Updated Global Soil Map for the WRF Model and Soil Moisture Initialization for the Noah LSM: SOIL MAP AND SOIL MOISTURE

Abstract

A meteorological model requires accurate initial conditions and boundary conditions to obtain realistic numerical weather predictions. The land surface controls the surface heat and moisture exchanges, which can be determined by the physical properties of the soil and soil state variables, subsequently exerting an effect on the boundary layer meteorology. The initial and boundary conditions of soil moisture are currently obtained via NCEP FNL (Final) Operational Global Analysis data, which are collected operationally in 1 degree by 1 degree resolutions every six hours. Another input to the model is the soil map generated by the FAO-UNESCO soil database, which combines several soil surveys from around the world. Both soil moisture from the FNL analysis data and the default soil map lack accuracy and feature coarse resolutions, particularly for certain areas of China. In this study, we update the global soil map with data from Beijing Normal University in 1 km by 1 km grids and propose an alternative method of soil moisture initialization. Simulations of the Weather Research and Forecasting Model show that spinning-up the soil moisture improves near-surface temperature and relative humidity prediction using different types of soil moisture initialization. Explanations of that improvement and improvement of the planetary boundary layer height in performing process analysis are provided.

Full text

Journal of Geophysical Research: Atmospheres
Updated global soil map for the Weather Research
and Forecasting model and soil moisture
initialization for the Noah land
surface model
C. Y. DY1and J. C. H. Fung1,2
1Department of Mathematics, Hong Kong University of Science and Technology, Hong Kong, 2Division of Environment,
Hong Kong University of Science and Technology, Hong Kong
Abstract A meteorological model requires accurate initial conditions and boundary conditions to obtain
realistic numerical weather predictions. The land surface controls the surface heat and moisture exchanges,
which can be determined by the physical properties of the soil and soil state variables, subsequently
exerting an effect on the boundary layer meteorology. The initial and boundary conditions of soil moisture
are currently obtained via National Centers for Environmental Prediction FNL (Final) Operational Global
Analysis data, which are collected operationally in 1∘by 1∘resolutions every 6 h. Another input to the model
is the soil map generated by the Food and Agriculture Organization of the United Nations - United Nations
Educational, Scientific and Cultural Organization (FAO-UNESCO) soil database, which combines several soil
surveys from around the world. Both soil moisture from the FNL analysis data and the default soil map lack
accuracy and feature coarse resolutions, particularly for certain areas of China. In this study, we update the
global soil map with data from Beijing Normal University in 1 km by 1 km grids and propose an alternative
method of soil moisture initialization. Simulations of the Weather Research and Forecasting model show
that spinning-up the soil moisture improves near-surface temperature and relative humidity prediction
using different types of soil moisture initialization. Explanations of that improvement and improvement of
the planetary boundary layer height in performing process analysis are provided.
1. Introduction
In meteorological models, land surface processes and planetary boundary layer (PBL) schemes are crucial
to simulating accurate numerical weather patterns, particularly within the boundary layer. Weather systems
especially conditions in the boundary layer are strongly influenced by land surface states and surface energy
fluxes. The state-of-the-art Weather Research and Forecasting (WRF) model [Skamarock et al., 2008] is a mete-
orological model that combines various physics schemes from the literature and computing technology to
provide a tool for simulating atmospheric conditions and enhancing meteorological studies. Land surface
models (LSMs) obtain radiation and cloud cover states from other physics schemes to simulate soil states
and surface fluxes, which are subsequently used by PBL schemes to predict the vertical profiles of several
atmospheric variables. The Noah LSM [Chen et al., 1996, 1997; Chen and Dudhia, 2001; Ek et al., 2003] in the
WRF model simulates skin temperature using an energy balance model at the surface layer [Mahrt and Ek,
1984] together with the classic diffusion equation used to simulate the soil temperature. The volumetric soil
moisture and intercepted canopy water in the Noah LSM follow the Richards equation for soil water move-
ment proposed in Richards [1931], Noilhan and Planton [1989], and Chen et al. [1996], with a determination of
evaporation made using an electric analogy [Jarvis, 1976; Noilhan and Planton, 1989; Jacquemin and Noilhan,
1990] and soil moisture runoffs [Schaake et al., 1996]. The physical properties of the soil affect the soil tem-
perature and volumetric soil moisture content. In the WRF model, the soil is classified into 16 categories
[Davis and Bennett, 1927; Soil Survey Division Staff , 1993], and each soil type is quantified with soil ther-
modynamic and hydraulic properties, namely, wilting point, field capacity, saturated diffusion coefficient,
heat capacity, and many other soil-related physical properties. The current WRF soil map is generated
from the State Soil Geographic (STATSGO)/Food and Agriculture Organization (FAO) soil database [Food and
Agriculture Organization of the United Nations - United Nations Educational, Scientific and Cultural Organiza-
tion (FAO-UNESCO), 1974, 1971– 1981; FAO , 1991], which has 30 arc-sec resolution in the U.S. and 5 arc min
RESEARCH ARTICLE
10.1002/2015JD024558
Key Points:
• Updated the WRF soil map with BNU
soil data set to show the long-term
effects on soil moisture
• A long spinning-up period which
depends on latitude of locations is
required for soil moisture
• Updated soil map and spun-up
soil moisture improves the 2 m
temperature by 1 degree in most
places
Correspondence to:
J. C. H. Fung,
majfung@ust.hk
Citation:
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,
doi:10.1002/2015JD024558.
Received 1 DEC 2015
Accepted 7 JUL 2016
Accepted article online 13 JUL 2016
Published online 3 AUG 2016
©2016. American Geophysical Union.
All Rights Reserved.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8777

