论文信息
论文解读
AI 辅助整理- 方法
- 综述被动微波遥感、LFMC与水势数据集及VOD方法。
- 发现
- 点-像元尺度不匹配与生态生理机制理解不足是主要障碍。
- 机会点
- 可发展多频微波与EPV曲线结合的方法,实现像元尺度水分含量与水势反演。
论文摘要
摘要 综述目的 理解植被水分动态对于监测生态系统健康、改进火灾风险评估以及模拟陆地-大气反馈至关重要。本综述考察卫星遥感如何帮助监测植被水分状态(包括含水量和水分势的信息),并识别当前阻碍卫星数据与实地测量数据协调的主要差距。本文旨在建立跨空间尺度的桥梁,以增强当前从区域到全球层面的植被水分状态监测框架。本文主要关注被动微波遥感,该方法适用于在区域至全球尺度上感知不同植被层的水分,而目前相关研究工作与数据产品仍然有限。近期研究发现 近期研究已建立并扩展了全球活体燃料含水率(LFMC)和水分势(Ψ)的实地与遥感数据集。此外,新的被动微波卫星产品在粗尺度上对冠层水分动态表现出更高的敏感性。然而,遥感数据与实地数据的协调工作受到点-像元尺度不匹配以及对卫星数据与生态生理过程之间关系理解有限的制约。新兴概念如生态系统压力-体积(EPV)曲线(具有足迹尺度代表性的潜力)以及基于卫星VOD的新方法为从太空推断含水量和水分势提供了有前景的路径。多频率微波遥感的发展进一步为监测植物水分状态的垂直梯度以及研究微波对不同植物组织水分的敏感性开辟了新机遇。总结 弥合遥感与生态水文学点尺度研究之间的差距需要协调数据集以进行产品验证与比较(本文重点关注LFMC产品),以及监测含水量和水分势的新框架。本文综述了面向这些目标的研究进展。这将有助于实现植被水分动态的大尺度监测,增进我们对整个生态系统植物干旱响应的理解,并改善生物圈-大气相互作用的空间分辨预测。
全文
Current Forestry ReportsREVIEW
Current Forestry Reports (2026) 12:34
https://doi.org/10.1007/s40725-026-00290-y
Extended author information available on the last page of the article
Abstract
Purpose of The Review Understanding vegetation water dynamics is crucial for monitoring ecosystem health, improving
fire risk assessment, and modelling land–atmosphere feedbacks. This review examines how satellite remote sensing can
help monitor the vegetation water status (with information about water content and water potential) and identifies the major
gaps that currently hinder reconciling satellite data with field-based measurements. It aims to establish bridges across spatial
scales to enhance the current monitoring framework for vegetation water status from regional to global levels. The paper
mostly focuses on passive microwave remote sensing, which is appropriate for sensing water across different vegetation lay-
ers at regional to global scales, and at which the amount of research work and data products are limited.
Recent Findings Recent research has established and expanded global in situ and remote sensing datasets of live fuel mois-
ture content (LFMC) and water potential (Ψ). Also, new passive microwave satellite products offer improved sensitivity to
canopy water dynamics at coarse scales. However, efforts to reconcile remote sensing and field data are hampered by point-
to-pixel scale mismatches and by a limited understanding of how satellite data relates to ecophysiological processes. Emerg-
ing concepts such as ecosystem pressure–volume (EPV) curves (with the potential of footprint-scale representativeness) and
new satellite VOD-based approaches provide promising pathways to infer water content and water potential from space.
Developments in multi-frequency microwave remote sensing further open new opportunities to monitor vertical gradients of
water status in plants and to study the sensitivity of microwaves to water across plant tissues.
Summary Bridging remote sensing and point-scale research in ecohydrology requires harmonized datasets for validation
and comparison of products (here we focus on LFMC products) and new frameworks to monitor water content and water
potential. In this paper, we review research progress towards these goals. This will enable large-scale monitoring of veg -
etation water dynamics, advancing our understanding of plant responses to drought for entire ecosystems and improving
spatially-resolved predictions of biosphere–atmosphere interactions.
Keywords Vegetation water content · Vegetation water potential · Live fuel moisture content · Passive microwave
remote sensing · Ecosystem hydraulics · Vegetation optical depth
Received: 11 February 2026 / Accepted: 21 September 2026
© The Author(s) 2026
Monitoring Water in Vegetation: Mind the Gap Between Satellite
Remote Sensing and Field Scale Approaches
David Chaparro1 · Thomas Jagdhuber2,3 · Marta Yebra4,5 · Christoph Rüdiger6 · François Jonard7 ·
Andrew F. Feldman8,9 · Alexandra G. Konings10 · Oliver Binks1,11
same time, drier vegetation and soils increase fuel flamma -
bility, contributing to more frequent, intense, and larger for-
est fires [ 8–10]. Water scarcity also affects the central role
of vegetation in providing ecosystem services, including
climate and weather regulation through land-atmosphere
feedbacks. This influences the global carbon, water and
energy cycles [ 11–16]. Hence, there is considerable inter -
est in monitoring water in vegetation and assessing global
vegetation health and function.
Monitoring water status and dynamics in vegetation is
important for several applications. For instance, the live fuel
moisture content (LFMC), which is the percentage of mass
Introduction
Climate change is causing a warmer and drier atmosphere
and more frequent and intense droughts, leading to higher
atmospheric water demand and in some cases drier soils
compared to prior decades [ 1–3]. These changes constrain
the vegetation’s ability to access water resources, which in
turn reduces the capacity of vegetation to assimilate carbon
and leads to drought-induced tree mortality [ 4–7]. At the
1 3
34 Page 2 of 24
Current Forestry Reports (2026) 12:34
of water relative to the plant’s dry mass, is a key parameter
for fire risk assessment since it is used to model fuel flam -
mability and fire behavior [ 17–21]. Plant water content (θ)
and water potential (Ψ, [MPa]) are key indicators of plant
water status. Water potential, a thermodynamic property
describing the free energy of water and thus its tendency to
flow, strongly influences plant function and physiological
stress, providing essential information about forest vulnera-
bility to drought. Increasing water stress is generally related
to lower water volume and more negative water potential.
Water potential is linked to the plant water content through
pressure-volume relationships. Enhanced monitoring of
vegetation water status can improve our representation of
the soil-plant-atmosphere water flux, which has a signifi -
cant effect on land-atmosphere feedbacks and hydrologi -
cal cycling. This can ultimately result in improved drought
monitoring and forecasting, understanding of plant hydrau-
lics, and climate and meteorological models [11, 22].
Vegetation water status is described by several related yet
distinct variables. Some quantify the amount of water stored
in leaves, stems, or canopies (e.g., θ and LFMC), whereas
others describe the energy state of that water and its tendency
to move through the soil–plant–atmosphere continuum (Ψ).
Table 1 summarizes these variables and highlights those
most relevant to remote sensing, ecophysiology, hydrol -
ogy, fire risk, and drought assessment. Three main meth -
ods exist for estimating these variables: in situ samples (e.g.
[19]), remote sensing [ 17, 18], and process-based models
[20, 21]. In this work, we review the first two and how they
are related across scales. In situ data typically have limited
spatial extent, are temporally sparse, and require substantial
collection effort because the measurements are destructive
and/or expensive. For instance, extensive LFMC sampling
networks have only been deployed at a regional scale in a
few regions [ 38–40]. At a global scale, datasets of in situ
LFMC [ 19, 41] and water potential [ 45] have only been
Table 1 Variables quantifying water in vegetation, classified in water content and water potential groups. The main variables addressed in this
study are indicated in italics. The other variables can be derived from the main ones or are conceptually related: EWT, CWS and RWC are linked
to VWS and θ1. The mg is convertible to/from LFMC [18]. Right-hand columns cite some examples of remote sensing products and in situ datasets/
measurements of each variable
Variable SI units Description Remote
sensing
In
situ
Water
content
Stem water content (θstem) m3/m3 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝑉𝑉𝑉𝑉 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
𝑆𝑆𝑉𝑉𝑉𝑉𝑉𝑉 𝑆𝑆𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉
N/A2 [23]
Leaf water content (θleaf) kg/m2 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹 𝐹 𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹
𝐿𝐿𝐹𝐹𝐿𝐿𝐿𝐿 𝐿𝐿𝐹𝐹𝐹𝐹𝐿𝐿
Vegetation Water Storage
(VWS)
kg/m2 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹 𝐹 𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹
𝑔𝑔𝐹𝐹𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑔𝑔𝐹𝐹𝐹𝐹𝑔𝑔
[24–27]
Equivalent water thickness
(EWT)
mm EWT (water column) is a synonim of VWS since water density is
1 g/cm
[28] [29]
Canopy water storage (CWS) kg/m2 𝜃𝜃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ⋅ 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 [30, 31]
Relative water content (RWC) % In plant physiology (RWCphys)1:
𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹
𝑇𝑇𝑇𝑇𝐹𝐹𝑇𝑇𝑇𝑇𝑇𝑇 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹
⋅ 100
In remote sensing (RWCRS)3:
𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑉𝑉𝑉𝑉𝑉𝑉)
𝑉𝑉𝑚𝑚𝑚𝑚(𝑉𝑉𝑉𝑉𝑉𝑉)𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑉𝑉𝑉𝑉𝑉𝑉)
⋅ 100
[32, 33] [34]
Live fuel moisture content
(LFMC)
% 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹 −𝐷𝐷𝐹𝐹𝐷𝐷 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹
𝐷𝐷𝐹𝐹𝐷𝐷 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹
⋅ 100 [17, 35–37],
Table 2
[19,
38–
41]4
Gravimetric moisture (mg) kg/kg 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹 − 𝐷𝐷𝐹𝐹𝐷𝐷 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹
𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝑙𝑙𝐹𝐹𝑙𝑙𝑙𝑙𝐹𝐹𝐹𝐹
[35, 42]
Water
potential
Leaf water potential (Ψ1) MPa
(MJ/m3) Potential energy of water in the leaf, measured in pressure units. [43, 44] [45]
Xylem water potential (Ψx) MPa
(MJ/m3) Potential energy of water in the xylem, in pressure units.
m = Mass
1RWCphys is related to θ and VWS through stem volume or leaf area but needs the turgid mass to be computed
2Only available using remote sensors in laboratory experiments [ 46], to our knowledge
3Note that, as we detail in Sect. "Assessing vegetation water content: Building a bridge between remote sensing and plant ecophysiology", RWC as
defined in remote sensing literature (RWCRS) is actually a relative signal strength, not RWC of the plant as defined in ecophysiology (RWCphys). In
RWCRS maximum and minimum of VOD can be substituted by percentiles 95 and 5, respectively, to reduce the impact of extreme values (outliers)
4LFMC = [mg/(1- mg)]·100 [18]
1 3
Page 3 of 24 34
Current Forestry Reports (2026) 12:34
established recently, with most measurements available in
Europe and the USA [ 19, 45]. Because these datasets rely
on direct measurements of plants, they can provide bench -
marks for the calibration of models [ 47] and the validation
of satellite retrievals [35, 36] (Table 1).
Satellites allow regular monitoring of the land surface from
local to global scales. They capture vegetation water dynam-
ics through water-related reflectance changes with optical sen-
sors [53], changes in leaf temperature with infrared sensors
measuring thermal emissions [17], or water-induced changes
in emission and backscatter with microwave sensors [ 18].
Importantly, different frequencies at which the sensors operate
have different sensing depths within the canopy. Optical and
thermal sensors, similarly to high frequency microwaves (e.g.,
Ku-band; ~18 GHz), are only sensitive to the top-of-the-can-
opy leaves, while low microwave frequencies (e.g., L-band at
1.4 GHz) are likely sensitive to the full canopy depth [54–57].
This frequency-depth relationship also depends on the canopy
density and homogeneity [58]. Thus, the choice of the fre -
quency and remote sensing technique to be used depends on
the system and tissue that one aims to study. High microwave
frequencies and some optical systems are appropriate for
grasslands and shrublands and for forest leaf conditions, while
low microwave frequencies (e.g., L-band) are appropriate to
study the full forest canopy depth and are also sensitive to the
woody tissues [54, 56].
