📄 earth_science

Hydrological Modelling and Runoff Estimation of the Wainganga River Basin Using HEC-HMS and Geospatial Clustering Techniques

This study demonstrates that an integrated framework combining hierarchical clustering, Thiessen polygon analysis, and HEC-HMS modeling effectively improves rainfall-runoff estimation and flood forecasting accuracy in the monsoon-prone Wainganga River basin, as evidenced by high performance metrics during calibration and validation.

Original authors: Rajesh Vijaykumar Kherde, Gauri Patil, Kiran More, Priyadarshi Sawant

Published 2026-07-31

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Rajesh Vijaykumar Kherde, Gauri Patil, Kiran More, Priyadarshi Sawant

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Earth as a giant, living sponge. When it rains, this sponge soaks up water, but if the rain comes too fast or the sponge is already full, the excess water spills over, racing down hills and valleys to form rivers. This race of water is called "runoff," and predicting how much water will rush through a river is a bit like trying to guess how fast a crowd will run through a maze after a fire alarm. Scientists who study this are called hydrologists, and they use special computer programs to build digital twins of river basins. These programs need two main things to work: a map of the land (to know where the water can go) and a record of how much rain fell (to know how much water is entering the system). The tricky part is that rain doesn't fall evenly; one hill might get a storm while the next valley stays dry. To solve this, scientists use clever tricks to figure out the "average" rain for a whole area and then feed that data into their computer models to see if they can predict floods before they happen.

This is exactly what a team of researchers did for the Wainganga River in central India, a place that often gets soaked by heavy monsoon rains. Their goal was to build a better "digital twin" of this river basin to predict floods more accurately. They used a powerful computer tool called HEC-HMS, which acts like a virtual water calculator, but they realized they needed to organize their rain data smarter first. Think of the rain gauges (the devices that measure rain) as a group of friends at a party. Some friends always talk about the same things, while others have different stories. The researchers used a technique called "Hierarchical Clustering" to group these rain gauges based on how similar their "stories" (rainfall patterns) were. It's like sorting your playlist by mood: you wouldn't mix a heavy metal song with a lullaby, so you group the similar ones together. By grouping the rain gauges this way, they could pick the best representatives for each area, ensuring their computer model wasn't confused by conflicting data.

Once the rain data was organized, they used another method called "Thiessen Polygons." Imagine drawing a map where every point on the ground belongs to the nearest rain gauge, kind of like drawing territories on a game board. If you are standing closer to the gauge in Seoni than the one in Balaghat, your rain is counted as Seoni's rain. This helped the team calculate the total amount of rain falling over the entire river basin, even in places where there were no gauges. They fed this organized rain data, along with maps of the soil and land (like forests and farms), into their HEC-HMS model. The model then simulated how the water would move through the river system, turning rain into flowing rivers.

The team tested their model by running simulations for two different time periods. First, they "calibrated" it using data from 2001 to 2010, which means they tweaked the model's settings until its predictions matched what actually happened in the real world during those years. Then, they "validated" it using data from 2011 to 2017, which is like taking a final exam with questions you've never seen before to prove you really understand the subject. The results were impressive. The model's predictions were very close to the real river flows. In the language of science, they measured this using numbers like the Nash–Sutcliffe Efficiency (NSE), which ranged from 0.897 to 0.98 during their tests. Since a perfect score is 1.0, these numbers suggest the model was doing an excellent job. The "Root Mean Square Error" (RMSE), which measures how far off the predictions were, stayed low, between 0.2 and 1.36.

The researchers found that by combining the smart grouping of rain gauges with the territory-mapping of Thiessen Polygons, they could predict the river's behavior much better than before. Their model successfully caught the big spikes in water flow during heavy monsoon storms, matching the real-life peaks almost perfectly. This suggests that their approach is a reliable way to forecast floods in the Wainganga basin. While the model isn't perfect—it still relies on the quality of the data it receives and assumes rain is evenly spread within those "territories"—it offers a strong tool for managing water resources and warning people about potential floods. The study concludes that this mix of smart data sorting and computer modeling can help keep communities safer and manage water better in this part of India.

Technical Summary: Hydrological Modelling and Runoff Estimation of the Wainganga River Basin Using HEC-HMS and Geospatial Clustering Techniques

Problem Statement
The Wainganga River basin in central India, a major tributary of the Godavari system, is highly susceptible to destructive seasonal flooding caused by intense monsoon rainfall, undulating terrain, and expanding human settlements. Effective flood risk management requires reliable forecasting systems; however, these are often hindered by gaps in rainfall data due to equipment malfunction, transmission failures, and the impracticality of maintaining continuous data from all stations in large, heterogeneous catchments. Furthermore, climate variability has altered rainfall intensity and runoff patterns, complicating predictions. The study addresses the need to optimize the rain gauge network and improve runoff estimation accuracy to support flood management and early warning systems in this monsoon-dominated region.

