\makeatletter
\@ifundefined{HCode}
{\documentclass[screen,CRGEOS,Unicode,biblatex,published]{cedram}
\addbibresource{crgeos20241011.bib}
\newenvironment{noXML}{}{}
\let\citep\parencite
\let\citet\textcite
\def\xcitealp#1#2{\citeauthor{#1}, \citelink{#1}{#2}}
\def\citealt#1#2{\citeauthor{#1}, \citelink{#1}{#2}}
\newcommand*{\citelink}[2]{\hyperlink{cite.\therefsection @#1}{#2}}
\def\defcitealias#1#2{}
\let\citetalias\textcite
\let\citepalias\parencite
\def\thead{\noalign{\relax}\hline}
\def\endthead{\noalign{\relax}\hline}
\def\tbody{\noalign{\relax}}
\def\tabnote#1{\vskip4pt\parbox{.95\linewidth}{#1}}
\def\tsup#1{$^{{#1}}$}
\def\tsub#1{$_{{#1}}$}
\def\ndash{\text{--}}
\def\hyphen{\text{-}}
\def\og{\guillemotleft}
\def\fg{\guillemotright}
\def\0{\phantom{0}}
\makeatletter
\g@addto@macro{\UrlBreaks}{\UrlOrds}
\gappto{\UrlBreaks}{\UrlOrds}
\RequirePackage{etoolbox}
\usepackage{upgreek}
\usepackage{pifont}
\def\inlinefig#1{\includegraphics{#1}}
\def\jobid{crgeos20241011}
%\graphicspath{{/tmp/\jobid_figs/web/}}
\graphicspath{{./figures/}}
\def\xmorerows#1#2{#2}
\newcounter{runlevel}
\usepackage{multirow}
\def\nrow#1{\@tempcnta #1\relax%
\advance\@tempcnta by 1\relax%
\xdef\lenrow{\the\@tempcnta}}
\def\morerows#1#2{\nrow{#1}\multirow{\lenrow}{*}{#2}}
\let\MakeYrStrItalic\relax
\def\refinput#1{}
\let\ubreak\break
\csdef{Seqnsplit}{\\}
\def\botline{\\\hline}
\def\back#1{}
\newcommand*{\hyperlinkcite}[1]{\hyper@link{cite}{cite.#1}}
\DOI{10.5802/crgeos.327}
\datereceived{2024-12-05}
\daterevised{2025-09-01}
\datererevised{2026-01-27}
\dateaccepted{2026-02-09}
\ItHasTeXPublished
}
{
\PassOptionsToPackage{authoryear}{natbib}
\documentclass[crgeos]{article}
\usepackage[T1]{fontenc}
\usepackage{pifont}
\def\CDRdoi{10.5802/crgeos.327}
\def\citelink#1#2{\citeyear{#1}}
\def\xcitealp#1#2{\citealp{#1}}
\def\xmorerows#1#2{\morerows{#1}{#2}}
\let\ubreak\relax
\makeatletter
\def\CDRsupplementaryTwotypes#1#2{}
\def\hyperlinkcite#1#2{\citeyear{#1}}
}
\makeatother

\usepackage{upgreek}

\dateposted{2026-04-07}
\begin{document}

\begin{noXML}

\CDRsetmeta{articletype}{research-article}

\TopicFR{G\'eophysique, g\'eod\'esie}
\TopicEN{Geophysics, geodesy}

\editornote{Article submitted by invitation}
\alteditornote{Article soumis sur invitation}

\title{Mapping water flow pathways in the Fengjiaping landslide using
self-potential and electrical resistivity tomography}

\alttitle{Cartographie des chemins d'\'{e}coulement de l'eau dans
le glissement de terrain de Fengjiaping \`{a} l'aide du potentiel
spontan\'{e} et de la tomographie de r\'{e}sistivit\'{e}
\'{e}lectrique}

\author{\firstname{Kaiyan} \lastname{Hu}\CDRorcid{0000-0002-6721-1705}}
\address{Hubei Subsurface Multi-Scale Imaging Key Laboratory, School of
Geophysics and Geomatics, China University of Geosciences, Wuhan
430074, China}

\author{\firstname{Qinghua} \lastname{Huang}\CDRorcid{0000-0002-1923-3002}\IsCorresp}
\address{State Key Laboratory of Earthquake Dynamics and Forecasting, Peking University,
Beijing 100871, China}
\address{Department of Geophysics, School of Earth and Space Sciences, Peking University,
Beijing 100871, China}
\email[Q. Huang]{huangq@pku.edu.cn}

\author{\firstname{Peng} \lastname{Han}\CDRorcid{0000-0002-9997-8505}}
\address{Department of Earth and Space Sciences, Southern University of
Science and Technology, Shenzhen 518055, China}

\author{\firstname{Tao} \lastname{Tao}\CDRorcid{0009-0004-4094-5444}}
\addressSameAs{4}{Department of Earth and Space Sciences, Southern
University of Science and Technology, Shenzhen 518055, China}

\author{\firstname{Shuangshuang} \lastname{Li}\CDRorcid{0000-0001-9829-2272}}
\addressSameAs{4}{Department of Earth and Space Sciences, Southern
University of Science and Technology, Shenzhen 518055, China}

\author{\firstname{Shuangling}\nobreakauthor \lastname{Mo}}
\addressSameAs{4}{Department of Earth and Space Sciences, Southern
University of Science and Technology, Shenzhen 518055, China}

\author{\firstname{Gexue} \lastname{Bai}}
\address{Gansu Institute of Engineering Geology, Gansu Provincial
Bureau of Geology and Mineral Exploration \& Development, Lanzhou
730000, China}

\author{\firstname{Yunlong} \lastname{Hou}}
\addressSameAs{5}{Gansu Institute of Engineering Geology, Gansu
Provincial Bureau of Geology and Mineral Exploration \& Development,
Lanzhou 730000, China}

\author{\firstname{Ruidong} \lastname{Li}}
\addressSameAs{5}{Gansu Institute of Engineering Geology, Gansu
Provincial Bureau of Geology and Mineral Exploration \& Development,
Lanzhou 730000, China}

\author{\firstname{Baofeng} \lastname{Wan}}
\addressSameAs{5}{Gansu Institute of Engineering Geology, Gansu
Provincial Bureau of Geology and Mineral Exploration \& Development,
Lanzhou 730000, China}

