Scientific African 13 (2021) e00887

Contents lists available at ScienceDirect

Scientific African
journal homepage: www.elsevier.com/locate/sciaf

Efficiency of camera sensors for flood monitoring and

warnings
Georgina Ufuoma a, Blessing Funmbi Sasanya b,∗, Perkins Abaje c,
Philip Awodutire a
a
Department of Mathematical Sciences, University of Africa, Toru Orua., Bayelsa State, Nigeria
b
Department of Crop and Soil Science, University of Port Harcourt, Rivers State, Nigeria
c
Department of Computer Engineering and Systems, University of Greenwich, Medway, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Adverse effects of flood on lives and properties has become an annual event in many de-
Received 8 February 2021 veloping countries. Flood events sometimes destroy agricultural lands and crops as well as
Revised 18 June 2021
public infrastructure and utilities. It is therefore of utmost importance to investigate the
Accepted 22 July 2021
efficiencies and performances of technologies which are capable of monitoring regular wa-
ter level changes. This study investigated the efficiency of Camera Sensor (CS) for water
Editor: DR B Gyampoh level monitoring at varying operating and image capturing distances of 1.0, 2.3 and 2.8 m
from the water bodies. Changes in water levels were captured at five (5) minutes time in-
Keywords:
tervals at varying tilting angles both at indoor and outdoor conditions. CS captured blur
Camera sensor distance
images at distance 1 m from the water body. Clear and visible images were however cap-
Water level monitoring
Flood mitigation tured when the CS was placed at distances 2.3 and 2.8 m from the experimental flume.
Data logger Distance 2.3 m was however recommended for CS operation; since it captured clearer and
Experimental flume more visible images. At distance 2.8 m, CS captured water level more accurately, since it
resulted in lower standard errors for all compared trials. CS’s angle of inclination has no
effect whatsoever on the quality of images captured. Water level monitoring by CS can
be adequately used for optimal flood monitoring, warnings and data gathering, because
changes in water level can be captured at chosen intervals. This will enable timely actions
for the mitigation of accompanying flood effects, especially in developing African countries
where most catchments are ungauged.
© 2021 The Authors. Published by Elsevier B.V. on behalf of African Institute of
Mathematical Sciences / Next Einstein Initiative.
This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)

Introduction

Water is very essential to all forms of life as it covers 71% of the Earth’s surface. However, despite water being vital for
the human existence, it could also result in dangerous disasters, such as tsunamis, flooding and typhoon; which can lead to
loss of lives and properties. Flooding is the overflow of water beyond normal limits, thus submerging land surfaces [1].

∗
Corresponding author.
E-mail addresses: [email protected] (B.F. Sasanya), [email protected] (P. Abaje).

https://doi.org/10.1016/j.sciaf.2021.e00887
2468-2276/© 2021 The Authors. Published by Elsevier B.V. on behalf of African Institute of Mathematical Sciences / Next Einstein Initiative. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