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
resolution elsewhere. Updating the soil map with a higher resolution is necessary to obtain a more accu-
rate representation of the physical properties of the land surface. Advances in soil research allowed Beijing
Normal University (BNU) to obtain a China soil data set [Shangguan et al., 2013], and the subsequent global
soil data set was released in 2014 [Shangguan et al., 2014]. The BNU China soil data combine several databases
and maps of China and actual measurements from the Chinese authorities covering the whole of China. The
BNU global soil data set is used to update the WRF default soil map in this study . The differences between the
default and updated soil map are analyzed, and their influence on soil moisture investigated. Another impor-
tant input to the WRF model is the initial soil moisture content. The value of that content is related to the
surface moisture transfer and heat transfer in land surface processes, which have a strong effect on the surface
temperature in the atmosphere and on PBL height. Currently, very few measurements of soil moisture values
are available, and initial conditions are typically obtained from analysis of the global circulation model, which
has a coarse resolution (1∘by 1∘). Therefore, we implement the WRF model for 5 year simulations to spin-up
the soil moisture field itself to obtain a more accurate soil moisture field for model initialization. A similar tech-
nique is illustrated by Lim et al. [2012], who suggest that the spinning-up period should be around 2 years.
For model evaluation, the spinning-up soil moisture technique and updated soil map are used in the WRF
model for simulations of January and July 2011. The goal of this paper is to provide the information on how
the soil type in WRF is updated, show that the soil type has impacts on the soil moisture at around 3 year time
scale, and analyze how the2mtemperature and relative humidity prediction benefited from the updated soil
map and soil moisture spinning-up technique. The remainder of the paper is organized as follows. Section 2
reviews the Noah LSM and land surface characteristics. The updated soil map for the WRF model and an alter-
native soil moisture initialization technique for the Noah LSM are presented in sections 3 and 4, respectively.
Details of the numerical experiments and the results are given in sections 5 and 6, respectively, and section 7
concludes the paper.
2. Noah Land Surface Model and Land Surface Characteristics
The Noah LSM [Chen et al., 1996, 1997; Chen and Dudhia, 2001; Ek et al., 2003] uses radiation and cumulus
information to model skin temperature, which is used as the boundary condition for modeling the soil tem-
perature. Soil temperature and soil moisture are dependent on one another [Mahrt and Ek, 1984; Noilhan and
Planton, 1989; Jacquemin and Noilhan, 1990; Chen et al., 1996] and hence affect the surface fluxes [Ek and
Mahrt, 1991; Garratt, 1993]. The soil state variables are soil temperature and volumetric soil moisture in the
soil layers and the intercepted canopy water content. The calculated surface fluxes include the sensible heat
flux, latent heat flux, and moisture flux at the surface. Surface fluxes are used by PBL schemes [Ek and Mahrt,
1991; Pleim and Xiu, 1995; Xie et al., 2012] to predict the vertical profiles and boundary layer atmospheric con-
ditions. The physical properties of the land surface quantify how much soil moisture can be held (saturated
volumetric soil moisture content), how fast thermal energy is transferred (saturation thermal conductivity),
and the fraction of solar energy that is absorbed or reflected (albedo). The soil is classified into 16 categories
based on the soil composition and the proportions of sand, silt, and clay (Table 1) [Davis and Bennett, 1927; Soil
Survey Division Staff , 1993]. For each soil category, the corresponding bulk physical properties are obtained
from Cosbyetal.[1984]. In the following subsections, we describe our formulation of the soil state variables
and their relation to land surface properties. Note that the formulation involving ice/snow is omitted because
it is beyond the scope of this paper.
2.1. Soil Temperature
The soil temperature Tsoil is modeled by the diffusion equation, namely,
cp,sl
𝜕Tsoil
𝜕t=𝜕
𝜕zKt
𝜕Tsoil
𝜕z,(1)
where cp,sl is the specific heat capacity under constant pressure of the soil layer and Ktis thermal conductivity.
The diffusion equation characterizes the energy transferred within the soil layer as a diffusion process from
the top soil layer to the deep soil layer. The quantity Kt∕cp,sl describes the speed of the diffusion process within
the soil layers. The equation requires the boundary condition at the air-soil interface, and the soil temperature
at the surface is equal to the skin temperature Tskin , which can be obtained from the surface energy balance
equation [Mahrt and Ek, 1984]:
(1−𝛼)S+L−𝜎T4
skin =G+SH +LH,(2)
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8778

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Tab le 1. Soil Composition and Hydraulic Properties Tablea
Type Soil Type Name % Sand % Silt % Clay Wilting Point Saturation Soil Moisture
1 Sand 92 5 3 0.010 0.339
2 Loamy sand 82 12 6 0.028 0.421
3 Sandy loam 58 32 10 0.047 0.434
4 Silt loam 17 70 13 0.084 0.476
5 Silt 10 85 5 0.084 0.476
6 Loam 43 39 18 0.066 0.439
7 Sandy clay loam 58 15 27 0.067 0.404
8 Silty clay loam 10 56 34 0.120 0.464
9 Clay loam 32 34 34 0.103 0.465
10 Sandy clay 52 6 42 0.100 0.406
11 Silty clay 6 47 47 0.126 0.468
12 Clay 22 20 58 0.138 0.468
13 Organic material 0 0 0 0.066 0.439
14 Water 0 0 0 0.000 1.000
15 Bedrock 0 0 0 0.006 0.200
16 Other 0 0 0 0.028 0.421
aThis table contains the reference composition of sand, silt, and clay for each soil type [Davis and Bennett, 1927;
Soil Survey Division Staff , 1993] and the volumetric soil moisture values at the wilting point and at saturation [Cosby
et al., 1984].
where 𝛼is the albedo, Sand Lare the downward shortwave and longwave radiation, respectively, 𝜎is the
Stefan-Boltzmann constant, Gis the ground heat flux, SH is the sensible heat flux, and LH is the latent heat
flux. All of these fluxes are evaluated at the air-soil interface.
2.2. Thermal Conductivity
Both the thermal conductivity and specific heat capacity under constant pressure of the soil layer in
equation (1) depend on the soil moisture and soil type, which control the speed of diffusion in the soil layers.
The formulation of thermal conductivity (Kt)is a linear interpolation between thermal conductivity under
dry condition (Kdry )and saturated condition (Ksat), with a weight of Kersten Number (KN). Kersten Number
(KN) depends on the degree of saturation [Peters-Lidard et al., 1998]. The formula for thermal conductivity
[Peters-Lidard et al., 1998] is given by
Kt=Kdry(1−KN)+Ksat (KN).(3)
In the dry condition, the empirical formula [Peters-Lidard et al., 1998] is
Kdry =0.135 ×2700(1−wsat)+64.7
2700 −0.947 ×2700(1−wsat),(4)
where wsat is the saturated soil moisture of the given soil type. In the saturated condition, thermal conductivity
is interpolated exponentially by the thermal conductivities of quartz (7.7), a nonquartz solid (2), and soil water
(0.57). The formula [Peters-Lidard et al., 1998] is
Ksat =7.7QZ ×2(1−QZ)(1−wsat )0.57wsat,(5)
where QZ is the quartz content percentage of the soil type and wsat is its saturated soil moisture.
2.3. Specific Heat Capacity
The specific heat capacity under constant pressure of the soil layer cp,sl is additive and is calculated by the
linear interpolation between the specific heat capacity of water (cwater =4.2×106J/m3/K), soil (for a rural
area, csoil =2×106J∕m3∕K; for an urban area, csoil =3×106J/m3/K), and air (cair =1004 J/m3/K). The specific
heat capacity under constant pressure is given by
cp,sl =w×cwater +(1−wmax)×csoil +(wmax −w)×cair .(6)
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8779

