterization in the EARR system.
The input error regarding time information of observation data from Japan [Automated Meteorological Data Acquisition System (AMeDAS)] is corrected in the EARR system.
Additionally, the ERA-I reforecast fields, provided by ECMWF, are also investigated to compare with the EARR reforecast fields. The ERA-I, whose horizontal resolution is approximately 80 km, is a global reanalysis, and the data assimilation method used to produce ERA-I is 4DVAR.
The reforecast fields are obtained by integrating reanalysis fields (i.e., initial condition) for 6 h. For two different periods, the EARR reforecast, OPER forecast, and ERA-I reforecast fields are obtained by the integration from their own analysis or reanalysis (i.e., initial condition) by their own NWP model used for the production of their analysis or reanalysis.
2) Monitoring during production
During the reanalysis production period, it is necessary to monitor if all the procedures work properly for every cycle. This is important because the reanalysis is sequentially generated based on cycling. Hence, this section introduces the monitoring system used for the EARR.
Figure 3 shows the real-time spatial distribution of observations for aircraft; sonde; surface observations; Global Positioning System Radio Occultation (GPSRO); scatterometer-derived winds (scatwind); satellite-derived winds (satwind); the Advanced TIROS Operational Vertical Sounder (ATOVS), which consists of High Resolution Infrared Radiation Sounder (HIRS) and the Advanced Microwave Sounding Unit-A (AMSU-A); and Infrared Atmospheric Sounding Interferometer (IASI) on 31 October 2014. This is useful, when there is some error regarding observations, to monitor if observation data are provided appropriately or to confirm what kind of observation has a problem.
The real-time spatial distribution of observations: (a) aircraft, (b) sondes, (c) surface, (d) GPSRO, (e) scatwind, (f) satwind, (g) ATOV, and (h) IASI at 1200 UTC 31 Oct 2014.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
The real-time observation minus background (O − B) and observation minus analysis (O − A) for surface, aircraft, and sonde observations are shown in Fig. 4. For surface observations (Figs. 4a–c) including land surface synoptic weather observations (SYNOP), surface weather observations and reports (METAR), sea surface weather observations by ship (Ship) and buoy (Buoy), and bogus observations generated by national meteorological centers (BOGUS) and aircraft observations (Figs. 4d–f), the time series of O − B and O − A are shown for every 0000 and 1200 UTC in July 2014. For sonde observations (Figs. 4g–i), O − B and O − A are shown as a function of model levels at the time of 0000 UTC 1 July 2014. The difference between O − B and O − A is used to confirm if all the observations are assimilated appropriately.
The real-time observation minus background (O − B; gray) and observation minus analysis (O − A; black) of surface observations, including land surface synoptic weather observations, surface weather observations and reports, sea surface weather observations by ship, sea surface weather observations by buoy, and bogus observations generated by national meteorological centers, for surface variables (a) u wind, (b) temperature, and (c) relative humidity in July 2014; aircraft observations for (d) u wind, (e) υ wind, and (f) theta in July 2014; and sonde observations for (g) u wind, (h) theta, and (i) relative humidity at 0000 UTC 1 Jul 2014.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
In addition, the time series of the initial 4DVAR cost or penalty function at every cycle from 16 December 2012 to July 2013 are shown in Fig. 5, which shows one example of the monitoring system where an error occurred during production for EARR. There are some drastic drops in the initial cost function for several periods (i.e., December 2012; January, February, and July 2013), indicating that some observations were not assimilated due to some systematic error. To rectify this omission of observations, the reanalysis for this period was produced again. Therefore, the time series of the initial cost function are useful to check if a certain observation is omitted, especially during the massive production of the reanalysis.
The real-time initial 4DVAR cost function for every cycle from 16 Dec 2012 to July 2013.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
b. Verification methods
1) Periods
In this study, two different verification periods are selected for 14 days in the summer and winter, respectively. Reforecasts (or forecasts) are verified twice a day at 0600 and 1800 UTC for the period from 10 to 23 July 2013 (hereinafter 201307) and the period from 1 to 14 February 2014 (hereinafter 201402).
2) Observations
For the verification of the precipitation variable, it may be fairer to use the gridded observation data as the truth rather than the point-based observation data, as indicated in Jermey and Renshaw (2016). Although there are a gauge-based analysis of daily precipitation over East Asia for the period from 1978 to 2003 (Xie et al. 2007) and a NOAA Climate Prediction Center (CPC) unified gauge-based analysis of global daily precipitation from 1979 to present (Chen et al. 2008), their spatial resolution is 0.5° × 0.5° and temporal resolution is daily. There is no very high-resolution gridded observation dataset in East Asia which can be used for verification. Therefore, validation against point-based observations in East Asia is examined in this study. For verification of 6-h accumulated precipitation over East Asia, global surface weather observations (PREPBUFR format), which are operationally collected by the National Centers for Environmental Prediction (NCEP), are used. The verification domain is denoted by the blue box in Fig. 1 (22°–53°N, 105°–148°E). The reforecast (or forecast for OPER) fields are interpolated to the location of observation using bilinear interpolation (Press et al. 1992).
In this section, the POD, FAR, ETS, and FBI are considered for the verification of precipitation. These verification scores are obtained based on the 2 × 2 contingency table (Wilks 2006) in Table 1. For two different periods (i.e., 201307 and 201402), 6-h accumulated precipitation fields based on the 6-h forecasts from the initial time at every 0000 and 1200 UTC are verified with respect to those verification scores.
