Abstract

In this study, the causes of the underestimated diurnal 2-m temperature range and the overestimated 2-m specific humidity in the winter of northern China in the Rapid-Refresh Multiscale Analysis and Prediction System–Short Term (RMAPS-ST) are investigated. Three simulations based on RMAPS-ST are conducted from 1 November 2016 to 28 February 2017. Further analyses show that the partitioning of surface upward sensible heat fluxes and downward ground heat fluxes might be the main contributing factor to the 2-m temperature forecast bias. In this study, two simulations are conducted to examine the effect of soil moisture initialization and soil hydraulic property on the 2-m temperature and 2-m specific humidity forecasts. First, the High-Resolution Land Data Assimilation System (HRLDAS) is used to provide an alternative soil moisture initialization. The results show that the drier soil moisture could lead to noticeable change in energy partitioning at the land surface, which in turn results in improved prediction of the diurnal 2-m temperature range, although it also enlarges the 2-m specific humidity bias in some parts of the domain. Second, a soil texture dataset developed by Beijing Normal University and the revised hydraulic parameters are applied to provide a more detailed description of soil properties, which could further improve the 2-m specific humidity bias. In summary, the combination of using optimized soil moisture initialization, an updated soil map, and revised soil hydraulic parameters can help improve the 2-m temperature and 2-m specific humidity prediction in RMAPS-ST.

1. Introduction

The Rapid-Refresh Multiscale Analysis and Prediction System–Short Term (RMAPS-ST), the operational short-range numerical weather prediction (NWP) system of Beijing Meteorological Service (BMS), is a WRF-based system developed by Institute of Urban Meteorology (BMS/IUM). However, noticeable systematic bias of 2-m temperature (T2) and 2-m specific humidity (Q2) forecasts are found in the RMAPS-ST operational forecasts in the winter. The near-surface temperature and humidity forecasts are important for economic development and people’s daily life, so to find out the causes of such forecasting bias is of great urgency.

A number of studies have investigated the reasons for the poor performance of near-surface temperature and humidity forecasts in NWP. Numerous hypotheses concerning the sources of these near-surface temperature forecast biases have been proposed, including inadequate horizontal or vertical resolution, inaccurate initialization and parameterization of boundary layer physics, and the imperfect land surface characteristics and processes (Hanna and Yang 2001; Mass et al. 2002; Cheng and Steenburgh 2005). Many studies have focused on the initialization and the parameterization of land surface characteristics and processes, which control the surface energy budget and can contribute to near-surface temperature errors due to the inaccurate partitioning of sensible, latent, and ground heat fluxes (Huang et al. 1996; Davis et al. 1999; Marshall et al. 2003; Reeves et al. 2011; Massey et al. 2014; Lin and Cheng 2016; Dy and Fung 2016).

For land surface models used in regional NWP systems, soil moisture and temperature are typically derived from global NWP model predicted fields, for example, National Centers for Environmental Prediction (NCEP) model or European Centre for Medium-Range Weather Forecasts (ECMWF) model. Generally, these models have a coarser horizontal resolution than the regional NWP system and the data are less representative when interpolated to the finer regional grids. Moreover, soil moisture evolution has a negligible impact on the mean near-surface thermodynamic variables in contrast to initial soil moisture (Trier et al. 2008). The improper initialization of soil moisture can directly bring about the near-surface temperature and humidity forecast errors (e.g., Massey et al. 2014; Dy and Fung 2016).

Past studies have also reached the conclusion that the snow cover and its amount are important components in the interaction between the land surface and atmosphere, affecting the near-surface air temperature due to snow’s insulating properties and the latent heat needed for snowmelt (Thomas 2008; Tomasi et al. 2017). It is clear that the accurate representation of snow cover in NWP models is vital for the calculation of surface fluxes over snow-covered surfaces and subsequent forecasts of atmospheric variables. It might be a cause of the systematic bias in the near surface forecasts in the winter.

The soil textures and hydraulic parameters are also important in soil state simulations. Soil texture directly affects soil hydraulic conductivity, field capacity, and wilting point, which are important soil hydraulic parameters for soil moisture simulation (Xia et al. 2015). The soil’s physical and hydraulic properties play an important part in the accurate simulation of land surface hydrological processes. The importance of soil textures on land surface hydrological processes has been demonstrated in several studies (Warrach-Sagi et al. 2008; Zhang et al. 2012; Xia et al. 2015; Lin and Cheng 2016).

RMAPS-ST employs the Noah land surface model (LSM; Chen and Dudhia 2001). The soil parameters used in Noah LSM are determined from a lookup table that is dependent on the corresponding soil texture. The default soil texture data in WRF-Noah utilizes two soil datasets, the State Soil Geographic (STATSGO) dataset from the U.S. Department of Agriculture (USDA; Miller and White 1998), which has 30-s resolution within the conterminous United States, and the Food and Agriculture Organization (FAO) soil dataset from the United Nations, which was published in 1991 with 5-min resolution over the entire globe (FAO and UNESCO 1974, 1981; FAO 1991). It is possible that this dataset may not accurately represent the detailed soil features in regions outside the United States, Shangguan et al. (2013) have developed the Beijing Normal University (BNU) China soil dataset based on a soil map at a scale of 1:1 000 000, obtained from the second national soil survey in China from 1979 to 1985, the national and provincial soil books, and 8979 distinct soil profiles. Shangguan et al. (2014) further used several soil databases spread over the world to create a global soil dataset, which has been widely used in recent studies (Zheng et al. 2015; Lin and Cheng 2016; Sun et al. 2017). Additionally, Kishné et al. (2017) compared default soil hydraulic parameters for Noah LSM with measured soil properties from a soil database that includes 6749 soil samples located within and around Texas, and they found that 95% of the default soil parameters are significantly different from the region-specific measured values. They then produced a revised soil parameter lookup table for Noah LSM. In general, these studies have provided a possibility of carrying out NWP simulations by using a more detailed soil map and an alternative set of soil parameters.

In this paper, the causes of the systematic bias of the T2 and Q2 forecasts in the winter of 2016/17 over northern China in RMAPS-ST are investigated. Several factors, such as snow cover, surface energy budget, and soil moisture initialization, are studied to find out which factor contributes most to the inaccurate diurnal temperature range and the obvious wet humidity bias. The High-Resolution Land Data Assimilation System (HRLDAS; Chen et al. 2007) is used to provide an alternative soil moisture initialization. In addition, updated soil texture dataset with revised hydraulic parameters is applied to provide a more detailed description of soil properties. Three simulations are conducted to investigate the impact of soil moisture, soil map, and soil parameters on the T2 and Q2 bias.

The remainder of this paper is organized as follows. Section 2 briefly introduces RMAPS-ST, as well as its 2-m temperature and 2-m humidity bias in winter. Detailed analyses of the reasons for the bias are also given in section 2. Simulations and the results are described and discussed in section 3. Section 4 summarizes the paper.

