Difficulty in Observed data Limitation

[Sarangan K]

Hello everyone,

I am currently working on SWAT model calibration and validation using SWAT-CUP (SUFI-2) for a semi-arid watershed in India.

During validation, I obtained the following performance statistics for FLOW_OUT_2:

p-factor = 0.26
r-factor = 0.70
R² = 0.01
NSE = −1.90
KGE = −0.12
PBIAS = −52

Although the simulated hydrograph follows the general seasonal pattern, the statistical indicators (especially NSE and R²) are very poor.

Some background information:
• Study period: 1994–2024
• Calibration: 2004–2017
• Validation: 2018–2024
• The observed inflow dataset contained missing daily values, which were filled using interpolation only within months having available data.
• The watershed is located in a semi-arid region with highly variable rainfall-runoff response.

My questions are:

1. I derived this inflow from the reservoir capacity and outflow releases as there is no available data for the calibration process in the basin. Will it affect this?

2. Is it acceptable to proceed with water balance analysis and scenario assessment when NSE is negative but the seasonal flow pattern is reasonably captured?

3. Could the data can be  interpolated ?

4. Some data values are zero but in realistic the flow will be generated for that day, how can I overcome that?

5. Are there any recommended approaches to improve calibration for such datasets with limited observed flow?

I would appreciate any suggestions or experiences from others who have faced similar issues. I would also like to share the data if you could look into it.

Thank you.