Considering the spatiotemporal domain of satellite sensors,
orbits of the satellites carrying optical and thermal sensors
allow sub-weekly overpasses at spatial resolutions down to
10 m (or submeter scales for some commercial satellites) [59,
60]. However, the effective temporal availability of data can
be reduced because these sensors cannot measure the surface
when it is covered by clouds. Active microwave satellites have
comparable spatial resolution of tens of meters but at mostly
biweekly time resolutions (e.g., Sentinel-1, ALOS-2-PAL -
SAR, or NISAR [61, 62]). Passive microwave satellite sensors
such as those on the Soil Moisture and Ocean Salinity (SMOS;
[63]), on the Soil Moisture Active-Passive (SMAP; [64]), and
the Advanced Microwave Scanning Radiometers (AMSR-
E/2/3; [65, 66]) have coarse resolutions (15–40 km) but higher
temporal resolution (2–3 days). Unlike optical sensors, micro-
wave remote sensing is largely unaffected by clouds.
Despite the recent advances in remote sensing for moni -
toring water in vegetation [35, 36, 42, 43] and the growth of
in situ datasets across more locations [ 19, 41, 45], signifi-
cant challenges remain in connecting remote sensing with
in situ data characterizing the water status of plants. This
is due to differences in spatial resolution (point-to-pixel
scale gap) and to the need to improve the methods used
to translate remote sensing observations into accurate and
reliable indicators of plant water status. For instance, [ 24]
details how satellite estimates of vegetation water storage
(VWS) can provide information linked to the water status
of plants, and highlights the need for upscaling field studies
in ecophysiology from plot to pixel scales by developing
pressure-volume curves [ 67] at the ecosystem scale (EPV
curves) [68, 69]. These are water content vs. water potential
relationships established beyond the tree level that can be
applied to derive vegetation water potential (Ψ) informa -
tion from satellite VWS estimates [ 24, 68, 69]. Also, [ 70]
coins the term ‘water potential information gap’, describing
the existing gap between sparse, irregular samples of in situ
vegetation Ψ (mostly based on pressure chamber measure -
ments) and the need for more regular and continuous data of
this variable. This gap could be bridged with a more exten -
sive use of psychrometers, by deploying Ψ networks like
PSInet [ 45], and by linking them to the targeted but still
unachieved large-scale Ψ estimation from remote sensing
[70]. The remainder of this manuscript will investigate this
and other research gaps in monitoring water in vegetation,
including water content, fuel moisture, and water potential.
To illustrate the main knowledge gaps, we conducted a
structured literature search for the period 2014–2024 (see
SI1 for the methodology) and summarized the results in
Fig. 1. As a result, Fig. 1a classifies the most relevant lit -
erature on remote sensing of vegetation water content and
water potential. We observe that, while local-scale research
on water in vegetation is abundant, important gaps remain
at regional to global scales (Fig. 1a).
We focus on large-scale gaps by exploring how regional
to global remote sensing of the focal variables (Table 1)
can be better linked with existing in situ datasets. Specific
research gaps for each variable (Fig. 1b) are explored in
Sects. "Monitoring live fuel moisture content (LFMC) " to
"A road towards monitoring vegetation water potential (ψ)
from remote sensing". Section "Monitoring live fuel mois -
ture content (LFMC) " provides, to our knowledge, the first
comparison among large-scale LFMC datasets and against in
situ data (see [71] for one smaller-scale example). The same
section analyses which strategies should be used to bridge
the spatial-scale mismatch between in situ LFMC data and
pixel-scale satellite estimates in validation procedures. Sec-
tion "Assessing vegetation water content: Building a bridge
between remote sensing and plant ecophysiology" addresses
the challenges related to water content. This includes empir-
ical simplified estimates of VWS [25], the lack of plot-scale
data to validate them, and the need to understand which eco-
physiological processes, canopy levels, and plant tissues are
represented in satellite vegetation water estimates. Finally,
Sect. "A road towards monitoring vegetation water potential
(ψ) from remote sensing" reviews new opportunities based
on passive microwave signals and multi-frequency sensor
synergies that can enable the assessment of spatial, temporal
and vertical variability of Ψ [44, 72].
1 3
34 Page 4 of 24
Current Forestry Reports (2026) 12:34
In summary, the research gaps addressed in this study can
be grouped into two main research questions. First, which
strategies are needed to achieve and/or enhance large-scale
products of water in vegetation? Second, how can ground-
based data be better integrated to establish a stronger link
with remote sensing estimates? (Fig. 1b). The goal of this
paper is to address these questions and provide guidance
on how the gaps illustrated in Fig. 1 can be reduced. This
review intentionally pays special -although not exclusive-
attention to passive microwave remote sensing for several
reasons. First, passive microwaves are particularly appropri-
ate to sense water across different canopy layers and have
demonstrated large potential for the assessment of vegetation
water status [ 18, 24]. Second, plot-scale microwave-based
approaches relying on Global Navigation Satellite Systems
(GNSS) [73, 74] are expanding and will complement in situ
samples for validation and interpretation of microwave sat-
ellite vegetation water estimates. Third, intensive research
has been conducted in the passive microwave remote sens -
ing field on the connection between satellite and GNSS mea-
surements with vegetation canopy water variables during the
last five years. We review this research and propose further
development.
This review aims to inform a wide range of research fields
and end-users working across scales, including plant ecophys-
iologists, fire risk assessment agencies (mainly working with
LFMC data), and Earth system modelers (e.g., using vegeta-
tion water status to improve land-atmosphere feedback esti-
mates). Our work is applicable to all vegetation types, even if
some discussion focuses on forests, where the availability of
different microwave frequencies has the advantage of being
sensitive to different canopy layers and tissues.
Monitoring Live Fuel Moisture Content
(LFMC)
LFMC Estimates from Remote Sensing: Physical
Background and Available Products
LFMC refers to the mass of water contained in living plant
matter, expressed relative to its dry mass. Historically, the
Fig. 1 a Number of remote
sensing papers selected ordered
by spatial scale (rows) and by
groups of variables in Table 1
(columns; ‘water content’ and
‘water potential’). Gaps are iden-
tified when ten or fewer papers
are found in each row-column
pair for the period 2014–2024.
The list of literature references
containing all the papers counted
in this figure is in Supplementary
References. b A text box for
each group of variables details
the main research gaps or chal-
lenges at large scales and the
section in the paper where they
are addressed. These challenges
lead us to the two main research
questions of our paper (red box).
Questions and challenges are
linked by green squares and red
triangles in the figure. Note that
“large scale” in Fig. 1b refers to
continental and global scales.
See SI1 for more details on the
methodology of the review
1 3
Page 5 of 24 34
Current Forestry Reports (2026) 12:34
optical domain has been used to infer LFMC. Water in
plants mainly absorbs the electromagnetic waves in the
short-wave infrared (SWIR) and, to a lesser extent, in the
near infrared (NIR) regions of the spectrum. Thus, varia -
tions in absolute water content in plants modify the NIR and
SWIR reflectances, leading to covariation between water
content and NIR/SWIR [ 17, 53]. In addition to this direct
effect, plant water stress can also indirectly influence reflec-
tance through the depletion of chlorophyll content, which
affects the visible region of the spectrum. LFMC retriev -
als from optical sensors generally rely on vegetation indi -
ces and statistical methods [ 75–78], machine learning [ 48,
49, 52, 79–81], or the inversion of radiative transfer models
(RTM) [50, 51, 82–85]. In the review period (2014–2024),
six large-scale LFMC estimates have been published using
only optical sensors [48–52]. These products are listed and
briefly described in Table 2.
The microwave domain of the spectrum is also sensi -
tive to changes in the water content of plants, since water
modifies the dielectric constant of the canopy, and thus
its capacity to emit, reflect and attenuate electromagnetic
waves [18]. Because of this, data from microwave sensors
are used to estimate the vegetation optical depth (VOD), a
variable quantifying the attenuation of the microwaves that
pass through vegetation after being emitted by the land sur-
face. VOD is not only sensitive to the water, but also to the
biomass and structure of different parts of the canopy, thus
being high for wet and dense canopies [86]. Recent research
has isolated the water component (expressing it as LFMC)
from the biomass-structure component of the VOD signal.
It has done so by applying semi-physical algorithms using
multi-sensor synergies both in the western USA [ 35] or
globally [87]. Also, large-scale LFMC products have been
established by combining Sentinel-1 radar backscatter with
Landsat 8 reflectance in the western USA [37], or have used
statistical approaches combining VOD and leaf area index
(LAI) to map LFMC globally [36] (Table 2).
Field Data Observations of LFMC
The most comprehensive in situ LFMC dataset has been
gathered in the Globe-LFMC database [ 41], which has
been recently updated in 2024 (Globe-LFMC v2; [19]). The
updated version contains ~ 294,000 LFMC samples from
Table 2 Large-scale LFMC datasets. All of them, except [48] and [49] are intercompared in this study
Product Region Period available
(Period
intercompared)
Sensors Res. Approach
[50] Global 2001–2019
(depending on the
region)
MODIS 0.05º Applies RTMs to simulate spectra for grasslands, shrublands and
forests. Then, LUT inversions by land cover minimize differences
between simulated and satellite data to estimate LFMC.
[36] Global 1997–2018
(depending on the
region)
TMI, Wind-
Sat, AMSR-
E, AMSR-2,
MODIS
0.25º VOD, LAI, and tree cover are found to be the best estimators of
LFMC based on field-scale regressions. Then, random forests are
applied to tailor the regression parameters and build LFMC maps.
[49]1 CONUS 2014–2017
(N/A)
MODIS 500 m. One year of MODIS reflectance data, meteorological data, and
field LFMC data are used in multi-modal, multi-temporal convolu-
tional neural networks (CNN) to estimate LFMC.
[48] 2 CONUS 2019–2021
(N/A)
MODIS 1 km. Field LFMC, MODIS reflectance, LST, soil saturation, ET, land
use and topography are used to calibrate a random forest model
for LFMC estimation.
[37] West US 2015–2019
(Apr. 2015 – Dec.
2018)
Landsat
8 OLI,
Sentinel-1
250 m. Landsat reflectance and Sentinel-1 radar backscatter are used in a
recurrent NN to estimate LFMC with help of auxiliary variables
(e.g., soil textures or canopy height).
[35] West US 2015–2018
(Apr. 2015 – Dec.
2018)
AMSR-2,
SMAP,
GEDI,
Sentinel-1
0.25º A semi-physical algorithm applies canopy height and radar
backscatter to estimate VOD. Differences with satellite VOD are
minimized to derive the dielectric constant of vegetation (ε). Then,
LFMC is derived from ε using dielectric models.
LFMC derived from X-band VOD is used in the comparison, since
it is the most consistent with leaves and twigs in situ samples [35].
[51] Australia 2001–2017
(2001–2017)
MODIS 500 m. Applies RTMs to simulate spectra for grasslands, shrublands and
forests. Then, LUT, land cover maps and a merit function are used
to estimate LFMC.
[52] Mediterranean
basin
2001–2021
(Jan. 2001 – Jul.
2017)
MODIS 500 m. Random forest models are built to estimate LFMC based on field
LFMC, MODIS reflectance, vegetation indices, surface tempera-
ture and the day of the year.
1Since the product is not publicly available it was not included in the product comparison. 2The study period in [ 48] does not match the study
period of other datasets in the region. Hence [ 48], is excluded from the comparison. Acronyms: LUT (look-up-table), GEDI (Global Ecosystem
Dynamics Investigation), TMI (Tropical Rainfall Measuring Mission’s Microwave Imager)
1 3
34 Page 6 of 24
Current Forestry Reports (2026) 12:34
2,245 sites collected across 15 countries over the period
1977–2023. However, it is worth noting that 80% of data
and 55% of sites are in the United States.
In situ LFMC data is measured in single leaves (only
occasionally in small twigs), usually at the top of the can -
opy, and most samples are taken during the summer [ 19,
41]. These measurements are destructive and labour inten -
sive. LFMC is an appropriate variable to predict the risk of
flammability and fire propagation, especially in short plants,
since the leaves’ moisture is representative of the entire
plant hydric conditions. In the case of forests, LFMC can -
not represent the water status of entire trees, since it is not
measured in stems or branches along the full trees’ heights.
Still, it accounts for the dryness of tree crowns and of the
understory, which are main vectors in fire propagation.