2. RMAPS-ST and its winter performance in northern China

a. RMAPS-ST

RMAPS-ST is a WRF-based regional NWP system. It employs 9- and 3-km one-way nested domains (see Fig. 1) covering the whole of China and northern China, respectively, and it has 51 vertical levels with a model top of 50 hPa. The physics packages include the RRTMG longwave and shortwave radiation parameterization (Iacono et al. 2008), the Noah LSM (Chen and Dudhia 2001), the Yonsei University PBL parameterization (Hong et al. 2006), the Thompson microphysics scheme, and the Kain–Fritsch cumulus parameterization (Kain 2004). The last one is only used in the 9-km domain.

Fig. 1.

(a) The nested domains of RMAPS-ST. (b) The inner domain of RMAPS-ST. The locations of the 53484 and 54706 observing sites and geographic locations referenced in the paper are marked. The shading indicates the terrain height.

Fig. 1.

(a) The nested domains of RMAPS-ST. (b) The inner domain of RMAPS-ST. The locations of the 53484 and 54706 observing sites and geographic locations referenced in the paper are marked. The shading indicates the terrain height.

RMAPS-ST operates in a rapid-updated partial cycling mode. Each cycle consists of nine runs: a cold start at 1800 UTC, providing 6-h forecasts for initializing the 0000 UTC run, and eight warm starts at subsequent times of 0000, 0300, 0600, 0900, 1200, 1500, 1800, and 2100 UTC, using forecasts from the previous run as the background. Data assimilated in RMAPS-ST include upper-air observations, aircraft data, surface observations, ground-based GPS zenith total delay, and radar observations. The initial and boundary conditions of cold runs and the boundary conditions of all runs are from ECMWF forecast at 0.25° grid spacing and 3-h time interval.

b. Winter T2 and Q2 performance

RMAPS-ST forecasts are validated against observations from meteorological stations. The surface weather observation data used for evaluating model performance in this study is obtained from the China meteorological data sharing system. Specifically, there are 744 surface observation stations in the inner domain. The measurements include hourly T2 and Q2. The bias between forecasts and observed daily 2-m maximum and minimum temperatures (tmax and tmin) is used to study the RMAPS-ST T2 forecast error. In addition, the Q2 bias at 0900 UTC is typical, so it is used to study the Q2 forecast error. Routine verifications in winter show that the RMAPS-ST underestimates the diurnal temperature range, with a warm bias tmin and a cold bias tmax in northern China. The 2-m tmin and tmax in winter usually appear early in the morning (about 2100 UTC) and in the afternoon (about 0600 UTC), respectively. Moreover, it is shown that there is an obvious wet (positive) bias in Q2. In contrast, RMAPS-ST shows a warm and dry bias during daytime, weak cold bias during nighttime in summer, which is different with forecast bias in winter. The causes of forecast bias in summer are distinct from that in winter. Both 9- and 3-km resolution domains present similar performances in northern China. Therefore, our investigation on the causes of the surface variable bias focuses on the 3-km resolution domain. To clarify the causes of the bias mentioned above, a BASELINE simulation from 1 November 2016 to 28 February 2017 is conducted based on RMAPS-ST. In the RMAPS-ST, data assimilation and warm start could impact the forecast and the impact needs to be analyzed individually. So the BASELINE simulation does not assimilate any observations. All simulations are cold starts initialized at 0000 UTC, and the initial and boundary conditions are from the same global forecast that is used in RMAPS-ST operational cold start runs.

The BASELINE simulation is validated against the same observed dataset used in RMAPS-ST routine verification. In the BASELINE simulation, the time-averaged 2-m tmax bias, 2-m tmin bias, and Q2 bias at 0900 UTC are shown in Fig. 2. The cool bias of 2-m tmax appears at most observing sites, especially those located around the northeast–southwest diagonal line of the domain. Most negative 2-m tmax bias less than −3.0°C appears at the observing sites in the middle mountainous areas. The warm bias of 2-m tmin occurs at most observing sites. The number of observing sites with warm bias greater than 3°C is far beyond that with cold bias smaller than −3°C. The warm bias at nighttime is more apparent than the cold bias in daytime; one of the causes is severe air pollution in the winter over the plain areas. The aerosol-induced effect on shortwave radiation is not coupled in this study, so there is excess shortwave radiation on the surface during daytime because T2 is warmer in the model forecast than that with the aerosol effect. The cause the bias of 2-m tmax is not apparent over the plains. Besides, the maxima of wet Q2 bias at 0900 UTC are over 0.6 g kg−1 at some observing sites. The general characteristics of daily T2 and Q2 bias from the BASELINE simulations are very similar to that from RMAPS-ST operational runs within the same duration. As the BASELINE simulation is conducted without data assimilation and in cold running mode, it is clear that such bias is dominated by physical processes.

Fig. 2.

(a) 2-m tmax bias, (b) 2-m tmin bias, and (c) 2-m specific humidity bias at 0900 UTC at observing sites from the BASELINE simulations, averaged between 1 Nov 2016 and 28 Feb 2017. (d) Accumulated number of days with snow cover at observing sites. (e) 2-m tmax and (f) 2-m tmin scatterplots; the interval of the blue line is 3°C.

Fig. 2.

(a) 2-m tmax bias, (b) 2-m tmin bias, and (c) 2-m specific humidity bias at 0900 UTC at observing sites from the BASELINE simulations, averaged between 1 Nov 2016 and 28 Feb 2017. (d) Accumulated number of days with snow cover at observing sites. (e) 2-m tmax and (f) 2-m tmin scatterplots; the interval of the blue line is 3°C.

The comparison of surface shortwave radiation forecasts in the BASELINE with the observation (figure not shown) indicates that the surface radiation is not the cause for such a bias over the mountain areas. The snow cover on the ground can increase the surface albedo, thereby altering the energy and mass balance; influencing surface heat fluxes, ground temperature, runoff, and soil moisture; and providing a feedback mechanism that modulates atmospheric variability. It is necessary to figure out if snow cover is the possible reason for such a bias.