\author{\firstname{Ning} \lastname{An}}
\addressSameAs{5}{Gansu Institute of Engineering Geology, Gansu
Provincial Bureau of Geology and Mineral Exploration \& Development,
Lanzhou 730000, China}

\shortrunauthors

\keywords{\kwd{Self-potential method}
\kwd{Electrical resistivity tomography}
\kwd{Groundwater flow}
\kwd{Landslide}
\kwd{Electrokinetic effect}}

\altkeywords{\kwd{M\'{e}thode du potentiel spontan\'{e}}
\kwd{Tomographie de r\'{e}sistivit\'{e} \'{e}lectrique}
\kwd{\'{E}coulement des eaux souterraines}
\kwd{Glissement de terrain}
\kwd{Effet \'{e}lectrocin\'{e}tique}}

\thanks{National Natural Science Foundation of China (42574089),  China
Postdoctoral Science Foundation (GZC20241598, 2024M753016),  China
Postdoctoral Science Foundation--Hubei Joint Support Program
(2025T045HB), ``CUG Scholar'' Scientific Research Funds at China
University of Geosciences (Wuhan) (Project No. 2023139),  Geological
Disaster Prevention Special Fund of the Gansu Provincial Department of
Natural Resources (Grant No. 20230209GYY)}

\begin{abstract}
The Fengjiaping Landslide, located at the transitional zone between the
western Qinling Mountains and the southwestern margin of the Loess
Plateau in China, is a reactivated loess--mudstone interface landslide.
Its complex evolution is influenced by geological, hydrological and
climatic factors, as well as human activities. The critical zone
regulates precipitation infiltration, which, in turn, controls soil
moisture and groundwater dynamics. Although excessive water
infiltration is recognized as the primary trigger, the landslide
exhibits heterogeneous deformation, with recurrent events not always
correlated with rainfall, making its reactivation mechanisms difficult
to understand and predict. Potential sliding zones in moisture-induced
landslides are typically characterized by high soil moisture and
elevated water fluxes, manifesting as low electrical resistivity and
enhanced streaming current densities. In this study, we applied an
integrated geophysical approach, combining direct-current electrical
resistivity tomography (ERT) and self-potential (SP) measurements, to
infer subsurface water pathways and identify zones potentially
contributing to slope instability. The joint interpretation of SP and
ERT data suggests preferential flow channels and groundwater activity
beneath scarps and cracks, highlighting their potential role as
conduits for infiltration and slope weakening. Despite these insights,
uncertainties remain due to limitations in data coverage, boundary
effects, and simplified assumptions in the inversion framework. Future
work should focus on continuous and time-lapse SP and ERT monitoring,
complemented by methods such as induced polarization and borehole
investigations, to better constrain subsurface hydrogeological
properties and improve the understanding of the processes governing
slope instability.
\end{abstract}

\begin{altabstract} 
Le glissement de terrain de Fengjiaping, situ\'{e} dans la zone de
transition entre les monts Qinling occidentaux et la marge sud-ouest du
plateau de Loess en Chine, est un glissement r\'{e}activ\'{e} \`{a}
l'interface l\oe{}ss--mudstone dont l'\'{e}volution
complexe est influenc\'{e}e par des facteurs g\'{e}ologiques,
hydrologiques, climatiques et anthropiques. La zone critique r\'{e}gule
l'infiltration des pr\'{e}cipitations, contr\^{o}lant ainsi
l'humidit\'{e} des sols et la dynamique des eaux souterraines. Bien
qu'une infiltration excessive soit reconnue comme le principal
facteur d\'{e}clenchant, le glissement pr\'{e}sente une d\'{e}formation
h\'{e}t\'{e}rog\`{e}ne, avec des r\'{e}activations qui ne sont pas
toujours corr\'{e}l\'{e}es aux pr\'{e}cipitations, rendant ses
m\'{e}canismes difficiles \`{a} comprendre et \`{a} pr\'{e}voir. Les
zones potentielles de rupture dans les glissements induits par
l'humidit\'{e} se caract\'{e}risent g\'{e}n\'{e}ralement par une
forte teneur en eau et des flux hydriques \'{e}lev\'{e}s, se traduisant
par une faible r\'{e}sistivit\'{e} \'{e}lectrique et une densit\'{e}
accrue de courants d'\'{e}coulement. Dans cette \'{e}tude, nous avons
appliqu\'{e} une approche g\'{e}ophysique int\'{e}gr\'{e}e combinant la
tomographie de r\'{e}sistivit\'{e} \'{e}lectrique en courant continu
(ERT) et des mesures de potentiel spontan\'{e} (SP) afin d'identifier
les chemins d'\'{e}coulement de l'eau en subsurface et les zones
susceptibles de contribuer \`{a} l'instabilit\'{e} du versant.
L'interpr\'{e}tation conjointe des donn\'{e}es SP et ERT met en
\'{e}vidence des chenaux d'\'{e}coulement pr\'{e}f\'{e}rentiels et
une activit\'{e} des eaux souterraines sous les escarpements et les
fissures, sugg\'{e}rant leur r\^{o}le potentiel comme conduits
d'infiltration et de d\'{e}stabilisation. Toutefois, des incertitudes
persistent en raison des limites de couverture des donn\'{e}es, des
effets de bord et des hypoth\`{e}ses simplifi\'{e}es du cadre
d'inversion. Des travaux futurs devraient privil\'{e}gier un suivi
continu et temporel des mesures SP et ERT, compl\'{e}t\'{e} par la
polarisation provoqu\'{e}e et des investigations par forage, afin de
mieux contraindre les propri\'{e}t\'{e}s hydrog\'{e}ologiques en
subsurface et d'am\'{e}liorer la compr\'{e}hension des processus
r\'{e}gissant l'instabilit\'{e} des versants.
\end{altabstract}

%\input{CR-pagedemetas}

\maketitle

\vspace*{-4pt}

\twocolumngrid

\end{noXML}

\defcitealias{Chambers2011}{ibid.}

\section{Introduction}\label{sec1}

Gansu Province, located in inland China, lies at the intersection of
the Qinghai-Tibet Plateau, the Loess Plateau, and the Inner Mongolia
Plateau. This region is characterized by its complex geological
structures, diverse landforms, and fragile ecological environment.
Frequent earthquakes and concentrated rainfall make it highly
susceptible to geological hazards such as collapses, landslides, and
debris flows \citep{Meng1998,ZhangHuang2018,Fan2019,Peng2019a,Xu2020}.
Among its cities, Tianshui is the second largest in Gansu Province and
lies on the southern edge of the Loess Plateau. The area's bedrock
primarily consists of mudstone, overlain by loess, forming a
geologically unstable setting highly sensitive to rainfall
\citep{Zhang2020,Qi2021}. Human activities such as irrigation and land
reclamation significantly influence soil--water interactions, further
destabilizing slopes \citep{ZhangWang2018,Jia2023,Lan2023}. The
Fengjiaping Landslide, located in the Tianshui region, is a
representative loess--mudstone interface landslide triggered by multiple
factors. Assessing its risk requires a comprehensive understanding of
the hydromechanical processes that drive its activity.