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 1. Domains 1 to 4. Computational domains of our simulations covering the Asia-Pacific region. Domain 1 covers China and neighboring countries.
Domain 2 focuses on southern China. Domain 3 covers Guangdong Province. Domain 4 covers the Pearl River Delta and Hong Kong. The grid sizes for domains
1to4are27km,9km,3km,and1kmrespectively.
2.4. Sensible Heat Flux and Latent Heat Flux
The sensible heat flux SH and latent heat flux LH at the surface [Garratt, 1993] is modeled by
SH =𝜌acp,a ChVa(Tg−Ta),(7)
LH =𝜌aCLHVa(qg−qa),(8)
where 𝜌ais the density of the air, cp,a is the volumetric heat capacity of the air under constant pressure, Ch
and CLH are the surface exchange coefficient of sensible heat and latent heat accordingly, Vais the air wind
speed immediate adjacent to the ground, and (Tg−Ta)and (qg−qa)are the temperature difference and vapor
mixing ratio difference between the ground and the air immediate adjacent to the ground, respectively.
In the daytime, shortwave radiation from the Sun that is received in part by the ground raises the soil tempera-
ture, and the ground emits longwave radiation to warm up the surface layer of the air. Heat is then transferred
upward from the lower atmosphere to the upper atmosphere by diffusion. Soil type and soil moisture control
the thermal conductivity and heat capacity of the soil related to soil temperature, soil moisture also controls
the evaporation rate and the latent heat flux, and thus both affect the surface sensible heat flux, near-surface
temperature, and relative humidity.
3. Updated Soil Map for WRF Model
The physical properties of the soil in different locations differ. The soil is categorized with bulk physical proper-
ties assigned to each soil classification. There are two major land surface classifications: land use classification
and soil classification.
3.1. Domain
Figure 1 is a plot of the computational domains. The domain setting for the simulations in this study is based
on the model used by the Hong Kong Environmental Protection Department. There are four domains in our
simulations. Domain 1 covers almost the whole of China and the South China Sea. Domain 2 covers Taiwan,
Hainan, and Guangdong Province. Domain 3 covers part of Guangdong Province, and Domain 4 just the Pearl
River Delta (PRD) region and Hong Kong. The grid spacing in the four domains is 27 km, 9 km, 3 km, and
1 km, respectively, and the static data used for them have resolutions of 10 arc min, 5 arc min, 2 arc min, and
30 arc sec. The projection method is Lambert projection, which is suitable for midlatitude simulations.
3.2. Land Use Categories
The vegetation parameters that are dependent on land use include the albedo maximums and minimums,
emissivity, roughness length, leaf area index, rooting depth, minimum stomatal resistance, and radiation
stress function parameters. Land uses are categorized into 24 classes, including urban and built-up land,
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8780

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 2. Default land use categories and updated land use categories. (a and b) The default land use categories and updated land use categories for domain 3
[Lametal., 2006]. (c and d) The default land use categories and updated land use categories for domain 4. The numbers in the legend refer to the USDA land use
categories. Type 1, for example, indicates an urban location.
several types of cropland and pasture land, mosaic, grassland and shrubland, forest, wetland, tundra, and
water. One of the land use data set used in the WRF community is the United States Geological Survey (USGS)
global land use classification, which was created using Advanced Very High Resolution Radiometer scanning
for April 1992 to March 1993 by the USGS. Given the rapid economic development and urbanization seen in
China in the past two decades, particularly in the PRD region, Lam et al. [2006] have updated the land use cat-
egories for the PRD region. Figure 2 shows the land use data set from the National Oceanic and Atmospheric
Administration and the updated land use data set from Lam et al. [2006] for the 2 arc min and 30 arc sec
resolutions covering part of Guangdong Province and the PRD region.
3.3. Soil Categories
The soil classification method used in this study is from United States Department of Agriculture (USDA)
16-class soil classification system [Davis and Bennett, 1927; Soil Survey Division Staff , 1993]. There are 12 soil
types (Figure 3) and four nonsoil types (Table 3), including organic material, water, bedrock, and others. Soil
type is classified based on the percentage of sand, silt, and clay in the soil. For each soil type, the WRF model
has a default soil parameter table that generalizes the hydraulic and thermal properties of the soil. The soil
parameters used in the Noah LSM include the saturated soil matric potential, saturated soil conductivity,
saturated soil diffusivities, wilting point soil moisture, soil quartz content, and saturated soil moisture content.
The hydraulic and thermal properties of the soil depend on the composition of the soil characteristics. The
default WRF static soil data utilize two soil data sets, the STATSGO soil data set from the USDA [Miller and White,
1998], which has 30 s resolution within the conterminous U.S., and the FAO soil data set from the UN Food
and Agriculture Organization (FAO), which was published in 1991 with 5 min resolution over the entire globe
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8781

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 3. Soiltexturetriangle[Soil Survey Division Staff , 1993]. The soil texture triangle indicates the soil classification
based on the percentage of sand, silt, and clay in the soil. This classification method was first introduced in Davis and
Bennett [1927] and was further refined in Soil Survey Division Staff [1993] and named USDA soil classification.
[FAO-UNESCO, 1974, 1971– 1981; FAO, 1991]. The accuracy of these soil data sets is questionable. The reference
soil composition and soil hydraulic parameters [Cosbyetal., 1984] for each soil type are listed in Table 1.
3.4. Updated Soil Map
The BNU China soil data are derived by Shangguan et al. [2013] based on a soil map at a scale of 1:1 million for
China obtained from the Second National Soil Survey in China from year 1979 to year 1985, and all soil books
at national and provincial levels together with 8979 distinct soil profiles measured. Shangguan et al. [2014]
further used several soil databases spread over the globe to create a global soil data set. We have updated the
hybrid STATSGO/FAO data set with the BNU global soil data set for use in the WRF model. The BNU global soil
data set has a horizontal resolution of 30 arc sec. Its data are in the form of matrices, with the percentage of
sand, silt, and clay and latitude and longitude information in NetCDF form. The depths of the eight soil layers
considered are 0– 0.045 m, 0.045–0.091 m, 0.091 –0.166 m, 0.166– 0.289 m, 0.289– 0.493 m, 0.493–0.829 m,
0.829–1.383 m, and 1.383 –2.296 m. The WRF default soil data are in two layers: 0 to 30 cm for the top soil
layer and 30 to 100 cm for the bottom soil layer. As of the WRF version 3.7.1, the input format of soil type is
fixed to two layers and only the bottom soil map is utilized by the WRF model in a bulk sense.
The new top layer uses data on the first five layers, whereas the newbottom layer uses data from layers 5 to 7.
The percentages of sand, silt, and clay are calculated by weighted averaging, with soil layer thickness used
as the weight, and the soil is then classified into soil categories in accordance with the USDA 16-class soil
classification system. Figure 3 shows the soil texture triangle [Soil Survey Division Staff, 1993]. A point in the
triangle indicates the percentage of sand, silt, and clay, which is categorized in accordance with the triangle.
The default soil map is used as the base, and a grid is updated only if the BNU global soil data set has data for
that grid. The water mask from the land use data is then placed onto the intermediate updated soil map, as
some of the water grids are classified as being nonwater soil types in the default soil classification.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8782

