Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Earth-Science Reviews 211 (2020) 103419

Contents lists available at ScienceDirect

Earth-Science Reviews
journal homepage: www.elsevier.com/locate/earscirev

Review article

Room for improvement: A review and evaluation of 24 soil thermal


conductivity parameterization schemes commonly used in land-surface,
hydrological, and soil-vegetation-atmosphere transfer models
Hailong He a, Dong He b, Jiming Jin b, c, Kathleen M. Smits d, e, f, Miles Dyck g, Qingbai Wu h,
Bingcheng Si b, i, *, Jialong Lv a, j, **
a
College of Natural Resources and Environment, Northwest A&F University, Yangling, Shaanxi 712100, China
b
Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas, Ministry of Education, Northwest A&F University, Yangling, Shaanxi 712100,
China
c
Department of Watershed Sciences, Quinney College of Natural Resources, Utah State University, Logan, UT 84322-5210, United States
d
Department of Civil Engineering, The University of Texas at Arlington, Arlington, TX 76019, United States
e
Department of Civil and Environmental Engineering, Colorado School of Mines, Golden, CO 80401, United States
f
Environmental Engineering at the US Air Force Academy, Colorado Spring, CO. United States
g
Department of Renewable Resources, University of Alberta, Edmonton, AB T6G 2H1, Canada
h
State Key Laboratory of Frozen Soil Engineering, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, Gansu 730000, PR
China
i
Department of Soil Science, University of Saskatchewan, Saskatoon, SK S7N5A8, Canada
j
Key Laboratory of Plant Nutrition and the Agri-Environment in Northwest China (Ministry of Agriculture), Northwest A&F University, Yangling, Shaanxi 712100, China

A R T I C L E I N F O A B S T R A C T :

Keywords: Effective thermal conductivity of soils (λeff) is a critical parameter for agriculture, environment science, and
Soil thermal conductivity scheme engineering. Functions to estimate λeff from readily available soil properties, known as soil thermal conductivity
Model evaluation (STC) schemes, are needed by land-surface models (LSMs), hydrological models, and soil-vegetation-atmosphere
Land-surface models
transfer (SVAT) models to study the land surface energy balance, heat flux, and soil thermal regime under various
Soil water content
Organic matter
climates and geographic regions. The selection of a STC scheme can result in large differences in temperature
Quartz content estimates in LSMs, sometimes masking the effects of climate change. Therefore, accurate selection of a STC
Matric potential scheme is critically important to LSM estimates. Although a number of STC schemes have been incorporated in
various LSMs, no study has systematically evaluated their performance. Therefore, the objectives of this study
were to review and evaluate STC schemes employed by LSMs by comparing (1) predicted and measured STCs and
(2) modelled land surface temperature (LST) using the Community Land Model at three selected sites and the
corresponding LST data from the moderate resolution imaging spectrometer (MODIS). In total, 24 STC schemes
were collated from 38 mainstream LSMs, SVAT, and hydrological models. They were divided into three cate­
gories based on model types: one physically-based scheme, eight linear/non-linear regression schemes, and 13
normalized schemes. We also include two schemes that express STC as a function of matric potential (ψ , hereafter
referred as λeff (ψ ) schemes). The first three types of STC schemes were evaluated with a large compiled dataset
consisting of 439 unfrozen and frozen measurements of λeff from 16 soils. The λeff (ψ ) schemes were evaluated
with simultaneously measured λeff (ψ ) from eight soils from two separate or independent studies. Results showed
that none of the STC schemes could be used to accurately predict λeff for all soil types. STC scheme performance
largely depended on the size (number of samples) and characteristics (e.g., soil types) of the data used for
comparison. Some STC schemes work well on certain types of soils, but care should be taken for larger scale
applications. LSM simulated LST for 24 STC schemes varied when compared with MODIS LST. In general, the STC
schemes performed better in medium- and coarse-textured soils than in fine-textured soils. However, large dis­
crepancies were observed on the estimated LST using different STC schemes for medium and coarse-textured

* Correspondence to: B. Si, Department of Soil Science, University of Sasktachewan, Saskatoon, SK, S7N5A8, Canada.
** Correspondence to: Jialong Lv, College of Natural Resources and Environment, Northwest A&F University, Yangling, Shaanxi 712100, China.
E-mail addresses: hailong.he@nwafu.edu.cn (H. He), 1558798954@nwafu.edu.cn (D. He), jiming.jin@usu.edu (J. Jin), kathleen.smits@uta.edu (K.M. Smits),
miles.dyck@ualberta.ca (M. Dyck), qbwu@lzb.ac.cn (Q. Wu), bing.si@usask.ca (B. Si), ljlll@nwsuaf.edu.cn (J. Lv).

https://doi.org/10.1016/j.earscirev.2020.103419
Received 17 March 2020; Received in revised form 20 October 2020; Accepted 25 October 2020
Available online 29 October 2020
0012-8252/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
H. He et al. Earth-Science Reviews 211 (2020) 103419

soils. We recommend that LSM modelers be mindful of the inherent bias in STC schemes on the surface tem­
perature estimates and hence overall model predictions. Orchestrated efforts are urgently needed on the part of
the soil science, hydrology, and climatology communities to develop a more extensive and systematic λeff
database for development and evaluation of improved STC schemes for wider and more accurate applications.

1. Introduction Vegetation-Atmosphere Transfer (SVAT) model. Luo et al. (2009b) and


Yang et al. (2018) demonstrated a temperature error up to 3 ◦ C resulting
Soil thermal properties influence how energy is partitioned within from two STC schemes (Farouki, 1981; Johansen, 1975) with the
the environment and are therefore critically important to many disci­ Common Land Model (CoLM) and Community Land Model (CLM4.5) on
plines including earth and environmental science, agronomy, and en­ the Tibetan plateau. Hu et al. (2017) evaluated three STC schemes
gineering. Soil thermal properties determine a variety of biological, (Farouki, 1981; Johansen, 1975; Luo et al., 2009b) in permafrost regions
chemical, and physical processes and properties in the Earth’s critical of the Tibetan plateau. Their results showed that the Johansen (1975)
zone, including soil’s response to forest fires, plant root respiration, and and Luo et al. (2009b) schemes performed well in frozen and unfrozen
methane exchange, and global climate and hydrological cycles (Cheng soils, respectively. Dai et al. (2019a) compared seven STC schemes
et al., 2020; Halloran et al., 2016; He et al., 2018b; Rau et al., 2014). (Balland and Arp, 2005; Côté and Konrad, 2005; de Vries, 1963; Farouki,
Furthermore, soil thermal properties alter soil and land-surface tem­ 1981; Johansen, 1975; Lu et al., 2007; Tarnawski and Leong, 2012) in
perature, thus affecting seed germination, seedling development, and CoLM and found that the scheme of Balland and Arp (2005) performed
stand establishment as well as plant water use and irrigation scheduling the best for estimating soil temperature.
(Abu-Hamdeh and Reeder, 2000; Campbell, 1985; Chung and Horton, The above-mentioned studies demonstrate that LSM outputs are
1987; Ghuman and Lal, 1985). They also play a key role in mass and sensitive to STC scheme selection (Clark and Arritt, 1995; Peters-Lidard
energy exchange through porous media and interactions between the et al., 1998). This is because soil thermal properties are often derived
ground and atmosphere, which in turn alter climate at regional and from the LSM STC scheme. Without independent STC validation, the
global scales (Ebel et al., 2019; Harris et al., 2009; Koster et al., 2004; effects of the STC scheme selection can be masked by driving forces or
Kurylyk et al., 2014; Lafrenière and Lamoureux, 2019; Li et al., 2019; other highly parameterized relationships such as soil water retention
Yang and Wang, 2019; Yang et al., 2018; Yang et al., 2010). characteristics (Clark et al., 2008; Duan et al., 2006; Kavetski et al.,
Soil thermal conductivity, volumetric heat capacity, and thermal 2006; Wagener and Wheater, 2006).
diffusivity are three key thermal properties (He et al., 2018b). Soil Unfortunately, many STC schemes selected for use in LSMs have not
thermal conductivity (STC), which is interchangeably referred as the been thoroughly validated with experimental measurements of λeff (Hu
effective thermal conductivity of soil (λeff, W m-1 ◦ C-1), defined as the et al., 2017; Luo et al., 2009a; Luo et al., 2009b) or only used a limited
quantity of heat that flows through a unit area in a unit time under a unit dataset representing certain soil types or limited ranges of water con­
temperature gradient. Soil volumetric heat capacity (Cv or ρc) defines tents (Peters-Lidard et al., 1998). Tong et al. (2016) and Dai et al.
the change in soil heat storage required to cause a unit rise in temper­ (2019a) were among the first to validate the performance of seven STC
ature. Soil thermal diffusivity (κ or α) describes the transmission rate of schemes with a relatively large dataset, but no experimental measure­
temperature changes within the soil or the ability to transmit heat over ments of λeff for frozen soil were included. Because over half of the
the ability to store heat. Cv can be calculated based on the volume and ground surface in the Northern Hemisphere undergoes annual freeze-
specific heat capacity of soil components (Wang et al., 2019), while κ is thaw processes cycles (Zhang et al., 2004), the testing and then subse­
the ratio of λeff to Cv. Therefore, λeff is a critical parameter requiring quent implementation of an STC scheme that represents both frozen and
either estimation or measurement for properly representing soil heat unfrozen environments are critical to ensure proper prediction under the
flux at or below the land surface. Because soil heat flux dictates the total span of environments. However, to the best of our knowledge, no
partition of net radiation into latent and sensible heat flux (Mauder study has verified STC schemes with both experimental measurements
et al., 2020; Russell et al., 2015; Verhoef et al., 2012), λeff is incorporated of λeff for unfrozen and frozen soils, yet measurements of λeff for unfrozen
into land-surface models (LSMs) to estimate the ground thermal regime and frozen soils are readily available in the literature. Therefore, there is
(e.g., soil temperature) or surface heat fluxes at different scales (Bristow, an urgent need to capitalize on availability of measured STC over a wide
2002; He et al., 2018a; He et al., 2018b; Peters-Lidard et al., 1998; range of soil properties to evaluate STC schemes incorporated in main­
Verhoef, 2004; Wang et al., 2012; Yang et al., 2005; Yang et al., 2007; stream LSMs, SVAT, and hydrological models.
Zhang et al., 2012). The objectives of this study, therefore, were to: (1) conduct an
Like soil hydraulic properties, STC depends strongly on factors such extensive review to collate STC schemes implemented in mainstream
as soil water content, bulk density, soil texture, organic matter content, LSMs, SVAT, and hydrological models; (2) compile a large dataset of
mineralogy, and temperature. Therefore, because of the vast variety of experimental λeff of soils with a wide range of properties (e.g., water
soil types and conditions, STC parameterization remains a challenge for content, soil texture, organic matter content, unfrozen and frozen tem­
different soils. Numerous STC functions or schemes are available in the peratures); and (3) fully evaluate the performance of the collated STC
literature (Cheng and Hsu, 1999; Dong et al., 2015; He et al., 2020d; schemes with the compiled dataset.
Zhang and Wang, 2017) and many of them have been incorporated into
various LSMs. Although much effort has been made to determine 2. Material and methods
appropriate STC schemes for LSMs by comparing different STC schemes
with measured soil temperature (Chen et al., 2012; Lawrence et al., 2.1. Review of STC schemes in numerical models
2011; Luo et al., 2017; Luo et al., 2009a; Peters-Lidard et al., 1998; Tong
et al., 2016; Wang et al., 2017), different STC schemes result in An extensive review returned 24 soil thermal conductivity (STC)
distinctive simulation results with LSMs (Dai et al., 2019a; Luo et al., schemes incorporated in 38 LSMs, SVAT, and hydrological models
2009b; Peters-Lidard et al., 1998; Yang et al., 2018). For example, Pe­ (hereafter referred to LSMs only). A summary of these 24 STC schemes is
ters-Lidard et al. (1998) reported a 1~3 ◦ C difference in soil temperature tabulated below (Table 1); a detailed description of these schemes can be
simulation and up to 56~74 W m-2 difference in simulation of latent and found in the supplementary file. To facilitate communication, the
sensible heat flux compared to experimental data using two STC models have been abbreviated using the initials of the author(s) and year
schemes (Johansen, 1975; McCumber and Pielke, 1981) with the Soil- of publication in the references cited in Table 1. For single-authored

2
H. He et al. Earth-Science Reviews 211 (2020) 103419

Table 1
Selected soil thermal conductivity (STC) schemes incorporated in land-surface models (LSMs). Additional details and variable definitions can be found in the sup­
plementary file.
No Source Abbrev. Expressions Parameters Land-surface Note
models host*

Physically-based STC scheme


1 de Vries DV1963 1 ∑[ (∑ ) Fi, ϕi, λs, gk BEST, SHAW, Theoretical-
∑N ∑N ]− 1 gk = 1; k = a, b, c
λeff = i=0 (Fi ϕi λi )/ i=0 Fi ϕi , Fi = 1 + (λi /λ0 − 1)gk ,
(1963) 3 k k STEMMUS discrete
model,
unfrozen and
frozen
Linear/non-linear regression STC schemes
2 Kersten KM1949 θg, ρb COUPModel, Empirical-
(1949) UC1 and SOIL, ST&TP- Linear
λeff =
UC2 are ES regression
unit model,
⎧ ( )
UC1 0.9logθg − 0.2 × 100.01UC2 ⋅ρb conversion Unfrozen and

⎪ Unfrozen silt/clay (a)
⎨ UC ( 0.01 × 100.022UC2 ⋅ρb + 0.085 × 100.008UC2 ⋅ρb θ )
⎪ factors frozen soils
1
[( ) ] g Frozen silt/clay (b)
⎪ UC1 0.7logθg + 0.4 100.01UC2 ⋅ρb
⎪ Unfrozen sandy soils (c)
⎪ [ ]

UC1 0.076 × 10 0.013UC2 ⋅ρb
+ 0.0032 × 10 0.0146UC2 ⋅ρb
θg frozen sandy soils (d)
3 de Vries DV1975 λeff, om = h1 + h2θv h1 , h2 , θ v CoLM, SOIL, Humus or
(1975) COUPModel, organic
STEMMUS matter
4 Camillo and CS1981 λeff = 0.4186[1.5(1 − P) + 1.3θv]/(0.75 + 0.65P − 0.4θv) P, θv SiB2
Schmugge
(1981)
5 Cass et al. CS1984 λeff = A1 + B2 ⋅ Sr − (A1 − D1) exp [− (C1 ⋅ Sr)E1] A1, B1, C1, UNSAT-H Originally for
(1984) D1, E1, Sr unfrozen soils
only, SVAT
6 Campbell CG1985 λeff = A2 + B2θv − (A2 − D2) exp [− (C2θv)E2] A2, B2, C2, SLI, SVATS, Also referred
(1985) 0.57 + 1.73fquartz + 0.93fother D 2 , E 2 , θv, as McInnes
A2 = − 2.8P(1 − P), B2 = 2.8(1 − P)
1 − 0.74fquartz − 0.49fother P, fquartz, (1981),
√̅̅̅̅̅̅̅̅
C2 = 1 + 2.6/ fclay , D2 = 0.03 + 0.7(1 − P)2, E2 = 4 fclay originally for
unfrozen soils
√̅̅̅̅̅
7 Chung and CH1987 λeff = a + bθv + c θv a, b, c, θv HDS-SPAC originally for
Horton unfrozen soils
(1987)
{ [ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ] }/
8 Becker et al. BB1992 a1, a2, a3, SiB2 Empirical
λeff = UC1 log [(100Sr )/a1 +Sinh(a4 ) ] + [(100Sr )/a1 + Sinh(a4 ) ]2 + 1 − a3 a2
(1992) a 4, S r model, for
both frozen
and unfrozen
states
( )
9 Hubrechts HL1998 − C3 θv, fclay, fom, MACRO
λeff = A3 + B3 exp , A3 = − 0.295 + 1.26fclay + 0.388ρb,
(1998) 100θv ρb
B3 = − 1.776 + 2.0476ρb + 12.4fom,C3 = exp (0.976 + 6.5fclay + 14fom)
STC schemes based on Normalized concept
10 Johansen JO1975 λeff = Ke(λsat − λdry) + λdry P, Sr, θlw, CABLE, Empirical-

