https://doi.org/10.31449/inf.v47i9.4890 Informatica 47 (2023) 109–122 109
Integrated Streamflow Forecasting System: A Step Towards Smart
Flood Management
Irfan Ardiansah
1*
, Alfonsus Mario Agung
2
, Chay Asdak
2
, Nurpilihan Bafdal
2
, Roni Kastaman
1
, Selly Harnesa Putri
1
and Desy Nurliasari Suparno
1
1
Department of Agro-Industrial Technology, Faculty of Agro-Industrial Technology, Universitas Padjadjaran,
Indonesia
2
Department of Agriculture Engineering and Biosystem, Faculty of Agro-Industrial Technology, Universitas
Padjadjaran, Indonesia
e-mail: irfan@unpad.ac.id
*
Coresponding author
Keywords: streamflow forecasting, regression model, precipitation data, water resource management, flood mitigation
Received: May 26, 2023
This study aims to help users manage water resources and prevent flooding by creating an online monthly
streamflow forecasting system. We have integrated a regression model into the system, using historical
information on rainfall and streamflow selectivity from a number of monitoring stations in the Upper
Cimanuk sub-basin. Users can access the online system to input and view rainfall and streamflow data
and enumerate monthly streamflow rate projections. To verify the system's forecast accuracy, we
compared it with manual calculations employing the velocity-area method and field observations. The
system provides reasonably accurate forecasts, as indicated by the system's high coefficient of
determination (R
2
) value of 0.91. Nevertheless, the differences between predictions and measurements
suggest there is scope to improve the accuracy of the system by including additional variables and more
comprehensive data. Future enhancements may include additional validation using a wider range of field
data, as well as the inclusion of precipitation intensity, duration, catchment shape and size. The developed
monthly streamflow forecasting system is a valuable tool for analyzing and forecasting streamflow rates,
providing a basis for informed decision making in water resource management and flood disaster
mitigation.
Povzetek: V članku je predstavljen spletni sistem za napovedovanje mesečnega pretoka rek (opozorilo za
poplave) v Cimanuku z visoko natančnostjo (R
2
= 0.91).
1 Introduction
[1] defines flooding as the inundation of an area that
occurs when water overflows beyond its drainage
capacity, causing physical, social, and economic losses.
Flooding happens when the river or canal overflows on
either its right or left side because the channel capacity
cannot manage the streamflow. Flooding occurs due to
overflowing to the left or right side of the river/canal
because the capacity of the river channel is not sufficient
for the streamflow [2]. The complexity of flooding in a
basin involves numerous main elements, which function
both as natural physical objects or targets, and as subjects
or actors utilized by humans. The elements interact and
mutually influence each other, leading to the hydrological
condition of the basin [3].
The Watershed Management Board of Cimanuk –
Citanduy recorded flood and landslide report data, noting
more than one flood event in the Upper Cimanuk Sub-
basin. Flooding struck Sindangsari Village, Garut
Regency on March 28th, 2014, with runoff reaching 165
cm. In 2015, floods took place between March 15 and 16
with a streamflow of 384 m
3
/s. A flash flood hit Sukakarya
Village, Tarogong Kidul Subdistrict, Garut Regency, on
September 20, 2016, with a runoff of 50 - 200 cm,
claiming three lives [4]. In terms of administration, two
governmental regions divide the Upper Cimanuk sub-
basin: Garut Regency, which covers the majority with 20
sub-districts, and Sumedang Regency, which covers the
remaining 12 sub-basins with an area of 156,020 hectares
[5].
Streamflow forecasting is used to try to anticipate
flooding. Basin rivers take longer to observe precipitation
observations than streamflow observations [6]. Studying
the relationship between the two variables is important.
We can express the relationship of hydrological variables
in mathematical formulas, which can then be used for
hydrological analysis, such as forecasting, extension,
repair, and data filling [7].
In linear regression analysis, we examine the
relationship between two or more variables. We use a
linear regression model to establish the relationship
between dependent and independent variables. Linear
regression comes in two types: simple and multiple,
depending on the number of independent variables.
Multiple linear regression involves more than one
independent variable, while simple linear regression
involves only one [8]. In this study, we use simple linear
110 Informatica 47 (2023) 109–122 I. Ardiansah et al.
regression, treating precipitation as the independent
variable.
Considering the rise in internet users in Indonesia and
the recurring floods in the Upper Cimanuk sub-basin, we
plan to develop a monthly online streamflow forecasting
system. This system will leverage widespread internet
access to share crucial information about potential floods,
thereby minimizing their adverse effects on local
communities. According to the Indonesian Internet
Service Providers Association survey, Indonesia has 204.7
million active internet users, making up 74.23% of the
total population. Java Island hosts the highest number of
internet users, comprising around 43.92% of Indonesian
users [9].
The relationship between precipitation and river flow
is unidirectional in both the dry and wet seasons; as
precipitation increases, so does river flow, and vice versa.
We plan to build an online monthly streamflow
forecasting system, considering the growing number of
Internet users and the flooding in the Upper Cimanuk
subbasin in Indonesia. We will implement a linear
regression model to forecast streamflow, using
precipitation as the independent variable and streamflow
as the dependent variable. The system will analyze both
variables to predict monthly streamflow. Given the
substantial number of internet users in Indonesia, we
believe this online system will be accessible to everyone
who needs it and will help reduce flooding.
The purpose of the study is to develop a web-based
monthly streamflow forecasting system for the Upper
Cimanuk sub-basin that users can access online to analyze
and forecast streamflow. The system calculates
streamflow forecasts using rainfall and streamflow
information and compares the precision of its forecasts to
manual calculations. The purpose of this study is to
provide a tool for analyzing and predicting streamflow to
serve as a foundation for decision-making in water
resource management and flood disaster prevention. This
research includes only the Upper Cimanuk sub-basin and
focuses on developing and evaluating an online
application that offers users access to precipitation and
streamflow data and allows them to forecast streamflow
based on the provided rainfall input.