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 4. Global soil map comparisons. The soil data set in the WRF model is in the form of two soil layers, a top layer and a bottom layer, each with resolutions
of 30 arc sec and 2, 5, and 10 arc min. (a and c) The global soil map using default WRF soil data obtained from the hybrid STATSGO/FAO data set for top layer and
bottom layer soil, respectively, at a resolution of 30 arc sec. (b and d) The updated global soil map using data from Shangguan et al. [2013] and Shangguan et al.
[2014] for the top layer and bottom layer soil, respectively, at a resolution of 30 arc sec.
Tab le 2. Global Soil Type Distribution Tablea
Top Layer Soil Bottom Layer Soil
Type Soil Type Name Original Soil Type Updated Soil Type Original Soil Type Updated Soil Type
1 Sand 2.7% 3.3% 2.7% 3.5%
2 Loamy sand 2.5% 6.2% 2.3% 4.3%
3 Sandy loam 18.0% 18.3% 14.5% 15.2%
4 Silt loam 6.0% 7.9% 4.9% 5.0%
5 Silt 0.0% 0.0% 0.0% 0.0%
6 Loam 40.2% 35.4% 26.8% 28.7%
7 Sandy clay loam 7.3% 7.3% 7.4% 8.0%
8 Silty clay loam 0.3% 0.7% 0.6% 1.1%
9 Clay loam 11.5% 9.5% 22.6% 17.2%
10 Sandy clay 0.1% 0.9% 1.0% 1.2%
11 Silty clay 0.1% 0.5% 0.5% 0.9%
12 Clay 6.2% 5.5% 11.6% 10.3%
13 Organic material 0.2% 0.0% 0.1% 0.1%
14 Water 0.0% 0.0% 0.0% 0.0%
15 Bedrock 0.2% 0.1% 0.3% 0.2%
16 Other 4.6% 4.3% 4.7% 4.4%
aThis table lists the percentages of each soil type over the globe for both the top layer (left side of the table) and
bottom layer (right side of the table) before and after the soil map update. The dominant soil type across the different
layers and versions is loam (type 6). The second most dominant type of top layer soil is sandy loam (type 3), whereas that
of bottom layer soil is clay loam (type 9). The percentages of the dominant soil type(s) are shown in boldface.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8783

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 5. China soil map comparisons. These are the same as in Figure 4 but for the Asia-Pacific region.
Figures 4a– 4d show the default global soil map and updated global soil map for the top and bottom soil
layers, respectively. Soil type updates for regions in China, Australia, South America, and Southern Africa can
clearly be seen for both the top and bottom layers. The global soil type distribution corresponding toFigure 4
(excluding water bodies) is listed in Table 2 for both layers. Loam is the dominant soil type over the Earth’s land
surface. Sandy loam and clay loam rank the second and the third for the top soil layer, whereas the positions
are reversed for the bottom soil layer. Figure 5 is the same as Figure 4 but is confined to the regions of com-
putational domain 1. It can be observed that the updated soil maps (Figures 5b and 5d) have greater spatial
variation than the default soil map (Figures 5a and 5c), particularly in southern China. The soil type distribu-
tion in domain 1 is summarized in Table 3. The dominant soil type for the top soil layer is loam both before
and after the update. That for the bottom soil layer, however, changes from clay loam to loam in updated
domain 1. Table 4 summarizes all of the major changes in domain 1 pointwise from the default soil map to the
updated soil map. For the top layer, 20.6% of the updated grids change from clay loam to loam. With reference
to Table 1, there is a decrease in saturated soil moisture from 0.465 to 0.439 and a decrease in the wilting point
from 0.103 to 0.066, meaning that there is also a decrease in both the maximum and minimum volumetric
soil moisture. In other words, the soil is dryer in the updated soil map in these grids.
4. Soil Moisture Initialization
Volumetric soil moisture is one of the model fields needed to drive the LSM. The soil moisture field is avail-
able in analysis data (e.g., final [FNL] analysis data) and is typically used for soil moisture initialization. The
soil moisture field in the FNL analysis data is obtained from the analysis step of the Global Forecast System
(GFS) model of the National Centers for Environmental Prediction (NCEP) [National Centers for Environmental
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8784

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Tab le 3. Domain 1 Soil Type Distribution Tablea
Top Layer Soil Bottom Layer Soil
Type Soil Type Name Original Soil Type Updated Soil Type Original Soil Type Updated Soil Type
1 Sand 2.9% 0.7% 2.9% 0.7%
2 Loamy sand 0.1% 5.7% 0.8% 4.7%
3 Sandy loam 5.3% 7.3% 0.7% 6.3%
4 Silt loam 1.5% 8.5% 0.6% 4.5%
5 Silt 0.0% 0.0% 0.0% 0.0%
6 Loam 49.7% 52.7% 21.8% 41.9%
7 Sandy clay loam 9.6% 6.4% 7.6% 5.2%
8 Silty clay loam 0.0% 0.2% 0.5% 1.3%
9 Clay loam 23.5% 10.9% 48.1% 25.0%
10 Sandy clay 0.0% 0.0% 1.1% 0.2%
11 Silty clay 0.1% 2.9% 0.2% 2.2%
12 Clay 5.8% 4.4% 14.1% 7.6%
13 Organic material 0.0% 0.0% 0.0% 0.0%
14 Water 0.0% 0.0% 0.0% 0.0%
15 Bedrock 0.0% 0.0% 0.0% 0.0%
16 Other 1.6% 0.3% 1.6% 0.4%
aThis table lists the percentages of each soil type over computational domain 1 for both the top layer (left side of the
table) and bottom layer (right side of the table) before and after the soil map update. The dominant soil type for the top
layer soil remains loam (type 6), whereas that for the bottom layer soil changes from clay loam (type 9) to loam (type 6).
The percentages of the dominant soil type(s) are shown in boldface.
Prediction, National Weather Service, NOAA, U.S. Department of Commerce, 2000]. Inadequate resolution is the
main problem with this type of soil moisture initialization. The soil moisture resolution is 1∘(around 100 km
at the equator). In our application, however, the smallest domain has a 1 km grid, meaning that one value in
the raw data represents the soil moisture of 100 by 100 grids, thus lacking realistic representation and spatial
variations. The accuracy of soil moisture initialization depends on the bias produced in the GFS model. Owing
to a lack of soil moisture measurements, a type of dynamic adjustment called model spin-up is needed for
Tab le 4. Major Changes in the Soil Map for Both the Top and Bottom Layers in Domain 1a
Rank Original Soil Type Updated Soil Type Percentage Over Updated Grids
Domain 1 Top Layer Soil Map
1ClayLoam(9) Loam (6) 20.6%
2Loam(6) Silt Loam (4) 10.1%
3Loam(6) Sandy Loam (3) 8.4%
4 Sandy Clay Loam (7) Loam (6) 6.2%
5 Loam (6) Loamy Sand (2) 5.3%
Domain 1 Bottom Layer Soil Map
1 Clay Loam (9) Loam (6) 41.3%
2 Clay Loam (9) Sandy Loam (3) 7.3%
3 Clay (12) Loam (6) 6.5%
4 Clay (12) Clay Loam (9) 4.4%
5 Sand (1) Loamy Sand (2) 3.8%
aThe top part of the table refers to the major changes in the soil map for the top layer soil in computational domain 1,
and the bottom part of the table to the major changes in that for the bottom layer soil in computational domain 1. The
major change in the soil maps for both layers of soil is from clay loam (type 9) to loam (type 6), which accounts for 20.6%
and 41.3%, respectively, over the updated grids. The 3 major changes (in boldface) in the top layer soil type correspond
to Figure 6 (a)-(c) respectively.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8785