(1975) ⎨ 0.7log10 Sr + 1 Sr > 0.05, coarse soil, unfrozen (a) ρb, fquartz, λs CLASS, ISBA- normalized
Ke = log10 Sr + 1 Sr > 0.1, fine soil, unfrozen (b) MEB, NCEP, model

Sr frozen soils (c) PRM

⎨ λPw λ1−
s
P
Unfrozen soil (a)
λsat = λs = λfquartz
quartz
λ1− fquartz
other
⎩ λP− θlw λ1− P λθlw Frozen soil (b)
ice s w

λdry = (0.135ρb + 0.0647)/(ρs − 0.947ρb)


{ ( )
11 Farouki FO1981 Ke λsat − λdry + λdry Sr > 10− 7 (a) T, Tf, Sr CLM4/CESM, Empirical-
λeff =
(1981) λdry Sr ≤ 10− 7 (b) CoLM, DOS- modified
{
log10 Sr + 1 T ≥ Tf (a) TEM, GCM, normalized
Ke =
Sr T < Tf (b) JULES, mode
λs = [8.8fsand + 2.92fclay]/[fsand + fclay] MOSES, UM

12 Lunardini LV1981 ⎨ λT+ ( ) T > Tf + ΔT (a) P, θice, T, Tf, LSM Empirical-
(1981) λeff = λ + (λT+ − λT− )(2ΔT)− 1 T − Tf + ΔT Tf − ΔT ≤ T ≤ Tf + ΔT (b) λT+ = modified
⎩ T−
λT− T < Tf − ΔT (c) normalized
( 1− P θ ) ( 1− P θice )
λs λw − 0.15 θ/P + 0.15λT− = λs λice − 0.15 θice /P + 0.15 mode
13 Verseghy VD1991 λeff = (λsat − λdry)θtotal/P + λdry, λsat = λP− θlw 1− P θlw
ice λs λw , λdry = λ1−
s
P
θtotal, θlw, CLASS Empirical-
(1991) θice, λs, P modified
normalized
model
14 Desborough DP1998 λeff = (λsat − λdry)θtotal/P* + λdry, P* = min (0.45, 0.388 + 0.0125Itexture), P*, θtotal, CAPS, BASE Empirical-
and Pitman λsat = λθwlww/θtotalλθiceice/θtotalλ1−
s
P
*, λdry = 0.25 + 0.05[(9 − Itexture)/8] Itexture, λsat, modified
(1998) λdry Verseghy,
1991 model
15 Peters- PL1998 λeff = Ke(λsat − λdry) + λdry, λsat = λP− θlw 1− P θlw
ice λs λw , P, θlw, λs, ρb UM, ISBA, Empirical-
Lidard et al. λdry = (0.135ρb + 0.0647)/(ρs − 0.947ρb) ± 20% Noah, JULES modified

(1998) ⎨ log10 Sr + 1 Unfrozen, Sr ≥ 0.1 (a) normalized
Ke = 0 Unfrozen, Sr < 0.1 (b) model

Sr Frozen soils (c)
(continued on next page)

3
H. He et al. Earth-Science Reviews 211 (2020) 103419

Table 1 (continued )
No Source Abbrev. Expressions Parameters Land-surface Note
models host*
√̅̅̅̅̅̅̅̅̅̅( )
16 Shmakin SA1998 λeff = λwp + θtotal λfc − λwp θtotal, θice, SPONSOR Empirical-
(1998) λfc, λwp modified
normalized
model
17 Cox et al. CM1999 λeff = (λsat − λdry)θtotal/P + λdry, λsat = λ1− P θlw P− θlw
λw λice , λdry = λ1− P P
P, θ’lw, θice, JULES Empirical-
’ ’
s s λair
(1999) λs modified
normalized
model, new
Ke
18 Côté and CK2005 k1 ⋅Sr k1, χ , η, Sr DOS-TEM Empirical-
λeff = Ke(λsat − λdry) + λdry, Ke =
Konrad 1 + (k1 − 1)Sr modified

(2005) ⎨ λPw λ1− P normalized
s Unfrozen soil (a)
λsat = , λdry = χ 10− η⋅P model
⎩ λP− θlw λ1− P λθlw Frozen soil (b)
ice s w
19 Lawrence LS2008 λeff = Ke(λsat − λdry) + λdry fom, λs,m, λs, CLM Empirical-

and Slater ⎨ 0.7lg10 Sr + 1 Sr > 0.05, coarse soil, unfrozen (a) om modified
(2008) Ke = lg S + 1 Sr > 0.1, fine soil, unfrozen (b) normalized
⎩ 10 r
Sr frozen soils (c) model by
λdry = (1 − fom)λdry, m + fomλdry, om, λs = (1 − fom)λs, m + fomλs, om combining
organic soil
with mineral
soil
20 Dharssi et al. DI2009 λeff = Ke(λsat − λdry) + λdry, λs = λfsand
sand fsilt fclay
λsiltλclay λdry, P, θlw JULES, UM Empirical-

(2009) ⎨ 0.7lg10 Sr + 1 Sr > 0.05, coarse soil, unfrozen (a) modified
Ke = lg S + 1 Sr > 0.1, fine soil, unfrozen (b) normalized
⎩ 10 r
Sr frozen soils (c) model
λsat = λP− P
ice λw λsat, o/λw,λsat, o = 1.58 + 12.4(λdry − 0.25)
θlw θlw

21 Luo et al. LS2009 k2, χ , η, Sr, CoLM Empirical-


(2009b) P, fquartz modified
JO1975,
CK2005, and
λeff = FO1981

[( ]
⎧ k2 ⋅Sr )
fquartz 1− fquartz 1− P θw
⎪ λquartz λother λlw − χ 10− η⋅P + χ 10− η⋅P 5


⎪ 1 + (k 2 − 1)Sr T ≥ Tf , Sr > 10− (a)





⎪ [( ]


⎨ k2 ⋅Sr )
fquartz 1− fquartz 1− P θlw θice
λquartz λother λlw λice − χ 10− η⋅P + χ10− η⋅P 5
1 + (k 2 − 1)Sr T < Tf , Sr > 10− (b)










⎪ 5

⎩ Sr ≤ 10− (c)
χ10− η⋅P
22 Chadburn CS2015 λeff = Ke(λsat − λdry) + λdry, λs = λfsand sand fsilt fclay
λsiltλclay λdry, fom, JULES Empirical-

et al. (2015) ⎨ 0.7log10 Sr + 1 Sr > 0.05, coarse soil, unfrozen (a) θlw, θice, θsat modified
Ke = log10 Sr + 1 Sr > 0.1, fine soil, unfrozen (b) normalized

Sr frozen soils (c) model (based
(θ θ )/(θlw +θice )
λw(θlw θsat )/(θlw +θice ) λiceice sat on DI2009)
λdry = λ1− fom fom
dry, mλdry, om, λsat = λsat,0
λθwsat

⎨ 0.5
[ ( ) ]/[ ( )] λdry < 0.06 (a)
λsat,0 = 1 − 0.0134ln λdry − 0.745 − ln λdry 0.06 ≤ λdry ≤ 0.3 (b)

2.2 λdry > 0.3 (c)
STC schemes with λeff as a function of matric potential
{
23 McCumber MP1981 418.4exp[-pF − 2.7] pF ≤ 5.1 (a) pF ECMWF, Eta- Empirical
λeff =
and Pielke 0.1714 pF > 5.1 (b) LSS, ISBA, model,
(1981) LEAF-OLAM, thermal
NCEP, Noah, conductivity
OSU-CAPS, vs soil water
SVATS, retention
UVMM, WRF characteristic
(SWRC)
24 Kutchment KL1983 λeff = λ(θlw)(1 + θice) ψ , θlw, θice NCEP Extended
et al. (1983) MP1981

*Abbreviations
BASE: The Best Approximation of Surface Exchange land-surface model; BEST: The Bare Essentials of Surface Transfer; CABLE: The Community Atmosphere Biosphere
Land Exchange model; CESM: Community Earth System Model; CLASS: Canadian Land Surface Scheme; CLM: Community Land Model, CLM1~4; CoLM: The Common
Land Model; COUPModel: Coupled heat and mass transfer model for the soil-plant-atmosphere system; CAPS: Coupled Atmosphere Plant Snow model; DOS-TEM
Terrestrial Ecosystem Model; ECMWF: European Centre for Medium-Range Weather Forecasts; Eta-LSS: Land-surface Scheme of the NCEP Meso-scale Eta Model;
GCM: Global Climate Model/Global Circulation Model; HDS-SPAC: A Hybrid Dual-Source Scheme Based Soil-Plant-Atmosphere Continuum Model; HYDRUS: Software
package for simulating water, heat, and solute movement in variably saturated media; ISBA: Interaction Soil Biosphere Atmosphere; ISBA-MEB: The Interactions
between Soil-Biosphere-Atmosphere (ISBA) land-surface model and Multi-Energy Balance (MEB); JULES: Joint UK Land Environment Simulator land-surface model;
LEAF: The Land Ecosystem-Atmosphere Feedback model; LPRM: Land parameter retrieval model; LSM: Land-surface model; MACRO: A model of water movement and

4
H. He et al. Earth-Science Reviews 211 (2020) 103419

studies or papers with three or more authors, the initials of the first from LSM to LSM; the original classification methods used in the LSM
author are used, for example the CG1985 and DI2009 are abbreviated were retained in the analysis of schemes in this study. More details and
for Campbell (1985) and Dharssi et al. (2009), respectively. For papers applicability of these STC schemes are discussed in the following and can
with two authors, the first letters of surnames are used, for example be found in Table 1 and/or in the supplementary file.
CS1981 and CH1987 are abbreviated for Camillo and Schmugge (1981) It is noteworthy that the 24 STC schemes consist of historical, semi-
and Chung and Horton (1987), respectively. current, and current STC schemes, because the inclusion of STC schemes
The 24 STC schemes were divided into four categories: (1) the in LSMs, SVATs, and hydrological models evolved with time. All 24 STC
theoretically/physically based STC scheme proposed by de Vries (1963). schemes were assessed to inform readers on the advantages, limitations,
The DV1963 (de Vries, 1963) assumes that the soil particles and air- and applications. In addition, although the STC schemes were originally
filled pores are ellipsoids dispersed in the continuous water host/back­ developed for various specific applications or soils, regions, or climates,
ground. de Vries (1963) takes a weighted average of the thermal con­ inclusion of them into LSMs usually entails application over a broader
ductivity of each soil phase with weighting factors accounting for set of environmental conditions. Therefore, in this work, these 24 STC
particle shape effects. Four modified forms of this scheme used in LSMs schemes will be systematically evaluated with a dataset covering soils of
were compared; (2) eight linear or nonlinear regression STC schemes; a wide range of properties, from dry to saturated, from deeply frozen
(3) thirteen STC schemes based on the normalized concept that in­ soils (e.g., -25◦ C) to unfrozen soils (e.g., 1~50 ◦ C), from fine-textured
terpolates λeff at any water content or degree of saturation from the soils to coarse sands, and from zero to 22% of organic matter content.
thermal conductivity of soils at a relative saturation of zero and full A detailed description of the dataset compiled for the evaluation of STC
saturation (He et al., 2020d; Wang et al., 2020b); and (4) two STC schemes is presented below.
schemes that relate λeff to soil matric potential (Kutchment et al., 1983;
McCumber and Pielke, 1981). The KL1983 is an extension of the
2.2. Data compilation and process
MP1981 by accounting for ice content.
Among the 24 STC schemes, only one (i.e., DV1963) is physics-based
2.2.1. Dataset for mineral soils
with the remaining 23 being empirically based. STC schemes vary in
Measured thermal conductivity data of different soil types under
both their applicability and complexity. For instance, the physically
different moisture contents are required for model evaluation. Published
based DV1963 scheme has more parameters with physical meaning, but
thermal conductivity data were strictly screened following four criteria:
these parameters are difficult to determine, even for well-characterized
(1) The λeff were measured by a transient-state method (e.g., thermal
soils. Moreover, some STC schemes focus mainly on mineral soils, while
probe, thermal conductivity probe, transient heat-flow method, single/
the others account for the influence of organic matter (Chadburn et al.,
dual-probe heat pulse method), because transient methods are less likely
2015; Hubrechts, 1998; Lawrence and Slater, 2008; Zeng and Su, 2013)
subject to errors compared to steady-state methods (de Vries, 1952; He
or were specifically developed for peat soils (de Vries, 1975). Addi­
et al., 2018a; He et al., 2015); (2) measurements of λeff at both positive
tionally, some STC schemes were developed for unfrozen soils (Camp­
and negative temperatures were available; (3) information on soil
bell, 1985; Cass et al., 1984; Chung and Horton, 1987; McCumber and
physical properties such as water content, texture (i.e., content of sand,
Pielke, 1981), while the others can be applied to both unfrozen and
silt and clay), organic matter content, soil particle density, bulk density,
frozen soils but were only tested for a limited range of soil textures
or porosity were available; and (4) detailed description of, and repeat­
(Becker et al., 1992; Côté and Konrad, 2005; Johansen, 1975; Kersten,
able, experimental design and setup were used with strict quality control
1949). It is also noted that threshold of soil texture classification (e.g.,
of the data. This screening process resulted in a total of 439 measure­
coarse-, medium- and fine-textured soils) for STC schemes may vary
ments on 16 soils from five studies (Gori and Corasaniti, 2003; Lu et al.,

solute transport in macro porous soils; MOSES: Modular Observation Solutions for Earth Systems land-surface model/Met Office Surface Exchange Scheme/System;
NCEP: National centers Environmental Prediction; Noah: The community Noah Land-Surface Model; OSU-CAPS: The Oregon State University Coupled Atmosphere-
Plant-Soil model; PRM: Purdue Regional Model; SHAW: The Simultaneous Heat and Water Model; SiB2: Simple Biosphere Model2; SLI: Soil-Litter-Iso model; SOIL:
Simulating Model for Soil Water and Heat Conditions; SPONSOR: Semi-distributed ParameterizatiON Scheme of the ORography-induced hydrology; ST&TP-ES: Soil
thermal & transport properties-expert system; STEMMUS: Soil water flow, Coupled transfer, Heat transfer, Evaporation, Arid, Semi-Arid; SVATS: Soil-vegetation-
atmosphere transfer schemes; UM: Unified model; UNSAT-H: Unsaturated Soil Water and Heat Flow Model; UVMM: University of Virginia Mesoscale Model; WRF:
The Weather Research and Forecasting Model.
Note: HYDRUS, SHAW, SOIL, SVAT, and UNSAT-H belong to hydrological or soil-vegetation-atmosphere transfer models.
Symbols:
η: Parameter of the Côté and Konrad (2005) scheme; λeff:Effective soil thermal conductivity; λ0: Thermal conductivity of 0th phase of soil components, W m-1 ◦ C-1; λquartz:
Thermal conductivity of quartz, 7.7 W m-1 ◦ C-1); λother: Thermal conductivity of other mineral, λother = 2.0 W m-1 ◦ C-1 for fquartz >20% otherwise λother = 3 W m-1 K-1 at
room temperature); λdry: Thermal conductivity of dry soil as a whole, W m-1 ◦ C-1; λdry,m: Thermal conductivity of dry mineral soil, W m-1 ◦ C-1; λdry,om: Thermal con­
ductivity of dry organic matter, W m-1 ◦ C-1; λsat: Thermal conductivity of saturated soil, W m-1 ◦ C-1; λw: Thermal conductivity of water, 0.56 W m-1 ◦ C-1 at room
temperature; λice: Thermal conductivity of ice, 2.24 W m-1 ◦ C-1; λair: Thermal conductivity of air, 0.025 W m-1 ◦ C-1; λlw: Thermal conductivity of unfrozen or liquid water,
W m-1 ◦ C-1; λs: Thermal conductivity of soil solids as a whole, W m-1 ◦ C-1; λs,m: Thermal conductivity of mineral soil solids, W m-1 ◦ C-1; λs,om: Thermal conductivity of
organic matter, W m-1 ◦ C-1; λsand: Thermal conductivity of sand, W m-1 ◦ C-1; λslit: Thermal conductivity of slit, W m-1 ◦ C-1; λclay: Thermal conductivity of clay, W m-1 ◦ C-1;
λfc: Thermal conductivity at field capacity, 2.5 W m-1 ◦ C-1; λwp: Thermal conductivity at wilting point, 0.25 W m-1 ◦ C-1; λT+: Thermal conductivity of soil at unfrozen
state, W m-1 ◦ C-1; λT-: Thermal conductivity of soil at frozen state, W m-1 ◦ C-1; ρb: Dry bulk density, g cm-3; ρs: Density of soil solid, g cm-3; ϕi: Volumetric fraction of ith
phase of soil, cm3 cm-3; ϕair: Volumetric fraction of air content, cm3 cm-3; ϕs: Volumetric fraction of mineral soil solid, cm3 cm-3; ϕw or θv: Volumetric water content, cm3
cm-3; θg: Gravimetric water content, g g-1; θice: Ice content, cm3 cm-3; θinit: Initial water content prior to soil freezing, cm3 cm-3; θlw: Volumetric liquid water content or
unfrozen water content, cm3 cm-3; θ’lw, The saturation concentration of liquid water for the current liquid water to ice mass ratio; θsat: Saturated water content, cm3 cm-
3
; θtotal: Total volumetric water and ice, cm3 cm-3; θv: Volumetric water content, cm3 cm-3; χ: Parameter of the Côté and Konrad (2005) scheme; ψ : Matric potential, cm
H2O; a,b,c: Parameters of the the Chung and Horton (1987) scheme; a1, a2, a3, a4: Parameters of the Becker et al. (1992) scheme; fquartz: Quartz content of the total solids
content at mass basis, %; fsand: Sand content, %; fslit: Silt content, %; fclay: Clay content, %; fom: Organic matter content, %; gk: depolarization factors (particle shape
factors); h1,h2: Parameters of the de Vries (1975) scheme; i: Subscript, ith phase of soil system; k: Subscript, kth depolarization factor in the de Vries (1963) scheme; k1:
Parameter of the Côté and Konrad (2005) scheme; k2: Parameter of the Luo et al. (2009b) scheme; pF: matric potential, pF=log10 (ψ ); A1, B1, C1,D1,E1: Parameters of the
Cass et al. (1984) scheme; A2, B2, C2,D2,E2: Parameters of the Campbell (1985) scheme; F: Weighting factors; Itexture: Soil texture index in the Desborough and Pitman
(1998) scheme; N: The number of dispersed phases in the de Vries (1963) scheme; T: Measured temperature, ◦ C; Tf: Freezing depression point, ◦ C; ΔT: Temperature
difference, ◦ C; UC1, UC2, UC3: Unit conversion factor; Ke: Kersten number; Sr: Degree of saturation; P: Soil porosity; P*: Assumed soil porosity in the Desborough and
Pitman (1998) scheme.

