INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 973

Analysis and Predictive Modeling of River Water Level in Surma-

Meghna River

Zubayer IBN Mostafa.,

Zayed IBN Mostafa.,

Khandaker Ferdous Rawnak.,

Presila Tanchangya

Sub-Assistant Engineer / Sectional Officer, Mithamain Water Development Section BWDB, Kishoreganj.

Institution: Presidency University

3,4

Institution: Shahjalal University of Science and Technology, Sylhet

*Corresponding Author

DOI: https://doi.org/10.51583/IJLTEMAS.2025.1408000126

Abstract: This study explores the hydrological dynamics of the Meghna-Surma basin in Kishoregonj, Bangladesh, with the

objective of analyzing and forecasting river water levels. Using 30 years of daily data (1995–2024) on rainfall, discharge, and

water levels, the study first applies statistical regression to examine correlations among the variables. Subsequently, the ARIMA

(Autoregressive Integrated Moving Average) model is employed to predict future water level trends up to 2029. Findings reveal a

strong correlation between rainfall and discharge, and the ARIMA (4,1,3) model demonstrated satisfactory short-term forecasting

performance. The study highlights the practicality of using ARIMA for real-time water level prediction in data-constrained

environments. These insights are crucial for flood risk mitigation, agricultural planning, and infrastructure development in river-

prone regions of Bangladesh.

Keyword: River water level forecasting , Meghna-Surma basin , ARIMA model , Hydrological correlation , Flood risk

management

I. Introduction

Bangladesh, being a riverine country, is shaped by its adjoining ecosystem which is constituted of the Ganges-Brahmaputra-

Meghna (GBM) delta system/ It possesses the most active and complex system of hydrology in the world. The nation’s socio-

economic activities are deeply connected with river water levels of more than 700 rivers, including Meghna, Surma and

Kushiyara (Kumar et al., 2022). The Meghna-Surma basin is a sensitive region with a quite hydrological dynamic. This is

particularly located around Kishoregonj District and Bhairab Upazila of Bangladesh. This area’s agriculture, fisheries,

transportation and rural livelihood are frequently disrupted by Meghna-Surma basin. This is why, proper understanding of its

water level, discharge and rainfall is vital for flood risk management and irrigation planning (Mojid, 2020).

Due to upstream flow and monsoonal precipitation, river water levels and discharge patterns get influenced. And, this leads to

uncertain flooding or water scarcity. Real-time analysis and prediction of these hydrological parameters is significant to reduce

disaster impacts and improve planning of water resource (Rahman et al., 2024). The forecast of water level trends based on

historical dynamics and correlation with rainfall and discharge, provides valuable insights for engineers and policymakers.

This research basically focuses on Meghna-Surma basin by analyzing water level, discharge and rainfall data of 30 years. Firstly,

simple statistical regression analysis was employed to determine the correlation among these parameters. Lastly, Autoregressive

Integrated Moving Average (ARIMA) was applied to predict future water levels. It has been proven that ARIMA is a robust and

effective time series forecasting method in river water level predictions (Noor et al., 2022).

Figure 1. Study Area

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 974

Figure 2- Flood prone Areas

II. Literature Review

There have been many studies where the statistical inter-relationships among hydrological parameters were illustrated. In a study,

a comprehensive analysis of Chari-Logone sub-basins where rainfall variations have an impact on river discharge (Mahamat Nour

et al., 2021). It has also been investigated that there is a strong positive correlation between rainfall and discharge in Surma river

which emphasizes the significance of regional statistical analyses for flood prediction(Akter et al., 2019). The rainfall-runoff

interactions affected by land use changes and groundwater exchanges and it induces more complex behavior (Gao et al., 2011).

Statistical regression and correlation approaches have been largely utilizing by researchers since many years. It is a basic method

and can be used to understand the inter-dependency among two or more parameters. Researchers have examined lagged

relationships between rainfall and river water levels using correlation techniques (Getirana et al., 2021). There is a liner

correlation between rainfall and discharge data in Kushiyara river basin and it was proved to be useful in predicting peak flood

(Uddin et al., 2018). These researches are showing the necessity of statistical techniques in correlating the dynamic relation with

respect to rainfall, discharges and water levels to areas which are flood-prone.