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 6. Average volumetric soil moisture time series of the top three changes in the updated soil map of China. The top three changes in the updated soil type
from the default WRF soil type across China is from clay loam to loam, loam to silt loam, and loam to sandy loam. (a– c) The 1 m deep volumetric soil moisture
simulated and averaged over all grids with specified soil type change for the 2006 to 2011 period, with the specified changes in soil type corresponding tothe
top three changes in the soil map of China. The blue curve indicates simulations with the original soil type, and the green simulations with the updated soil type.
The red curve indicates the absolute difference between the two, scaled up by 3 times to clearly show the difference on the same graph. (d–f ) Locations withthe
specified change in soil type.
the preforecast period to minimize bias in the soil moisture field. The effect of changing the soil map on the
spun-up soil moisture is discussed in this section, and suggestions for the spin-up period are given.
4.1. Spinning-Up Soil Moisture Initialization
Soil moisture changes very slowly in dry seasons and rapidly in wet seasons if there is precipitation. In dry
seasons, even after 10 or 20 days of simulations, the soil moisture varies by less than 2% relative to that in the
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8786

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 7. Different soil moisture initialization fields for January and July 2011. (a and b) The volumetric soil moisture of the FNL analysis data at 8:00 A.M. (LT) on
31 December 2010 and 30 June 2011. (c and d) The volumetric soil moisture using the default soil map with 5 year spinning-up at 8:00 A.M. (LT) on 31 December
2010 and 30 June 2011. (e and f) The volumetric soil moisture using the updated soil map with 5 year spinning-up at 8:00 A.M. (LT) on 31 December 2010 and
30 June 2011.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8787

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Tab le 5. Configuration of Four Sets of Simulationsa
Experiments Sim-DF Sim-DS Sim-UF Sim-US
Soil map Default Default Updated Updated
Soil moisture initialization method FNL Spin-up FNL Spin-up
aFour sets of simulations using the same physics schemes but different combinations of soil maps (default WRF
soil map and updated soil map) and soil moisture initializations (from FNL analysis and spinning-up soil moisture
initialization).
initial condition. As soil moisture maintains a roughly constant value throughout the simulation, the accuracy
of soil moisture initialization is important. Accordingly, a long simulation period is needed to spin-up soil
moisture. The spinning-up period for soil moisture should be more than 2 years, according to Lim et al. [2012].
To illustrate the long-term effect of a change in the soil map on the volumetric soil moisture, a 5 year period
is chosen as the spinning-up period in this study.
The simulations performed to spin-up the volumetric soil moisture consist of 1 month run segments with
1 day of overlap for the spinning-up of other meteorological fields from 2006 to 2011. They are performed
separately using the default soil map and updated soil map. To minimize the effect of incorrect wind predic-
tion, observation nudging on the surface wind field in domains 3 and 4 and analysis nudging on the wind field
above PBL in domain 1 are employed. With regard to simulating the physical process, the WRF Single-Moment
three-class scheme, Noah land surface scheme, ACM2 PBL scheme, MM5 similarity surface layer scheme,
Dudhia Longwave scheme, and Rapid Radiative Transfer Model shortwave scheme are selected.
The soil moisture field passes from one run segment to another. Note that the soil temperature has not been
passed from one run segment to another as the soil temperature can be spun-up in a few hours. For the top
three major changes in soil type in domain 1 (see Table 4), the spatially averaged time series of 1 m deep
volumetric soil moisture over all the grids with specified soil type change is shown in Figures 6a–6c, and
the corresponding spatial distribution of the changes from one type to another is shown in Figures 6d–6f.
Tab le 6. Monthly Average Improvement in 2 m Temperature Absolute Error of Sim-US Over Sim-DF at Each Hour
Between 10:00 and 15:00 (LT) for Both the January and July 2011 Casesa
SIM-US Improved by SIM-US Worsen by
January 2011 Case
>0.3 0.2– 0.3 0.1–0.2 0 –0.1 subtotal subtotal 0–0.1 0.1 –0.2 0.2– 0.3 >0.3
10:00 3 14 61 73 151 (79%) 41 (23%) 32 8 1 0
11:00 18 30 56 56 160 (83%) 32 (17%) 27 5 0 0
12:00 21 38 57 42 158 (82%) 34 (18%) 28 5 1 0
13:00 25 49 53 34 161 (84%) 31 (16%) 23 7 1 0
14:00 39 39 53 35 166 (86%) 26 (14%) 20 6 0 0
15:00 37 53 43 31 164 (85%) 28 (15%) 22 6 0 0
July 2011 Case
>0.3 0.2– 0.3 0.1–0.2 0 –0.1 subtotal subtotal 0 –0.1 0.1– 0.2 0.2– 0.3 >0.3
10:00 4 13 45 74 136 (71%) 56 (29%) 40 14 2 0
11:00 4 12 71 66 153 (80%) 39 (20%) 32 7 0 0
12:00 10 24 67 65 166 (86%) 26 (14%) 23 3 0 0
13:00 13 36 59 56 164 (85%) 28 (15%) 21 7 0 0
14:00 27 43 48 48 166 (86%) 26 (14%) 16 7 1 2
15:00 37 39 53 37 166 (86%) 26 (14%) 17 6 3 0
aFor the January 2011 case (shown in the top table), 192 stations are classified into eight categories. The average 2 m
temperature improvement shown in Sim-US is (1) more than 0.3∘, (2) 0.2– 0.3∘, (3) 0.1–0.2∘, and (4) 0– 0.1∘. For (5) to (8),
there is a corresponding worsening in Sim-US relative to Sim-DF. A similar categorization of the 192 stations for the July
2011 case is given in the bottom table. The dominant group(s) is presented in boldface.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8788

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 8. (a– f) Difference in average error in 2 m temperature at each station between Sim-DF and Sim-US for the July 2011 case. The average mean absolute
error for Sim-US and Sim-DF is calculated at each station and plotted here. Negative values (in blue) indicate an average improvement at the observation station
for Sim-US versus Sim-DF, with the degree of improvement shown by the size of the dots.
The time series in blue and green indicate the spun-up soil moisture obtained using the default soil map
and updated soil map, respectively. The curve in red indicates the scaled (by 3 times) absolute difference
in averaged volumetric soil moisture between the simulations run with the default and updated soil maps,
respectively. In Figure 6a, the time series of the spinning-up soil moisture with loam in the updated soil map
have smaller values than those with clay loam in the default soil map in the specified locations, which is con-
sistent with the bulk soil properties in Table 1. Both the wilting point and saturated soil moisture values are
smaller for loam than for clay loam. The two curves of the spun-up soil moisture diverge in the first year,
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8789

