Questions regarding calibration workflow and sensitivity analysis in SWAT+ (Discharge Data)

[Nadya Salma Azzahra]

Currently, I am working on a watershed modeling project using SWAT+, and the only observation data I have is discharge data.

Regarding the next steps, I have a few questions. I would greatly appreciate some insights from those of you with more experience:

1. What is the best reference for determining the minimum and maximum values for sensitive parameters?

2. The toolbox provides several methods such as Latin Hypercube, Sobol, FAST, RBD-FAST, and Delta Moment. What are the fundamental differences between these methods, and which one is most recommended for a general case?

3.  When conducting a sensitivity analysis, what is the guideline for determining the ideal number of seeds/samples?

4. There are algorithm options like CALSI, DDS, and DREAM. What are the characteristics and differences of each method? Which one is generally the most stable for limited data?

5. When performing automatic calibration, is there a rule of thumb regarding how many iterations (simulations) to input so that the results converge while keeping the computation efficient?

6. Is it recommended to strictly use automatic calibration, or is the best practice to mix it with manual calibration?

7. For the model validation stage in the SWAT+ ecosystem, where exactly is it usually done in the workflow, and how do you set it up?