https://doi.org/10.

5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Modelling rock glacier velocity and ice content, Khumbu and Lhotse
Valleys, Nepal
Yan Hu1,2,3, Stephan Harrison2, Lin Liu1,3, Joanne Laura Wood2
1
Earth System Science Programme, Faculty of Science, The Chinese University of Hong Kong, Hong Kong SAR, China
2
5 College of Life and Environmental Sciences, University of Exeter, Penryn, Cornwall, TR10 9EZ, UK
3
Institute of Environment, Energy and Sustainability, The Chinese University of Hong Kong, The Chinese University of Hong
Kong, Hong Kong SAR, China

Correspondence to: Yan Hu (huyan@link.cuhk.edu.hk)

10 Abstract. Rock glaciers contain significant amounts of ground ice and serve as important freshwater resources as mountain
glaciers melt in response to climate warming. However, current knowledge about ice content in rock glaciers has been acquired
mainly from in situ investigations in limited study areas, which hinders a comprehensive understanding of ice storage in rock
glaciers situated in remote mountains over local to regional scales. In this study, we develop an empirical rheological model
to infer ice content of rock glaciers using readily available input data, including rock glacier planar shape, surface slope angle,
15 active layer thickness, and surface creep rate. The model is calibrated and validated using observational data from the Chilean
Andes and Swiss Alps. We apply the model to infer the ice content of five rock glaciers in Khumbu and Lhotse Valleys, north-
eastern Nepal. The velocity constraints applied to the model are derived from Interferometric Synthetic Aperture Radar
(InSAR) measurements. The inferred volumetric ice fraction in Khumbu and Lhotse Valleys ranges from 71% to 75.3% and
water volume equivalents lie between 1.40 to 5.92 million m3 for individual landforms. Considering previous mapping results
20 and extrapolating from our findings to the entire Nepalese Himalaya, the total amount of water stored in rock glaciers could
be in the magnitude of 10 billion m3, equivalent to a ratio of 1:17 between rock glacier and glacier reservoirs. Due to the
accessibility of the input parameters of the model developed in this study, it is promising to apply the approach to permafrost
regions where previous information about ice content of rock glaciers is lacking, and ultimately to estimate the water storage
potential of the remotely located rock glaciers.

25 1 Introduction

Rock glaciers are valley-floor and valley-side landforms consisting of ice–rock mixtures and are common in all arid and cold
mountain regions. Recent research has suggested that they represent significant hydrological resources in areas where glaciers
are undergoing recession in the face of climate change (Azócar and Brenning, 2010; Jones et al., 2018a; Munroe, 2018;
Rangecroft et al., 2014). The potential hydrological value of rock glaciers, and thus their importance in terms of hydrological

1
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

30 research, was first noted by Corte (1976); despite this, research on the role of rock glaciers in maintaining hydrological stores
in mountain catchments remains limited.
In regions such as the Himalaya, recent research has argued that rock glaciers might represent the end member of an
evolutionary process where some glaciers transition to debris-covered glaciers, a proportion of which will then undergo further
transition to rock glaciers (Jones et al., 2019; Knight et al., 2019). This process would be triggered by the paraglacial response
35 of high mountain slopes as glaciers undergo downwasting, which produces rock slope failures and mountainside collapses,
increasing the flux of rock debris to glacier surfaces. Depending on the debris cover thickness, this would be expected to limit
ice melting and increase the resilience of glaciers to climate change (e.g., Reznichenko et al., 2010).
Recent work (Jones et al., 2021) was the first to show that around 25,000 rock glaciers exist in the Himalayas, covering 3747
km² and containing 51.80 ± 10.36 km³ of water volume equivalent. The comparative importance of rock glacier ice content
40 versus that in glaciers in the region was 1:25, ranging from 1:42 to 1:17 in the Eastern and Central Himalaya and falling to 1:9
in Nepal. Importantly, we expect these existing ratios to reduce significantly as glaciers melt and undergo transitions to rock
glaciers; yet the rates of transition from glacier to rock glacier are not understood. We also expect rock glaciers to provide
water supplies long after glaciers have melted; in other high arid mountains, such as the Andes, ice-cored rock glaciers have
persisted in valleys long after glacier recession (Azócar and Brenning, 2010; Monnier and Kinnard, 2015a). However, there
45 lacks modelling studies to test these postulations and to assess the likelihood of glacier–rock glacier transition and the
hydrological implications of this process.
A significant gap in our understanding of the likely future hydrological role of rock glaciers in arid mountains is the absence
of quantitative information concerning their ice content. Currently, estimates of ice content in rock glaciers have focused on
empirical information from drilling cores and boreholes (Hausmann et al., 2007; Monnier and Kinnard, 2013, 2015a, b; Fukui
50 et al., 2007; Arenson et al., 2002; Berthling et al., 2000; Croce and Milana, 2002; Florentine et al., 2014; Fukui et al., 2008;
Guglielmin et al., 2004; Guglielmin et al., 2018; Haeberli et al., 1998; Haeberli et al., 1999; Krainer et al., 2015; Leopold et
al., 2011; Steig et al., 1998), and from geophysical surveys (e.g., for reviews see: Hauck, 2013; Kneisel et al., 2008; Scott et
al., 1990). However, these approaches are costly, time-consuming and extremely difficult to apply to rock glaciers at high
altitudes and in remote mountains. It is therefore desirable to develop alternative approaches to understanding the likely ice
55 content of rock glaciers, especially for regional scale estimates.
Ice content is one factor controlling the movement of rock glaciers by influencing the driving force and the rheological
properties of materials which constitute the permafrost core (Arenson and Springman, 2005a; Cicoira et al., 2020), thus it is
feasible to infer ice content using rheological modelling and observed kinematic data. Here we adapt an empirical rheological
model by integrating rheological properties of rock glaciers derived from laboratory experiments (Arenson and Springman,
60 2005a), and parameterise the rheological model based on the structure and composition data of Las Liebres rock glacier
(Monnier and Kinnard, 2015b; Monnier and Kinnard, 2016). We then apply the model to simulate surface velocities of four
rock glaciers with known ice content in the Swiss Alps and evaluate the modelling results to determine a suitable
parameterisation scheme. Finally, we present results from modelling the kinematic responses of the coherently moving part of

2
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

five rock glaciers in the study region of north-eastern Nepal and assess the modelled movement as a proxy for ice content by
65 using remote sensing-derived downslope velocities as constraints.

2 Study area

Our study area comprises the Khumbu and Lhotse valleys in north-eastern Nepal (Fig. 1a). The glaciers draining Everest and
Lhotse (e.g., Khumbu and Lhotse glaciers) are the highest in the world and have well defined debris-covered snouts. The
70 tributary valleys contain a variety of rock glaciers and composite landforms where glaciers are transitioning to rock glaciers
(Jones et al., 2019; Knight et al., 2019). There are five rock glaciers in the study area, namely Kala-Patthar, Kongma, Lingten,
Nuptse, and Tobuche (Fig. 1b). The five rock glaciers examined in this study are situated at 4900–5090 m a.s.l., near the
altitudinal boundary of discontinuous permafrost in the region. Previous seismic refraction surveys conducted on active rock
glaciers indicate that the lower limit of permafrost occurrence in this region to be ≈5000–5300 m a.s.l. (Jakob, 1992), which
75 is consistent with an earlier estimate of 4900 m a.s.l. based on ground temperature measurements (Fujii and Higuchi, 1976).
Meteorological data provided by the Pyramid Observatory Laboratory near Lobuche village on the western side of the Khumbu
Glacier (5050 m a.s.l.) reveal that the dominating climate of this area is the South Asian Summer Monsoon. For the period of
1994–2013, recorded accumulated annual precipitation is 449 mm yr-1, with 90% of the precipitation concentrated during
June–September (Salerno et al., 2015). The mean annual air temperature is –2.4 °C (Salerno et al., 2015).

3
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

80
Figure 1: (a) Location of the study site; (b) Google Earth images showing the spatial distribution of the active ice–debris landforms,
including rock glaciers (RG) in red outlines and debris-covered glaciers (DCG) in blue boundaries. RGs are delineated based on
Google Earth images (Jones et al., 2018b).

3 Methods

85 Our methods are divided into two parts and detailed in the subsections below. First, we derived surface kinematics of rock
glaciers using Interferometric Synthetic Aperture Radar (InSAR; Sect. 3.1). Second, we developed a rheological model for
estimating ice content of rock glaciers by using surface velocity as a constraint. Sect. 3.2 describes the model design,
calibration, and validation procedures, as well as application of the model.