The global availability of in situ LFMC has two main
advantages for scientific research and satellite product
development. First, the large amount of data is an essential
source of information for machine learning approaches to
produce more robust large-scale satellite LFMC estimates.
Indeed, MODIS and Globe-LFMC have been combined to
build LFMC maps across both the continental USA [48, 81]
and the Mediterranean basin [ 52]. Secondly, Globe-LFMC
has also been used as benchmark data to validate large-scale
satellite-derived LFMC datasets [35, 36, 52, 81]. However,
each of these products has been validated individually, and
limited effort has been put into using global in situ LFMC
data to compare them (one example is [71], but it is confined
to Israel). Consequently, we have limited knowledge of how
the different products and their underlying techniques per -
form comparatively. We aim to close this gap by conducting
filtering and comparison analyses in the next section.
Comparison of LFMC Datasets and Assessment of
the Point-to-Pixel Scale Mismatch
A central obstacle in evaluating LFMC products is that the
field measurement and the satellite retrieval do not repre -
sent the same spatial scales of the vegetation components.
They also represent different components of the canopy:
field LFMC is usually measured on selected leaves or small
twigs, whereas satellite products integrate over mixed pixels
and, depending on the sensor, over different canopy compo-
nents. We therefore use the available product comparison
not only to rank datasets, but to identify which kinds of fil -
tering improve the biological representativeness of valida -
tion sites.
The literature review finds eight datasets published in the
period 2014–2024, of which two are global, two span the
entire conterminous United States (CONUS), two cover the
western USA, and one each in the Mediterranean basin and
Australia (Table 2). We will compare them against in situ
data and explore the best quality filtering approaches in this
section. We acknowledge that some of these products [36, 48,
51] have been calibrated using in situ LFMC data. This may
impact the results by increasing correlations of these satellite
products with in situ LFMC and by decreasing the error and
bias statistics. Note that it is not possible for us to distinguish
which sites were applied to the calibration (and, consequently,
it is not feasible to discard them in the comparison).
The validation and comparison of LFMC estimates using
in situ data is hampered by representativeness issues. There
exists a point-to-pixel spatial scale mismatch, which, in the
case of LFMC, refers to the capacity of a single point mea -
surement to describe the moisture state and dynamics of
plants in a pixel tens of meters to kilometers across that can
be heterogeneous usually in terms of land cover (e.g., [ 35,
36]), but also in terms of the species sampled or in terms
of the physical representativeness of the individual sample
within a canopy [35, 41]. To deal with these issues, different
filters exist to select the most representative sites for vali -
dation using, for example, the spatial variability of NDVI,
the land cover heterogeneity in a satellite pixel, or metrics
to determine the diversity of species sampled across a site
(Table 3). To compare filters and products, we use the stan-
dard metrics Pearson’s correlation coefficient (r), RMSE
and bias. Correlation is only meaningful when the in situ
LFMC time series has sufficient temporal variability ([ 35,
37], Stdsite in Table 3, and SI2).
Here, we first evaluate how homogeneity and quality fil-
ters reported in the literature (Table 3) affect the comparison
statistics of selected large-scale LFMC datasets (Table 2)
in the western USA and the Mediterranean Basin (Figure
S1). The SI2 provides methodological details. Correlations
between remote sensing and in situ datasets in the western
USA responded differently depending on the filter applied
(Fig. 2). Filters based on land cover homogeneity (D LC and
LC) or vegetation greenness (CV NDVI, Max ) generally had
minimal effects, with correlation changes either absent or
small (Δr < + 0.1). In contrast, filters representing the homo-
geneity of sampled species (Sp1 and Sp2) consistently led
to stronger improvements, with correlations increasing by
roughly 0.1 to 0.3 (Fig. 2c and d). These filters ensure that
the species sampled are always the same or at least consistent
with the pixel-dominant vegetation type. Future research
should explore the impact of intra-specific variability
across climate, soils and topography, which are relevant to
explain moisture variations in plants [89]. Also, it is impor-
tant to consider the variability in time of the in situ data. As
expected, sites with high Std site show a higher correlation
than sites in which time-variability of in situ LFMC is low
(Fig. 2f). Generally, the results presented are consistent with
those obtained in the Mediterranean basin (Figure S2) and
show that three filters are promising to achieve consistent
1 3
Page 7 of 24 34
Current Forestry Reports (2026) 12:34
screening across products and regions: Sp1, Sp2 and Std site
(see Table 3 for more details on each filter). Still, in some
cases, applying filters Sp1 and/or Std site≥35% increases the
bias and/or the RMSE (Figures S3 to S6), suggesting that
Sp1 may not be fully appropriate and that a lower Std site of
20% enhances the selection criteria.
Combining the species homogeneity or time variability
filters with land cover ones does not change the results sig -
nificantly (not shown). This confirms how accounting for land
cover homogeneity is not enough to bridge the point-to-pixel
scale mismatch. This representativity issue is particularly
important in short-statured, agricultural, and mixed shrub–
grass systems. In such ecosystems, water content and water
potential can often be measured directly using destructive
sampling and pressure-chamber measurements. Therefore, the
main limitation is not necessarily the feasibility of field mea-
surements, but whether those measurements represent the veg-
etation component sensed by the satellite pixel. For example,
coarse pixels may include mixtures of bare soil, litter, grasses,
shrubs, crops, irrigated areas, or vegetation at different pheno-
logical stages. In low-biomass systems, the vegetation water
signal can also be more strongly affected by soil background,
surface wetting, canopy cover, and management practices than
in closed forests. Future research should therefore combine
LFMC or water-potential measurements with information on
fractional vegetation cover, species or crop type, phenologi-
cal stage, management or irrigation status where relevant, and
repeated sampling across the satellite footprint.
In a second step, we undertake our comparison among
LFMC products in the Western USA, the Mediterranean, and
Australia using the best filtering strategy identified earlier (Fig-
ure S1 and Table S3). According to the results on filters, we
simultaneously apply filters Sp2 = 75% and Stdsite≥20% and
discard other combinations (SI2). Overall agreement between
the products and the in situ samples was moderate to strong
across the three regions (median r: 0.5–0.9, RMSE < 30%,
median bias < 60%; Fig. 3). Note that the RMSE and bias per-
centages are in LFMC units, which means that, for example, a
bias of 60% represents 15% of the maximum range of LFMC.
This is based on the fact that LFMC can reach ~ 400% accord-
ing to [19] and assumes a minimum LFMC of zero.
In the western USA, three products have high median
correlations, ranging from 0.72 to 0.84, while the product
in [ 35] shows a smaller median correlation ( r = 0.5), but
a lower bias (< 20%). We suggest that this lower correla -
tion in time may be due to the different timescales used in
the dynamic parameters (δ and VOD) applied in the semi-
physical model used in [ 35]. A low bias is also found in
the region in [36]. Low biases in both VOD-based products
suggest that accounting for vegetation structure using LAI
[36] or removing the biomass and structure signals in VOD
using canopy height and radar data [35] can isolate the veg-
etation moisture component from the passive microwave
signal. Products based on optical combined with radar data
[37] or without it [ 50] show the highest correlations in the
western USA and keep the median r ~ 0.8 in the Mediterra-
nean, like the product in [52]. This might be influenced by the
higher original spatial resolution of optical products compared
to that of the VOD-based ones, even if an aggregation has been
conducted (SI2).
Table 3 Homogeneity and quality filters reported in the literature which impact the validation results. In the column “Type” the spatial filters screen
out pixels or sites with large spatial heterogeneity, the species filters aim at ensuring consistency between the plant species measured at a site and
the dominant vegetation within a pixel, and the temporal filter screens out sites with low time-variability of in situ LFMC. Columns “More strict”
and “Less strict” indicate the range of conditions studied in Fig. 2 for each filter
Name Type Description Literature
values
More
strict
Less
strict
CVNDVI, Max Spatial Maximum Coefficient of Variation (CV; [%]) in time of monthly averaged NDVI,
computed on an area of 500 m × 500 m around the site’s coordinates, provided by
[41]1.
15–30%
[50]
26% [36]
25% [35]
15% 40%
DLC Spatial Land cover distance between 1 km and 0.25° pixels. It uses the percentages of tree
cover (TC) and short vegetation (SV) from [88] at both resolutions:
𝐷𝐷𝐿𝐿𝐿𝐿 = √(𝑇𝑇𝑇𝑇25𝑘𝑘𝑘𝑘 − 𝑇𝑇𝑇𝑇 1𝑘𝑘𝑘𝑘)2 + (𝑆𝑆𝑆𝑆25𝑘𝑘𝑘𝑘 − 𝑆𝑆𝑆𝑆 1𝑘𝑘𝑘𝑘)2
< 10%
[36]
5% 20%
Sp1 Species Exclude multi-species LFMC sites unless species LFMC time series are correlated
(Pearson r ≥ 0.5).
[37] Applied2 Not
applied
Sp2 Species/
Spatial
Sites where x% of the sampled species are of the same vegetation type (tree, shrub
or grass) as the dominant in the pixel.
100%
[36]
100% 0%
LC Spatial Sites where the dominant land cover spans through ≥ x% of the pixel area are chosen 60% [35] 100% 0%
Stdsite Temporal Sites where standard deviation of LFMC in time is > x are chosen. In sites with low
variability, correlation is less useful because variability can be driven by noise [35,
37]. This metric is applied in [35] but originally based on [37], where the authors
applied the coefficient of variation in time (instead of the standard deviation).
- 0% ≥ 50
1Not available in Globe-LFMC v2, so we use Globe-LFMC v1. 2We also test excluding all sites with > 1 species (regardless of correlations).
Seasonalities are omitted due to poor information on how the authors built them
1 3
34 Page 8 of 24
Current Forestry Reports (2026) 12:34
Results in Australia and the Mediterranean show that the
product in [ 36] (passive microwaves) and optical products
[50, 51] are similar in correlation. The bias and RMSE in
Australia are low for [ 51] and [ 50] which, together with
their high resolution, make these products suitable for
LFMC estimations there. The correlations in Australia
(0.55 < r < 0.63) compared to those obtained in the Mediter-
ranean basin and the western USA (0.72 < r < 0.90 for four
out of five products) would likely be greater if quality fil -
ters could have been applied. Finally, it is worth noting that
we have not found a specific technique (i.e., random for -
ests, neural networks, RTMs, or semi-physical algorithms)
outperforming others in the estimation of LFMC (see Fig. 3
and Table 2).
Assessing Vegetation Water Content:
Building a Bridge Between Remote Sensing
and Plant Ecophysiology
Monitoring the Water Status of Vegetation
While LFMC is principally applied in fire risk assessment,
water in an ecosystem can also be measured using variables
Fig. 2 Western USA. Effect of
progressively stricter site-screen-
ing criteria (as in Table 3) on the
agreement (median correlation)
between satellite-derived LFMC
products and in situ LFMC
measurements. The satellite
LFMC products are represented
by coloured lines. The legend
provides the products’ refer-
ences. The subfigures show: a
the maximum spatial coefficient
of variation of NDVI at a site
(CVNDVI, Max), b the land cover
‘distance’ between 1-km and
0.25º pixels (DLC), c the filter
removing LFMC sites with more
than one species sampled (Sp1),
d the percentage of sampled
species which are of the same
vegetation type as the dominant
in the pixel (Sp2), e land cover
homogeneity as the percentage of
dominant land cover in the pixel
(LC) and f standard deviation of
in situ LFMC (Stdsite)
1 3
Page 9 of 24 34
Current Forestry Reports (2026) 12:34
that rep resent the water status of plants and are directly
linked with ecosystem functions including the fluxes of
water, carbon and energy. The vegetation water storage
(VWS) [
𝑘𝑘𝑘𝑘𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝑚𝑚2𝑔𝑔𝑤𝑤𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ] is the water contained in the vegetation relative to ground area. It is directly convertible to the absolute
mass of water when multiplied by the area (e.g., pixel size), or
to the water column when divided by the water density. The
latter is also named equivalent water thickness (EWT [mm];
Table 1). VWS or EWT are extensive variables (i.e., variables
depending on the size of the system; [68]) which scale linearly
with biomass. Generally, a high-biomass forest contains more
water and thus a larger VWS than a similarly hydrated forest
with less biomass. On the other hand, water content relative
to wood volume (θstem; [
𝑚𝑚3𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝑚𝑚3𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 ]) and water content relative
to leaf area (θleaf; [
𝑘𝑘𝑘𝑘𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝑚𝑚2𝑙𝑙𝑤𝑤𝑤𝑤𝑙𝑙 ]) are intensive variables (i.e., inde-
pendent of the size of the system) representing physical
properties of the tissues (wood and leaves) [ 68]. Ecophysi-
ologists also define water content relative to a tissue-level
saturated value (RWC phys) (Table 1). θ and RWC phys are
used in ecophysiological field research -together with Ψ- to
study the plant water status and its vulnerability to drought
[67].