In this paper, snow cover is simply defined as the presence of snow at 0000 UTC at observing sites. The number of snow cover days at each observing site is counted and shown in Fig. 2d. For about 50% of the observing sites used in the validation, the number of days with snow cover is smaller than 5. During the winter, only 10.7% of all the stations (744 sites in 120 days) are covered by snow. For the scatterplots in Figs. 2e and 2f, the bias of stations with no snow cover is similar to that with snow cover. The plots also indicate that the warm bias of tmin is more severe than the cold bias of tmax. The results demonstrate that the snow cover is not the crucial cause of the inaccurate prediction of 2-m diurnal temperature amplitude.

c. Surface heat fluxes

Figure 3 displays the spatial distribution of daytime (0300–0600 UTC averaged) and nighttime (1800–2100 UTC averaged) averaged heat fluxes from the BASELINE simulations. During the daytime, upward sensible heat flux (SHF) is dominate, downward ground heat flux (GHF) comes second, and the latent heat flux (LHF) is much smaller than the other two fluxes. The highest SHF occurs in the western domain where the land cover is shrubland and barren vegetation with lower LHF. Comparing the heat flux over the areas that have noticeable cold bias of 2-m tmax with that over other areas, the upward SHF is lower at a magnitude greater than 20 J m−2 (Fig. 3a). The downward GHF is clearly higher over southwestern mountain areas in daytime, with a maximum value up to 60 J m−2 (Fig. 3c). Surface energy partitioned to upward SHF heats the near-surface atmosphere in daytime. Lower daytime upward SHF could lead to less heating and lower 2-m temperature in daytime. During the nighttime, the upward GHF is much higher than the other two fluxes. The SHF comes from the atmosphere to land surface, while SHF in most area is slightly positive. It is clear that the higher upward GHF corresponds to noticeable warm 2-m tmin bias. The more surface energy partitioned to downward GHF in daytime, the more surface energy is transported to the atmosphere in the form of upward GHF at nighttime. Higher nighttime upward GHF could lead to more heating of the land surface and help to maintain the warm surface temperature that causes higher 2-m temperature at night. Higher downward GHF in daytime and upward GHF at nighttime coincide with the cold T2 in daytime and warm T2 at nighttime.

Fig. 3.

Daytime and nighttime averaged (a),(d) SHF, (b),(e) LHF, and (c),(f) GHF from the BASELINE simulations, averaged between 1 Nov 2016 and 28 Feb 2017. To clearly express the spatial distribution of the heat flux, subplots have different color bars.

Fig. 3.

Daytime and nighttime averaged (a),(d) SHF, (b),(e) LHF, and (c),(f) GHF from the BASELINE simulations, averaged between 1 Nov 2016 and 28 Feb 2017. To clearly express the spatial distribution of the heat flux, subplots have different color bars.

In Noah LSM, ground heat flux G is computed by a flux–gradient relationship (Rosero et al. 2010):

 
G=DF1STC1T10.5×ZSOIL(1).
(1)

STC1 is the temperature at the center of the first soil layer (0.5 × ZSOIL(1)) and T1 is the surface temperature. DF1 is the heat conductivity of the surface soil layer, which is dominated by the soil water content and soil porosity. It indicates that soil moisture and texture jointly determine the soil heat conductivity and the ground heat flux. The decreasing (increasing) DF1 will bring about a smaller (greater) downward ground heat flux in daytime and thereby a smaller (greater) upward ground heat flux at night. Furthermore, it can result in more (less) energy allocated to increase SHF and the near-surface temperature in the day and less (more) energy allocated to increase near-surface temperature at night.

In addition, there is a mismatch between the cold bias in daytime and high downward GHF over the plain areas. One sound reason is the air pollution over the plain areas in the winter. The aerosol-induced effect on shortwave radiation is not considered in this study.

d. Soil moisture

The initial soil moisture fields for the RMAPS-ST system are provided by ECMWF model with a coarser resolution than RMAPS-ST. As atmospheric boundary layer thermodynamic properties are always related to soil moisture, soil variables might be considered to be a parameter to be tuned to compensate for various model bias (Di Giuseppe et al. 2011). Being highly dependent on the atmospheric model formulation, soil moisture from one model might be unsuitable for another.

The initial top 10-cm soil temperature (hereafter referred to as surface soil temperature) and initial top 10-cm soil moisture (hereafter referred to as surface soil moisture) from the BASELINE simulation are averaged and shown in Figs. 4a and 4b. Averaged surface soil temperature is below freezing over the mountain areas and the northeastern domain. The surface soil moisture in most domains is higher than 0.24 m3 m−3, except the western domain. The three wettest soil moisture areas are over the northeast, the middle mountain, and the south domain. In addition, over the south domain soil water is liquid. While over the other two areas, soil temperature is under 0°C, and the frozen water proportion is nearly linear to soil temperature, and as a result, soil moisture contains a large fraction of frozen water as shown in Fig. 4d. As the thermal conductivity of frozen soil is higher than that of unfrozen soil (Zhang et al. 2007), DF1 is high and could lead to more downward GHF in daytime. The pattern of soil frozen water corresponds to the pattern of GHF in daytime (Fig. 3c), except in the northern domain covered by snow. Moreover, the spatial patterns of soil moisture and frozen water have a strong connection to the T2 bias. It shows that the areas with 2-m tmax bias greater than −3°C (Fig. 2a) agree well with the areas with surface soil moisture greater than 0.3 m3 m−3 and surface soil temperature in the range from −4° to 2°C near the middle of the domain, where the warm bias of T2 is also apparent.

Fig. 4.

(a) Surface soil temperature; the black line denotes the averaged 0°C contour. Surface soil moisture of (b) total, (c) liquid, and (d) solid.

Fig. 4.

(a) Surface soil temperature; the black line denotes the averaged 0°C contour. Surface soil moisture of (b) total, (c) liquid, and (d) solid.

The analyses bring about a hypothesis that the noticeable T2 bias might tightly relate to unsuitable soil moisture initialization. Additional simulations adopting a more suitable soil moisture initialization should be conducted to clarify to what degree the initial soil moisture contributes to such noticeable T2 bias.

3. Numerical experiments and analyses

a. Impact of soil moisture

1) Experiment outline

Related to the surface moisture and heat transfer in land surface processes, soil moisture content has a strong effect on the surface temperature. It is clearly shown in section 2 that the soil moisture initialization in the BASELINE simulation is closely related to the noticeable 2-m temperature forecasting bias in northern China. The SOILMOIS simulation is therefore conducted and analyzed in this section. The SOILMOIS experiment is the same as the BASELINE simulation except for a different initial soil moisture field. It is expected to find out to what extent the initial soil moisture variation might impact T2 bias. HRLDAS is adapted to provide a different soil moisture initialization.

The HRLDAS is run in an uncoupled model on the same horizontal grids, with the same terrain/land-use/soil texture fields and LSM parameters as RMAPS-ST system. It allows the HRLDAS soil moisture to be directly ingested into RMAPS-ST system without spatial interpolation. The HRLDAS is driven by real-time meteorological observations with a dense gauge network, for example, 2-m temperature, 2-m water vapor mixing ratio diagnosed from relative humidity, 10-m wind speed, station pressure, precipitation, and downward longwave and shortwave radiation. Quasi-equilibrium conditions of soil moisture and temperature are obtained by using a long spinup time up to 18 months in HRLDAS.

2) Soil moisture initialization

The initial surface soil moisture for all SOILMOIS and BASELINE simulations is averaged respectively. Figure 5a shows the averaged initial surface soil moisture in SOILMOIS. It generally varies smoothly from low to high from northwestern to southeastern part of the domain. The distribution pattern of soil moisture in SOILMOIS follows the pattern of rainfall (figure not shown) well, as the precipitation is an important parameter that determines the soil moisture distribution. The evaluation against observation also demonstrate the reliability of the soil moisture from HRLDAS [section 3a(5)]. The averaged initial surface soil moisture difference between SOILMOIS and BASELINE is presented in Fig. 5b. In most areas, the averaged initial surface soil moisture from SOILMOIS is drier than that from BASELINE, with the driest areas in the middle mountains and the northeastern domain. There are also some positive differences in the western domain.