4
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

90 3.1 Deriving surface kinematics with InSAR

Nineteen L-band ALOS PALSAR images and twenty-one ALOS-2 PALSAR-2 images acquired during 2006–2010 and 2015–
2020, respectively, were used to form more than fifty interferograms to measure the surface displacements of the landforms in
the study area (Table 1). We selected SAR data to achieve high interferometric coherence by following the criteria such as:
(1) short temporal spans (less than 92 days for ALOS pairs and 70 days for ALOS-2 pairs); (2) short perpendicular baselines
95 (smaller than 800 m for ALOS pairs and 400 m for ALOS-2 pairs). We estimated and removed the topographic phase with the
1-arcsec digital elevation models (DEM) produced by the Shuttle Radar Topography Mission (SRTM) (spatial resolution ~30
m). Multi-looking operation and adaptive Goldstein filter (8×8 pixels) were applied to the interferometric processing, which
was implemented by the open-source software ISCE version 2.4.2 (available at https://github.com/isce-framework/isce2). The
interferograms were unwrapped using the SNAPHU software (Chen and Zebker, 2002). We randomly selected three pixels
100 located at flat and stable ground near each ice–debris landform and averaged their phase values to re-reference the unwrapped
phases measured within the landforms. By doing so, atmospheric artefacts including the water vapour delay and ionospheric
effects can be effectively removed because these are spatially long-wavelength features and can be assumed as constant within
the range of our study objects (Hanssen, 2001).
We then derived the surface velocities along the SAR satellite line-of-sight (LOS) direction from the unwrapped interferograms
105 and projected the LOS velocities to the downslope direction of the landforms. Uncertainties were quantified by considering
the error propagation of the InSAR measurements and associated geometry parameters (Hu et al., 2021).
After that, to ensure high data quality, we selected the InSAR observations meeting the following criteria as valid results for
further analyses: (1) the pixels showing low coherence (<0.3) are masked out before velocity statistics, and more than 40% of
pixels remain after the masking procedure; (2) the mean velocity of the landform is larger than 5 cm yr-1.
110 Next, we defined and outlined the coherently moving part of the landform by considering the time series of downslope velocity
of each pixel acquired during all the observational periods. If the InSAR-measured velocity is higher than 5 cm yr-1 in more
than half of the periods at a given pixel, it was included into the coherently moving part of the landform.
Finally, we analysed the velocity values of all pixels within the coherently moving part of the landform and selected the mean,
median, and maximum values for each observation to characterise the surface kinematics of the landforms.
115 Table 1. List of ALOS PALSAR and ALOS-2 PALSAR-2 data used in this study.

Satellite Acquisition interval Period Path/frame Orbit direction No. of

(days) interferograms
ALOS 46 Dec 2007 to Feb 2010 507/540 Ascending 8
ALOS 46 Dec 2007 to Feb 2010 507/550 Ascending 6
ALOS 46 Jun 2007 to Feb 2010 508/540 Ascending 4
ALOS 46 May 2006 to Jul 2006 511/540 Ascending 1
ALOS-2 14 Mar 2015 48/3050 Descending 1
ALOS-2 14 Jun 2015 to Feb 2020 156/550 Ascending 20

5
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

3.2 Estimating ice content from a surface-velocity-constrained model

This subsection describes the process of model development. The workflow is illustrated in Fig. 2.

Figure 2: Diagram of the workflow conducted in this study to develop and apply a modelling approach for inferring ice content of
120 rock glaciers (RG).

3.2.1 Model design and assumptions

Active rock glaciers creep as a result of two internal processes: plastic deformation of the ice-rich permafrost core, and
deformation at the shear horizon at depth (e.g., Arenson et al., 2002; Berthling, 2011; Cicoira et al., 2019b; Haeberli, 2000;
Kenner et al., 2019). Many previous modelling studies depict the deformation mechanism of rock glaciers based on Glen’s
125 flow law (e.g., Arenson and Springman, 2005a; Cicoira et al., 2020; Whalley and Azizi, 1994), which essentially relates strain
rate (𝜀̇) with effective shear stress (𝜏) and describes the rheology of ice flow (Glen, 1955):
𝜀̇ = 𝐴𝜏 ! , (1)
where 𝐴 and 𝑛 are parameters reflecting variations in environmental conditions (mainly including temperature and pressure),
material properties (such as composition, structure, and texture), and operating creep mechanisms (e.g., diffusion and
130 dislocation).
In this study, we primarily adopted a creep model of ice–debris mixture, proposed by Moore (2014), based on Glen’s flow law:
𝜀̇ = 𝐸𝐴[(𝜏 − 𝜏"# )𝛤]! , (2)
where 𝐸 is a strain enhancement factor; 𝛤 is a parameter reflecting the strength of the ice–debris mixture, associated with the
volumetric debris content (𝜃$ ). When 𝜃$ is less than a critical volumetric debris content (𝜃$% ), the strength of the mixture is
135 governed by interparticle friction, and the value of 𝛤 equals one. Theoretically, 𝜃$% is around 0.52 (Moore, 2014). 𝜏"# is a
threshold stress imparted by the frictional strength between debris particles, also depending upon the volumetric debris content
(𝜃$ ).
6
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Assuming that 𝜏"# ≪ 𝜏, 𝜃$ < 𝜃$% , and 𝛤 = 1, Eq. 2 can be reduced to the following form (Monnier and Kinnard, 2016):
& !
𝜀̇ = 3 4 , (3)
'
!
( +
"
140 where 𝐵 is the effective viscosity and is equal to 3 4 .
)*

Following a common setup in glaciology (Cuffey and Paterson, 2010), we consider each rock glacier as a slab with uniform
width and thickness and a semi-elliptical cross-section, resting on a bed of constant slope. It consists of two layers: an active
layer and a permafrost core. The active layer is a mixture of debris and air, and the permafrost core consists of ice, water,
debris and air. Both layers are assumed as homogeneous. Movement of rock glaciers is caused by the steady creep of the
145 permafrost core in the plane parallel to the bed slope. The active layer moves passively along with the inner core, which has
been validated by observations (Arenson et al., 2002; Haeberli, 2000). From Eq. 3 and the structure and geometry illustrated
in Fig. 3, we have:
$, & !
$-
= 2 3' 4 , (4)
$,
where $- is the velocity derivative relative to the depth 𝑧 in the permafrost core.

150 At a given depth 𝑧, the driving stress 𝜏 is imparted, taking into account the loading of the above material and the effect of
frictional drag occurring between the lateral margins and surrounding bedrocks, which is represented by a shape factor 𝑆.
(Cuffey and Paterson, 2010):
𝜏(𝑧) = 𝑆. sin𝛼(𝜌/0 𝑔ℎ/0 + 𝜌%123 𝑔𝑧) , (5)
where 𝛼 is the slope angle; 𝑔 is the gravitational acceleration; 𝜌/0 and 𝜌%123 are the densities of the active layer and the
155 permafrost core, respectively; ℎ/0 is the active layer thickness.
The shape factor is expressed as (Oerlemans, 2001):
4 6
𝑆. = arctan 3574 , (6)
5

where 𝑊 and 𝑇 are the width and thickness of the rock glacier, respectively.
The integration of the velocity profile (Eq. 4 and 5) is expressed as:
- 8# 9 :;< = ! -
160 ∫> 𝑑𝑢 = −2 3 '
4 ∫> (𝜌/0 ℎ/0 + 𝜌%123 𝑧)! 𝑑𝑧, (7)
!
5(A$% #$% B A&'() -)"*! 8# 9 :;< =
𝑢(𝑧) = 𝑢? − 3 4 , (8)
A&'() (!B() '

where 𝑢? is the surface velocity as illustrated in Fig. 3. When 𝑧 is set as the thickness of the ice core (ℎ%123 ) and basal sliding
is assumed to be absent, 𝑢? is then expressed as:
!
5(A$% #$% B A&'() #&'() )"*! 8# 9 :;< =
𝑢? = (!B()
3 4 , (9)
A&'() '

165 The densities of the active layer (𝜌/0 ) and the permafrost core (𝜌%123 ) are given as:
𝜌/0 = 𝜃$,/0 𝜌$ + 𝜃/,/0 𝜌/ , (10)

7
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

𝜌%123 = 𝜃$,%123 𝜌$ + 𝜃/,%123 𝜌/ + 𝜃F,%123 𝜌F + 𝜃G,%123 𝜌G , (11)

where 𝜃$,/0 and 𝜃/,/0 are the volumetric contents of debris and air in the active layer, respectively. The volumetric contents of
the components in the inner core, namely debris, air, ice and water, are expressed as 𝜃$,%123 , 𝜃/,%123 , 𝜃F,%123 , and 𝜃G,%123 ,
170 respectively. 𝜌$ , 𝜌/ , 𝜌F , and 𝜌G are the densities of debris, air, ice, and water, respectively.
For the flow law exponent (𝑛), we first used an empirical average value as assumed in numerous glaciological studies:
𝑛=3, (12)
We also adopted a linear relationship between 𝑛 and the volumetric ice content (𝜃F,%123 ) based on laboratory experiments
undertaken on borehole samples from two rock glaciers (Arenson and Springman, 2005a):
175 𝑛 = 3𝜃F,%123 , (13)

Figure 3: Schematic geometry, structure, stress status, and composition of rock glaciers. The rock glacier consists of a permafrost
core underlying the active layer. Parameters involved in the model include surface slope (𝜶), active layer thickness (𝒉𝒂𝒍 ), thickness
of permafrost core (𝒉𝒄𝒐𝒓𝒆 ), driving stress at the base of the active layer (𝝉𝟎 ), driving stress at depth z (𝝉𝒛 ), surface velocity (𝒖𝒔 ),
180 velocity at depth z (𝒖𝒛 ). 𝜽𝒅,𝒂𝒍 and 𝜽𝒂,𝒂𝒍 refer to the debris fraction and air fraction of the active layer. 𝜽𝒅,𝒄𝒐𝒓𝒆 , 𝜽𝒊,𝒄𝒐𝒓𝒆 , 𝜽𝒘,𝒄𝒐𝒓𝒆 , and
𝜽𝒂,𝒄𝒐𝒓𝒆 are the fractions of debris, ice, water, and air in the permafrost core, respectively.