Currently, satellite-based mapping of water content
relies mostly on estimating the VWS (e.g., microwave
measurements are more sensitive to VWS than to intensive
Fig. 3 Comparison between satellite and in situ LFMC time series. Columns show the regions; rows correspond to temporal correlation, RMSE,
and bias. Boxplots represent median (black line), 25th and 75th percentiles (limits of the boxes) and 5th and 95th percentiles (whiskers)
1 3
34 Page 10 of 24
Current Forestry Reports (2026) 12:34
variables). This limits our ability to fully quantify water sta-
tus of forests independently of the size of the system, as
needed in ecophysiological research. To our knowledge,
remote sensing of θstem and θleaves has only been conducted
in laboratory experiments using terahertz spectroscopy,
which is not comparable to satellite techniques [ 46]. The
lack of direct measures of θ is a clear limitation to enable
monitoring of forest water status from space.
RWC, and its approximation in the remote sensing com-
munity (here called RWC RS), is poorly related to the eco -
physiological meaning and definition of RWC phys. RWCRS
is computed by normalizing the sensor signal (e.g., VOD)
between percentiles 5 and 95 of the time-series, usually con-
sidering periods of relatively stable biomass to ensure that
changes in VOD are only due to variations in the water content
[24, 32, 33, 43, 72]. Thus, it represents a fractional variation
within the sensed range of values, that is, a relative measure of
signal strength. An RWC in the ecophysiological sense would
require remote observations to determine when the vegetation
is at the saturation point and when it is completely dry (see the
RWC formulas in Table 1). Importantly, RWCRS, which var-
ies between 0% and 100%, is different in range from RWC-
phys, which typically varies between > 70% and < 100%. This
minimum value of ~ 70% is explained by mortality thresholds
in woody tissues being around 70% to 90% of the saturated
value [90] and has considerable variation. Moreover, the range
obtained by RWCRS is dependent on the conditions during
the measurement period (e.g., the lowest/highest values in the
remote sensing time-series are assumed to be representative
of extreme drought/wetness situations, which may not be the
case). Still, RWCRS has been used to explain regional-scale
forest mortality [32] and has been related to Ψ (Sect. "A road
towards monitoring vegetation water potential (ψ) from remote
sensing"). This can be explained because RWCRS and RWCphys
can be proportional and because normalizing a remote sensing
signal like VOD allows comparison of different ecosystems
across space. Notwithstanding, there exists a relevant concep-
tual and naming gap between RWCphys and RWCRS.
Still, estimating VWS, RWC or θ alone is not enough for
monitoring the water status of vegetation. This is because:
(i) plant growth, ecosystem dynamics, and carbon and water
cycles are mediated by Ψ, (ii) pressure-volume curves,
which define the relationship between θ (or RWC phys) and
Ψ, have not been established beyond the tissue scale (i.e.,
leaves and sapwood), and (iii) these relationships likely vary
along the vertical profile of plants [ 24]. These are relevant
gaps between large-scale remote sensing and the current
local-scale capacity to assess the water status of vegeta -
tion in ecophysiology, which is almost exclusively based on
field data. To help bridge this gap we propose a workflow to
derive not only water content, but also water potential, from
remote sensing VOD data (Fig. 4). In the following sections,
we review the current satellite VWS research (Sect. " Map-
ping VWS from space ") and address the gaps pointing to
the potential of developing satellite θ measurements (Sect.
"Linking satellite and in situ water content information "),
the relevance of vertical gradients (Sect. " Relationship of
microwave sensing depth to the water status from different
plant heights and tissues"), and how satellite-derived Ψ esti-
mates can be approached (Sect. "A road towards monitoring
vegetation water potential (ψ) from remote sensing").
Mapping VWS from Space
Top-of-canopy VWS (i.e., CWS [
𝑘𝑘𝑘𝑘𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝑚𝑚2𝑔𝑔𝑤𝑤𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ]; Table 1) has been
retrieved by establishing a relationship with satellite opti -
cal reflectances using RTMs and machine learning [30, 31].
Also, VWS has been estimated based on NDVI and stem
factors (i.e., the product between the average height of a
land cover class and the ratio of sapwood to leaf area [104]).
In both cases, the use of optical data constrains the sensing
depth to top-canopy leaves and branches, at least in closed
canopies. With these approaches, dryness in canopies can
be detected, but it cannot be understood whether it is caused
by (i) steep vertical gradients of water status (i.e., dry leaves
but wet stems, which is common during periods of high
transpiration) or (ii) entirely dry vegetation under stress,
reaching critical conditions.
Passive microwave remote sensing uses the linear rela -
tionship between VOD and VWS to estimate VWS. This
direct relationship is enabled by a constant called ‘b-param-
eter’ [86] (Fig. 4). However, the values of the b-parame-
ter are generally assumed to be time-invariant and mostly
based on a land cover look-up table [104]. The b-parameter
fields have limited spatial variability and do not consider
phenological changes [ 105, 106]. Consequently, they are
subject to uncertainty and debate in the microwave retrieval
algorithm community [106, 107]. Thus, a better understand-
ing of the spatio-temporal variation of the b-parameter is
needed to propose enhanced formulations to retrieve it.
Alternatively, semi-empirical models and artificial neural
networks (ANNs) based on multi-frequency microwave
remote sensing provide good results when estimating VWS
at high microwave frequencies [ 27]. VWS has also been
mapped with radars [ 101, 102] and with global navigation
satellite systems reflectometry (GNSS-R) measurements,
for instance from the CYGNSS satellite [ 103] (Fig. 4). In
fact, CYGNSS was used to estimate tropical and sub-trop -
ical VWS using linear models and ANNs, being the only
quasi-global-scale VWS map based on L-band data which
does not use the b-parameter [103]. This suggests an inter-
esting complementarity of AI with new signals of opportu -
nity to monitor VWS.
1 3
Page 11 of 24 34
Current Forestry Reports (2026) 12:34
Fig. 4 Workflow linking VOD estimations with the plant water sta -
tus through VWS at in situ and plot scales (left side) and at satellite
(coarse) scale (right side). Black arrows and boxes indicate variables
and the fluxes between them. Red lines and boxes indicate measure -
ments, while blue ones indicate models. Sections and tables where
each step is described are referred above or below these lines. The
workflow shows how VOD is retrieved from satellites using the τ−ω
model (or at the plot level using GNSS-T). VWS is then retrieved from
satellite VOD using the b-parameter approach, but alternative tech -
niques at satellite (e.g., GNSS-R) and plot (Eq. (1) to (3)) scales have
been proposed. We show how VWS and Ψ are related via θ by using
EPV curves [68], which can theoretically allow Ψ estimations at both
plot and satellite scales. Both θ and Ψ can be measured in situ for vali-
dation purposes or, in the case of Ψ, estimated with vegetation models.
The workflow also shows that RWC phys can be derived from VWS or
θ, and that normalizing VOD leads to a remote-sensing-based proxy
of (but not an equivalent quantity of) RWC (i.e., RWC RS). Note that
LFMC is not included in the workflow for the sake of simplicity (but
check Eqs. (3), (5) and (6) in the text)
1 3
34 Page 12 of 24
Current Forestry Reports (2026) 12:34
Linking Satellite and In situ Water Content
Information
A. Building Plot-Scale VWS Estimates to Validate and
Interpret Satellite VWS
The main limitation of satellite-based VWS estimates is that
independent data for validation is lacking. In situ VWS mea-
surements have been published only in few cases (e.g. [29]).
Meanwhile, the US National Ecological Observatory Network
(NEON) database contains canopy water indices obtained from
airborne optical reflectance measurements [108]. Also, recent
developments have led to global navigation satellite systems
transmissometry (GNSS-T) sensors enabling plot-scale (up to
~ 100 m in diameter) estimates of VOD. These estimates are
based on the difference in signal-to-noise ratios between anten-
nas placed above and below the canopy [73]. However, these
are still indirect estimates of local-scale VWS, not retrievals
of it. This lack of point-to-plot scale VWS data also limits our
understanding of plant-water relations across scales. To bridge
this gap, a potential approach can be the estimation of plot-scale
VWS derived from GNSS-T VOD data, which has already been
used to estimate vegetation moisture [73] (Fig. 4).
Here, we also propose another approach taking advan -
tage of in situ measurements of θstem and θleaf available from
ecophysiological research sites [23, 95, 96]. These measure-
ments have to be upscaled and transformed into plot-scale
VWS estimates, which can serve for validation of satellite
VWS or CWS (Table 1; Fig. 4) if the vegetation in the plot
is representative of that in the pixel (Sect. " Monitoring live
fuel moisture content (LFMC) ").
The purpose of the following equations is not to propose
a single operational retrieval algorithm, but to explicitly
state which field measurements are needed to connect tis -
sue-scale water content with plot- or pixel-scale VWS. Stem
and leaf water stores require different scaling approaches.
Thus, for stems (adapted from [68]):
𝑉𝑉𝑉𝑉𝑉𝑉 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 =
∑
𝑖𝑖 𝑠𝑠𝑖𝑖
𝑖𝑖𝑖𝑖
−
𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 .𝜌𝜌𝑤𝑤𝑤𝑤𝑠𝑠𝑠𝑠𝑤𝑤 .𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 𝑠𝑖𝑖
𝐴𝐴𝑔𝑔𝑤𝑤𝑤𝑤𝑔𝑔𝑔𝑔𝑤𝑤
(1)
where
−
𝜃𝜃 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is the average θstem measured in situ for each
𝑖𝑖 species, 𝜌𝜌 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 is the density of liquid water [1000 kg/
m3], 𝐴𝐴𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 is the plot ground area [m2] and 𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 is the
volume of wood [ 𝑚𝑚3𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 ] of each species. Note that θ stem
can be measured by ecophysiologists using stem sensors
based on time- or frequency-domain reflectometry [ 109]
or electric resistance sensors [ 99]. Networks of such sen -
sors can become a reality soon since there has been prog -
ress in developing general standards for their calibration
[110]. Plot-scale 𝑉𝑉𝑉𝑉𝑉𝑉 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 can be used for validation and
interpretation of VWS derived from L-band VOD because:
(i) L-band has large canopy sensing depth, and (ii) its long
wavelength (~ 21 cm) makes it much more sensitive to
larger structural components (stems, branches, trunks) than
to leaves (see Sect. " A road towards monitoring vegetation
water potential (ψ) from remote sensing " and [ 111]). Still,
note that reliably deconvoluting the contributions of leaves
and stems within the L-band signal is not possible. We
acknowledge this as an inherent limitation of this approach.
To estimate plot-level VWS in leaves, one has to use
Eq. (2) (adapted from [68]):
𝑉𝑉𝑉𝑉𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝐿𝐿𝐿𝐿𝐿𝐿𝐿
𝑖𝑖 𝑙𝑙𝑖𝑖
∑
𝑖𝑖𝑖𝑖
−
𝜃𝜃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑖𝑖 ⋅ 𝐿𝐿𝑙𝑙𝑙𝑖𝑖 (2)
where 𝐿𝐿𝐿𝐿𝐿𝐿 is the leaf area index [𝑚𝑚2𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 / 𝑚𝑚2𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ],
−
𝜃𝜃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
is the average θ leaf measured in situ for each 𝑖𝑖 species and
𝐴𝐴𝑓𝑓𝑓𝑓𝑓 is the leaf area fraction covered by each 𝑖𝑖 species in
the plot. Importantly, plot-scale 𝐿𝐿𝐿𝐿𝐿𝐿 and leaf area fractions,
as well as 𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 in Eq. (1), can be estimated with terrestrial
laser scanners (TLS) [112–114] or UA V-based LiDAR sen-
sors [115]. In cases where the plot under study is sufficiently
representative, plot-scale 𝑉𝑉𝑉𝑉𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 can be used for valida-
tion and interpretation of satellite CWS derived from high
VOD microwave frequencies [27] (X- and Ku-bands; Sect.
