Applications of Advanced Technologies in the Development of Urban Flood Models

Abstract

Over the past 10 years, urban floods have increased in frequency because of extreme rainfall events and urbanization development. To reduce the losses caused by floods, various urban flood models have been developed to realize urban flood early warning. Using CiteSpace software’s co-citation analysis, this paper reviews the characteristics of different types of urban flood models and summarizes state-of-the-art technologies for flood model development. Artificial intelligence (AI) technology provides an innovative approach to the construction of data-driven models; nevertheless, developing an AI model coupled with flooding processes represents a worthwhile challenge. Big data (such as remote sensing, crowdsourcing geographic, and Internet of Things data), as well as spatial data management and analysis methods, provide critical data and data processing support for model construction, evaluation, and application. The further development of these models and technologies is expected to improve the accuracy and efficiency of urban flood simulations and provide support for the construction of a multi-scale distributed smart flood simulation system.

Full text

Water 2023, 15, 622. https://doi.org/10.3390/w15040622 www.mdpi.com/journal/water
Review
Applications of Advanced Technologies in the Development of
Urban Flood Models
Yuna Yan
1
, Na Zhang
1,2,
* and Han Zhang
1
1
College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China
2
Beijing Yanshan Earth Critical Zone National Research Station, University of Chinese Academy of Sciences,
Beijing 101408, China
* Correspondence: zhangna@ucas.ac.cn
Abstract: Over the past 10 years, urban floods have increased in frequency because of extreme rain-
fall events and urbanization development. To reduce the losses caused by floods, various urban
flood models have been developed to realize urban flood early warning. Using CiteSpace software’s
co-citation analysis, this paper reviews the characteristics of different types of urban flood models
and summarizes state-of-the-art technologies for flood model development. Artificial intelligence
(AI) technology provides an innovative approach to the construction of data-driven models; never-
theless, developing an AI model coupled with flooding processes represents a worthwhile chal-
lenge. Big data (such as remote sensing, crowdsourcing geographic, and Internet of Things data), as
well as spatial data management and analysis methods, provide critical data and data processing
support for model construction, evaluation, and application. The further development of these mod-
els and technologies is expected to improve the accuracy and efficiency of urban flood simulations
and provide support for the construction of a multi-scale distributed smart flood simulation system.
Keywords: artificial intelligence technology; remote sensing; crowdsourcing geographic data; Inter-
net of Things; spatial data management and analysis; distributed smart flood simulation system
1. Introduction
In recent years, frequent extreme rainfall events have caused severe flooding world-
wide. In addition, the rapid increase in impervious areas associated with urbanization has
aggravated urban floods. For example, the peak flow of urban floods is several times or
even dozens of times higher than that of natural rivers [1]. Frequent urban floods have
seriously affected the normal functioning of cities. During and after rainstorms, rapidly
accumulated floods cannot be discharged within a short time, which may have a signifi-
cant impact on traffic, human lives, and even cause casualties. Given that urban areas
have highly dense populations, numerous buildings, developed industries, and com-
merce activities, the impacts of floods are much more severe in urban areas than in natural
areas. According to the Emergency Events Database, flooding is responsible for over one-
third and more than half of global emergency event-related economic losses and fatalities,
respectively [2]. For example, a flood in Queensland in September 2010 affected more than
200,000 people and led to a loss of $2.38 billion [3]. A total of 641 out of 654 cities have
been exposed to urban floods in China [4]. The ‘7.21′ flash flood of Beijing in 2012 affected
nearly 1.9 million people, caused a disaster area of 14,000 km
2
, and produced losses of
approximately $1.6 billion [5]. The extraordinary rainstorms and floods in Henan Prov-
ince, China that took place on July 16, 2021 affected 150 counties, 1616 towns, and 1.47
million people and caused direct losses of $17.96 billion [6]. Moreover, a large amount of
pollutants is washed off during runoff periods and enters rivers and groundwater, which
consequently degrades the water environment. To reduce disaster consequences while
guaranteeing the normal functioning of cities, flooding prevention and mitigation
Citation: Yan, Y.; Zhang, N.;
Zhang, H. Applications of Advanced
Technologies in the Development of
Urban Flood Models. Water 2023, 15,
622.
https://doi.org/10.3390/w15040622
Academic Editors: Jorge Leandro
and Mingfu Guan
Received: 18 December 2022
Revised: 1 February 2023
Accepted: 2 February 2023
Published: 5 February 2023
Copyright: © 2023 by the authors. Li-
censee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and con-
ditions of the Creative Commons At-
tribution (CC BY) license (https://cre-
ativecommons.org/licenses/by/4.0/).

Water 2023, 15, 622 2 of 24
measures must be formulated, and the timely warning of urban floods is an indispensable
basis. Spatial assessments of infrastructure flood exposure are a necessity for rational
flood risk assessments, targeted mitigation measure implementation, and for ensuring lo-
cal community preparedness [7]. To that end, large-scale assessments of flood-exposed
areas and their spatial inequalities have attracted research attention [8].
Urban flood simulation is a generally used and valid approach for timely flood warn-
ing and spatial flooding assessments [1,9], and associated models are constantly evolving.
The widely used models include the storm water management model (SWMM) [10–12],
the MIKE FLOOD model [13], the InfoWorks ICM (integrated catchment management)
model [14,15], and the Hydrologic Engineering Center’s hydrologic modeling system
(HEC-HMS) and river analysis system (HEC-RAS) [16]. These models can simulate sur-
face runoff and overflow processes, and the results can be used to derive rainstorm flood
information. However, more accurate simulations are required to realize more reliable
flood warning, including the refinement of inputs with higher temporal and spatial reso-
lution; the collection, processing, management, and analysis of a large amount of multi-
source spatially heterogeneous data; the finer demarcation of sub-catchments or land
cover patches; the calibration and automatic adjustment of a variety of parameters; and
more detailed descriptions of hydrological and hydrodynamic processes. Moreover, the
actual flooding process is difficult to observe during severe rainstorms, and measured
flooding data are not easily obtained from government departments or companies, which
are issues that hinder the validation of modeled results.
The accuracy of flood model data for inputs and validation can be improved by ap-
plying advanced technologies, such as remote sensing [17], crowdsourcing geographic
data, and Internet of Things (IoT) technologies [18]. However, a higher simulation accu-
racy generally implies a lower computational efficiency. Hence, models involving detailed
processes at a fine spatiotemporal grain level are not suitable for urgent flood warning
systems. Data-driven models based on artificial intelligence (AI) technology or models
fused with AI can address this contradiction by ignoring or attenuating the interior mech-
anisms of hydrological and hydrodynamic processes and only if there is sufficient high-
quality data. Geographic information systems (GIS) are integrated platforms that play an
important role in data management, analysis, and simulation. In recent decades, GIS use
has been well established, and Earth observation data have become increasingly available,
which has increased the feasibility of large-scale evaluations at fine resolutions. Ap-
proaches and methods of spatial analysis are helpful for extracting multi-scale spatial in-
formation on input and output data and for linking multi-scale processes. In summary,
these advanced technologies can be used at each level of model development, including
constructing multi-scale flood models or fusing different models; obtaining, selecting,
processing, managing, and analyzing model input variables and parameters; validating
modeled flooding processes; and presenting spatial patterns of modeled flooding status.
This review first summarizes the characteristics of different types of main urban
flood models. Approximately ten studies and review articles have already performed
comparisons of different urban flood models [19–21]; therefore, this study focuses on the
recent applications of advanced technologies in urban flood simulation and discusses in-
sights into promising opportunities for future research.
2. Methods
The process of collecting and analyzing the literature consists of three steps. First, we
selected “urban flood” and “model” as the key strings for the topic search. The literature
published during 1980–2022 (up to March 9, 2022) was searched in the Web of Science
(WOS). All publications containing either of these key strings in their abstracts, titles, or
keywords were retrieved from the bibliography. We assumed that the publications in the
WOS represented influential works capable of demonstrating the trends in the research
field; however, the retrieved works did not cover all related studies, such as those in non-
English languages. Second, using the co-citation analysis tool in CiteSpace software, 120

Water 2023, 15, 622 3 of 24
important publications with higher co-citation intensity were selected from the biblio-
graphic records. Moreover, some recently published articles were also reviewed. Finally,
from these important publications, we extracted information related to urban flood mod-
els and advanced technology applications.
3. Descriptions of the Main Urban Flood Models
Urban flood models have been developed to understand various hydrological or hy-
drodynamic processes of floods in urban areas and to obtain flooding conditions, such as
inundation area, flood depth, and volume.
The main simulated processes include surface runoff, surface convergence, pipe con-
vergence, and flood inundation (Figure 1) [22,23]. Surface runoff is determined based on
the amount of rain falling to the surface and the subtraction of the amount of evaporation
from rain intercepted by the canopy, absorbed through infiltration, and detained in de-
pressions [24]. The most commonly used methods for simulating surface runoff include
the full storage runoff method, the infiltration curve subtraction method, the index deduc-
tion method, and the runoff coefficient method [25]. Surface convergence is the process of
surface flow into drainage networks and rivers, and it is usually simulated using hydro-
logical and hydrodynamic methods [26,27]. Hydrological methods include isochronal, lin-
ear, and nonlinear reservoir algorithms. Hydrodynamic methods are based on micro-
scopic physical laws by numerically solving continuity and momentum equations. The
most widely used method is the 2D shallow water equation. Pipe convergence refers to
the confluence process of surface runoff after it enters rainwater pipe networks, and it is
commonly simulated using the kinematic wave, diffusion wave, dynamic wave, instanta-
neous unit hydrograph, and Muskingum methods [28]. The flood inundation process de-
scribes the overflow of water in rivers or pipelines, and is mainly simulated using the 2D
hydrodynamic method, the terrain submergence method, and the cellular automata
method [29].
Figure 1. Urban flooding simulation processes (adapted from [30]). The simulations involve hydro-
logical processes, from precipitation, to surface runoff and hydrodynamic processes, that generate
flood inundation.
Distinct urban flood models involve different hydrological and hydrodynamic pro-
cesses. Based on previous classifications [20,21,31], we summarized these models as sim-
plified flood models, physical flood models, and data-driven models according to the
complexity of the model structure, the described processes, and the amounts of required
data (Figure 2). Simplified flood models require less input data, owing to their simple
structure and relatively few model parameters. Physical flood models have the most com-
plex structures and largest number of model parameters, and they require more data for
parameter calibration and model validation. Data-driven models do not have complex
structures but require massive data on the characteristics of flood events.