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 9. The average top 10% difference in mean absolute error of 2 m temperature between Sim-US and Sim-DF for the (a) January and (b) July 2011 cases.
The lowest 10% difference in mean absolute error is averaged at each station and plotted in the figure. Negative values (in blue) indicate an average
improvement at the observation station in Sim-US relative to Sim-DF, with the degree of improvement shown by the size of the dots.
and their difference then oscillates over the years. Annual cycles are observed in the soil moisture time series
in which the peaks located in the summer indicate precipitation, which is the major source of soil moisture.
In Figure 6b, the two soil moisture time series differ by almost a constant value after 3 years of spinning-up,
which is longer than the 1 year period in Figure 6a because the locations used for averaging in Figure 6b range
in latitude from 35∘north to 45∘north. Hence, locations with less precipitation take longer for the model to
adjust dynamically. Figure 6c shows the differences in soil moisture time series with higher fluctuation values
during the summer, with fluctuation stabilizing after 2.5 years of spinning-up. The corresponding changes in
Tab le 7. Monthly Average Improvement in 2 m Relative Humidity Absolute Error of Sim-US Over Sim-DF at Each Hour
Between 10:00 and 15:00 (LT) for Both the January and July 2011 Cases
SIM-US Improved by SIM-US Worsen by
January 2011 Case
>3% 2%– 3% 1%–2% 0% –1% subtotal subtotal 0%–1% 1%–2% 2% –3% >3%
10:00 0 6 38 30 74 (67%) 36 (33%) 25 10 1 0
11:00 0 4 32 36 72 (65%) 38 (35%) 23 12 3 0
12:00 0 3 28 34 65 (59%) 45 (41%) 29 12 4 0
13:00 0 3 24 44 71 (65%) 39 (35%) 25 10 4 0
14:00 1 3 21 47 72 (65%) 38 (35%) 26 9 3 0
15:00 1 3 22 54 80 (73%) 30 (27%) 21 8 1 0
July 2011 Case
>3% 2%– 3% 1%–2% 0% –1% subtotal subtotal 0%–1% 1%–2% 2% –3% >3%
10:00 0 0 12 53 65 (59%) 45 (41%) 34 11 0 0
11:00 0 1 8 50 59 (54%) 51 (46%) 42 9 0 0
12:00 0 1 11 52 64 (58%) 46 (42%) 43 3 0 0
13:00 0 0 15 53 68 (62%) 42 (38%) 39 3 0 0
14:00 0 0 12 56 68 (62%) 42 (38%) 37 5 0 0
15:00 0 0 8 60 68 (62%) 42 (38%) 38 4 0 0
aFor the January 2011 case (shown in the top part of the table), 110 stations are classified into eight categories. The
average 2 m relative humidity improvement shownin Sim-US is (1) more than 3%, (2) 2% – 3%, (3) 1%– 2%, and (4) 0% –1%.
For (5) to (8), there is a corresponding worsening in Sim-US relative to Sim-DF. A similar categorization of the 110 stations
for the July 2011 case is given in the bottom table. The dominant group(s) is presented in boldface.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8790

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 10. (a– f) Difference in average error in 2 m relative humidity at each station between Sim-DF and Sim-US for the July 2011 case. The average mean
absolute error for Sim-US and Sim-DF is calculated at each station and plotted here. Negative values (in blue) indicate an average improvement at the
observation station for Sim-US versus Sim-DF, with the degree of improvement shown by the size of the dots.
soil type are concentrated at latitudes between 30∘north and 35∘north. The implications of Figure 6 are that
soil type controls the long-term mean of soil moisture and that the spinning-up period is dependent on the
latitudes of the simulation regions. Regions with less precipitation need a longer period of time to spin-up the
soil moisture. The suggested length of the soil moisture spinning-up period is 3 years for locations at latitudes
ranging from 15∘north to 45∘north.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8791

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 11. The average top 10% difference in mean absolute error of 2 m relative humidity between Sim-US and Sim-DF for the (a) January and (b) July 2011
cases. The lowest 10% difference in mean absolute error is averaged at each station and plotted in the figure. Negative values (in blue) indicate an average
improvement at the observation station in Sim-US relative to Sim-DF, with the degree of improvement shown by the size of the dots.
5. Numerical Experiments
5.1. Experiment Outline
With regard to the changes in soil moisture initialization and soil map input discussed in section 4, numerical
simulations are required to determine whether the soil map changes and spinning-up method lead to sub-
stantial improvements in the model simulations. The main objective of this section is to show the effects of
the proposed methods established in section 4.
5.2. Data
The main data needed to drive the WRF model [Skamarock et al., 2008] are analysis data. In this study, FNL
analysis data are used. These data [National Centers for Environmental Prediction, National Weather Service,
NOAA, U.S. Department of Commerce, 2000] constitute the analysis step in the GFS model, which has a reso-
lution of 1∘by 1∘(around 100 km at the equator). They are available for a daily 6 h period from 20 July 1997
to a near-current date. The data set used in the simulations is in GRIB1 format and contain 40 analysis fields.
The meteorological fields used are temperature, velocity, relative humidity, height, surface pressure, sea level
pressure, soil moisture, soil temperature, sea ice flag, land-sea flag, skin temperature, and water equivalent
snow depth.
Our observation data are obtained from local meteorological authorities, including the Guangdong
Meteorological Service, Hong Kong Observatory (HKO), and Macao Meteorological and Geophysical Bureau
(MASMG). The meteorological fields in the observations include air temperature, relative humidity, velocity,
surface pressure, and sea level pressure. One hundred ninety-two temperature observations and 110 relative
humidity observations are utilized for evaluation.
5.3. Simulation Period
Four sets of simulations are performed for January 2011 and July 2011 with 4 day run segments to demon-
strate the air-land interaction in dry and wet conditions, respectively. Only days without precipitation are
considered, as our aim is to demonstrate the effect of spinning-up the soil moisture map.
5.4. Soil Moisture Initializations
We use three different soil moisture initializations in our simulations. The initial soil moisture conditions for
January 2011 and July 2011 are presented in Figure 7. The left- and right-hand sides of Figure 7 show the
three initial conditions for the January 2011 and July 2011 cases, respectively: Figures 7a and 7b are from
the FNL analysis data with coarse resolution, Figures 7c and 7d are the 5 year spun-up soil moisture using
the default soil map; and Figures 7e and 7f are the 5 year spun-up soil moisture using the updated soil map.
With the spinning-up runs, regardless of the soil map used, the soil moisture field has a higher resolution and
more spatial variations than when the FNL data are used. For the spun-up soil moisture using the default and
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8792

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 12. Monthly average difference (Sim-US - Sim-DF) in 1 m deep volumetric soil moisture at each hour between 10:00 A.M. and 3:00 P.M. (LT) for the July
2011 case. (a– f) The monthly average difference in volumetric soil moisture between Sim-US and Sim-DF over the 6 h period.
updated soil maps, similar structures can be observed in the eastern part of Tibet, as there are no changes
in soil type in that region. Greater differences can be observed in southern China, where the major change is
from clay loam (type 9) to loam (type 6). In terms of the physical properties of the soil, loam holds less water
than clay loam, as there is a smaller percentage of clay in loam (18%) than in clay loam 34% (refer to Table 1).
In southern China, the average spun-up soil moisture using the default soil map is around 0.4, whereas the
value for the updated soil map is around 0.3. These results are consistent with the updated soil map holding
less water in southern China. Similarly, for the July case, the right-hand side of Figure 7 shows the two spun-up
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8793