"A road towards monitoring vegetation water potential (ψ)
from remote sensing") or from optical satellite sensors [30].
In addition, they can be used to compare plot-scale VOD and
derived estimates based on GNSS-T sensors [73] (Fig. 4).
An alternative for the estimation of 𝑉𝑉𝑉𝑉𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 , in this case at
the plant scale, can be using data from in situ LFMC networks,
which are mostly based on samples of leafy tissues (only occa-
sionally of small twigs) [19, 41]. 𝑉𝑉𝑉𝑉𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 and LFMC can be
linked using in situ estimates of LAI and specific leaf area (SLA;
[m2leaf area/kgleaf dry biomass]) [116], as shown in Eq. (3):
𝑉𝑉𝑉𝑉𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 =𝜃𝜃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ⋅ 𝐿𝐿𝐿𝐿𝐿𝐿 =
(
𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿
100
) ⋅ 𝐿𝐿𝐿𝐿𝐿𝐿
𝑉𝑉𝐿𝐿𝐿𝐿
(3)
Since SLA and LAI are not contained in LFMC network
databases, we suggest adding them in future versions as
linking variables between the research fields of remote sens-
ing, fire risk assessment, and plant ecophysiology. Note that
(3) will be applied using individual LFMC measures, and
experimental SLA and LAI estimates, not variables at plot-
aggregated scales as in (1) and (2). Still, we acknowledge
the fact that LAI and SLA are variables that can be mea -
sured using different techniques in the field, such as direct
destructive methods or indirect optical methods (e.g., plant
canopy analyzer LAI-2000). Sampling issues and assump -
tions for the optical methods may be especially relevant
in short vegetation systems due to small leaves. To partially
1 3
Page 13 of 24 34
Current Forestry Reports (2026) 12:34
overcome these limitations in our approach, we suggest using the
standardized methods for ground LAI and SLA measurements
which are defined in the TRY database [34]. In these systems,
validation should also account for fractional cover, phenologi-
cal stage, exposed soil, and management practices such as irri-
gation or harvest, because these factors affect the relationship
between field samples and the remotely sensed pixel. Also, we
note that Eqs. (2) and (3) are the ones to be used in grasslands
and croplands.
B. Towards estimating 𝜽𝜽 from satellite remote sensing
Rearranging Eq. (1, 2 and 3) also provides potential frame-
works for estimating 𝜃𝜃 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 and 𝜃𝜃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 from space (Fig. 4).
For stems, L-band VOD-derived VWS ( 𝑉𝑉𝑉𝑉𝑉𝑉 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 ) has to
be used, although we acknowledge again the impossibility
of fully isolating the stems’ signal from that of leaf tissues
within satellite observations, and the simplification of bas -
ing VWS estimates on the b-parameter approach (Sect.
"Mapping VWS from space "). To estimate 𝜃𝜃 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 it has to
be isolated from (1):
𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 =
𝑉𝑉𝑉𝑉𝑉𝑉 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 ⋅ 𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑠𝑠𝑝𝑝
𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 ⋅ 𝜌𝜌𝑤𝑤𝑤𝑤𝑠𝑠𝑠𝑠𝑤𝑤
𝑉𝑉𝑉𝑉𝑉𝑉 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 ⋅ 𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑠𝑠𝑝𝑝 ⋅ 𝜌𝜌𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
𝜌𝜌𝑤𝑤𝑤𝑤𝑠𝑠𝑠𝑠𝑤𝑤 ⋅ 𝐵𝐵𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
(4)
This approach can use large-scale maps of: (i) wood density
( 𝜌𝜌 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 ) which is modeled globally [ 117], (ii) current [ 25]
or enhanced L-band VWS maps, and (iii) wood biomass
( 𝐵𝐵𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 ) which can be obtained from global above-ground
biomass maps (e.g. [ 118]), assuming that woody biomass
largely dominates over leafy biomass (Table 4). Since this
assumption is held in woody systems [ 125] and 𝜌𝜌 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
maps are only for woody vegetation [ 117], the 𝜃𝜃 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 esti-
mates will be here limited to woody plants as well.
𝜃𝜃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 can eventually be obtained from (2) and (3):
𝜃𝜃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑉𝑉𝑉𝑉𝑉𝑉𝐿𝐿𝐿𝐿𝐿𝐿 = 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑉𝑉𝐿𝐿𝐿𝐿 (5)
LAI maps are commonly available from a range of multispec-
tral satellite sources (e.g. [119]), and VWS for the top of the
canopy (CWS) can be derived from optical remote sensing
[30] or high-frequency VOD [27]. Also, LFMC (Table 2) and
satellite-based SLA maps [120] can be used (Table 4).
Finally, by using L-band VOD-based LFMC (𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 )
based on the approximation of [35], 𝜃𝜃 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 can also be approxi-
mated as in (6):
𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = (𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 /100) .𝜌𝜌𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝜌𝜌𝑤𝑤𝑤𝑤𝑠𝑠𝑠𝑠𝑤𝑤 (6)
Relationship of Microwave Sensing Depth to the
Water Status from Different Plant Heights and
Tissues
The Vertical Water Potential Gradient in the Forest Canopy
Going beyond the estimation of water content in the spatio-
temporal domain, it is paramount to consider the vertical dis-
tribution of water content and water potential in the canopy,
Table 4 Data needed to: (i) estimate θ as in Eq. (3) to (6), (ii) build PV curves, and (iii) build plot- and pixel-scale EPV curves that can be applied to estimate
Ψ from remote sensing (Eqs. (7), (8))
Data Plot scale approach Pixel scale analogy
θ estimation
(Eqs. 3 to 6)
LFMC Stems: likely from tower radiometry [43] or GNSS-
T (but not done yet).
Leaves: e.g., tower spectrometers
Stems: L-band-VOD-derived LFMC [35]
Leaves: products in Table 2
VWS N/A1 Stems: L-band VOD-derived VWS [25]
Leaves: [27] or [30]
ρwood Stems: In situ data Global map based on machine learning
and ~ 30,000 in situ measurements [117]
Bwood Stems: In situ data Global biomass maps (e.g., [118])
LAI In situ data or TLS Global LAI maps [119]
SLA In situ data Global database of leaf traits based on
remote sensing [120, 121]
Traits for PV curves
(Figure S7)
Ψthr In situ data, literature (e.g., [67]), and phylogeny
[122]
Literature (e.g., [67]) and phylogeny [122]
θ(Ψmax)
Ψmax Canopy height (see Table S1 in [68]) Canopy height maps [123, 124]
EPV curves Number or propor-
tions of species
Field knowledge on the number of trees for each
species
Pixel-weighted means of species propor-
tions based on NFIs
Stem volume In situ data or TLS Biomass / 𝜌𝜌 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
[117, 118]
Leaf area In situ data or TLS LAI · (pixel size)
Area Plot area Pixel grid size
1Using Eqs. (1) and (2) would introduce redundancy since θ is needed to estimate VWS and vice versa
1 3
34 Page 14 of 24
Current Forestry Reports (2026) 12:34
which is responsible for driving the water flux through
the soil-plant-atmosphere system. Before dawn, plants are
typically near hydraulic equilibrium. This means that roots,
xylem and leaves should reach similar water potential val -
ues due to overnight rehydration, when root uptake outpaces
transpiration [126]. In contrast, the maximum water demand
generally occurs in the middle of the day, when the maxi -
mum transpiration rate and water potential gradient between
the soil and the leaves are reached (Fig. 5b). Under drought,
the stomatal closure reduces transpiration, which progres -
sively diminishes the water potential gradient until reaching
a drought-induced equilibrium [ 68] and potentially leads
to increased plant mortality and canopy die-back [ 127].
Hence, the absence of a water potential gradient during the
day should by extension represent drought stress (Fig. 5a
and [68]).
The use of VOD from multiple frequencies may provide
insights into the relative magnitude of the gradient from
the upper canopy (high frequencies) to the entire vertical
profile (L-band) [ 24]. Under non-stressed conditions, we
expect a big difference to emerge in the ratios of different
VOD frequencies between predawn (minimal gradient) and
midday (significant gradient). In contrast, we expect the dif-
ference between predawn and midday ratios to be close to
zero under drought conditions. This is because a reduced
daily gradient would indicate the onset of drought (Fig. 5a).
One challenge of using the ratios of different VOD bands
is that, ideally, the retrievals at each band should have been
taken simultaneously. As depicted in Fig. 5b, satellite-based
L-band measurements are available at dawn (SMOS/SMAP,
~ 6 a.m.), while C- to Ku-band acquisitions are available
overnight (AMSR-E/2&3; ~1:30 a.m.). Interpreting those in
the context of hydraulic equilibrium requires the assump -
tion that vegetation has reached equilibrium with the soil
by ~ 1:30 a.m [129]. Midday overpasses are available from
some optical sensors (e.g., MODIS, ~ 1:30 p.m.) and from
C-/X-/Ku-band data as well (AMSR-E/2&3; ~1:30 p.m.).
Thus, comparing midday with night ratios between frequen-
cies is only currently feasible with the high-frequency mea-
surements of AMSR-E/2&3. As of 2029, the Copernicus
Imaging Microwave Radiometer satellite mission (CIMR)
is expected to provide pre-dawn multi-frequency over -
passes, ensuring collocated measurements at L- to Ku-bands
which should facilitate the analysis of hydraulic gradients
in VOD data due to coinciding, multi-frequency measure -
ments, allowing the reconstruction of the vertical profile
across the vegetation [130].
In contrast, L-band measurements (entire vegetation pro-
file) [ 54–56], are not currently available in the middle of
the day (Fig. 5b), thus limiting the assessment of gradients
in forest stems using currently available passive microwave
satellites. The launch of sub-daily missions has been theo -
retically proposed in the form of a multi-satellite L-band
synthetic aperture radar constellation [ 24, 131]. The future
realization of this or other lower frequency sub-daily mis -
sions [24] can improve our capacity to monitor vegetation
function and drought stress.
Sensitivity of Different Microwave Frequencies to Different
Plant Tissues
Importantly, while the vertical sensing depth of microwaves
through canopies is understood to vary with frequency
[54–56], the depth of plant tissue that is represented by the
signal is largely unknown (Fig. 5c and [68]; however, also
see [132] and [133] for exemplary studies). Plant stems and
branches comprise tissues arranged concentrically. First,
there is a thin layer of bark and phloem on the outer surface
(< 10 mm). The phloem transports carbohydrates from the
leaves to the stem and roots. Second, the xylem includes
sapwood conducting water upwards from roots to leaves.
Finally, the heartwood (‘inactive’ xylem) is in the center
of woody plants (Fig. 5c). Understanding the link between
microwave remote sensing and tissue-specific conditions,
combined with information on plant vertical water gradi -
ents, can bring invaluable capacity to assess vegetation con-
ditions and functions in ecophysiology.
L-band measurements have been found to correlate with
xylem water potential [ 43, 44], but little is known about
whether this and other frequencies are also sensitive to
processes in the phloem. Further research will have to be
conducted to determine the relationship between the micro-
waves’ dielectric constant and attenuation, and the flux of
water and sugars through the phloem. Such an analysis will
need to account for the impact of solute concentration in
dielectric models, which is only assumed as a constant and
empirical value in [ 134], and ignored in other models (e.g.
[135]). Note that the role of phloem and solute transpor -
tation may be essential in plant drought-induced mortality,
which can be driven by xylem cavitation but also by carbon
starvation [136]. Another gap in the current state-of-the-art
is whether microwave emissions and backscatter are sensi -
tive to the water stored in the heartwood. This is especially
the case of branches or stems thicker than 5 cm, a depth to
which L-band has been found sensitive [132] (Fig. 5c).