The findings of this study will grant convenient access
to information on streamflow and precipitation for users
in the Upper Cimanuk sub-watershed, benefiting planning
and growth activities in the area.
2 Related works
In recent years, there has been some interest in the field of
water resource management and flood disaster mitigation.
By integrating advanced modeling techniques and data-
driven approaches, these systems have paved the way for
flood management strategies. Streamflow forecasting
stands as water resource management and flood disaster
mitigation tool. An array of methodologies has been
cultivated for streamflow forecasting, spanning statistical
methods, hydrological models, and machine learning
techniques. This chapter offers an overview of researches
within the integrated streamflow forecasting systems
domain.
The exploration into streamflow prediction has led to
a series of investigations. Regression models have been a
prominent focus, revealing their potential to forecast
streamflow dynamics. [10] study investigates the marginal
advantage of a different methods using initial hydrologic
conditions (IHC), focusing on seasonal water supply
forecasts (WSF) with case studies on five watersheds
located in the US Pacific Northwest region. The
researchers found that climate information can increase
the reliability of forecasts from IHC, but strict control over
sample size must be observed to avoid overtrained
forecast solutions.
The rise of online platforms has significantly
impacted water resource management. [11] study
introduces a serious gaming framework to assist
stakeholders in the decision-making process for water
resources scheming and disaster mitigation. The
framework includes a Multi-Hazard Tournament (MHT)
and a web-based decision support tool. The framework
was evaluated in a case study and found to be effective in
increasing collective understanding and awareness of
water-related hazards and mitigation strategies.
The validation of forecasting systems plays a crucial
role in establishing their credibility. [12] introduced a
machine learning model that can be used to predict
drought events in the eastern Mediterranean. The bagging
algorithm was the most accurate in the training stage, but
the bagging and random forest algorithms were more
dynamic in drought capturing. The results of the research
can help decision-makers with drought mitigation plans.
Spatial and temporal factors emerge as pivotal
determinants of forecast accuracy. [13] developed a
distributed hydro-meteorological forecasting approach to
provide information at unexplored sites. The system was
validated with respect to actual road inundations and the
results are promising. The system could be used to identify
areas at risk and adopt appropriate safety and rescue
measures.
The incorporation of additional variables has emerged
as a promising avenue. [14] built a method for
disaggregating daily rainfall observations into hourly
rainfall. The method was applied in Singapore and was
found to produce intensity–duration–frequency curves
with significantly improved accuracy.
One contemporary example of a forecast system
model is presented by [15]. By employing a cross-
disciplinary collaboration between life scientists and
expert users of Earth system models will greatly enhance
the likelihood of developing robust evidence to address
climate change challenges. This will make caveats more
explicit and place decisions regarding potential tradeoffs
in the hands of the user.
Another notable example of how the use of big data
and machine learning technologies has the potential to
impact many facets of environmental and water
management comes from the following research by [16].
They found that big data and machine learning have the
potential and benefits to enable data-driven research in
environmental and water management, provide an
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 111
overview of key concepts and approaches in big data and
machine learning, and discuss key issues and challenges.
Table 1 provides a summarized overview of key
findings from various literature reviews concerning
streamflow forecasting systems, encompassing those
utilizing regression models and online platforms. The
compilation underscores that while these state-of-the-art
methodologies offer valuable insights, they often exhibit
limitations in terms of accuracy and suitability for specific
hydrological contexts.
The significance of our work becomes apparent
within the context of streamflow forecasting, owing to the
following attributes:
• Integrated Approach: Our study adopts an integrated
methodology, amalgamating multiple techniques to
enhance accuracy and resilience.
• Online Accessibility: Implementation through an
online platform expands accessibility to a diverse user
base.
• Real-world Validation: Evaluation using historical
data from a practical watershed affirms the
practicality of our approach.
The outcomes of this endeavor suggest that our
proposed system holds potential to furnish precise and
trustworthy streamflow forecasts for an array of
applications. Its implications extend to refining water
resource management, mitigating flood disasters, and
safeguarding lives and assets against inundation risks.
3 Research methodology
In this study we used (1) 10 years of precipitation data as
the source of the database, collected from the Watershed
Management Office West Java Region, (2) 10 years of
streamflow data from the Watershed Management Office
West Java Region to form the basis for streamflow
forecasting, (3) a 1:50,000 scale map of the Upper
Cimanuk Sub-basin, obtained from the Cimanuk -
Table 1: Summary of: Related Works
Author Results Advantages Disadvantages
[10] Investigated seasonal water supply
forecasts in five US Pacific Northwest
watersheds. Found that climate information
enhances forecast skill but cautioned
against over-trained solutions.
Climate info improves seasonal
forecasts
Insights into
handling sample size limitations
Limited to specific
regions Risk of over-
training
[11] Introduced a serious gaming framework for
water resource planning and hazard
mitigation. Evaluated effectiveness in a
case study, enhancing awareness and
understanding of hazards and mitigation
strategies.
Engages stakeholders through
serious gaming
Enhances
collective awareness
Specific to decision-
making contexts, May
require technological
infrastructure
[12] Introduced a machine learning model for
drought prediction in the eastern
Mediterranean. Bagging and random forest
algorithms were dynamic in drought
capturing.
Machine learning for dynamic
drought prediction
Insights
for mitigation planning
Focus on drought
prediction, Algorithm
complexity
[13] Developed a distributed hydro-
meteorological forecasting approach to
identify ungauged sites at risk of road
inundation. Promising results for safety
and rescue measures.
Provides info for ungauged
sites
Identifies at-risk areas
for inundation
Specific to road
inundation, May require
data infrastructure
[14] Built a method to disaggregate daily
rainfall into hourly observations. Applied
in Singapore, producing improved
accuracy in intensity–duration–frequency
curves.