Floods of various origins, if not handled properly can be classed amongst the most dangerous natural disasters; because
of its effects on lives and properties [2]. It negatively impacts lives, agricultural productivity, economic sustainability, cul-
tural heritages, material possessions and infrastructural amenities in both rural and urban areas [3]. Flood disasters are
attributable to environmental imbalances caused by blocked drainages, deforestation, forest exploitation and intense rainfall
which leads to overflow of river banks. [4,5]
In Nigeria, flooding occurs in three main forms namely; river, urban and coastal flooding. Flooding in developing coun-
tries is attributable to long duration, high intensity rainfall coupled with poor drainage system networks which are unable to
sustain temporarily increased water volumes. Flood occurrences between September and October 2012 in Nigeria for exam-
ple were not only attributed to heavy rainfall, but also to the failure of the major Lagdo dam in the neighbouring Northern
Cameroon
Floods are experienced especially in the southern part of Nigeria, mostly during the rainy seasons between May and
October. These had led to serious flood events in most part of the country in past years. Some of the flood prone areas in
Nigeria include the Niger Delta regions along the coast. Aside the Niger-Delta regions, the northern and western parts of
the nation also experience flooding more often than not. The map on SM1 shows the Niger Delta states in Nigeria. Figure
SM 2 shows the map of Bayelsa state indicating the flood affected areas during the flood event of 2012. The encircled areas
on Fig. SM 2 experienced the highest flood impacts in 2012. The flood event in 2012 affected six of the eight districts in
Bayelsa state.
Flood extents caused in urban and rural areas can be predicted from and hydrologic and hydraulic models (with geo-
morphic analysis) by investigating the topography of the floodplain, channel and other boundary conditions [3,6]. Although,
flood events cannot be stopped but the accompanying adverse effects can be mitigated [5]. It is thus of utmost importance
to alert the people at higher risks and the public at large; so that the effects of these disasters can be reduced or eliminated
through germane precautionary measures. This can only be achieved through effective flood monitoring.
Water level monitoring carried out by individuals will not stop the effects of flooding, but generated data can be invalu-
able for designs for flood control structures and policy formulations. In lieu of these and for the purpose of curtailing flood
disasters by improving flood monitoring and readiness, major initiatives are being taken. Parts of such initiatives include
the development of secured and timely flood warning systems through the deployment of the sensor technology for regular
water level monitoring. These enables real-time flood monitoring as a major tool for flood data collection. The importance
of accurate real-time hydrologic variables for flood monitoring, forecasting and warning systems optimization for damage
mitigation is therefore of utmost importance [7]. An efficient flood warning system requires the use of automated systems
for data acquisition, key parameter data analysis and the issuing of real-time warnings for flood predictions [8].
Some previous researches had developed real time data collection systems in the past. Parts of such researches devel-
oped and tested suitable algorithm to perform near real-time assimilation of Synthetic Aperture Radar (SAR) - derived flood
water levels into a flood forecasting model by [6] The concepts of an efficient and standardized SAR based flood monitor-
ing involved the combination of radio metric thresholding and region growing approach which enabled automatic objective
and reliable flood extent extraction from SAR images [9]. According to [6], river gauges gives very little information on the
spatial variations in the flood levels, which is of importance in urban areas. Spatial information (spatially distributed water
levels) is however contained in flood extents images captured by SAR satellites. This was highlighted as the best source for
assimilation of data from hydraulic models. Water levels measurements by SAR is of high advantage because of its useful-
ness in ungauged catchments by employing the use of assimilation techniques such as Ensemble Kalman Filter (EnKF) which
assimilates water levels from a subset of points along a flood extent [6].
[10] made use of ultrasonic sensor for the detection of drain water level. This system sends warning notification to users’
phone based on three categories of water level which are safe, warning and critical level. [7] researched the design, develop-
ment and deployment of a real-time monitoring system for hydrological applications. This employed a wireless data logger
device (datalogVI) which combined the low consumption of motes and the reliable communication of most powerful multi-
sensor data logger stations; to satisfy the requirements of flood warning system services for water level data monitoring.
In the same vein, [11] proposed an intelligent sensor system which relies on 3D map reconstruction for the computation
of high precision real time data without human intervention. [12] researched the process of water levels and water flows
data collection and transmission from remote hydrological station locations to the data centre through the use of mobile
phone networks using GPRS technology. The research presented data collection for countries which include Cambodia, Laos,
Thailand and Vietnam. The study made use of internet-based systems for real-time monitoring and visualisation of water
levels and rainfall at the main stations; which comprised of flood warning systems made of five major layers.
However, none of these researches dealt with the accuracy of Camera Sensors (CS) for immediate water level monitoring
to make informative data on water level changes and movement available. There is a dearth of knowledge on appropriate
distances and angles which CS can be deployed and operated for water level monitoring in real time scenarios. This study
therefore aims to investigate the efficiency of camera sensor for water level monitoring at sequential distances of 1.0, 2.3
and 2.8 m from intended water body using 5 minutes’ capture time interval, during the day and also at night. The sensor
camera systems are aimed to be deployed for immediate water level data collection and monitoring. This will give adequate
knowledge on water level movement at every given point in time. This was achieved through laboratory studies using
experimental flumes, reflective stickers, plastic rulers, a tripod stand to adjust the angle of the camera sensor and a data
logger.

2
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

Materials and methods

Materials Description

The data quality and durability of the camera sensor (CS) (Fig. SM 3) were investigated in the hydraulics laboratory of the
University of Greenwich, Medway, United Kingdom. An artificial deep gully flume was used to simulate real environment in
conjunction with reflective marked measuring plastic ruler (which enables long distance and night visibilities), tripod stands
and a water level data logger labelled HOBO U20.