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 13. Monthly average difference (Sim-US - Sim-DF) in soil temperature at each hour from 10:00 A.M. to 3:00 P.M. (LT) for the July 2011 case.
(a– f) The monthly average difference in soil temperature between Sim-US and Sim-DF over the 6 h period.
soil moistures to have more spatial variations. In the eastern part of Tibet, the two soil moisture fields have a
similar soil moisture structure. Less water can be held using the updated soil map than the default soil map
in southern China.
5.5. Numerical Simulation Configurations
The simulations are conducted to determine whether use of the updated soil map and spinning-up soil mois-
ture initialization can improve near-surface temperature and relative humidity prediction. The observation
nudging applied to the wind and physics schemes is identical to that in section 4.1. There are four sets of
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8794

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 14. Monthly average difference (Sim-US - Sim-DF) in latent heat flux at each hour from 10:00 A.M. to 3:00 P.M. (LT) for the July 2011 case.
(a– f) The monthly average difference in sensible heat flux between Sim-US and Sim-DF over the 6 h period.
simulations for each simulation period. We label the four experiments Sim-DF, Sim-DS, Sim-UF, and Sim-US.
The main difference among the four sets is the soil type used and the method of soil moisture initialization.
The configurations of the four sets of simulations are listed in Table 5.
6. Simulation Results
The improvement in 2 m temperature and relative humidity prediction is discussed in this section, although
with analysis of the reasons for that improvement.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8795

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 15. Monthly average difference (Sim-US - Sim-DF) in sensible heat flux at each hour from 10:00 A.M. to 3:00 P.M. (LT) for the July 2011 case.
(a– f) The monthly average difference in sensible heat flux between Sim-US and Sim-DF over the 6 h period.
For both the January and July cases, 2 m temperature data from 192 observation stations and relative humid-
ity data from 110 observation stations across the PRD are used for error analysis. Table 6 shows the average
improvement in the 2 m temperature error of Sim-US over Sim-DF for each hour of the day from 10:00 A.M.
(LT) to 3:00 P.M. (LT) in January and July 2011. At different times of the day between those hours, 79% to 86%
of the stations in January 2011 and 71% to 86% of the stations in July 2011 showed improvement in 2 m
temperature. The average improvement in Sim-US over Sim-DF across the stations is around 0.2∘.
Root-mean-square error (RMSE) of 2 m temperature is 1.94∘for SIM-DF, average improvement of 0.2∘refers
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8796

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 16. Monthly average difference (Sim-US - Sim-DF) in 2 m temperature at each hour from 10:00 A.M. to 3:00 P.M. (LT) for the July 2011 case.
(a– f) The monthly average difference in 2 m temperature between Sim-US and Sim-DF over the 6 h period.
to 10% of the error in 2 m temperature has been improved. Figures 8a–8f show the difference in average 2 m
temperature error between the Sim-DF and Sim-US at each station from 10:00 A.M. to 3:00 P.M. for the entire
month of July 2011. The majority of the observations (blue dots) show Sim-US to achieve superior2mtem-
perature prediction to Sim-DF, with an average improvement of more than 1∘observed at some locations.
Figures 9a and 9b show the average top 10% difference in mean absolute error of 2 m temperature of Sim-US
over Sim-DF for January 2011 and July 2011, respectively. Stations in the west of domain 4 show average
improvements in excess of 1 ∘10% of the time in both months.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8797

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
Figure 17. Monthly average difference (Sim-US - Sim-DF) in PBL height at each hour from 10:00 A.M. to 3:00 P.M. (LT) for the July 2011 case. (a –f ) The monthly
average difference in PBL height between Sim-US and Sim-DF over the 6 h period.
Table 7 showsthe average improvement in the2mrelativehumidity error of Sim-US over Sim-DF for each hour
of the day from 10:00 A.M. (LT) to 3:00 P.M. (LT) in January and July 2011. At different times of the day between
those hours, 59% to 73% of the stations in January 2011 and 54% to 62% of the stations in July 2011 showed
improvement in 2 m relative humidity. The average improvement in Sim-US over Sim-DF across the stations
is around 1%. RMSE of relative humidity is 12.57% for SIM-DF; average improvement of 1% refers to 8% of the
error in relative humidity has been improved. Figures 10a–10f show the difference in average 2 m humidity
error between the Sim-DF and Sim-US at each station from 10:00 A.M. to 3:00 P.M. for the entire month of
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8798

