Kharun river at Mahadev Ghat, Raipur
Makarand Purohit
In central Chhattisgarh, the Kharun River has rarely commanded national attention. It does not roar through mountains or vanish into desert sands. Instead, it has long been counted on, steady enough to irrigate rice fields, recharge aquifers, supply expanding towns, and quietly drain monsoon excess into the Mahanadi system. Its strength has always been its predictability. That predictability is beginning to fray.
A new peer-reviewed study in Environmental and Sustainability Indicators by Vipin Kumar Mishra and colleagues applies advanced machine-learning methods to examine how climate change and land-use transformation are reshaping water availability in the Kharun River Basin. The findings challenge conventional assumptions. The basin is projected to receive more total water in the future, yet become less secure.
This emerging paradox is stark: higher runoff, increased flooding, and greater sediment loads alongside intensified dry-season stress. In other words, abundance without assurance. The Kharun’s future mirrors a wider pattern unfolding across India’s monsoon heartland, where climate variability and landscape change are altering not just how much water flows, but when and how reliably it can be used.
The basin and what is at stake
The Kharun River Basin spans about 4,143 square kilometres across Balod, Durg, Dhamtari, and Raipur districts in Chhattisgarh. Farming dominates the landscape, with rice as the main crop. Forests, scrublands, and wetlands have traditionally helped regulate river flows by absorbing rainfall, slowing runoff, and supporting groundwater recharge.
Over the past three decades, this balance has shifted. Raipur has expanded rapidly, and agriculture has intensified. Built-up areas have grown nearly five times since the early 1990s. Forest cover has declined, and fallow lands have been converted into farms or settlements. These changes are significant because rivers respond not only to how much rain falls but also to how the land manages that rain. When vegetation is removed or soil is compacted, water runs off more quickly. Climate change interacts with these altered conditions, strengthening their impact.
Until recently, most basin studies in central India relied on process-based hydrological models. These models often struggle in regions with limited data and complex interactions between rainfall, land cover, soils, and river flow. This study takes a different approach. It asks whether machine learning, using long-term climate records, satellite-based land-use maps, soil data, slope information, and river observations, can better explain how the basin is changing.
Why machine learning offers new insight
The researchers evaluated five machine learning methods: Random Forest, Artificial Neural Networks, Support Vector Machines, k-Nearest Neighbours, and Extreme Gradient Boosting, known as XGBoost. Among them, XGBoost performed best. It showed high accuracy in predicting both water yield and sediment yield, with a Nash-Sutcliffe Efficiency of 0.89 and an R-squared value of 0.95 during validation.
This is important because river systems rarely respond in simple or proportional ways. A small rise in rainfall intensity can produce a much larger increase in runoff when forests are cleared or soils are hardened. XGBoost is particularly effective at detecting these nonlinear responses and threshold effects.
The researchers used the model in two stages. First, they reconstructed changes in water yield between 1992 and 2022 as land use evolved. Second, they projected future conditions using CMIP6 climate scenarios, specifically SSP2 4.5 and SSP5 8.5, combined with realistic land use pathways.
What the past three decades reveal
The historical analysis shows that the basin is becoming more prone to sudden and intense flows. During the monsoon, especially in August and September, water yield has increased significantly. This rise reflects greater surface runoff from expanding urban areas and more intensively farmed land.
At the same time, dry season flows have declined. In winter months such as January and December, baseflows have dropped sharply, in some places by more than 40 percent. This pattern signals a weakening ability of the landscape to store and slowly release water. When rainfall occurs, more of it now moves quickly across the surface instead of soaking into the soil. When the rains end, there is less stored water to maintain steady river flow.
Sediment levels have also increased. Faster runoff over exposed or disturbed land carries more soil into the river. This raises sediment transport and speeds up siltation in downstream reservoirs and irrigation systems. These changes have practical consequences. They mean higher flood risks, shorter reservoir lifespans, rising costs for water treatment, and greater dependence on groundwater at a time when natural recharge is declining.