Methodology
The research employs an integrated framework combining geospatial techniques, statistical clustering, and semi-distributed hydrological modeling:

Data Acquisition: The study utilized rainfall and discharge data from 2001 to 2017. Topographical data (DEM) was sourced from the National Remote Sensing Centre (NRSC), while hydrological discharge data was obtained from the Central Water Commission (CWC). Rainfall data was sourced from the Maharashtra State Agriculture Department and Madhya Pradesh Land Records department.
Rain Gauge Network Optimization:
- Hierarchical Clustering (HC): An agglomerative (bottom-up) approach using R software was applied to group 17 rain gauge stations based on rainfall similarity, elevation, and geographical proximity. This identified two primary clusters: an upper high-intensity zone (hilly regions like Seoni, Balaghat) and a middle-to-lower regulated zone (plains).
- Thiessen Polygon Method: Used to spatially divide the watershed into influence zones for each station. This allowed for the calculation of area-weighted average rainfall, assigning weights based on the polygon area surrounding each gauge.
Geospatial Processing (HEC-GeoHMS & ArcGIS):
- DEM data was processed to fill sinks, determine flow direction, and delineate sub-basins.
- Land Use/Land Cover (LULC) maps and soil data (FAO DSMW) were integrated to generate Curve Number (CN) grids using the SCS method.
- The basin was subdivided into hydrologic units, and physical parameters (length, slope, centroid) were extracted to create the HEC-HMS basin model.
Hydrological Modeling (HEC-HMS):
- The HEC-HMS (version 4.3) model was configured using the SCS Curve Number method for loss and the SCS Unit Hydrograph for transform.
- Calibration: The model was calibrated using observed monsoon data from 2001–2010 at the Asthi outlet station. Parameters such as CN and Lag Time were iteratively adjusted.
- Validation: The model was validated against independent data from 2011–2017 without further parameter modification.
Performance Evaluation: Model accuracy was assessed using Nash–Sutcliffe Efficiency (NSE), the coefficient of determination ( $R^2$ ), and Root Mean Square Error (RMSE).

Key Results
The study achieved high simulation precision across both calibration and validation periods:

Calibration (2001–2010): NSE values ranged from 0.895 to 0.962, with $R^2$ values consistently above 0.895. RMSE values varied between 0.2056 and 1.3620, with lower errors generally observed in years with less extreme peak fluctuations.
Validation (2011–2017): The model maintained robust performance, with NSE values ranging from 0.906 to 0.980 and $R^2$ values exceeding 0.90 for all years. RMSE values remained low, ranging from 0.3296 to 0.8707.
Hydrograph Comparison: Visual analysis confirmed that the simulated hydrographs closely matched observed discharge trends, accurately replicating rising limbs, peak discharge magnitudes, and recession curves for major monsoon events.
Clustering Utility: The hierarchical clustering successfully categorized stations into hydrologically distinct zones, justifying the use of differentiated parameter initialization for the upper hilly catchments versus the lower plains.

Significance and Claims
The paper claims that the combined application of hierarchical clustering, Thiessen polygon analysis, and HEC-HMS modeling significantly improves runoff estimation and flood forecasting accuracy in the Wainganga basin. The study demonstrates that:

Identifying representative rain gauge stations through clustering allows for more efficient network utilization and gap-filling in data-scarce scenarios.
The developed framework provides a reliable tool for operational flood forecasting, early warning systems, and water resource planning in monsoon-dominated river basins.
The model's ability to replicate complex monsoon dynamics suggests it can support flood management and mitigation planning for researchers and policymakers in central India.

Limitations
The authors acknowledge several constraints affecting the model's predictive accuracy:

Data Reliability: The model's output is intrinsically dependent on the quality and continuity of historical field logs from external agencies.
Station Density: While sufficient for validation, the current network density may not fully capture micro-climatic variations in the rugged Satpura Range; increasing gauge density would reduce interpolation errors.
Thiessen Assumptions: The method assumes uniform rainfall distribution within polygons, which may overlook orographic lifting effects in hilly upper catchments.
Temporal Scope: The study is limited to the 2001–2017 period; longer time series would be required to assess performance under extreme prolonged droughts or catastrophic historical flood stages.

More like this