8
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

3.2.2 Model calibration

Combining Eq. 9–11 with Eq. 12 or 13, we formulated several models depicting the relationship between the surface velocity
and properties of rock glaciers, including their composition, structure, and geometry. We then calibrated the models to
185 determine the curve of best fit between the effective viscosity (𝐵) and the volumetric ice content (𝜃F,%123 ) using observational
data of Las Liebres rock glacier in Central Chilean Andes (collected by Monnier and Kinnard, 2015b). This dataset includes
information of structure (ℎ%123 and ℎ/0 ), geometry (𝛼 and 𝑆. ), and composition (𝜃$,%123 , 𝜃/,%123 , 𝜃F,%123 , and 𝜃G,%123 ), all of
which were derived from Ground Penetrating Radar (GPR) measurements. Surface velocities (𝑢? ) were provided by a
Differential Global Positioning System (DGPS) along the central creep line at 14 locations on Las Liebres rock glacier,
190 (detailed in Monnier and Kinnard, 2015b & 2016).
First, we adopted the exponential 𝐵– 𝜃F,%123 relation estimated by Monnier & Kinnard (2016) with the same dataset and a
constant creep parameter 𝑛 (Eq. 12) (Fig. 4a). Then by integrating the relationship between 𝑛 and ice content (Eq. 13), we
applied both a 2nd-degree polynomial regression model and an exponential regression model to determine the 𝐵– 𝜃F,%123
relationship (Fig. 4b, c). The polynomial regression model is used to capture the subtle increase in effective viscosity when
195 the ice fraction becomes larger at the ice-rich end. This trend is also depicted in the laboratory experiment conducted by
Arenson and Springman (2005a) as a parabolic relationship between the minimum axial creep strain rate and the volumetric
ice content. Finally, we obtained three candidate parameterisation schemes expressed as:
H
5(A$% #$% B A&'() #&'() )+ 8# 9 :;< =
𝑢? = 3 ,..!/ 4 , (14)
A&'() (!B() HIH>>3 0,&'()

2/0,&'() *! HM0,&'()
5(A$% #$% B A&'() #&'() ) 8# 9 :;< =
𝑢? = A&'() (!B()
MJ(KHLHIM , N , (15)
0,&'() +NILHINOM0,&'() BHH55OHJ

2/0,&'() *! 8# 9 :;< = HM0,&'()

5(A$% #$% B A&'() #&'() )
200 𝑢? = A&'() (!B()
3 34.,5/0,&'() 4 , (16)
I5(JN>I3

For simplicity, the parameterisation scheme proposed in Monnier & Kinnard (2016) is labelled as Scheme 1 (Eq. 14). The
parameterisation scheme considering the empirical relation between 𝑛 and 𝜃F,%123 (Eq. 13) and parameterisation scheme
derived from the polynomial and exponential relationship between 𝐵 and 𝜃F,%123 are marked as Scheme 2 and Scheme 3 (Eq.
15 and 16), respectively.

9
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

205
Figure 4: (a)–(c) Relationships between the ice fraction (𝜽𝒊,𝒄𝒐𝒓𝒆 ) and the effective viscosity (B) estimated from the three regression
equations and parameterisation schemes (Eq. 14, 15, and 16, respectively). (d) Relationship between debris fraction (𝜽𝒅,𝒄𝒐𝒓𝒆 ) and
water fraction (𝜽𝒘,𝒄𝒐𝒓𝒆 ). The observational data are derived from the GPR and DGPS measurements in Monnier and Kinnard
(2015b & 2016).

210 3.2.3 Model validation

The three parameterisation schemes (Eq. 14–16) were validated using observational data of four rock glaciers in the Swiss
Alps, namely Murtèl-Corvatsch, Gruben, Muragl, and Schafberg (Cicoira et al., 2019a; Arenson et al., 2002; Hoelzle et al.,
1998; Barsch et al., 1979). We simulated the surface velocity (𝑢? ) of each rock glacier by varying volumetric ice content

10
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

(𝜃F,%123 ) of the permafrost core and inferred its ice fraction by comparing the modelled velocity and the measured velocity
215 from Terrestrial Geodetic Surveys (PERMOS, 2019). We then referred to the previously estimated ice content of the selected
rock glaciers to validate our predicted results.
To derive the input parameters, we first outlined the boundaries of the four rock glaciers, from which their shapes and areal
extents can be extracted. An empirical relationship established by Brenning (2005b) was then applied to calculate the rock
glacier thickness (𝑇) from its areal extent (𝐴29 ):
220 𝑇 = 50𝐴29 >.5 , (17)
where the area (𝐴29 ) is in km2. The width of each glacier was quantified as the width of its minimum envelop rectangle. We
took the mean value of the active layer thickness obtained from borehole measurements in the PERMOS network as the input
parameter ℎ/0 for each rock glacier. The surface slope (𝛼) was calculated based on the SRTM DEM with a spatial resolution
of 30 m. Table 2 lists the values of the above parameters. The permafrost core thickness (ℎ%123 ) can be obtained by subtracting
225 ℎ/0 from the total thickness 𝑇 calculated using Eq. 17.
We assumed the volumetric ice content (𝜃F,%123 ) of the permafrost core to be between 40% and 100%, considering the
prerequisites of the modified ice–debris mixture flow law (Eq. 3) that the debris fraction (𝜃$,%123 ) should be less than the
threshold (𝜃$% ) (Sect. 3.2.1). We varied the ice content (𝜃F,%123 ) by 1% in each step to model the corresponding surface
velocities (𝑢? ). We fixed the air content in the permafrost core as 7.5%, which is a mean value of the air fraction in ice-rich
230 permafrost samples (Arenson and Springman, 2005b). At near 0 °C, the volumetric content of water (𝜃G,%123 ) displays a
positive correlation with the debris fraction (𝜃$,%123 ) (Monnier and Kinnard, 2016). Thus, we calculated the 𝜃$,%123 – 𝜃G,%123
correlation based on the data published in Monnier and Kinnard (2015b) and assumed the constitution of the selected rock
glaciers followed the same linear relationship (Fig. 4d). The debris density (𝜌$ ) was given as 2450 kg/m3 (Monnier and Kinnard,
2016). The density of air (𝜌/ ) is determined by the elevation of each rock glacier. The ice density (𝜌F ) is 916 kg/m3 and the
235 water density (𝜌G ) is 1000 kg/m3.
Table 2. Summary of the geometric and structural parameters used in the validation.

Rock glacier Area (𝑨𝒓𝒈 ) (km2) Width (W) (m) Active layer thickness (𝐡𝐚𝐥 ) (m) Surface slope (𝛂) (°)
Murtèl-Corvatsch 0.06487 29 3.0 16
Gruben 0.7422 47 2.0 10
Muragl 0.02666 24 4.5 12
Schafberg 0.02715 24 4.8 16

3.2.4 Sensitivity analysis

To explore how uncertainties of the input parameters contribute to the final output of the developed approach, we tested the
response of the model to varying input parameters by performing a series of synthetic sensitivity experiments. For these
240 experiments, we simulated surface velocities of the rock glacier with varying ice fractions and inferred the current ice content

11
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

from the velocity constraint. The parameters explored here are detailed in Table 3. A reference scenario is set up with the
parameters of Murtèl-Corvatsch rock glacier and labelled as Sc-1.0. We designed eight scenarios extending from Sc-1.0,
naming each scenario after a multiplication factor which indicates the ratio between the parameter in each scenario and that in
the reference scenario; with the exception of two parameters, namely debris density (ρQ ) and debris fraction in the active layer
245 (θQ,RS ), where we changed the upper or lower boundary of the value range to be consistent with the usual value range in reality.
We performed the sensitivity experiments by varying one parameter at a time while keeping the other variables constant.
Table 3. Parameters of the sensitivity experiments. Scn-1.0 is the reference scenario that adopts the parameters of Murtèl-Corvatsch
rock glacier. The other scenarios are designed by multiplying the reference value of each variable with the corresponding factor in
their scenario labels.