Moisture-induced landslides are often triggered by water infiltration
in the unsaturated zone, groundwater table fluctuations, preferential
flow paths, and their complex interactions
\citep{Sorbino2013,Fan2017,Whiteley2019,Xu2020}. During rainfall or
artificial irrigation events, infiltrating water rapidly increases pore
water pressure and soil moisture, weakening the shear strength and
raising the likelihood of slope failure. Notably, shallow landslides
may initiate before full soil saturation is reached
\citep{Sorbino2013,Li2016,Yang2017,Filho2019,Hu2021}. Conversely,
landslides may also occur under fully saturated conditions due to
aquifer variations caused by water recharge, even in the absence of
rainfall \unskip\break\citep{Tu2010}. The critical roles of groundwater flow and
soil moisture in these processes underscore the importance of
monitoring their spatial and temporal variations to identify high-risk
zones and support risk management.

Understanding landslides requires an integrated approach that considers
the complex interactions within the critical zone, particularly in
regions like Tianshui. Monitoring rainwater infiltration and
hydrodynamic processes often involves hydrological and deformation
measurements. Conventional techniques include measuring pore water
pressure, matric suction, soil moisture, and groundwater levels, as
well as stress--strain and surface deformation monitoring 
\citep[e.g.,][]{Lourenco2006,Schulz2009,Terajima2014,Jiang2017,Lu2024}.
However, these approaches are largely point-based or surface
measurements, offering limited resolution of the internal dynamics of
landslides. Given the heterogeneous nature of water flow within sliding
masses, no single traditional technique can fully capture their spatial
and temporal complexity.

Geophysical methods provide valuable tools for imaging subsurface
structures and delivering higher spatial information than point-based
measurements. Among these, Direct-Current Electrical Resistivity
Tomography (DC ERT) is widely applied in landslide studies due to its
sensitivity to water content. The availability of open-source and
user-friendly inversion software such as Res2DInv
\citep{Loke1996,Loke2020}, BERT \citep{Rucker2006,Gunther2006}, pyGIMLi
\citep{Rucker2017}, and ResIPy \citep{Blanchy2020}, has further
promoted its use in landslide research
\citep[e.g.,][]{Caris1991,Lapenna2003,Lapenna2005,Colangelo2006,Perrone2014,Uhlemann2017,Peng2019a}. 
For instance, \citet{Gance2015} applied the open-source BERT framework
to correct fissure effects in \mbox{resistivity} pseudosections, improving
interpretation reliability. More recently, \citet{Jabrane2023} {combined}
ERT with vertical electrical soundings using pyGIMLi and BERT to
delineate sliding surfaces, underscoring ERT's potential for landslide
monitoring and hazard assessment. Low resistivity is often associated
with highly saturated soils and therefore related to slope instability
\citep{Szalai2017,Zhao2020,Wang2022,Zhang2022}. In the Loess Plateau,
however, resistivity is also influenced by high salinity, clay content,
and structural features such as fissures and loess caves
\citep{Waxman1968,Jougnot2010,Revil2017,Zhang2019,Mendieta2021,QiWu2022}. 
These complexities highlight the need to combine geophysical methods
with localized geological knowledge for reliable interpretation.

In addition to ERT, the self-potential (SP) method provides a passive
approach for monitoring hydrogeological processes. SP measurements
capture natural electric fields generated by infiltrating water,
offering unique insights into subsurface water flow \citep{Thony1997}.
Since the early 20th century, SP has been recognized for its potential
in landslide monitoring, particularly due to its sensitivity to water
saturation and water flow \citep{Bogoslovsky1977,Corwin1990}. Recent
studies show its promise as a supporting tool for early warning of
rainfall-induced landslides, as it can detect water infiltration and
preferential flow patterns
\citep{Haas2009,Yamazaki2017,Whiteley2019,Guo2022,Hu2024,Hu2025a,deAraujo2025}.
For example, in southern Italy, SP mapping has been successfully
applied to investigate groundwater flow direction in landslide zones
\citep{Lapenna2003,Colangelo2006}. Additionally, \citet{Chambers2011},
demonstrated the combined use of SP and ERT methods for investigating
landslides in mudstone--sandstone formations.

Although this study shares methodological similarities with
\citetalias{Chambers2011}, it differs in several key aspects. First, the
Fengjiaping Landslide represents a loess--mudstone contact surface
landslide, dominated by Malan loess and fractured mudstone debris,
where moisture infiltration and preferential flow were primary drivers
of instability. This differs fundamentally from mudstone--sandstone
settings as \citetalias{Chambers2011}. Second, we applied an SP inversion
approach based on the current continuity equation and incorporate
ERT-derived conductivity models, in contrast to cross-correlation
techniques used \citep[e.g.][]{Colangelo2006,Chambers2011,Hu2024}. This
approach ensures a more physically consistent interpretation of
streaming current sources. Third, we integrated in-situ soil property
measurements, including temperature, volumetric water content, and bulk
electrical conductivity, which provided a stronger basis for
interpreting geoelectrical\break anomalies.

In this study, we integrate passive SP and active ERT methods to
investigate subsurface water flow paths and delineate possible
water-rich zones that contribute to slope instability. By performing SP
inversion combined with ERT results, we analyze SP source distributions
and their implications for landslide processes. Using the Fengjiaping
Landslide as a case study, our objective is to map subsurface
water-storage areas and preferential pathways that may trigger
reactivation. These geoelectrical approaches enhance the understanding
of hydrological controls on landslides and provide a framework for
improving risk assessment and mitigation strategies within the Earth's
critical zone.