5
H. He et al. Earth-Science Reviews 211 (2020) 103419

2018; Penner et al., 1975; Suzuki et al., 2002; Zhang et al., 2018). The 2.3. Community Land Model (CLM) simulation
selected physical properties of the soils are listed in Table 2. For a more
detailed description of the experimental procedures and setup, the 2.3.1. Description of the CLM
reader is refereed to respective studies. The CLM is a critical component of the Community Earth System
Because λeff measurements of organic soils or peat are rare, two soils Model developed by the National Center for Atmospheric Research and
from Suzuki et al. (2002) containing organic matter of 13 and 22%, has been extensively evaluated and validated (Bonan et al., 2013;
respectively, were used to test the effects of organic matter on STC Brunke et al., 2016; Dickinson et al., 2006; Li et al., 2011; Makushev
schemes. et al., 2016; Okalebo et al., 2016; Wang and Yang, 2018). It has also
been widely applied at regional or global scales to investigate a wide
2.2.2. Dataset containing matric potential range of soil physical, hydrological, and biological processes, land-cover
MP1981 and KL1983 are the only two STC schemes that require change, and climate response (Cao et al., 2009; Du et al., 2016; Ghimire
input of matric potential, and therefore the datasets of Smits et al. et al., 2016; Han et al., 2014; Wu et al., 2014; Xie et al., 2019). The CLM
(2010) and Wallen et al. (2016) were selected. These datasets contain consists of a single-layer vegetation scheme with up to 15 plant func­
soil water content, thermal conductivity, and matric potential simulta­ tional types (PFTs) to better resolve the surface heterogeneity of com­
neously measured in the same soils with a Tempe cell or hanging water plex vegetated environments (Bonan et al., 2002), a snow scheme of up
column setup. The datasets consist of eight sands with measured λeff as a to 5 layers (Oleson et al., 2013; Swenson and Lawrence, 2012), and a soil
function of matric potential (Table 2). scheme of 10 layers. The spatial distribution and seasonal climatology of
those PFTs for CLM are derived from the moderate resolution imaging
2.2.3. Water content spectrometer (MODIS) satellite land-surface data products that are
Water content (θ) can be easily measured for unfrozen conditions described in the following section (Lawrence and Chase, 2007). The
and is provided typically in the literature. However, unfrozen liquid CLM version 4.5 was used to assess the sensitivity of soil/land surface
water content (θlw, m3 m-3 or cm3 cm-3) at subfreezing temperatures is temperature (LST) to the STC schemes in this study.
not generally available, because measurement of θlw requires special
equipment such as nuclear magnetic resonance (Oliphant and Tice, 2.3.2. Site descriptions and the MODIS land surface temperature data
1982; Smith and Tice, 1988; Watanabe and Wake, 2009), gas dilatom­ Three sites with different soil textures were selected to demonstrate
etry (Spaans and Baker, 1995), or differential scanning calorimetry the effects of STC schemes on the estimated LST. The first site (34.55◦ N,
(Oliphant and Tice, 1982). Data containing both λeff and θlw are rare. 99.35◦ E) is a clay dominated soil with 51% clay and 20% sand and the
Hence, we estimated θlw by the Saxton et al. (1986) method: land cover is composed of 17% bare and 83% C3 arctic grass. The second
site (40.50◦ N, 92.50◦ E) is a loamy soil with 51% sand and 13% clay and
ψ = Aθblw , where A = ψ e θblw (1)
the land cover is composed of 67% bare, 15% C3 non-arctic grass and
where A and b are empirical coefficients. 18% C4 grass. The third site (40.45◦ N, 102.05◦ E) is sand with 90% sand
Rearranging Eq. (1) gives and 5% clay and the land cover is composed of 90% bare, 5% C3 non-
arctic grass and 5% C4 grass.
θlw = exp[ln(ψ /A)/b ] (2) The accuracy and applications of MODIS LST products have been
where ψ is matric potential that is calculated with the modified well documented (Droogers et al., 2000; Salomonson et al., 1987; Wan,
Clausius-Clapeyron equation (Kojima et al., 2018; Kurylyk and Wata­ 2008; Wan, 2014; Wang et al., 2008). The daily MODIS LST data of the
nabe, 2013) abovementioned three sites over a period of 7 years ranging from 2003
( ) to 2010 were used as observational data of LST to assess the LST sim­
Lρ T ulations with different STC schemes. The differences between CLM-
ψ= f w (3)
g 273.15 estimated LST and MODIS-observed LST was calculated based on a
where Lf is latent heat of fusion (334 J ), ρw is density of water (g cm- grid size of 0.05◦ ×0.05◦ or 5 km ×5 km.
3
), T is temperature ( C), and g is gravitational acceleration (9.8 m s− 2).

Equation (3) assumes that ice and osmotic pressure are zero so that soils 2.4. Performance metrics
must be unsaturated and solute-free (Fuchs et al., 1978).
Empirical coefficients A (Mpa) and b are related to soil texture The performance of different thermal conductivity models was
(Saxton et al., 1986) compared with measured and predicted λeff values using a 1:1 plot
( ) (Piñeiro et al., 2008), ±10% and ±30% lines off the 1:1 line and three
2
A = 0.1exp − 4.396 − 7.15fclay − 4.880fsand 2
− 4.285fsand fclay (4)
additional statistical indices:
(1) root mean square error (RMSE)
2
b = − 3.140 − 22.2fclay 2
− 3.484fsand fclay (5)
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√∑
where the value of 0.1 was included to convert unit bar to MPa and f √n ( )2
√ X − Xm
√i=1 p
is the mass fraction of the indicated particle-size class. RMSE = (6)
We calculated the ice content of frozen soils based on the mass bal­ n
ance θice = 1.09(θ-θ lw). It should be noted that sometimes the Saxton (2) the average deviations (AD)
et al. (1986) method, which determines θlw from soil temperature and
texture, gives negative estimates of θlw. Because some STC schemes (e.g., 1∑ n
( )
AD = Xp − Xm (7)
KM1949, LV1981, MP1981, BB1992, SA1998, and CK2005) do not n i=1
require the input of θlw, they thus are not affected by θlw. For STC
(3) the Nash-Sutcliffe Efficiency (NSE)
schemes of DV1975, CS1981, CG1985, CH1987, and HL1998 that
require the input of total water content, the sum of the liquid water and n (
∑ )2
Xp − Xm
ice content was used (θ = θlw+ θice).
NSE = 1-∑
n (
i− 1
)2 (8)
Xp − Xmean− m
i− 1

Xp values are predicted by different thermal conductivity models, Xm


are measured values, Xmean-m are mean measured values, n is the number

6
Table 2
Selected physical properties of soils used to evaluate the 24 soil thermal conductivity (STC) schemes.

H. He et al.
Literature Abbrev. Soil name No. of Meas. Temperature Texture Organic Particle Bulk Porosity Water content Degree of
source matter density density (P) (θinit) saturation (Sr)
Unfrozen Frozen Gravel Sand Silt Clay
(ρs (ρb)

C >2 2~0.05mm 0.05~0.002 <0.0002 -g cm-3- - cm3 cm-3 -
mm

Dataset 1
Gori and GC2003 Olivine 11 3 -17.0~50 - 1.00 0.00 0.00 - 2.65 1.67 0.37 0.00~0.36 0.00~1.00
Corasaniti
(2003)
Lu et al. (2018) LY2018 Aeolian 40 46 -20~20 - 0.88 0.12 0.00 - 2.67 1.40~1.65 0.38~0.48 0.04~0.25 0.09~0.65
sand
Penner et al. PE1975 Soil #1 3 6 5, -5, -15 - 0.32 0.49 0.19 - 2.75 1.66~1.89 0.32~0.41 0.10~0.26 0.26~0.83
(1975)
Penner et al. PE1975 Soil #2 3 6 5, -5, -15 - 0.64 0.21 0.15 - 2.73 1.73~1.93 0.30, 0.36 0.10~0.25 0.29~0.85
(1975)
Penner et al. PE1975 Soil #3 5 6 5, -5, -15 - 0.90 0.06 0.04 - 2.74 1.71~1.97 0.28~0.37 0.07~0.20 0.18~0.72
(1975)
Penner et al. PE1975 Soil #4 3 6 5, -5, -15 - 0.22 0.60 0.19 - 2.75 1.61~1.79 0.35~0.41 0.10~0.30 0.23~0.83
(1975)
Penner et al. PE1975 Soil #5 3 6 5, -5, -15 - 0.06 0.71 0.23 - 2.75 1.70~1.86 0.32~0.39 0.08~0.25 0.21~0.77
(1975)
Penner et al. PE1975 Soil #6 3 6 5, -5, -15 0.05 0.06 0.65 0.24 - 2.74 1.64~1.83 0.33~0.40 0.11~0.26 0.27~0.79
(1975)
Penner et al. PE1975 Soil #7 3 6 5, -5, -15 0.17 0.27 0.31 0.25 - 2.75 1.96~2.07 0.25~0.28 0.09~0.26 0.33~0.92
(1975)
Penner et al. PE1975 Soil #8 3 6 5, -5, -15 - 0.47 0.00 0.53 - 2.73 1.39~1.57 0.43~0.49 0.08~0.30 0.16~0.70
(1975)
Penner et al. PE1975 Soil #9 3 6 5, -5, -15 - 0.00 0.68 0.32 - 2.66 1.43~1.64 0.40~0.45 0.08~0.29 0.19~0.72
(1975)
7

Penner et al. PE1975 Soil #10 4 8 5, -5, -15, -25 - 0.00 0.44 0.56 - 2.74 1.40~1.63 0.41~0.49 0.10~0.33 0.20~0.81
(1975)
Suzuki et al. SS2002 Hokudai 7 35 -10~1 - 0.45 0.33 0.22 0.13 2.55 1.02, 1.22 0.6, 0.52 0.23~0.58 0.44~0.97
(2002) soil*
Suzuki et al. SS2002 Kuroboku 3 15 -10~1 - 0.51 0.27 0.22 0.22 2.37 0.71 0.7 0.48~0.66 0.69~0.94
(2002) soil*
Suzuki et al. SS2002 Sand 3 15 1 - 0.97 0.01 0.02 - 2.80 1.48 0.47 0.14~0.43 0.30~0.91
(2002)
Zhang et al. ZM2018 Silty clay 118 48 -19~18 - 0.39 0.54 0.07 - 2.65 1.50~1.75 0.34~0.43 0.20~0.35 0.45~1.00
(2018)
Dataset 2
Smits et al. SK2010 Sand20/30 743 - Room temp - 1.00 0.00 0.00 - 2.65 1.78 0.33 0.03~0.33 0.09~1.00
(2010)
Wallen et al. WB2016 C10F00** 266 - Room temp - 1.00 0.00 0.00 - 2.65 1.81 0.32 0.00~0.32 0.00~1.00
(2016)
Wallen et al. WB2016 C09F01** 353 - Room temp - 1.00 0.00 0.00 - 2.65 1.96 0.26 0.01~0.26 0.04~1.00
(2016)

Earth-Science Reviews 211 (2020) 103419


Wallen et al. WB2016 C08F02** 359 - Room temp - 1.00 0.00 0.00 - 2.65 2.09 0.21 0.02~0.21 0.10~1.00
(2016)
Wallen et al. WB2016 C07F03** 328 - Room temp - 1.00 0.00 0.00 - 2.65 2.15 0.19 0.00~0.19 0.00~1.00
(2016)
Wallen et al. WB2016 C05F05** 328 - Room temp - 1.00 0.00 0.00 - 2.65 2.02 0.24 0.01~0.24 0.04~1.00
(2016)
Wallen et al. WB2016 C02F08** 281 - Room temp - 1.00 0.00 0.00 - 2.65 1.86 0.30 0.01~0.30 0.03~1.00
(2016)
Wallen et al. WB2016 C00F10** 382 - Room temp - 1.00 0.00 0.00 - 2.65 1.76 0.34 0.01~0.34 0.03~1.00
(2016)
*
Hokudai soil and Kuroboku soil are volcanic ashes
**
mixture of Accusand (Coarse-gained, mean diameter 1.04 mm) and Ottawa sand (fine-grained, mean diameter 0.12 mm) at seven coarse to fine (C/F) ratio, for example, C09F01 indicates ratio of 9:1 for coarse-grained
Accusand to fine-grained Ottawa sand
H. He et al. Earth-Science Reviews 211 (2020) 103419

of measured values and p is the number of parameters in the model. and air) with spherical components, gk=[1/3, 1/3, 1/3] that is depo­
RMSE describes the deviation between the predicted values and the larization factors (particle shape factors); (2) DV1963II, soil is a four-
measured values. AD represents the deviation from the predicted value phase system (i.e., soil solids, liquid water, ice, and air) with plate-
and the measured value. NSE determines the relative magnitude of the shaped components, gk=[0, 0, 1]; (3) DV1963III, soil is a four-phase
residual variance between measured and predicted values (Nash and system (i.e., soil solids, liquid water, ice, and air) with needle-shaped
Sutcliffe, 1970). A smaller difference between the predicted and components, gk=[1/2, 1/2, 0]; (4) DV1963IV, soil is a five-phase sys­
measured values is expected when NSE approaches 1. tem (i.e., liquid water, air, quartz, other mineral other than quartz, and
other organic materials) as used in STEMMUS (Zeng and Su, 2013).
3. Results and discussion Fig. 1 and Table 3 show that the performances of all four forms of the
DV1963 vary significantly with data source and state (frozen or unfro­
This section is organized according to the four categories of STC zen). For instance, DV1963I, III, and IV overestimated data of GC2003
schemes as stated in section 2.1. More discussion on the uncertainty of and SS2002. DV1963II generally underestimated PE1975 and LY.
evaluating the STC schemes is presented in section 3.5. The effects of DV1963III overestimated while the remaining three schemes under­
STC schemes on land surface temperature is demonstrated in Section estimated data of LY2018. DV1963IV gave much lower estimates of
3.6. unfrozen data of LY2018 compared to frozen data of LY2018, while the
other three schemes generally showed similar underestimates or over­
3.1. Physically based STC scheme and its modified forms estimates of unfrozen or frozen data.
The DV1963 scheme, which has been used to predict λeff in CLMs
Four cases of the DV1963 STC (de Vries, 1963), in which soil water is (Tian et al., 2016; Yan et al., 2019), performed unsatisfactorily relative
assumed to be the continuous medium (i.e., 0th phase), were tested: (1) to other schemes (Table 3). The reason may be that many of the required
DV1963I, soil is a four-phase system (i.e., soil solids, liquid water, ice, parameters, though having physical meaning, are difficult to determine;

Fig. 1. Comparison of measured λeff and predicted λeff with physically based soil thermal conductivity (STC) scheme of de Vries (1963) with different parame­
terizations: (a) DV1963I with spherical particles (gk=[1/3, 1/3, 1/3]), (b) DV1963II with disc-shaped particles (gk=[0, 0, 1]), (c) DV1963III with needle-shaped
particles (gk=[0.5, 0.5, 0]), and (d) DV1963IV with parameters used in STEMMUS (Zeng and Su, 2013). Closed (T+) and open (T-) symbols indicate unfrozen
and frozen λeff, respectively. GC2003 (Gori and Corasaniti, 2003), LY2018 (Lu et al., 2018), PE1975 (Penner et al., 1975), SS2002 (Suzuki et al., 2002) and ZM2018
(Zhang et al., 2018) are the five data sources.