Water 2023, 15, 622 4 of 24
Figure 2. Classification and characteristics of the main urban flood models
based on the complexity
of the model structure and modeled processes and the amount of required data.
3.1. Simplified Flood Models
Simplified flood models can be regarded as gray-box models that do not clearly de-
scribe the internal laws of urban flooding. They have been widely used when the physical
mechanisms of floods are unclear [32] and simulate flood processes by solving simplified
equations (e.g., shallow water equations) that omit one or two acceleration terms [33]. The
final or maximum flood range and related depth are obtained by solving the water balance
equation.
These models have obvious advantages, such as reduced computational time. Jamali
et al. developed a simplified model and showed that it could reduce the simulation time
by two orders of magnitude when compared with MIKE FLOOD, although differences
were observed in the simulated flood depth [33]. When the required simulation accuracy
is not high or the situation is urgent, a simplified model can provide a stop-gap approach
for rapid flood prediction.
3.2. Physical Flood Models
Compared to simplified flood models, physical flood models can be regarded as
white-box models that describe the internal mechanisms of flood generation and are more
suitable for the refined simulations of complex urban underlying surfaces.
Typical physical flood models include SWMM [10,34], MIKE FLOOD [35,36], In-
foWorks ICM [15,37], and HEC-HMS/RAS [16,38]. Urban hydrological models have also
been developed based on cellular automata, which we call UHCA models [39–41] (Figure
2). Most physical flood models use the catchment as the basic hydrological unit, whereas
UHCA models use grid cells as units to achieve the fine division of underlying surfaces.
Except for the SWMM, most physical flood models can simulate the flow of floods on a
2D surface. Moreover, some models include a pipe network module, some include a GIS
module to achieve efficient processing of the driving variables and parameters, and some
include a low-impact development (LID) module to describe local eco-hydrological pro-
cesses in detail. The SWMM takes less time owing to more simplifications, and it may
have lower simulation accuracy. By contrast, InfoWorks ICM and MIKE FLOOD can ob-
tain more accurate surface flood distributions, although their calculation speeds are rela-
tively low. The calculation speed and accuracy of the HEC-HMS/RAS model are relatively
low [38]. UHCA models have the advantages of the above models and can simulate a

Water 2023, 15, 622 5 of 24
flood event at the second to 10 min level [41]. They also present high accuracy based on
the use of high-precision data at the grid cell scale and high efficiency based on the use of
parallel computing rules. Hence, UHCA models can support large-scale and high-preci-
sion spatial calculations in urban areas (Table 1).
Table 1. Comparisons of the main physical flood models.
Model SWMM *
InfoWorks ICM
* MIKE FLOOD
HEC-
HMS/RAS
*
UHCA *
Basic unit catchment catchment catchment catchment grid cell
Dimension 1D 2D 2D 2D 2D
GIS module ×
√
√
× ×
Pipe module √ × √ × √ or ×
LID * module √ × imperfect × √ or ×
Calculation speed ***** *** ** * ****
Accuracy ** **** **** ** *****
Open source √ × × × √ or ×
Note(s): * SWMM, storm water management model; InfoWorks ICM, InfoWorks integrated catch-
ment management model; HEC-HMS/RAS, Hydrologic Engineering Center’s hydrologic modeling
system and river analysis system; UHCA, urban hydrological models based on cellular automata;
LID, low impact development. The larger number of the character of “*” means the higher
calculation speed or the higher accuracy.
Developing a hybrid model by utilizing the distinct advantages of different models
or modules is a feasible and efficient approach. For example, we developed and validated
a cellular automata-based distributed hydrological model for urban surface runoff (CA-
DUSRM) [42], and then the simulated surface runoff was input into the pipe convergence
module in SWMM for pipe overflow. Ultimately, the simulated overflow was input into
a developed flood inundation module for the occurrence and expansion of inundation.
This hybrid model has been validated in a small area and a large area using the measured
and crowdsourced data.
These physical flood models describe the complex hydrological and hydrodynamic
processes of floods in cities through physical laws formulated by differential equations,
such as the Saint Venant, shallow water, and Boltzmann equations [43]. To simulate the
temporal dynamics and spatial patterns of runoff and flood processes, these models gen-
erally require various data, such as topography, soil types, sewer conveyance networks,
infiltration conditions, flow curves, and runoff parameters [43]. If these required data are
difficult to obtain, it becomes impossible to use a physical model, especially for urban
catchments with complex underlying surface conditions, boundary conditions, and intri-
cate drainage systems. In addition, iterative calculations in simulations of urban runoff
and flood processes require substantial time, and parallel computation in UHCA models
requires adequate preparation for massive computing performance, which contradicts the
short duration of urban flooding. All of these factors have greatly restricted the timely
application of these models.
3.3. Data-Driven Flood Models
Data-driven flood models can be regarded as black box models without presenting
the internal mechanism of flooding. These models have the most straightforward ap-
proach to forecasting the occurrence and magnitude of floods, which are based on histor-
ical data [20,21,31,44–49]. The black box nature of AI technology makes it a core algorithm
for data-driven models.
Data-driven models exhibit good self-learning and continuous evolution capabilities
[50]. In particular, they do not require an understanding of internal specific mechanisms

Water 2023, 15, 622 6 of 24
and can avoid describing complicated hydrological and hydrodynamic processes. Data-
driven flood models can automatically better extract flood features than simplified flood
models. Moreover, they have a higher calculation efficiency and much lower require-
ments for flooding parameters than physical flood models. Therefore, data-driven flood
models may provide rapid forecasts for urgent urban floods. However, these models re-
quire considerable historical data to train the simulation process and validate the simula-
tion results.
To realize accurate prediction and control over an entire study area, flood models
should contain more spatial elements; thus, spatially explicit models may be necessary
[51]. Even if the model is not spatially explicit or developed for local sites or single sub-
catchments, the spatial heterogeneous nature of flooding conditions and associated driv-
ing variables must be characterized. Obviously, both the development of spatial models
and the identification of spatial patterns can benefit from emerging technological ad-
vances.
4. Development of Urban Flood Models Using Advanced Technologies
The development of urban flood models requires a consideration of model construc-
tion, evaluation, and application. In addition to the model structure and algorithms, data
issues penetrate throughout the process of model development.
4.1. Model Construction
AI algorithms have been utilized in urban flood fields to develop data-driven models
that can provide timely prediction of the imminent incidences of flash floods [52]. Many
AI systems rely on machine learning methods, and deep learning is a new method for
machine learning (Figure 3). The deep learning method analyzes and learns by establish-
ing and simulating the neural network of the human brain. Compared to traditional ma-
chine learning methods, in which features are first manually extracted from data, deep
learning methods can learn these features while training a large amount of data [53].
Figure 3. Categories of artificial intelligence algorithms that can be applied in urban flood simula-
tions.
Common AI models involve three main layers: input, hidden, and output layers (Fig-
ure 4). The output layer contains data on flood variables, such as inundation area [54],
flood volume [9], and flood depth. The input layer contains data on feature factors that

Water 2023, 15, 622 7 of 24
may impact flood variables, such as precipitation [9], topography [55], normalized differ-
ence vegetation index (NDVI) [50], soil moisture, land use type (or imperviousness) [56],
and urbanization level, which can be identified by correlation analysis. Although various
input data can be provided to train a model, a systematic investigation should be consid-
ered to determine which inputs are optimal to predict flood variables [49]. AI algorithms
and network structures are trained in one or multiple hidden layers to determine the ap-
propriate weights that describe the relationships between the input and output variables.
Flood data that are independent of those used for model training are used to validate the
simulation results. After the model is validated to be reliable and accurate, model evalu-
ations can be conducted, including sensitivity analysis of the model parameters, uncer-
tainty analysis of the simulation results, and comparisons with other models. Further-
more, the validated models can be used to provide flood warnings, flood risk assessments,
and advice for landscape planning.
Figure 4. General flow chart for constructing an artificial intelligence model.
Over the last decade, a tremendous increase in the use of AI technologies has been
observed for urban hydrological simulation, which has provided early warnings and real-
time forecasts to decrease flood risks and losses [57]. AI algorithms have been used at
single sites (1D) and achieved better results than traditional physical models. For example,
Ke et al. used a rainfall threshold to differentiate flood events from non-flood events based
on a support vector machine (SVM), which improved the accuracy of flood identification
to 96.5% when applied to Shenzhen, China [58]. Tian et al. used a recurrent neural network
for 1D flood runoff prediction, and the Nash coefficient was greater than 0.8 [59]. Further-
more, the prediction of a 2D flood inundation area and flood depth is more difficult, be-
cause there are many more feature factors and more complex nonlinear relationships be-
tween the input and output variables. To this end, exploratory studies have been con-
ducted. For example, Berkhahn et al. implemented an ensemble training approach and
network growing algorithms for an artificial neural network (ANN) for 2D distributed
maximum water level prediction in urban areas [47]. Chang et al. used a convolutional
neural network to predict potential rainfall events from hyetograph feature parameters
and to further estimate flood depth [52]. Hou et al. established a model that can rapidly
estimate urban flood inundation based on machine learning algorithms (random forest

Water 2023, 15, 622 8 of 24
and K-nearest neighbor) and showed that these algorithms can effectively improve the
estimation accuracy; for example, the mean relative errors between the modeled and
measured inundation area and flood depth values were less than 5%, and the mean rela-
tive errors between the modeled and measured flood volume values could be controlled
to within 10% [60]. Moreover, AI-based systems have been developed. Ye et al. suggested
an AI-driven platform for urban flood prevention and warning [61], and Goyal et al. cre-
ated a system for real-time post-flood management using an ANN to analyze the areas
affected by floods, and they found that this system was better than previous flood control
systems [62].
The application of AI in urban flood simulations has greatly improved the efficiency
and intelligence of estimations, as well as provided important support for the develop-
ment of smart water cycle systems. AI algorithms can numerically formulate the nonline-
arity of urban flood dynamics based solely on data. In addition, they can avoid iterative
calculations of complex physical equations. Therefore, data-driven AI algorithms are
promising tools for rapidly training and validating models.
However, the advantage of AI models also implies a disadvantage associated with a
limitation. We cannot generate an explicit understanding of the underlying physical pro-
cesses from either a data-driven AI model or its simulation results, which may lead to
doubts in the model reliability, even if the simulation results have higher accuracy. More-
over, conflicts may occur between the simulation results and known physical principles
[63]. To overcome this limitation, new AI theories and methods that couple data with a
priori knowledge represent promising approaches [64]. For example, after the hydrologi-
cal-related values modeled by physical models under various rainfall scenarios are vali-
dated, they can be used as training and test samples for an AI model. Bermúdez et al. used
a hydrological model to generate discharge, which was input for SVM models with ap-
proximately 25,000 control points in an urban catchment for the output of the maximum
flood depth and velocity [65]. Lin et al. generated a synthetic event database using the
hydrological model LARSIM (large area runoff simulation model) [48] for discharge hy-
drographs and the HEC-RAS 2D hydrodynamic model for flood inundation map, and,
then, these databases were used for training and validating an ANN. This is considered a
loose coupling approach.
When the model is used over a large spatial extent, close coupling between a spatially
distributed flood model and an AI algorithm becomes necessary. This type of coupling
model has higher requirements for the training algorithms and network structures be-
cause of the complex nonlinear relationships between the spatial input and output varia-
bles that may exist at distinct scales. The structure of an AI flood model (such as the net-
work topology and convolution core) can be adjusted according to the generation and
evolution of spatially distributed flooding processes [64]. Moreover, prior knowledge can
be used to constrain the range of model outputs to enhance the authenticity of simulation
results. At present, the development of AI models driven by data that are guided and
constrained by prior knowledge is still in the stage of theoretical exploration [66]. Alt-
hough few related examples are available in the field of flooding, this idea has been ap-
plied in geosciences. For example, Chen and Zhang proposed a mechanism-mimicking AI
network structure designed according to geomechanical equations [67].
In addition to AI algorithms, this type of coupling model has the requirements for
the physical mechanisms that describe hydrological and hydrodynamics processes. There
are certain uncertainties in the physical mechanisms. For example, the Saint Venant,
shallow water, and Boltzmann equations generally used in physical models are more
suitable for continuous processes on free surfaces [21].For noncontinuous processes on a
surface with large undulations or complex and heterogeneous land covers (typically in
cities), there might be big errors. Although UHCA models can address this problem, our
study showed that there is much room for the improvement of rules that describe
intercellular water exchanges in multiple flow directions [42]. The incomplete laws and