Fig. 5.

The top-layer initial soil moisture averaged from November 2016 to February 2017 (a) from SOILMOIS and (b) the difference of that between SOILMOIS and BASELINE. Averaged heat conductivity at initial time in (c) BASELINE and (d) SOILMOIS simulations.

Fig. 5.

The top-layer initial soil moisture averaged from November 2016 to February 2017 (a) from SOILMOIS and (b) the difference of that between SOILMOIS and BASELINE. Averaged heat conductivity at initial time in (c) BASELINE and (d) SOILMOIS simulations.

In the SOILMOIS simulation, bigger (smaller) DF1 (Figs. 5c,d) occurs in area with decreasing (increasing) surface soil moisture than that in the BASELINE simulation, except in urban areas where the values are the same (greater than 2.0) in both simulations. Noticeable changes occur in the middle and northeastern domain, with the value dropping from greater than 1.2 to about 0.6. In the western domain, DF1 decreases from greater than 2.0 in BASELINE to about 1.6 in SOILMOIS. Remarkable changes of DF1 also occur in Mongolia, but it is not discussed here for the lack of observations. Significantly, there are some unnatural hot spots over the western area in the heat conductivity of BASELINE. These hot spots are in the area where the soil texture is sand with wet soil moisture. When the soil moisture is initialized with HRLDAS and the soil texture is updated, most of the hot spots vanish. The general decrease of DF1 around the diagonal area of the 3-km domain in SOILMOIS is expected to change the surface heat fluxes, which will be discussed in the following.

3) Impact on surface heat fluxes

Figure 6 shows heat flux differences between SOILMOIS and BASELINE. Here, heat fluxes averaged during 0300–0600 UTC are taken as representative values of daytime, and heat fluxes averaged during 1800–2100 UTC are for nighttime. Figure 6a shows that the averaged daytime SHF increases (decreases) by the amount up to 50 J m−2 in the area with decreased (increased) soil moisture. The maximum change of SHF agrees well with the largest change in soil moisture and heat conductivity. Figure 6b shows that the averaged daytime LHF decreases in the area with decreased soil moisture by as much as 50 J m−2. The largest flux changes collocate in the area with soil moisture lowered by more than 0.2 m3 m−3. Reduced soil moisture results in less water available to be transported to the surface for evaporation and transpiration, leading to a decrease in LHF. The downward daytime ground flux also decreases (increases) in the area with decreased (increased) soil moisture (Fig. 6c), while the upward ground flux of nighttime decreases (increases) in the area with decreased (increased) soil moisture (Fig. 6d). It means that less energy is transferred into ground during the day, and therefore less energy is available to be transferred into surface from ground during the night. Smaller amount of upward ground fluxes at night may lead to less error in the prediction of 2-m tmin. Although the GHF variation caused by soil moisture change is smaller than that for SHF and LHF, the change in near-surface temperature is expected to be slightly higher at nighttime than that in daytime.

Fig. 6.

Averaged heat flux difference of (a) SHF of daytime, (b) LHF of daytime, (c) GHF of daytime, and (d) GHF of nighttime between BASELINE and SOILMOIS simulations.

Fig. 6.

Averaged heat flux difference of (a) SHF of daytime, (b) LHF of daytime, (c) GHF of daytime, and (d) GHF of nighttime between BASELINE and SOILMOIS simulations.

4) Impact on averaged bias at observing sites

The averaged bias in 2-m tmax, tmin, and Q2 from the SOILMOIS run is plotted in Fig. 7. When compared with Fig. 2a from BASELINE, the cold bias of 2-m tmax is lower in SOILMOIS. In the area that the initial surface soil moisture is reduced in SOILMOIS, the 2-m tmax prediction changes from an evident underprediction to a slight underprediction or even a slight overprediction. The area with a cold bias of 2°C or more in 2-m tmax shrinks. Reduced soil moisture also improves the prediction of 2-m tmin, and there are fewer observing sites with warm 2-m tmin bias higher than 2°C (Fig. 7b) when compared with those in the BASELINE simulation (Fig. 2b). The positive Q2 bias at 0900 UTC in the BASELINE simulation (Fig. 2c) changes to negative bias in most parts of the domain due to the reduced soil moisture shown in Fig. 7c. Clearly, the soil moisture initialization from HRLDAS produces an expected effect on the T2 prediction. Meanwhile, it makes positive Q2 bias smaller in most areas, but also causes a drier bias in some areas, especially in the plain areas.

Fig. 7.

(a) 2-m tmax bias, (b) 2-m tmin bias, and (c) 2-m specific humidity bias at 0900 UTC at observing sites from the SOILMOIS simulation, averaged between 1 Nov 2016 and 28 Feb 2017.

Fig. 7.

(a) 2-m tmax bias, (b) 2-m tmin bias, and (c) 2-m specific humidity bias at 0900 UTC at observing sites from the SOILMOIS simulation, averaged between 1 Nov 2016 and 28 Feb 2017.

5) Evaluation of time series at surface observing sites

In addition to time-averaged bias at the observing sites, the time series of bias is also worth attention. Surface stations 53484 and 54705 (shown in Fig. 1b) are selected due to their soil moisture variability and the prediction errors in the diurnal near-surface temperature amplitude. The initial soil temperature of either BASELINE or SOILMOIS matches well with the observed soil temperature variation with a positive bias at both stations (Fig. 8a). The initial soil moisture of BASELINE, coming from the ECMWF forecast, is much wetter than the observation at both sites. The initial soil moisture from SOILMOIS, coming from HRLDAS, mostly agrees well with the observed value at site 53484, while at site 54705 it is about 0.05 m3 m−3 wetter than the observation (Fig. 8b). It is noted that the observed surface soil moisture drops suddenly on 20 November 2016 at station 53484, and at station 54705 it shows a relatively low value during the second half of January 2017. Coincidentally, these are times when the surface soil temperature falls below 0°C. As most soil moisture devices malfunction when soils are frozen, the soil moisture measurements may be underestimated (Hallikainen et al. 1985).

Fig. 8.

From top to bottom: time series of initial surface soil temperature, initial surface soil moisture, 2-m tmax bias, 2-m tmin bias, and 2-m specific humidity bias at surface stations (left) 53484 and (right) 54705 from BASELINE and SOILMOIS. Stars on the top plot denote the days with snow cover.

Fig. 8.

From top to bottom: time series of initial surface soil temperature, initial surface soil moisture, 2-m tmax bias, 2-m tmin bias, and 2-m specific humidity bias at surface stations (left) 53484 and (right) 54705 from BASELINE and SOILMOIS. Stars on the top plot denote the days with snow cover.