Experimental setup

The experimental flume (Fig. SM 4) used as the laboratory pilot scale stream made use of laminar and turbulent stream
flows at different time intervals; since stream flows can be unstable and unsteady. The flume has control buttons which
controls the rate of flow. Series of tests with turbulent flume flows were done to obtain data which assesses CS’s visibility
and water level capturing efficiency. Clear and turbid water were allowed to flow in the flume to test the efficiency of CS.
The flow was increased to 29.9 cm (measured from the ruler) and decreased at intervals. A measuring ruler laced with
reflective stickers was placed behind the flume. This facilitated the monitoring of water movement during increased and
decreased flow from 29.9cm. The tripod stands (on which the CS was mounted) was used to obtain the appropriate height
and most ideal angle for water level capturing. The pressure and temperature data of water in the flume were measured as
the water level increased from 0.00 cm to 29.9 cm using the data logger. An average of 8 minutes elapsed for the flume to
reach the maximum water level.
The CS was experimented at varying distances 1, 2.3 and 2.8 m from the experimental flume. CS efficiency at these
varying distances were tested and trials were repeated between 8 and 10 times, with averages of water levels obtained
for each test. Each trial carried out, lasted for 45 minutes with water level images captured at 5 minutes’ time intervals.
Different tilting angles (120 , 240 , 360 …1800 ) of CS were tested to determine the optimum for water level measurement.
The arrangement of the experimental set up is shown on Fig. SM 5. The data logger (Fig. SM 6) was firmly placed in
the flume for monitoring of the pressure and temperature of the stream. The HOBOware Pro Software was used for easy
conversion of the data; for the determination of accurate barometric pressure, temperature and water density.

Data analysis

Data obtained from water level in the experimental gully, captured by CS was subjected to statistical analysis using
descriptive statistics such as mean, standard deviation, range and standard error. Excel 2016 statistical package was used in
this respect.

Outdoor testing

Outdoor testing of the CS was also carried out using the same artificial watercourse (the deep gully flume). These facil-
itated 3G efficiency test on the CS system. One CS was placed in front of the artificial watercourse and another placed at
a height of 3.0 m from the experimental flume. The one directly placed in front of the artificial watercourse captured the
water level movement while the one placed at the height of 3.0 m away from the watercourse captured the surrounding
environment. Fig. 1 summaries the stepwise approach to the reported experiment.

Results and discussion

Features of camera sensors

Image camera sensors has relatively low procurement cost; portability; high efficiency, durability, low maintenance cost
and low power consumption (it can be powered by solar panels). Sensor networks are energy efficient [13]. The image
sensor also communicates with a 3G network module. The sensor has speed downloading rates of 7.2 Mbps and uploading
rates of 5.5 Mbps. Some other features of the CS are 22 high powers infrared (IR) LED’s to implement the night vision, a
fish eye lens and a light sensor. Images were captured by CS at intervals were achieved up to a tilted angle of 1800 .

Capturing angle of camera sensors

At an angle of 120 , CS was observed to give equal visibility and clarity with all other experimented angles. Other water
level measurements made without tilting the CS are equal to measurements made from tilted CS for all tested distances;
when clear water was in the flume. It was however nearly impossible to capture a clear image of turbid water level. This
was however made possible when the top level of the reflective metre rule was visibly captured by the camera. This im-
plies stages for water level measurement should be long enough, when CS is deployed to capture flood water level in real
situations.

3
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

Fig. 1. Stepwise Summary of Methodology.