Water 2023, 15, 622 9 of 24
imperfect algorithms in numerical simulation and deep learning increase the uncertainties
of simulation results.
4.2. Data for Model Construction and Application
The construction of both pure data-driven models and data-driven models coupled
with mechanisms require massive amounts of data for model training, and it also requires
a large amount of data for model inputs and validation when a constructed model is ap-
plied over a large space. Therefore, collecting, processing, and identifying high-quality
data have become major issues in both model construction, evaluation, and application.
In the era of big data, the amount of data related to urban flooding has sharply in-
creased with the development of information collection technology. For instance, a large
amount of meteorological data (4–5 PB) have been collected, with an annual increment of
1 PB, and frequent exchanges occur among countries worldwide [68]. Representative big
data include crowdsourcing geographic data and IoT data, in addition to remote sensing
and traditional site data (Figure 5). Various data have been used as inputs to provide me-
teorological data and flood-related data for urban flood simulation during both model
training and application, and some can provide information for the calibration of param-
eters and the validation of modeled results. Model-predicted flood information (e.g., in-
undation area and flood depth) is generally communicated to mobile instruments or web
users. Big data technology essentially unifies data from different sources and structures
to provide support for urban flood simulations.
Figure 5. Big data technology involved in the construction, parameter calibration, validation, and
application of urban flood models.
4.2.1. Remote Sensing Data
Remote sensing technology has become increasingly significant in the field of flood
simulation, especially with the continuous emergence of multi-platform, multi-temporal,
and high-spatial-resolution remote sensing products. This technology can be used to de-
tect broad-scale land surface data at a low cost and high efficiency, which may compensate
for the shortcomings of traditional in situ measurements. Moreover, it can also provide

Water 2023, 15, 622 10 of 24
information on underlying surfaces, meteorological factors, and water flows to develop
and implement flood laws, although such technology is currently mainly applied in nat-
ural watersheds.
1. Underlying Surface Information
In urban flood models, underlying surface information with an effect on floods
mainly includes topographic and land surface characteristics (e.g., buildings and land
cover) [69].
The depth and speed of floods are easily affected by urban micro-topographical char-
acteristics. Thus, small topographic errors may lead to significantly distorted simulation
results. The complex variation in microtopography in urban areas may involve various
combinations of altitude differences, slopes, undulations, and overland flow widths,
which can generate nonlinear water flow processes [70]. For example, the altitude differ-
ence between two sites directly affects the water flow direction, which, thus, implies the
expected position of the outlet within a sub-catchment and determines the discretization
of sub-catchments. The combination of terrain undulation (concaveness/convexity of ter-
rain) and slope affects the water flow speed and, thus, the water storage within a sub-
catchment, which ultimately has an effect on runoff, convergence, and flooding [42].
Therefore, obtaining terrain data with a high spatial resolution is critical for revealing the
flood flow mechanisms influenced by the terrain.
Currently, commonly used digital elevation model (DEM) data sources include the
shuttle radar topography mission (SRTM, 30–90 m resolution), the advanced space-borne
thermal emission and reflection radiometer (ASTER, 30 m resolution), and Google Earth
(from multi-source data fusion and interpolation at approximately 7 m resolution) [71]. In
addition, the latest remote sensing technologies, such as synthetic aperture radar interfer-
ometry (InSAR) and light detection and ranging (LiDAR), have further improved the ac-
curacy and speed of terrain data acquisition. InSAR and LiDAR can provide terrain data
with meter-level resolution over a large range. Fine LiDAR data are used to identify highly
heterogeneous microtopography and building contours, which can facilitate the rapid de-
velopment of high-spatial-resolution flood simulations [72,73]. However, this high-reso-
lution data may not be necessary for a sub-catchment with small terrain undulations. Our
studies showed that the modeled surface runoff processes for a sub-catchment without
apparent depressions did not change significantly when the spatial grain size increased
from 1 m to approximately half the sub-catchment area [42].
Land cover is another indispensable input data type when modeling runoff and
flooding processes. Pervious and impervious covers have significantly different effects on
overland flows, because they have different capacities for water retention, storage, ab-
sorption, and infiltration, which are determined by their physical characteristics (e.g.,
roughness, depression storage, and permeability), as well as biological and chemical at-
tributes (e.g., plant utilization and soil absorption) [74]. Thus, the pervious or impervious
area ratio is critical for the generation and dynamics of surface runoff and flooding. The
spatial configuration among pervious covers, impervious covers, and outlets determines
the overland flow routing and is also considered in some models, such as SWMM and
MIKE FLOOD [75]. For plant cover, distinguishing herbs and shrubs from trees and de-
termining their leaf areas are also necessary to better understand the effects of different
vegetation types in mitigating flooding by more accurately simulating eco-hydrological
processes, such as canopy interception of rainfall [76]. For hardened surfaces, different
underlying surface materials (e.g., asphalt, cement, and brick) with different roughness
values and permeabilities must be identified. Other land use types, especially those
closely associated with human activities in urban areas, such as LID facilities, need to be
identified because of their different eco-hydrological effects. The complexity and diversity
of urban land cover and land use require data with a high spatial or spectral resolution to
identify the types and materials of the underlying surface and to discretize sub-

Water 2023, 15, 622 11 of 24
catchments. In some dynamic urban research, land use data with a high temporal resolu-
tion are also needed [77].
In the past few decades, remote sensing techniques have been used to obtain land
cover and land use data through feature extraction and classification based on distinct
electromagnetic wave information of the land surface objects. In contrast, hyperspectral
remote sensing technology has greatly improved the ability to distinguish and recognize
urban features, because it reduces the occurrence of phenomena in which different objects
may have the same spectrum and the same object may have different spectra [78,79]. How-
ever, the spatial resolution of an image is generally low when its spectral resolution is
high, which increases the difficulty of extracting and classifying complex features in urban
areas. Generally, easily obtained high-spatial-resolution images are used as supplements
to hyperspectral data. Globally, the mainstream remote sensing satellites with high spatial
resolution include IKONOS, SPOT-5, the Earth Observation System, QuickBird,
WorldView-1/2, Beijing-1, and Gaofen, and their spatial resolutions can reach the meter
level [80].
2. Meteorological information
Meteorological data represent important inputs for urban flood models. With the de-
velopment of remote sensing technology, many meteorological-related data, such as rain-
fall [81], evapotranspiration [82], and soil moisture [83,84], can be rapidly quantified using
retrieval algorithms [85,86].
Precipitation observations based on remote sensing images represent the most effec-
tive method for obtaining spatial and temporal distribution characteristics and changes in
urban rainfall. Satellite and fusion precipitation products have been widely used in urban
flood simulations [87]. Satellite precipitation information primarily depends on the radi-
ation characteristics of clouds in the atmosphere acquired from onboard sensors [88]. The
radiation characteristics of clouds are directly related to the state parameters of the cloud
layer, such as the cloud layer height and thickness, cloud top temperature, and cloud ex-
pansion rate. These parameters can be characterized using brightness temperature data
obtained from sensors. At present, the cloud index and cloud life history methods are
widely used to retrieve precipitation data based on the radiation characteristics of the
cloud layer from visible/infrared images [89]. A precipitation calculation method based
on statistical data and cloud radiation models was established for the retrieval of micro-
wave observations [90].
Various satellite precipitation products have been produced, including tropical rain-
fall measuring mission (TRMM) multisatellite precipitation analysis (TMPA, 0.25°, 3 h),
global satellite mapping for precipitation (GSMap, 0.1°, 30 min–1 day), and integrated
multisatellite retrievals for global precipitation measurement (IMERG, 0.1°, 30 min). These
products can provide precipitation data with a wide variety of spatial and temporal reso-
lutions for urban flood simulations. Their application in urban flood simulations have
provided useful insights regarding the temporal evolution and spatial variation of the
precipitation [91–94]. However, all satellite precipitation products failed to provide accu-
rate rainfall records [92]. In general, these products have large systematic errors and ran-
dom errors, which are affected by atmospheric attenuation, sensor performance, spatial
resolution, and inversion algorithms [95]. In particular, IMERG data cannot reproduce
local scale extreme precipitation, with an error of more than 70 mm, because of their coarse
spatial resolution. Compared to satellite data, ground data have higher accuracy, and cou-
pling with ground weather station data is a critical approach to obtain high-accuracy pre-
cipitation products. For example, the hourly precipitation grid dataset issued by the China
Meteorological Data Service Center was coupled with CMORPH precipitation products
based on data from the automatic weather stations from over all of China. However, these
products cannot meet the requirement for the simulation of rapidly occurring and chang-
ing flash flood processes because of their low temporal resolution [91]. Given that ground
weather stations and weather radar can provide the massive precipitation data

Water 2023, 15, 622 12 of 24
continuously measured at minute or second levels, some studies developed the methods
of fusing satellite data with ground data to generate precipitation products that can have
a spatial resolution of 1 km and temporal resolution of 6 min under ideal conditions [95].
Evapotranspiration represents the water flux that diffuses from the endothermic
phase change of surface water and vegetation water to the atmosphere, and it includes
vegetation transpiration, soil evaporation, and canopy interception evaporation. Evapo-
transpiration may be ignored during rainstorms. However, it can greatly affect the water
cycle and the ultimate soil water condition of an ecosystem before rainfall, which then
influences infiltration, surface runoff, and flood genesis during rainfall. In urban areas,
land use changes from permeable and moist natural surfaces (e.g., soil and vegetation) to
large areas of impervious artificial surfaces (e.g., roads and buildings) fundamentally in-
fluence the thermodynamics, water circulation, and aerodynamic characteristics of the
underlying surface and, thus, have direct consequences on surface evaporation and veg-
etation transpiration [96]. Currently, remote-sensing sensors cannot directly observe
evapotranspiration and mainly measure surface parameters that are closely related to
evapotranspiration, such as surface temperature, specific emissivity, surface type, surface
albedo, vegetation index, and vegetation coverage. These parameters can be used to esti-
mate evapotranspiration.
Typical evapotranspiration products include the global evapotranspiration product
MOD16, which is based on the moderate-resolution imaging spectrometer (MODIS), on
global evapotranspiration products based on the advanced very high-resolution radiom-
eter, and on global land evaporation Amsterdam model products. However, continuous
evapotranspiration is difficult to estimate in space and time because of the transient nature
of remote sensing images and cloud interference.
Soil water conditions can determine the water exchange between the atmosphere and
land surface. Hence, soil moisture is an important parameter or an initial variable in flood
models. Soil moisture can be retrieved from remote sensing images based on the spectral
reflectance or emission characteristics of the soil and vegetation. For example, soil mois-
ture estimates from the temperature vegetation dryness index rely on the land surface
temperature and vegetation index, which can be easily retrieved from thermal infrared
and visible near-infrared spectrum images, respectively [97]. Soil moisture estimates from
soil thermal inertia depend on land surface albedo and the diurnal temperature range of
topsoil, which can be retrieved from the multispectral reflectance and thermal infrared
emissivity of topsoil at different times, respectively [98]. Some soil moisture estimates are
retrieved from the brightness temperature obtained from microwave remote sensing im-
ages [99].
Existing soil moisture remote sensing products include essential climate-variable soil
moisture (0.25°), soil moisture and ocean salinity (25 km), advanced microwave scanning
radiometer 2 (25 km), and soil moisture active/passive detection satellite products (36 km)
[100]. Although great progress has been made in the remote sensing inversion of soil mois-
ture, many aspects need to be strengthened. For example, existing soil moisture data
mainly represent moisture information of the top soil layer (usually 0–20 cm), which has
limited applications in eco-hydrology studies, because the roots of most plants (especially
trees) can reach far below the top layer [101]. In particular, current soil moisture data have
low spatial resolutions and, thus, have limited applications in urban flood models. The
development of future soil moisture inversion methods based on physical principles will
be challenging because of the high number of factors (such as soil physical and chemical
properties) that may affect soil moisture and the limited number of parameters that can
be retrieved from remote sensing data under the current technical situation.
3. Flood information
Remote sensing technology has been widely used to identify water bodies (including
floods) since the late 1970s, especially through the Earth Resources Technology Satellite
and Landsat series [102]. Floods were identified based on the spectral characteristics of