3) Verification metrics
(i) POD and FAR
The 2 × 2 contingency table for dichotomous (yes/no) events.
(ii) ETS and FBI
(iii) Comparison of spatial fields
In the previous sections, verification scores are calculated to access objectively the accuracy of the reforecast or forecast from EARR, OPER, and ERA-I; thus, the spatial fields of precipitation are also depicted to understand or confirm the results of verification scores intuitively and to analyze subjectively the characteristics and uncertainty of the predicted precipitation.
3. Results
a. Equitable threat score
The point-based ETSs for two different periods (201307 and 201402) are displayed in Figs. 6a and 6c, respectively. For both of those periods, the EARR ETS is greater than the ERA-I ETS, suggesting that EARR, whose resolution is relatively high, outperforms ERA-I with respect to the precipitation. In addition, the EARR ETS is also greater than the OPER ETS for the two periods.
(a),(c) The ETS and (b),(d) FBI of the 6-h accumulated precipitation, (top) from 10 to 23 Jul 2013 and (bottom) from 1 to 14 Feb 2014, for the reforecast fields of EARR (blue), OPER (red), and ERA-I (black).
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
For the summer period (201307), the ETSs of EARR and OPER are higher than those of ERA-I. Overall, the ETSs of EARR and OPER for the summer period decrease as a threshold increases, which indicates that the reforecast accuracy of EARR and OPER at lower thresholds is higher than that at higher thresholds (Fig. 6a). This might be because lower threshold events tend to cover larger spatial scales, which are more comparable with the point data (i.e., observation stations), as Jermey and Renshaw (2016) mentioned. Although the ETS difference between EARR and OPER is not distinct compared to the ETS difference between EARR and ERA-I, the EARR ETS is greater than the OPER ETS for the summer period.
For the winter period (201402), the ETSs of EARR and OPER are greater than those of ERA-I for all thresholds (Fig. 6c). For all thresholds, the EARR ETS is greater than the OPER ETS for the winter period, although the ETS difference between EARR and OPER at thresholds of 8 and 16 mm (6 h)−1 is small.
Overall, ETS for the winter period is greater than that for the summer period. This seasonal variation is similar to the results shown in Ebert et al. (2003) and Jermey and Renshaw (2016). As Jermey and Renshaw (2016) described, precipitation in the winter is primarily controlled by the large-scale atmospheric motions, which are much easier for a model to capture than small-scale ones (i.e., convection) that mostly occur in the summer.
b. Frequency bias index
The point-based FBI for two different periods (201307 and 201402) is displayed in Figs. 6b and 6d, respectively. FBI is considered to be closer to observations as FBI is closer to 1. Overall, the EARR FBI is closer to 1 than the ERA-I FBI for all thresholds.
For the summer period (Fig. 6b), the FBIs of EARR, OPER, and ERA-I are greater than 1 for lower thresholds and are less than 1 for higher thresholds. In particular, the ERA-I FBI is much greater than 1 for lower thresholds, which indicates that ERA-I tends to overforecast the occurrence of light precipitation more than EARR and OPER do. Moreover, for higher thresholds, the ERA-I FBI is much less than 1, implying that ERA-I tends to underforecast the occurrence of heavy precipitation more than EARR and OPER do. The EARR FBI is much closer to 1 compared to the OPER FBI at lower thresholds, whereas the OPER FBI is much closer to 1 compared to the EARR FBI at higher thresholds. Overall, the FBI of EARR and OPER is a lot closer to 1 compared to that of ERA-I, which might be related to higher ETSs of EARR and OPER.
For the winter period (Fig. 6d), similar to the summer period, EARR, OPER, and ERA-I tend to overforecast for lower thresholds and underforecast for higher thresholds. However, the FBIs of EARR and OPER are closer to 1, compared to that of ERA-I, which indicates that EARR and OPER are much closer to observation compared to ERA-I. The difference between EARR FBI and OPER FBI is not distinct, compared to that for the summer period.
For lower thresholds, the FBIs of EARR, OPER, and ERA-I for the summer period (Fig. 6b) are much larger than those for the winter period (Fig. 6d), although those are already greater than 1. That is why the ETSs of EARR, OPER, and ERA-I at lower thresholds for the summer period (Fig. 6a) are much lower than those for the winter period (Fig. 6c).
c. Probability of detection and false alarm ratio
The point-based POD and FAR for two different periods are shown in Fig. 7. Overall, POD shows a decline toward higher rainfall intensities for two periods.
(a),(c) The POD and (b),(d) FAR of the 6-h accumulated precipitation, (top) from 10 to 23 Jul 2013 and (bottom) from 1 to 14 Feb 2014, for EARR (blue), OPER (red), and ERA-I (black).
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
For the summer period (201307), ERA-I shows higher POD than the others at lower thresholds (Fig. 7a). As discussed in the previous section, ERA-I tends to overforecast light precipitation. Due to the tendency of ERA-I to overforecast at lower thresholds, probability of precipitation detection of ERA-I is higher than that of EARR and the ERA-I FAR is higher than the EARR FAR at lower thresholds. High POD and FAR of ERA-I at lower thresholds imply that ERA-I is likely to overforecast the occurrence of light precipitation. Overall, EARR shows lower POD and FAR than OPER and ERA-I at lower thresholds. Meanwhile, at higher thresholds, POD and FAR of EARR and OPER are higher than those of ERA-I, similar to the trend of FBI. For the summer period, the trends of FBI, POD, and FAR are similar to each other.