\section{Materials and methods}\label{sec2} 

\subsection{Study area}\label{sec21} 

The Fengjiaping Landslide is an ancient landslide, shaped like a
``circular chair'', measures approximately 840 m in length and an
average width of 423~m, covering an area of $2.32 \times
10^{5}$~m$^{2}$  (Figure~\ref{fig1}). It has an average thickness of 16
m, a total volume of $3.78 \times 10^{6}$~m$^{3}$, and a primary
sliding direction of 245\textdegree. The elevation at the front edge of
the landslide is 1390~m, while the rear edge rises to 1570 m, resulting
in a relative elevation difference of 180 m. The overall slope of the
landslide is 15.7\textdegree.

\begin{figure*}
\includegraphics{fig01}
\vspace*{-4pt}
\caption{\label{fig1}(a)~Location of the investigation site; (b)~the
ancient Fengjiaping landslide marked by white line to indicate its
boundary and its reactivation body with yellow shaded area (image from
Google Earth); (c)~the drill log (the red dot in b) conducted in 
April~22--23 of 2022. $Q_{4}^{\mathrm{ml}}$, $Q_{4}^{\mathrm{del}}$ and N
denote artificially filled silt, Holocene deposits and the Neogene
system, respectively.}
\vspace*{-6pt}
\end{figure*}


This ancient landslide has been reactivated due to the combined effects
of rainfall, groundwater activity, and land use. The reactivated
Fengjiaping Landslide is a medium-sized loess--mudstone contact surface
landslide, displaying a ``long tongue'' shape planform and a stepped
cross-sectional profile (Figure~\ref{fig1}b). It features well-developed
gullies along both sides of the landslide body (H1 and H2). The
reactivated landslide has a front edge elevation of 1395 m and a rear
edge elevation of 1532~m, with a relative height difference of 137 m.
It spans 450 m in length, with an average width of 64.4 m, covering an
area of $2.90 \times 10^{4}$~m$^{2}$. The reactivated landslide has
an average thickness of approximately 8 m, a total volume of 
$2.32 \times 10^{5}$~m$^{3}$, a main sliding direction of
225\textdegree, and an average slope of 14.7\textdegree. The
thicknesses of both the ancient and reactivated landslides were
estimated based on borehole logging data. Since only two boreholes were
available, located respectively within the H1 and H2 bodies, these
thickness estimates are subject to uncertainty. Nevertheless, they
provide first-order constraints on the internal structure of the
landslide.

The study area is predominantly covered by Holocene strata of the
Quaternary system. A borehole was drilled in the central region of the
reactivated landslide body (34\textdegree24$'$36.641268$''$N,
105\textdegree41$'$13.539480$''$E) at an elevation of 1646.10 m above sea
level on April 22--23, 2022 (Figure~\ref{fig1}c). The drill logs
indicate that the shallow zone primarily consists of artificially
filled silt (Q$_{4}^{\mathrm{ml}}$), with \mbox{increasing} water
content and heterogeneous soil texture at greater depths. The landslide
accumulation mainly derives from Holocene deposits
(Q$_{4}^{\mathrm{del}}$), predominantly composed of silty clay.
Beneath these deposits, moderately weathered mudstone and sandy
mudstone of the Neogene system (N) are exposed (Figure~\ref{fig1}c). The
integrity of the rock formation has been compromised by multiple
vertical joints along the slope.

\begin{figure*}
\vspace*{-2pt}
\includegraphics{fig02}
\vspace*{-2pt}
\caption{\label{fig2}The photos of (a)~H2 body recorded by 
Da-Jiang Innovations (DJI) M300 drone, 
(b) and (c):~house cracks, and (d)~exposed mudstone layer in the gully of the
landslide.}
\vspace*{-2pt}
\end{figure*}

\begin{figure*}
\vspace*{-2pt}
\includegraphics{fig03}
\vspace*{-3pt}
\caption{\label{fig3}The observation system in the Fengjiaping
landslide. (a)~The digital elevation model (DEM) with measurement
points using different techniques \citep{Hu2025a}; (b)~Da-Jiang
Innovations (DJI) M300 drone; (c)~NI Compact DAQ for SP measurements;
(d)~ERT measurement.}
\vspace*{-4pt}
\end{figure*}

The deformation caused by the landslide has led to significant damage
to residential properties (Figure~\ref{fig2}). A visible gap between a
house and its foundation underscores the ground displacement and
movement caused by the landslide (Figure~\ref{fig2}b). Within the
landslide zone, one building shows extensive structural damage, with
multiple fissures and large cracks along the walls, accompanied by a
vertical displacement of approximately 1.70~m
\mbox{(Figure~\ref{fig2}c)}. The exposed blue-gray mudstone layer in
the gully of the slope is highly sensitive to water \mbox{infiltration}.
Recharge into this layer reduces its mechanical strength, thereby
accelerating slope destabilization. The interaction between
infiltrating water and the mudstone layer plays a critical role in the
reactivation and progression of the landslide. Understanding this
relationship is essential for assessing the stability of the slope and
predicting future risks.

The primary trigger of the Fengjiaping landslide was attributed to
rainfall. However, the presence of deep gullies and cracks complicates
groundwater flow paths, leading to failures that are not always
synchronized with rainfall events. Soils that are fully or partially
saturated with water exhibit \mbox{significantly} lower electrical resistivity
compared to their surroundings. Moreover, groundwater movement within
the slope naturally generates SP signals. To identify the convergence
zones of rainwater infiltration and detect hidden channels that
facilitate dominant groundwater flow, we conducted geoelectrical
measurements on the slope. By integrating ERT and SP methods, we
inferred subsurface water flow paths, offering valuable insights into
the hydrological processes that influence slope instability.

\subsection{Data and methods}\label{sec22} 

Drone-based LiDAR surveys were conducted across the Fengjiaping region
to capture detailed \mbox{topographic} and surface feature data
(Figure~\ref{fig3}). The DJI M300 drone, equipped with Real-Time
Kinematic (RTK) positioning and a DJI Zenmuse L1 LiDAR sensor, was used
for data acquisition (Figure~\ref{fig3}b). This LiDAR system can
differentiate surface features such as soil, residential buildings,
tree branches, and vegetation canopies by analyzing multiple echoes.
The system provides an elevation accuracy of about 5~cm. The acquired
LiDAR data were processed in DJI Terra software to generate a filtered
point-cloud dataset, with buildings and vegetation removed. The
processed data show topographic variability, with elevation differences
in the study area exceeding 100~m (Figure~\ref{fig3}a).

