SWAT+ Toolbox sensitivity analysis returns NaN while model and auto-calibration run successfully

[Namraj Jaishi]

Hello SWAT+ Toolbox  family

I am having an issue with the SWAT+ Toolbox sensitivity analysis and would appreciate any advice.

I am using SWAT+ Editor 3.2.3 and SWAT+ Toolbox v3.2.0. My project runs successfully in the SWAT+ Editor, and the model output is imported into the database without errors. I also confirmed that the model produces valid monthly channel output for the channel I am using.

Project summary:

• Simulation period: 2003–2025

• Warm-up period: 3 years

• Evaluation/observation period used: 2000–2025

• Observation: Channel 17 Monthly River Flow

• Observation file: monthly USGS flow in m³/s

• Sensitivity method: Random Balance Designs Fourier Amplitude

• Objective functions tested: NSE and MARE

The model runs successfully, and auto-calibration also runs fine, producing statistics. However, when I run Sensitivity Analysis using observations, the Toolbox aborts and reports that the objective function evaluation yielded NaN values.

The error message says something like:

“Sensitivity Analysis Aborted!
NSE evaluation for observations for Channel 17 Monthly River Flow yielded ‘NaN’ values.”

I also tried MARE, but I received the same type of NaN error.

I checked the following:

• The observation CSV loads correctly and plots in the Toolbox.

• The observation dates are monthly and formatted as month-end dates, for example, 2003-01-31, 2003-02-28, etc.

• The observation file has no missing or zero values.

• The simulated monthly channel output exists for Channel 17.

• The simulated and observed data overlap from 2003-01-31 to 2025-12-31.

• The model run and auto-calibration run successfully.

Has anyone seen this issue before? Are there any additional files, settings, or output print options I should check for sensitivity analysis with monthly channel flow observations?

Thank you,

Namraj