For the winter period (201402), EARR shows higher POD than the others at all thresholds. The difference between EARR POD and ERA-I POD increases as a threshold increases. Overall, FARs of EARR, OPER, and ERA-I for the winter period are smaller than those for the summer period, although the ranges of POD for each period are similar to each other. These results are associated with higher ETS for the winter period than that for the summer period.
d. Comparison of spatial fields
The 6-h accumulated precipitation fields of observation and three products’ reforecast (or forecast) are shown at the specific time when heavy precipitation occurred during the summer and the winter period in Figs. 8–10.
The 6-h accumulated precipitation of the (left) observation and reforecasts [(center left) ERA-I, (center right) EARR, and (right) OPER]. The initial times are selected every 12 h from (a)–(d) 0000 UTC 11 Jul to (u)–(x) 1200 UTC 13 Jul 2013.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
As in Fig. 8, but the initial times are selected every 12 h from 0000 UTC 18 Jul to 1200 UTC 20 Jul 2013.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
As in Fig. 8, but the initial times are selected every 12 h from 1200 UTC 6 Feb to 0000 UTC 9 Feb 2014.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
In the left column of Figs. 8–10, filled dots indicate observation stations. The color of the dot represents the intensity of precipitation at the station, so small white dots indicate no precipitation at the station.
1) The summer period (201307)
The 6-h accumulated precipitation fields of observation, two reforecasts, and one operational forecast during the summer period are investigated. In this section, two different shorter periods in the summer period are chosen and each part is shown in Figs. 8 and 9, respectively.
For the first part, the initial times are chosen every 12 h from 0000 UTC 11 July to 1200 UTC 13 July 2013, as shown in Fig. 8. For the precipitation accumulated for 6 h from 0000 UTC 11 July 2013 (Figs. 8a–d), EARR, OPER, and ERA-I capture well heavy precipitation over North Korea. However, ERA-I tends to overforecast light precipitation over south-central and east China, and only EARR is able to represent small-scale precipitation over Shandong Province, which is the northern part of east China. For the precipitation accumulated for 6 h from 0000 UTC 12 July 2013 (Figs. 8i–l), EARR and OPER successfully simulate heavy rainfall over the eastern end of the Shandong Peninsula and the eastern area of North Korea, whereas ERA-I fails and tends to overforecast light precipitation. Furthermore, for the precipitation on 13 July 2013 (Figs. 8s,w), only EARR simulates well the location and intensity of the thin, long shape of heavy precipitation over Shandong Province in east China or the Korean Peninsula.
To analyze the later part in the summer period, the initial times are chosen every 12 h from 0000 UTC 18 July to 1200 UTC 20 July 2013 (Fig. 9). For the precipitation accumulated for 6 h from 0000 UTC 18 July 2013 (Figs. 9a–d), ERA-I wrongly forecasts light precipitation over south-central and east China and the Korean Peninsula, and EARR and OPER forecast heavier precipitation than ERA-I. For the precipitation accumulated for 6 h from 1200 UTC 18 July 2013 (Figs. 9e–h), although the location of precipitation captured by EARR and OPER is slightly displaced northward over China and EARR and OPER simulate stronger intensity of precipitation over China compared to observations, EARR and OPER are able to better simulate the strong intensity and location of the long, narrow shape of precipitation over China than ERA-I. For the precipitation accumulated for 6 h from 0000 UTC 19 July 2013 (Figs. 9i–l), ERA-I wrongly simulates light precipitation over southern China. In addition, all reforecasts and operational forecasts capture well the small-scale precipitation over northeast China, but EARR and OPER are able to capture its structure, which has two cores, in greater detail than ERA-I.
For the precipitation accumulated for 6 h from 1200 UTC 19 July 2013 (Figs. 9m–p), EARR and OPER simulate heavier precipitation over the northern part of northeast China, compared to observations. However, EARR and OPER represent the location and strong intensity of small-scale precipitation over northeast China more exactly than ERA-I in Figs. 9q–t as well as Figs. 9m–p. For the precipitation accumulated for 6 h from 0000 UTC 20 July 2013 (Figs. 9q–t), ERA-I overforecasts light precipitation over southern China and misses heavy precipitation over the northern part in south-central China. In contrast, even though the location of precipitation from EARR and OPER is slightly displaced, EARR and OPER capture the strong intensity of it. For the precipitation accumulated for 6 h from 1200 UTC 20 July 2013 (Figs. 9u–x), two reforecasts and one operational forecast represent well heavy precipitation over North Korea.
2) The winter Period (201402)
The 6-h accumulated precipitation fields of observation, two reforecasts, and one operational forecast during the winter period (201402) are shown in Fig. 10. The initial times are selected every 12 h from 1200 UTC 6 February to 0000 UTC 9 February 2014.