8
H. He et al. Earth-Science Reviews 211 (2020) 103419

Table 3
Performance metrics of the 24 soil thermal conductivity (STC) schemes compared with experimental data. Three scenarios of quartz content were assumed: (1) quartz
content equals half of sand content, fquartz =0.5 fsand; (2) quartz content equals sand content, fquartz = fsand; and (3) quartz content equals half of the sum of sand and silt
content, fquartz=0.5(fsand+fsilt).*, **
STC schemes fquartz =0.5 fsand fquartz = fsand fquartz=0.5(fsand+fsilt)

RMSE AD NSE RMSE AD NSE RMSE AD NSE

W m-1 ◦ C-1 - W m-1 ◦ C-1 - W m-1 ◦ C-1 -

Physically-based STC scheme


DV1963a 0.62 -0.35 -0.16 0.62 -0.41 -0.16 0.67 -0.36 -0.33
DV1963b 0.84 +0.01 -1.13 0.81 -0.09 -0.95 0.91 -0.00 -1.51
DV1963c 0.73 -0.57 -0.62 0.85 -0.73 -1.14 0.78 -0.55 -0.81
DV1963d 0.79 -0.24 -0.85 0.99 -0.44 -1.94 0.93 -0.38 -1.59
Linear/non-linear regression STC schemes
KM1949 0.75 +0.07 -0.69 - - - - - -
DV1975 1.34 +1.21 -4.42 - - - - - -
CS1981 0.90 +0.72 -1.43 - - - - - -
CS1984 0.61 +0.30 -0.13 - - - - - -
CG1985 0.89 +0.68 -1.36 - - - - - -
CH1987 0.65 +0.16 -0.27 - - - - - -
BB1992 0.61 -0.05 -0.10 - - - - - -
HL1998 0.55 -0.17 0.09 - - - - - -
STC schemes based on Normalized concept
JO1975 0.60 -0.29 -0.09 0.61 -0.37 -0.12 0.66 -0.29 -0.32
FO1981 1.41 -1.03 -4.99 - - - - - -
LV1981 0.83 +0.02 -1.04 - - - - - -
VD1991 1.14 -1.04 -2.88 1.30 -1.20 -4.09 1.12 -1.01 -2.79
DP1998 1.19 -0.46 -3.26 1.52 -0.64 -5.94 1.31 -0.50 -4.11
PL1998 0.77 -0.41 -0.79 0.79 -0.50 -0.86 0.83 -0.42 -1.07
SA1998 0.60 -0.05 -0.08 - - - - - -
CM1999 0.67 -0.33 -0.35 0.69 -0.42 -0.44 0.73 -0.34 -0.62
CK2005 0.58 -0.26 -0.01 0.61 -0.35 -0.11 0.64 -0.26 -0.23
LS2008 0.59 -0.28 -0.06 0.60 -0.36 -0.06 0.65 -0.28 -0.27
DI2009 1.07 -0.76 -2.45 - - - - - -
LS2009 0.82 -0.65 -1.03 0.98 -0.82 -1.85 0.86 -0.63 -1.24
CS2015 0.58 -0.12 -0.02 - - - - - -
*
DV1963a~d are special scenarios when gk=[1/3, 1/3, 1/3], gk=[0, 0, 1], gk=[0.5, 0.5, 0], and approach used in STEMMUS, respectively. More details can be found
in the supplementary file.
**
empty values ("-") of RMSE, AD and NSE for fquartz = fsand and fquartz=0.5(fsand+fsilt) indicate STC schemes were not affected by the quartz content.

thereby, in practice, the scheme is often simplified, resulting in a (Fig. 2c~e and Table 3). This could be because these three STC schemes
weakened capability and applicability. In contrast, the empirical STC were developed for specific soils at unfrozen states. For example, the
schemes (in section 3.2 through 3.4 below) require fewer parameters, CS1984 was developed based on two soils (i.e., Lysimeter sand and
and can be readily incorporated into numerical simulation programs (He Portneuf silt loam) from south/central Washington state and southern
et al., 2017). This could be among the critical attributes as to why Idaho, USA (Cass et al., 1984), while the CG1985 was for five soils (i.e.,
empirical STC schemes are preferred by LSMs. Quincy fine sand, Ritzville silt loam, Walla Walla silt loam, Palouse silt
loam, and Naff silt loam) from eastern Washington state, USA (Camp­
bell, 1985; McInnes, 1981). It should be noted that CG1985 was first
3.2. Linear or non-linear regression STC schemes developed by McInnes (1981) and Campbell (1985): Both established
pedotransfer functions to estimate the parameters. McInnes (1981) and
Among the eight linear/non-linear regression STC schemes, the Campbell (1985) are referred to interchangeably in LSMs and indicate
HL1998 (Hubrechts, 1998) performed best with a RMSE = 0.55 W m-1 the same STC scheme. While the CS1984 has the similar equation as
◦ -1
C , AD = -0.17 W m-1 ◦ C-1 and NSE = 0.09 (Fig. 2h and Table 3); when CG1985, the water content is replaced with degree of saturation,
applied to the unfrozen data of LY2018 only (Lu et al., 2018), the per­ resulting in a different set of parameters as recommended by Cass et al.
foramnce was even better. Following HL1998, are the BB1992 (Becker (1984).
et al., 1992) (Fig. 2g and Table 3) and CS1984 (Cass et al., 1984) (Fig. 2d CH1987 (Chung and Horton, 1987), developed using soil samples
and Table 3). This agrees with previous studies that showed BB1992 ranging from residual to full saturation levels, is not valid for soils with
gave better estimates than other STC schemes (He et al., 2021; He et al., saturations below residual (Fig. 2f). It also has problems in expressing
2020c). The DV1975 (de Vries, 1975) had the worst performance and the λeff(θ) relationship for soils over a wide range of textures or bulk
significantly underestimated the λeff of all soils (Fig. 2b and Table 3). But densities (Lu et al., 2014).
this is not surprising as DV1975 was developed for predicting λeff of peat There are other issues associated with linear/non-linear regression
soils with significantly different thermal properties compared to mineral STC schemes. For instance, the CG1985 cannot be applied to soils with
soils (Brovka and Rovdan, 1999; Pernitsky et al., 2016). zero or small clay content. Otherwise numeric errors (e.g., division by
The KM1949 (Kersten, 1949) generally underestimated frozen λeff of zero) arise. Although CG1985 has been extended to fine-textured soils
all soils (Fig. 2a and Table 3), particularly for the coarse-textured (Tien et al., 2005) and frozen soils (Hansson et al., 2004), these modi­
LY2018 and volcanic SS2002 soils (Suzuki et al., 2002). However, its fications frequently give unreasonable estimates of λeff (data not shown).
overall performance is better as compared to the unfrozen data of CS1984 was developed for sand and silt loam soils (Cass et al., 1984),
LY2018 (Fig. 2a and Table 3), and overestimated both frozen and un­ but was subsequently extended to all soil types, both frozen and unfro­
frozen λeff for the fine- textured ZM2018. zen (Fayer, 2000). However, we found that only the set of parameters for
CS1981 (Camillo and Schmugge, 1981), CS1984 (Cass et al., 1984) sand gave numerically correct results, which might be the reason why
and CG1985 (Campbell, 1985) generally underestimated λeff of all soils

9
H. He et al. Earth-Science Reviews 211 (2020) 103419

Fig. 2. Comparison of measured λeff and predicted λeff with linear/non-


linear regression based soil thermal conductivity (STC) schemes: (a)
KM1949 (Kersten, 1949), (b) DV1975 (de Vries, 1975), (c)CS1981
(Camillo and Schmugge, 1981), (d) CS1984 (Cass et al., 1984), (e)
CG1985 (Campbell, 1985), (f) CH1987 (Chung and Horton, 1987), (g)
BB1992 (Becker et al., 1992), and (h) HL1998 (Hubrechts, 1998).
Closed (T+) and open (T-) symbols indicate unfrozen and frozen λeff,
respectively. GC2003 (Gori and Corasaniti, 2003), LY2018 (Lu et al.,
2018), PE1975 (Penner et al., 1975), SS2002 (Suzuki et al., 2002) and
ZM2018 (Zhang et al., 2018) are five data sources.

only one set of parameters was used in the UNSAT-H program (Fayer, soils (λsat) as the lower and upper limits of λeff, respectively. The λeff at
2000). Note that these problems are not unique to CS1984 and CS 1985, intermediate water contents was interpolated with the Kersten function
because similar numeric error (e.g., division by zero) was encountered (Ke). The JO1975 scheme can be used for both unfrozen and frozen soils,
in dry soils for the HL1998. however the λsat and Ke parameters differ between frozen and unfrozen
soils (He et al., 2020d; He et al., 2017; Johansen, 1975; Yan et al., 2019).
3.3. Normalized STC schemes The complexity of the 12 modified forms of the JO1975 varies,
mainly due to the multiple ways used to calculate Ke, λs (thermal con­
The JO1975 (Johansen, 1975) scheme is expressed asλeff = Ke(λsat − ductivity of soil solid), λdry, and λsat (He et al., 2020d). More details about
λdry) + λdry, with thermal conductivity of dry soils (λdry) and saturated these STC schemes are presented in the supplementary file. At present,

10
H. He et al. Earth-Science Reviews 211 (2020) 103419

the JO1975 and its modified versions are the primary STC schemes used and NSE ≈ 0 (Table 3). The FO1981 (Farouki, 1981) ranked last with
in LSMs, hydrological and SVAT models. To avoid too many models in RMSE = 1.41 W m-1 ◦ C-1, AD = -1.03 W m-1 ◦ C-1 and NSE = -4.99 (Fig. 3b
one figure, the 13 normalized STC schemes were arbitrarily divided into and Table 3). Most of the 13 normalized schemes could better predict
two groups based on whether they were published before or after 2000. data from PE1975 and LY2018, while they generally overestimated
Among the 13 normalized STC schemes, the JO1975 (Fig. 3a), unfrozen and frozen data of SS2012 and ZM2018.
SA1998 (Shmakin, 1998) (Fig. 3g), CK2005 (Côté and Konrad, 2005) The scheme of FO1981 differs from JO1975 in two aspects: (1)
(Fig. 4b), LS2008 (Lawrence and Slater, 2008) (Fig. 4c), and CS2015 FO1981 calculates soil solid thermal conductivity based on the fraction
(Chadburn et al., 2015) (Fig. 4f) ranked as the top performing schemes of sand and clay, (λs = [8.8fsand + 2.92fclay]/[fsand + fclay] vs λs =
of this categories, with RMSE ≈ 0.6 W m-1 ◦ C-1, AD ≈ -0.17 W m-1 ◦ C-1 λfquartz
quartz
λ1− fquartz
other ), where fquartz is mass fraction of quartz content. The

Fig. 3. Comparison of measured λeff and predicted λeff with soil ther­
mal conductivity (STC) schemes based on the normalized concept: (a)
JO1975 (Johansen, 1975), (b) FO1981 (Farouki, 1981), (c) LV1981
(Lunardini, 1981), (d) VD1991 (Verseghy, 1991), (e) DP1998 (Des­
borough and Pitman, 1998), (f) PL1998 (Peters-Lidard et al., 1998), (g)
SA1998 (Shmakin, 1998), and (h) CM1999 (Cox et al., 1999). Closed
(T+) and open (T-) symbols indicate unfrozen and frozen λeff, respec­
tively. GC2003 (Gori and Corasaniti, 2003), LY2018 (Lu et al., 2018),
PE1975 (Penner et al., 1975), SS2002 (Suzuki et al., 2002) and ZM2018
(Zhang et al., 2018) are five data sources.

11
H. He et al. Earth-Science Reviews 211 (2020) 103419

Fig. 4. Comparison of measured λeff and predicted λeff with soil thermal conductivity (STC) schemes based on the normalized concept: (a) JO1975 (Johansen, 1975),
(b) CK2005 (Côté and Konrad, 2005), (c) LS2008 (Lawrence and Slater, 2008), (d) DI2009 (Dharssi et al., 2009), (e) LS2009 (Luo et al., 2009b), and (f) CS2015
(Chadburn et al., 2015). Closed (T+) and open (T-) symbols indicate unfrozen and frozen λeff, respectively. GC2003 (Gori and Corasaniti, 2003), LY2018 (Lu et al.,
2018), PE1975 (Penner et al., 1975), SS2002 (Suzuki et al., 2002) and ZM2018 (Zhang et al., 2018) are five data sources.

12
H. He et al. Earth-Science Reviews 211 (2020) 103419

proposed approach of FO1981 does not require the input of quartz


content and constrains λs to 2.92 to 8.8 W m-1 ◦ C-1. However, a small
variation in sand or clay content may significantly change the value of λs
when the sum of sand and clay content (fsand+fclay) is small; when
fsand+fclay =0 or fsilt =1, numerical problem (e.g., division by zero) oc­
curs; (2) FO1981 oversimplified the equation of Ke for unfrozen and
frozen soils and did not differentiate unfrozen soils of various textures.
The other normalized STC schemes generally oversimplified the
equations for calculating Ke, λsat or λdry, perhaps to reduce programming
load (Dharssi et al., 2009) or computation time. However, these over­
simplifications largely contribute to the underperformance of these STC
schemes. For example, SA1998 (Shmakin, 1998) used an oversimplified
√̅̅̅̅̅̅̅̅̅̅
Ke = θtotal to calculate λeff. While VD1991 (Verseghy, 1991), DP1998
(Desborough and Pitman, 1998), and CM1999 (Cox et al., 1999)
assumed that Ke = Sr for all soils at both frozen and unfrozen states,
where Sr is the degree of saturation. In a similar manner, DI2009
(Dharssi et al., 2009) and CS2015 (Chadburn et al., 2015) assumed that
λsat ranges from 0.5 to 2.2 W m-1 ◦ C-1, SA1998 (Shmakin, 1998) used a
constant λsat =2.5 W m-1 ◦ C-1, while DP1998 (Desborough and Pitman,
1998) resulted in an unrealistically small values of λsat.
Models also vary in their choices of λdry. LV1981 (Lunardini, 1981)
assumed a constant small value of λdry =0.15 W m-1 ◦ C-1, SA1998 used
λdry =0.25 W m-1 ◦ C-1, CM1999 used λdry = 0.218 W m-1 ◦ C-1, 0.227 W m-
1 ◦ -1
C , and 0.272 W m-1 ◦ C-1 for fine-, medium- and coarse-textured soils,
respectively, while VD1991 (Verseghy, 1991) used λdry = λ1− s
P
that
resulted in a very high dry soil thermal conductivity, where P is porosity.
Heuristic treatments also appear in some models. For example, when
the estimates of λeff failed to match experimental λeff, a correction factor
was introduced, making the results more arbitrary. For example, one
recommendation is to multiply λeff by 1.3 if all soil water becomes frozen
or by values of 1 ~ 1.3 if soil water is partially frozen for the SA1998
(Kalyuzhny and Pavlova, 1981).