Scenario 𝐀𝐫𝐠 W 𝜶 𝒉𝒂𝒍 𝛒𝐝 𝛉𝐝,𝐚𝐥 𝛉𝐚,𝐜𝐨𝐫𝐞

(km ) 2 (m) (°) (m) (kg/m3) (%) (%)
Scn-0.2 0.01297 40 3.2 0.6 1450 13 1.5
Scn-0.4 0.02594 80 6.4 1.2 1700 26 3.0
Scn-0.6 0.03892 120 9.6 1.8 1950 39 4.5
Scn-0.8 0.05189 160 12.8 2.4 2200 52 6.0
Scn-1.0 0.06487 200 16 3.0 2450 65 7.5
Scn-1.2 0.07784 240 19.2 3.6 2700 72 9.0
Scn-1.4 0.09081 280 22.4 4.2 2950 79 10.5
Scn-1.6 0.10379 320 25.6 4.8 3200 86 12.0
Scn-1.8 0.11677 360 28.8 5.4 3450 93 13.5

250 3.2.5 Model application

We applied the validated model with the optimal parameterisation scheme to infer ice content of the coherently moving part
of five rock glaciers in the Khumbu and Lhotse Valleys. The geometric and structural data used as input parameters are detailed
in Table 4. Area, width, and slope angle are quantified using the same method as described in Sect. 3.2.3. Active layer thickness
was determined as the mean value over the extent of each rock glacier during 2006–2017 from the European Space Agency
255 Permafrost Climate Change Initiative Product (ESA CCI) (Obu et al., 2020). The same empirical relation for calculating rock
glacier thickness as used in the validation procedure was adopted here to obtain the thickness parameter. The surface velocity
constraint is the range of InSAR-derived downslope velocity during the observed period (Sect. 3.2.2); except for Tobuche RG
where the abnormal value in 2015 is removed from the range (see Sect. 4.1 for details). Finally, we calculated the water volume
equivalent to estimate the amount of water stored in rock glaciers by considering the inferred ice content, areal extent, and
260 permafrost core thickness.

12
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Table 4. Summary of the geometric and structural parameters used for inferring ice content of the coherently moving part of rock
265 glaciers in the study area.

Rock glacier Area (𝐀𝐫𝐠 ) (km2) Width (W) (m) Active layer thickness (𝐡𝐚𝐥 ) (m) Surface slope (𝛂) (°)
Kala-Patthar 0.074 240 0.68 9
Kongma 0.077 300 0.83 13
Lingten 0.094 240 0.65 20
Nuptse 0.234 400 0.30 13
Tobuche 0.128 400 1.67 16

4 Results

In this section we first summarise the surface kinematic characteristics of rock glaciers in Khumbu and Lhotse Valleys
measured by InSAR. Then we show the results of model validation and sensitivity experiments. Finally, we present the inferred
ice contents and estimated water storage of rock glaciers in the study area.

270 4.1 InSAR-derived surface kinematics of rock glaciers

We used InSAR to derive the downslope surface velocities of five rock glaciers situated in the study region. Surface kinematics
of debris-covered glaciers were also quantified and presented in Fig. S1 and S2 in the supplementary materials.
Figure 5 shows the time series of the InSAR-derived surface velocities of the rock glacier coherently moving parts. We observe
that the median and mean velocities of each landform have similar values, and both are capable of characterising the kinematic
275 status of the landforms. By selecting the mean velocity as the representative value, most rock glaciers, except for Tobuche,
moved at a nearly stable rate, ranging from 5 cm yr-1 to 30 cm yr-1 during the observational period, with the largest standard
deviation being 3.4 cm yr-1 for Lingten. The maximum velocity represents the local extreme of downslope motion and was as
high as 112.1±12.4 cm yr-1 for Lingten during 2019/07/15–2019/08/26. Tobuche displayed similar stable behaviour before
2010 but had accelerated by more than four times from 14.9±0.2 cm yr-1 to 81.4±2.4 cm yr-1 since 2010. The maximum velocity
280 reached was 181.0±57.4 cm yr-1 for the period 2015/03/18–2015/03/22. However, the associated uncertainties during this
period were high: the relative uncertainties of mean, median, and maximum velocity were 2.9%, 38.2%, and 31.7%,
respectively. Therefore, the acceleration of Tobuche cannot be confidently revealed by our data. The extents of coherently
moving parts of the five rock glaciers are presented in Fig. 6, with the average velocities derived from the interferograms
obtained during the past several years.

13
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

285
Figure 5: Time series of the InSAR-derived downslope velocities of the landforms. The spatial mean velocities and uncertainties
during each period are shown (red squares and error bars) as well as the median (blue) and maximum (orange) velocities.

14
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Figure 6: Velocity field maps show the average movement rates of the coherently moving parts of the five rock glaciers (blue outlines)
290 in the study area. The boundaries of the landforms delineated in previous inventorying work are in red polygons. The background
are Google Earth images.

4.2 Model validation

With the input parameters and model setup as detailed in Sect. 3.2.3, we simulated the surface velocities (𝑢? ) of each rock
glacier using Schemes 1–3. Uncertainties, generated through the statistical analysis used to establish the model (as shown in
15
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

295 Fig. 4), have all been considered in the simulation. We used the annual mean surface velocities, calculated from the Terrestrial
Ground Survey data (PERMOS, 2019), as the constraint for inferring the ice content.
For each rock glacier, an inferred ice content range is derived based on the velocity constraint and modelled 𝑢? – 𝜃F,%123
relationship. The median of the range is selected as the inferred ice content and compared with the reference ice content, which
is taken as the average value of the estimated ice content based on previous field measurements (Cicoira et al., 2019a; Arenson
300 et al., 2002; Hoelzle et al., 1998; Barsch et al., 1979).
Comparing the reference and inference ice content from the three schemes, Scheme 2 is the optimal one for the following two
reasons: (1) the reference ice content is within the range inferred from Scheme 2 (Fig. 8); (2) Scheme 2 gives the smallest
average bias (8.4%) compared with Scheme 1 (12.9%) and Scheme 3 (13.3%) (Table 5). However, the above bias is not
statistically useful for correcting the modelling results due to the limited amount of validation data.

16
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

305
Figure 7: Modelled relationships (grey shaded areas) between the ice fraction (𝜽𝒊,𝒄𝒐𝒓𝒆 ) and the surface velocity (us) of 95% confidence
intervals for the four RGs monitored in the PERMOS network with model parameterisation Scheme 1. The ranges of the observed
velocities (yellow bands) are used as velocity constraints for inferring ice content from the modelled relationships. Also shown are
the reference ice content obtained from previous field-based surveys (blue lines). The inference ice contents are the mean values
310 (solid black lines) with the estimated ranges (dash-dotted black lines).

17
https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Figure 8: Similar to Fig. 7, but showing results obtained based on model parameterisation Scheme 2.

18
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Figure 9: Similar to Fig. 7 and 8, but showing results derived from model parameterisation Scheme 3.

315

19
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

320 Table 5. Summary of the reference and inference ice contents derived from the three model parameterisation schemes. The values
in brackets following the inference ice contents give the corresponding bias from the reference ice contents.

Rock glacier 𝐑𝐞𝐟𝐞𝐫𝐞𝐧𝐜𝐞 (%) Inference and bias

Scheme 1(%) Scheme 2 (%) Scheme 3 (%)
Murtèl-Corvatsch 85 91 (7) 75 (–10) 81 (–4)
Gruben 60 85 (25) 71 (11) 80 (20)
Muragl 50 56 (6) 59 (9) 66 (16)
Schafberg 65 79 (14) 69 (4) 78 (13)
Mean bias – 12.9 8.4 13.3

4.3 Model sensitivity

The results of sensitivity experiments are shown, normalised to the corresponding values of the reference scenario (Scn-1.0),
in Fig. 10. We observe that the inference result remains stable in response to most varying parameters, with a bias of less than
325 5%, relative to the reference scenario (Scn-1.0). Surface slope angle influences the result most: in the extreme scenario (Scn-
0.2), the inferred ice content can be altered by 15%. In non-extreme cases (e.g., Scn-0.8, Scn-0.6), the influences of varying
slope angles can be well constrained within the 5% range. In general, the model is insensitive to the uncertainties of any single
input parameter.

330 Figure 10: Normalised inference ice fractions from sensitivity experiments with different parameter scenarios. The varying
parameters include rock glacier area (blue line), width (orange line), surface slope (green line), active layer thickness (brown line),
debris density (purple line), debris fraction in the active layer (pink line), and air faction in permafrost core (black line).