Improved accuracy in rainfall
curves
Valuable for
hydrological modeling
Focus on rainfall
disaggregation, Regional
applicability
[15] Emphasized cross-disciplinary
collaboration for robust climate change
evidence. Advocated for informed
decision-making by users through explicit
caveats and tradeoff considerations.
Integrates Earth system models
and expert users, Enhances user
decision-making
Reliant on
interdisciplinary
collaboration, Potential
for complex
communication
[16] Highlighted big data and machine
learning's potential in environmental and
water management research. Discussed
concepts, approaches, benefits, and
challenges.
Enabling data-driven research,
Overview of key concepts and
approaches
General overview, Lack
of specific applications
112 Informatica 47 (2023) 109–122 I. Ardiansah et al.
Citanduy Watershed Management Board to create
Thiessen polygons, and (4) Flood and Landslide Reports
from the Cimanuk - Citanduy Watershed Management
Board to determine the minimum streamflow required
during flood events.
The deliberate selection of the Upper Cimanuk sub-
basin as the focal point of our study is a result of careful
deliberation guided by multifaceted considerations.
Acknowledging the sub-basin's proclivity towards flood
occurrences due to its heterogeneous topography, land use
patterns, and hydrological attributes, our deliberate choice
is anchored in the intention to confront the tangible
challenges entwined with flood management and
optimization of water resources. This regional spotlight
provides an impeccable crucible for evaluating the
efficacy of our integrated streamflow forecasting system
amidst the complexities and dynamism of the
environment.
The Upper Cimanuk sub-basin offers a manifold of
advantages stemming from its hydrological diversity,
rendering it an invaluable paradigm for comprehending
the dynamics of our system's performance across an array
of terrains and land utilization. Simultaneously, the
region's susceptibility to floods accentuates the criticality
of precise streamflow forecasts as a bedrock for potent
flood disaster mitigation strategies.
Our methodology involved the comprehensive
development and evaluation of a monthly streamflow
forecasting system tailored specifically for the Upper
Cimanuk sub-basin. At the heart of our Integrated
Streamflow Forecasting System lies the intricate
framework of the regression model. Our methodology
seamlessly integrates a straightforward linear regression
model, meticulously combining historical rainfall and
streamflow data gathered from a diverse array of
monitoring stations scattered throughout the expansive
Upper Cimanuk sub-basin. This particular model is
thoughtfully crafted to illuminate the complex interplay
between rainfall patterns and the resulting streamflow
rates unique to this region.
To establish the relationship between rainfall and flow
data, we incorporated a simple linear regression model
into our system design process. This model helps to
understand and describe the relationship between rainfall
and river flow dynamics. To make things easy and
accessible for users, we've developed this application
using web-based technology. This lets you connect to the
system from anywhere by simply using the internet.
To ensure the system is functioning accurately, we
execute manual computations employing the area velocity
technique as a standard for precision in forecasting the
river's flow. The velocity-area approach is a customary
means of approximating the velocity of the river's current
by gauging the dimensions of the river's cross-section and
determining the speed of the water at various locations.
We subdivided the length of the river into ten segments
and gauged the swiftness of the water at particular depths
(at 20% of the complete depth and at 80% of the entire
depth) [17]. This technique depends on the principles of
fluid dynamics to provide us with an estimate of the speed
at which the fluid is traveling.
The velocity-area approach is a frequently employed
and comparatively straightforward method to employ.
Nevertheless, its precision can be affected by numerous
factors such as disparities in flow velocity at distinct
segments of the river, turbulence, and errors committed by
individuals. To guarantee the tool functions correctly in
measuring river flow, we perform a form of testing known
as black box testing. This approach aids us in verifying
that every facet of the tool operates precisely and in
accordance with user requirements [18]. We employ black
box testing to assess: user registration pages, sign-in
pages, user-submitted flow forecast pages, and flow
prediction pages.
Application of the approach entails statistical
analysis, software development, and system assessment
through juxtaposition with manual computations, field
inspection tasks, and black box testing. The objective is to
furnish a user-friendly monthly river flow prediction
system for planning and development endeavors in the
Cimanuk Hulu sub-basin.
3.1 Data collection
In this study, we utilized a quantitative research method.
By obtaining secondary data from the Cimanuk - Citanduy
Watershed Management Board, the Upper Cimanuk Sub
Basin map, and a Flood and Landslide Report, which
includes 10 years of rainfall data and water flow data.
3.2 Determination of simple linear
regression equation
The coefficient of determination of a simple linear
regression equation is used to study the relationship
between two variables, the independent variable (X) and
the dependent variable (y). This allows researchers to
develop a linear mathematical model to predict the value
of the dependent variable, Y, based on the value of the
independent variable, X [8], [19].
We are using rainfall as the independent variable (X),
while flow is representing the dependent variable (Y). We
develop the model to investigate the relationship between
rainfall and runoff and predict runoff [20].
The steps to determine the simple linear regression
equation include:
1. Obtaining precipitation (X) and streamflow (y) data
from relevant sources, such as the West Java
Provincial Watershed Management Office.
2. Calculating mean values for the independent variable
(X) and dependent variable (y).
3. Measuring the correlation coefficient (R) between the
independent variable (X) and dependent variable (y)
to determine the strength of the linear relationship
between the two variables.
We calculate regression coefficients (a and b) using a
predetermined formula. Coefficient a is a constant, while
coefficient b is the slope of the regression line [19].
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 113
a = R (
σ
𝑦 σ
𝑥 )
b = R (
σ
𝑥 σ
𝑦 )
We can construct a simple linear regression equation
with the calculated regression coefficients (a and b) in the
form of:
y = a + bX
By evaluating the fit of the regression model, we can
calculate the coefficient of determination (R
2
). The value
of R
2
ranges from 0 to 1, where a higher value indicates a
better model in explaining the variation in the data. After
forming the simple linear regression equation, we validate
and interpret the resulting model. We will use the
formulated regression model to forecast streamflow based
on observed precipitation and aid in water resources
planning and management in the Upper Cimanuk sub-
basin region.