For the precipitation accumulated for 6 h from 1200 UTC 6 February 2014 and from 0000 UTC 7 February 2014 from the ERA-I reforecast (Figs. 10b,f), ERA-I wrongly predicts light precipitation over south China, similar to the summer period. However, this is more noticeable in the summer period than in the winter period. For the precipitation accumulated for 6 h from 1200 UTC 7 February 2014 (Figs. 10i–l), ERA-I overforecasts light precipitation over inland South Korea, where there is no precipitation, whereas EARR and OPER are able to forecast no precipitation over that area. For the precipitation accumulated for 6 h from 1200 UTC 8 February 2014 (Figs. 10q–t), EARR and OPER can represent more detailed features of precipitation over southern China than ERA-I.
e. Case studies
In this section, two extreme precipitation events that occurred in the Korean Peninsula for the summer and winter are investigated to demonstrate how EARR can improve the accuracy of high-impact weather forecasts. Two cases are selected among two periods examined in this study. To investigate more detailed features of heavy precipitation in the Korean Peninsula, data from 479 Automatic Weather Stations (AWSs) provided by the KMA are used for verification. The 6-h accumulated precipitation fields from observed data, two reforecasts, and one operational forecast in the Korean Peninsula are examined.
1) Case on 12 July 2013
As the first case, the observed precipitation accumulated for 6 h from 1200 UTC 12 July 2013 has a peak value of 111 mm in Seoul, the capital city of South Korea, which is located in the middle part of the Korean Peninsula (Fig. 11a). This heavy rain occurred during the East Asian summer monsoon season, known as changma (in Korea), mei-yu (in China), and baiu (in Japan). During this period, the North Pacific high was extended to the southern area in the Korean Peninsula, and atmospheric instability was intensified due to an upper-tropospheric trough.
As in Fig. 8, but the initial times are (a)–(d) 1200 UTC 12 Jul 2013 and (e)–(h) 0000 UTC 7 Feb 2014. South Korea is investigated.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
The location of heavy precipitation predicted by ERA-I, EARR, and OPER is slightly displaced compared to the observed fields where a strong intensity of precipitation is concentrated on the western area in the middle part of the Korean Peninsula (Figs. 11a–d). Nonetheless, ERA-I, EARR, and OPER well capture the overall precipitation area concentrated in the middle part of the Korean Peninsula. Although ERA-I, EARR, and OPER are not able to simulate locally extreme precipitation, EARR and OPER are able to represent the stronger intensity of precipitation than ERA-I. Even though OPER captures heavier precipitation than EARR, the position of peak precipitation of OPER is still misplaced, and overall ETSs of EARR at this time are greater than those of OPER (not shown).
Because of the difficulty in predicting convective activities, resolving precipitation activities at short time scales and their propagation has been a challenge. Wang et al. (2009) found that the skill scores for precipitation forecasts associated with midtropospheric perturbation (MP) were low as a result of the propagation speed bias of MPs and underpredicted precipitation amounts. In this study, because the resolutions of regional reanalyses and their assimilation are much higher than those of global reanalysis and its assimilation, much improvement in resolving convective rainfall can be obtained due to higher resolution, which may be a major advantage of regional reanalysis.
2) Case on 7 February 2014
The second case is a heavy snowfall event that occurred in the eastern coast of the Korean Peninsula and the eastern side of the Taebaek Mountains, which are located along the eastern edge of the Korean Peninsula (Fig. 11e). The precipitation fields (Figs. 11e–h) were accumulated for 6 h from 0000 UTC 7 February 2014.
In this case, the synoptic conditions were as follows. In the lower atmosphere, the easterly airflow was intensified by a north–south-oriented high-to-low pressure system over the eastern Korea. With easterly airflow, the expansion of cold air into the relatively warm East Sea of South Korea generated a great difference of temperature between the ocean and the air and a large amount of water vapor was provided. This easterly flow was blocked by the Taebaek Mountains and converged to the eastern side of the Taebaek Mountains and the eastern coast of the Korean Peninsula. Consequently, heavy precipitation was concentrated on the eastern side of the Taebaek Mountains.
Although ERA-I simulates the precipitation fields, which occurred on the eastern coast of the Korean Peninsula, ERA-I captures a smaller precipitation area and a weaker intensity of precipitation than the observed field (Fig. 11f). Meanwhile, EARR and OPER are able to accurately capture the position and the intensity of precipitation. Precipitation fields of EARR and OPER are very similar to observed precipitation fields and more accurate than those of ERA-I.
f. Evaluation of precipitation reforecast from the same NWP model
The results in the previous sections indicate that precipitation fields of EARR and OPER are more accurate than those of ERA-I. In this section, the causes of these differences in accuracy are examined. In general, reanalysis and reforecast datasets are developed based on one system. For this reason, datasets are dependent on their own system. The majority of the previous studies which focus on intercomparison of reanalyses or reforecasts did not use the same NWP model, but evaluated datasets themselves (Bukovsky and Karoly 2007; Bosilovich et al. 2008; Lorenz and Kunstmann 2012; Peña-Arancibia et al. 2013; Chen et al. 2014; Lader et al. 2016; Jermey and Renshaw 2016; Nkiaka et al. 2017).
However, in reality, reanalysis data can be used as initial and boundary conditions with whichever NWP model a researcher wants to use. In this section, 6-h accumulated precipitation forecast fields are generated using the same NWP model, which is the Weather Research and Forecasting (WRF, v3.7.1) Model, initialized by EARR, OPER, and ERA-I analyses with boundary conditions from ERA-I. It needs to be noted that when the WRF Model is used, there are the following weaknesses for EARR and OPER, compared to ERA-I. For integrating forecast fields initialized from EARR and OPER, boundary conditions are provided by ERA-I, not by EARR or OPER. In addition, some near-surface variables, soil moisture, and soil temperature are provided by ERA-I because of the limitations of UM output format to the WRF Model.