3.4. STC schemes as a function of matric potential

The KL1983 (Kutchment et al., 1983) scheme is an extension of the


MP1981 (McCumber and Pielke, 1981) by adding an ice content
Fig. 5. Comparison of measured and predicted thermal conductivity (λeff) as a
weighted factor, i.e., KL1983= (1+θice) MP1981. However, measure­ function of matric potential (pF) with soil thermal conductivity (STC) scheme of
ments of λeff as a function of matric potential (ψ , cm water) in frozen soils MP1981 (McCumber and Pielke, 1981) and KL1983 (Kutchment et al., 1983):
are not available. For the purpose of this study, the KL1983 and the (a) four soils of sand 20/30, C10F00, C09F01, and C05F05; (b) four soils of
MP1981 were treated as the same scheme and evaluated with the C08F02, C07F03, C02F08 and C00F10. Data used were from Smits et al. (2010)
dataset of Smits et al. (2010) and Wallen et al. (2016); only the predicted and Wallen et al. (2016), more details can be found in Table 2. For unfrozen
λeff (ψ ) with MP1981 is shown. The logarithm scale (log10) for both the x- soils, the KL1983 is the same as MP1981 and only results predicted with
and y-axis were used, and pF=log10 (ψ ). MP1981 are shown. pF = log10(ψ ), ψ is matric potential in cm water, Y-axis is
Large discrepancies were found between the measured λeff (ψ ) and in Log10 scale to show all the predicted values of λeff.
predicted λeff(ψ ) with MP1981 (Fig. 5 and Table 4). This agrees well with
the study of Peters-Lidard et al. (1998) who found that the MP1981
Table 4
generally overestimates λeff of wet soils, and underestimates λeff of dry
Performance metrics of the soil thermal conductivity (STC) scheme of MP1981
soils, consequently resulting in large errors in estimated surface heat (McCumber and Pielke, 1981) compared to experimental data of Smits et al.
flux. Other studies (Béhaegel et al., 2007; Chen and Dudhia, 2001; Lu (2010) and Wallen et al. (2016).
et al., 2019) also reported unacceptable estimates of λeff (ψ ) at high
No. Soil name RMSE AD NSE
water contents or over the full water content range. But the merit of the
MP1981 scheme is that it enables the study of coupled water and heat W m-1 ◦ C-1 -
1 Sand20/30 16.59 -15.38 -720.66
transport; over 10 LSM models have incorporated MP1981 (Table 1),
2 C10F00 24.42 -10.07 -706.43
which makes it historically the most widely used scheme in land-surface 3 C09F01 6.98 -3.04 -135.76
modeling (Chen and Dudhia, 2001; Fernando et al., 2013; Massey et al., 4 C08F02 14.26 -2.53 -1093.93
2014; Noilhan and Mahfouf, 1996; Noilhan and Planton, 1989; Smir­ 5 C07F03 9.05 -1.76 -606.79
nova et al., 1997). However, few models continue to incorporate 6 C05F05 20.73 -3.16 -3799.36
7 C02F08 1.30 -0.83 -18.45
MP1981 nowadays. 8 C00F10 16.01 -4.29 -4862.76
The unsatisfactory performance of the MP1981 scheme may be
attributed to the fact that it incorrectly assumes that matric potential is
independent of soil type (Chang et al., 1999; He et al., 2020a; Noilhan logarithm function, MP1981 describes the second part well, but not the
and Planton, 1989; Walko et al., 2000). Moreover, air entry value (ψ b) first part (He et al., 2020a). This explains why the MP1981 scheme failed
divides the soil water retention characteristics (SWRC) and λeff (ψ ) into to predict λeff(ψ ) at low pF range for all soils (Fig. 5 and Table 4).
two parts: when ψ < ψ b SWRC shows a plateau and when ψ > ψ b there is
a steep decrease of water content with increase of matric potential. As a

13
H. He et al. Earth-Science Reviews 211 (2020) 103419

3.5. Uncertainty analysis of STC schemes

3.5.1. Quartz content


Previous studies have shown that quartz content plays a critical role
in calculating λs and estimating λeff. However, the measurement of
quartz content requires special instruments including X-ray fluorescence
(XRF) or X-ray diffraction (XRD) which are not always practical to apply
(Nikoosokhan et al., 2015; Schönenberger et al., 2012). There are nine
physically-based or normalized STC schemes requiring λs input (e.g.,
DV1963I~IV, JO1975, VD1991, DP1998, PL1998, CM1999, CK2005,
LS2008, and LS2009). For the three investigated scenarios of quartz
content: (1) quartz content equals half of sand content, fquartz =0.5 fsand;
(2) quartz content equals sand content, fquartz = fsand; and (3) quartz
content equals half of the sum of sand and silt content, fquartz=0.5
(fsand+fsilt). The assumption of fquartz =0.5 fsand generally gave better
estimates than the other two (Table 1), which confirms that fquartz = 0.5
fsand is generally a better approximation of quartz content when fquartz
measurement is not available (Chen et al., 2012; He et al., 2021; Peters-
Lidard et al., 1998). However, care should be taken for applying this
approximation for soils with mostly quartz sand.

3.5.2. Effects of organic matter


Among the 24 STC schemes, only seven schemes (i.e., DV1975,
BB1992, CK2005, DV1963IV, HL1998, LS2008 and CS2015) consider
the influence of peat or organic matter. The STC scheme of DV1975 (de
Vries, 1975; Zeng and Su, 2013) aims to predict λeff of pure peat. The STC
schemes of BB1992 (Becker et al., 1992) and CK2005 (Côté and Konrad,
2005) presented a set of parameters for peat but did not recommend an
approach to calculate λeff of mineral soils with organic matter. There­
fore, the change of organic matter content does not affect the results of
DV1975, BB1992, and CK2005 and they were therefore excluded from
further analysis in this review.
The four STC schemes investigated (i.e., DV1963IV (de Vries, 1963),
HL1998 (Hubrechts, 1998), LS2008 (Lawrence and Slater, 2008), and
CS2015 (Chadburn et al., 2015)), could not satisfactorily predict λeff of
the two soils from Suzuki et al. (2002). However, the overall perfor­
mance of the first three schemes was improved when the effect of Fig. 6. Effects of organic matter on λeff estimated with soil thermal conduc­
tivity (STC) schemes of DV1963IV (Zeng and Su, 2013), HL1998 (Hubrechts,
organic matter was included (Table 5). But the fourth scheme DV1963IV
1998), LS2008 (Lawrence and Slater, 2008), and CS2015 (Chadburn et al.,
did not show significant differences in λeff with and without the inclusion
2015). Closed and open symbols indicate λeff accounting for and excluding
of organic matter (Fig. 6a and b). Moreover, inclusion of organic matter
organic matter, respectively. Two vocanic soils (i.e., Hokudai soil and Kuroboku
would result in smaller values of λeff predicted by LS2008 and CS2015, soil) from SS2002 (Suzuki et al., 2002) are used.
and the differences for Hokudai soil with 13% organic matter content
(Fig. 6a) were greater than that of Kuroboku soil with 22% organic
in a sharp increase in thermal conduction and thus λeff (Singh and Devid,
matter (Fig. 6b). Conversely, the HL1998 scheme resulted in a greater
2000). In the same manner, cementation, such as colloidal precipitation
λeff by accounting for the effects of organic matter from both soils
or organic matter at particle contacts, may increase the contact area and
(Fig. 6a and b). Furthermore, all four schemes performed better (i.e., less
thus increase the λeff (Adivarahan et al., 1962; Tarnawski et al., 2002).
off the l:1 line) at low water contents than at greater water contents (or
However, Abu-Hamdeh and Reeder (2000) reported that organic matter
λeff range > ~1 W m-1 ◦ C-1, Fig. 6).
(e.g., treatments with 5~30% organic matter at 5% increment) had a
It is well known that λeff of unsaturated soils increases with water
negative influence on thermal conductivity of a clay loam soil; the dry
content because the thickening of the water films “bridges” soil particles
soil thermal conductivity decreases with an increase in organic matter
and increases the heat flow path (or inter-particle contacts). This results
content. This is understandable because thermal conductivity of organic
matter is about 0.25 W m-1 ◦ C-1, which is smaller than other components
such as water (~ 0.56 W m-1 ◦ C-1) and soil solids (2~8 W m-1 ◦ C-1). Zhao
Table 5 and Si (2019) and Zhao et al. (2019) demonstrated that λeff of sand-peat
Effects of organic matter (om) content on performance of four soil thermal
mixture decreased with increasing ratios of peat to sand (e.g., treatments
conductivity (STC) schemes (DV1963IV (de Vries, 1963; Zeng and Su, 2013),
with 0~100% organic matter at 20% increment) at full water content
HL1998 (Hubrechts, 1998), LS2008 (Lawrence and Slater, 2008), CS2015
(Chadburn et al., 2015) based on data of SS2002 (Suzuki et al., 2002). range. From this aspect, the LS2008 and CS2015 can reflect the general
trend of change in λeff when organic matter is included.
STC Schemes With organic matter Without organic matter

RMSE AD NSE RMSE AD NSE 3.5.3. Selection of data for evaluation


W m-1 ◦ C-1 - W m-1 ◦ C-1 - Sections 3.1 through 3.4 showed that the 24 STC schemes incorpo­
DV1963IV 0.99 -0.98 -6.61 1.30 -1.28 -11.98 rated in LSMs varied in their ability to compare with the experimental
HL1998 0.42 -0.35 0.64 +0.46 -2.18 data used for evaluation. Therefore, we further investigated how the
choice of dataset affects the estimated λeff for each STC scheme. Three
+0.29
LS2008 0.53 -0.48 -1.17 0.60 -0.55 -1.81
CS2015 0.35 +0.20 0.05 0.32 +0.13 0.22 scenarios for each of the five data sources were compared based on: (1)

14
H. He et al. Earth-Science Reviews 211 (2020) 103419

both unfrozen and frozen data points (T+&T-); (2) unfrozen data points Therefore, the CLM model outputs can catch the general trend (seasonal
only (T+); (3) frozen data points only (T-). For the nine STC schemes and interannual) as indicated by the nigh NSE values, but not the small-
requiring a quartz content, fquartz =0.5 fsand was assumed based on the time scale variation as implied in the large RMSE and AD values. This
analysis in section 3.5.1. Results are shown in Fig. 7 and Table S-3. echoes the findings in sections 3.1 through 3.5 that when validated
The performance of the STC schemes largely depended on the size of against measured STC, the performance of any STC scheme is at best
the dataset (number of samples or measurements). For example, RMSE comparable to that of an average STC, as indicated by the generally
could be as small as 0.1 W m-1 ◦ C-1 for KM1949, DV1975, CH1987, negative NSE values (Table 3). The RMSE of simulated temperature
LV1981, and SA1998 for frozen data of GC2003 (three data points), but obtained in our study is similar to that of Yang et al. (2018) who showed
for data from other studies, RMSE are generally much larger (Fig. 7a~c the RMSE of simulated soil temperature at 10 and 20 cm below ground
and Table S-3). The linear/non-linear regression STC schemes tended to surface with CLM 4.5, ranges from 1.92 to 5.73 ◦ C. Moreover, there is
underestimate λeff for all frozen and unfrozen data (Fig. 7d~f). Most STC soil-texture-dependent performance: CLM with all the 24 schemes
schemes overestimated the unfrozen data of GC2003 and under­ overestimated MODIS LST in clay (Fig. 9b) and underestimated MODIS
estimated the unfrozen data of LY2018 (Fig. 7e). LST in loam (Fig. 9e) and sand (Fig. 9h).Note that the magnitude of the
Many STC schemes performed better (with NSE >0.8) on a single RMSE and AD may also result from the biased forcing data of the study
data source (Fig. 7g~i, Table S-3) than on the composite of all five data sites (e.g., precipitation, temperature, wind, humidity, and radiation) in
sources (Table 3). These schemes include (1) CS1984, JO1975, SA1998, addition to the STC schemes. Therefore, it is prudent to also evaluate the
CM1999, CK205, LS2008 on all data or unfrozen data of GC2003; (2) discrepancies between the STC schemes and soils.
CS2015 on unfrozen data of SS2002, DV1963I and III, CM199; (3) In general, averages of RMSE, NSE and AD over all the 24 STC
CK2005 on frozen data of PE1975; (4) CG1985 on unfrozen data of schemes, show that loam had the best performance(Fig. 9d-f) and clay
ZM208;(5) LS2009 on unfrozen data of LY2018. Figure 8 demonstrates had the worst among the three soil texture classes (Fig. 9a-c). Moreover,
that greater RMSE values are generally associated with larger datasets, the variability of RMSE, NSE and AD over the 24 STC schemes show that
which highlights that there is no universal STC scheme that can be they performed similarly in clay soil (Fig. 9a-c) but quite differently in
applied to a wide range of datasets, soil types, or regions. sand (Fig. 9g-i) or loam (Fig. 9d-f). This indicates that CLM simulated
Most STC schemes worked better for frozen soil data of LY2018 soil temperature is sensitive to the STC schemes in sand and loam, but
(blue) and PE1975 (olive) than that of GC2003 and ZM2018 (Fig. 7i), less so in clay. This is because when validated against measured STC (He
due, in part, to the relatively small sample size (14 points) of GC2003, et al., 2020b), all STC schemes performed poorly in clay. These STC
-and having eight of 14 measured at 35~50 ◦ C. schemes could cause large estimation errors in simulated temperature
It is important to note that a few previous studies investigated effects from a LSM, which could mask the difference between STC schemes in a
of temperature on λeff (Abdulagatova et al., 2009; Campbell et al., 1994; LSM because of low sensitivity or over sensitivity.
Mengistu et al., 2017; Smits et al., 2013). However, the effects of tem­ The four groups of STC schemes in Fig. 9g performed differently. The
perature were generally not considered in any of the 24 STC schemes. In DV1963I-IV (the first group in Fig. 9) performed the worst, suggesting
addition, as a frequently reported phenomenon in frozen soil studies (He that the physically based schemes are less likely to represent the soil
and Dyck, 2013; Spaans and Baker, 1996), and characterized in the data system well. The matric potential based STC schemes (last group in
of ZM2018 (Zhang et al., 2018), freeze-thaw hysteresis has not been Fig. 9, i.e., MP1981 and KL1983) had a similarly poor performance
accounted for in any STC schemes, to the best of our knowledge. perhaps for a different reason: Currently, only limited data are available
Previous studies evaluating STC schemes are generally based on to develop and validate these models (He et al., 2020a). The normalized
unfrozen soils of small sample size or limited data sources (e.g., <100 STC schemes (the third group in Fig. 9) performed slightly better than
data points) (Wang et al., 2020a), which may not provide a thorough the linear or non-linear regression STC schemes (the second group in
overview of the performance of STC schemes. In addition, empirical STC Fig. 9). This may be because the normalized STC schemes restrict the
schemes, including linear/non-linear regression and normalized STC change of STC between two realistic values, the saturated and dry STC
schemes, theoretically cannot be applied to soils beyond the soil types or (He et al., 2020d).
water contents used to develop those schemes (He et al., 2017). There­ The STC schemes have evolved over time and the latest schemes in
fore, a STC scheme developed or calibrated with a large dataset may general performed better than the older schemes as evidenced by the
enable wider and more accurate applications (He et al., 2017; Yan et al., much better performance of the LS2009 and CS2015 than that of others.
2019). However, a sufficiently large and comprehensive database on However, an exception is that VD1991 did not show a satisfactory per­
measured λeff, especially λeff of frozen soils, is not available. It is there­ formance for the STC estimates (Fig. 3d and Table 3), but surprisingly
fore urgent to establish such a database for evaluating or developing STC displayed the best performance for the LST estimates (Fig. 9).
schemes. Moreover, as demonstrated here, no single correlation or
technique currently used in LSMs is capable of accurately predicting λeff 4. Conclusions and perspectives
for all soil types and states (Cheng and Hsu, 1999; Progelhof et al.,
1976). Consequently, before modeling with LSMs one should carefully In total, 24 soil thermal conductivity (STC) schemes incorporated in
evaluate whether the STC scheme incorporated in the LSMs meets the 38 land-surface models (LSMs), SVAT, and hydrological models were
modeling requirement. reviewed and compared with a large compiled dataset consisting of both
unfrozen and frozen measurements of soil thermal conductivity. The
3.6. Demonstration of the effects of STC schemes on land surface STC schemes were also incorporated into the CLM 4.5 and assessed at
temperature three sites with contrasting soil texture using seven-year MODIS land-
surface temperature data. Results showed that none of the STC
The RMSE, NSE and AD of the CLM soil temperature outputs at 0.01 schemes could satisfactorily predict λeff or land surface temperature for
m depth below the ground surface with the 24 STC schemes (including the wide range of soil types and states as presented in this review. The
four modified versions of DV1963 scheme) for clay, loam, and sand are performance of the STC schemes largely depended on the dataset size
presented in Fig. 9. All 24 STC schemes resulted in large RMSE values and characteristics (e.g., unfrozen or frozen data, fine-, medium- or
(Fig. 9 a, d, g), suggesting a poor performance from all 24 STC schemes. coarse-textured soils). For instance, a greater mean value of RMSE,
The NSE values for clay are generally less than 0.5, while NSE values calculated from the 24 STC schemes, is generally associated with a larger
generally ranges from 0.65 to 0.85 for sand and loam. This indicates that dataset. This may indicate that there is a lack of universal STC scheme or
the model prediction bias is large, as indicated by RMSE and AD values, that good performance of a STC scheme may be restricted to certain
but it is still small relative to the overall variation in the measured data. dataset, soil types or regions. In many examples in the literature, STC