Water 2023, 15, 622 13 of 24
water compared with other land cover types. Flood information (range, depth, and dura-
tion) was extracted from the electromagnetic spectrum reflected by water to realize the
dynamic monitoring of flood processes. This information can be used as samples for train-
ing urban flood models, calibrating parameters, validating simulated flood range or
depth, and assimilating models [103].
Different types of remote sensing data have different advantages and limitations for
identifying floods (Table 2). Optical remote sensing data have a large range of spatial and
temporal resolutions and, thus, can meet the requirements for identifying flood range,
depth, and duration at different scales. For example, the Gaofen-2 satellite data have a
spatial resolution finer than 10 m, and the WorldView commercial satellite data even
reach the submeter level. MODIS data can capture the dynamic process of water inunda-
tion during floods well because of their prominent advantages in high temporal resolution
(0.5 d), high spatial resolution (250 m), and wide spectral range (0.4–14 μm) [104]. How-
ever, optical sensors are easily affected by clouds and weather, and obtaining effective
data during flood events is difficult due to overcasting [105].
Table 2. Application of different types of remote sensing data to flood information identification for
urban flood models.
Remote Sensing Data Type
(Temporal and Spatial Res-
olution)
Flood
Information Characteristic Satellite/Sensor
Optical data
(0.5–26 days, <500 m)
Flood range
Flood depth
Flood duration
Vulnerable to clouds and
weather
MODIS *
Landsat TM *
Passive microwave data
(twice a day, 10–70 km)
Flood range
Flood duration
Less affected by weather;
low spatial resolution;
high temporal resolution
AMSR-E *
TRMM/TMI *
SSM/I *
Active microwave data
(14–28 days, 1–10 m)
Flood range
Flood depth
Unaffected by weather
high spatial resolution;
low temporal resolution
TerraSAR-X *
COSMO-SkyMed
*
ALOS *-2
Sentinel-1
Note(s): * MODIS, moderate resolution imaging spectroradiometer; TM, thematic mapper; AMSR-
E, advanced microwave scanning radiometer for the Earth Observation System; TRMM, tropical
rainfall measurement mission; TMI, TRMM microwave imager; SSM/I, special sensor microwave
imager; TerraSAR-X, terra synthetic aperture radar-X; COSMO-SkyMed, constellation of small sat-
ellites for Mediterranean basin observation-SkyMed; ALOS, advanced land observation satellite
phased array L-band synthetic aperture radar.
Passive microwaves can penetrate clouds or vegetation and are less affected by
weather conditions or land cover (Table 2). Passive microwave remote sensing sensors
have a short revisit period, and their observation frequency can reach twice a day over
most parts of the world; thus, they are suitable for monitoring broad-scale flood processes
and ranges. Brakenridge et al. demonstrated that electromagnetic radiation reflected or
emitted from the surface of floods (water) and land was very easy to distinguish when the
band frequency of the horizontal polarization of the sensor was 36.5 GHz or 37 GHz [106].
However, passive microwave systems are rarely used alone to extract urban flood infor-
mation because of the coarse spatial resolution of the data [107]. In contrast, active micro-
wave systems have a high spatial resolution (1–10 m); thus, more accurate flood range and
depth information can be obtained. Microwaves transmitted from such sensors can pene-
trate cloud cover, haze, and dust, regardless of the weather conditions, which means that
flood events can be observed during the day and night [108,109]. Studies have developed
automated flood detection algorithms or systems based on synthetic aperture radar (SAR)

Water 2023, 15, 622 14 of 24
images. However, SAR may not be able to observe most of the land surface in urban areas
with high buildings or tall trees because of its side-viewing nature, as well as the corner
reflection that frequently appears for urban buildings with a rectangular surface structure
[107,110].
The use of satellite remote sensing technology in flood monitoring and disaster emer-
gency response is limited because of the transient transit time of satellites [111]. Un-
manned aerial vehicle (UAV) technology is a new emerging technology that can compen-
sate for this limitation. UAVs capable of carrying different sensors have become an im-
portant means of monitoring flood disasters over the past 10 years, because UAV surveys
are flexible and convenient and can generate high-precision (submeter level) data [112].
Feng et al. used a UAV carrying an RGB digital camera to obtain optical images and mon-
itor serious urban waterlogging in Yuyao, Zhejiang Province, China [113]. Researchers
have also used UAVs carrying LiDAR (mainly composed of a global position system, an
inertial measurement unit, and digital cameras) to reconstruct high-precision 3D models
and DEMs, and they also combined UAVs with GIS spatial analysis technology to extract
flood range data and assess flood risk [114]. In addition, the simultaneous localization and
mapping (SLAM) technique with LiDAR equipment can quickly reconstruct the real-time
surrounding scenes without time and place limitations; thus, it can obtain more precise
flood-related data [115].
4.2.2. Crowdsourcing Geographic Data
Crowdsourcing geographic data refer to open geographic data obtained by a large
number of voluntary nonprofessionals and professionals and provided to the public in
the form of text, pictures, and videos via the Internet (Figure 5) [116]. In current society,
the public can play an unexpected role in the issue of real-time flood information, which
increases the data size considerably.
Crowdsourced data are a type of unstructured data that belong to natural language
text. Flood information extraction from crowdsourced data mainly rely on natural lan-
guage processing, which is a type of AI algorithm. Fohringer et al. were the first to man-
ually derive flood inundation area maps from social media posts on Twitter [117]. Li et al.
used flood-related text information released on Sina Weibo to monitor rainstorm events
in Wuhan and Shenzhen, China [118]. In addition to monitoring and presenting the cur-
rent flood occurrence, these data may be used to validate the simulated flood situations
and further develop a model with high reliability and accuracy [21]. Furthermore, the vul-
nerability to urban flood disasters or the risk of urban flooding can be assessed based on
real-time crowdsourcing data [119].
4.2.3. Internet of Things Data
IoT refers to a network that combines various sensors, such as infrared sensors, laser
scanners, and global positioning systems, via the Internet or telecommunications net-
works [120]. In the field of flood studies, common IoT devices include wireless sensors,
cameras, mobile phones, and automatic weather stations (Figure 5). They are usually lo-
cated at specific locations, such as water-prone road sections, low-lying areas, under-
passes of overpasses, and tunnels. By using these devices, the network can automatically
identify, locate, track, and monitor real-time flood events [121], from which we can gen-
erate a large amount of continuous flood-related data (such as rainfall, flood flow speed,
and flood depth) that can be used to construct or evaluate urban flood models. For exam-
ple, an electronic water gauge can accurately monitor the dynamics of flood depth on a
road at a minute level. The entire process of surface flooding and receding water in wa-
terlogging events can be obtained from video surveillance systems; then, flood inundation
information can be retrieved using computer image recognition technology. Based on the
aforementioned technologies, high-definition cameras can detect real-time floods within
a few seconds [117,122].

Water 2023, 15, 622 15 of 24
IoT systems are suitable for the early forecasting of urgent events, such as urban
floods, because they can efficiently collect, manage, analyze, and share real-time data. Im-
minent rainstorm events can be predicted from the current meteorological conditions
monitored by an IoT system and can then be used as inputs to predict the range and in-
tensity of potential waterlogging through urban flood models. Subsequently, the IoT sys-
tem can issue this information to the public via the Internet.
Hyper-resolution monitoring of urban floods that relies on crowdsourced geographic
data and IoT data can supplement the deficiencies in traditional urban flood data and be
expected to reveal more phenomena. However, the current application of these big data
methods in urban flood simulation is limited. First, collecting a large amount of data is
difficult because of the protection of users’ personal privacy or intellectual property and
the inability to deploy IoT devices over an entire urban area. Second, processing these big
data requires a considerable amount of time. There are considerable redundancies and
noise contained within these big data, and considerable data cleaning work is required,
including checking the credibility of information and evaluating the uncertainty of flood
location and depth [123]. Data cleaning relies on more advanced cloud computing tech-
nology for data resources and a more comprehensive big data computing framework for
algorithm support.
4.3. Spatial Data Management and Analysis
Although current advanced technologies have provided rich data for urban flood
simulations, they have also introduced new challenges. First, rapid increases in various
high-resolution data necessitate strict requirements for high-performance data pro-
cessing, storage, and management. Second, approaches for fusing massive multi-source
data and forming a comprehensive basic database should be developed. Most critically,
the data obtained at multiple spatial and temporal levels also creates issues related to scal-
ing, such as determining the relationships that occur between adjacent-scale or cross-scale
data or information. Tackling these challenges requires more effective spatial data man-
agement and analytical methods.
4.3.1. Spatial Data Platform
GIS has been widely used in flood simulations to provide spatial data management
functions. More importantly, given that most urban flood models cannot express the spa-
tial distribution of floods, the strong spatial analysis capabilities of GIS become indispen-
sable.
Basic data analysis tools in GIS can be used to calculate the various parameters re-
quired for urban flood models. For example, hydrological analysis tools (e.g., basin, flow
direction, and flow accumulation in ArcGIS software) have been used to determine the
flow direction and water outlets in a sub-catchment using DEM data, based on which the
sub-catchments become discretized [124]. The 3D analyst tool in ArcGIS has been used to
calculate various parameters, such as the average slope of a sub-catchment. Zonal statis-
tics in ArcGIS have been used to calculate the sub-catchment area and the pervious to
impervious surface ratio based on land use data [125]. By using the topology function in
ArcGIS, a drainage pipe network layout that represents the actual situation can be estab-
lished, which is crucial for improving the simulation accuracy of pipe convergence pro-
cesses [126]. Based on these tools, ModelBuilder programming techniques in ArcGIS can
be used to develop automatic workflows of calculation processes, which can greatly im-
prove the efficiency of processing data [124].
GIS can also provide a platform that integrates different technologies or models, such
as spatial information technology, big data technology, AI technology, and hydrological
models, to build an integrated urban flood information system. Jing et al. developed a
model based on 2D unsteady flow theory using the ArcView software platform to con-
struct a rainstorm and waterlogging monitoring and early warning system in Harbin, Hei-
longjiang Province, China [127]. Wang et al. used 1D open channel unsteady flow and 2D