Similarly, it is likely that C-band VOD is sensitive to
the water in the sapwood and phloem in branches, how -
ever, the full extent of its sensitivity through the branch
thickness is not known and needs more research, while
Ku-VOD may only be representative of fine twigs or thin
branches (Fig. 5c). These different assumptions have direct
physiological implications because, to relate an amount of
1 3
Page 15 of 24 34
Current Forestry Reports (2026) 12:34
Fig. 5 Conceptual representation
of remote sensing of vegetation
moisture from passive micro-
waves. a Drought evolution,
which can be sensed by micro-
wave satellites, progressively
leads to increased water stress
and finally to drought-mediated
equilibrium due to large water
depletion [68], which involves
high mortality risk. b The data
availability of different satellite
microwave frequencies and their
sensing depth through the canopy
is represented in colored waves
(see the bottom legend), together
with the distribution of the water
vertical profile in forests at the
satellite time overpasses (in a
24-h clock, with sunrise, midday,
sunset and night represented).
Note the higher drought stress
at midday, when L-band data
is unavailable. At midday,
static atmospheric conditions
are assumed. c The drawing
represents microwave interac-
tions at different frequencies
with decreasing diameters. This
includes a trunk (left), a branch
(center) and a twig or an herba-
ceous stem (right; note the mm
scale). Microwave sensing depth
across the thickness of forest tis-
sues is known only in some cases
(~ 5 cm for L-band according to
[128]). It requires deeper research
to determine which frequencies
(and for which thicknesses) can
sense the physiologically active
fractions of vegetation stems (red
dashed waves). d Microwave
sensing depth for increasing bio-
mass and vegetation water con-
tent, and the limitations linked to
each vegetation type
1 3
34 Page 16 of 24
Current Forestry Reports (2026) 12:34
water to the current physiological state of the vegetation,
one needs to know which part of the system is sensed. As
discussed above, this is a complex issue because the behav-
ior of microwaves varies not only as a function of tissue
and observation frequency but also depends on the viewing
geometry (incidence and aspect) and polarization.
Need for Enhanced Dielectric and Retrieval Models
Likewise, different tissues and plant types may require the
application of specific dielectric models to retrieve their
water content from microwave remote sensing. However,
we have found that this is not the case in the literature. For
instance, the same dielectric model by Ulaby & El Rayes,
based on laboratory permittivity measurements of free water
and sucrose-based solutions [ 134], has been used to link
the dielectric constant with vegetation moisture either in a
wheat field [42] or in both the trunk’s xylem and the leaves
of an oak forest [ 43]. Furthermore [ 35], used two models
based on dielectric measurements of leaves [ 128, 135] in
comparison with Ulaby & El Rayes [134] and applied those
to different ecosystems and variations of leafy or woody tis-
sues to retrieve LFMC. Although all three models showed
a reasonable performance in estimating LFMC ( r > 0.5 at
X-band), they also showed significant differences in the
relationship between vegetation moisture and dielectric per-
mittivity (see the supplementary information of [35]). Thus,
the application of dielectric models to retrieve water in veg-
etation seems to be arbitrary so far. Consequently, there is a
need for improvement and implementation of tissue-specific
dielectric models, especially with a focus on assessing the
dielectric constant in woody tissues (e.g. [137]), and its rela-
tionship to the xylem and phloem conditions.
There are various limitations in monitoring water in veg-
etation with VOD that should be understood. First, VOD
retrieval models consider only single scattering effects in
forests, while multiple scattering is expected. Taking it into
account significantly increases VOD values [ 138]. Second,
in dense forests there is a lower retrieval robustness due to
noise in single-angle radiometer instruments like SMAP.
Solving this requires modifying the current regulariza -
tion approaches in VOD retrievals [ 139]. Third, the sur -
face canopy water from interception and dew may prevent
VOD from providing information about plant water stress
in tropical forests [ 140] (Fig. 5d). Finally, it should be fur -
ther explored whether soil emissions and roughness are cor-
rectly accounted for in VOD retrieval algorithms, especially
in short vegetation (Fig. 5d). All these may compromise
the quality of vegetation water estimates and need further
research to improve current retrieval models.
A Road Towards Monitoring Vegetation
Water Potential (ψ) from Remote Sensing
Linking Water Content to Water Potential: From In
situ Measurements to Remote Sensors
Water potential mechanistically relates to physiological
processes such as stomatal conductance (influencing rates
of photosynthesis [ 141]) and to hydraulic traits including
conductivity thresholds (influencing drought induced mor -
tality [136]). Therefore, Ψ is a central component of plant
hydraulic monitoring using models [142], and it is needed to
improve drought response predictions [ 143]. The relation -
ship between θ (or RWCphys) and Ψ, known as the pressure-
volume (PV) curve (Figs. 4 and S7), has been characterized
for leaves of many species and in woody tissues of some
species [67, 144]. In the region of the PV curve of minimal
water stress, the PV relationship tends to be almost linear,
after which there is an inflection point indicating a thresh -
old towards water stress (i.e., the turgor loss point in leaves
[67], or the cavitation threshold in the xylem [ 145, 146];
Figure S7).
While PV relationships are increasingly well understood
in isolated plant tissues, less is known about how Ψ and θ
or RWCphys vary across tissues and spatiotemporal scales
[68] and it is thus challenging to create an ecosystem-level
PV curve. It has been common practice to measure leaf
water potential at predawn and midday in field sites [ 127],
but less common to measure leaf θ or stem/branch Ψ or θ.
Moreover, the measurement of leaf Ψ tends to focus on
sampling extreme values (e.g., upper sunlit leaves), rather
than ecosystem average values [68]. Consequently, we have
limited understanding of the relationship between θ and Ψ
beyond tissue scales, and similarly for how Ψ varies across
systems [70]. Recently, a new global network of in situ Ψ
measurements (PSInet) has been published. It will facilitate
the potential to calibrate, compare and even validate future
remote sensing Ψ estimates [45].
Until now, most Ψ-related remote sensing research has
focused on studying correlations between optical indices
and plant Ψ, mainly at the local scale (Fig. 1). These studies
usually assess crops (see ‘Supplementary references’). Con-
cerning microwaves, the relationship between VOD and
Ψ has been explored in a forest stand using a tower-based
radiometer [44]. Also, the joint effects of Ψ, biomass, and
LAI on the VOD signal have been characterized in [ 147].
However, only recent research has provided estimates of
water potential based on a synergy between remote sens -
ing retrievals of plant water content and models of plant
water relations [43, 72]. In these cases, water potential was
1 3
Page 17 of 24 34
Current Forestry Reports (2026) 12:34
estimated with plot-scale radiometers in a wheat field [ 72]
and in a forest stand [ 43] following four steps. First, the
effect of the dielectric constant of vegetation (εveg) on VOD
was disentangled from the effect of temporal variations in
biomass. Previous studies have done this by either consider-
ing only sufficiently short periods when biomass variations
were low [43] or by using Sentinel-1 radar in a semi-phys -
ical model accounting for biomass variations [ 35]. Second,
dielectric models were applied to estimate the gravimetric
vegetation moisture (mg; Table 1) from εveg [35, 42]. Third,
mg was normalized to calculate RWC RS [%] (instead of
directly normalizing VOD as in Table 1). Fourth, Ψ was
estimated from RWCRS by using a sigmoidal function with
empirically determined parameters defined in [ 148] which
aims at approximating the behavior of PV curves. Other
examples assume linear relationships between VOD, RWC
and Ψ [44], which may exist within the normal physiologi -
cal range of plant function, but cannot estimate Ψ responses
beyond the turgor loss point [68] (Figure S7).
Bridging the Fields of Ecophysiology and Remote
Sensing to Retrieve Water Potential
Despite these recent efforts, further research is needed to
close the gap between remote sensing and ecophysiology.
First, we recall that RWC RS is by no means equivalent to
RWCphys (Sect. "Monitoring the water status of vegetation"),
which limits its applicability in pressure-volume relation -
ships. This limitation can be overcome by using θ and VWS
[68] (Sect. "Linking satellite and in situ water content infor-
mation") as the basis to estimate Ψ (Figure S7), instead of
using RWC.
A second issue is that the new remote sensing research
pursuing Ψ estimates has approximated the concept of PV
curves by using empirical sigmoidal or linear functions [43,
44, 72]. This can be enhanced by applying recently defined,
more realistic ecosystem-scale PV curves (EPV curves) to
upscale water relations to the plot level [ 24, 68, 69, 149].
Here we propose a theoretical framework applying EPV
curves for a remote sensing-based estimation of Ψ in plots
(e.g., GNSS-T sites of up to ~ 100 m in aerial diameter).
We also suggest a theoretical approach to Ψ estimates at the
scale of passive microwave pixels (0.25º). For these pur -
poses the required data are listed in Table 4, and three steps
should be followed:
- Obtain data on hydraulic traits: First, hydraulic traits
have to be measured at plot or site level or obtained from
literature or auxiliary information for the species present
in that site ([ 67, 68, 122], and Table 4). These traits are
the basis for building tissue- or tree-scale PV relation -
ships (Figure S7). At the pixel level, the same literature
and auxiliary sources could be used. In addition, knowl-
edge on which species are present inside the pixel and
their spatial distribution can be obtained from national
forest inventories (NFIs). - Define the EPV curves: Second, it is necessary to math -
ematically define the function linking volume and pres -
sure. To do so at the plot level one can refer to [ 68],
where the relationship between the plot-scale VWS and
Ψ is established based on the θ of trees and uses the wood
volume ( 𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 , Eq. (7)) and the LAI (Eq. (8)) as upscal-
ing variables:
𝑉𝑉𝑉𝑉𝑉𝑉 (Ψ)𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 =𝜌𝜌 𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 .
𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑡𝑡𝑡
∑
𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑡𝑡
𝑉𝑉𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 .𝜃𝜃 (Ψ)𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤 /𝐴𝐴𝑝𝑝𝑝𝑝𝑤𝑤𝑤𝑤 (7)
𝑉𝑉𝑉𝑉𝑉𝑉 (Ψ)𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 (Ψ)𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (8)
where 𝑉𝑉𝑉𝑉𝑉𝑉 (Ψ ) is the EPV plot-scale function and 𝜃𝜃 (Ψ )
is the PV curve of the wood or leaves (Figure S7). We note
that, in non-woody ecosystems, only Eq. (8) should be
applied.
At the remote sensing pixel (0.25º) scale, the upscaling
variables wood volume (i.e., based on biomass and wood
density) or LAI can be derived from datasets in Table 4. The
fact that (i) these datasets and the species presence maps
are available at < 1 km resolution and (ii) the traits infor -
mation is available at the species level, allows us to: (i)
approximate with a certain detail the spatial distribution of
traits within a 0.25º pixel, and (ii) build weighted means of
traits at the 0.25º scale using LAI, volume and species spa -
tial information. As a result, pixel-scale EPV curves could
be approximated also based on ( 7) and ( 8) considering the
pixel area instead of the plot area. Although this approach
is purely theoretical, data on species abundances have been
recently used to create pixel-weighted means of plant traits
[121], confirming the potential of this concept. While this
approach is species-based, trends in intraspecific variation
by region or climate should also be explored in the future
[89]. - Plot- or pixel-scale Ψ estimates: Finally, the EPV curves
defined in the previous step can be used to derive plot-
or pixel-scale Ψ estimates. This requires using remote
sensing plot- (GNSS-T) or pixel-scale (satellite-based)
estimates of VWS or θ (Sect. " Linking satellite and in
situ water content information" and Fig. 4).
Importantly, note that scaling hydraulic traits per ground area
is challenging, with uncertainty associated with each data
source (Table 4). This represents the forefront of ecophysi -
ological field research. However, it is conceptually similar
1 3
34 Page 18 of 24
Current Forestry Reports (2026) 12:34
to the big-leaf models used for decades to estimate fluxes in
soil-plant-atmosphere models [150], but with explicit repre-
sentation of the vegetation water storage component.
Conclusions and Recommendations
In this paper, we review the literature available on remote
sensing of water in vegetation across scales and variables,
including LFMC, water content (θ, VWS, RWC) and water
potential (Ψ). We find that, independently of the variable
studied, local-scale research largely dominates the litera -
ture, underlining that knowledge of large-scale properties
is still limited (Fig. 1a). We have identified several research
gaps and highlighted possible methods to improve large-
scale monitoring of water in vegetation, especially from
satellite technologies and their synergy with emerging net -
works of in situ data [19, 41, 45]. These networks have not
yet been used to their full potential to link remote sensing
with ground-based research and to validate or intercompare
remote sensing products for the assessment of vegetation
water status (Fig. 1b). These gaps can be summarized in
two research questions: (i) Which strategies are needed to
achieve and/or enhance large-scale products of water in veg-
etation? (ii) How can ground-based data be better integrated
to establish a stronger link with remote sensing estimates?