20
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

4.4 Modelled ice contents

Figure 11 and Table 6 present the inference ice contents of rock glaciers in the study area. The inferred average ice fractions
335 of the landforms are in the range of 71.0–75.3%; the water volume equivalent ranges from 1.40 to 5.92 million m3 for individual
landforms. The maximum range of ice fraction is estimated to be 57.5–92.0%; the corresponding water volume equivalent
ranges from 1.13 to 7.24 million m3. Nuptse stores the most ice by volume due to its largest dimensions. The total amount of
water stored in each of the five rock glaciers lies between 10.61 and 16.54 million m3, with an average value of 13.57 million
m3.

21
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

340
22
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Figure 11: Modelled relationships between the ice fraction (𝜽𝒊,𝒄𝒐𝒓𝒆 ) and the surface velocity (𝒖𝒔 ) of 95% confidence intervals for the
five RGs in Khumbu Valley with model parameterisation Scheme 2 (grey shaded areas). The ranges of the InSAR-derived velocities
are shown (yellow bands), which are used as the velocity constraints for inferring ice contents from the modelled relationships. The
upper and lower boundaries of the estimated ice contents are within the range outlined by the dash-dotted black lines and the solid
345 black lines show the mean values representing the inference ice contents.

Table 6. Modelled average ice contents, as well as the minimum and maximum estimates (in brackets) of rock glaciers in Khumbu
and Lhotse Valleys and the corresponding water volume equivalents.

Rock glacier Inference ice content (%) Water volume equivalent (million m3)
Kala-Patthar 71.0 (57.5–84.5) 1.40 (1.13–1.66)
Kongma 73.3 (58.0–88.5) 1.50 (1.19–1.82)
Nuptse 74.5 (58.0–91.0) 5.92 (4.61–7.24)
Lingten 75.3 (58.5–92.0) 1.98 (1.54–2.42)
Tobuche 75.0 (58.0–92.0) 2.77 (2.14–3.40)

5 Discussion

The following subsections firstly investigate the potential water storage of rock glaciers over the Nepalese Himalaya by
350 extrapolating the estimated water storage in the Khumbu and Lhotse Valleys (Sect. 5.1). We then discuss the validity of the
approach developed in this study for inferring ice content by analysing the general assumptions and design of the surface-
velocity-constrained model (Sect. 5.2), based upon which we discuss the application of the method to large-scale regions and
further improvements (Sect. 5.3).

5.1 Potential water storage in rock glaciers in the Nepalese Himalaya

355 The inferred average ice content of the five rock glaciers in the study area lies within a narrow range (71.0–75.3%), mainly
due to their similar observed downslope velocities (5–30 cm yr-1), used as modelling constraints (Fig. 5; Fig. 6; Sect. 4.1). In
general, rock glaciers typically creep at a rate ranging from decimetre to several meters per year (Delaloye and Echelard, 2020),
thus the average ice content of the five rock glaciers may not be able to represent the motion of all rock glaciers situated in the
entire mountain range.
360 A previous study has compiled an inventory including 4226 intact rock glaciers over the Nepalese Himalaya, and a first-order
estimate has indicated that these landforms contain between 16.72 and 25.08 billion m3 of water (Jones et al., 2018b). By
extrapolating from the estimated water storage in the five rock glaciers found in this study, the total amount of water stored in
all the intact rock glaciers ranges from 8.97 to 13.98 billion m3 over the entire Nepalese Himalaya, which is in the same
magnitude predicted by the previous research (Jones et al., 2018b). In the Nepalese Himalaya, the ratio between the amount
365 of water stored in rock glaciers (11.47 billion m3) and in glaciers (197.63 billion m3) is 1:17. By using the minimum and
maximum inference values, the estimated ratio ranges between 1:21 and 1:13. Our modelling-based results are lower than

23
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

earlier estimates (1:9), yet reveal higher hydrological importance than across the Himalayas (1:24) (Jones et al., 2018b; Jones
et al., 2021).

5.2 Validity of general assumptions and model design

370 Our aim was to develop a rheological model that allows for inferring ice content of rock glaciers in areas where in situ
investigations are scarce. To achieve this objective, certain simplifications have been applied to the model setup. Here we
discuss the validity of the method through the following six aspects of the model assumptions and design, including (1)
assumption of a steady-state rock glacier creep, (2) homogeneous warm permafrost hypothesis, (3) neglect of shear horizon in
the model design, (4) accuracy of rock glacier thickness derivation, (5) identification of the coherently moving part of rock
375 glaciers, and (6) generalisation of statistical relationships derived from Las Liebres rock glacier.

5.2.1 Steady-state creep of rock glaciers

By using the adapted form of Glen’s flow law (Eq. 2), we primarily assume the rock glacier movement to be steady-state creep
driven by viscoelastic deformation of the ice–debris mixture (Moore, 2014). This premise indicates that our method is
applicable to rock glaciers currently moving at a relatively stable rate. Recent research has reported abrupt and significant
380 acceleration of rock glaciers triggered by abnormal surface warming events (Delaloye et al., 2013). These destabilised rock
glaciers are beyond the applicability of our method. In this study, we quantified surface kinematics of rock glaciers over
multiple years to quantify the stability of the rock glacier motion. The seasonal variations in creep rate are neglected and
sudden acceleration events are excluded in the velocity range used as the model constraint (Sect. 4.1).
We also exclude the component of basal sliding processes in our model design (Fig. 3), which operates at the base of some
385 rock glaciers, as observed in the Tien Shan (Harrison unpubl.). This is not surprising; many clean glaciers undergo basal sliding
in certain situations (e.g., Vivian and Bocquet, 1973) although, a debris-rich layer with high water content at the base undergoes
enhanced deformation and some sliding (e.g., Boulton and Jones, 1979; Echelmeyer and Wang, 1987). This process also occurs
in rock glaciers with a high ice content and is accompanied by the disruption of sediment and vegetation at the front of such
features. The kinematics of these rock glaciers cannot be appropriately simulated by our current approach.

390 5.2.2 Homogeneous warm permafrost

Ground temperature is one of the factors controlling the creep parameter (𝐴) (Eq. 1), as described by the Arrhenius relation
(Mellor and Testa, 1969). As ground temperature changes with depth, primarily due to heat diffusion, creep parameter (𝐴)
varies along the vertical profile of the rock glacier. Previous studies implemented different relationships between creep
parameter and temperature, and integrated a heat diffusion model (proposed by Carslaw and Jaeger, 1959) to consider this
395 effect (Azizi and Whalley, 1996; Arenson and Springman, 2005a; Kääb et al., 2007; Ladanyi, 2003).
In our model design, we use the effective viscosity (𝐵) to absorb the intricate effects of strain enhancement factor (𝐸), threshold
stress (𝜏"# ), and most importantly, the creep parameter (𝐴) (Eq. 3), which reduces the number of input parameters and allows
24
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

for developing an empirical relationship between effective viscosity and ice content based on an existing observational dataset
and laboratory findings (Eq. 13). The data and relationship between variables we used for model calibration are derived from
400 observations of rock glaciers situated in a warm permafrost environment (>–3°C) (Monnier and Kinnard, 2016; Arenson and
Springman, 2005a)
Measurements of ground temperature in the study area are scarce in general. However, we infer that these rock glaciers develop
in a warm permafrost environment for the following reasons: (1) the landforms are located near or below the altitudinal limit
of permafrost distribution in Nepal (Sect. 2) (Fujii and Higuchi, 1976; Jakob, 1992), indicating that the local environment is
405 at the critical condition of permafrost occurrence; (2) based on empirical relationships between mean annual ground
temperature (MAGT), mean annual air temperature, latitude, and altitude, the estimated MAGT is >0.5°C, which suggests that
permafrost in this area is in a warm and unstable condition (Nan et al., 2002; Zhao and Sheng, 2015).

5.2.3 Neglect of shear horizon

The shear horizon is discovered from borehole investigations and is defined as the thin layer situated at more than ten meters
410 deep where the majority of internal deformation takes place (Arenson et al., 2002; Buchli et al., 2018; Haeberli et al., 1998).
Field observations and numerical modelling suggest that unfrozen water within the shear horizon plays an important role in
controlling the seasonal variations in rock glacier creep (Buchli et al., 2018; Cicoira et al., 2019b; Kenner et al., 2019). This
short-term feature of rock glacier kinematics is insignificant to modelling the relationship between ice content and multi-annual
average movement velocity in our study.
415 Previous studies have considered the enhanced deformation occurring in the shear horizon additionally, but it requires detailed
knowledge of the internal structure, i.e., the depth of shear horizon (Frehner et al., 2015; Ladanyi, 2003). To tackle the issue
of data insufficiency of internal rock glacier structure in this study area – as with most permafrost areas – we neglect the
distinct rheology in the shear horizon and assume a constant effective viscosity instead. This simplification has also been
adopted in other research aiming at studying rock glacier dynamics over a large-scale extent (Cicoira et al., 2020).