3.3 Determining the thiessen polygon value
We use the Polygon Thiessen technique in spatial analysis
to estimate variables in a region based on values observed
at specific measurement points [21]. In this research, we
employ Thiessen polygons to ascertain the scope of the
Upper Cimanuk sub-basin by considering the location of
weather stations in the region.
We split the Cimanuk Hulu sub-basin into sections
based on the position of the weather stations using ArcGIS
software. This process entails developing Thiessen
polygons, where each weather station point becomes the
nucleus of the polygon. Every Thiessen polygon
represents a region nearest to its core weather station.
After Thiessen polygons are generated, we calculate
each polygon area to determine the monthly precipitation
at the closest weather station to the streamflow station
[13]. As a result, past data from weather stations can be
used to estimate precipitation and flow at different points
in the Upper Cimanuk catchment area.
The Thiessen polygon method is critical for regional
water resource management, strategy development and
decision making, as it provides accurate precipitation and
runoff information on the sub-basin distribution [21], [22].
3.4 Overview of system design
The main system architecture objective is to enable people
in the Upper Cimanuk sub-basin to evaluate rainfall and
streamflow data for flood forecasting and management
purposes. Users must either log in (if they already have an
account) or register (if they do not) to access the database.
By visiting the database page, users can search for data
and observe how it has changed over time. They can also
calculate streamflow by selecting that option from the
streamflow calculation menu, which aids in planning and
determining flood management and water resource
administration in the region.
Users can access the streamflow calculation menu
page without logging in. This page offers a tool that
calculates streamflow based on user-provided inputs, such
as precipitation and station information, and generates
estimated streamflow to assist in planning and decision-
making. Figure 1 displays the system's site map.
Therefore, the system provides easy and quick access
to relevant precipitation and streamflow information and a
Figure 1: Site map of streamflow forecast system
(a)
(b)
Figure 2: (a) Data context diagram (b) Data flow
diagram of streamflow forecast system
114 Informatica 47 (2023) 109–122 I. Ardiansah et al.
useful streamflow calculation tool for users like students,
lecturers, researchers, and practitioners in the field of
water resources management.
4 Proposed system
The system forecasts streamflow, particularly in the Upper
Cimanuk sub-basin. Users can access the database only
after logging in. As for streamflow forecasting, users do
not need to log in. The user's experience with the system
starts with the Data Context Diagram (DCD), shown in
Figure 2a. The DCD visually represents the framework
and primary components of the monthly streamflow
forecasting system. It shows the interaction between users
and the system, as well as the underlying processes that
occur when the system processes user input [23]. The
DCD helps describe the flow of information and
interactions between the user and the system, and provides
an overview of how the monthly streamflow forecasting
system works.
Figure 2b shows the Data Flow Diagram (DFD) of the
system development. The DFD models the flow of data
through a system or process, as well as how the data is
processed and stored. The DFD is more detailed than the
DCD in describing the system flow process. In the
monthly streamflow forecasting system, users must log in
to access the database. The system matches the username
and password entered by the user with the registered data
in the user database. A successful login provides users
with streamflow and precipitation data according to the
desired annual data of the selected station. The DFD helps
acknowledge how data flows through the system and how
the system produces the desired output. It also makes it
easy to point out and enhance the processes that occur in
the monthly streamflow forecasting system [24].
Figure 3: Home page of online streamflow forecast system
Figure 4: Entity relationship diagram of streamflow forecast system
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 115
The Entity Relationship Diagram (ERD) defines the
data structure and system entities relationships. The ERD
helps identify the main entities and their interaction with
each other [25]. Figure 4 illustrates the relationship
between user, precipitation, and streamflow entities; its
interpretation is as follows (1) After successful login,
users have access to precipitation data and streamflow
data, (2) Precipitation and streamflow data are used in the
calculation process of streamflow forecasting. The ERD
model makes it easier to explain the data structure and
relationships between entities in the monthly streamflow
forecasting system. With ERD, system developers can
design and optimize the system's data structure more
efficiently.
5 Results and discussion
An online web application is the monthly streamflow
forecasting system. Users can use Internet-connected
desktops, laptops, tablets, and smartphones to access the
system. Users can access the system anytime and
anywhere as needed. When first accessing the system,
users will see a home view that contains general
information about the monthly streamflow forecasting
system. Figure 3 illustrates the home view. To access the
features in the system, users must log in by entering their
registered username and password. If the user does not
have an account, they can register through the join menu
available on the login page. The upper center of the web
application will show log on username signifying that they
have successfully entered the system and can use its
features as shown in Figure 5.
Users can access precipitation and streamflow
database menus through the monthly streamflow
forecasting system after logging in successfully as shown
in Figure 6. Users will see the feature menu located in the
upper center corner of the system page. Users will have
access to precipitation data, streamflow data, and
streamflow calculations. This menu provides access to
precipitation data based on weather stations in the Upper
Cimanuk sub-basin and the selected year. By selecting
from the menu in the middle, users can also access the
streamflow database. Streamflow menu has the same
presentation as the precipitation database, including the
choice of annual streamflow station data.
When users search for the desired data, they will get
a tabular perspective of precipitation data from the annual
elected weather station data as shown in Figure 7. In
addition, users will also see a small map showing the area
of the weather station. With this online monthly
streamflow forecasting system, users can easily access
information on precipitation and streamflow in the Upper
Cimanuk sub-basin for analysis, planning, and
development purposes in various related fields.
5.1 System evaluation
We assess the monthly streamflow forecasting system by
comparing its calculation results with manual calculations.
The results obtained are in accordance with the manual
method, indicating that the system works accurately. The
system conducts the calculation process from user input.
The user inputs precipitation and streamflow data twelve
times, representing monthly data, and then compares the
data with manual calculations. After the user inputs the
data, the system displays the final calculation table used to
determine the regression equation, correlation coefficient
(R), and coefficient of determination (R
2
). The system
provides a limit of two decimal digits to facilitate data
reading by the user.