For site 53484, both the 2-m tmax and 2-m tmin bias from SOILMOIS are closer to zero than those from BASELINE on most days during the simulation, and the improvement of 2-m tmax bias is approximately equal to the improvement of 2-m tmin bias (left column, Figs. 8c,d). While for site 54705, the negative 2-m tmax bias and the positive 2-m tmin bias from SOILMOIS are smaller than those from BASELINE. But the positive 2-m tmax bias and the negative 2-m tmin bias from SOILMOIS are slightly bigger than those from BASELINE (right column, Figs. 8c,d). On most days in the simulation, Q2 bias at 0900 UTC for both observing sites from BASELINE is positive, while that from SOILMOIS is negative.

b. Soil map and soil parameter

1) Experiment outline

Analyses in section 3a clearly show that a better soil moisture initialization can actually improve the 2-m diurnal temperature amplitude by increasing the daily maximum temperature and lowering the daily minimum temperature, and can reduce wet 2-m specific humidity bias closer to zero in most areas. However, it makes the dry bias even larger in some areas. To look for possible ways to improve the dry Q2 forecast bias, we turn to soil texture and soil hydraulic parameters used in the LSM, as they represent the physical properties of the soil. In this section, a simulation, named SOILPROPERTY, is performed, which examines the effect of using a different soil map and soil hydraulic parameters on the prediction of surface variables.

2) Updated soil texture and revised soil parameters

The default soil texture dataset used in the BASELINE and SOILMOIS simulations may not accurately reflect the detailed soil features in the areas outside the United States (Gao et al. 2008). In this study, the Chinese 30-arc-s-resolution soil texture dataset developed by BNU (Shangguan et al. 2014) is used to replace the default soil texture dataset for Noah LSM.

Figure 9 shows the default soil texture and the BNU soil texture fields in the 3-km domain. It can be observed that the BNU soil maps show greater spatial variation than the default soil map. Large area in the right lower triangular domain with soil texture of clay loam in the default soil map is replaced by loam in the BNU soil map, which happens to be the area where many stations have a noticeable dry Q2 bias in SOILMOIS (Fig. 7c).

Fig. 9.

Soil texture from (a) default dataset of Noah LSM and (b) BNU dataset. Distribution of α = 1/(Θref − Θw) at initial time in (c) SOILMOIS and (d) SOILPROPERTY.

Fig. 9.

Soil texture from (a) default dataset of Noah LSM and (b) BNU dataset. Distribution of α = 1/(Θref − Θw) at initial time in (c) SOILMOIS and (d) SOILPROPERTY.

The soil hydraulic parameters can affect the state of soil temperature and soil moisture in an LSM. Since these parameters have not been updated for many years for Noah LSM, a more updated soil parameter lookup table may be worth investigating. With the understanding that there is uncertainty in the specification of these parameters, the revised soil parameter lookup table from Kishné et al. (2017) is tested. The default and revised soil parameters of those soil textures are presented in Table 1.

Table 1.

Updated and default (in parentheses) soil hydraulic parameter table of the Noah LSM. BB is the parameter in hydraulic functions related to water potential and volumetric water content. DRYSMC is soil water content threshold for ceasing evaporation from the surface soil layer. F11 is soil thermal diffusivity or conductivity coefficient. MAXSMC is soil water content at saturation or soil porosity. REFSMC is soil water content at field capacity. SATPSI is saturation matric potential or air entry water potential. SATDK is saturated hydraulic conductivity. SATDW is saturated hydraulic diffusivity. WILTSMC is soil water content at the wilting point. QTZ is quartz content.

Updated and default (in parentheses) soil hydraulic parameter table of the Noah LSM. BB is the parameter in hydraulic functions related to water potential and volumetric water content. DRYSMC is soil water content threshold for ceasing evaporation from the surface soil layer. F11 is soil thermal diffusivity or conductivity coefficient. MAXSMC is soil water content at saturation or soil porosity. REFSMC is soil water content at field capacity. SATPSI is saturation matric potential or air entry water potential. SATDK is saturated hydraulic conductivity. SATDW is saturated hydraulic diffusivity. WILTSMC is soil water content at the wilting point. QTZ is quartz content.
Updated and default (in parentheses) soil hydraulic parameter table of the Noah LSM. BB is the parameter in hydraulic functions related to water potential and volumetric water content. DRYSMC is soil water content threshold for ceasing evaporation from the surface soil layer. F11 is soil thermal diffusivity or conductivity coefficient. MAXSMC is soil water content at saturation or soil porosity. REFSMC is soil water content at field capacity. SATPSI is saturation matric potential or air entry water potential. SATDK is saturated hydraulic conductivity. SATDW is saturated hydraulic diffusivity. WILTSMC is soil water content at the wilting point. QTZ is quartz content.

3) Change in moisture availability

As it is in winter and the vegetation cover fraction in the domain is low, the direct evaporation from the ground surface plays a dominant role in the land surface–atmosphere interaction. In Noah LSM, the direct evaporation from the ground surface is computed as

 
Edir=(1σf)βEp,
(2)
 
β=Θ1ΘwΘrefΘw.
(3)

Here, Ep is the potential evaporation, which is based on the Penman energy balance approach, and σf is the greenness vegetation fraction, which is a critical parameter for partitioning between bare soil direct evaporation and canopy transpiration. Parameters Θref and Θw are the field capacity and wilting point, respectively. The field capacity is defined as the soil moisture when the drainage rate through the bottom of the root zone is 0.5 mm day−1 (Hillel 1980), and the wilting point is the threshold of soil moisture at or below which the evaporation will stop. The parameter β, ratio of soil available moisture for evaporation to soil available water content, varies from 0 to 1 and represents the moisture availability. To help with understanding these parameters, we define 1/(Θref − Θw) as α hereafter, which depends only on soil texture and hydraulic parameter lookup table.

Figures 9a and 9b respectively show the α distribution in SOILMOIS with the default soil texture and default soil hydraulic parameter lookup table and in SOILPROPERITY with the BNU soil texture and revised soil hydraulic parameter lookup table.

The parameter α varies moderately in the whole domain in SOILMOIS, with a value of 3–4 in areas with soil types of loam, clay loam, and sand, a value of 4–5 in areas of sand and sandy clay loam, and a value smaller than 3 in areas of sandy loam. This shows that the moisture availability is very similar among different soil types using the default soil hydraulic parameters. Applying a different soil dataset only leads to a subtle variation of α.

The range of α values in the SOILPROPERTY experiment is much larger when compared with that in SOILMOIS. It is clear that the revised soil hydraulic parameter table permits a larger moisture availability range among different soil textures. It also assigns much bigger moisture availability to certain soil textures when compared to the default soil hydraulic parameter table. Taking the soil type loam as an example, the value of α is 4 in SOILMOIS and 7 in SOILPROPERTY. This suggests that the soil map and parameter table applied in the SOILPROPERTY run may favor more direct evaporation from the ground surface.