Water level measurement by camera Sensor

Captured water levels from CS distance 2.3 m from the flume at 5 minutes’ time intervals (repeated eight (8) times) are
shown on Fig. 2. The water level captured within 45 minutes has range 5.62 cm, mean of 4.46 cm and Standard Deviation
(SD) of 1.18 cm. In the same vein, the captured water levels within a period of 45 minutes (at 5 minutes’ time interval) has
a range 12.6 cm, mean 14.08 cm and SD 4.49 cm at CS distance 2.8 m (Fig. 3).
These Figs. show the rates at which water level varies within the flume and it gives credence to the efficiency of CS
for effective monitoring of water level changes at regular time intervals. The recorded water levels at CS distances 2.3 m
(Fig. 2) and 2.8 m (Fig. 3) were observed to have closely related change pattern. These gave credence to the fact that CS
performed quite satisfactorily in water level monitoring. This shows both distances are adequate for the operation of such
cameras. However, [14] recommended the operation of ultrasonic sensors within the range of 3 m. Accuracies of measure-
ment were claimed to be affected beyond such distance. This was attributed to degradation caused by noise in free space
loss. [15] as well recommended placing image sensors at distances away from water body being monitored; as this will
encourage adequate field of view and will also protect the sensor from flood disasters.
Water levels captured at CS distance 1 m from the experimental gully flume blurred. The distance 1 m is therefore not
appropriate for the operation of the CS in water level capturing. Samples of images captured by CS are shown on Fig. SM 7
through SM 10. Figures SM 7 and SM 8 show the images captured by CS at distance 2.3 m during the day and night time
respectively. Figures SM 9 and SM 10 respectively captures water levels at distance 2.8 m during the day and night time.
From these Figs., captured images are obviously bolder and clearer at shorter distance 2.3 m than those captured at 2.8 m
CS distance; however, images capture at 2.8 m seems more accurate. Images captured during the day time obviously have
better visibility than captures made during night time. The night time images at distance 2.3 m are however bolder, clearer
and more visible. Visual sensing approach is therefore more reliable for monitoring of fluctuation in water level, elevation
measurement and flood invasion using real world coordinates [15,16].

4
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

Fig. 2. Water level measurement at 5 minutes’ interval and CS distance 2.3 m.

Fig. 3. Water level measurement at time interval of 5 minutes and CS distance of 2.8 m.

Mean water level data

The averages of water levels measured for the trials at different CS distances 2.3 m and 2.8 m are presented on Figs. 4
and 5 respectively. From Fig. 4, the first trial averaged 15.8 cm, while the last trial resulted in an average of 15.1 cm. The
range of the averages is 7.62 cm and the Standard Error (SE) is 0.84 cm. Similarly, the first and last trials for CS distance of
2.8 cm from the water body has mean value 13 cm and 14.1 cm respectively. The range and SE are respectively 4.9 cm and
0.71 cm (Fig. 5).
Fig. 6 compares the variations of recorded water level averages at CS distances 2.3 and 2.8 m from the water body. The
blue bars represent the mean water levels at CS distance 2.3 m while the orange bars represent mean water levels at CS
distance 2.8 m. The ranges and SEs for 2.3 m and 2.8 m CS distances are 3.5 cm and 0.40 cm as well as 2.7 cm and 0.27
cm respectively. These values show water levels measured at both CS distances do not have much difference. However, the
water level movements captured at CS distance 2.3 m are higher than those captured at CS distance 2.8 m. In real-time
scenario, this implies that smaller areas have higher probability of being flooded than larger areas for the same level of
water flows. Flood warnings could thus be sent to high risks prone areas.

5
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

Fig. 4. Averages of Water Levels Measured at CS distance 2.3.

Fig. 5. Averages of Water Level Measured at CS distance 2.8 m.

According to [17], traditional flood monitoring devices can be adversely affected by weather changes and this could lead
to avoidable errors. CS can therefore efficiently and adequately replace the traditional water level and flood monitoring
methods, since it can be free from human and operational errors. Unavailability of individuals to take needed water level
records cannot hamper the operation of CS from time to time.

Outdoor water level measurement

At outdoor conditions, CS placed before the watercourse and that at height 3.0 m above the water course captured
clearly visible images of both clear / turbid waters and the surrounding environment respectively. The images captured at
outdoor operations are shown on Figs. SM 11 and SM 12. Similar to indoor conditions, captured images at CS distance 2.3
m are bolder, clearer and more visible than images captured from CS distance 2.8 m. The data was transmitted remotely
using the available 3G network in the area. The system’s 3G functionality to remotely transfer obtained data is efficient and
effective. The 3G data transmission viability of CS can be harnessed to inform residents in flood prone areas of imminent

6
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

14
12

WATER LEVELS (CM)

10
8
6
4
2
0
1 2 3 4 5 6 7 8
TRIALS
Fig. 6. Comparison of Recorded Water Levels for different trials at CS Distances 2.3 m and 2.8 m.