15
H. He et al. Earth-Science Reviews 211 (2020) 103419

Fig. 7. Effects of data size and type on three performance metrics of the 22 soil thermal conductivity (STC) schemes. Three metrics: RMSE indicates root mean
squared errors (a~c), AD indicates average deviation (d~f) and NSE indicate Nash-Sutcliff Efficiency (g~i, NSE<-10 were truncated). a, d and g are performance of
STC schemes on all data measured at both unfrozen and frozen temperatures for each of the five data sources; b, e, h are performance of STC schemes on unfrozen
data (T+); c, f and i are performance of STC schemes on frozen data (T-). STC schemes including four types of the de Vries (1963) with different parameterizations
(DV1963I ~ IV), KM1949 (Kersten, 1949), DV1975 (de Vries, 1975), CS1981 (Camillo and Schmugge, 1981), CS1984 (Cass et al., 1984), CG1985 (Campbell, 1985),
CH1987 (Chung and Horton, 1987), BB1992 (Becker et al., 1992), HL1998 (Hubrechts, 1998), JO1975 (Johansen, 1975), FO1981 (Farouki, 1981), LV1981
(Lunardini, 1981), VD1991 (Verseghy, 1991), DP1998 (Desborough and Pitman, 1998), PL1998 (Peters-Lidard et al., 1998), SA1998 (Shmakin, 1998), CM1999 (Cox
et al., 1999), CK2005 (Côté and Konrad, 2005), LS2008 (Lawrence and Slater, 2008), DI2009 (Dharssi et al., 2009), LS2009 (Luo et al., 2009b) and CS2015
(Chadburn et al., 2015). Five data sources includes GC2003 (Gori and Corasaniti, 2003), LY2018 (Lu et al., 2018), PE1975 (Penner et al., 1975), SS2002 (Suzuki
et al., 2002) and ZM2018 (Zhang et al., 2018). Values of NSE less than -10 were truncated. Data in this figure can also be found in the supplementary table S-3.

16
H. He et al. Earth-Science Reviews 211 (2020) 103419

Fig. 8. Effects of data size and soil frozen condition on performance of root-mean-squared errors (RMSE). The solid circles of red (unfrozen measurements of a certain
dataset), blue (frozen measurements of a certain dataset), and olive (both unfrozen and frozen measurements of a certain dataset) are mean of RMSE for 24 soil
thermal conductivity schemes based on the specified dataset. The details of corresponding dataset can be found in Table 2.

Fig. 9. Performance measures of the 24 soil thermal conductivity schemes on soil surface temperature at 0.01 m modelling in three sites with soil texture of clay,
loam and sand using Community Land Model 4.5 (CLM). STC schemes including four types of the de Vries (1963) with different parameterizations (DV1963I ~ IV),
KM1949 (Kersten, 1949), DV1975 (de Vries, 1975), CS1981 (Camillo and Schmugge, 1981), CS1984 (Cass et al., 1984), CG1985 (Campbell, 1985), CH1987 (Chung
and Horton, 1987), BB1992 (Becker et al., 1992), HL1998 (Hubrechts, 1998), JO1975 (Johansen, 1975), FO1981 (Farouki, 1981), LV1981 (Lunardini, 1981),
VD1991 (Verseghy, 1991), DP1998 (Desborough and Pitman, 1998), PL1998 (Peters-Lidard et al., 1998), SA1998 (Shmakin, 1998), CM1999 (Cox et al., 1999),
CK2005 (Côté and Konrad, 2005), LS2008 (Lawrence and Slater, 2008), DI2009 (Dharssi et al., 2009), LS2009 (Luo et al., 2009b), CS2015 (Chadburn et al., 2015),
MP1981 (McCumber and Pielke, 1981), and KL1983 (Kutchment et al., 1983).

schemes with specific applications to certain soil types have been arbi­ evaluate current STC schemes and to develop new STC schemes with
trarily extended or modified for much wider or even universal applica­ wider and more accurate applications. The following areas need
tions without critical evaluation or calibration with large datasets, a particular attention:
condition that will undoubtedly lead to large error in model simulations.
There is a general insufficiency of data on STC at present and we (1) Measured data for clay and frozen soils. Clay soil is subject to
recommend development of a systematic, comprehensive database to structural and volume changes, dictated by minerology and soil

17
H. He et al. Earth-Science Reviews 211 (2020) 103419

management. Yet, there is limited measured STC data available Adivarahan, P., Kunii, D., Smith, J.M., 1962. Heat transfer in porous rocks through which
single-phase fluids are flowing. Soc. Pet. Eng. J. 2 (3), 290–296.
for both clay soils and frozen soils. This largely forgotten area of
Balland, V., Arp, P.A., 2005. Modeling soil thermal conductivities over a wide range of
study should be the focus of STC measurement and theoretical conditions. J. Environ. Eng. Sci. 4 (6), 549–558.
investigations and is of great importance for food security and Becker, B.R., Misra, A., Fricke, B.A., 1992. Development of correlations for soil thermal
greenhouse gas emission studies. conductivity. Int. Commun. Heat Mass Transf. 19 (1), 59–68.
Béhaegel, M., Sailhac, P., Marquis, G., 2007. On the use of surface and ground
(2) Measured data are only available in limited geographic areas. A temperature data to recover soil water content information. J. Appl. Geophys. 62 (3),
concerted effort and network are needed for sampling and 234–243.
measuring soil STC worldwide. Bonan, G.B., Oleson, K.W., Vertenstein, M., Levis, S., Zeng, X., Dai, Y., Dickinson, R.E.,
Yang, Z.-L., 2002. The land surface climatology of the community land model
(3) The available STC data have been generally collected under coupled to the ncar community climate model*. J. Clim. 15 (22), 3123–3149.
controlled laboratory settings (He et al., 2020d; Lu et al., 2007; Bonan, G.B., Hartman, M.D., Parton, W.J., Wieder, W.R., 2013. Evaluating litter
Tarnawski et al., 2015; Wang et al., 2020a). There is no in situ decomposition in earth system models with long-term litterbag experiments: an
example using the Community Land Model version 4 (CLM4). Glob. Chang. Biol. 19
equipment for long-term monitoring of STC (He et al., 2018b; He (3), 957–974.
et al., 2020e). Therefore, it is recommended to include STC sen­ Bristow, K.L., 2002. 5.3 Thermal Conductivity. In: Dane, J.H., Topp, C.G. (Eds.), Methods
sors into meteorological stations. This would allow us to couple of Soil Analysis: Part 4 Physical Methods. SSSA Book Series. Soil Science Society of
America, Madison, WI, pp. 1209–1226.
STC and other soil properties (e.g., soil temperature, water/ Brovka, G., Rovdan, E., 1999. Thermal conductivity of peat soils. Eurasian Soil Sci. 32
moisture content) and climatic data used to validate or calibrate (5), 533–537.
LSMs. Similar to datasets of hydraulic properties (Rahmati et al., Brunke, M.A., Broxton, P., Pelletier, J., Gochis, D., Hazenberg, P., Lawrence, D.M.,
Leung, L.R., Niu, G.-Y., Troch, P.A., Zeng, X., 2016. Implementing and evaluating
2018; Zhang et al., 2020), precipitation, and temperature (He
variable soil thickness in the community land model, Version 4.5 (CLM4.5). J. Clim.
et al., 2020f; Lembrechts et al., 2020; Peng et al., 2019), future 29 (9), 3441–3461.
work on developing high-resolution dataset of soil thermal Camillo, P., Schmugge, T., 1981. A computer program for the simulation of heat and
properties at regional to global scales for land-surface modeling is moisture flow in soils.
Campbell, G.S., 1985. Soil physics with BASIC: Transport Models for Soil-Plant Systems.
desired (Dai et al., 2019b; Zhao et al., 2018). Developments in soil science: 14. Elsevier, New York, p. 14.
(4) It should be noted that the issue of thermal conductivity has Campbell, G.S., Jungbauer Jr., J.D., Bidlake, W.R., Hungerford, R.D., 1994. Predicting
received relatively little attention among the land surface com­ the effect of temperature on soil thermal conductivity. Soil Sci. 158 (5), 307–313.
Cao, L., Bala, G., Caldeira, K., Nemani, R., Ban-Weiss, G., 2009. Climate response to
munity (Dharssi et al., 2009). LSM modelers need to be mindful of physiological forcing of carbon dioxide simulated by the coupled Community
the inherent bias in a scheme when making predictions using Atmosphere Model (CAM3.1) and Community Land Model (CLM3.0). Geophys. Res.
LSMs for different regions. Fortunately, parts of the land surface Lett. 36 (10).
Cass, A., Campbell, G.S., Jones, T.L., 1984. Enhancement of thermal water-vapor
community are aware of the importance of STC schemes and are diffusion in soil. Soil Sci. Soc. Am. J. 48 (1), 25–32.
currently working with soil physicists, hydrologists, and meteo­ Chadburn, S., Burke, E., Essery, R., Boike, J., Langer, M., Heikenfeld, M., Cox, P.,
rologists to address the issue. For instance, the International Soil Friedlingstein, P., 2015. An improved representation of physical permafrost
dynamics in the JULES land-surface model. Geosci. Model Dev. 8 (5), 1493–1508.
modelling Consortium (ISMC) has initiated a “Soil Thermal Chang, S., Hahn, D., Yang, C.-H., Norquist, D., Ek, M., 1999. Validation study of the CAPS
properties Working group project” to improve descriptions of model land surface scheme using the 1987 Cabauw/PILPS dataset. J. Appl. Meteorol.
thermal soil properties and related global parameter sets for 38 (4), 405–422.
Chen, F., Dudhia, J., 2001. Coupling an Advanced Land Surface–Hydrology Model with
LSMs.
the Penn State–NCAR MM5 Modeling System. Part I: Model Implementation and
Sensitivity. Mon. Weather Rev. 129 (4), 569–585.
Chen, Y., Yang, K., Tang, W., Qin, J., Long, Z., 2012. Parameterizing soil organic carbon’s
impacts on soil porosity and thermal parameters for Eastern Tibet grasslands. Sci.
Declaration of Competing Interest
China Earth Sci. 55 (6), 1001–1011.
Cheng, P., Hsu, C., 1999. The effective stagnant thermal conductivity of porous media
The authors declare that they have no known competing financial with periodic structures. J. POROUS MEDIA 2 (1), 19–38.
interests or personal relationships that could have appeared to influence Cheng, Y., Wang, J., Wang, J., Wang, S., Chang, S.X., Cai, Z., Zhang, J., Niu, S., Hu, S.,
2020. Nitrogen deposition differentially affects soil gross nitrogen transformations in
the work reported in this paper. organic and mineral horizons. Earth Sci. Rev. 201, 103033.
Chung, S., Horton, R., 1987. Soil heat and water flow with a partial surface mulch. Water
Resour. Res. 23 (12), 2175–2186.
Acknowledgements Clark, C.A., Arritt, P.W., 1995. Numerical simulations of the effect of soil moisture and
vegetation cover on the development of deep convection. J. Appl. Meteorol. 34 (9),
Datasets used in this study were digitized from the published liter­ 2029–2045.
Clark, M.P., Slater, A.G., Rupp, D.E., Woods, R.A., Vrugt, J.A., Gupta, H.V., Wagener, T.,
ature or obtained by personal communication. The authors are indebted Hay, L.E., 2008. Framework for Understanding Structural Errors (FUSE): A modular
to the referred authors for their valuable data and efforts. Thanks go to framework to diagnose differences between hydrological models. Water Resour. Res.
Jie Yang for running the CLM and processing the temperature outputs. 44 (12).
Côté, J., Konrad, J.-M., 2005. A generalized thermal conductivity model for soils and
Funding for this research was provided in part by the Strategic Pri­
construction materials. Can. Geotech. J. 42 (2), 443–458.
ority Research Program of Chinese Academy of Sciences (No. Cox, M.P., Betts, A.R., Bunton, B.C., Essery, H.R.L., Rowntree, R.P., Smith, J., 1999. The
XDA20040202), the national Natural Science Foundation of China impact of new land surface physics on the GCM simulation of climate and climate
sensitivity. Clim. Dyn. 15 (3), 183–203.
(NSFC Grant No. 41877015, 42077135 and 41877017), the National
Dai, Y., Wei, N., Yuan, H., Zhang, S., Shangguan, W., Liu, S., Lu, X., Xin, Y., 2019a.
Key Research and Development Program of China (No. Evaluation of soil thermal conductivity schemes for use in land surface modeling.
2017YFD0200205), Natural Science Foundation of Shaanxi Province J. Adv. Model. Earth Syst. 11 (11), 3454–3473.
(2020JM-169), China Postdoctoral Science Foundation Dai, Y., Xin, Q., Wei, N., Zhang, Y., Shangguan, W., Yuan, H., Zhang, S., Liu, S., Lu, X.,
2019b. A global high-resolution data set of soil hydraulic and thermal properties for
(2018M641024), the State Key Laboratory of Frozen Soil Engineering land surface modeling. J. Adv. Model. Earth Syst. 11 (9), 2996–3023.
(SKLFSE201905), the Training Program Foundation for the Young Tal­ de Vries, D.A., 1952. A nonstationary method for determining thermal conductivity of
ents of the Northwest A&F University, and the 111 project (No.B12007). soil in situ. Soil Sci. 73 (2), 83–90.
de Vries, D.A., 1963. Thermal properties of soil. In: van Dijk, W.R. (Ed.), Physics of Plant
Environment. North Holland Publishing, Amsterdam, pp. 210–235.
Reference de Vries, D.A., 1975. Heat transfer in soils. In: de Vries, D.A., Afgan, N.H. (Eds.), Heat
and mass transfer in the biosphere, I. Transfer processes in the plant environment.
Scripta Book Co, Washington, D.C., pp. 5–28
Abdulagatova, Z., Abdulagatov, I.M., Emirov, V.N., 2009. Effect of temperature and
Desborough, C.E., Pitman, A.J., 1998. The BASE land surface model. Glob. Planet. Chang.
pressure on the thermal conductivity of sandstone. Int. J. Rock Mech. Min. Sci. 46
19 (1-4), 3–18.
(6), 1055–1071.
Dharssi, I., Vidale, P.L., Verhoef, A., Macpherson, B., Jones, C., Best, M., 2009. New soil
Abu-Hamdeh, N.H., Reeder, R.C., 2000. Soil thermal conductivity effects of density,
physical properties implemented in the Unified Model at PS18. Met Office, Exeter,
moisture, salt concentration, and organic matter. Soil Sci. Soc. Am. J. 64 (4),
UK.
1285–1290.