420 5.2.4 Accuracy of rock glacier thickness derivation

The accuracy of rock glacier thickness is discussed here because it influences the surface kinematics most significantly. As
shown in Eq. 8, the surface velocity is proportional to the thickness to the power of 𝑛 + 1, resulting from the vertical integration
of Eq. 7. We adopt the empirical relationship between rock glacier area and thickness (Eq. 17) based on field observations in
the Andes (Brenning, 2005a). The derived thicknesses of the four rock glaciers in the Swiss Alps, used for validation, are
425 consistent with previous in situ measurements. However, another rock glacier, namely Ritigraben, situated in the same region,
does not follow this empirical relationship and has a bias as large as ten meters compared with the field estimates. We also test
a linear relationship between surface slope angle and thickness, recently established by Cicoira et al. (2020). The accuracies
of the results turn out to be at the same level, with four out of five rock glaciers having good estimation results (Table 7). Thus,

25
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

the uncertainty introduced by thickness derivation when applied to rock glaciers without known information of structure is
430 unavoidable.
Table 7. Estimated rock glacier thickness derived from the thickness–area relationship used in this study (𝑻𝒂𝒓𝒆𝒂 ) (Eq. 17), and the
corresponding bias relative to in situ measured thickness (Barsch et al., 1979; Cicoira et al., 2019a; Arenson et al., 2002; Hoelzle et
al., 1998). The rock glacier thickness derived from slope angle (𝑻𝒔𝒍𝒑 ) proposed by Cicoira et al. (2020), and the associated bias.

Rock glacier 𝑻𝒂𝒓𝒆𝒂 and bias (m) 𝑻𝒔𝒍𝒑 and bias (m) 𝑻𝒓𝒆𝒇 (m)
Murtèl-Corvatsch 29 (2) 26.2 (–0.8) 27
Gruben 47 (–3) 28 (–22) 50
Muragl 24 (4) 19 (–1) 20
Schafberg 24 (–1) 20.8 (–4.2) 25
Ritigraben 28 (10) 12.7 (–5.3) 18

5.2.5 Identification of the coherently moving part of rock glaciers

435 Taking advantage of the multi-temporal observations and continuous spatial coverage of the InSAR measurements, we define
the coherently moving part of each rock glacier in our study area (Fig. 6; Sect. 3.1) and infer the ice content of the coherently
moving part using the developed modelling approach. We introduce this concept because it corresponds with the general model
setup (Fig. 3). Moreover, with the assistance of displacement maps generated by InSAR, the defined boundary of coherently
moving rock glacier avoids the ambiguities involved when delineating rock glaciers solely based on geomorphologic features
440 from a highly dynamic environment where complex glacial, periglacial, and paraglacial processes take place (Jones et al.,
2019; Delaloye and Echelard, 2020). In addition, defining the coherently moving part associates with the uncertainty in the
rock glacier area, which is unlikely to affect the modelling result significantly due to the insensitive response of our model to
this input parameter (Fig. 10). Finally, employing this kinematics-based definition, may also contribute to the currently
inevitable uncertainties in thickness estimation (Sect. 5.2.4), as Eq. 17 (proposed by Brenning, 2005b) uses rock glacier area
445 as a variable, without considering the landform activity.

5.2.6 Generalisation of statistical relationships derived from Las Liebres rock glacier

The model developed in this study essentially relies on statistical relationships, especially the effective viscosity–ice content
relationship, derived from geophysical surveys conducted on Las Liebres rock glacier in the Andes (Monnier and Kinnard,
2015b). Applying the calibrated model to rock glaciers in other areas is primarily based on the assumption that a common flow
450 law governs the rheology of rock glaciers developed in warm permafrost environment, irrespective of locality. In addition to
the general hypothesis, we tackled this issue in two ways. First, we validated the model using samples from a different region,
i.e., the Swiss Alps (Sect. 3.2.3). Secondly, the uncertainties introduced by the statistical analysis have all been quantified in
the validation and inference procedures, which leads to a wide range of the inference ice contents (Sect. 3.2.3 and 4.4).

26
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

5.3 Application of the model and outlook

455 By adopting the assumptions and setup as discussed in Sect. 5.2, we firstly present an approach for estimating ice content of
rock glaciers with simple input parameters, including the planar shape (area and width), surface slope angle, active layer
thickness, and surface velocity, all of which can be obtained directly or derived from remote sensing techniques and products.
Therefore, this established method for estimating the amount of ice stored in numerous rock glaciers with well-quantified
uncertainties is ready to be applied to many remote alpine environments.
460 To improve the performance of the approach, more data obtained from field and geophysical investigations, especially detailed
data of rock glacier composition, can be integrated in the future to calibrate and validate the empirical rheological model. More
reliable methods for estimating rock glacier thickness will also improve the accuracy of the modelling results.

6 Conclusions

We develop an empirical rheological model for inferring ice content of rock glaciers and apply it to estimate the water storage
465 of rock glaciers situated in the Khumbu and Lhotse Valleys using surface-velocity-constraints derived from InSAR
measurements. The main findings are summarised as follows:
(1) An empirical rheological model is presented in this study for estimating ice content of rock glaciers using five input
parameters, namely rock glacier area, width, surface slope angle, active layer thickness, and surface velocity, all of which can
be obtained from readily available remote sensing products or emerging datasets.
470 (2) Mean downslope velocities of the rock glaciers situated in Khumbu and Lhotse Valleys ranged from 5 cm yr-1 to 30 cm yr-
1
and mostly remained stable during the observational period (2006–2020).
(3) The inferred average ice contents of rock glaciers in Khumbu and Lhotse Valleys are in the range of 71.0–75.3%; the water
volume equivalent ranges from 1.40 to 5.92 million m3 for individual landforms. Nuptse RG stores the most ice due to its
largest dimensions among the five studied rock glaciers.
475 (4) The inference range of ice content of the landforms lies between 57.5 and 92.0%. Total amount of water stored in the five
rock glaciers in Khumbu and Lhotse Valleys ranges from 10.61 to 16.54 million m3, with an average value of 13.57 million
m3.
(5) Considering previous estimates and extrapolating from our inference results in Khumbu and Lhotse Valleys, the total
amount of water stored in rock glaciers over the Nepalese Himalaya is in the magnitude of 10 billion m3, and the ratio between
480 water storage in rock glaciers and glaciers ranges from 1:13 to 1:21, averaging at 1:17.
This study demonstrates the effectiveness of inferring ice content of rock glaciers by using a surface-velocity-constrained
model. The estimated ice content and water storage in the study area highlights the hydrological significance of rock glaciers
in the Nepalese Himalaya. The approach developed here can be readily applied to other alpine permafrost regions where rock
glaciers are widespread for a preliminary water resource evaluation.

27
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

485 Code and data availability

The source code of ISCE is available at https://github.com/isce-framework/isce2. The ALOS PALSAR and ALOS-2 PALSAR-
2 data are copyrighted and provided by the Japan Aerospace Exploration Agency through the EO-RA2 project ER2A2N081.
Data for the rock glacier kinematics in the Swiss Alps are available at http://www.permos.ch/data.html. The ESA CCI
permafrost data are available at http://catalogue.ceda.ac.uk/uuid/1f88068e86304b0fbd34456115b6606f. The code of the
490 modelling approach for estimating ice content will be provided by Yan Hu upon request.

Author contribution

YH developed the code, performed the data analysis and interpretation, visualised the results, and wrote the majority of the
manuscript. SH conceptualised the research goal, supervised the study, and wrote Sect. 1 of the draft. LL advised YH and
actively helped the investigation process. JLW helped formulate the initial framework of the method and collect research data.
495 All the authors contributed to the reviewing and editing of the manuscript.

Competing interests

The authors declare that they have no conflict of interest.

Acknowledgements

We thank Juliet Ermer for helping to digitalize landform boundaries used in this work.

500 Financial support

This work is supported by CUHK Global Scholarship Programme, CUHK-Exeter Joint Centre for Environmental
Sustainability and Resilience (ENSURE, 4930821), the Hong Kong Research Grants Council (CUHK14303417 and HKPFS
PF16-03859), and CUHK Direct Grant for Research (405348).

References

505 Arenson, L., Hoelzle, M., and Springman, S.: Borehole deformation measurements and internal structure of some rock glaciers in
Switzerland, Permafrost Periglac., 13, 117-135, https://doi.org/10.1002/ppp.414, 2002.

Arenson, L., and Springman, S.: Mathematical descriptions for the behaviour of ice-rich frozen soils at temperatures close to 0 C, Can.
Geotech. J., 42, 431-442, https://doi.org/10.1139/t04-109, 2005a.

28
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Arenson, L., and Springman, S.: Triaxial constant stress and constant strain rate tests on ice-rich permafrost samples, Can. Geotech. J., 42,
510 412-430, https://doi.org/10.1139/t04-111, 2005b.