4) Impact on surface heat fluxes

While the change in heat fluxes are up to 50 J m−2 in the SOILMOIS run by applying HRLDAS soil moisture initialization (Fig. 6), the heat flux differences introduced by applying updated soil map and revised soil parameters in most areas of the domain are between −20 and 20 J m−2 (Fig. 10). For the same initial soil moisture, soil texture modification and hydraulic parameters can only introduce moderate modification of heat fluxes. Most remarkable variations in SHF and GHF occur in the western domain, where soil textures are changed from sand and loam in the default soil map to loamy sand in the BNU soil map. The replacement of sand soil by loamy sand soil changes the surface energy partitioning by favoring more upward SHF and less downward GHF. When loam soil is replaced by loamy sand soil, the situation is then the opposite. Soil texture changes mentioned above do not have much impact on LHF in this area, because the soil moisture is lower than 0.1 m3 m−3. Modification of LHF brought by soil texture modification and hydraulic parameter change occurs mainly in the plain areas, with a magnitude of 10 J m−2, which is much lower than that brought by soil moisture initialization modification. Corresponding to the variation of LHF and GHF, more soil water is being evaporated from the top shallow soil layer to the atmosphere, so that the surface soil moisture becomes a little drier (Fig. 10d), and the surface soil temperature decreases by a moderate range (Fig. 10e).

Fig. 10.

Differences of (a) SHF, (b) LHF, (c) GHF, (d) surface soil moisture, and (e) surface soil temperature at 0600 UTC between SOILMOIS and SOILPROPERTY, averaged between 1 Nov 2016 and 28 Feb 2017.

Fig. 10.

Differences of (a) SHF, (b) LHF, (c) GHF, (d) surface soil moisture, and (e) surface soil temperature at 0600 UTC between SOILMOIS and SOILPROPERTY, averaged between 1 Nov 2016 and 28 Feb 2017.

5) Impact on surface variables

Figure 11 shows how applying updated soil map and hydraulic parameters can impact T2 and Q2 prediction in SOILPROPERTY when compared to SOILMOIS. At 0600 UTC, a representative daytime, it is shown in Fig. 11a that the T2 decreases at a magnitude of about 0.4°C in the plain areas, where a 2-m tmax warm bias exists in SOILMOIS. Meanwhile, the T2 increases at a magnitude of about 0.2°C in the southwestern domain, where a 2-m tmax cold bias exists in SOILMOIS. The T2 decreases (increases) at a magnitude more than 0.4°C in the western domain, where 2-m tmax cold bias exists in SOILMOIS and the soil type changes from sand (loam) in SOILMOIS to loamy sand in SOILPROPERTY. In general, applying an updated soil map and hydraulic parameters could improve the daytime T2 forecast in SOILPROPERTY compared with SOILMOIS. At 2100 UTC, a representative nighttime, it is shown in Fig. 11b that T2 decreases in the plain areas, where 2-m tmin warm bias exists in SOILMOIS. But T2 increases in the southwestern domain, where 2-m tmin warm bias exists in SOILMOIS. For the areas that soil type changes from sand (loam) in SOILMOIS to loamy sand in SOILPROPERTY, the T2 decreases (increases) more than 0.4°C. Q2 at 0900 UTC increases at a magnitude of 0.1 g kg−1 in a large part in the domain, and at a magnitude of over 0.2 g kg−1 in the southern domain and the plain areas, while a Q2 dry bias exists in large area in SOILMOIS. There is little difference for the tmax and tmin bias between SOILPROPERTY (Figs. 11d,e) and SOILMOIS. On the other hand, Q2 bias (Fig. 11f) apparently decreases. Although the Q2 bias is still dry, the Q2 forecast error in SOILPROPERTY is smaller than that in SOILMOIS and BASELINE.

Fig. 11.

Difference of (a) 2-m temperature at 0600 UTC, (b) 2-m temperature at 2100 UTC, and (c) 2-m specific humidity at 0900 UTC between SOILMOIS and SOILPROPERTY. (d) 2-m tmax bias, (e) 2-m tmin bias, and (f) 2-m specific humidity bias at 0900 UTC at observing sites from the SOILPROPERTY simulation, averaged between 1 Nov 2016 and 28 Feb 2017.

Fig. 11.

Difference of (a) 2-m temperature at 0600 UTC, (b) 2-m temperature at 2100 UTC, and (c) 2-m specific humidity at 0900 UTC between SOILMOIS and SOILPROPERTY. (d) 2-m tmax bias, (e) 2-m tmin bias, and (f) 2-m specific humidity bias at 0900 UTC at observing sites from the SOILPROPERTY simulation, averaged between 1 Nov 2016 and 28 Feb 2017.

As SOILPROPERTY is performed based on SOILMOIS, the changes of surface heat fluxes resulting from the soil properties are commingled with those from the different soil moisture initialization, and another simulation based on BASELINE is run to display the changes just from soil property. When updating the soil map and soil parameters by BASELINE, the change pattern for surface heat fluxes is similar to that in SOILPROPERTY, just with a smaller magnitude (figure not shown).

4. Summary

The cold 2-m tmax bias, warm 2-m tmin bias, and wet Q2 bias of the RMAPS-ST forecasts in the winter of 2016/17 are fixed in this study. To simplify the investigation, a BASELINE experiment is carried out. Further analyses show that initial soil moisture might be the main contributing factor. Therefore, SOILMOIS and SOILPROPERTY simulations are performed to examine to what degree soil moisture and soil’s hydraulic properties can impact T2 and Q2 forecast.

The soil moisture from HRLDAS products is utilized in SOILMOIS simulations to provide better soil moisture initialization. The initial soil moisture of the SOILMOIS experiment from HRLDAS product has finer details that match with the resolution of the 3-km grid and is drier than the data from ECMWF in many parts of the domain except for a small area in the northwest. Soil moisture changes can lead to changes in soil thermal conductivity, and hence impact surface energy partition. Drier soil results in lower soil thermal conductivity and therefore less downward GHF, less LHF and more SHF in daytime, and less upward GHF at nighttime. As a result, it leads to various reductions of the cold 2-m tmax bias, warm 2-m tmin bias, and positive Q2 bias in the BASELINE experiment.

The BNU dataset and revised soil parameters from Kishné et al. (2017) are used in the SOILPROPERTY experiment. The soil texture determines the soil’s physical and hydraulic properties, and soil texture change and the application of revised soil parameters lead to a general increase in soil moisture availability and a change in surface energy partition. The utilization of updated soil map and soil hydraulic parameter lookup table leads to a slight reduction of the cold 2-m tmax bias and warm 2-m tmin bias in most of the 3-km domain and a clear reduction of the dry Q2 bias in the SOILMOIS simulation.