In the WRF Model, the horizontal resolution used is 12 km with 430 × 330 grid points. The 6-h precipitation forecasts are integrated every 12 h (initialized from 0000/1200 UTC) for 4 months. For the summer and winter seasons, two months in July 2013 and 2014 (hereinafter summer) and two months in January and February 2014 (hereinafter winter) are investigated, respectively. These new periods include two different periods of 2 weeks investigated in the previous sections. For verification, the same verification domain (blue box in Fig. 1) and the same observation dataset are used.
For the period of two months in the summer season, ETSs of EARR, OPER, and ERA-I have similar trend and values (Fig. 12a). Meanwhile, compared to Fig. 6a, the values of ETS are reduced, especially for EARR and OPER, suggesting that the performance of EARR and OPER is much better at representing convective rainfall for the summer season when the UM is used compared to the WRF Model. In addition, the error of EARR reforecasts in the WRF Model may be generated by boundary conditions (i.e., ERA-I) that are from different the model compared to the model used to generate the EARR, whereas ERA-I reforecasts in the WRF Model are generated using ERA-I for both initial and boundary conditions. Furthermore, compared to Fig. 6a, the difference of ETS between EARR and ERA-I is reduced, which implies that the same model may simulate similar features of precipitation forecasts. Nevertheless, EARR ETS is higher than OPER ETS (Fig. 12a), which is the same as the previous result. For higher thresholds [(4, 8, and 16 mm (6 h)−1], EARR ETS is higher than ERA-I ETS, although EARR has weaknesses mentioned above when integrated in the WRF Model, compared to ERA-I.
As in Fig. 6, but for (top) summer in July 2013 and 2014 and (bottom) winter in January and February 2014, for the reforecast fields of EARR (blue), OPER (red), and ERA-I (black) based on the WRF Model.
Citation: Journal of Hydrometeorology 20, 2; 10.1175/JHM-D-18-0068.1
In addition, for the summer season, FBIs of EARR, OPER, and ERA-I (Fig. 12b) are similar to each other and much higher than 1, compared to Fig. 6b. Even for higher thresholds, all FBI values are greater than 1. The FBI values that are greater than 1 for all thresholds indicate that the WRF Model (or convective parameterization scheme used in this section) tends to overestimate the frequency of precipitation. Even though the FBI difference between EARR and ERA-I is indistinct, EARR FBI is closer to 1 than ERA-I FBI for higher thresholds.
For the winter season, while the values of ETS (Fig. 12c) are a bit reduced compared to Fig. 6c, especially for EARR and OPER, they show a pronounced difference between EARR and ERA-I for all thresholds. Like Fig. 6c, EARR ETS is higher than ERA-I ETS for all thresholds. FBIs of EARR, OPER, and ERA-I for the winter season are smaller than 1 for all thresholds (Fig. 12d), implying that the WRF Model tends to underestimate the occurrence of precipitation at all thresholds for the winter season regardless of initial conditions. For all thresholds, EARR FBI is closer to 1 than ERA-I FBI.
4. Summary and conclusions
Reanalyses are important multidecadal and gridded datasets for the applications in a great variety of fields. Although global reanalysis datasets have been widely used, the need for regional reanalyses with higher resolution emerged around the world. Hence, a short period of regional reanalysis in East Asia was recently developed based on the UM and was evaluated against ERA-I and observation data (Yang and Kim 2017). In this study, the quality of regional reanalyses is evaluated through the accuracy of short-range reforecasts of precipitation integrated from a reanalysis which is an initial condition.
In this study, the OPER forecast and ERA-I reforecast fields are also evaluated to compare with the EARR reforecast fields. The OPER is not able to benefit from additional datasets, the latest updates with the model, and the reprocessing.
During the reanalysis production period, it is necessary to monitor whether all the procedures work properly for every cycle. For the production of EARR, the difference between O − B and O − A and the time series of initial 4DVAR cost function are used to monitor observations assimilated.
For the verification of 6-h accumulated precipitation reforecasts, two different verification periods are selected for 14 days in the summer (201307) and winter (201402), respectively.
For both of the two periods, the EARR and OPER ETS are greater than the ERA-I ETS, suggesting that EARR and OPER outperform ERA-I with relatively low resolution. Although the ETS difference between the EARR and OPER is not distinct compared to the ETS difference between the EARR and ERA-I, generally the EARR ETS is greater than the OPER ETS for two periods. Overall, ETS for the winter period is greater than that for the summer period since precipitation in the winter is mainly predicted by the large-scale atmospheric motions, which is easier to capture than convection activities in the summer.
With respect to FBI, overall, the EARR and OPER FBI are closer to 1 than the ERA-I FBI for all thresholds, indicating that the EARR and OPER are much closer to the observation compared to ERA-I, and EARR, OPER, and ERA-I tend to overforecast for lower thresholds and underforecast for higher thresholds. For the period 201307, the ERA-I FBI, which is much greater than 1 for lower thresholds (smaller than 1 for higher thresholds), implies that ERA-I tends to overforecast (underforecast) the occurrence of light (heavy) precipitation more than EARR and OPER do. For lower thresholds during the period 201307, greater FBIs of EARR, OPER, and ERA-I may be associated with lower ETSs of EARR, OPER, and ERA-I, compared to FBIs and ETSs during the period 201402.