Azizi, F., and Whalley, W. B.: Numerical modelling of the creep behaviour of ice-debris mixtures under variable thermal regimes, in:
International Offshore and Polar Engineering Conference Proceedings, The Sixth International Offshore and Polar Engineering Conference,
Los Angeles, 1996, WOS:A1996BF95F00053, 362-366,

Azócar, G. F., and Brenning, A.: Hydrological and geomorphological significance of rock glaciers in the dry Andes, Chile (27°-33°S),
515 Permafrost Periglac., 21, 42-53, https://doi.org/10.1002/ppp.669, 2010.

Barsch, D., Fierz, H., and Haeberli, W.: Shallow core drilling and borehole measurements in the permafrost of an active rock glacier near
the Grubengletscher, Wallis, Swiss Alps, Arctic Alpine Res., 11, 215-228, https://doi.org/10.2307/1550646, 1979.

Berthling, I., Etzelmüller, B., Isaksen, K., and Sollid, J. L.: Rock glaciers on Prins Karls Forland. II: GPR soundings and the development
of internal structures, Permafrost Periglac., 11, 357-369, https://doi.org/10.1002/1099-1530, 2000.

520 Berthling, I.: Beyond confusion: Rock glaciers as cryo-conditioned landforms, Geomorphology, 131, 98-106,
https://doi.org/10.1016/j.geomorph.2011.05.002, 2011.

Boulton, G. S., and Jones, A. S.: Stability of temperate ice caps and ice sheets resting on beds of deformable sediment, J. Glaciol., 24, 29-
43, https://doi.org/10.3189/S0022143000014623, 1979.

Brenning, A.: Geomorphological, hydrological and climatic significance of rock glaciers in the Andes of Central Chile (33-35 degrees S),
525 Permafrost Periglac., 16, 231-240, https://doi.org/10.1002/ppp.528, 2005a.

Brenning, A.: Climatic and geomorphological controls of rock glaciers in the Andes of Central Chile, Humboldt-Universität zu Berlin,
Mathematisch-Naturwissenschaftliche Fakultät II, 2005b.

Buchli, T., Kos, A., Limpach, P., Merz, K., Zhou, X. H., and Springman, S. M.: Kinematic investigations on the Furggwanghorn Rock
Glacier, Switzerland, Permafrost Periglac., 29, 3-20, https://doi.org/10.1002/ppp.1968, 2018.

530 Carslaw, H. S., and Jaeger, J. C.: Conduction of heat in solids, Oxford Science Publications. Clarendon Press, Oxford, 1959.

Chen, C. W., and Zebker, H. A.: Phase unwrapping for large SAR interferograms: statistical segmentation and generalized network models,
IEEE T. Geosci Remote., 40, 1709-1719, https://doi.org/10.1109/TGRS.2002.802453, 2002.

Cicoira, A., Beutel, J., Faillettaz, J., Gartner-Roer, I., and Vieli, A.: Resolving the influence of temperature forcing through heat conduction
on rock glacier dynamics: a numerical modelling approach, The Cryosphere, 13, 927-942, https://doi.org/10.5194/tc-13-927-2019, 2019a.

29
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

535 Cicoira, A., Beutel, J., Faillettaz, J., and Vieli, A.: Water controls the seasonal rhythm of rock glacier flow, Earth Planet Sc. Lett., 528,
https://doi.org/10.1016/j.epsl.2019.115844, 2019b.

Cicoira, A., Marcer, M., Gärtner‐Roer, I., Bodin, X., Arenson, L. U., and Vieli, A.: A general theory of rock glacier creep based on in‐situ
and remote sensing observations, Permafrost Periglac., https://doi.org/10.1002/ppp.2090, 2020.

Corte, A.: The Hydrological Significance of Rock Glaciers, J. Glaciol., 17, 157-158, https://doi.org/10.3189/s0022143000030859, 1976.

540 Croce, F. A., and Milana, J. P.: Internal structure and behaviour of a rock glacier in the Arid Andes of Argentina, Permafrost Periglac., 13,
289-299, https://doi.org/10.1002/ppp.431, 2002.

Cuffey, K., and Paterson, W. S. B.: The physics of glaciers, 4th ed., Butterworth-Heinemann/Elsevier, Burlington, MA, 2010.

Delaloye, R., Morard, S., Barboux, C., Abbet, D., Gruber, V., Riedo, M., and Gachet, S.: Rapidly moving rock glaciers in Mattertal, Geogr.
Helv., 21-31, 2013.

545 Delaloye, R., and Echelard, T.: Towards standard guidelines for inventorying rock glaciers: Baseline concepts version 4.0,
https://bigweb.unifr.ch/Science/Geosciences/Geomorphology/Pub/Website/IPA/Guidelines/V4/200117_Baseline_Concepts_Inventorying_
Rock_Glaciers_V4.pdf, 2020.

Echelmeyer, K., and Wang, Z.: Direct Observation of Basal Sliding and Deformation of Basal Drift at Sub-Freezing Temperatures, J.
Glaciol., 33, 83-98, https://doi.org/10.3189/S0022143000005396, 1987.

550 Florentine, C., Skidmore, M., Speece, M., Link, C., and Shaw, C. A.: Geophysical analysis of transverse ridges and internal structure at Lone
Peak Rock Glacier, Big Sky, Montana, USA, J. Glaciol., 60, 453-462, https://doi.org/10.3189/2014jog13j160, 2014.

Frehner, M., Ling, A. H. M., and Gärtner-Roer, I.: Furrow-and-ridge morphology on rockglaciers explained by gravity-driven buckle folding:
A case study from the Murtèl rockglacier (Switzerland), Permafrost Periglac., 26, 57-66, https://doi.org/10.1002/ppp.1831, 2015.

Fujii, Y., and Higuchi, K.: Ground Temperature and its Relation to Permafrost Occurrences in the Khumbu Region and Hidden Valley,
555 Journal of Japanese Society of Snow and Ice., 38, 125-128, https://doi.org/10.5331/seppyo.38.Special_125, 1976.

Fukui, K., Sone, T., Strelin, J. A., Torielli, C. A., and Mori, J.: Ground penetrating radar sounding on an active rock glacier on James Ross
Island, Antarctic Peninsula region, Pol Polar Res., 13-22, 2007.

Fukui, K., Sone, T., Strelin, J. A., Torielli, C. A., Mori, J., and Fujii, Y.: Dynamics and GPR stratigraphy of a polar rock glacier on James
Ross Island, Antarctic Peninsula, J. Glaciol., 54, 445-451, https://doi.org/10.3189/002214308785836940, 2008.

560 Glen, J. W.: The creep of polycrystalline ice, Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences
(1934-1990), 228, 519-538, https://doi.org/10.1098/rspa.1955.0066, 1955.

30
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Guglielmin, M., Camusso, M., Polesello, S., and Valsecchi, S.: An old relict glacier body preserved in permafrost environment: The Foscagno
rock glacier ice core (Upper Valtellina, Italian central Alps), Arctic AntArctic Alpine Res., 36, 108-116, https://doi.org/10.1657/1523-0430,
2004.

565 Guglielmin, M., Ponti, S., and Forte, E.: The origins of Antarctic rock glaciers: periglacial or glacial features?, Earth Surf Proc Land., 43,
1390-1402, https://doi.org/10.1002/esp.4320, 2018.

Haeberli, W., Hoelzle, M., Kaab, A., Keller, F., Vonder Mühll, D., and Wagner, S.: Ten years after drilling through the permafrost of the
active rock glacier Murtél, Eastern Swiss Alps: answered questions and new perspectives, Seventh International Conference on Permafrost,
Yellowknife, 1998, 403-410,

570 Haeberli, W., Kaab, A., Wagner, S., Vonder Mühll, D., Geissler, P., Haas, J. N., Glatzel-Mattheier, H., and Wagenbach, D.: Pollen analysis
and C-14 age of moss remains in a permafrost core recovered from the active rock glacier Murtel-Corvatsch, Swiss Alps: Geomorphological
and glaciological implications, J. Glaciol., 45, 1-8, https://doi.org/10.3189/s0022143000002975, 1999.

Haeberli, W.: Modern research perspectives relating to permafrost creep and rock glaciers: A discussion, Permafrost Periglac., 11, 290-293,
https://doi.org/10.1002/1099-1530(200012)11:4, 2000.

575 Hanssen, R. F.: Radar interferometry : data interpretation and error analysis, edited by: NetLibrary, I., Kluwer Academic, Dordrecht Boston,
2001.

Hauck, C.: New concepts in geophysical surveying and data interpretation for permafrost terrain, Permafrost Periglac., 24, 131-137,
https://doi.org/10.1002/ppp.1774, 2013.

Hausmann, H., Krainer, K., Bruckl, E., and Mostler, W.: Internal structure and ice content of reichenkar rock glacier (Stubai alps, Austria)
580 assessed by geophysical investigations, Permafrost Periglac., 18, 351-367, https://doi.org/10.1002/ppp.601, 2007.