Comparison results show that the R, R
2
, and
regression equation values obtained from both methods
are quite close, indicating that the system is accurate. The
system displays streamflow forecasting based on previous
calculations by substituting precipitation values into the
regression equation. However, the system has the
disadvantage of not being able to provide
recommendations for users in preventing flood hazards.
The system can only display monthly streamflow
forecasting based on precipitation values that will occur.
Figure 5: Dashboard of online streamflow forecast system as seen by registered user
116 Informatica 47 (2023) 109–122 I. Ardiansah et al.
The system calculates the streamflow forecasting
result for each unique streamflow station. We perform
ground check operations to verify that the predicted
streamflow values are consistent with the measured values
in the field. These ground check activities serve to verify
the accuracy of the data and ensure consistency between
the data analysis results and the real-world conditions in
the field.
We used the current meter to gauge the speed of the
river's flow while measuring the streamflow using the
velocity-region technique. As part of the ground check
procedures, we take streamflow field measurements to
utilize the velocity-region technique. For this study, we
estimated the average velocity of the river by taking
readings at two depths (0.2d and 0.8d) from the total
depth. We determined the surface region of each segment
by looking at whether the segment was triangular or
trapezium shaped. Using the below formula, we
determined the streamflow at each individual segment:
Q = A * V
(streamflow = surface region * flow velocity)
and then summed the value of each segment to obtain the
total streamflow. By comparing the forecasted streamflow
values of the system with the measured values, we can
gain insights into the validation of our streamflow
forecasting results. Table 2 presents a comprehensive
comparison between field-measured streamflow values
and the forecasted streamflow values produced by our
system. The results affirm that our system offers relatively
accurate forecasting outcomes, yet disparities exist
between the forecasted and observed results.
Table 2: Comparison of measurement results at several
weather station
Location Qground (m
3
/s) Qforecast (m
3
/s)
Bayongbong 2.46 3.37
Bojongloa 3.01 4.49
Cibatu 4.14 11.96
Leuwidaun 5.29 19.79
Leuwigoong 4.85 12.26
Analyzing the data in Table 2, we observe distinct
variations among different weather stations. For instance,
at the Bayongbong streamflow station, the observed
streamflow (Q) value is recorded as 2.46 m3/s, while the
corresponding estimated Q value is 3.37 m3/s, indicating
a slight difference between the two. Conversely, at the
Leuwidaun streamflow station, the observed Q value is
5.29 m3/s, significantly deviating from the estimated Q
value of 19.79 m3/s, marking the largest discrepancy in
this comparison.
The observed discrepancies can be attributed to a
variety of factors. One significant factor is the absence of
certain data points used in our study. Additionally, other
variables that were not included in the regression process
for streamflow forecasting, such as precipitation intensity,
duration of precipitation time, and basin shape and size,
play a role in these differences. These unaccounted
variables can influence the accuracy of our calculations,
leading to discrepancies between forecasted and measured
streamflow values.
This underscores the importance of considering a
broader spectrum of variables that impact streamflow. By
incorporating these additional variables into our
forecasting model, we can enhance the accuracy of our
streamflow predictions. As part of further research, we
recommend the collection of supplementary data and a
comprehensive analysis to refine the accuracy of our
streamflow forecasting system. Incorporating these
relevant variables will bring our forecasting results closer
to real-world field conditions, ultimately providing more
precise and valuable insights for a range of analytical,
planning, and developmental applications in related
domains.
s
Figure 6: Rainfall database of online streamflow forecast system as seen by registered user
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 117
From Figure 8, a high value of 0.91 for R² shows that
the forecasting results in this study have high accuracy and
are close to the original data value, because an R² value
close to one indicates that the regression model used can
explain the data variation. Based on the comparison
between the value of the system forecasting calculation
and field measurements, we conclude that the system
produces good forecasts. Although differences exist
between the forecasted results and the field measurements,
the system can produce streamflow estimates that are
useful in the context of this study.
The limitations and challenges of the system include
the absence of data used in the study and other variables
not included in the regression process of streamflow
forecasting, such as precipitation intensity, duration of
precipitation time, basin shape and size, among others.
Researchers can collect additional data and conduct a
more comprehensive analysis to enhance the accuracy of
the streamflow forecasting system, as well as incorporate
additional relevant variables in the forecasting model.
Additionally, regular updates of the system with new data
can help improve its accuracy over time. Integrating the
current system with other tools or systems is possible to
provide additional, more comprehensive information and
recommendations for flood hazard prevention and water
resource management.
5.2 Black box testing
Black box testing on this monthly streamflow forecasting
system aims to check each part of the system's functions
to see if they run well or not. This test includes:
• User register page: If a user leaves fields blank, the
system displays an error message, limits character
filling in the username field, and checks username
availability,
• Login page: If a user leaves fields blank or the
username and password don't match, the system
displays an error message. The forgotten password
and unregistered functions also work well,
• Flow-rate forecasting page from the user: The
calculate flow rate button works well. The system
displays a warning if there are empty fields or input
other than numbers. If everything checks out, the
system previews a monthly forecast of streamflow.
• Flow rate forecasting page from the database: All
buttons work properly, including the streamflow
station select box and the beginning and ending year
select boxes. The final year select box adjusts to the
selected initial year select box, facilitating user
selection and preventing errors.
With this black box test, the monthly streamflow
forecasting system successfully ensures that all functions
run properly and in accordance with user needs.
5.3 Discussion
In the realm of integrated streamflow forecasting systems,
our study embarks on an expansive discourse, intricately
juxtaposing our freshly conceived forecasting system with
the landscape outlined in Table 2. This illuminating
comparative analysis serves as a prism, elucidating
profound insights into the diverse echelons of forecast
accuracy achieved through various methodologies. Most
notably, it unveils the distinctive advantages harnessed
from the harmonious amalgamation of a regression model
with historical rainfall and streamflow data.
