Seven New Runoff Prediction Models

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 and validated with ground-level primary data and also outputs are compared with the same output from other models. AutoRegressive Integrated Moving Average(ARIMA) was found to be the most used model for comparison.

In the case of selection of RPMs, various statistical indices are used like Root Mean Square Error(RMSE), Mean Absolute Error (MAE), Mean Relative Error(MRE), Nash Sutcliffe Coefficient(NS), Nash Sutcliffe Efficiency(NSE), BIAS, etc. among which NSE and MAR were found to be the most widely used error functions.

Seven Most Recent RPM

The seven most recent Runoff Prediction Models(RPM) are selected based on their accuracy, reliability, ease of use, and recentness. 

1. Non Homogeneous Markov Chain Runoff Prediction Model(NHMC-RPM)
2. Optimized SWAT-ANN
3. Particle Swarm Optimization(PSO)- Long Short-Term Memory (LSTM) Model
4. Support Vector Machine(SVM)
5. Identification of unit Hydrograph and Component flows from Rainfall, Evapotranspiration, and Streamflow(IHACRES)-Group Method of Data Handling(GMDH) Hybrid
6. Combined Forecasting Model(CFM) : Extreme Learning Machine(ELM) with PSO 
7. Auto-Regressive Moving Average(ARMA)

Click to view the schematics of each model.

If you had found some interesting RPMs published within the last year(2021-22) share them by posting a comment to this post.

Thanks for reading,

@Mrinmoy's Page
Share your data requirement, 

@Merchandises or @Shop
Become my friend in Listy/Pearltrees/Twitter