Table 2 summarizes the forecast errors in the BASELINE, SOILMOIS, and SOILPROPERTY experiments. Improved soil moisture initialization in SOILMOIS has definitely improved the prediction of diurnal temperature range. It reduces the domain-averaged 2-m tmax bias and root-mean-square error (RMSE) by 83% and 27%, respectively, while the domain averaged 2-m tmin bias and RMSE are lowered by 44% and 30%, respectively. Although the domain averaged bias changes from positive to negative, the domain averaged RMSE for Q2 is improved by 4%. In experiment SOILPROPERTY, the utilization of up-to-date soil map and revised soil hydraulic parameters helps to make more improvement to Q2 by reducing bias and RMSE by 47% and 18%, respectively. In summary, the improved soil moisture initialization, the utilization of up-to-date soil map and revised soil hydraulic parameters have been shown to significantly improve the T2 and Q2 forecast in RMAPS-ST.

Table 2.

Domain-averaged forecasting error of 2-m tmax, 2-m tmin, and 2-m specific humidity in the 3-km resolution domain from the BASELINE, SOILMOIS, and SOILPROPERTY simulations. The improvement range of SOILMOIS and SOILPROPERTY simulations relative to the BASELINE simulation is listed in parentheses.

Domain-averaged forecasting error of 2-m tmax, 2-m tmin, and 2-m specific humidity in the 3-km resolution domain from the BASELINE, SOILMOIS, and SOILPROPERTY simulations. The improvement range of SOILMOIS and SOILPROPERTY simulations relative to the BASELINE simulation is listed in parentheses.
Domain-averaged forecasting error of 2-m tmax, 2-m tmin, and 2-m specific humidity in the 3-km resolution domain from the BASELINE, SOILMOIS, and SOILPROPERTY simulations. The improvement range of SOILMOIS and SOILPROPERTY simulations relative to the BASELINE simulation is listed in parentheses.

However, some forecast errors in the diurnal temperature range and the humidity bias in RMAPS-ST still remain. There are still noticeable warm biases and RMSEs of 2-m tmin. Further investigations are needed. As shown in Fig. 8, the initial soil moisture on site 53484 is approximately equal to the observed soil moisture, but it contains only liquid water content in the soil. It is clear that a soil moisture initialization for frozen soil still presents a challenge. Snow cover days at some observing sites in northern China are nearly half of the winter days. Therefore, further investigation on the impact of snow physics on screen-level temperature and humidity and more accurate representation of snow cover in regional NWP models will be conducted. Moreover, air pollution is a serious problem in some areas in the 3-km domain. Uncertainties in the forecast (especially the diurnal temperature) due to aerosol-induced effect on shortwave/longwave radiation are expected to have some impact. However, aerosol–meteorology feedback is not considered in RAMPS-ST currently and is expected to be included in the future.

Acknowledgments

This work was supported by National Key Research and Development Program of China (2018YFC1506804), National Natural Science Foundation of China (41705087 and 41705076), and Beijing Natural Science Foundation (8184073).