Hoelzle, M., Wagner, S., Kääb, A., and Vonder Mühll, D.: Surface movement and internal deformation of ice-rock mixtures within rock
glaciers in the Upper Engadin, Switzerland, Proceedings of the 7th International Conference on Permafrost, 465-471, 1998.

Hu, Y., Liu, L., Wang, X., Zhao, L., Wu, T., Cai, J., Zhu, X., and Hao, J.: Quantification of permafrost creep provides kinematic evidence
for classifying a puzzling periglacial landform, Earth Surf Proc Land., 46, 465-477, https://doi.org/10.1002/esp.5039, 2021.

585 Jakob, M.: Active rock glaciers and the lower limit of discontinuous alpine permafrost, Khumbu-Himalaya, Nepal, Permafrost Periglac., 3,
253-256, 1992.

Jones, D. B., Harrison, S., Anderson, K., and Betts, R. A.: Mountain rock glaciers contain globally significant water stores, Sci Rep., 8,
https://doi.org/10.1038/s41598-018-21244-w, 2018a.

31
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Jones, D. B., Harrison, S., Anderson, K., Selley, H. L., Wood, J. L., and Betts, R. A.: The distribution and hydrological significance of rock
590 glaciers in the Nepalese Himalaya, Global Planet Change., 160, 123-142, https://doi.org/10.1016/j.gloplacha.2017.11.005, 2018b.

Jones, D. B., Harrison, S., and Anderson, K.: Mountain glacier-to-rock glacier transition, Global Planet Change., 181,
https://doi.org/10.1016/j.gloplacha.2019.102999, 2019.

Jones, D. B., Harrison, S., Anderson, K., Shannon, S., and Betts, R. A.: Rock glaciers represent hidden water stores in the Himalaya, Sci.
Total Environ., 145368, https://doi.org/10.1016/j.scitotenv.2021.145368, 2021.

595 Kääb, A., Frauenfelder, R., and Roer, I.: On the response of rockglacier creep to surface temperature increase, Global Planet Change., 56,
172-187, https://doi.org/10.1016/j.gloplacha.2006.07.005, 2007.

Kenner, R., Pruessner, L., Beutel, J., Limpach, P., and Phillips, M.: How rock glacier hydrology, deformation velocities and ground
temperatures interact: Examples from the Swiss Alps, Permafrost Periglac., 31, 3-14, https://doi.org/10.1002/ppp.2023, 2019.

Kneisel, C., Hauck, C., Fortier, R., and Moorman, B.: Advances in geophysical methods for permafrost investigations, Permafrost Periglac.,
600 19, 157-178, https://doi.org/10.1002/ppp.616, 2008.

Knight, J., Harrison, S., and Jones, D. B.: Rock glaciers and the geomorphological evolution of deglacierizing mountains, Geomorphology,
324, 14-24, https://doi.org/10.1016/j.geomorph.2018.09.020, 2019.

Krainer, K., Bressan, D., Dietre, B., Haas, J. N., Hajdas, I., Lang, K., Mair, V., Nickus, U., Reidl, D., Thies, H., and Tonidandel, D.: A
10,300-year-old permafrost core from the active rock glacier Lazaun, southern Ötztal Alps (South Tyrol, northern Italy), Quaternary Res.,
605 83, 324-335, https://doi.org/10.1016/j.yqres.2014.12.005, 2015.

Ladanyi, B.: Rheology of ice/rock systems and interfaces, The Eighth International Conference on Permafrost, Zurich, Switzerland, 2003,
WOS:000185049300110, 621-626,

Leopold, M., Williams, M. W., Caine, N., Völkel, J., and Dethier, D.: Internal structure of the Green Lake 5 rock glacier, Colorado Front
Range, USA, Permafrost Periglac., 22, 107-119, https://doi.org/10.1002/ppp.706, 2011.

610 Mellor, M., and Testa, R.: Effect of temperature on the creep of ice, J. Glaciol., 8, 131-145, 1969.

Monnier, S., and Kinnard, C.: Internal structure and composition of a rock glacier in the Andes (upper Choapa valley, Chile) using borehole
information and ground-penetrating radar, Ann. Glaciol., 54, 61-72, https://doi.org/10.3189/2013AoG64A107, 2013.

Monnier, S., and Kinnard, C.: Reconsidering the glacier to rock glacier transformation problem: New insights from the central Andes of
Chile, Geomorphology, 238, 47-55, https://doi.org/10.1016/j.geomorph.2015.02.025, 2015a.

32
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

615 Monnier, S., and Kinnard, C.: Internal structure and composition of a rock glacier in the Dry Andes, Inferred from ground-penetrating radar
data and its artefacts, Permafrost Periglac., 26, 335-346, https://doi.org/10.1002/ppp.1846, 2015b.

Monnier, S., and Kinnard, C.: Interrogating the time and processes of development of the Las Liebres rock glacier, central Chilean Andes,
using a numerical flow model, Earth Surf Proc Land., 41, 1884-1893, 10.1002/esp.3956, 2016.

Moore, P. L.: Deformation of debris-ice mixtures, Rev. Gepphys., 52, 435-467, https://doi.org/10.1002/2014rg000453, 2014.

620 Munroe, J. S.: Distribution, evidence for internal ice, and possible hydrologic significance of rock glaciers in the Uinta Mountains, Utah,
USA, Quaternary Res., 90, 50-65, https://doi.org/10.1017/qua.2018.24, 2018.

Nan, Z., Li, S., and Liu, Y.: Mean annual ground temperature distribution on the Tibetan plateau: Permafrost distribution mapping and
further application, Journal of Glaciology and Geocryology, 24, 142-148, 2002.

Obu, J.; Westermann, S.; Barboux, C.; Bartsch, A.; Delaloye, R.; Grosse, G.; Heim, B.; Hugelius, G.; Irrgang, A.; Kääb, A.M.; Kroisleitner,
625 C.; Matthes, H.; Nitze, I.; Pellet, C.; Seifert, F.M.; Strozzi, T.; Wegmüller, U.; Wieczorek, M.; Wiesmann, A. (2020): ESA Permafrost
Climate Change Initiative (Permafrost_cci): Permafrost version 2 data products. Centre for Environmental Data Analysis, date of
citation. http://catalogue.ceda.ac.uk/uuid/1f88068e86304b0fbd34456115b6606f

Oerlemans, J.: Glaciers and climate change, A.A. Balkema Publishers, Lisse, 2001.

PERMOS: Permafrost in Switzerland 2014/2015 to 2017/2018, edited by: Noetzli, J., Luethi, R., and Staub, B., the Cryospheric Commission
630 of the Swiss Academy of Sciences, Glaciological Report (Permafrost) No. 12–15, 104 pp., 2019.

Rangecroft, S., Harrison, S., Anderson, K., Magrath, J., Castel, A. P., and Pacheco, P.: A first rock glacier inventory for the Bolivian Andes,
Permafrost Periglac., 25, 333-343, https://doi.org/10.1002/ppp.1816, 2014.

Reznichenko, N., Davies, T., Shulmeister, J., and McSaveney, M.: Effects of debris on ice-surface melting rates: an experimental study, J.
Glaciol., 56, 384-394, https://doi.org/10.3189/002214310792447725, 2010.

635 Salerno, F., Guyennon, N., Thakuri, S., Viviano, G., Romano, E., Vuillermoz, E., Cristofanelli, P., Stocchi, P., Agrillo, G., Ma, Y., and
Tartari, G.: Weak precipitation, warm winters and springs impact glaciers of south slopes of Mt. Everest (central Himalaya) in the last 2
decades (1994–2013), The Cryosphere, 9, 1229-1247, https://doi.org/10.5194/tc-9-1229-2015, 2015.

Scott, W. J., Sellmann, P. V., and Hunter, J. A.: Geophysics in the study of permafrost, in: Geotechnical and Environmental Geophysics:
Volume I, Review and Tutorial, 355-384, 1990.

640 Steig, E. J., Fitzpatrick, J. J., Potter, j. N., and Clark, D. H.: The geochemical record in rock glaciers, Geogr. Ann. A., 80, 277-286,
https://doi.org/10.1111/j.0435-3676.1998.00043.x, 1998.

33
 https://doi.org/10.5194/tc-2021-110
Preprint. Discussion started: 14 July 2021
c Author(s) 2021. CC BY 4.0 License.

Vivian, R., and Bocquet, G.: Subglacial cavitation phenomena under the glacier D’Argentière, Mont Blanc, France, J. Glaciol., 12, 439-451,
https://doi.org/10.3189/S0022143000031853, 1973.

Whalley, W. B., and Azizi, F.: Rheological models of active rock glaciers - evaluation, critique and a possible test, Permafrost Periglac., 5,
645 37-51, https://doi.org/10.1002/ppp.3430050105, 1994.

Zhao, L., and Sheng, Y.: Permafrost survey manual, Science Press, Beijing, 2015.

34