Beyond the correlation analysis, the use of ARIMA models has managed to predict the hydrological variables such as river water

levels, streamflow and rainfall. It has been demonstrated that the application of ARIMA to forecast the streamflow in non-

stationary environments whereby the aptitude of utility in changeable climatic situations (Adamowski & Chan, 2011). In

Southeast Asia, ARIMA has been used to predict short-term water levels in the Mekong River and it showed high accuracy (Tran

et al., 2022). ARIMA was also applied in predicting water levels in Jakarta, where it surpassed Long Short-Term Memory

(LSTM) in Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) (Kaburuan et al., 2024)

ARIMA has been utilized in water level forecasting in a bigger hydrologic system such as the Ganges and subsidiary

Brahmaputra systems in the case of Bangladesh with some promising outcomes (Rashid et al., 2022). However, local or regional

studies are less common in critical hydrological basin like Meghna-Surma basin in Kishoregonj-Bhairab. This research bridges

this gap by including statistical regressiona analysis as a means of identifying correlations with ARIMA-based forecasting.

III. Methodology

Study Area and Data Collection

This study focuses on the Meghna-Surma basin within Kishoregonj District, specifically targeting the Bhairab Upazila area,

which is prone to seasonal hydrological fluctuations. The dataset comprises daily water level (WL in mMSL), rainfall (mm), and

discharge (m³/s) records spanning from 1995 to 2024. Data were obtained from national hydrological monitoring agencies and

included time-stamped entries with uniform temporal resolution.

Data Preprocessing

 The dataset was first cleaned to handle missing values, ensuring continuity.

 Outliers were assessed using boxplots and visual inspections, and anomalies beyond three standard deviations were

removed.

 To align datasets temporally, all variables were aggregated to daily averages, ensuring consistency in time series length.

Correlation Analysis: Simple Statistical Regression Approach

The first objective was to evaluate the interrelationship among water level, discharge, and rainfall through Pearson’s correlation

coefficient (r). This linear regression-based analysis quantifies the strength and direction of association:

𝑟 =

∑

(𝑥

𝑖

− 𝑥)(𝑦

𝑖

− 𝑦)

√

∑

(𝑥

𝑖

− 𝑥)

∑

(𝑦

𝑖

− 𝑦)

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 975

Where 𝑥 and 𝑦 represent paired observations (e.g., water level and rainfall), and 𝑥, 𝑦 are their respective means. Correlation

matrices were generated to visualize relationships between variables.

ARIMA Modeling for Water Level Prediction

The Autoregressive Integrated Moving Average (ARIMA) model was employed to forecast future water level trends in the

Meghna-Surma basin at Bhairab, Kishoregonj. ARIMA is one of the most widely applied statistical models in hydrological time

series forecasting due to its simplicity, adaptability, and robustness in capturing linear dependencies within temporal datasets

(Adamowski & Chan, 2011; Box, 2013).

Rationale for Using ARIMA

Hydrological processes such as river water level fluctuations often exhibit autocorrelation, seasonality, and trends over time.

While physical-based models require extensive spatial-temporal data (e.g., soil moisture, evapotranspiration), data-driven models

like ARIMA can forecast future values solely based on historical observations (Wang et al., 2009). Given the availability of a 30-

year daily water level dataset (1995–2024) and the data-limited context of the region, ARIMA is an appropriate choice for

modeling medium-term water level predictions (Akter et al., 2019).

Model Identification and Parameter Selection

To build an effective ARIMA model, it is essential to determine the three core parameters:

 p (Autoregressive Order): Represents the influence of past values (lags) on current water levels.

 d (Differencing Order): Ensures stationarity by eliminating trends.

 q (Moving Average Order): Accounts for the correlation of forecast errors.

a) Stationarity Check

The time series data were initially plotted and tested using the Augmented Dickey-Fuller (ADF) Test, which revealed non-

stationarity due to inherent seasonal and trend components. First-order differencing (d=1) was applied to stabilize the mean and

remove trend components, which is a common practice in hydrological time series modeling (Khodakhah et al., 2022).

b) PACF and ACF Analysis

 The Partial Autocorrelation Function (PACF) plot showed significant lags up to the 4th order, indicating that water

levels at a given time point are influenced by values from the past four days.

 The Autocorrelation Function (ACF) plot exhibited significant spikes up to lag 3 in residual errors, guiding the

selection of q=3.

 Thus, the chosen ARIMA structure was ARIMA (4,1,3).

These parameter selections align with established hydrological ARIMA modeling protocols, where PACF is used to determine p,

ACF for q, and differencing to achieve stationarity (Valipour et al., 2016, Water Resources Management).