Table 5 addresses these questions by providing a list of rec-
ommendations derived from this paper.
Overall, our review highlights the need to further com -
bine in situ data and satellite remote sensing retrievals. This
joint improvement includes (i) accounting for more sophis-
ticated filters of in situ data for validation strategies of satel-
lite products, (ii) deriving water content information at plot
and pixel scales, (iii) validating satellite VWS estimates
using upscaled water content information from in situ and
GNSS-T-derived VWS [74], and (iv) reconciling the remote
sensing approach to estimate RWC RS with ground-based
ecophysiological variables (RWC phys). In addition, remote
sensing of water in vegetation needs to account for the vari-
ability of plant water status across the vertical gradients of
the plant. The gradients change throughout the day and in
drought conditions. To limit the effect of vertical profiles of
water status on variations in observations, multiple frequen-
cies to estimate VOD need to be exploited, emphasizing
the need for future passive microwave missions providing
data at sub-daily timescales to capture predawn and midday
plant water status in forests. The capacity of microwave fre-
quencies to sense different tissues depending on their sens -
ing depth is also discussed. Future research has to be carried
out to determine which microwave frequencies are best
suited to sense the xylem and the phloem, as a new step to
further understand forest responses to drought stress (Sect.
Table 5 Recommendations derived from this paper for each variable. Symbols are consistent with those in Fig. 1 and refer each one to a research
question: Which strategies are needed to achieve and/or enhance large-scale products of water in vegetation? How can ground-based data be better
integrated to establish a stronger link with remote sensing estimates?
Var. Symbol Recommendation
LFMC ◼ Different technīques of LFMC satellite products based on machine learning, RTM and VOD-LAI models capture
well time-dynamics in the western USA and in the Mediterranean (Table 2). In the western USA, consider using the
VOD-based semi-physical algorithm by [35] or the optics-radar synergy in [37] to capture absolute LFMC values,
sincethey have the lowest RMSE and bias in the region. In Australia, onecan apply [50] and [51] which are the best
performing products.
▲ For validation purposes of LFMC satellite products, consider thehomogeneity of species as well as if they are rep-
resentative of the dominant vegetation type in the pixel. Also, consider the variability in time of the in situ data. We
hypothesize that these filters can be applicable also to future satellite vegetation water products.
VWS ◼ Improve satellite VWS estimates by (i) enhancing estimates of the b-parameter, (ii) applying different VOD frequen-
cies to produce maps of VWS for different canopy fractions, and (iii) further exploring the synergy between signals
of opportunity (e.g., GNSS) with Al.
▲ To validate satellite VWS maps, use VWSstem and VWSleaf plot-scaleestimates upscaled from in situ 0 measure-
ments and TLS data ofwood volume and leaf area (see Eqs. (1) and (2)), or develop plot-scale VWS estimates based
on GNSS-T VOD data [74]. Vegetation in the benchmark plots should be representative of the vegetation in the pixel.
▲ Add field SLA and LAI data to in situ LFMC databases. It can be applied to derive Oleaf data, serving as a linking
variable between the research fields of remote sensing, fire risk assessment and plant ecophysiology.
θ ◼ Develop 0 datasets based on satellite VWS or LFMC, and on biomass, wood density and/or LAI and SLA maps (Eqs.
3 to 6 and Table 4).
◼ Plan satellite L-band missions with sub-daily overpasses (e.g., SLAINTE), to complete our capacity to assess vertical
gradients ofwater in forests.
Ψ ▲ Conduct laboratory and field experiments to determine the sensitivity of microwave frequencies to the water content
in different plant tissues, and for different thicknesses of branches and trunks. Develop more specific dielectric mod-
els tailored to, for instance, specific plant tissues or to species composition in study pixels.
◼ Further explore the theoretical framework proposed to develop plot-scale and satellite remote sensing estimates of Y
based on EPV curves, weighted means of plant traits and remote sensing estimates of water content (Section 4 and
Table 4). Validate the estimates with in situ PSInet data where available [45].
1 3
Page 19 of 24 34
Current Forestry Reports (2026) 12:34
"Relationship of microwave sensing depth to the water sta -
tus from different plant heights and tissues").
Finally, a new theoretical framework is proposed to esti-
mate Ψ from remote sensing at plot (e.g., from GNSS-T
VOD) and satellite scales (from satellite VOD). Remote
sensing-based Ψ estimates will require, as a first step, to
monitor VWS (Sect. " Linking satellite and in situ water
content information "). Afterwards, novel approximations
like the use of plot- or pixel-scale EPV curves should be
applied to derive Ψ from these VWS estimates. When con -
ducted at the satellite scale, this approach requires merging
new global maps of wood density [ 117] and biomass [118]
with forest inventories and ground-based ecophysiological
information on plant hydraulic traits to upscale pressure-
volume relationships (Table 4).
The proposed frameworks and recommendations can
improve understanding of LFMC products for fire risk
assessment and open new perspectives on plant ecophysiol-
ogy, extending it from current point-scale measurements to
potential large-scale estimates of key variables regularly in
time and space. This can represent a significant step towards
determining forest mortality thresholds, enhancing our
understanding of plant water relations and including infor -
mation on plant water status in Earth system mode .
Key References
Papers of particular interest, published recently, are high -
lighted here:
•• Of major importance
● [19] Yebra M, Scortechini G, Adeline K, Aktepe N, Al-
moustafa T, Bar-Massada A, et al. Globe-LFMC 2.0,
an enhanced and updated dataset for live fuel moisture
content research. Sci Data. 2024;11(1):332. This paper
details the updated first-time global dataset of in situ
LFMC samples. The dataset has been essential for cali -
bration of LFMC estimates and is used here for the com-
parison of LFMC datasets in Section 2.
● [45] Restrepo-Acevedo AM, Guo JS, Kannenberg
SA, Benson MC, Beverly D, Diaz R, et al. PSInet: A
new global water potential network. Tree Physiol.
2024;44(10):tpae110. This is the first global dataset of in
situ vegetation water potential measurements. It should
serve as the key calibration and validation tool for po -
tential remote-sensing-based estimates of Ψ.
● [68] Binks O, Meir P, Konings AG, Cernusak L, Christ-
offersen BO, Anderegg WR, et al. A theoretical frame -
work to quantify ecosystem pressure–volume relation -
ships. Glob Change Biol. 2024;30(11):e17567. This
paper proposes and applies a framework to develop EPV
curves at plot scales, which is essential for future large-
scale research of plant water relations and for potential
estimations of water content and water potential from
remote sensing sources.
• Of importance
● [35] Chaparro D, Jagdhuber T, Piles M, Jonard F,
Fluhrer A, Vall-Llossera M, et al. Vegetation mois -
ture estimation in the Western United States using ra -
diometer–radar–lidar synergy. Remote Sens Environ.
2024;303:113993. This paper presents the first LFMC
estimates for different microwave frequencies (i.e., for
different canopy layers). The study is conducted in the
western USA. Its main innovation is the fact that it iso -
lates the water component of VOD from the biomass/
structure component of this signal.
● [36] Forkel M, Schmidt L, Zotta RM, Dorigo W, Yebra
M. Estimating leaf moisture content at global scale from
passive microwave satellite observations of vegetation
optical depth. Hydrol Earth Syst Sci. 2023;27(1):39–68.
This paper presents the first microwave-based global
LFMC estimates. It implicitly separates water and bio -
mass effects in the VOD signal by combining VOD data
with LAI information.
● [37] Rao K, Williams AP, Flefil JF, Konings AG. SAR-
enhanced mapping of live fuel moisture content. Remote
Sens Environ. 2020;245:111797. This work presents an
innovative synergy between radar and optical satellite
remote sensing to provide enhanced high-resolution
LFMC estimates in the western USA.
● [43] Jagdhuber T, Schmidt AS, Fluhrer A, Chaparro D,
Jonard F, Piles M, et al. Estimation of forest water poten-
tial from ground-based L-band radiometry. IEEE J Sel Top
Appl Earth Obs Remote Sens. 2025;18:5509–22. This pa-
per estimates, for the first time, the water potential of a for-
est stand based on local-scale microwave remote sensing.
● [74] Feldman AF. Emerging methods to validate re -
motely sensed vegetation water content. Geophys Res
Lett. 2024;51(14):e2024GL110505. This paper propos -
es different ways (including in situ and GNSS-T data)
to validate satellite VOD products, bridging the gap be-
tween in situ, plot and satellite scales for VOD accuracy
assessment.
Supplementary Information The online version contains supplementary
material available at h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 4 0 7 2 5 - 0 2 6 - 0 0 2 9 0 - y .
Acknowledgements This research has been funded by the Junior Leader
fellowships from the “La Caixa” foundation LCF/BQ/PI25/12100008
(D. Chaparro) and LCF/BQ/PI23/11970013 (O. Binks), as well as
by the project H2020 FORGENIUS (Improving access to FORest
GENetic resources Information and services for end-USers) #862221
1 3
34 Page 20 of 24
Current Forestry Reports (2026) 12:34
and by the European Union’s Horizon Europe research and innovation
programme under the Marie Skłodowska-Curie grant agreement No
101204227. A. Feldman was supported by a NASA Ecohydrology Pro-
gram Grant (Grant Number 80NSSC25K0156). A.G. Konings was sup-
ported by the NASA SMAP Science Team (grant 80NSSC25K7055)
and by the Alfred P. Sloan Foundation. A.G. Konings acknowledges
funding from the Gordon and Betty Moore Foundation through the
Western Fire and Forest Resilience Collaborative under grant 13283.
We are thankful to the authors of large-scale LFMC products who
shared their datasets online. We also thank Dr. Lynn Miller (Monash
University) for helping us in the interpretation of LFMC maps and Drs.
Miquel de Cáceres and Víctor Flo (CREAF) for sharing their expertise.
We thank the reviewers and the editor of this manuscript for their time
and for providing insightful revision which improved the manuscript.
Authors' Contributions Conceptualization: D.C., O.B., T.J., M.Y .,
C.R.; Data curation: D.C., M.Y .; Formal analysis: D.C.; Funding ac-
quisition: D.C., O.B.; Methodology: D.C., T.J., O.B., M.Y .; Writing
- original draft: D.C., T.J., O.B., M.Y ., C.R., F.J. and Writing - review
& editing: all authors.
Funding Open Access funding provided by CERCA through the
CRUE-CSIC agreement with Springer Nature.
Data Availability The compared LFMC products are publicly available
following data availability statements in references of Table 2 and/or
upon request to the authors.
Declaration
Competing Interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adaptation,
distribution and reproduction in any medium or format, as long as you
give appropriate credit to the original author(s) and the source, provide
a link to the Creative Commons licence, and indicate if changes were
made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in
a credit line to the material. If material is not included in the article’s
Creative Commons licence and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this
licence, visit h t t p : / / c r e a t i v e c o m m o n s . o r g / l i c e n s e s / b y / 4 . 0 / .
References
- Vicente-Serrano SM, McVicar TR, Miralles DG, Yang Y , Tomas‐
Burguera M. Unraveling the influence of atmospheric evapora -
tive demand on drought and its response to climate change. Wiley
Interdiscip Rev Clim Change. 2020;11(2):e632. - Chiang F, Mazdiyasni O, AghaKouchak A. Evidence of anthropo-
genic impacts on global drought frequency, duration, and inten -
sity. Nat Commun. 2021;12(1):2754. - Balting DF, AghaKouchak A, Lohmann G, Ionita M. Northern
Hemisphere drought risk in a warming climate. NPJ Clim Atmos
Sci. 2021;4(1):61. - Senf C, Buras A, Zang CS, Rammig A, Seidl R. Excess forest
mortality is consistently linked to drought across Europe. Nat
Commun. 2020;11(1):6200. - Chaparro D, Vayreda J, Vall-Llossera M, Banqué M, Piles M,
Camps A, et al. The role of climatic anomalies and soil moisture
in the decline of drought-prone forests. IEEE J Sel Top Appl
Earth Obs Remote Sens. 2016;10(2):503–14. - Allen CD, Breshears DD, McDowell NG. On underestimation of
global vulnerability to tree mortality and forest die-off from hotter
drought in the Anthropocene. Ecosphere. 2015;6(8):1–55. - Hammond WM, Williams AP, Abatzoglou JT, Adams HD, Klein
T, López R, et al. Global field observations of tree die-off reveal
hotter-drought fingerprint for Earth’s forests. Nat Commun.