Water 2023, 15, 622 16 of 24
unsteady flow equations, combined with real-time rainfall monitoring data from regional
automatic stations, to refine precipitation forecasts, as well as construct a dynamic forecast
and early warning system on the Meteo GIS platform for urban waterlogging manage-
ment in Langfang, Hebei Province, China [128]. Urban flood models have also been de-
veloped based on GIS platforms, such as MIKE Urban, InfoWorks ICM, and Digital Water
DS. These platforms can convey detailed flood physical mechanisms, have fast processing
capabilities for spatially distributed data, and present efficient modeling capabilities.
Some studies mixed AI models and GIS analysis to extrapolate local forecasts to flooded
areas at sub-catchment scales [129].
Future research will mainly focus on supporting real-time data formats [130], and 3D
GIS will become a support platform for ensuring the accuracy and credibility of flood
simulations [131]. Such research has considerable room for improvement.
4.3.2. Spatial Data Analysis
Spatial analysis methods (especially spatial statistics methods) can play indispensa-
ble roles in determining suitable data inputs, in analyzing the spatial patterns of simulated
flooding processes, in linking the patterns and processes among different scales, and in
developing multi-scale distributed flood models.
The selection of spatial data depends mainly on the research purpose, scale of discre-
tized catchments, and average size of historical inundation areas [132]. If we are concerned
only with large inundation areas in a large region, then coarse-resolution data and catch-
ments are practicable. In contrast, if we are concerned with all the inundation areas re-
gardless of their sizes or depths, then the fine-resolution data must be selected to identify
catchments at a fine scale. Regardless of the type of concern, the spatial grain size of the
data should be smaller than the average size of the inundated patches [133]. Thus, a
method for determining the average patch size before performing spatial data selection is
required. The average size of inundated patches (and their connectivity, proximity, isola-
tion, aggregation, and dispersion) can be calculated using spatial pattern analysis soft-
ware, such as Fragstats 4.2 [134], based on a generated raster image with various patch
types. For a large and spatially complicated urban area, some spatial statistical methods
(such as lacunarity index, spatial autocorrelation analysis, and scale variance analysis or
scale variance analysis coupled with Moran’s I scalogram) can help identify the average
patch size [135,136].
The identification of multi-scale spatial patterns of flood-related variables (such as
rainfall, soil moisture, evapotranspiration, and surface runoff) relies on spatial statistical
methods. In addition to the methods mentioned above, which are used for patch-type
data, wavelet analysis is widely used for quantitative variables. From the spatial statistical
analysis, we may determine the nested hierarchical structure of these variables across
scale domains in a large range, such as the small flooding aggregation occurring at local
sites, medium aggregation covering multiple land covers, and large aggregation across
adjacent sub-catchments. These spatial statistical methods can also identify whether a var-
iable is spatially random, uniform, aggregated, or dispersed within each scale domain, as
well as whether the two variables have a spatially positive or negative association. Within
a limited range or scale domain, if the spatial pattern analysis shows that a variable pre-
sents a fractal structure rather than a hierarchical structure, then the complex flooding
processes can be dealt with in a simple way based on power laws [137].
The multi-scale spatial pattern analysis of inundated patches and flood-related vari-
ables can help us not only explore the spatial patterns of patches and variables and their
associations with hydrological processes within each scale domain, but also explore the
pattern–process relationships across scales. Thus, spatial pattern analysis is the basis for
developing a multi-scale distributed flood model across cell (or site), patch, sub-catchment
(or catchment), and landscape (or region) scales [40,137]. If water flow interactions among
cells are fully comprehended, then upscaling from cells to patches can be realized, and if
water flow routings among adjacent patches are determined, then hydrological processes

Water 2023, 15, 622 17 of 24
can be intrinsically upscaled from patches to sub-catchments or catchments [138]. We de-
veloped CA-DUSRM, a cellular automata-based distributed hydrological model for urban
surface runoff, in which the nonlinear reservoir algorithm was improved and combined
with a weighting system to simulate surface runoff processes. We assumed that each cell
interacted with other cells in its neighborhood via local surface water exchange rules. Such
rules realize the upscaling of runoff, first from cells to their neighborhoods, and then from
cell neighborhoods to their distinct nearest outlets. Ultimately, runoff converged to mul-
tiple outlets is lumped to a sub-catchment [42]. The whole upscaling is realized within the
bottom-up simulation framework of cellular automata. The similar upscaling approach
can be implemented when simulating the spatial expansion of water overflowed from
pipe network nodes to their nearest neighborhoods and then to farther surroundings.
More heuristic simulation approaches or sophisticated spatial technologies are required
for future model development, which represents one of the most rewarding challenges.
4.3.3. Uncertainty in Spatial Analysis
Although advanced technologies have facilitated flood model development, uncer-
tainties exist in the spatial analysis methods for such models. Original data have errors,
and estimated data have uncertainties [19]. For example, fine urban meteorological data
and underlying surface parameters (e.g., depression storage, Manning’s roughness coef-
ficient, and soil texture) are mainly obtained from local ground stations and sample sites.
Uncertainties always occur when such data are spatially extrapolated to a large area, due
to insufficient sample density and imperfect upscaling methods [138]. When multiple data
with different types and sources are input to a flood model, data transformation to unified
formats (e.g., the same resolution and geographical project) will inevitably produce devi-
ations. Scale effect is another source of uncertainty for model outputs. Different spatio-
temporal grains for the same variables or parameters may generate great discrepancies in
simulated runoff and flood processes; for example, our study showed that increasing tem-
poral grains can result in aliasing in hydrographs [42]. In these cases, the uncertainties of
model outputs over a space can be estimated based on the statistical distributions of input
data (and correlations between inputs) from their actual values or generated random val-
ues that conform to their statistical characteristics.
In addition, although certain technologies and approaches can theoretically be used
to obtain the information on flood inundation range and depth, there might be still many
places and times for which the information has not been covered. Such cases will make
model validation difficult and uncertainty evaluations incomplete, especially for small
flooded sites.
5. Conclusions
Urban flood models can be classified into three categories: simplified, physical, and
data-driven. With the application of advanced technologies in flood models, the simula-
tion accuracy and efficiency have improved markedly, thus providing support for the es-
tablishment of a smart and integrated model framework for urban floods.
Although a perfect urban flood model has not yet been developed, the advantages of
distinct models can be combined to produce an optimal model. The designed perfect
model framework should be a multi-scale distributed smart simulation system that con-
siders cells as the basic units and supports simulations within and across distinct scales,
such as cells, patches, sub-catchments, and catchments.
The optimal system involves three main parts: inputs, models (algorithms, functions
or rules), and outputs. For model input variables and parameters, remote sensing,
crowdsourcing geographic data, and IoT big data technologies and methods that have
been sufficiently developed and improved should be considered via the spatial data man-
agement functions of GIS platforms. By using image processing software and 3D GIS tech-
nologies, the synthesis and presentation of real-time and authentic flooding maps should
be a critical output part of systems for the early warning and control of floods.

Water 2023, 15, 622 18 of 24
Regarding the model, the expected models should integrate two types of general
strategies that have separate requirements for flood simulation. The first general strategy
focuses on simulation accuracy. For accuracy, modules associated with eco-hydrological
and hydrodynamic processes at multiple scales, such as overland flow and inundation at
the cell or local scale, runoff generation and confluence at patch-scale LID facilities, and
overflow in pipe networks at the sub-catchment scale, should be considered and com-
bined into the system using a common GIS platform. The modules should be continuously
updated using the developed mechanisms. The second general strategy focuses on simu-
lation efficiency. For efficiency, AI models that are trained using advanced algorithms
based on a large amount of meteorological, topographic, land cover, and flood data are
required. Given that we must simultaneously address the issues of simulation accuracy
and efficiency during urgent urban waterlogging, the expected models should not be com-
pletely mechanism models or completely non-mechanism (or statistical) models; rather,
they should effectively couple mechanism and non-mechanism models.
Although many advanced spatial technologies have been developed for the construc-
tion of the expected multi-scale distributed smart simulation system, challenges remain.
The most difficult task is to determine the scaling laws of eco-hydrological and hydrody-
namic processes based on the spatial pattern characteristics of processes and pattern–pro-
cess relationships across scales. Resolving these issues will not only require advanced
methods in spatial analysis, but will also require an understanding of the dynamics and
mechanisms at multiple scales. The second most difficult task is to develop models that
can couple calculations driven by flooding processes with AI algorithms that are driven
by data to realize high simulation accuracy and efficiency.
We suggest that a multi-scale distributed smart simulation system will provide more
accurate flood information in a highly efficient manner, and such characteristics will be
essential for the wide application of such systems for the early warning and control of
floods.
Author Contributions: Funding acquisition, N.Z.; Writing—original draft, Y.Y., N.Z.; Writing—re-
view and editing, N.Z., H.Z. All authors have read and agreed to the published version of the man-
uscript.
Funding: This research was supported by the Beijing Natural Science Foundation [8181001] and the
Special Fund for Scientific Research Cooperation between Colleges and Institutes of the University
of Chinese Academy of Sciences [Y65201NY00].
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data will be made available upon request.
Conflicts of Interest: No potential conflict of interest was reported by the authors.
References
1. Chen, G.-Z.; Hou, J.-M.; Zhou, N.; Yang, S.-X.; Tong, Y.; Su, F.; Huang, L.; Bi, X. High-resolution urban flood forecasting by
using a coupled atmospheric and hydrodynamic flood models. Front. Earth Sci. 2020, 8, 545612.
https://doi.org/10.3389/feart.2020.545612.
2. EM-DAT: The International Disaster Database. 2018. Available online: http://www.emdat.be (accessed on 18 December
2022).
3. Carbone, D.; Hanson, J. Floods: 10 of the deadliest in Australian history. Australian Geographic. 2012, March 8.
https://www.australiangeographic.com.au/topics/history-culture/2012/03/australias-worst-floods/ (accessed on 18
December 2022).
4. Task Force on Urban Flooding Problem and Solution Investigation (TFUFPSI). China’s urban flooding program and solution.
China Floods Droughts Prot. 2014, 24, 46–48, 65. (In Chinese)
5. Qiu, J. Urbanization contributed to Beijing storms. Nature 2012, 10, 11086. https://doi.org/10.1038/nature.2012.11086.