danger. These can be achieved in conjunction with operators in telecommunication industries by sending alerts to mobile
data transmission enabled telephone devices.
These results show that CS can be efficiently deployed for flood monitoring especially in developing African countries
such as Nigeria where several catchments are ungauged and flood events are experienced annually. Advantages such as rel-
atively low procurement cost; portability; high efficiency, durability, low cost of maintenance and low power consumption
makes CS viable for water level monitoring for flood warnings in developing African countries. Applications of these tech-
nologies are capable of providing real time overview, so that needed and timely actions can be taken for flood protection in
places where the technology are deployed [18,19]. CS is capable of giving immediate water level data which gives adequate
knowledge of water level changes at every given point in time. This system is also capable of monitoring culverts blockages
which result is head losses and instantaneous water level rise. The use of CS for water level measurement is a very reliable
method of water level monitoring [16,18].
This study’s importance is due to severe occurrences of floods in southern Nigeria; a developing country which suf-
fers from different urban and coastal flooding types in the southern lagoon area. These sterns from facts that the present
drainage system networks are incapable of handling temporarily increased water volumes. Although flood extents in urban
and rural areas can be predicted from hydrologic and hydraulic models in this parts of Nigeria, like other places, but flood
events cannot be stopped, but the accompanying negative effects can be largely mitigated [3,5,6] by adequate monitoring,
early detection and warning alerts. The utilization of the sensor based technology for water level data capturing can thus
play significant roles in flood readiness, warnings and predictions in developing countries.

Conclusion

The utilization of the sensor based technology for water level data capturing plays significant roles in flood readiness,
warnings and predictions since CS can capture water level images at regular time intervals. Advantages such as low power
and energy consumption as well as clarity of captured images makes CS viable and recommended as an alternative technol-
ogy for flood monitoring, which could adequately replace the traditional system of water level monitoring. This would give
room for adequate and robust water level data collection which can aid prior warnings on flood events.
Having tested the efficiency of CS and its accuracy in taking constant images of water levels at regular intervals, it
was observed that installation distances is one of the several important factors affecting the optimal efficiency of CS. The
visibility of images captured at all tested angles are the same. Angle of inclination of the CS are therefore observed not to
have any effect on the clarity of captured images. Water level images captured at CS distance of 1 m from the artificial gully
are very blur while those captured at 2.3 and 2.8 m CS distance are clear and visible. Comparison between images captured
by CS at distances 2.3 and 2.8 m from the water body however makes the recommendation of CS distance 2.3 m viable and
optimal; since images captured at this distance are clearer, bolder and more visible both during the day and also at night.
Even though water level images captured at CS distance 2.8 m seems more accurate, since they have lower SE, the clearer
visibility of CS images captured at 2.3 m makes it optimal and recommended for the operation of CS at all conditions. The
same performances of the CS were observed both at the indoor and other operating conditions. In summary, the results
from this research affirmed CS should be placed at distance 2.3 m from water bodies being monitored, so that adequate
water levels data can be obtained for all flow conditions. CS can thus be set to record water levels at required time intervals
and distance 2.3 m from water bodies, for generation of reliable data. Adequate and informed plans to mitigate the effects
of flood occurrences through individual actions, government policies and stakeholders’ interventions can thus be made.

7
G. Ufuoma, B.F. Sasanya, P. Abaje et al. Scientific African 13 (2021) e00887

This research was carried out in the laboratory as a pilot scale experiment. Further researches on the performances and
efficiency of CS and data logger in real life situations (at outdoor conditions and at the peak of rainy season) is recom-
mended. This will facilitate the determination of CS durability under harsh operating conditions.

Declaration of Competing Interest

The authors declare no conflict of interests on this research and the accompanying write up.

Funding statement

The authors acknowledge the Bayelsa State Government, Nigeria for funding this research and sponsoring the trip to
University of Greenwich. The efforts of the members of staff in the hydraulics Laboratory at the University of Greenwich are
as well appreciated.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.sciaf.2021.e00887.