Our crafted monthly streamflow forecasting system, a
testament to our study's rigor, proudly bears the emblem
of precision, conspicuously manifested by the substantial
R² value of 0.91 underscored in Figure 8. This robust
correlation serves as an unequivocal testament to the
prowess of our regression model, proficiently dissecting
and explicating a significant fraction of the data variance.
The upshot is an assemblage of forecasts that
harmoniously converge with authentic streamflow values.
Though discernible disparities linger between our
Figure 7: Weather station location and streamflow database as seen by registered user
118 Informatica 47 (2023) 109–122 I. Ardiansah et al.
foreseen outcomes and the tangible field observations, the
prognostications furnished by our system remain germane
within the contextual contours of our study's purview.
In juxtaposition to antecedent inquiries, our study
discerns its stronghold through a nuanced interplay of
advantages. Within the annals of streamflow prediction,
the focus on regression models, as evidenced by the
endeavors of [10], has unfurled a promising tapestry of
potential. Noteworthy is their aptitude in illuminating the
dynamics of streamflow, as proven by their ability to
forecast with precision. Our comparative analysis with
this lineage underscores a pivotal divergence, attributed to
the infusion of a regression model with historical rainfall
and streamflow data. This synergy of variables serves as
our study's fulcrum, endowing our system with an elevated
plane of accuracy and efficacy.
Moreover, the rise of online platforms, as propounded
by [11], has ushered an epoch of transformation in water
resource management. Our innovative solution further
amplifies this transformative spirit by seamlessly
integrating an online platform. This dynamic interplay
empowers stakeholders, offering not only forecasting
insights but also interactive tools for decision-making. By
enhancing collective awareness and understanding of
water-related hazards and mitigation strategies, our
system underscores a pivotal shift in the approach to
resource management, surpassing mere prediction and
ushering in the era of informed, preemptive action.
Validation, a cornerstone of credibility, as
emphasized by [12], forms a fundamental tenet of our
approach. We champion this by grounding our system in
meticulous validation processes, ensuring its robustness
and reliability. Furthering the thread of augmentation,
spatial and temporal factors, as underscored by [13], etch
their influence onto our methodology. The systematic
inclusion of these factors enriches the predictive prowess
of our system, enabling us to identify at-risk areas and
extend timely safety measures, thereby fostering a
paradigm of proactive risk mitigation.
Incorporating additional variables, as exemplified by
[14], kindles a beacon of promise. Our study inherently
heeds this call by embracing historical rainfall and
streamflow data as pivotal components in our predictive
framework. This infusion enhances our accuracy, aligning
our forecasts more closely with real-world dynamics.
Moreover, the fusion of cross-disciplinary collaboration
and the insights of expert users, as illuminated by [15],
epitomizes our journey. We seamlessly weave life
sciences and Earth system models, fusing empirical
evidence with expert intuition. By doing so, we endow
decision-makers with a holistic and nuanced
understanding, empowering them to navigate complex
climate change challenges with clarity and astuteness.
Yet another facet, resonating with [16], unearths the
transformative potential of big data and machine learning
technologies. Our system, founded upon these very
principles, encapsulates the transformative spirit of data-
driven environmental management. In doing so, we bridge
the gap between theoretical concepts and practical
application, rendering our solution not merely a scholarly
endeavor but a dynamic instrument to revolutionize the
very landscape of water resource management.
In synthesis, our study's discourse within the domain
of integrated streamflow forecasting unfolds as a
testament to the fusion of innovation, validation, and
cross-disciplinary insights. By synergizing the potency of
regression models with historical data, augmenting our
platform with dynamic online tools, and embracing the
nuances of spatial, temporal, and additional variables, our
approach stands as a beacon of advancement. It is an
embodiment of the metamorphosis in water resource
management, forging an informed and empowered future,
where foresight marries action to harmoniously navigate
the dynamic tapestry of our hydrological world.
Figure 8: Comparison of measurement results between field measurements and system
measurements
y = 5,3231x - 10,652
R² = 0,9125
0
2
4
6
8
10
12
14
16
18
20
2 2,5 3 3,5 4 4,5 5 5,5
Application Streamflow Calculation (m
3
/S)
Field Streamflow Calculation (m
3
/s)
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 119
The novelty of our integrated streamflow forecasting
system, which transcends its technical intricacies, lies in
its profound impact within the domains of flood
management and water resource optimization. A hallmark
of our solution is the seamless integration of an online
platform, endowing users with real-time interactive
capabilities for data input and visualization. This aspect
significantly amplifies the pragmatic utility and versatility
of our forecasting system.
Central to the efficacy of the online platform is its
ability to provide facile access to monthly streamflow and
precipitation data. This resource-rich accessibility
empowers decision-makers to make well-informed
choices concerning water resource development within
the Upper Cimanuk sub-basin. Furthermore, our platform
serves as a conduit for timely initiation of proactive flood
disaster mitigation strategies, including early warning
systems and evacuation plans, all rooted in accurate and
promptly delivered information.
Beyond its applications in flood management, our
forecasting system bears relevance to diverse sectors
encompassing agriculture, irrigation, and hydropower. By
delivering precise streamflow predictions, it optimizes
irrigation scheduling, enabling farmers to maximize crop
yields while conserving water resources.
The harmonization of an online platform with our
regression model engenders a synergy, positioning our
integrated streamflow forecasting system as an instrument
in the realm of smart flood management. The intuitive user
interface augments the accessibility and usability of our
findings, rendering our approach highly pertinent in the
context of contemporary water resource management
practices.
6 Conclusion
The monthly streamflow forecasting system generates
proportionate precise forecasting results that resemble
values of original data. This is evident from the 100%
coefficient of determination (R²) value, which indicates
that the regression model can adequately explain the
differences in the data. Nevertheless, there is potential to
develop and improve the accuracy and precision of
forecasting. There may be discrepancies between
predicted and measured results due to lack of data and the
exclusion of other variables from the streamflow
regression prediction process. These include the intensity
and duration of precipitation, also the shape and size of the
catchment area.