In terms of POD and FAR, generally, POD for two periods shows a decline as a threshold increases. For the period 201307, high POD and FAR of ERA-I at lower thresholds are associated with the tendency of ERA-I to overforecast the occurrence of light precipitation. Furthermore, EARR shows lower FAR than the others at lower thresholds. For the period 201402, EARR shows higher POD than the others at all thresholds. Overall, FARs of EARR, OPER, and ERA-I for the winter period are smaller than those for the summer period, which may be related to higher ETS for the winter period.
In terms of spatial precipitation fields, EARR, OPER, and ERA-I generally capture heavy precipitation. However, ERA-I tends to overforecast light precipitation, especially in the period 201307, and EARR and OPER tend to predict heavier precipitation than ERA-I. Although the location of precipitation captured by EARR and OPER is slightly displaced, EARR and OPER are able to simulate the approximate location and strong intensity of the long, narrow shape of precipitation more accurately than ERA-I.
In addition, two extreme precipitation events that occurred in the Korean Peninsula are investigated. For high-impact weather forecasts, EARR and OPER are better at representing heavy precipitation in the Korean Peninsula than ERA-I. Due to the difficulty in representing convective activities, resolving precipitation activities and their propagation has been a challenge. In this study, much improvement in resolving convective precipitation can be obtained due to higher resolution, which may be a major advantage of regional reanalysis.
Because some users of reanalyses may use reanalysis data as initial and boundary conditions with an NWP model, the 6-h accumulated precipitation forecast fields initialized from EARR, OPER, and ERA-I are additionally generated every 12 h for 4 months using the WRF Model for more consistent results. Although the differences of ETS and FBI between EARR and ERA-I for the summer season are not distinct, the results with ETS and FBI based on the WRF Model for the summer and winter periods are consistent with those based on their own models.
Based on several evaluations, the reforecasts of precipitation of EARR are confirmed to be more accurate than those of OPER and ERA-I in East Asia for two different periods. In addition, EARR is better at representing precipitation fields than OPER, but the performance difference is not distinct compared to that between EARR and ERA-I. Regional reanalysis with higher resolution and better quality over East Asia would provide better representation of regional weather and climate phenomena over East Asia, which is highly vulnerable to the high-impact weather phenomena.
Acknowledgments
The authors appreciate three reviewers for their valuable comments. This study was supported by a National Research Foundation of Korea (NRF) grant funded by the South Korean government (Ministry of Science and ICT) (Grant 2017R1E1A1A03070968). The authors appreciate the Numerical Modeling Center and the National Center for Meteorological Supercomputer of the Korea Meteorological Administration and the Met Office for providing computer facility support and resources for this study. Please contact the corresponding author to obtain the datasets used in this study.
REFERENCES
Barker, D., R. Renshaw, and P. Jermey, 2013: Regional reanalysis. MOSAC and SRG Meetings 2013, Exeter, United Kingdom, Met Office, MOSAC PAPER 18.12, https://www.metoffice.gov.uk/binaries/content/assets/mohippo/pdf/library/mosac/mosac_18.12_barker.pdf.
Barker, D., R. Renshaw, and P. Jermey, 2015: Towards a UM-based regional reanalysis for Europe. 9th Annual CAWCR Workshop, Melbourne, Australia, Bureau of Meteorology, 24 pp., http://www.cawcr.gov.au/events/AWS9/PDF/Barker_Wed_CAWCR.pdf.
Borsche, M., A. K. Kaiser-Weiss, P. Undén, and F. Kaspar, 2015: Methodologies to characterize uncertainties in regional reanalyses. Adv. Sci. Res., 12, 207–218, https://doi.org/10.5194/asr-12-207-2015.
Bosilovich, M. G., 2008: NASA’s Modern Era Retrospective-Analysis for Research and Applications: Integrating Earth observations. Earthzine, https://earthzine.org/nasas-modern-era-retrospective-analysis/.
Bosilovich, M. G., J. Chen, F. R. Robertson, and R. F. Adler, 2008: Evaluation of global precipitation in reanalyses. J. Appl. Meteor. Climatol., 47, 2279–2299, https://doi.org/10.1175/2008JAMC1921.1.
Bromwich, D., Y.-H. Kuo, M. Serreze, J. Walsh, L. S. Bai, M. Barlage, K. Hines, and A. Slater, 2010: Arctic System Reanalysis: Call for community involvement. Eos, Trans. Amer. Geophys. Union, 91, 13–14, https://doi.org/10.1029/2010EO020001.
Bukovsky, M. S., and D. J. Karoly, 2007: A brief evaluation of precipitation from the North American Regional Reanalysis. J. Hydrometeor., 8, 837–846, https://doi.org/10.1175/JHM595.1.
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, M., W. Shi, P. Xie, V. B. S. Silva, V. E. Kousky, R. Wayne Higgins, and J. E. Janowiak, 2008: Assessing objective techniques for gauge-based analyses of global daily precipitation. J. Geophys. Res., 113, D04110, https://doi.org/10.1029/2007JD009132.
Compo, G. P., J. S. Whitaker, and P. D. Sardeshmukh, 2008: The 20th Century Reanalysis Project. Third WCRP Int. Conf. on Reanalysis, Tokyo, Japan, University of Tokyo, Japan, 5 pp., http://wcrp.ipsl.jussieu.fr/Workshops/Reanalysis2008/Documents/V5-511_ea.pdf.