2022;13(1):1761. - Ruffault J, Limousin JM, Pimont F, Dupuy JL, De Cáceres M,
Cochard H, et al. Plant hydraulic modelling of leaf and canopy
fuel moisture content reveals increasing vulnerability of a Medi -
terranean forest to wildfires under extreme drought. New Phytol.
2023;237(4):1256–69. - Pausas JG, Keeley JE. Wildfires and global change. Front Ecol
Environ. 2021;19(7):387–95. - Jolly WM, Cochrane MA, Freeborn PH, Holden ZA, Brown TJ,
Williamson GJ, et al. Climate-induced variations in global wild -
fire danger from 1979 to 2013. Nat Commun. 2015;6(1):7537. - Miralles DG, Gentine P, Seneviratne SI, Teuling AJ. Land–
atmospheric feedbacks during droughts and heatwaves: state
of the science and current challenges. Ann N Y Acad Sci.
2019;1436(1):19–35. - Wright JS, Fu R, Worden JR, Chakraborty S, Clinton NE, Risi
C, et al. Rainforest-initiated wet season onset over the southern
Amazon. Proc Natl Acad Sci U S A. 2017;114(32):8481–6. - Ruiz-Vásquez M, O S, Arduini G, Boussetta S, Brenning A,
Bastos A, et al. Impact of updating vegetation information
on land surface model performance. J Geophys Res Atmos.
2023;128(21):e2023JD039076. - Grossiord C, Buckley TN, Cernusak LA, Novick KA, Poulter B,
Siegwolf RT, et al. Plant responses to rising vapor pressure defi -
cit. New Phytol. 2020;226(6):1550–66. - Gentine P, Green JK, Guérin M, Humphrey V , Seneviratne SI,
Zhang Y , et al. Coupling between the terrestrial carbon and water
cycles—a review. Environ Res Lett. 2019;14(8):083003. - Jonard F, Feldman AF, Short Gianotti DJ, Entekhabi D. Observed
water and light limitation across global ecosystems. Biogeosci -
ences. 2022;19(23):5575–90. - Yebra M, Dennison PE, Chuvieco E, Riaño D, Zylstra P, Hunt ER
Jr, et al. A global review of remote sensing of live fuel moisture
content for fire danger assessment: moving towards operational
products. Remote Sens Environ. 2013;136:455–68. - Konings AG, Rao K, Steele-Dunne SC. Macro to micro: micro -
wave remote sensing of plant water content for physiology and
ecology. New Phytol. 2019;223(3):1166–72. - Yebra M, Scortechini G, Adeline K, Aktepe N, Almoustafa
T, Bar-Massada A, et al. Globe-LFMC 2.0, an enhanced and
updated dataset for live fuel moisture content research. Sci Data.
2024;11(1):332. - Nolan RH, Foster B, Griebel A, Choat B, Medlyn BE, Yebra M,
et al. Drought-related leaf functional traits control spatial and
temporal dynamics of live fuel moisture content. Agric Meteorol.
2022;319:108941. - Griebel A, Boer MM, Blackman C, Choat B, Ellsworth DS, Mad-
den P, et al. Specific leaf area and vapour pressure deficit control
live fuel moisture content. Funct Ecol. 2023;37(3):719–31. - Flo V , Martínez-Vilalta J, Mencuccini M, Granda V , Anderegg
WR, Poyatos R. Climate and functional traits jointly mediate tree
water-use strategies. New Phytol. 2021;231(2):617–30. - Matheny AM, Bohrer G, Garrity SR, Morin TH, Howard CJ,
V ogel CS. Observations of stem water storage in trees of oppos-
ing hydraulic strategies. Ecosphere. 2015;6(9):1–13. - Konings AG, Saatchi SS, Frankenberg C, Keller M, Leshyk V ,
Anderegg WR, et al. Detecting forest response to droughts with
1 3
Page 21 of 24 34
Current Forestry Reports (2026) 12:34
global observations of vegetation water content. Glob Change
Biol. 2021;27(23):6005–24.
25. Chan S, Bindlish R, Hunt R, Jackson T, Kimball J. Vegetation
Water Content. Pasadena (CA): Jet Propulsion Laboratory, Cali -
fornia Institute of Technology; 2013.
26. Wang H, Wigneron JP, Ciais P, Yao Y , Fan L, Liu X, et al. Sea-
sonal variations in vegetation water content retrieved from micro-
wave remote sensing over Amazon intact forests. Remote Sens
Environ. 2023;285:113409.
27. Santi E, Paloscia S, Pampaloni P, Pettinato S, Nomaki T, Seki
M, et al. Vegetation water content retrieval by means of multi -
frequency microwave acquisitions from AMSR2. IEEE J Sel Top
Appl Earth Obs Remote Sens. 2017;10(9):3861–73.
28. Miao J, Wang J, Zhao D, Shen Z, Xiang H, Gao C, et al. Modeling
strategies and influencing factors in retrieving canopy equivalent
water thickness of mangrove forest with Sentinel-2 image. Ecol
Indic. 2024;158:111497.
29. McNairn H, Jackson TJ, Powers J, Bélair S, Berg A, Bullock P,
et al. SMAPVEX16 Database Report. Ottawa (ON): Agriculture
and Agri-Food Canada; 2017.
30. García-Haro FJ, Campos-Taberner M, Moreno Á, Tagesson HT,
Camacho F, Martínez B, et al. A global canopy water content
product from A VHRR/Metop. ISPRS J Photogramm Remote
Sens. 2020;162:77–93.
31. Liu L, Li S, Yang W, Wang X, Luo X, Ran P, et al. Forest canopy
water content monitoring using radiative transfer models and
machine learning. Forests. 2023;14(7):1418.
32. Rao K, Anderegg WR, Sala A, Martínez-Vilalta J, Konings AG.
Satellite-based vegetation optical depth as an indicator of drought-
driven tree mortality. Remote Sens Environ. 2019;227:125–36.
33. Hellwig FM, Fluhrer A, Dubois C, Schellenberg K, Chaparro D,
Zehner M… Jagdhuber, T. Tracking water dynamics of a temper-
ate forest under drought and non-drought conditions using active
and passive microwave and optical remote sensing. J Geophys
Research: Biogeosciences. 2026;131(5):e2025JG009240.
34. Kattge J, Bönisch G, Díaz S, Lavorel S, Prentice IC, Leadley P, et
al. TRY plant trait database – enhanced coverage and open access.
Glob Change Biol. 2020;26(1):119–88.
35. Chaparro D, Jagdhuber T, Piles M, Jonard F, Fluhrer A, Vall-
Llossera M, et al. Vegetation moisture estimation in the Western
United States using radiometer–radar–lidar synergy. Remote
Sens Environ. 2024;303:113993.
36. Forkel M, Schmidt L, Zotta RM, Dorigo W, Yebra M. Estimating
leaf moisture content at global scale from passive microwave sat-
ellite observations of vegetation optical depth. Hydrol Earth Syst
Sci. 2023;27(1):39–68.
37. Rao K, Williams AP, Flefil JF, Konings AG. SAR-enhanced
mapping of live fuel moisture content. Remote Sens Environ.
2020;245:111797.
38. Martin-StPaul N, Pimont F, Dupuy JL, Rigolot E, Ruffault J, Far-
geon H, et al. Live fuel moisture content (LFMC) time series for
multiple sites and species in the French Mediterranean area since
1996. Ann Sci. 2018;75:1–8.
39. Ivison K, Little K, Orpin A, Lewis CHM, Dyer N, Keyzor L, et al.
A national-scale sampled temperate fuel moisture database. Sci
Data. 2024;11(1):973.
40. United States Forest Service (USFS). Fire Environment Mapping
System (FEMS) Field Sample Database User Guide. Version 2.1.
2024.
41. Yebra M, Scortechini G, Badi A, Beget ME, Boer MM, Brad -
stock R, et al. Globe-LFMC, a global plant water status database
for vegetation ecophysiology and wildfire applications. Sci Data.
2019;6(1):1–8.
42. Meyer T, Jagdhuber T, Piles M, Fink A, Grant J, Vereecken H,
et al. Estimating gravimetric water content of a winter wheat
field from L-band vegetation optical depth. Remote Sens.
2019;11(20):2353.
43. Jagdhuber T, Schmidt AS, Fluhrer A, Chaparro D, Jonard F, Piles
M, et al. Estimation of forest water potential from ground-based
L-band radiometry. IEEE J Sel Top Appl Earth Obs Remote Sens.
2025;18:5509–22.
44. Holtzman NM, Anderegg LD, Kraatz S, Mavrovic A, Sonnentag
O, Pappas C, et al. L-band vegetation optical depth as an indica -
tor of plant water potential in a temperate deciduous forest stand.
Biogeosciences. 2021;18(2):739–53.
45. Restrepo-Acevedo AM, Guo JS, Kannenberg SA, Benson MC,
Beverly D, Diaz R, et al. PSInet: A new global water potential
network. Tree Physiol. 2024;44(10):tpae110.
46. Gente R, Born N, V oß N, Sannemann W, Léon J, Koch M, et
al. Determination of leaf water content from terahertz time-
domain spectroscopic data. J Infrared Millim Terahertz Waves.
2013;34:316–23.
47. Balaguer-Romano R, Díaz-Sierra R, De Cáceres M, Cunill-Cam-
prubí À, Nolan RH, Boer MM, et al. A semi-mechanistic model
for predicting daily variations in species-level live fuel moisture
content. Agric Meteorol. 2022;323:109022.
48. McCandless TC, Kosovic B, Petzke W. Enhancing wildfire
spread modelling by building a gridded fuel moisture con -
tent product with machine learning. Mach Learn Sci Technol.
2020;1(3):035010.
49. Miller L, Zhu L, Yebra M, Rüdiger C, Webb GI. Projecting live
fuel moisture content via deep learning. Int J Wildland Fire.
2023;32(5):709–27.
50. Quan X, Yebra M, Riaño D, He B, Lai G, Liu X. Global fuel
moisture content mapping from MODIS. Int J Appl Earth Obs
Geoinf. 2021;101:102354.
51. Yebra M, Quan X, Riaño D, Rozas Larraondo P, Van Dijk AIJM,
Cary GJ. A fuel moisture content and flammability monitoring
methodology for continental Australia based on optical remote
sensing. Remote Sens Environ. 2018;212:260–72.
52. Cunill-Camprubí À, González-Moreno P, Resco De Dios V . Live
fuel moisture content mapping in the Mediterranean basin using
random forests and combining MODIS spectral and thermal data.
Remote Sens. 2022;14(13):3162.
53. Knipling EB. Physical and physiological basis for the reflectance
of visible and near-infrared radiation from vegetation. Remote
Sens Environ. 1970;1(3):155–69.
54. Chaparro D, Duveiller G, Piles M, Cescatti A, Vall-Llossera M,
Camps A, et al. Sensitivity of L-band vegetation optical depth to
carbon stocks in tropical forests: a comparison to higher frequen-
cies and optical indices. Remote Sens Environ. 2019;232:111303.
55. Frappart F, Wigneron JP, Li X, Liu X, Al-Yaari A, Fan L, et
al. Global monitoring of the vegetation dynamics from the
Vegetation Optical Depth (VOD): A review. Remote Sens.
2020;12(18):2915.
56. Olivares-Cabello C, Chaparro D, Vall-llossera M, Camps A,
López-Martínez C. Global unsupervised assessment of multifre -
quency vegetation optical depth sensitivity to vegetation cover.
IEEE J Sel Top Appl Earth Obs Remote Sens. 2022;16:538–52.
57. Kesselring J, Humphrey V , Niederberger M, Feigenwinter I,
Shekhar A, Buchmann N, et al. Divergence between in situ and
satellite-based estimates of forest canopy water content. Remote
Sens Environ. 2026;332:115097.
58. Baur MJ, Jagdhuber T, Feldman AF, Akbar R,
论文原文与出版信息
Current Forestry Reports