To improve the quality of the system's predictions, we
recommend adding more variables affecting flow and
collecting more data. Extensive field data validation can
also optimize the performance of the forecasting system.
The developed monthly flow forecasting system can be
considered as a valuable tool in the decision making
process through flow analysis and forecasting in water
resource management and flood prevention.
We highlight several limitations and challenges
associated with the developed system. Firstly, the system
accuracy is affected by the exclusion of other variables,
such as rainfall intensity, rainfall duration, basin shape and
size, etc. Next, the system only displays monthly
streamflow forecasts based on past rainfall, and does not
provide recommendations or next steps to prevent flood
hazards.
To address these limitations and challenges, we
propose multiple methods for improving the performance
of the system. Initially, researchers can collect additional
data, conduct a more thorough analysis, and incorporate
additional relevant variables into the forecasting model.
This can enhance the accuracy and precision of the
forecasting. Second, improved system capabilities can
provide users with recommendations for the next steps to
prevent flood hazards. For instance, the system can
recommend flood prevention measures based on the
anticipated streamflow values. Thirdly, we are conducting
field data validation to optimize system performance. This
will help confirm the forecasting accuracy and provide
valuable selective data for examination, strategy
development, and the enhancement of related fields.
7 Implications and future work
Our integrated streamflow forecasting system holds
significant potential for various real-world water resource
management scenarios. By focusing on the Upper
Cimanuk sub-basin, a region characterized by
heterogeneous topography and hydrological attributes
prone to flooding, our system directly confronts the
challenges inherent in flood management and water
resource optimization. This region serves as a testing
ground for evaluating the system's efficacy within the
complexities of its environment, making it an ideal
crucible for assessing performance across diverse terrains
and land use patterns. Moreover, the sub-basin's
susceptibility to floods accentuates the critical role of
precise streamflow forecasts in underpinning effective
flood disaster mitigation strategies.
The intricacies of our methodology revolve around
the development and evaluation of a bespoke monthly
streamflow forecasting system tailored to the Upper
Cimanuk sub-basin. Central to this system is a
sophisticated regression model that seamlessly blends
historical rainfall and streamflow data gleaned from an
extensive network of monitoring stations across the sub-
basin. This model artfully elucidates the intricate interplay
between regional rainfall patterns and the resulting
streamflow dynamics.
Our methodology deliberately incorporates a
straightforward linear regression model to establish a
coherent link between rainfall and streamflow data. This
model serves to illuminate the complex relationships
underpinning these variables. At the same time, we ensure
seamless access and engagement through our user-centric
web application. This platform allows users to interact
with our forecasting system from any internet-connected
device, making forecasting easy and convenient to use.
7.1 Challenges and limitations
It is important to recognise the potential challenges of real-
world implementation, although our system offers
120 Informatica 47 (2023) 109–122 I. Ardiansah et al.
promising results. Missing data points and unaccounted
for variables such as rainfall intensity, rainfall duration, or
basin shape and size can affect the accuracy of the
calculations. These factors contribute to the difference
between the predicted value and the manual measurement
value. In addition, although widely used, the velocity-area
method is difficult to detect the dynamics of flow
variations when extreme events or irregular river shapes
occur.
Other important considerations include scalability,
performance and data availability. Forecast accuracy and
real-time data acquisition become critical as systems scale
to larger areas. To guarantee forecast accuracy, it is
important to have consistent and reliable data sources.
7.2 Areas of improvement
We suggest several alternatives to improve system
accuracy. Incorporating additional variables such as
rainfall intensity, duration and watershed characteristics,
can improve forecast accuracy and minimize gaps
between predictions and observations. In addition,
exploring the integration of machine learning and artificial
intelligence techniques could be a promising technology
for enhancing predictive capabilities. The system can
provide more dynamic and accurate predictions,
especially for extreme events, by training machine
learning models on large data sets.
Reference
[1] M. Tanoue, R. Taguchi, S. Nakata, S. Watanabe,
S. Fujimori, and Y. Hirabayashi, “Estimation of
Direct and Indirect Economic Losses Caused by a
Flood With Long‐Lasting Inundation: Application
to the 2011 Thailand Flood,” Water Resour. Res.,
vol. 56, no. 5, May 2020, doi:
10.1029/2019WR026092.
[2] Z. Qin, M. Storozum, H. Liu, X. Zhang, and T. R.
Kidder, “Investigating environmental changes as
the driving force of agricultural intensification in
the lower reaches of the Yellow River: A case
study at the Sanyangzhuang site,” Quat. Int., vol.
521, pp. 25–34, Jun. 2019, doi:
10.1016/j.quaint.2019.06.033.
[3] T. Mai et al., “Defining flood risk management
strategies: A systems approach,” Int. J. Disaster
Risk Reduct., vol. 47, p. 101550, Aug. 2020, doi:
10.1016/j.ijdrr.2020.101550.
[4] E. Savitri and I. Pramono, “ANALISIS BANJIR
CIMANUK HULU 2016,” J. Penelit.
pengelolaan Drh. Aliran Sungai, vol. 1, no. 2, pp.
97–110, Oct. 2017, doi:
10.20886/jppdas.2017.1.2.97-110.
[5] M. Ardiansyah, R. A. Nugraha, L. O. S. Iman, and
S. D. Djatmiko, “Impact of Land Use and Climate
Changes on Flood Inundation Areas in the Lower
Cimanuk Watershed, West Java Province,” J.
Ilmu Tanah dan Lingkung., vol. 23, no. 2, pp. 53–
60, Dec. 2021, doi: 10.29244/jitl.23.2.53-60.
[6] F. A. Hirpa et al., “Calibration of the Global Flood
Awareness System (GloFAS) using daily
streamflow data,” J. Hydrol., vol. 566, pp. 595–
606, Nov. 2018, doi:
10.1016/j.jhydrol.2018.09.052.