Compo, G. P., and Coauthors, 2011: The Twentieth Century Reanalysis Project. Quart. J. Roy. Meteor. Soc., 137, 1–28, https://doi.org/10.1002/qj.776.
Courtier, P., J.-N. Thépaut, and A. Hollingsworth, 1994: A strategy for operational implementation of 4D-Var, using an incremental approach. Quart. J. Roy. Meteor. Soc., 120, 1367–1387, https://doi.org/10.1002/qj.49712051912.
Dahlgren, P., T. Landelius, P. Kållberg, and S. Gollvik, 2016: A high-resolution regional reanalysis for Europe. Part 1: Three-dimensional reanalysis with the regional HIgh-Resolution Limited-Area Model (HIRLAM). Quart. J. Roy. Meteor. Soc., 142, 2119–2131, https://doi.org/10.1002/qj.2807.
Davies, T., M. J. P. Cullen, A. J. Malcolm, M. H. Mawson, A. Staniforth, A. A. White, and N. Wood, 2005: A new dynamical core for the Met Office’s global and regional modelling of the atmosphere. Quart. J. Roy. Meteor. Soc., 131, 1759–1782, https://doi.org/10.1256/qj.04.101.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Dee, D. P., M. Balmaseda, G. Balsamo, R. Engelen, A. J. Simmons, and J.-N. Thépaut, 2014: Toward a consistent reanalysis of the climate system. Bull. Amer. Meteor. Soc., 95, 1235–1248, https://doi.org/10.1175/BAMS-D-13-00043.1.
Ebert, E. E., U. Damrath, W. Wergen, and M. E. Baldwin, 2003: The WGNE assessment of short-term quantitative precipitation forecasts. Bull. Amer. Meteor. Soc., 84, 481–492, https://doi.org/10.1175/BAMS-84-4-481.
Ebita, A., P. Kållberg, and Coauthors, 2011: The Japanese 55-year Reanalysis “JRA-55”: An interim report. SOLA, 7, 149–152, https://doi.org/10.2151/sola.2011-038.
Gibson, J. K., P. Kållberg, S. Uppala, A. Nomura, A. Hernandez, and E. Serrano, 1997: ERA description. ECMWF Re-Analysis Final Rep. Series 1, 71 pp.
Gilbert, G. F., 1884: Finley’s tornado predictions. Amer. Meteor. J., 1, 166–172.
Hersbach, H., C. Peubey, A. Simmons, P. Berrisford, P. Poli, and D. Dee, 2015: ERA-20CM: A twentieth-century atmospheric model ensemble. Quart. J. Roy. Meteor. Soc., 141, 2350–2375, https://doi.org/10.1002/qj.2528.
Ho, C.-H., and Coauthors, 2011: A projection of extreme climate events in the 21st century over East Asia using the Community Climate System Model 3. Asia-Pac. J. Atmos. Sci., 47, 329–344, https://doi.org/10.1007/s13143-011-0020-0.
Hu, Y., S. Maskey, and S. Uhlenbrook, 2012: Trends in temperature and rainfall extremes in the Yellow River source region, China. Climatic Change, 110, 403–429, https://doi.org/10.1007/s10584-011-0056-2.
IPCC, 2007: Climate Change 2007: Impacts, Adaptation and Vulnerability. M. L. Parry et al., Eds., Cambridge University Press, 976 pp.
IPCC, 2012: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press, 582 pp.
IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., https://doi.org/10.1017/CBO9781107415324.
Jermey, P. M., and R. J. Renshaw, 2016: Precipitation representation over a two-year period in regional reanalysis. Quart. J. Roy. Meteor. Soc., 142, 1300–1310, https://doi.org/10.1002/qj.2733.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1643, https://doi.org/10.1175/BAMS-83-11-1631.
Karl, T. R., and D. R. Easterling, 1999: Climate extremes: Selected review and future research directions. Climatic Change, 42, 309–325, https://doi.org/10.1023/A:1005436904097.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247–267, https://doi.org/10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.
Kobayashi, S., and Coauthors, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48, https://doi.org/10.2151/jmsj.2015-001.
Lader, R., U. S. Bhatt, J. E. Walsh, T. S. Rupp, and P. A. Bieniek, 2016: Two-meter temperature and precipitation from atmospheric reanalysis evaluated for Alaska. J. Appl. Meteor. Climatol., 55, 901–922, https://doi.org/10.1175/JAMC-D-15-0162.1.
Lorenz, C., and H. Kunstmann, 2012: The hydrological cycle in three state-of-the-art reanalyses: Intercomparison and performance analysis. J. Hydrometeor., 13, 1397–1420, https://doi.org/10.1175/JHM-D-11-088.1.
Mahmood, S., P. Jermey, and R. Renshaw, 2014: Methods for evaluating model performance of IMDAA. Forecasting Research Tech. Rep. 594, Met Office, 24 pp., https://www.metoffice.gov.uk/binaries/content/assets/mohippo/pdf/migrated/frtr594.compressed.pdf.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, https://doi.org/10.1175/BAMS-87-3-343.
MoES and NOAA, 2010: South Asian Regional Reanalysis (SARR). SARR Scoping Workshop Rep., Ministry of Earth Sciences, Government of India, and NOAA, 17 pp., http://www.ncmrwf.gov.in/SARR-workshop-report-final.pdf.