18
H. He et al. Earth-Science Reviews 211 (2020) 103419

Dickinson, R.E., Oleson, K.W., Bonan, G., Hoffman, F., Thornton, P., Vertenstein, M., Hubrechts, L., 1998. Transfer functions for the generation of thermal properties of
Yang, Z.-L., Zeng, X., 2006. The community land model and its climate statistics as a Belgian soils. Catholic University of Leuven, Belgium, 236 pp.
component of the community climate system model. J. Clim. 19 (11), 2302–2324. Johansen, O., 1975. Varmeledningsevne av jordarter (Thermal conductivity of soils),
Dong, Y., McCartney, J.S., Lu, N., 2015. Critical review of thermal conductivity models University of Trondheim, Trondheim, Norway., US Army Corps of Engineers, Cold
for unsaturated soils. Geotech. Geol. Eng. 33 (2), 207–221. Regions Research and Engineering Laboratory, Hanover, N.H. CRREL Draft English.
Droogers, P., Kite, G., Murray-Rust, H., 2000. Use of simulation models to evaluate Translation 637.
irrigation performance including water productivity, risk and system analyses. Irrig. Kalyuzhny, I.L., Pavlova, K.K., 1981. Formirovanie poter’ talogo stoka (forming of melt
Sci. 19 (3), 139–145. runoff losses) (in russian). Gidrometeoizdat, Leningrad.
Du, E., Vittorio, A.D., Collins, W.D., 2016. Evaluation of hydrologic components of Kavetski, D., Kuczera, G., Franks, S.W., 2006. Bayesian analysis of input uncertainty in
community land model 4 and bias identification. Int. J. Appl. Earth Obs. Geoinf. 48, hydrological modeling: 2. Application. Water Resour. Res. 42 (3).
5–16. Kersten, M.S., 1949. Thermal properties of soils, Minnesota University Institute of
Duan, Q., Schaake, J., Andréassian, V., Franks, S., Goteti, G., Gupta, H.V., Gusev, Y.M., Technology. Minneapolis, MN.
Habets, F., Hall, A., Hay, L., Hogue, T., Huang, M., Leavesley, G., Liang, X., Kojima, Y., Heitman, J.L., Noborio, K., Ren, T., Horton, R., 2018. Sensitivity analysis of
Nasonova, O.N., Noilhan, J., Oudin, L., Sorooshian, S., Wagener, T., Wood, E.F., temperature changes for determining thermal properties of partially frozen soil with
2006. Model Parameter Estimation Experiment (MOPEX): An overview of science a dual probe heat pulse sensor. Cold Reg. Sci. Technol. 151, 188–195.
strategy and major results from the second and third workshops. J. Hydrol. 320 (1- Koster, R.D., Dirmeyer, P.A., Guo, Z., Bonan, G., Chan, E., Cox, P., Gordon, C.T.,
2), 3–17. Kanae, S., Kowalczyk, E., Lawrence, D., Liu, P., Lu, C.-H., Malyshev, S.,
Ebel, B., Koch, J., Walvoord, M., 2019. Soil physical, hydraulic, and thermal properties in McAvaney, B., Mitchell, K., Mocko, D., Oki, T., Oleson, K., Pitman, A., Sud, Y.C.,
interior Alaska, USA: implications for hydrologic response to thawing permafrost Taylor, C.M., Verseghy, D., Vasic, R., Xue, Y., Yamada, T., 2004. Regions of strong
conditions. Water Resour. Res. 55 (5), 4427–4447. coupling between soil moisture and precipitation. Science 305 (5687), 1138–1140.
Farouki, O.T., 1981. The thermal properties of soils in cold regions. 0165-232X. Kurylyk, B.L., Watanabe, K., 2013. The mathematical representation of freezing and
Fayer, M.J., 2000. UNSAT-H Version 3.0:Unsaturated Soil Water and Heat Flow Model: thawing processes in variably-saturated, non-deformable soils. Adv. Water Resour.
Theory, User Manual, and Examples, United States. 60, 160-177.
Fernando, H.J.S., Verhoef, B., Di Sabatino, S., Leo, L.S., Park, S., 2013. The Phoenix Kurylyk, B.L., MacQuarrie, K.T.B., McKenzie, J.M., 2014. Climate change impacts on
Evening Transition Flow Experiment (TRANSFLEX). Bound.-Layer Meteorol. 147 (3), groundwater and soil temperatures in cold and temperate regions: Implications,
443–468. mathematical theory, and emerging simulation tools. Earth Sci. Rev. 138, 313–334.
Fuchs, M., Campbell, G.S., Papendick, R.I., 1978. An analysis of sensible and latent heat Kutchment, L.S., Demidov, V.N., Motovilov, Y.G., 1983. River runoff generation (in
flow in a partially frozen unsaturated soil. Soil Sci. Soc. Am. J. 42 (3), 379–385. Russian). Russ. Acad. of Sci, Moscow.
Ghimire, B., Riley, W.J., Koven, C.D., Mu, M., Randerson, J.T., 2016. Representing leaf Lafrenière, M.J., Lamoureux, S.F., 2019. Effects of changing permafrost conditions on
and root physiological traits in CLM improves global carbon and nitrogen cycling hydrological processes and fluvial fluxes. Earth Sci. Rev. 191, 212–223.
predictions. J. Adv. Model. Earth Syst. 8 (2), 598–613. Lawrence, P.J., Chase, T.N., 2007. Representing a new MODIS consistent land surface in
Ghuman, b.S., Lal, r., 1985. Thermal conductivity, thermal diffusivity, and thermal the Community Land Model (CLM 3.0). J. Geophys. Res. Biogeosci. 112 (G1).
capacity of some Nigerian soils. Soil Sci. 139 (1), 74–80. Lawrence, D.M., Slater, A.G., 2008. Incorporating organic soil into a global climate
Gori, F., Corasaniti, S., 2003. Temperature variation inside dry and partially frozen mars model. Clim. Dyn. 30 (2-3), 145–160.
soils, Aerospace Conference, 2003. Proceedings. 2003 IEEE, pp. In: 2_603-2_612. Lawrence, D.M., Oleson, K.W., Flanner, M.G., Thornton, P.E., Swenson, S.C.,
Halloran, L.J.S., Rau, G.C., Andersen, M.S., 2016. Heat as a tracer to quantify processes Lawrence, P.J., Zeng, X., Yang, Z.L., Levis, S., Sakaguchi, K., 2011. Parameterization
and properties in the vadose zone: A review. Earth Sci. Rev. 159, 358–373. improvements and functional and structural advances in version 4 of the Community
Han, X., Franssen, H.-J.H., Montzka, C., Vereecken, H., 2014. Soil moisture and soil Land Model. J. Adv. Model. Earth Syst. 3 (1).
properties estimation in the Community Land Model with synthetic brightness Lembrechts, J., Aalto, J., Ashcroft, M., Frenne, P., Kopecký, M., Lenoir, J., Luoto, M.,
temperature observations. Water Resour. Res. 50 (7), 6081–6105. Maclean, I., Roupsard, O., Fuentes-Lillo, E., García, R., Pellissier, L., Pitteloud, C.,
Hansson, K., Simunek, J., Mizoguchi, M., Lundin, L.-C., van Genuchten, M.T., 2004. Alatalo, J., Smith, S., Björk, R., Muffler, L., Cesarz, S., Gottschall, F., 2020. SoilTemp:
Water flow and heat transport in frozen soil: Numerical solution and freeze-thaw a global database of near-surface temperature. Glob. Chang. Biol. 26, 6616–6629.
applications. Vadose Zone J. 3 (2), 693–704. https://doi.org/10.1111/gcb.15123.
Harris, C., Arenson, L.U., Christiansen, H.H., Etzelmüller, B., Frauenfelder, R., Gruber, S., Li, H., Huang, M., Wigmosta, M.S., Ke, Y., Coleman, A.M., Leung, L.R., Wang, A.,
Haeberli, W., Hauck, C., Hölzle, M., Humlum, O., Isaksen, K., Kääb, A., Kern- Ricciuto, D.M., 2011. Evaluating runoff simulations from the Community Land
Lütschg, M.A., Lehning, M., Matsuoka, N., Murton, J.B., Nötzli, J., Phillips, M., Model 4.0 using observations from flux towers and a mountainous watershed.
Ross, N., Seppälä, M., Springman, S.M., Vonder Mühll, D., 2009. Permafrost and J. Geophys. Res.-Atmos. 116 (D24).
climate in Europe: Monitoring and modelling thermal, geomorphological and Li, R., Zhao, L., Wu, T., Wang, Q., Ding, Y., Yao, J., Wu, X., Hu, G., Xiao, Y., Du, Y.,
geotechnical responses. Earth Sci. Rev. 92 (3–4), 117–171. Zhu, X., Qin, Y., Yang, S., Bai, R., Du, E., Liu, G., Zou, D., Qiao, Y., Shi, J., 2019. Soil
He, H., Dyck, M., 2013. Application of multiphase dielectric mixing models for thermal conductivity and its influencing factors at the Tanggula permafrost region
understanding the effective dielectric permittivity of frozen soils. Vadose Zone J. 12 on the Qinghai–Tibet Plateau. Agric. For. Meteorol. 264, 235–246.
(1). Lu, S., Ren, T., Gong, Y., Horton, R., 2007. An improved model for predicting soil thermal
He, H., Dyck, M., Wang, J., Lv, J., 2015. Evaluation of TDR for quantifying heat-pulse- conductivity from water content at room temperature. Soil Sci. Soc. Am. J. 71 (1),
method-induced ice melting in frozen soils. Soil Sci. Soc. Am. J. 79 (5), 1275–1288. 8–14.
He, H., Zhao, Y., Dyck, M., Si, B., Jin, H., Lv, J., Wang, J., 2017. A modified normalized Lu, Y., Lu, S., Horton, R., Ren, T., 2014. An empirical model for estimating soil thermal
model for predicting effective soil thermal conductivity. Acta Geotech. 12 (6), conductivity from texture, water content, and bulk density. Soil Sci. Soc. Am. J. 78
1281–1300. (6), 1859–1868.
He, H., Dyck, M., Horton, R., Li, M., Jin, H., Si, B., 2018a. Distributed temperature Lu, Y., Yu, W., Hu, D., Liu, W., 2018. Experimental study on the thermal conductivity of
sensing for soil physical measurements and its similarity to heat pulse method. In: aeolian sand from the Tibetan Plateau. Cold Reg. Sci. Technol. 146, 1–8.
Sparks, D.L. (Ed.), Advances in Agronomy. Advances in Agronomy. Academic Press, Lu, S., Lu, Y., Peng, W., Ju, Z., Ren, T., 2019. A generalized relationship between thermal
Cambridge, pp. 173–230. conductivity and matric suction of soils. Geoderma 337, 491–497.
He, H., Dyck, M.F., Horton, R., Ren, T., Bristow, K.L., Lv, J., Si, B., 2018b. Development Lunardini, V.J., 1981. Heat transfer in cold climates. Van Nostrand Reinhold Company,
and application of the heat pulse method for soil physical measurements. Rev. New York, 731 pp.
Geophys. 56 (4), 567–620. Luo, S., Lü, S., Zhang, Y., 2009a. Development and validation of the frozen soil
He, H., Dyck, M., Lv, J., 2020a. A new model for predicting soil thermal conductivity parameterization scheme in Common Land Model. Cold Reg. Sci. Technol. 55 (1),
from matric potential. J. Hydrol. 589, 125167. 130–140.
He, H., Dyck, M., Lv, J., 2020b. Review and evaluation of 26 predictive models for Luo, S.Q., Lv, S.H., Zhang, Y., Hu, Z.Y., Ma, Y.M., Li, S.S., Shang, L.Y., 2009b. Soil themal
thermal conductivity of clays [submitted for review]. conductivity parameterization establishment and application in numerical model of
He, H., Flerchinger, G., Kojima, Y., He, D., Hardegree, P.S., Dyck, M., Horton, R., Si, B., central Tibetan Plateau [in Chinese with English abstract]. Chin. J. Geophys. 52,
Feng, H., Lv, J., 2020c. Evaluation of 14 frozen soil thermal conductivity 919~928.
parameterizations with experimental measurements and SHAW model simulation Luo, S., Fang, X., Lyu, S., Zhang, Y., Chen, B., 2017. Improving CLM4.5 simulations of
[submitted for review]. land–atmosphere exchange during freeze–thaw processes on the Tibetan Plateau.
He, H., Noborio, K., Johansen, Ø., Dyck, M., Lv, J., 2020d. Normalized concept for J. Meteorol. Res. 31 (5), 916–930.
effective soil thermal conductivity modelling from dryness to saturation. Eur. J. Soil Makushev, K., Lagutin, A., Volkov, N., Mordvin, E.Y., 2016. Validation of the RegCM4/
Sci. 71 (1), 27–43. CLM4.5 regional climate modeling system over the Western Siberia. XXII
He, H., Dyck, M., Lv, J., 2020e. The heat pulse method for soil physical measurements: A International Symposium Atmospheric and Ocean Optics. Atmospheric Physics
bibliometric analysis. Appl. Sci. 10 (18), 6171. 10035. SPIE.
He, J., Yang, K., Tang, W., Lu, H., Qin, J., Chen, Y., Li, X., 2020f. The first high-resolution Massey, J.D., Steenburgh, W.J., Hoch, S.W., Knievel, J.C., 2014. Sensitivity of near-
meteorological forcing dataset for land process studies over China. Sci. Data 7 (1), surface temperature forecasts to soil properties over a sparsely vegetated dryland
1–11. region. J. Appl. Meteorol. Climatol. 53 (8), 1976–1995.
He, H., Flerchinger, G., Kojima, Y., Dyck, M., Lv, J., 2021. A review and evaluation of 39 Mauder, M., Foken, T., Cuxart, J., 2020. Surface-energy-balance closure over land: a
thermal conductivity models for frozen soils. Geoderma. 382 (114694). review. Bound.-Layer Meteorol. 177 (2-3), 395–426.
Hu, G., Zhao, L., Wu, X., Li, R., Wu, T., Xie, C., Pang, Q., Zou, D., 2017. Comparison of the McCumber, M.C., Pielke, R.A., 1981. Simulation of the effects of surface fluxes of heat
thermal conductivity parameterizations for a freeze-thaw algorithm with a multi- and moisture in a mesoscale numerical model: 1. Soil layer. J. Geophys. Res. 86
layered soil in permafrost regions. CATENA 156, 244–251. (C10), 9929–9938.