Model Fitting and Validation

The ARIMA (4,1,3) model was calibrated on the training dataset using Maximum Likelihood Estimation (MLE). The resulting

model coefficients were:

 AR terms: ar.L1 = -0.3781, ar.L2 = -0.5645, ar.L3 = -0.0704, ar.L4 = -0.1047

 MA terms: ma.L1 = -0.6385, ma.L2 = 0.1407, ma.L3 = -0.4931

Model performance metrics were:

 MSE = 5.5788

 RMSE = 2.3620

 MAE = 2.3079

 R² = -88.36 (Low R² values are not uncommon in hydrological ARIMA models due to the inherently stochastic nature of

river systems and limitations of purely linear models in capturing nonlinear hydrological processes.)

However, for short-term water level prediction in data-limited regions, ARIMA has been proven to perform adequately,

especially when combined with noise simulation to represent uncertainty.

Forecasting and Uncertainty Simulation

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 976

Forecasts were generated for the period 2025–2029 (60 months). To incorporate the inherent randomness of environmental

processes, Gaussian white noise was added to the forecasted values. This step was taken to reflect real-world variability, a

common practice in hydrological ARIMA applications to produce realistic forecast intervals.

Visualizations plotted both historical data and forecasted trends, indicating expected seasonal peaks during monsoon periods and

gradual recession during dry seasons. The forecasting model is valuable for short to medium-term water level projections,

providing critical insights for flood preparedness, irrigation scheduling, and infrastructure planning in Kishoregonj-Bhairab.

IV. Result and Discussion

The results of this study confirm the strong interrelationships among rainfall, river discharge, and water level in the Meghna-

Surma basin. The Pearson correlation analysis indicated a significant positive relationship between rainfall and discharge, which

aligns with previous findings in similar hydrological contexts (Akter et al., 2019; Uddin et al., 2018). This reinforces the

importance of rainfall as a leading indicator in predicting water level fluctuations in the region.

The application of the ARIMA (4,1,3) model demonstrated the model’s effectiveness in capturing seasonal trends and short-term

variability in river water levels. Although the R² value was negative (-88.36), which might initially indicate poor model

performance, it’s important to understand that low R² values are common in hydrological time series forecasting due to the

inherently stochastic and nonlinear nature of river systems (Khodakhah et al., 2022). Instead, the model's performance should be

evaluated based on other error metrics. In this study, the model achieved an RMSE of 2.36 and an MAE of 2.30, indicating

reasonable accuracy for short-term forecasts in data-limited settings.

The inclusion of Gaussian white noise in the ARIMA forecasts adds a layer of realism by simulating environmental uncertainty,

which is particularly useful for practical planning scenarios. The forecasted water level trends for 2025–2029 showed expected

seasonal peaks during the monsoon and recessions in dry months—patterns that align with historical data.

Importantly, the study fills a knowledge gap by focusing on the Meghna-Surma basin—a region less explored in the context of

predictive modeling, despite its high vulnerability to flooding. By combining simple regression with ARIMA forecasting, this

research provides a replicable framework for other hydrologically sensitive regions in Bangladesh and beyond.

In conclusion, the research validates ARIMA as a reliable and practical tool for water level forecasting in river basins where data

availability is limited. The findings can support more proactive flood response strategies, irrigation planning, and rural

infrastructure development, particularly in regions like Kishoregonj-Bhairab where seasonal river fluctuations have high socio-

economic impacts.

Author’s Contribution:

All authors contributed to the study conception and design. The manuscript has been read and approved by the authors and there

are no other people who satisfied the criteria for authorship are listed. The order of authors listed in the manuscript has been

approved by all of us.

V. Funding

This research received no external funding.

Acknowledgement

N/A

Conflict of Interest

The authors declare no conflict of interest.

Figure 3 – Correlation Analysis

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 977

Figure 4 – Predictive water level

Figure 5. Wettest and Driest year (Discharge level)

Figure 6: Wettest and Driest year (Discharge level)

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 978

Figure 7: Wettest and Driest year(Rainfall)

References

1. Adamowski, J., & Chan, H. F. (2011). A wavelet neural network conjunction model for groundwater level forecasting.

Journal of Hydrology, 407(1–4), 28–40. https://doi.org/10.1016/j.jhydrol.2011.06.013

2. Akter, S., Howladar, M. F., Ahmed, Z., & Chowdhury, T. R. (2019). The rainfall and discharge trends of Surma River

area in North-eastern part of Bangladesh: an approach for understanding the impacts of climatic change. Environmental

Systems Research, 8(1), 28. https://doi.org/10.1186/s40068-019-0156-y

3. Box, G. (2013). Box and Jenkins: Time Series Analysis, Forecasting and Control. In A Very British Affair (pp. 161–215).