Water 2023, 15, 622 19 of 24
6. Ye, M.-H.; Chen, K. Urban catastrophe risk: Meteorological characteristics, loss status and optimization of management coun-
termeasures: Taking Zhengzhou “7.20” heavy rainstorm and Typhoon “Fireworks” as an example. Shanghai Insur. 2021, 08, 18–
22.
7. Stefanidis, S.; Alexandridis, V.; Theodoridou, T. Flood exposure of residential areas and infrastructure in Greece. Hydrology
2022, 9, 145. https://doi:10.3390/hydrology9080145.
8. Koks, E.E.; van Ginkel, K.C.H.; van Marle, M.J.E.; Lemnitzer, A. Brief communication: Critical infrastructure impacts of the 2021
mid-July Western European flood event. Nat. Hazards Earth Syst. Sci. 2022, 22, 3831–3838. https://doi:10.5194/nhess-22-3831-
2022.
9. Contreras, M.T.; Gironás, J.; Escauriaza, C. Forecasting flood hazards in real time: A surrogate model for hydrometeorological
events in an Andean Watershed. Nat. Hazards Earth Syst. Sci. 2020, 20, 3261–3277. https://doi.org/10.5194/nhess-20-3261-2020.
10. Agarwal, S.; Kumar, S. Urban flood modeling using SWMM for historical and future extreme rainfall events under climate
change scenario. Indian J. Ecol. 2020, 47(11), 48–53.
11. Rossman LA, Huber WC. Hydrology. In Storm Water Management Model Reference Manual; EPA/600/R-15/162A; United States
Environmental Protection Agency: Cincinnati, OH, USA, 2016; Volume 1.
12. Rossman, L.A.; Dickinson, R.E.; Schade, T.; Chan, C.C.; Burgess, E.; Sullivan, D.; Lai, F.H. SWMM 5—The next generation of
EPA’s storm water management model. J. Water Manag. Model. 2004, 12, 339–358. https://doi.org/10.14796/JWMM.R220-16.
13. Anni, A.H.; Cohen, S.; Praskievicz, S. Sensitivity of urban flood simulations to stormwater infrastructure and soil infiltration. J.
Hydrol. 2020, 588, 125028. https://doi.org/10.1016/j.jhydrol.2020.125028.
14. Wu, W.-L.; Lu, L.-J.; Huang, X.-F.; Shangguan, H.-D.; Wei, Z.-Q. An automatic calibration framework based on the InfoWorks
ICM model: The effect of multiple objectives during multiple water pollutant modeling. Environ. Sci. Pollut. Res. 2021, 28, 31814–
31830. https://doi.org/10.1007/s11356-021-12596-4.
15. InfoWorks ICM. Version 3.0.; Innovyze: Wallingford, UK, 2012.
16. HEC-RAS River Analysis System. Version 5.0.; 2D Modeling User’s Manual; Hydrologic Engineering Center: Davis, CA, USA,
2016.
17. Xu, Z.-X.; Cheng, T.; Hong, S.-Y.; Wang, L.-X. Review on applications of remote sensing in urban flood modeling. Chin. Sci. Bull.
2018, 63, 2156–2166. https://doi.org/10.1360/N972018-00239.
18. Qi, L.; Li, J.; Wang, Y.; Gao, X.-B. Urban observation: Integration of remote sensing and social media data. IEEE J. Sel. Top. Appl.
Earth Obs. Remote Sens. 2019, 12, 4252–4264. https://doi.org/10.1109/JSTARS.2019.2908515.
19. Teng, J.; Jakeman, A.J.; Vaze, J.; Croke, B.F.W.; Dutta, D.; Kim, S. Flood inundation modelling: A review of methods, recent
advances and uncertainty analysis. Environ. Modell. Softw. 2017, 90, 201–216. https://10.1016/j.envsoft.2017.01.006.
20. Abdulkareem, J.H.; Pradhan, B.; Sulaiman, W.N.A.; Jamil, N.R. Review of studies on hydrological modelling in Malaysia. Model.
Earth Syst. Environ. 2018, 4, 1577–1605. https://doi.org/10.1007/s40808-018-0509-y.
21. Qi, W.-C.; Ma, C.; Xu, H.-S.; Chen, Z.-F.; Zhao, K.; Han, H. A review on applications of urban flood models in flood mitigation
strategies. Nat. Hazards 2021, 108, 31–62. https://doi.org/10.1007/s11069-021-04715-8.
22. Fan, Y.-Y.; Ao, T.-Q.; Yu, H.-J.; Huang, G.-R.; Li, X.-D. A Coupled 1D-2D hydrodynamic model for urban flood inundation. Adv.
Meteorol. 2017, 2017, 1–12. https://doi.org/10.1155/2017/2819308.
23. Salvadore, E.; Bronders, J.; Batelaan, O. Hydrological modelling of urbanized catchments: A review and future directions. J.
Hydrol. 2015, 529, 62–81. https://doi.org/10.1016/j.jhydrol.2015.06.028.
24. Li, X.-L.; Wang, E.-J.; Zhao, J.; Cheng, M.-G.; Fan, Z.-Q. Waterlogging modeling and simulating for urban crisis management
wargaming system. In Proceedings of the 33rd Chinese Control Conference, Nanjing, China, 28 July 2014; pp. 6375–6381.
25. Zhao, L.-L.; Liu, C.-M.; Sobkowiak, L.; Wu, X.-X.; Liu, J.-F. A review of underlying surface parameterization methods in hydro-
logic models. J. Geogr. Sci. 2019, 29, 1039–1060. https://doi.org/10.1007/s11442-019-1643-9.
26. Zou, Y.-J. Discussion on urban surface runoff model. Sci. Surv. Mapp. 2013, 38, 5.
27. Ren, B.-Z.; Deng, R.-J. Analyses of properties and calculation methods of urban surface rainwater conflux. China Water
Wastewater 2006, 22, 39–42. https://doi.org/10.3321/j.issn:1000-4602.2006.14.012.
28. Todini, E. On the convergence properties of the different pipe network algorithms. In Proceedings of the Water Distribution
Systems Analysis Symposium, Cincinnati, OH, USA, 27–30 August 2006; pp. 1–16.
29. Song, L.-X.; Xu, Z.-X. Coupled hydrologic-hydrodynamic model for urban rainstorm water logging simulation: Recent ad-
vances. J. Beijing Norm. Univ. 2019, 55, 581. https://doi.org/10.16360/j.cnki.jbnuns.2019.05.005.
30. Yue, M.-L. Study on Risk Simulation and Early Warning of Urban Rainstorms—A Case Study of Zhengzhou City. Master’s
Thesis, North China University of Water Resources and Electric Power, Zhengzhou, China, 2019.
31. Jiang, Y.; Zevenbergen, C.; Ma, Y.-C. Urban pluvial flooding and stormwater management: A contemporary review of China’s
challenges and “sponge cities” strategy. Environ. Sci. Policy 2018, 80, 132–143. https://doi.org/10.1016/j.envsci.2017.11.016.
32. Pender, G. Briefing: Introducing the flood risk management research consortium. In Proceedings of the Institution of Civil Engi-
neers-Water Management; Thomas Telford Ltd.: London, UK, 2006; Volume 159, pp. 3–8.
33. Jamali, B.; Löwe, R.; Bach, P.M.; Urich, C.; Arnbjerg-Nielsen, K.; Deletic, A. A rapid urban flood inundation and damage assess-
ment model. J. Hydrol. 2018, 564, 1085–1098. https://doi.org/10.1016/j.jhydrol.2018.07.064.

Water 2023, 15, 622 20 of 24
34. Niazi, M.; Nietch, C.; Maghrebi, M.; Jackson, N.; Bennett, B.R.; Tryby, M.; Massoudieh, A. Storm water management model:
Performance review and gap analysis. J. Sustain. Water Built Environ. 2017, 3, 04017002. https://doi.org/10.1061/JSWBAY.0000817.
35. DHI. Collection System. User Guide; Danish Hydraulic Institute: Horsholm, Denmark, 2011.
36. Bisht, D.S.; Chatterjee, C.; Kalakoti, S.; Upadhyay, P.; Sahoo, M.; Panda, A. Modeling urban floods and drainage using SWMM
and MIKE URBAN: A case study. Nat. Hazards 2016, 84, 749–776. https://doi.org/10.1007/s11069-016-2455-1.
37. Sidek, L. M., Jaafar, A. S., Majid, W. H. A. W. A., Basri, H., Marufuzzaman, M., Fared, M. M., & Moon, W. C. High-resolution
hydrological-hydraulic modeling of urban floods using infoworks ICM. Sustainability, 2021, 13(18), 10259.
38. Verma, A.K.; Jha, M.K.; Mahana, R.K. Evaluation of HEC-HMS and WEPP for simulating watershed runoff using remote sens-
ing and geographical information system. Paddy Water Environ. 2010, 8, 131–144. https://doi.org/10.1007/s10333-009-0192-8.
39. Guidolin, M.; Chen, A.S.; Ghimire, B.; Keedwell, E.C.; Djordjević, S.; Savić, D.A. A weighted cellular automata 2D inundation
model for rapid flood analysis. Environ. Modell. Softw. 2016, 84, 378–394. https://doi.org/10.1016/j.envsoft.2016.07.008.
40. Feng, C.-H.; Zhang, N.; Habiyakare, T.; Yan, Y.-N.; Zhang, H. Prospects of eco-hydrological model for sponge city construction.
Ecosyst. Health Sustain. 2021, 7, 1994885. https://doi.org/10.1080/20964129.2021.1994885.
41. Jamali, B., Bach, P.M., Cunningham, L., Deletic, A. A cellular automata fast flood evaluation (CA-Ff´e) model. Water Resour. Res.
2019, 55, 4936–4953. https://doi.org/10.1029/2018WR023679.
42. Feng, C.-H. Eco-Hydrological Effects of Urban Landscape Spatial Pattern Change and Simulation. Ph.D. Thesis, University of
Chinese Academy of Sciences, Beijing, China, 2022.
43. Agudelo-Otálora, L.M.; Moscoso-Barrera, W.D.; Paipa-Galeano, L.A.; Mesa-Sciarrotta, C. Comparison of physical models and
artificial intelligence for prediction of flood levels. Tecnol. Cienc. Agua 2018, 9, 209–235. https://doi.org/10.24850/j-tyca-2018-04-
09.
44. Mignot, E.; Li, X.-F.; Dewals, B. Experimental modelling of urban flooding: A review. J. Hydrol. 2019, 568, 334–342.
https://doi.org/10.1016/j.jhydrol.2018.11.001.
45. Khac-Tien Nguyen, P., & Hock-Chye Chua, L. The data-driven approach as an operational real-time flood forecasting model.
Hydrol. Process, 2012, 26(19), 2878-2893. https://doi.org/10.1002/hyp.8347.
46. Zhu, Z.-D.; Oberg, N.; Morales, V.M.; Quijano, J.C.; Landry, B.J.; Garcia, M.H. Integrated urban hydrologic and hydraulic mod-
elling in Chicago, Illinois. Environ. Modell. Softw. 2016, 77, 63–70. https://doi.org/10.1016/j.envsoft.2015.11.014.
47. Berkhahn, S.; Fuchs, L.; Neuweiler, I. An ensemble neural network model for real-time prediction of urban floods. J. Hydrol.
2019, 575, 743–754. https://doi.org/10.1016/j.jhydrol.2019.05.066.
48. Lin, Q.; Leandro, J.; Wu, W.; Bhola, P.; Disse, M. Prediction of maximum flood inundation extents with resilient backpropagation
neural network: Case study of Kulmbach. Front. Earth Sci. 2020, 8(332), 1–8.
49. Löwe, R.; Böhm, J.; Jensen, D.G.; Leandro, J.; Rasmussen, S.H. U-FLOOD—Topographic deep learning for predicting urban
pluvial flood water depth. J. Hydrol. 2021, 603, 126898. https://doi.org/10.1016/j.jhydrol.2021.126898.
50. Zounemat-Kermani, M.; Matta, E.; Cominola, A.; Xia, X.-L.; Zhang, Q.; Liang, Q.-H.; Hinkelmann, R. Neurocomputing in sur-
face water hydrology and hydraulics: A review of two decades retrospective, current status and future prospects. J. Hydrol.
2020, 588, 125085. https://doi.org/10.1016/j.jhydrol.2020.125085.
51. Hunter, N.M.; Bates, P.D.; Horritt, M.S.; Wilson, M.D. Simple spatially-distributed models for predicting flood inundation: A
review. Geomorphology 2007, 90, 208–225. https://doi.org/10.1016/j.geomorph.2006.10.021.
52. Chang, D.-L.; Yang, S.-H.; Hsieh, S.-L.; Wang, H.-J.; Yeh, K.-C. Artificial intelligence methodologies applied to prompt pluvial
flood estimation and prediction. Water 2020, 12, 3552. https://doi.org/10.3390/w12123552.
53. Zanchetta, A.; Coulibaly, P. Recent advances in real-time pluvial flash flood forecasting. Water 2020, 12, 570.
https://doi.org/10.3390/w12020570.
54. Wei, M.; She, L.; You, X.-Y. Establishment of urban waterlogging pre-warning system based on coupling RBF-NARX neural
networks. Water Sci. Technol. 2020, 82, 1921–1931. https://doi.org/10.2166/wst.2020.477.
55. Chapi, K.; Singh, V.; Shirzadi, A.; Shahabi, H.; Bui, D.; Pham, B.; Khosravi, K. A novel hybrid artificial intelligence approach for
flood susceptibility assessment. Environ. Modell. Softw. 2017, 95, 229–245. https://doi.org/10.1016/j.envsoft.2017.06.012.
56. Khosravi, K.; Panahi, M.; Golkarian, A.; Keesstra, S.D.; Saco, P.M.; Bui, D.T.; Lee, S. Convolutional neural network approach for
spatial prediction of flood hazard at national scale of Iran. J. Hydrol. 2020, 591, 125552.
57. Cheng, J.Y.; Xiang, C.; Ma, Y. AI application on LID stormwater management and urban planning in Guam, USA, and Southern
China, PRC. In Proceedings of the International Low Impact Development Conference 2020; American Society of Civil Engi-
neers; Reston, VA, USA, 20–24 July 2020; pp. 188–200.
58. Ke, Q.; Tian, X.; Bricker, J.; Tian, Z.; Guan, G.-H.; Cai, H.; Huang, X.; Yang, H.; Liu, J. Urban pluvial flooding prediction by
machine learning approaches—A case study of Shenzhen City, China. Adv. Water Resour. 2020, 145, 103719.
https://doi.org/10.1016/j.advwatres.2020.103719.
59. Tian, Y.; Xu, Y.-P.; Yang, Z.; Wang, G.; Zhu, Q. Integration of a parsimonious hydrological model with recurrent neural net-
works for improved streamflow forecasting. Water 2018, 10, 1655. https://doi.org/10.3390/w10111655.
60. Hou, J.; Zhou, N.; Chen, G.; Huang, M.; Bai, G. Rapid forecasting of urban flood inundation using multiple machine learning
models. Nat. Hazards 2021, 108, 2335–2356. https://doi.org/10.1007/s11069-021-04782-x.

