Automatic Calibration
Switch on Automatic Calibration by clicking Automatic at the bottom of the Calibration section. The Automatic Calibration page is similar to the Manual Calibration page, but with a few key differences. It allows you to streamline and automate the process of adjusting model parameters to achieve optimal performance.
Basic Configuration
The following settings apply to all algorithms:
Processes
Specify the number of parallel processes to be used during calibration. Increasing this value may speed up the calibration but could also require more computational resources.
Calibration Observation
Select the observation against which you are calibrating. This is typically the dataset that you want your model to replicate accurately.
Multi-Site Calibration
Enable this option if you wish to optimise several observations at once. When doing so, you will need to provide weighting factors for each observation to balance their relative importance.
Objective Function
Select the objective function that will be used to measure calibration performance. Options include: - pBias - NSE (Nash–Sutcliffe Efficiency) - KGE (Kling–Gupta Efficiency) - RMSE (Root Mean Squared Error) - MSE (Mean Squared Error)
Choose the one most appropriate for your calibration goals.
Algorithms
There are three algorithms available in the SWAT+ Toolbox for Automatic Calibration:
CALSI Algorithm
Further details on the CALSI Algorithm will follow (a paper is in progress).
DREAM Algorithm
The DiffeRential Evolution Adaptive Metropolis (DREAM) algorithm is a Markov Chain Monte Carlo (MCMC) method designed to robustly explore a high-dimensional parameter space. It uses multiple chains and adapts the proposal distribution based on sampling history, which helps it converge more reliably on optimal parameter sets. For an in-depth explanation, refer to the work of Vrugt et al. (2009) or an equivalent online resource.
DDS Algorithm
The Dynamically Dimensioned Search (DDS) algorithm systematically searches the parameter space by progressively narrowing its focus on the most sensitive parameters. As the calibration progresses, DDS refines its search to improve the model’s performance in fewer iterations. For more information, see Tolson and Shoemaker (2007) or consult an online reference.
Monitoring Progress
During the Automatic Calibration process, the following tools help you track and assess ongoing performance:
Status
Displays all the steps involved in the calibration process and indicates the current step. This helps in understanding the progression and any potential bottlenecks.
Progress
Reports the model’s performance, based on the selected objective function, for each parallel process. You can use this to compare how different runs are evolving.
Charts
Presents graphs for the best parameter set as the calibration proceeds. These charts offer real-time insights into how the chosen parameter set is improving (or not) over time.
Dotty Plots
Continuously updates to illustrate how each parameter behaves and reveals its relative sensitivity. This visual aid helps identify which parameters drive the most significant changes in the model’s performance.