[7] Y. Gao, C. Merz, G. Lischeid, and M. Schneider,
“A review on missing hydrological data
processing,” Environ. Earth Sci., vol. 77, no. 2, p.
47, Jan. 2018, doi: 10.1007/s12665-018-7228-6.
[8] P. C. Austin and E. W. Steyerberg, “The number
of subjects per variable required in linear
regression analyses,” J. Clin. Epidemiol., vol. 68,
no. 6, pp. 627–636, Jun. 2015, doi:
10.1016/j.jclinepi.2014.12.014.
[9] Reza Pahlevi, “APJII: Penetrasi Internet
Indonesia Capai 77,02% pada 2022,” katadata,
2022.
https://databoks.katadata.co.id/datapublish/2022/
06/10/apjii-penetrasi-internet-indonesia-capai-
7702-pada-2022
[10] P. A. Mendoza et al., “An intercomparison of
approaches for improving operational seasonal
streamflow forecasts,” Hydrol. Earth Syst. Sci.,
vol. 21, no. 7, pp. 3915–3935, Jul. 2017, doi:
10.5194/hess-21-3915-2017.
[11] A. Teague, Y. Sermet, I. Demir, and M. Muste, “A
collaborative serious game for water resources
planning and hazard mitigation,” Int. J. Disaster
Risk Reduct., vol. 53, p. 101977, 2021, doi:
https://doi.org/10.1016/j.ijdrr.2020.101977.
[12] S. Mohammed et al., “A comparative analysis of
data mining techniques for agricultural and
hydrological drought prediction in the eastern
Mediterranean,” Comput. Electron. Agric., vol.
197, p. 106925, 2022, doi:
https://doi.org/10.1016/j.compag.2022.106925.
[13] J.-P. Naulin, O. Payrastre, and E. Gaume,
“Spatially distributed flood forecasting in flash
flood prone areas: Application to road network
supervision in Southern France,” J. Hydrol., vol.
486, pp. 88–99, 2013, doi:
https://doi.org/10.1016/j.jhydrol.2013.01.044.
[14] Y. Sun, D. Wendi, D. E. Kim, and S.-Y. Liong,
“Deriving intensity–duration–frequency (IDF)
curves using downscaled in situ rainfall
assimilated with remote sensing data,” Geosci.
Lett., vol. 6, no. 1, p. 17, 2019, doi:
10.1186/s40562-019-0147-x.
[15] D. S. Schoeman et al., “Demystifying global
climate models for use in the life sciences,”
Trends Ecol. Evol., vol. 38, no. 9, pp. 843–858,
Sep. 2023, doi: 10.1016/j.tree.2023.04.005.
[16] A. Y. Sun and B. R. Scanlon, “How can Big Data
and machine learning benefit environment and
water management: a survey of methods,
applications, and future directions,” Environ. Res.
Lett., vol. 14, no. 7, p. 73001, 2019, doi:
10.1088/1748-9326/ab1b7d.
[17] S. Otuagoma, E. Ogujor, and P. Kuale,
“Comparative Measurement of Stream Flow in the
Ethiope River for Small Hydropower
Development,” Niger. J. Technol., vol. 34, no. 1,
Integrated Streamflow Forecasting System: A Step Towards Smart Flood Management Informatica 47 (2023) 109–122 121
p. 184, Dec. 2014, doi: 10.4314/njt.v34i1.23.
[18] L. Yuge and T. Badarch, “Research on
Contemporary Software Development Life Cycle
Models,” Am. J. Comput. Sci. Technol., vol. 6, no.
1, pp. 1–9, Mar. 2023, doi:
10.11648/j.ajcst.20230601.11.
[19] M. Jena and S. Dehuri, “DecisionTree for
Classification and Regression: A State-of-the Art
Review,” Informatica, vol. 44, no. 4, Dec. 2020,
doi: 10.31449/inf.v44i4.3023.
[20] I. Ardiansah, N. Bafdal, E. Suryadi, and A. Bono,
“Design of micro-climate data monitoring system
for tropical greenhouse based on arduino UNO
and raspberry pi,” IOP Conf. Ser. Earth Environ.
Sci., vol. 757, no. 1, 2021, doi: 10.1088/1755-
1315/757/1/012017.
[21] A.-A. Hussein, V. Govindu, and A. G. M.
Nigusse, “Evaluation of groundwater potential
using geospatial techniques,” Appl. Water Sci.,
vol. 7, no. 5, pp. 2447–2461, Sep. 2017, doi:
10.1007/s13201-016-0433-0.
[22] G. B. Adane, B. A. Hirpa, C.-H. Lim, and W.-K.
Lee, “Evaluation and Comparison of Satellite-
Derived Estimates of Rainfall in the Diverse
Climate and Terrain of Central and Northeastern
Ethiopia,” Remote Sens., vol. 13, no. 7, p. 1275,
Mar. 2021, doi: 10.3390/rs13071275.
[23] Y. Wu and P. Xie, “Exploration of Enterprise
Audit Information Management System Model
Based on Data Flow Diagram,” in 2021
International Wireless Communications and
Mobile Computing (IWCMC), IEEE, Jun. 2021,
pp. 1997–2001. doi:
10.1109/IWCMC51323.2021.9498870.
[24] R. Arifin, G. N. Syaifuddiin, D. Desriyanti, Z. U.
Rosyidin, and G. A. Buntoro, “A Highly Accurate
Internet-Based Fake Information Detection Tool
for Indonesian Twitter,” Informatica, vol. 46, no.
9, Jan. 2023, doi: 10.31449/inf.v46i9.4416.
[25] T. A. Vakaliuk, O. V. Korotun, and S. O.
Semerikov, “The selection of cloud services for
ER-diagrams construction in IT specialists
databases teaching,” CTE Work. Proc., vol. 8, pp.
384–397, Mar. 2021, doi: 10.55056/cte.295.
122 Informatica 47 (2023) 109–122 I. Ardiansah et al.