\subsubsection{Monitoring data}\label{sec221} 

Throughout 2023, we collected monitoring data, including atmospheric
parameters from the Beidao station of National Oceanic and Atmospheric
Administration (NOAA), water levels measured by a pressure-type water
level gauge and elevation and azimuth angles recorded by inclinometers.
The data show a correlation between temporal increases in precipitation
and rising water table levels (Figure~\ref{fig4}). Additionally,
groundwater replenishment may also be influenced by the discharge from
the Weihe River, a tributary of the Yellow River \citep{Peng2015}.
\looseness=1

\begin{figure*}
\vspace*{-2pt}
\includegraphics{fig04}
\vspace*{-2pt}
\caption{\label{fig4}The monitoring data collected in the study area.
(a)~The hourly precipitation data collected from the Beidao station of
NOAA; (b)~the hourly displacement components measured by GNSS; (c)~the
depth of water table data; (d)~the angle of azimuth and elevation data
measured by two inclinometers; (e) the measured fracture width data by
a crack meter located at lnc01.}
\end{figure*}

\begin{table*}[t!]%tab1
\caption{\label{tab1}Configuration of ERT surveys and inversion misfits\vspace*{-2pt}}
\begin{tabular}{ccccc}
\thead
Line no. & \parbox[t]{2cm}{\centering Number of electrodes} & 
\parbox[t]{2.2cm}{\centering Length of survey line (m)} & 
\parbox[t]{2cm}{\centering Number of datum points} & 
\parbox[t]{2.3cm}{\centering Resulting misfit (RMS)} \vspace*{2pt}\\
\endthead
L1 & 60 & 118 & 570 & 1.95  \\
L2 & 20 & \019 & \057 & 1.53 \\
L3 & 30 & \029 & 135 & 1.22 \\
L4 & 30 & \029 & 135 & 1.93
\botline
\end{tabular}
\vspace*{-2pt}
\end{table*}


A Global Navigation Satellite System (GNSS) measuring station was
installed in the southwestern part of the study area
(Figure~\ref{fig3}a). Two inclinometers (Inc01 and Inc02), located 47 m
apart behind the residential buildings, recorded azimuth angles
averaging 224.12\textdegree~and 189.47\textdegree, respectively. The
average elevation angles were 3.81\textdegree~at Inc01 and
5.42\textdegree~at Inc02. In October 2023, a notable rise in the water
table (Figure~\ref{fig4}c) coincided with an increase in fracture width
near Inc01 (Figure~\ref{fig4}e). By November, the northeastern edge of
the slope exhibited general stability, while the central section of the
study area experienced sliding (Figure~\ref{fig4}b). This monitoring
data highlights the spatiotemporal complexity of slope deformation,
which appears to be strongly linked to the rising water levels.
\vspace*{-1pt}

\subsubsection{Electrical resistivity tomography}\label{sec222} 
\vspace*{-1pt}

To analyze the electrical structure of the Fengjiaping landslide, a
primary ERT survey line (L1) was laid out downslope, with three
additional survey lines (L2--L4) arranged perpendicularly to L1
(Figure~\ref{fig3}a). The ERT instrument manufactured by Chongqing
Jingfan Technology Co., Ltd. (China) was used in this experiment
(Figure~\ref{fig3}d). Before measurements, the grounding resistance of
each electrode was checked to ensure all values were below 
1~k${\Omega}$, minimizing data quality issues due to poor electrode
contact. \mbox{Table~\ref{tab1}} provides details of the ERT profiles, including the
number of electrodes, profile lengths, the number of data points, and
the root mean square (RMS) misfit of inversion. Specifically, Line L1
employed a dipole--dipole array with a nominal electrode spacing of
2~m; however, due to the topographic variations in the steep-slope
area, the actual electrode spacing was slightly larger. Lines L2--L4
used the Wenner array with an electrode spacing of 1~m.

For data processing, measurement points were corrected for elevation
and distance, and the apparent resistivity values were calculated based
on the actual electrode coordinates obtained using the RTK positioning.
The ERT data were inverted using the open-source software ResIPy, which
applies regularization inversion with linear filtering
\citep{Blanchy2020}. In this framework, the inversion seeks a
compromise between fitting the observed apparent resistivity data and
minimizing spatial variations in the resistivity model. The choice of
the initial reference model influences the convergence behavior and
stability of the inversion. While ResIPy can automatically estimate an
initial resistivity model from the apparent resistivity data, a uniform
initial resistivity of 10~${\Omega}{\cdot}$m was assigned to all survey
lines to ensure a consistent starting point and facilitate comparison
among profiles. Such a choice avoids introducing artificial contrasts
related to differing initial models and promotes stable convergence 
\citep[e.g.,][]{Tao2024a}.

The regularization parameter ${\lambda}$ controls the trade-off
between data misfit and model smoothness. Larger ${\lambda}$ values
favor smoother models at the expense of data fitting, whereas smaller
values allow greater model variability but may lead to instability or
overfitting \citep{Loke2014,Tao2024b}. In this study, ${\lambda}$ was
determined by the default adaptive optimization strategy in ResIPy,
which automatically updates the regularization strength during the
iterative inversion to balance data misfit reduction and model
roughness. This approach provides a robust and objective means of
selecting ${\lambda}$ without manual tuning. The RMS errors of
inversion on Lines L1, L2, L3, and L4 are 1.95, 1.53, 1.22, and 1.93,
respectively. The raw ERT measurements on each survey line can be found
in \citet{Hu2025a}.

\subsubsection{Self-potential}\label{sec223} 

To investigate subsurface water flow within the deformation zone of the
Fengjiaping landslide, a total of 86 sensors were strategically placed
across the study area (Figure~\ref{fig3}a). Given the extensive
measurement area---spanning over 120~m---the site was divided into
five smaller sub-regions. Each \mbox{sub-region} was equipped with one
reference electrode and approximately 16 measurement electrodes 
(Figure~S2 in the Supplementary Material). Continuous measurements were
conducted in each sub-area at a sampling rate of 10 Hz for 20~min.