Water 2023, 15, 622 21 of 24
61. Ye, X.-Y.; Wang, S.-H.; Lu, Z.-P.; Song, Y.; Yu, S.-Y. Towards an AI-driven framework for multi-scale urban flood resilience
planning and design. Comput.Urban Sci. 2021, 1, 11. https://doi.org/10.1007/s43762-021-00011-0.
62. Goyal, H.R.; Ghanshala, K.K.; Sharma, S. Post flood management system based on smart IoT devices using AI approach. Mater.
Today Proc. 2021, 46, 10411–10417. https://doi.org/10.1016/j.matpr.2020.12.947.
63. Chen, Y.; Zhang, D. Integration of knowledge and data in machine learning. arXiv 2022, arxiv:2202.10337.
64. Raissi, M.; Perdikaris, P.; Karniadakis, G.E. Physics-informed neural networks: A deep learning framework for solving forward
and inverse problems involving nonlinear partial differential equations. J. Comput. Phys. 2019, 378, 686–707.
https://doi.org/10.1016/j.jcp.2018.10.045.
65. Bermúdez, M.; Cea, L.; Puertas, J. A rapid flood inundation model for hazard mapping based on least squares support vector
machine regression. J. Flood Risk Manage. 2019, 12, e12522.
66. Felsberger, L.; Koutsourelakis, P.S. Physics-constrained, data-driven discovery of coarse-grained dynamics. arXiv 2018,
arxiv:1802.03824.
67. Chen, Y.; Zhang, D. Physics-constrained deep learning of geomechanical logs. IEEE Trans. Geosci. Remote Sens. 2020, 58, 5932–
5943. https://doi.org/10.1109/TGRS.2020.2973171.
68. Guo, J.; Wu, X.-H.; Wei, G. A new economic loss assessment system for urban severe rainfall and flooding disasters based on
big data fusion. Environ. Res. 2020, 188, 109822. https://doi.org/10.1016/j.envres.2020.109822.
69. Gashaw, T.; Tulu, T.; Argaw, M.; Worqlul, A.W. Modeling the impacts of land use–land cover changes on soil erosion and
sediment yield in the Andassa Watershed, Upper Blue Nile Basin, Ethiopia. Environ. Earth Sci. 2019, 78, 679.
https://doi.org/10.1007/s12665-019-8726-x.
70. Farooq, M.; Shafique, M.; Khattak, M.S. Flood hazard assessment and mapping of river swat using HEC-RAS 2D model and
high-resolution 12-m TanDEM-X DEM (WorldDEM). Nat. Hazards 2019, 97, 477–492. https://doi.org/10.1007/s11069-019-03638-
9.
71. Ejikeme, J.; Igbokwe, J.; Igbokwe, E.; Aweh, D. Evaluation of horizontal and vertical accuracies of SRTM and ASTER GDEMs
for topographic and hydrological modeling in Onitsha, South East Nigeria. Int. J. Innov. Res. Sci. Eng. Technol. 2017, 4, 634–643.
https://doi.org/10.21276/ijirem.2017.4.2.5.
72. Anthony, E.J. Shore Processes and Their Palaeoenvironmental Applications; Elsevier: Amsterdam, The Netherlands, 2008; ISBN 978-
0-444-52733-2.
73. Piégay, H.; Arnaud, F.; Belletti, B.; Bertrand, M.; Bizzi, S.; Carbonneau, P.; Dufour, S.; Liébault, F.; Ruiz-Villanueva, V.; Slater,
L. Remotely sensed rivers in the Anthropocene: State of the art and prospects. Earth Surf. Process. Landf. 2020, 45, 157–188.
https://doi.org/10.1002/esp.4787.
74. Maitre, D.C.L.; Kotzee, I.M.; O’Farrell, P.J. Impacts of land-cover change on the water flow regulation ecosystem service: Inva-
sive alien plants, fire and their policy implications. Land Use Policy 2014, 36, 171–181.
https://doi.org/10.1016/j.landusepol.2013.07.007.
75. Zhang, N.; Luo, Y.-J.; Chen, X.-Y.; Li, Q.; Jing, Y.-C.; Wang, X.; Feng, C.-H. Understanding the effects of composition and con-
figuration of land covers on surface runoff in a highly urbanized area. Ecol. Eng. 2018, 125, 11–25.
https://doi.org/10.1016/j.ecoleng.2018.10.008.
76. Huang, F.; Zhang, Y.; Zhang, D.; Chen, X. Environmental groundwater depth for groundwater-dependent terrestrial ecosystems
in arid/semiarid regions: A review. Int. J. Environ. Res. Public Health 2019, 16, 763. https://doi.org/10.3390/ijerph16050763.
77. Deng, J.S.; Wang, K.; Hong, Y.; Qi, J.G. Spatio-temporal dynamics and evolution of land use change and landscape pattern in
response to rapid urbanization. Landsc. Urban Plan 2009, 92, 187–198. https://doi.org/10.1016/j.landurbplan.2009.05.001.
78. Bioucas-Dias, J.M.; Plaza, A.; Camps-Valls, G.; Scheunders, P.; Nasrabadi, N.; Chanussot, J. Hyperspectral remote sensing data
analysis and future challenges. IEEE Geosci. Remote Sens. Mag. 2013, 1, 6–36. https://doi.org/10.1109/MGRS.2013.2244672.
79. Chen, Y.; Su, W.; Li, J.; Sun, Z. Hierarchical object oriented classification using very high resolution imagery and LIDAR data
over urban areas. Adv. Space Res. 2009, 43, 1101–1110. https://doi.org/10.1016/j.asr.2008.11.008.
80. Liu, B.; Zhang, Y.; Chen, T.; Song, Y. Urban impervious surface extraction based on the GF-2 satellite imagery. Geomat. World
2017, 24, 103–107.
81. Qi, M.; Huang, H.-B.; Liu, L.; Chen, X. Spatial heterogeneity of controlling factors’ impact on urban pluvial flooding in Cincin-
nati, US. Appl. Geogr. 2020, 125, 102362. https://doi.org/10.1016/j.apgeog.2020.102362.
82. Zhang, Y.; Li, L.; Chen, L.-Q.; Liao, Z.-H.; Wang, Y.-C.; Wang, B.-Y.; Yang, X.-Y. A Modified multi-source parallel model for
estimating urban surface evapotranspiration based on ASTER thermal infrared data. Remote Sens. 2017, 9, 1029.
https://doi.org/10.3390/rs9101029.
83. Njoku, E.G.; Jackson, T.J.; Lakshmi, V.; Chan, T.K.; Nghiem, S.V. Soil moisture retrieval from AMSR-E. IEEE Trans. Geosci. Re-
mote Sens. 2003, 41, 215–229. https://doi.org/10.1109/TGRS.2002.808243.
84. Zhou, P.; Ding, J.-L.; Wang, F.; Ubul, G.; Zhang, Z.-G. Retrieval methods of soil water content in vegetation covering areas based
on multi-source remote sensing data. J. Remote Sens. 2010, 5, 15. https://doi.org/10.1109/IITA-GRS.2010.5604231.
85. Wang, H.; Sun, X.-B.; Yang, L.-K.; Zhao, M.-R.; Lui, P.; Du, W.-B. Aerosol retrieval algorithm based on adaptive land–atmos-
pheric decoupling for polarized remote sensing over land surfaces. J. Quant. Spectrosc. Radiat. Transf. 2018, 219, 74–84.
https://doi.org/10.1016/j.jqsrt.2018.08.011.