19
H. He et al. Earth-Science Reviews 211 (2020) 103419

McInnes, K.J., 1981. Thermal conductivities of soils from dryland wheat regions of Smirnova, T.G., Brown, J.M., Benjamin, S.G., 1997. Performance of different soil model
Eastern Washington. Washington State University, Pullman, Washington. configurations in simulating ground surface temperature and surface fluxes. Mon.
Mengistu, A.G., van Rensburg, L.D., Mavimbela, S.S.W., 2017. The effect of soil water Weather Rev. 125 (8), 1870–1884.
and temperature on thermal properties of two soils developed from aeolian sands in Smith, M.W., Tice, A.R., 1988. Measurement of the unfrozen water content of soils: a
South Africa. CATENA 158, 184–193. comparison of NMR and TDR methods. 88-18.
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I - Smits, K.M., Sakaki, T., Limsuwat, A., Illangasekare, T.H., 2010. Thermal conductivity of
A discussion of principles. J. Hydrol. 10 (3), 282–290. sands under varying moisture and porosity in drainage–wetting cycles. Vadose Zone
Nikoosokhan, S., Nowamooz, H., Chazallon, C., 2015. Effect of dry density, soil texture J. 9 (1), 172–180.
and time-spatial variable water content on the soil thermal conductivity. Geomech. Smits, K.M., Sakaki, T., Howington, S.E., Peters, J.F., Illangasekare, T.H., 2013.
Geoeng. 11 (2), 149–158. Temperature dependence of thermal properties of sands across a wide range of
Noilhan, J., Mahfouf, J.F., 1996. The ISBA land surface parameterization scheme. Glob. temperatures (30-70◦ C).
Planet. Chang. 13 (1-4), 145–159. Spaans, E.J.A., Baker, J.M., 1995. Examining the use of time domain reflectometry for
Noilhan, J., Planton, S., 1989. A simple parameterization of land surface processes for measuring liquid water content in frozen soil. Water Resour. Res. 31 (12),
meteorological models. Mon. Weather Rev. 117 (3), 536–549. 2917–2925.
Okalebo, J.A., Oglesby, R.J., Feng, S., Hubbard, K., Kilic, A., Hayes, M., Hays, C., 2016. Spaans, E.J.A., Baker, J.M., 1996. The soil freezing characteristic: its measurement and
An Evaluation of the Community Land Model (Version 3.5) and Noah Land Surface similarity to the soil moisture characteristic. Soil Sci. Soc. Am. J. 60 (1), 13–19.
Models for Temperature and Precipitation Over Nebraska (Central Great Plains): Suzuki, S., Kashiwagi, J., Nakagawa, S., Soma, K., 2002. Mechanism of hysteresis in
Implications for Agriculture in Simulations of Future Climate Change and thermal conductivity of frozen soils between freezing and thawing processes [In
Adaptation. In: Leal Filho, W., Musa, H., Cavan, G., O’Hare, P., Seixas, J. (Eds.), Japanese with English abstract]. Transactions of the Japanese Society of Irrigation,
Climate Change Adaptation, Resilience and Hazards. Climate Change Management. Drainage and Reclamation Engineering (Japan).
Springer International Publishing, Cham, pp. 21–34. Swenson, S.C., Lawrence, D.M., 2012. A new fractional snow-covered area
Oleson, K., Lawrence, D., Bonan, G., Drewniak, B., Huang, M., Koven, C., Levis, S., Li, F., parameterization for the Community Land Model and its effect on the surface energy
Riley, W., Subin, Z., Swenson, S., Thornton, P., Bozbiyik, A., Fisher, R., Heald, C., balance. J. Geophys. Res.-Atmos. 117 (D21).
Kluzek, E., Lamarque, J.-F., Lawrence, P., Leung, L., Yang, Z.-L., 2013. Technical Tarnawski, V.R., Leong, W., 2012. A series-parallel model for estimating the thermal
description of version 4.5 of the Community Land Model (CLM). conductivity of unsaturated soils. Int. J. Thermophys. 33 (7), 1191–1218.
Oliphant, J.L., Tice, A.R., 1982. Comparison of unfrozen water contents measured by Tarnawski, V.R., Leong, W.H., Gori, F., Buchan, G.D., Sundberg, J., 2002. Inter-particle
DSC and NMR. In: Proceedings of the third international symposium on ground contact heat transfer in soil systems at moderate temperatures. Int. J. Energy Res. 26
freezing, Hanover, NH., pp. 115-121. (15), 1345–1358.
Peng, S.Z., Ding, Y.X., Liu, W.Z., Li, Z., 2019. 1 km monthly temperature and Tarnawski, V.R., Momose, T., McCombie, M.L., Leong, W.H., 2015. Canadian field soils
precipitation dataset for China from 1901 to 2017. Earth System Science Data 11 (4), III. Thermal-conductivity data and modeling. Int. J. Thermophys. 36 (1), 119–156.
1931-1946. Tian, Z., Lu, Y., Horton, R., Ren, T., 2016. A simplified de Vries-based model to estimate
Penner, E., Johnston, G.H., Goodrich, L.E., 1975. Thermal conductivity laboratory thermal conductivity of unfrozen and frozen soil. Eur. J. Soil Sci. 67 (5), 564–572.
studies of some mackenzie highway soils. Can. Geotech. J. 12 (3), 271–288. Tien, Y.M., Chu, C.A., Chuang, W.S., 2005. The prediction model of thermal conductivity
Pernitsky, T., Hu, W., Si, B.C., Barbour, L., 2016. Effects of petroleum hydrocarbon of sand-bentonite based buffer material.
concentration and bulk density on the hydraulic properties of lean oil sand Tong, B., Gao, Z., Horton, R., Li, Y., Wang, L., 2016. An empirical model for estimating
overburden. Can. J. Soil Sci. 96 (4), 435–446. soil thermal conductivity from soil water content and porosity. J. Hydrometeorol. 17
Peters-Lidard, C.D., Blackburn, E., Liang, X., Wood, E.F., 1998. The effect of soil thermal (2), 601–613.
conductivity parameterization on surface energy fluxes and temperatures. J. Atmos. Verhoef, A., 2004. Remote estimation of thermal inertia and soil heat flux for bare soil.
Sci. 55 (7), 1209–1224. Agric. For. Meteorol. 123 (3-4), 221–236.
Piñeiro, G., Perelman, S., Guerschman, J.P., Paruelo, J.M., 2008. How to evaluate Verhoef, A., Ottlé, C., Cappelaere, B., Murray, T., Saux-Picart, S., Zribi, M., Maignan, F.,
models: Observed vs. predicted or predicted vs. observed? Ecol. Model. 216 (3), Boulain, N., Demarty, J., Ramier, D., 2012. Spatio-temporal surface soil heat flux
316–322. estimates from satellite data; results for the AMMA experiment at the Fakara (Niger)
Progelhof, R.C., Throne, J.L., Ruetsch, R.R., 1976. Methods for predicting the thermal supersite. Agric. For. Meteorol. 154-155, 55–66.
conductivity of composite systems: A review. Polym. Eng. Sci. 16 (9), 615–625. Verseghy, D.L., 1991. CLASS—A Canadian land surface scheme for GCMS. I. Soil model.
Rahmati, M., Weihermüller, L., Vanderborght, J., Pachepsky, Y.A., Mao, L., Sadeghi, S. Int. J. Climatol. 11 (2), 111–133.
H., Moosavi, N., Kheirfam, H., Montzka, C., Van Looy, K., Toth, B., Hazbavi, Z., Al Wagener, T., Wheater, H.S., 2006. Parameter estimation and regionalization for
Yamani, W., Albalasmeh, A.A., Alghzawi, M.Z., Angulo-Jaramillo, R., Antonino, A.C. continuous rainfall-runoff models including uncertainty. J. Hydrol. 320 (1-2),
D., Arampatzis, G., Armindo, R.A., Asadi, H., Bamutaze, Y., Batlle-Aguilar, J., 132–154.
Bechet, B., Becker, F., Blöschl, G., Bohne, K., Braud, I., Castellano, C., Cerdà, A., Walko, R.L., Band, L.E., Baron, J., Kittel, T.G.F., Lammers, R., Lee, T.J., Ojima, D.,
Chalhoub, M., Cichota, R., Císlerová, M., Clothier, B., Coquet, Y., Cornelis, W., Pielke Sr., R.A., Taylor, C., Tague, C., 2000. Coupled Atmosphere-Biophysics-
Corradini, C., Coutinho, A.P., de Oliveira, M.B., de Macedo, J.R., Durães, M.F., Hydrology Models for Environmental Modeling. J. Appl. Meteorol. 39 (2000),
Emami, H., Eskandari, I., Farajnia, A., Flammini, A., Fodor, N., Gharaibeh, M., 931–944.
Ghavimipanah, M.H., Ghezzehei, T.A., Giertz, S., Hatzigiannakis, E.G., Horn, R., Wallen, B.M., Smits, K.M., Sakaki, T., Howington, S.E., Deepagoda, T.K.K.C.., 2016.
Jiménez, J.J., Jacques, D., Keesstra, S.D., Kelishadi, H., Kiani-Harchegani, M., Thermal conductivity of binary sand mixtures evaluated through full water content
Kouselou, M., Kumar Jha, M., Lassabatere, L., Li, X., Liebig, M.A., Lichner, L., range. Soil Sci. Soc. Am. J. 80 (3), 592–603.
López, M.V., Machiwal, D., Mallants, D., Mallmann, M.S., de Oliveira Marques, J.D., Wan, Z., 2008. New refinements and validation of the MODIS Land-Surface
Marshall, M.R., Mertens, J., Meunier, F., Mohammadi, M.H., Mohanty, B.P., Temperature/Emissivity products. Remote Sens. Environ. 112 (1), 59–74.
Moncada, M.P., Montenegro, S., Morbidelli, R., Moret-Fernández, D., Moosavi, A.A., Wan, Z., 2014. New refinements and validation of the collection-6 MODIS land-surface
Mosaddeghi, M.R., Mousavi, S.B., Mozaffari, H., Nabiollahi, K., Neyshabouri, M.R., temperature/emissivity product. Remote Sens. Environ. 140, 36–45.
Ottoni, M.V., Ottoni Filho, T.B., Pahlavan Rad, M.R., Panagopoulos, A., Peth, S., Wang, C., Yang, K., 2018. A New Scheme for Considering Soil Water-Heat Transport
Peyneau, P.E., Picciafuoco, T., Poesen, J., Pulido, M., Reinert, D.J., Reinsch, S., Coupling Based on Community Land Model: Model Description and Preliminary
Rezaei, M., Roberts, F.P., Robinson, D., Rodrigo-Comino, J., Rotunno Filho, O.C., Validation. J. Adv. Model. Earth Syst. 10 (4), 927–950.
Saito, T., Suganuma, H., Saltalippi, C., Sándor, R., Schütt, B., Seeger, M., Wang, W., Liang, S., Meyers, T., 2008. Validating MODIS land surface temperature
Sepehrnia, N., Sharifi Moghaddam, E., Shukla, M., Shutaro, S., Sorando, R., products using long-term nighttime ground measurements. Remote Sens. Environ.
Stanley, A.A., Strauss, P., Su, Z., Taghizadeh-Mehrjardi, R., Taguas, E., Teixeira, W. 112 (3), 623–635.
G., Vaezi, A.R., Vafakhah, M., Vogel, T., Vogeler, I., Votrubova, J., Werner, S., Wang, G., Liu, G., Li, C., Yang, Y., 2012. The variability of soil thermal and hydrological
Winarski, T., Yilmaz, D., Young, M.H., Zacharias, S., Zeng, Y., Zhao, Y., Zhao, H., dynamics with vegetation cover in a permafrost region. Agric. For. Meteorol. 162-
Vereecken, H., 2018. Development and analysis of soil water infiltration global 163 (0), 44–57.
database. Earth System Science Data 10 (3), 1237–1263. Wang, L., Zhou, J., Qi, J., Sun, L., Yang, K., Tian, L., Lin, Y., Liu, W., Shrestha, M.,
Rau, G.C., Andersen, M.S., McCallum, A.M., Roshan, H., Acworth, R.I., 2014. Heat as a Xue, Y., Koike, T., Ma, Y., Li, X., Chen, Y., Chen, D., Piao, S., Lu, H., 2017.
tracer to quantify water flow in near-surface sediments. Earth Sci. Rev. 129, 40–58. Development of a land surface model with coupled snow and frozen soil physics.
Russell, E.S., Liu, H., Gao, Z., Finn, D., Lamb, B., 2015. Impacts of soil heat flux Water Resour. Res. 53 (6), 5085–5103.
calculation methods on the surface energy balance closure. Agric. For. Meteorol. Wang, Y., Lu, Y., Horton, R., Ren, T., 2019. Specific heat capacity of soil solids:
214-215, 189–200. Influences of clay content, organic matter, and tightly bound water. Soil Sci. Soc.
Salomonson, V.V., Barnes, W.L., Maymon, P.W., Montgomery, H.E., Ostrow, H., 1987. Am. J. 83 (4), 1062–1066.
MODIS: advanced facility instrument for studies of the Earth as a system. Geoence Wang, J., He, H., Dyck, M., Lv, J., 2020a. A review and evaluation of predictive models
Remote Sens. IEEE Trans. 27 (2), 145–153. for thermal conductivity of sands at full water content range. Energies 13 (5), 1083.
Saxton, K.E., Rawls, W.J., Romberger, J.S., Papendick, R.I., 1986. Estimating generalized Wang, J., He, H., Li, M., Dyck, M., Si, B., Lv, J., 2020b. A review and evaluation of
soil-water characteristics from texture. Soil Sci. Soc. Am. J. 50 (4), 1031–1036. thermal conductivity models of saturated soils. Arch. Agron. Soil Sci. 1–13.
Schönenberger, J., Momose, T., Wagner, B., Leong, W.H., Tarnawski, V.R., 2012. Watanabe, K., Wake, T., 2009. Measurement of unfrozen water content and relative
Canadian field soils I. Mineral composition by XRD/XRF measurements. Int. J. permittivity of frozen unsaturated soil using NMR and TDR. Cold Reg. Sci. Technol.
Thermophys. 33 (2), 342–362. 59 (1), 34–41.
Shmakin, A.B., 1998. The updated version of SPONSOR land surface scheme: PILPS- Wu, D.D., Anagnostou, E.N., Wang, G., Moges, S., Zampieri, M., 2014. Improving the
influenced improvements. Glob. Planet. Chang. 19 (1), 49–62. surface-ground water interactions in the Community Land Model: Case study in the
Singh, D.N., Devid, K., 2000. Generalized relationships for estimating soil thermal Blue Nile Basin. Water Resour. Res. 50 (10), 8015–8033.
resistivity. Exp. Thermal Fluid Sci. 22 (3-4), 133–143.

20
H. He et al. Earth-Science Reviews 211 (2020) 103419

Xie, Z., Hu, Z., Ma, Y., Sun, G., Gu, L., Liu, S., Wang, Y., Zheng, H., Ma, W., 2019. Zeng, Y., Su, Z., 2013. STEMMUS : Simultaneous Transfer of Engery, Mass and
Modeling Blowing Snow Over the Tibetan Plateau With the Community Land Model: Momentum in Unsaturated. ITC-WRS Report, University of Twente, Soil.
Method and Preliminary Evaluation. J. Geophys. Res.-Atmos. 124 (16), 9332–9355. Zhang, N., Wang, Z., 2017. Review of soil thermal conductivity and predictive models.
Yan, H., He, H., Dyck, M., Jin, H., Li, M., Si, B., 2019. A generalized model for estimating Int. J. Therm. Sci. 117, 172–183.
effective soil thermal conductivity based on the Kasubuchi algorithm. Geoderma Zhang, T., Barry, R.G., Armstrong, R.L., 2004. Application of Satellite Remote Sensing
353, 227–242. Techniques to Frozen Ground Studies. Polar Geogr. 28 (3), 163–196.
Yang, K., Wang, C., 2019. Water storage effect of soil freeze-thaw process and its impacts Zhang, X., Gao, Z., Wei, D., 2012. The sensitivity of ground surface temperature
on soil hydro-thermal regime variations. Agric. For. Meteorol. 265, 280–294. prediction to soil thermal properties Using the Simple Biosphere Model (SiB2). Adv.
Yang, K., Koike, T., Ye, B., Bastidas, L., 2005. Inverse analysis of the role of soil vertical Atmos. Sci. 29 (3), 623–634.
heterogeneity in controlling surface soil state and energy partition. J. Geophys. Res.- Zhang, M., Lu, J., Lai, Y., Zhang, X., 2018. Variation of the thermal conductivity of a silty
Atmos. 110 (D8) n/a-n/a. clay during a freezing-thawing process. Int. J. Heat Mass Transf. 124, 1059–1067.
Yang, K., Watanabe, T., Koike, T., Li, X., Fujii, H., Tamagawa, K., Ma, Y., Ishikawa, H., Zhang, Y., Schaap, M.G., Wei, Z., 2020. Development of hierarchical ensemble model and
2007. Auto-calibration system developed to assimilate AMSR-E data into a land estimates of soil water retention with global coverage. Geophys. Res. Lett. 47 (15)
surface model for estimating soil moisture and the surface energy budget. e2020GL088819.
J. Meteorol. Soc. Japan. Ser. II 85A, 229–242. Zhao, Y., Si, B., 2019. Thermal properties of sandy and peat soils under unfrozen and
Yang, M., Nelson, F.E., Shiklomanov, N.I., Guo, D., Wan, G., 2010. Permafrost frozen conditions. Soil Tillage Res. 189, 64–72.
degradation and its environmental effects on the Tibetan Plateau: A review of recent Zhao, H., Zeng, Y., Lv, S., Su, Z., 2018. Analysis of soil hydraulic and thermal properties
research. Earth Sci. Rev. 103 (1-2), 31–44. for land surface modeling over the Tibetan Plateau. Earth System Science Data 10
Yang, K., Wang, C., Li, S., 2018. Improved simulation of frozen-thawing process in Land (2), 1–40.
Surface Model (CLM4.5). J. Geophys. Res.-Atmos. 123 (23), 13,238–13,258. Zhao, Y., Si, B.C., Zhang, Z., Li, M., He, H., Hill, L.R., 2019. A new thermal conductivity
model for sandy and peat soils. Agric. Forest Meterol. 274, 95–105.

21

You might also like