Data acquisition was facilitated using the National Instruments system
(Figure~\ref{fig3}c), where the input module connected sensors by
differential connections to minimize common-mode interference. The
control of SP measurements was achieved using an embedded LabVIEW
program, which enabled the development of a multi-channel landslide
monitoring system capable of synchronous observation and real-time data
transmission. SP measurements utilized custom-built Pb-PbCl$_{2}$
non-polarizable electrodes  (Figure~S4 in the Supplementary Material),
which were pretested in the laboratory and showed excellent stability
with minimal electrode potential differences \citep{Hu2025b,Li2025}.

Since SP signals are inherently weak and susceptible to environmental
interference, several measures were implemented to ensure the accuracy
and reliability of the data. First, the initial electrode range was
measured in a saturated saline solution prior to each measurement, with
electrodes submerged until readings stabilized. Second, to reduce
surface noise, electrode pits were dug using a soil-lifting device, and
electrodes were buried at a depth of 50~cm. After installation and
wiring, electrical potential differences were checked with a multimeter
to verify the accuracy of synchronous measurements before formal data
collection. Finally, the electrical potential differences between
reference electrodes in each sub-region were measured and used to
calibrate all measurement points to a unified reference electrode. The
raw monitoring and temperature- and electrode-corrected SP data are
archived in \citet{Hu2025a}. Further details of the SP pre-processing
workflow are provided in the Supplementary \mbox{Material}.

In theory, in the absence of external electrical currents and other
contributing mechanisms, the governing equation for SP is expressed as
\citep{Sill1983}:
{\begin{equation}\label{eq1}
\nabla \cdot \frac{\nabla SP}{\rho }=\nabla \cdot \mathbf{J}_{\mathrm{s}},
\end{equation}}\unskip
where $\nabla $ denotes gradient operator, $\rho $ (${\Omega}{\cdot}$m) 
denotes the electrical resistivity, and $\mathbf{J}_{\mathrm{s}}$
(A$\cdot$m$^{-2}$) denotes the streaming current density with
{\begin{equation}\label{eq2}
\mathbf{J}_{\mathrm{s}}=\hat{Q}_{v}\mathbf{u},
\end{equation}}\unskip
where $\hat{Q}_{v}$ (C${\cdot}$m$^{-3}$) and $\mathbf{u}$ 
(m${\cdot}$s$^{-1}$) denote the
effective excess charge density and Darcy velocity, respectively
\citep{Revil2003}. Thus, in the absence of other sources, gradients in
measured SP data are expected to align with the direction of
groundwater Darcy velocity. When no external current sources are
present and the electrical resistivity is known, SP gradient can
provide valuable insights into the direction of subsurface water flow
\citep[e.g.,][]{Colangelo2006}. Additionally, by inverting the measured
SP data, the distribution of current sources can be reconstructed
\citep[e.g.,][]{Jardani2006,Soueid2013,Rittgers2015,Alarouj2021}.

In this study, we combined the electrical conductivity model obtained from ERT
results with the SP data along L1 to perform the streaming source term
inversion (Equation~(\ref{eq1})). The SP data at ERT electrode positions were obtained
via natural neighbor interpolation of the three-dimensional SP dataset 
(Figure~S6 of the Supplementary Material). Inversion employed an IRLS
scheme with Gauss--Newton updates
\citep[e.g.,][]{LiOldenburg1996,Minsley2007}, which adaptively refines
model weights to recover sharper and more localized sources compared
with conventional $L_2$ inversions \citep{Reymond2010}. This SP
inversion allowed us to derive the distribution of streaming current
sources and infer the pathways of groundwater flow within the
landslide-affected area.

\subsubsection{Experimental procedure}\label{sec224} 

The field measurements were conducted following a sequential and
controlled acquisition protocol to minimize mutual interference between
different geophysical methods. First, all electrode locations were
prepared by digging shallow holes. At each location, soil temperature,
volumetric water content, and bulk electrical conductivity were
measured using a portable soil multi-parameter sensor  (Figure~S5,
SN-3001-TRREC-N01, Shandong Renke Control Technology Co., Ltd.). These
measurements were performed prior to SP acquisition to support
subsequent interpretation and temperature-drift correction.

As mentioned in Section~\ref{sec223}, the study area was divided into five
sub-regions, each equipped with a local reference electrode 
(Figure~S2). Before SP acquisition, the electrical potential
differences between reference electrodes in different sub-regions were
measured to allow all data to be later corrected to a unified
reference. SP signals were then recorded sequentially sub-region by
sub-region.

To avoid possible contamination of SP measurements by injected currents
or electrode polarization effects, ERT surveys were conducted after the
completion of all SP and soil parameter measurements. This acquisition
order ensured that the passive SP data mainly reflect natural
electrokinetic signals associated with groundwater flow rather than
artefacts induced by active electrical measurements.

\vspace*{-4pt}

\section{Results}\label{sec3} 

\vspace*{-2pt}

\subsection{Soil parameters}\label{sec31} 

\vspace*{-3pt}

To account for temperature drift in the electrodes, we measured soil
parameters at all SP measurement points, including soil temperature,
volumetric water content, and bulk electrical conductivity
(Figure~\ref{fig5}). The northeast and southwest regions of the
measurement area exhibited lower temperatures compared to the central
section (Figure~\ref{fig5}a--b). This can be explained by higher soil moisture
and the vegetation coverage in the upper residential area, and reduced
solar radiation within deep gullies in the lower slope
(Figure~\ref{fig2}a). Both the upper and middle--lower parts of the
slope showed elevated moisture values (${>}$0.3;
Figure~\ref{fig5}c). Bulk electrical conductivity reached a maximum of
700~${\upmu}$S/cm, with an average of 412~${\upmu}$S/cm
(Figure~\ref{fig5}d). Zones of high conductivity coincide with high
soil moisture, indicating water-rich characteristics 
(Figure~\ref{fig5}c--d).

\begin{figure*}
\includegraphics{fig05}
\vspace*{-3pt}
\caption{\label{fig5}Surface elevation and soil parameters measured 50
cm below the ground surface. (a)~Elevation; (b)~temperature;
(c)~volumetric water content; (d)~bulk electrical conductivity.
Triangles indicate measurement points.}
\vspace*{-3pt}
\end{figure*}

\vspace*{-4pt}

\subsection{ERT results}\label{sec32} 

\vspace*{-3pt}

The measured data were processed using the open-source package ResIPy
\citep{Blanchy2020} incorporating topographic modeling. Two-dimensional
inversions were performed for one primary profile (L1) along the slope
and three perpendicular profiles (L2--L4). The presented results have
accounted for the depth of investigation (DOI). The inversion result
for the profile along the slope (L1) is presented in the next section,
as it was used for the SP source inversion. We also evaluated the
reliability of the inversion results using the $R$ index
proposed by \citet{OldenburgLi1999}, which indicates that the presented
imaging regions are reliable except for the near-surface layer (depths
shallower than ${\sim}$0.5 m). Details about the depth of
investigation can be found in the Supplementary Material (Figures~S7,
S8). 

