Hydrology Newsletter

The story of modern hydrology does not begin with satellites, supercomputers, or deep learning models. It begins with a simple, yet revolutionary question: Is rainfall alone enough to explain the flow of a river? The answer, first demonstrated quantitatively in the Seine basin, quietly transformed hydrology from a discipline of speculation into a measurement‑driven science. hydrogeek.substack ​ A recent HydroGeek article, “Do you know who performed the first quantitative rainfall–runoff estimates?”, revisits this turning point and the scientist behind it, often described as a founder of experimental hydrology. This blog builds on that piece—zooming out to show why that early work still shapes how hydrologists think, model, and manage water today. hydrogeek.substack ​ The world before quantitative rainfall–runoff Before this pioneering study, many scholars and engineers doubted whether rainfall alone could sustain the perennial flow of major rivers. Explanations often invoked vag...

Engineering Optimization focuses on developing algorithms and techniques to find the best possible solution for complex engineering problems, such as optimizing the design of structures or processes. On the other hand, Polynomial Neural Networks explore the use of polynomial activation functions in neural networks, which can enhance their learning capabilities and improve their performance in certain tasks. Both areas contribute to advancing the capabilities of artificial intelligence and have significant applications in various industries.  The above video walkthrough tries to demonstrate how to use the advancement of polynomial neural networks in engineering optimization.  For example, in the field of civil engineering, optimization techniques can be applied to design more efficient and cost-effective buildings or bridges. By using algorithms and mathematical models, engineers can analyze different design variables such as material selection, structural configurations, and c...

Runoff Prediction Models Runoff Prediction Models(RPM) are those models which predict the runoff or flood peak of a watershed or input runoff to a dam etc. Generally, such models require climatic parameters, geomorphology, soil characteristics, and land cover as input against which a runoff model can predict the monthly, weekly, daily, or even hourly runoff. These models can be spatially distributed or lumped, temporally long or short, data-driven or conceptual.  In recent years due to the massive development in data-driven and smart concepts like artificial neural networks(ANN), decision trees, and evolutionary algorithms, application of such techniques are now common to develop RPMs. Among data-driven techniques, ANN is the most popular, followed by evolutional algorithms. But compared to the standalone application of neural networks, hybrid models where ANN with conceptual models like HyMOD or HEC is found to be more successful. Hydrologic models like RPMs need to be calibrated ...

The principal objective of hydrologic models is to forecast the runoff of a surface water body, especially dendritic systems like rivers, streams, etc. The inputs to these models are generally Rainfall/Precipitation, Soil Characteristics, and other Climatic parameters like evapotranspiration, humidity, etc. LULC and geo-morphology are also used as the required input parameters of the hydrologic models. Both input and output of these models are temporally as well as spatially variable. Now the resolution varies with different models. Some models consider all the sub-basins to be a single watershed and determine the output based on the characteristics of this single watershed(lumped).In contrast, some other models will consider the {impact|effect} of each of the sub-basins on the central outflow of the watershed(distributed).In a few models, the entire watershed is divided into grids or units of uniform dimension. However, the accuracy is highest for the models, which considers the {impa...