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
July 2011. The majority of the observations (blue dots) show Sim-US to achieve superior2mrelativehumidity
prediction to Sim-DF, with an average improvement of more than 1% observed at some locations. Figures 11a
and 11b show the average top 10% difference in mean absolute error of2mrelativehumidity of Sim-US over
Sim-DF for January 2011 and July 2011, respectively. Stations in Hong Kong show average improvements in
excess of 5% of the time in both months.
Figures 12– 17 are spatial plots to examine the regional meteorological response of Sim-US from Sim-DF for
July case. Figure 12 illustrates the spatial distribution of the volumetric soil moisture of 1 m deep soil of Sim-DF
and Sim-US. For the west of domain 4, the average soil moisture decreases more than 30% from Sim-DF to
Sim-US. According to equations (3) and (6), the thermal diffusivity in Sim-US is much higher than that in
Sim-DF, and the specific heat is lower in Sim-US than in Sim-DF. Hence, the diffusion speed (Kt∕cp) is faster.
Figure 13 shows that the soil temperature rises much faster in Sim-US than in Sim-DF in the west of domain
4, which leads to the increase in soil temperature from 10:00 A.M. to 3:00 P.M. (LT). Figure 14 shows the latent
heat flux of the 6 h period that is almost constant as the decreased volumetric soil moisture (Figure 12) limits
the rate of evaporation. Reduction in latent heat flux allowed more energy partition into sensible heat flux.
It can be seen in Figure 15 that the heat flux in the west of domain 4 is generally larger in Sim-US than in
Sim-DF due to the greater difference between soil temperature and air temperature related to equation (7).
Hence, the temperature increases faster in Sim-US than in Sim-DF in the west of domain 4 during the daytime,
as shown in Figure 16. The improved temperature prediction in the west of domain 4 reduces the negative
bias in temperature, thus explaining the improvement in2mtemperature prediction in Sim-US compared
to Sim-DF. The PBL height is also greater, which affects air quality modeling. Figure 17 shows the greater PBL
height prediction of Sim-US over Sim-DF during the daytime in the west of domain 4.
For January case, the relative higher soil moisture value (with respect to the soil type) will keep more energy
from the radiation, slowing down the rate of increase of soil temperature, and decrease the rate of increase
of sensible heat flux, 2 m temperature, and PBL height. The simulated results with the updated soil type can
reduce the overprediction in2mtemperature.
7. Summary
In this study, an updated global soil map for the WRF model is generated by combining the STATSGO/FAO soil
data set and BNU global soil data set, which gives a 30 s resolution over the globe. The differences between
the default WRF soil map and updated soil map are examined herein. The updated soil map produces drier
soil properties than the default soil map in domain 1. A 5 year (from 2006 to 2011) simulation is performed to
spin-up the volumetric soil moisture content. As illustrated in section 4, the soil map exerts a strong effect on
the volumetric soil moisture content over a long period of time. A 3 year spinning-up period for such content
is recommended. The updated soil map affects obvious improvements in spatial resolution and representa-
tion of the physical properties of soil, as revealed by the simulation results. The prediction of both the 2 m
temperature and 2 m relative humidity is improved with the updated soil map and the spun-up soil moisture.
The improvement in2mtemperature prediction can be explained by the dynamic relations among soil mois-
ture, soil temperature, surface heat fluxes, and 2 m temperature and the subsequent effect on PBL height,
which is crucial to air quality modeling.
References
Chen, F., and J. Dudhia (2001), Coupling an advanced land surface-hydrology model with the Penn State-NCAR MM5 modeling system.
Part I: Model implementation and sensitivity, Mon. Weather Rev.,129(4), 569– 585.
Chen, F., K. Mitchell, J. Schaake, Y. Xue, H.-L. Pan, V. Koren, Q. Y. Duan, M. Ek, and A. Betts (1996), Modeling of land surface evaporation
by four schemes and comparison with Fife observations, J. Geophys. Res.,101, 7251 –7268.
Chen, F., Z. Janji´
c, and K. Mitchell (1997), Impact of atmospheric surface-layer parameterizations in the new land-surface scheme of the
NCEP mesoscale Eta model, Boundary Layer Meteorol.,85(3), 391– 421.
Cosby, B., G. Hornberger, R. Clapp, and T. Ginn (1984), A statistical exploration of the relationships of soil moisture characteristics to the
physical properties of soils, Water R esour. R es.,20(6), 682 –690.
Davis, R. O. E., and H. H. Bennett (1927), Grouping of Soils on the Basis of Mechanical Analysis, vol. 419, 1– 15.
Ek, M., and L. Mahrt (1991), OSU 1-D PBL model user’s guide, version, 1(120),97, 331– 2209.
Ek, M., K. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. Tarpley (2003), Implementation of Noah land surface model
advances in the National Centers for Environmental Prediction operational mesoscale Eta model, J. Geophys. Res.,108(D22), 8851,
doi:10.1029/2002JD003296.
FAO (1991), The digitized soil map of the world, World Soil Resource Rep. 67, FAO, Rome.
FAO-UNESCO (1971– 1981), Soil Map of the World (1:5,000,000), vol. 1– 10, UNESCO, Paris, France.
Acknowled gments
We would like to express our deepest
gratitude to the Editor and anonymous
reviewers for their insightful and
useful comments. We also thank
W. Shangguan for providing us with
access to the global soil database on
the website of the Land-Atmospheric
Interaction Research Group at BNU
(http://globalchange.bnu.edu.cn/
research/soilw). Thanks are also
due the Guangdong Meteorological
Service, HKO, and MASMG for
providing the observation data. All
of the other data used are properly
cited and listed in the references.
All data shown in the results are
available at http://goo.gl/kh3DOz.
This work is supported by Research
Grants Council in Hong Kong with
grant number: 16300715.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8799

Journal of Geophysical Research: Atmospheres 10.1002/2015JD024558
FAO-UNESCO (1974), Soil Map of the World (1:5,000,000), vol. 1 legend, UNESCO, Paris, France.
Garratt, J. (1993), Sensitivity of climate simulations to land-surface and atmospheric boundary-layer treatments— A review, J. Climate,6(3),
419– 448.
Jacquemin, B., and J. Noilhan (1990), Sensitivity study and validation of a land surface parameterization using the HAPEX-MOBILHY data set,
Boundary Layer Meteorol.,52(1– 2), 93– 134.
Jarvis, P. (1976), The interpretation of the variations in leaf water potential and stomatal conductance found in canopies in the field, Philos.
Trans. R. Soc. London, Ser. B,273(927), 593– 610.
Lam, J., A. Lau, and J. Fung (2006), Application of refined land-use categories for high resolution mesoscale atmospheric modelling,
Boundary Layer Meteorol.,119(2), 263– 288.
Lim, Y.-J., K.-Y. Byun, T.-Y. Lee, H. Kwon, J. Hong, and J. Kim (2012), A land data assimilation system using the MODIS-derived land data and its
application to numerical weather prediction in East Asia, Asia-Pac. J. Atmos. Sci.,48(1), 83– 95.
Mahrt, L., and M. Ek (1984), The influence of atmospheric stability on potential evaporation, J. Clim. Appl. Meteorol.,23, 222– 234.
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(2), 1 –26.
National Centers for Environmental Prediction/National Weather Service/NOAA/U.S. Department of Commerce (2000), NCEP FNL
Operational Model Global Tropospheric Analyses, continuing from July 1999, Research Data Archive at the National Center for
Atmospheric Research, Computational and Information Systems Laboratory, Boulder, Colo., doi:10.5065/D6M043C6, [updated daily.]
Noilhan, J., and S. Planton (1989), A simple parameterization of land surface processes for meteorological models, Mon. Weather Rev.,117(3),
536– 549.
Peters-Lidard, C., E. Blackburn, X. Liang, and E. Wood (1998), The effect of soil thermal conductivity parameterization on surface energy
fluxes and temperatures, J. Atmos. Sci.,55(7), 1209– 1224.
Pleim, J. E., and A. Xiu (1995), Development and testing of a surface flux and planetary boundary layer model for application in mesoscale
models, J. Appl. Meteorol.,34(1), 16– 32.
Richards, L. A. (1931), Capillary conduction of liquids through porous mediums, J. Appl. Phys.,1(5), 318– 333, doi:10.1063/1.1745010.
Schaake, J. C., V. I. Koren, Q.-Y. Duan, K. Mitchell, and F. Chen (1996), Simple water balance model for estimating runoff at different spatial
and temporal scales, J. Geophys. Res.,101, 7461 –7475.
Shangguan, W., et al. (2013), A China data set of soil properties for land surface modeling, J. Adv. Model. Earth Syst.,5(2), 212 –224.
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(1),
249– 263.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, M. G. Duda, X.-Y. Huang, W. Wang, and J. G. Powers (2008), A description of
the Advanced Research WRF version 3, Tech. Rep. Note NCAR/TN-475+STR, DTIC Document, Boulder, Colo.
Soil Survey Division Staff (1993), Soil Survey Manual, Soil Conservation Service, vol. 18, USDA, U.S.A.
Xie, B., J. C. Fung, A. Chan, and A. Lau (2012), Evaluation of nonlocal and local planetary boundary layer schemes in the WRF model,
J. Geophys. Res.,117, D12103, doi:10.1029/2011JD017080.
DY AND FUNG SOIL MAP AND SOIL MOISTURE 8800