Nkiaka, E., N. R. Nawaz, and J. C. Lovett, 2017: Evaluating global reanalysis precipitation datasets with rain gauge measurements in the Sudano-Sahel region: Case study of the Logone catchment, Lake Chad Basin. Meteor. Appl., 24, 9–18, https://doi.org/10.1002/met.1600.
Onogi, K., and Coauthors, 2007: The JRA-25 reanalysis. J. Meteor. Soc. Japan, 85, 369–432, https://doi.org/10.2151/jmsj.85.369.
Peña-Arancibia, J. L., A. I. van Dijk, L. J. Renzullo, and M. Mulligan, 2013: Evaluation of precipitation estimation accuracy in reanalyses, satellite products, and an ensemble method for regions in Australia and South and East Asia. J. Hydrometeor., 14, 1323–1333, https://doi.org/10.1175/JHM-D-12-0132.1.
Peterson, T. C., X. Zhang, M. Brunet-India, and J. L. Vázquez-Aguirre, 2008: Changes in North American extremes derived from daily weather data. J. Geophys. Res., 113, D07113, https://doi.org/10.1029/2007JD009453.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, 1992: Numerical Recipes in Fortran 77: The Art of Scientific Computing. 2nd ed. Cambridge University Press, 933 pp.
Rawlins, F., S. P. Ballard, K. J. Bovis, A. M. Clayton, D. Li, G. W. Inverarity, A. C. Lorenc, and T. J. Payne, 2007: The Met Office global four-dimensional variational data assimilation scheme. Quart. J. Roy. Meteor. Soc., 133, 347–362, https://doi.org/10.1002/qj.32.
Renshaw, R., P. Jermey, D. Barker, A. Maycock, and S. Oxley, 2013: EURO4M regional reanalysis system. Forecasting Research Tech. Rep. 583, Met Office, 30 pp., https://www.metoffice.gov.uk/binaries/content/assets/mohippo/pdf/o/4/frtr583.pdf.
Rienecker, M. M., J. Pfaendtner, and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648, https://doi.org/10.1175/JCLI-D-11-00015.1.
Rood, R. B., and M. G. Bosilovich, 2010: Reanalysis: Data assimilation for scientific investigation of climate. Data Assimilation, W. Lahoz, B. Khattatov, and R. Menard, Eds., Springer, 623–646, https://doi.org/10.1007/978-3-540-74703-1_23.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, https://doi.org/10.1175/2010BAMS3001.1.
Schubert, S., J. Pfaendtner, and R. Rood, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc., 74, 2331–2342, https://doi.org/10.1175/1520-0477(1993)074<2331:AADFES>2.0.CO;2.
Shin, H.-C., H.-M. Noh, D.-J. Kim, T.-K. Jang, E. H. Lee, and S. Joo, 2015: Improvement of the numerical prediction system of the Korea Meteorological Administration to improve the forecasting performance of high-impact weather events in East Asia (in Korean). Proc. Autumn Meeting of Korean Meteorological Society, Daegu, South Korea, Korean Meteorological Society, 244–245.
Smith, G. C., F. Roy, P. Mann, F. Dupont, B. Brasnett, J.-F. Lemieux, S. Laroche, and S. Bélair, 2014: A new atmospheric dataset for forcing ice–ocean models: Evaluation of reforecasts using the Canadian global deterministic prediction system. Quart. J. Roy. Meteor. Soc., 140, 881–894, https://doi.org/10.1002/qj.2194.
Uppala, S. M., and Coauthors, 2005: The ERA-40 re-analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012, https://doi.org/10.1256/qj.04.176.
Wang, S. Y., T. C. Chen, and S. E. Taylor, 2009: Evaluations of NAM forecasts on midtropospheric perturbation-induced convective storms over the U.S. northern plains. Wea. Forecasting, 24, 1309–1333, https://doi.org/10.1175/2009WAF2222185.1.
Wang, Q. X., X. H. Fan, Z. D. Qin, and M. B. Wang, 2012: Change trends of temperature and precipitation in the Loess Plateau region of China, 1961–2010. Global Planet. Change, 92–93, 138–147, https://doi.org/10.1016/j.gloplacha.2012.05.010.
Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. 2nd ed. International Geophysics Series, Vol. 100, Academic Press, 648 pp.
Wilson, L., 2010: Verification of severe weather forecasts in support of the “SWFDP Southern Africa” project. Rep. for the WMO, 21 pp., www.wmo.int/pages/prog/www/BAS/documents/Doc-7-Verification.doc.
WMO, 2013: The global climate 2001–2010: A decade of climate extremes—Summary report. WMO-1119, World Meteorological Organization, 15 pp., https://library.wmo.int/pmb_ged/wmo_1119_en.pdf.
Xie, P., M. Chen, S. Yang, A. Yatagai, T. Hayasaka, Y. Fukushima, and C. Liu, 2007: A gauge-based analysis of daily precipitation over East Asia. J. Hydrometeor., 8, 607–626, https://doi.org/10.1175/JHM583.1.
Yang, E.-G., and H. M. Kim, 2017: Evaluation of a regional reanalysis and ERA-Interim over East Asia using in situ observations during 2013–14. J. Appl. Meteor. Climatol., 56, 2821–2844, https://doi.org/10.1175/JAMC-D-16-0227.1.