REFERENCES

REFERENCES
Chen
,
F.
, and
J.
Dudhia
,
2001
:
Coupling an advanced land surface-hydrology model with Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity
.
Mon. Wea. Rev.
,
129
,
569
585
, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Chen
,
F.
, and et al
,
2007
:
Description and evaluation of the characteristics of the NCAR high-resolution land data assimilation system
.
J. Appl. Meteor. Climatol.
,
46
,
694
713
, https://doi.org/10.1175/JAM2463.1.
Cheng
,
W. Y. Y.
, and
W. J.
Steenburgh
,
2005
:
Evaluation of surface sensible weather forecasts by the WRF and the Eta models over the western United States
.
Wea. Forecasting
,
20
,
812
821
, https://doi.org/10.1175/WAF885.1.
Davis
,
C.
,
T.
Warner
,
E.
Astling
, and
J.
Bowers
,
1999
:
Development and application of an operational, relocatable, mesogamma-scale weather analysis and forecasting system
.
Tellus
,
51A
,
710
727
, https://doi.org/10.3402/tellusa.v51i5.14490.
Di Giuseppe
,
F.
,
D.
Cesari
, and
G.
Bonafé
,
2011
:
Soil initialization strategy for use in limited-area weather prediction systems
.
Mon. Wea. Rev.
,
139
,
1844
1860
, https://doi.org/10.1175/2011MWR3279.1.
Dy
,
C. Y.
, and
J. C.-H.
Fung
,
2016
:
Updated global soil map for the weather research and forecasting model and soil moisture initialization for the Noah land surface model
.
J. Geophys. Res. Atmos.
,
121
,
8777
8800
, https://doi.org/10.1002/2015JD024558.
FAO
,
1991
:
The Digitized Soil Map of the World (release 1.0). Food and Agriculture Organization, Rome, Italy
.
FAO and UNESCO
,
1974
:
Legend. Vol. 1, Soil Map of the World (1:5,000,00), UNESCO, 59 pp.
, http://www.fao.org/3/as360e/as360e.pdf.
FAO and UNESCO
,
1981
:
Soil Map of the World (1:5,000,000). Vol. 1–10, UNESCO
, http://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/faounesco-soil-map-of-the-world/en/.
Gao
,
Y.
, and et al
,
2008
:
Enhancement of land surface information and its impact on atmospheric modeling in the Heihe River Basin, northwest China
.
J. Geophys. Res.
,
113
,
D20S90
, https://doi.org/10.1029/2008JD010359.
Hallikainen
,
M. T.
,
F. T.
Ulaby
,
M. C.
Dobson
,
M. A.
El-Rayes
, and
L.-K.
Wu
,
1985
:
Microwave dielectric behavior of wet soil—Part I: Empirical models and experimental observations
.
IEEE Trans. Geosci. Remote Sens.
,
23
,
25
34
, https://doi.org/10.1109/TGRS.1985.289497.
Hanna
,
S. R.
, and
R.
Yang
,
2001
:
Evaluations of mesoscale models’ simulations of near-surface winds, temperature gradients, and mixing depths
.
J. Appl. Meteor.
,
40
,
1095
1104
, https://doi.org/10.1175/1520-0450(2001)040<1095:EOMMSO>2.0.CO;2.
Hillel
,
D.
,
1980
:
Applications of Soil Physics. Academic Press, 385 pp
.
Hong
,
S.-Y.
,
Y.
Noh
, and
J.
Dudhia
,
2006
:
A new vertical diffusion package with explicit treatment of entrainment processes
.
Mon. Wea. Rev.
,
134
,
2318
2341
, https://doi.org/10.1175/MWR3199.1.
Huang
,
J.
,
H. M.
van den Dool
, and
K. P.
Georgarakos
,
1996
:
Analysis of model-calculated soil moisture over the United States (1931–1993) and applications to long-range temperature forecasts
.
J. Climate
,
9
,
1350
1362
, https://doi.org/10.1175/1520-0442(1996)009<1350:AOMCSM>2.0.CO;2.
Iacono
,
M. J.
,
J. S.
Delamere
,
E. J.
Mlawer
,
M. W.
Shephard
,
S. A.
Clough
, and
W. D.
Collins
,
2008
:
Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models
.
J. Geophys. Res.
,
113
,
D13103
, https://doi.org/10.1029/2008JD009944.
Kain
,
J. S.
,
2004
:
The Kain–Fritsch convective parameterization: An update
.
J. Appl. Meteor.
,
43
,
170
181
, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Kishné
,
A. S.
,
Y. T.
Yimam
,
C. L. S.
Morgan
, and
B. C.
Dornblaser
,
2017
:
Evaluation and improvement of the default soil hydraulic parameters for the Noah land surface model
.
Geoderma
,
285
,
247
259
, https://doi.org/10.1016/j.geoderma.2016.09.022.
Lin
,
T.-S.
, and
F.-Y.
Cheng
,
2016
:
Impact of soil moisture initialization and soil texture on simulated land–atmosphere interaction in Taiwan
.
J. Hydrometeor.
,
17
,
1137
1355
, https://doi.org/10.1175/JHM-D-15-0024.1.
Marshall
,
C. H.
,
K. C.
Crawford
,
K. E.
Mitchell
, and
D. J.
Stensrud
,
2003
:
The impact of the land surface physics in the operational NCEP Eta Model on simulation the diurnal cycle: Evaluation and testing using Oklahoma Mesonet data
.
Wea. Forecasting
,
18
,
748
768
, https://doi.org/10.1175/1520-0434(2003)018<0748:TIOTLS>2.0.CO;2.
Mass
,
C. F.
,
D.
Ovens
,
K.
Westrick
, and
B. A.
Colle
,
2002
:
Does increasing horizontal resolution produce more skillful forecasts? The result of two years of real-time numerical weather prediction over the Pacific Northwest
.
Bull. Amer. Meteor. Soc.
,
83
,
407
430
, https://doi.org/10.1175/1520-0477(2002)083<0407:DIHRPM>2.3.CO;2.
Massey
,
J. D.
,
W. J.
Steenbrugh
,
S. W.
Hoch
, and
J. C.
Knievel
,
2014
:
Sensitivity of near-surface temperature forecasts to soil properties over a sparsely vegetated dryland region
.
J. Appl. Meteor. Climatol.
,
53
,
1976
1995
, https://doi.org/10.1175/JAMC-D-13-0362.1.
Miller
,
D. A.
, and
R. A.
White
,
1998
:
A conterminous United States multilayer soil characteristics dataset for regional climate and hydrology modeling
.
Earth Interact.
,
2
, https://doi.org/10.1175/1087-3562(1998)002<0001:ACUSMS>2.3.CO;2.
Reeves
,
H. D.
,
K. L.
Elmore
,
G. S.
Manikin
, and
D. J.
Stensrud
,
2011
:
Assessment of forecasts during persistent valley cold pools in the Bonneville basin by the North American Mesoscale Model
.
Wea. Forecasting
,
26
,
447
467
, https://doi.org/10.1175/WAF-D-10-05014.1.
Rosero
,
E.
,
Z.-L.
Yang
,
T.
Wagener
,
L. E.
Gulden
,
S.
Yatheendradas
, and
G.-Y.
Niu
,
2010
:
Quantifying parameter sensitivity, interaction, and transferability in hydrologically enhanced versions of the Noah land surface model over transition zones during the warm season
.
J. Geophys. Res.
,
115
,
D03106
, https://doi.org/10.1029/2009JD012035.
Shangguan
,
W.
, and et al
,
2013
:
A China data set of soil properties for land surface modeling
.
J. Adv. Model. Earth Syst.
,
5
,
212
224
, https://doi.org/10.1002/jame.20026.
Shangguan
,
W.
,
Y.
Dai
,
Q.
Duan
,
B.
Liu
, and
H.
Yuan
,
2014
:
A global soil data set for earth system modeling
.
J. Adv. Model. Earth Syst.
,
6
,
249
263
, https://doi.org/10.1002/2013MS000293.
Sun
,
S. B.
,
B.
Chen
,
Q.
Shao
,
J.
Chen
,
J.
Liu
,
X.-j.
Zhang
,
H.
Zhang
, and
X.
Lin
,
2017
:
Modeling evapotranspiration over China’s landmass from 1979 to 2012 using multiple land surface models: Evaluations and analyses
.
J. Hydrometeor.
,
18
,
1185
1203
, https://doi.org/10.1175/JHM-D-16-0212.1.
Thomas
,
M.
,
2008
:
On the role of snow cover in depressing air temperature
.
J. Appl. Meteor. Climatol.
,
47
,
2008
2022
, https://doi.org/10.1175/2007JAMC1823.1.
Tomasi
,
E.
,
L.
Giovannini
,
D.
Zardi
,
M.
de Franceschi
,
2017
:
Optimization of Noah and Noah_MP WRF land surface schemes in snow-melting conditions over complex terrain
.
Mon. Wea. Rev.
,
145
,
4727
4745
, https://doi.org/10.1175/MWR-D-16-0408.1.
Trier
,
S. B.
,
F.
Chen
,
K. W.
Manning
,
M. A.
LeMone
, and
C. A.
Davis
,
2008
:
Sensitivity of the PBL and precipitation in 12-day simulations of warm-season convection using different land surface models and soil wetness conditions
.
Mon. Wea. Rev.
,
136
,
2321
2343
, https://doi.org/10.1175/2007MWR2289.1.
Warrach-Sagi
,
K.
,
V.
Wulfmeyer
,
R.
Grasselt
,
F.
Ament
, and
C.
Simmer
,
2008
:
Streamflow simulations reveal the impact of the soil parameterization
.
Meteor. Z.
,
17
,
751
762
, https://doi.org/10.1127/0941-2948/2008/0343.
Xia
,
Y.
,
T.
Ford
,
Y.
Wu
,
S. M.
Quiring
, and
M. B.
Ek
,
2015
:
Automated quality control of in situ soil moisture from the north American soil moisture database using NLDAS-2 products
.
J. Appl. Meteor. Climatol.
,
54
,
1267
1282
, https://doi.org/10.1175/JAMC-D-14-0275.1.
Zhang
,
C.
,
Y.
Wang
,
A.
Lauer
, and
K.
Hamilton
,
2012
:
Configuration and evaluation of the WRF Model for the study of Hawaiian regional climate
.
Mon. Wea. Rev.
,
140
,
3259
3277
, https://doi.org/10.1175/MWR-D-11-00260.1.
Zhang
,
X.
,
S. F.
Sun
, and
Y.
Xue
,
2007
:
Development and testing of a frozen soil parameterization for cold region studies
.
J. Hydrometeor.
,
8
,
690
701
, https://doi.org/10.1175/JHM605.1.
Zheng
,
D. H.
,
R.
van der Velde
,
Z.
Su
,
X.
Wang
,
J.
Wen
,
M. J.
Booij
,
A. Y.
Hoekstra
, and
Y.
Chen
,
2015
:
Augmentations to the Noah model physics for application to the Yellow River source area. Part I: Soil water flow
.
J. Hydrometeor.
,
16
,
2659
2676
, https://doi.org/10.1175/JHM-D-14-0198.1.
For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).