Water 2023, 15, 622 22 of 24
86. Mu, Q.; Heinsch, F.A.; Zhao, M.; Running, S.W. Development of a global evapotranspiration algorithm based on MODIS and
global meteorology data. Remote Sens. Environ. 2007, 111, 519–536. https://doi.org/10.1016/j.rse.2007.04.015.
87. Ma, M.-H.; Wang, H.-X.; Jia, P.-F.; Tang, G.-Q.; Wang, D.-C.; Ma, Z.-Q.; Yan, H.-M. Application of the GPM-IMERG products in
flash flood warning: A case study in Yunnan, China. Remote Sens. 2020, 12, 1954. https://doi.org/10.3390/rs12121954.
88. Kidd, C.; Levizzani, V. Status of satellite precipitation retrievals. Hydrol. Earth Syst. Sci. 2011, 15, 1109–1116.
https://doi.org/10.5194/hess-15-1109-2011.
89. Varma, A.K. Measurement of precipitation from satellite radiometers (visible, infrared, and microwave): Physical basis, meth-
ods, and limitations. In Remote Sensing of Aerosols, Clouds, and Precipitation; Elsevier: Amsterdam, The Netherlands, 2018; pp.
223–248.
90. Kummerow, C.; Giglio, L. A passive microwave technique for estimating rainfall and vertical structure information from space.
Part I: Algorithm description. J. Appl. Meteorol. Clim. 1994, 33, 3–18. https://doi.org/10.1175/1520-0450(1994)033<0019:AP-
MTFE>2.0.CO;2.
91. Cristiano, E.; ten Veldhuis, M.-C.; van de Giesen, N. Spatial and temporal variability of rainfall and their effects on hydrological
response in urban areas—A review. Hydrol. Earth Syst. Sci. 2017, 21, 3859–3878. https://doi.org/10.5194/hess-21-3859-2017.
92. Kazamias, A.-P.; Sapountzis, M.; Lagouvardos, K. Evaluation of GPM-IMERG rainfall estimates at multiple temporal and spatial
scales over Greece. Atmos. Res. 2022, 269, 106014. https://doi.org/10.1016/j.atmosres.2021.106014.
93. Tegos, A.; Ziogas, A.; Bellos, V.; Tzimas, A. Forensic hydrology: A complete reconstruction of an extreme flood event in data-
scarce area. Hydrology 2022, 9, 93. https://doi.org/10.3390/hydrology9050093.
94. Giannaros, C.; Dafis, S.; Stefanidis, S.; Giannaros, T.M.; Koletsis, I.; Oikonomou, C. Hydrometeorological analysis of a flash
flood event in an ungauged mediterranean watershed under an operational forecasting and monitoring context. Meteorol. Appl.
2022, 29, e2079. https://doi.org/10.1002/met.2079.
95. Salles, L.; Satgé, F.; Roig, H.; Almeida, T.; Olivetti, D.; Ferreira, W. Seasonal effect on spatial and temporal consistency of the
new GPM-Based IMERG-v5 and GSMaP-v7 satellite precipitation estimates in Brazil’s Central Plateau Region. Water 2019, 11,
668. https://doi.org/10.3390/w11040668.
96. He, X.-C.; Gao, W.-J.; Wang, R. Impact of urban morphology on the microclimate around elementary schools: A case study from
Japan. Build. Environ. 2021, 206, 108383. https://doi.org/10.1016/j.buildenv.2021.108383.
97. Tao, L.; Ryu, D.; Western, A.; Boyd, D. A new drought index for soil moisture monitoring based on MPDI-NDVI trapezoid
space using MODIS data. Remote Sens. 2020, 13, 122. https://doi.org/10.3390/rs13010122.
98. Hulley, G.C.; Hook, S.J.; Baldridge, A.M. Investigating the effects of soil moisture on thermal infrared land surface temperature
and emissivity using satellite retrievals and laboratory measurements. Remote Sens. Environ. 2010, 114, 1480–1493.
https://doi.org/10.1016/j.rse.2010.02.002.
99. Kornelsen, K.C.; Coulibaly, P. Advances in soil moisture retrieval from synthetic aperture radar and hydrological applications.
J. Hydrol. 2013, 476, 460–489. https://doi.org/10.1016/j.jhydrol.2012.10.044.
100. Yadav, V.P.; Prasad, R.; Bala, R.; Srivastava, P.K. Synergy of vegetation and soil microwave scattering model for leaf area index
retrieval using C-band sentinel-1A satellite data. IEEE Geosci. Remote Sens. Lett. 2020, 19, 1–5.
101. Shulin, C.; Yuanbo, L.; Zuomin, W. Satellite retrieval of soil moisture: An overview. Anvances Earth Sci. 2012, 1192–1203.
https://doi.org/10.1109/LGRS.2020.3034420.
102. McGinnis, D.; Rango, A. Earth resources satellite systems for flood monitoring. Geophys. Res. Lett. 1975, 2, 132–135.
https://doi.org/10.1029/GL002i004p00132.
103. Giustarini, L.; Matgen, P.; Hostache, R.; Montanari, M.; Plaza, D.; Pauwels, V.R.N.; Pfister, L.; Hoffmann, L.; Savenije, H.H.G.
Assimilating SAR-derived water level data into a hydraulic model: A case study. Hydrol. Earth Syst. Sci. 2011, 15, 2349–2365.
https://doi.org/10.5194/hessd-15-2349-2011.
104. Brakenridge, R.; Anderson, E. Modis-based flood detection, mapping and measurement: The potential for operational hydro-
logical applications. In Nato Science Series: IV: Earth and Environmental Sciences; Springer: Berlin/Heidelberg, Germany, 2006; pp.
1–12. https://doi.org/10.1007/1-4020-4902-1_1.
105. Gao, W.; Shen, Q.; Zhou, Y.-H.; Li, X. Analysis of flood inundation in ungauged basins based on multi-source remote sensing
data. Environ. Monit. Assess 2018, 190, 129. https://doi.org/10.1007/s10661-018-6499-4.
106. Brakenridge, G.R.; Nghiem, S.V.; Anderson, E.; Mic, R. Orbital microwave measurement of river discharge and ice status. Water
Resour. Res. 2007, 43. https://doi.org/10.1029/2006WR005238.
107. Rees, W.G. Physical Principles of Remote Sensing; Cambridge University Press: Cambridge, UK, 2013; ISBN 978-1-107-00473-3.
108. Mason, D.C.; Giustarini, L.; Garcia-Pintado, J.; Cloke, H.L. Detection of flooded urban areas in high resolution synthetic aperture
radar images using double scattering. Int. J. Appl. Earth Obs. Geoinf. 2014, 28, 150–159. https://doi.org/10.1016/j.jag.2013.12.002.
109. Lin, Y.N.; Yun, S.-H.; Bhardwaj, A.; Hill, E.M. Urban flood detection with sentinel-1 multi-temporal Synthetic Aperture Radar
(SAR) observations in a Bayesian framework: A case study for Hurricane Matthew. Remote Sens. 2019, 11, 1778.
https://doi.org/10.3390/rs11151778.
110. Bhatt, C.; Thakur, P.K.; Singh, D.; Chauhan, P.; Pandey, A.; Roy, A. Application of active space-borne microwave remote sensing
in flood hazard management. In Geospatial Technologies for Land and Water Resources Management; Springer: Berlin/Heidelberg,
Germany, 2022; pp. 457–482.

Water 2023, 15, 622 23 of 24
111. Zhu, Z.J.; Jiang, A.Z.; Lai, J.; Xiang, Y.; Baird, B.; McBean, E. Towards efficient use of an unmanned aerial vehicle for urban flood
monitoring. J. Water Manag. Model. 2017, 26, C433. https://doi.org/10.14796/JWMM.C433.
112. Colomina, I.; Molina, P. Unmanned aerial systems for photogrammetry and remote sensing: A review. ISPRS J. Photogramm.
2014, 92, 79–97. https://doi.org/10.1016/j.isprsjprs.2014.02.013.
113. Feng, Q.; Liu, J.; Gong, J. Urban flood mapping based on unmanned aerial vehicle remote sensing and random forest classifier—
A case of Yuyao, China. Water 2015, 7, 1437–1455. https://doi.org/10.3390/w7041437.
114. Şerban, G.; Rus, I.; Vele, D.; Breţcan, P.; Alexe, M.; Petrea, D. Flood-prone area delimitation using UAV technology, in the areas
hard-to-reach for classic aircrafts: Case study in the North-East of Apuseni Mountains, Transylvania. Nat. Hazards 2016, 82,
1817–1832. https://doi.org/10.1007/s11069-016-2266-4.
115. Fohringer, J.; Dransch, D.; Kreibich, H.; Schröter, K. Social media as an information source for rapid flood inundation mapping.
Nat. Hazards Earth Syst. Sci. 2015, 15, 2725–2738. https://doi.org/10.5194/nhess-15-2725-2015.
116. Sui, D.; Elwood, S.; Goodchild, M. Crowdsourcing Geographic Knowledge: Volunteered Geographic Information (VGI) in Theory and
Practice; Springer: Berlin/Heidelberg, Germany, 2012; ISBN 978-94-007-4587-5.
117. Filonenko, A.; Wahyono; Hernández, D.C.; Seo, D.; Jo, K.-H. Real-time flood detection for video surveillance. In Proceedings of
the IECON 2015—41st Annual Conference of the IEEE Industrial Electronics Society, Yokohama, Japan, 9–12 November 2015;
pp. 004082–004085.
118. Li, J.; He, Z.; Plaza, J.; Li, S.-T.; Chen, J.-F.; Wu, H.; Wang, Y.; Liu, Y. Social media: New perspectives to improve remote sensing
for emergency response. Proc. IEEE 2017, 105, 1900–1912. https://doi.org/10.1109/JPROC.2017.2684460.
119. Wu, Z.-N.; Shen, Y.-X.; Wang, H.-L.; Wu, M.-M. Quantitative assessment of urban flood disaster vulnerability based on text
data: Case study in Zhengzhou. Water Supply 2020, 20, 408–415. https://doi.org/10.2166/ws.2019.171.
120. Ma, H.-D. Internet of things: Objectives and scientific challenges. J. Comput. Sci. Technol. 2011, 26, 919–924.
https://doi.org/10.1007/s11390-011-1189-5.
121. Poser, K.; Dransch, D. Volunteered geographic information for disaster management with application to rapid flood damage
estimation. Geomatica 2010, 64, 89–98. https://doi.org/10.1088/1755-1315/57/1/012015.
122. Dhaya R, Kanthavel R. IoT based urban flooding high definition surveillance using concurrent multipath wireless system. Earth
Sci Inform, 2022, 15(3): 1407-1416. https://doi.org/10.1007/s12145-022-00817-4.
123. Furht, B.; Villanustre, F. Introduction to big data. In Big Data Technologies and Applications; Furht, B., Villanustre, F., Eds.;
Springer: Cham, Switzerland, 2016; pp. 3–11, ISBN 978-3-319-44550-2.
124. Gericke, O.J.; du Plessis, J.A. Catchment parameter analysis in flood hydrology using GIS applications. J. South Afr. Inst. Civ.
Eng. 2012, 54, 15–26.
125. Magesh, N.S.; Chandrasekar, N.; Kaliraj, S. A GIS based automated extraction tool for the analysis of basin morphometry. Bon-
fring Int. J. Ind. Eng. Manag. Sci. 2012, 2, 4.
126. Anselin, L.; Getis, A. Spatial statistical analysis and geographic information systems. Ann. Reg. Sci. 1992, 26, 19–33.
https://doi.org/10.1007/BF01581478.
127. Jing, X.-Y.; Liu, Y.-F.; Wang, Y.-B.; Shen, Y.-Y.; Zhang, S.-F.; Tao, G.-H. Construction of the urban waterlogging monitoring and
forecasting system of Harbin City. J. Catastrophology 2009, 24, 54–57.
128. Wang, Q.-C.; Shou, S.-W.; Zhang, S.-H.; Huo, D.-S.; Zhou, Y.-D. A study on the early warning system for rainstorm Waterlog-
gingin Langfang City of Hebei Province. J. Arid. Meteorol. 2013, 31, 609–615. https://doi.org/10.11755 /j.issn.1006-7639(2013)-03-
0609.
129. 129. Chang, M. -J.; Chang, H.-K.; Chen, Y.-C.; Lin, G.-F.; Chen, P.-A.; Lai, J.-S. A Support vector machine forecasting model for
typhoon flood inundation map-ping and early flood warning systems. Water 2018, 10, 734. https://doi.org/ 10.3390/W10121734.
130. Van Ackere, S.; Verbeurgt, J.; De Sloover, L.; Gautama, S.; De Wulf, A.; De Maeyer, P. A review of the internet of floods: Near
real-time detection of a flood event and its impact. Water 2019, 11, 2275. https://doi.org/10.3390/w11112275.
131. Rong, Y.; Zhang, T.; Zheng, Y.; Hu, C.; Peng, L.; Feng, P. Three-dimensional urban flood inundation simulation based on digital
aerial photogrammetry. J. Hydrol. 2020, 584, 124308. https://doi.org/10.1016/j.jhydrol.2019.124308.
132. Jiménez-Jiménez, S.I.; Ojeda-Bustamante, W.; Ontiveros-Capurata, R.E.; de Marcial-Pablo, M.J. Rapid urban flood damage as-
sessment using high resolution remote sensing data and an object-based approach. Geomat. Nat. Haz. Risk 2020, 11, 906–927.
https://doi.org/10.1080/19475705.2020.1760360.
133. Dungan, J.L.; Perry, J.N.; Dale, M.R.T.; Legendre, P.; Citron-Pousty, S.; Fortin, M.-J.; Jakomulska, A.; Miriti, M.; Rosenberg, M.S.
A balanced view of scale in spatial statistical analysis. Ecography 2002, 25, 626–640. https://doi.org/10.1034/j.1600-
0587.2002.250510.x.
134. McGarigal, K.; Marks, B.J. FRAGSTATS: Spatial Pattern Analysis Program for Quantifying Landscape Structure.; U.S. Department
of Agriculture, Forest Service, Pacific Northwest Research Station: Portland, OR, USA, 1995; p. PNW-GTR-351.
135. Zhang, N.; Zhang, H. Scale variance analysis coupled with Moran’s I scalogram to identify hierarchy and characteristic scale.
Int. J. Geogr. Inf. Sci. 2011, 25, 1525–1543. https://doi.org/10.1080/13658816.2010.532134.
136. Zhang, N. Scale issues in ecology: Concepts of scale and scale analysis. Acta Ecol. Sin. 2006, 2340–2355.

Water 2023, 15, 622 24 of 24
137. Blöschl, G.; Sivapalan, M. Scale issues in hydrological modelling: A review. Hydrol. Process. 1995, 9, 251–290.
https://doi.org/10.1002/hyp.3360090305.
138. Zhang, N. Scale issues in ecology: Upscaling. Acta Ecol. Sin. 2007, 27(10), 4252–4266.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury
to people or property resulting from any ideas, methods, instructions or products referred to in the content.