The climate signal: Intensification without security
Future projections using CMIP6 climate scenarios deepen these concerns. Under the SSP2 4.5 pathway, annual water yield in the Kharun Basin could rise by as much as 135 percent by the end of the century compared to historical levels. Under the higher emission SSP5 8.5 scenario, increases may reach nearly 120 percent before stabilising later in the century.
However, when this water arrives is as important as how much arrives. June, which marks the beginning of the monsoon and is crucial for agriculture, shows declining water yield in several projected periods. This suggests weaker or delayed early monsoon flows. In contrast, late monsoon and post-monsoon months show very large increases in runoff. September flows rise sharply, signalling greater flood risk.
Dry season patterns also become less stable. Some winter months show increases in simulated water yield, while others, especially under SSP5 8.5, show declines. This variability complicates planning for irrigation, drinking water supply, and reservoir management.
Climate change is therefore pushing the basin towards greater extremes, with excess water during periods that are difficult to manage and uncertainty during periods when water is most needed.
Why more water can reduce resilience
The study points to a major policy gap. Water security is often measured by total annual availability. In monsoon-dependent systems, timing and storage capacity are more critical than totals alone.
Higher annual water yield without corresponding storage and infiltration capacity leads to:
More frequent and damaging floods, especially in urban and peri-urban zones.
Faster sedimentation of tanks and reservoirs, reducing effective storage.
Declining groundwater recharge, increasing dry-season vulnerability.
Greater inequality, as downstream towns face floods while upstream farmers face water stress.
This is not a future problem. It is already unfolding, and climate change is accelerating it.
What needs to change: From drainage to retention
The study’s greatest value lies in its implications for action. Several clear priorities emerge.
Reframe land use as water infrastructure: Forests, fallows, wetlands, and low-intensity agricultural lands function as hydrological regulators. Protecting and restoring them is as important as building canals or reservoirs. In the Kharun Basin, mid-catchment recharge zones and riparian buffers should be formally identified and safeguarded through zoning regulations.
Shift urban planning from evacuation to absorption: Cities like Raipur have treated stormwater as a nuisance to be drained away. That approach is no longer viable. Permeable pavements, bioswales, rain gardens, and infiltration parks can slow runoff, reduce flooding, and enhance recharge. These interventions are cheaper and more flexible than hard flood-control structures.
Massively scale decentralised storage: The projections show that monsoon surpluses are increasing. Capturing even a fraction of this water through farm ponds, check dams, percolation tanks, and revived traditional tanks could transform dry-season availability. Storage must be distributed, not centralised, to reduce evaporation and sediment risks.
Redesign irrigation and cropping strategies: Traditional calendars assume a predictable monsoon onset and gradual recession. That assumption no longer holds. Irrigation advisories, crop choices, and sowing dates must be updated using seasonal forecasts and basin-specific projections such as those generated by this model.
Integrate sediment management into water planning: Rising sediment yield threatens every storage structure. Catchment treatment—contour bunding, vegetative barriers, reduced tillage—must be treated as core water-security investments, not peripheral environmental add-ons.
Institutionalise machine-learning decision support: XGBoost and similar models should be embedded in basin-level decision-support systems. Used alongside rainfall forecasts and ground observations, they can support flood warnings, reservoir operation, and land-use planning at district and block scales.
Democratise monitoring and data: The authors point to participatory monitoring and citizen science as critical complements to modelling. Local rainfall observers, community sediment monitoring, and open data platforms can improve model accuracy while building public trust in water governance.
Beyond the Kharun: A national warning
The Kharun River is not an exception. It is a preview. Across central and eastern India, monsoon-fed basins are experiencing the same convergence of climate intensification and land-use change. What distinguishes this study is its clarity: climate change will not simply reduce water availability. It will rearrange it—seasonally, spatially, and socially.
Machine learning offers a powerful lens to see this rearrangement in advance. But foresight only matters if it informs action. Without a decisive shift towards retention-based, landscape-sensitive water management, rivers like the Kharun will continue to grow louder during the monsoon—and fall silent when communities need them most. The science is no longer the bottleneck. Governance is.