Triangular meshing was applied during inversion, with mesh refinement
around the electrode locations and near the surface, while gradually
coarsening with depth (Figure~\ref{fig6}). The resistivity model is
defined on this mesh, where resistivity values are assigned to mesh
vertices and rendered using linear interpolation within each triangular
element. The resistivity images show a general decrease in resistivity
with depth along all profiles. A pronounced transition from relatively
high to low resistivity is observed at shallow depths, which is
interpreted as reflecting the contrast between the unsaturated
near-surface materials and deeper, more conductive zones associated
with increased moisture content. This transition is inferred to
indicate the approximate position of the shallow groundwater table,
occurring at depths of roughly 2--5 m below the ground surface. This
interpretation is consistent with the recorded water level data 
(Figure~\ref{fig4}d). Overall, the groundwater level rises from the
slope toe toward the head. 

\begin{figure*}
\vspace*{1pt}
\includegraphics{fig06}
\vspace*{.5pt}
\caption{\label{fig6}ERT inversion results for horizontal survey lines
on (a)~L4, (b)~L3, and (c)~L2, respectively. The white triangles
represent the inversion mesh.}
\end{figure*}

\subsection{SP results}\label{sec33} 

The reference potential was fixed at the slope toe, with all SP values
corrected for electrode drift and temperature (Figure~\ref{fig7}).
Overall, SP values are negative, indicating a groundwater flow trend
toward the slope toe (Figure~\ref{fig7}a). The lowest SP values are
found in the upper-middle section of the slope (Zone 3) beneath the
residential houses (Figure~\ref{fig2}a). SP gradients show stronger
north--south than east--west variations, with a maximum spatial
gradient below 3~mV/m (Figure~\ref{fig7}b,c). Negative SP anomalies
cluster in Zones 3 and 4, where 42 SP \mbox{electrodes} were \mbox{concentrated.}
These areas also coincide with higher soil conductivity and steeper SP
gradients (Figures~\ref{fig5}d, \ref{fig7}b,c), implying stronger streaming
currents and higher groundwater velocities  (Equation~(\ref{eq1})). 

\begin{figure*}
\vspace*{2pt}
\includegraphics{fig07}
\vspace*{1pt}
\caption{\label{fig7}SP profiles: (a)~SP profile corrected for
temperature drift; (b)~east--west SP gradient; (c)~north--south SP
gradient. Colored triangles indicate SP electrodes in different zones,
while dashed grey lines denote the boundaries between those zones. 
The area enclosed by the blue line represents the high-conductivity zone.}
\end{figure*}

For quantitative interpretation, we performed SP inversion along L1.
The final inversion (85 iterations) achieved an RMS misfit of 0.0032 mV
(Figure~\ref{fig8}). Along the L1 profile, the average SP value is
${-}$10.4~mV, with two notable negative anomalies (Figure~\ref{fig9}a).
The SP gradient, driven by electrokinetic effects, aligns with the
expected groundwater flow direction. Negative SP anomalies are
typically associated with infiltration and downward water flux 
\citep[e.g.,][]{Bogoslovsky1977,Zlotnicki1998}. A minimum SP value
(${{\sim}{-}}22$~mV) occurs at ${\sim}$92 m near the main scarp, coinciding
with a sink zone identified from the SP inversion (Zone S1 in
Figure~\ref{fig9}b). This location also corresponds to the shallow
conductive anomaly C1 (Figure~\ref{fig9}c). A secondary SP anomaly
appears toward the southwestern end of the profile (${\sim}$30 m),
associated with another localized streaming sink feature (Zone S2).
However, interpretation in this sector remains less certain, as the
inversion results are affected by reduced ERT coverage and potential
boundary effects.

\begin{figure*}
\includegraphics{fig08}
\vspace*{-2pt}
\caption{\label{fig8}Convergence curve of the SP inversion showing RMS
misfit reduction during IRLS iterations.}
\vspace*{-2pt}
\end{figure*}

\begin{figure*}
\includegraphics{fig09}
\vspace*{2pt}
\caption{\label{fig9}SP inversion along the survey line L1. (a)~SP data
and elevation of observations; (b)~inverted streaming source
distribution; (c)~corresponding ERT conductivity model. Major negative
SP anomalies (Zones S1 and S2) align with streaming sinks and
conductive anomalies (Zone C1). Dashed white line denotes the proposed
water level from ERT results.}
\vspace*{2pt}
\end{figure*}

\section{Discussions}\label{sec4} 

Overall, the SP measurements reveal a general gradient directed toward
the slope toe (Figure~\ref{fig7}). The near-surface water content and
electrical conductivity exhibit a trend of higher values in the middle
and lower portions, decreasing toward the edge of the sliding mass.
This spatial variation in soil properties is consistent with the
topographic features of the slope. Notably, areas with high gradients
in SP, water content, and electrical conductivity tomography coincide
with zones of intense deformation (Figures~\ref{fig5}c,d and
Figure~\ref{fig7}).

A particularly weak SP signal is observed within the vegetable plot
\ding{172} (Figure~\ref{fig10}), which may be partly influenced by
plant transpiration, a process known to generate localized electrical
potentials \citep[e.g.,][]{Hu2025c}. In addition, two irrigation water
storage containers \ding{175}, located above the vegetable plot, likely
enhance local water infiltration. By analyzing SP gradients, we
inferred potential seepage pathways (Figure~\ref{fig10}). These suggest
that groundwater flows not only along the slope toward the gullies but
also migrates along preferential subsurface flow channels. 

The negative SP anomaly in this area, expressed as a downward gradient,
indicates active groundwater migration downslope, thereby contributing
to landslide dynamics. In contrast, the low-permeability and
high-conductivity mudstone layer in the central sector \ding{174} acts
as a barrier (Figures~\ref{fig9}c and~\ref{fig10}), impeding
groundwater movement and creating spatial heterogeneity in the
hydrogeological response of the landslide body.