Palgrave Macmillan UK. https://doi.org/10.1057/9781137291264_6

4. Gao, H., Bohn, T. J., Podest, E., McDonald, K. C., & Lettenmaier, D. P. (2011). On the causes of the shrinking of Lake

Chad. Environmental Research Letters, 6(3), 034021. https://doi.org/10.1088/1748-9326/6/3/034021

5. Getirana, A., Kumar, S. V., Konapala, G., & Ndehedehe, C. E. (2021). Impacts of Fully Coupling Land Surface and

Flood Models on the Simulation of Large Wetlands’ Water Dynamics: The Case of the Inner Niger Delta. Journal of

Advances in Modeling Earth Systems, 13(5). https://doi.org/10.1029/2021MS002463

6. Kaburuan, E. R., Mutu Manikam, R., Kucharov Olimjon Ruzimurodovich, O. K., & Rianggi Boyo, A. (2024).

Comparison of LSTM and ARIMA Algorithms in Predicting Water Levels in Jakarta. 2024 International Conference on

Orange Technology (ICOT), 1–8. https://doi.org/10.1109/ICOT64290.2024.10936979

7. Khodakhah, H., Aghelpour, P., & Hamedi, Z. (2022). Comparing linear and non-linear data-driven approaches in

monthly river flow prediction, based on the models SARIMA, LSSVM, ANFIS, and GMDH. Environmental Science and

Pollution Research, 29(15), 21935–21954. https://doi.org/10.1007/s11356-021-17443-0

8. Kumar, A., Mondal, S., & Lal, P. (2022). Analysing frequent extreme flood incidences in Brahmaputra basin, South Asia.

PLOS ONE, 17(8), e0273384. https://doi.org/10.1371/journal.pone.0273384

9. Mahamat Nour, A., Vallet‐Coulomb, C., Gonçalves, J., Sylvestre, F., & Deschamps, P. (2021). Rainfall-discharge

relationship and water balance over the past 60 years within the Chari-Logone sub-basins, Lake Chad basin. Journal of

Hydrology: Regional Studies, 35, 100824. https://doi.org/10.1016/j.ejrh.2021.100824

10. Mojid, M. A. (2020). Climate change-induced challenges to sustainable development in Bangladesh. IOP Conference

Series: Earth and Environmental Science, 423(1), 012001. https://doi.org/10.1088/1755-1315/423/1/012001

11. Noor, F., Haq, S., Rakib, M., Ahmed, T., Jamal, Z., Siam, Z. S., Hasan, R. T., Adnan, M. S. G., Dewan, A., & Rahman,

R. M. (2022). Water Level Forecasting Using Spatiotemporal Attention-Based Long Short-Term Memory Network.

Water, 14(4), 612. https://doi.org/10.3390/w14040612

12. Rahman, M. M., Chowdhury, M. M. I., Al Amran, M. I. U., Malik, K., Abubakar, I. R., Aina, Y. A., Hasan, M. A.,

Rahman, M. S., & Rahman, S. M. (2024). Impacts of climate change on food system security and sustainability in

Bangladesh. Journal of Water and Climate Change, 15(5), 2162–2187. https://doi.org/10.2166/wcc.2024.631

13. Rashid, A., Alamgir, M., Ahmed, M. T., Salam, R., Islam, A. R. Md. T., & Islam, A. (2022). Assessing and forecasting of

groundwater level fluctuation in Joypurhat district, northwest Bangladesh, using wavelet analysis and ARIMA modeling.

Theoretical and Applied Climatology, 150(1–2), 327–345. https://doi.org/10.1007/s00704-022-04160-y

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIV, Issue VIII, August 2025

www.ijltemas.in Page 979

14. Tran, T. T., Duy Nguyen, L., Bao Pham, Q., Thanh Tran, T., Ngoc Hoai, P., Thi Thanh Huyen, P., Phuong Dong, N.,

Hoang Hieu, H., & Thu Hien, N. (2022). Long Short-Term Memory (LSTM) neural networks for short-term water level

prediction in Mekong river estuaries. Article in Songklanakarin Journal of Science and Technology, 44(4), 1057–1066.

https://doi.org/10.14456/sjst-psu.2022.138

15. Uddin, M., Akter, S., Uddin, M., & Diganta, M. (2018). Trend Analysis Variations and Relation Between Discharge and

Rainfall: a Study on Kushiyara River. Journal of Environmental Science and Natural Resources, 10(2), 121–132.

https://doi.org/10.3329/jesnr.v10i2.39025

16. Wang, W.-C., Chau, K.-W., Cheng, C.-T., & Qiu, L. (2009). A comparison of performance of several artificial

intelligence methods for forecasting monthly discharge time series. Journal of Hydrology, 374(3–4), 294–306.

https://doi.org/10.1016/j.jhydrol.2009.06.019