References

[1] K. Auynirundronkool, N. Chen, C. Peng, Flood detection and mapping of the Thailand Central plain using RADARSAT and MODIS under a sensor web
environment, Int. J. Appl. Earth Obs. Geoinf. 14 (1) (2012) 245–255.
[2] R. Sulistyowati, H.A. Sujono, A.K. Musthofa, A river water level monitoring system using android-based wireless sensor networks for a flood early
warning system, in: Lecture Notes in Electrical Engineering, vol. 365, 2016.
[3] C.C. Abon, C. Primo, C. David, G.Q. Tabios, Community-based monitoring for flood early warning system, Disaster Prev. Manag. An Int. J. 21 (1) (2012),
doi:10.1108/09653561211202728.
[4] D. Satria, S. Yana, R. Munadi, and S. Syahreza, ‘Design of information monitoring system flood based internet of things (IoT)’, (2018).
[5] S.I. Abdullahi, M.H. Habaebi, T.S. Gunawan, M.D. Rafiqul Islam, Miniaturized water flow and level monitoring system for flood disaster early warning,
IOP Conference Series: Materials Science and Engineering, 2017 vol. 260, no. 1„ doi:10.1088/1757-899X/260/1/012019.
[6] D. Mason, G. Schuman, J. Neal, J.B.P. Garcia-Pintado, Automatic near real-time selection of flood water levels from high resolution synthetic aperture
radar images for assimilation into hydraulic models: a case study, Remote Sens. Environ. 124 (2012) 705–716.
[7] R. Marin-Perez, J. Garcia-Pintado, A.S. Gomez, A real-time measurement system for long-life flood monitoring and warning applications, Open Access
Sens. 12 (2012) 4213–4236.
[8] H. Hassan, M.I.Q. Mazlan, T.N.T. Ibrahim, M.F. Kambas, IOT system: water level monitoring for flood management, IOP Conference Series: Materials
Science and Engineering, 2020 vol. 917, no. 1„ doi:10.1088/1757-899X/917/1/012037.
[9] P. Matgen, R. Hostache, G. Schumann, L. Pfister, L. Hoffmann, H.H.G. Savenije, Towards an automated SAR-based flood monitoring system: lessons
learned from two case studies, Phys. Chem. Earth, Parts A/B/C 36 (7-8) (2011) 241–252.
[10] M.I. Hadi, et al., Designing early warning flood detection and monitoring system via IoT, IOP Conference Series: Earth and Environmental Science, 2020
vol. 479, no. 1, doi:, doi:10.1088/1755-1315/479/1/012016.
[11] M. Ancona, et al., On the design of an intelligent sensor network for flash flood monitoring, diagnosis and management in urban areas position paper,
Proc. Comput. Sci. 32 (2014), doi:10.1016/j.procs.2014.05.515.
[12] S. Keoduangsine, R. Goodwin, P. Gardner-Stephen, A study of an SMS-based flood warning system for flood risk areas in Laos, Int. J. Futur. Comput.
Commun. 3 (3) (2014), doi:10.7763/ijfcc.2014.v3.292.
[13] O.T. Arulogun, A.A. Adigun, O.O. Okediran, R.A. Ganiyu, Design and development of a security surveillance system based on wireless sensor network,
Int. J. Innov. Sci., Eng. Technol. 1 (4) (2014) 284–291.
[14] S.L. Mohammed, A. Al-Naji, M.M. Farjo, J. Chahl, Highly accurate water level measurement system using a microcontroller and an ultrasonic sensor,
IOP Conference Series: Materials Science and Engineering, 2019 vol. 518, no. 4„ doi:10.1088/1757-899X/518/4/042025.
[15] S.W. Lo, J.H. Wu, F.P. Lin, C.H. Hsu, Visual sensing for urban flood monitoring, Sensors (Switzerland) 15 (8) (2015), doi:10.3390/s150820 0 06.
[16] I. Hasan, M. Satzger, M. Sattler, R. Duester, T. Hies, E. Jose, An effective camera based water level recording technology for flood monitoring, INTER-
PRAEVENT 2016 - Conf. Proc. 1 (2016).
[17] G. Tang, Y. Zhu, G. Wu, J. Li, Z.L. Li, J. Sun, Modelling and analysis of hydrodynamics and water quality for rivers in the northern cold Region of China,
Int. J. Environ. Res. Public Health 13 (4) (2016) 1–15, doi:10.3390/ijerph13040408.
[18] I. Hasan, M. Satzger, M. Sattler, R. Duester, T. Hies, E. Jose, An effective camera based water level recording technology for flood monitoring, in:
INTERPRAEVENT 2016 - Conference Proceedings, 2016 vol. 1.
[19] M.M. Bhosale, Review on flood monitoring and early warning system, Int. J. Res. Appl. Sci. Eng. Technol. 7 (1) (2018), doi:10.22214/ijraset.2019.1079.