The distribution of negative source terms also highlights potential
seepage inlets \citep[e.g.,][]{Guo2022}. For example, scarp \ding{173}
near Zone S1 appears to facilitate vertical infiltration, which is
subsequently redirected underground to form a preferential flow channel
(Zone C1). Because the subsurface consists of loess and clay, zones of
high conductivity partly reflect surface conductivity and pore-water
salinity effects, rather than being solely controlled by water content
\citep{Revil1998}. When considered together, the co-location of
negative SP anomalies (Figures~\ref{fig7}a and~\ref{fig9}a), the sink area (Zone
S1 in Figure~\ref{fig9}b), elevated near-surface water content and
reduced soil temperature (Figure~\ref{fig5}b--c), and the shallow high
conductive anomaly (Figure~\ref{fig5}d and Zone C1 in
Figure~\ref{fig9}c) supports the interpretation of a preferential
seepage pathway emerging near the scarp.

Nevertheless, SP is a potential-field method, which inherently limits
the depth resolution and the ability to uniquely constrain the depth
extent of subsurface sources. In addition, to account for topography
and the underlying bedrock, the modeling domain was extended from the
air layer down to the basal boundary (Figure~\ref{fig9}b). This domain
extension may introduce additional uncertainty in the recovered source
magnitudes and affect the quantitative accuracy of the inversion
results. The absence of borehole data within the study area limits
vertical constraints on the electrical conductivity structure and the
depth distribution of streaming current sources. Moreover, the
effective excess charge density that controls electrokinetic coupling
depends on factors such as pore size distribution, water content, and
pore-water chemistry
\citep[e.g.,][]{Linde2007,Revil2007,Jougnot2019,Jougnot2020,Hu2020,Hu2025b,Solazzi2022},
and thus likely varies spatially across the landslide body. A full
quantification of Darcy flux would require specific assumptions
regarding both the electrokinetic coupling coefficient and the excess
charge density \citep[e.g.,][]{Minsley2007}. Therefore, while our
results allow us to infer groundwater flow pathways, they cannot yet
provide absolute flow rates.

To improve the understanding of subsurface water dynamics and enhance
the resolution of streaming current analyses, future work should
integrate SP monitoring with complementary methods such as time-lapse
ERT and induced polarization, along with borehole investigations (e.g.,
clay content, electrokinetic coupling coefficient, permeability) and
provide improved vertical resolution. Combining these approaches would
enable a more robust characterization of subsurface structures and
groundwater flow pathways over time.

\begin{figure*}
\vspace*{-2pt}
\includegraphics{fig10}
\vspace*{-1pt}
\caption{\label{fig10}The DEM covered with SP data (contour map) and
areas with slopes exceeding 30 degrees (red polygons) to indicate the
scarps caused by the landslide. The dashed line arrows indicate the
proposed groundwater flow pathways.}
\vspace*{-2pt}
\end{figure*}

\section{Conclusion}\label{sec5} 

This study employed a geoelectrical framework to investigate the
Fengjiaping Landslide, integrating SP measurements with ERT to infer
subsurface water pathways and identify zones potentially related to
slope reactivation. The ERT results provided conductivity models that
suggest the presence of a shallow water table and delineate possible
aquifer distributions within the study area. The joint interpretation
of SP and ERT data along the slope revealed negative SP source
anomalies concentrated beneath the main scarps in the accumulation
zone. These anomalies, together with cracks and scarps, likely act as
preferential pathways for infiltration. The inferred coexistence of
shallow water-enriched zones and clay-rich deposits appears to play a
key role in governing slope instability.

Nonetheless, several limitations remain. The interpolation of SP data
onto ERT electrode positions may introduce local uncertainties,
particularly in sparsely covered sectors. Inversion results, especially
near profile edges, are less constrained due to boundary effects and
limited resistivity sensitivity. Furthermore, the adopted inversion
framework assumes simplified source--flow relationships, which may not
fully capture the complex, coupled hydrogeophysical processes in
heterogeneous and deforming slopes.

Future work should focus on integrating time-lapse SP and ERT
monitoring with complementary methods, such as induced polarization and
borehole investigations, to improve vertical resolution and help
constrain key parameters, including clay content, permeability, and
electrokinetic coupling coefficients. Applying this integrated
framework to other landslide-prone regions may enhance our ability to
map preferential flow pathways and support the development of
early-warning and mitigation \mbox{strategies}.

\section*{Acknowledgements}

This work has been partly supported by the National Natural Science
Foundation of China (42574089), the Postdoctoral Fellowship Program of
China Postdoctoral Science Foundation (GZC20241598, 2024M753016), the
China Postdoctoral Science Foundation--Hubei Joint Support
Program (2025T045HB), the ``CUG Scholar'' Scientific Research Funds at
China University of Geosciences (Wuhan) (Project No. 2023139), and the
Geological Disaster Prevention Special Fund of the Gansu Provincial
Department of Natural Resources (Grant No. 20230209GYY). The authors
thank Dr.~Jing Xie at Central South University and Dr.~Wenwu Tang at
East China University of Technology for discussions on the SP
inversions. The dataset collected during the study have been made
publicly available on the Hydrogeophysics Community of Zenodo
at \url{https://doi.org/10.5281/zenodo.17010708}. The authors thank the
editor and two anonymous reviewers for their comments and
recommendations, which help us to improve this manuscript.

\CDRGrant[NNSFC]{42574089}
\CDRGrant[CPSF]{GZC20241598}
\CDRGrant[CPSF]{2024M753016}
\CDRGrant[CPSF]{2025T045HB}
\CDRGrant[CUG]{2023139}
\CDRGrant[DNR]{20230209GYY}

\section*{Declaration of interests}
The authors do not work for, advise, own shares in, or receive funds
from any organization that could \mbox{benefit} from this article, and
have declared no affiliations other than their research organizations.

\section*{Supplementary materials}

Supporting information for this article is available on the journal's
website under \printDOI\ or from the author.

\CDRsupplementaryTwotypes{supplementary-material}{\cdrattach{crgeos-327-suppl.pdf}}

\back{}

\printbibliography
\refinput{crgeos20241011-reference.tex}

\end{document}