Unexpected streamflow spike when switching forcing data

[Noah H]

Hello WRF-Hydro users,
I am currently working on a flash flood simulation using WRF-Hydro v5.1.1, and I am experiencing an unexpected issue related to streamflow behavior when switching forcing datasets.

Details
- Model: WRF-Hydro v5.1.1 (standalone offline mode)
- Domain size: 800 km²
- Simulation period: 2023-01-01 to 2023-12-31 (hourly forcing input & hourly CHRTOUT output)
- Objective: Flash flood simulation

Previously, I used GLDAS forcing data, and the model produced reasonable streamflow results without any major issues. After replacing GLDAS forcing with GLDAS ensemble forcing (processed to match WRF-Hydro input format), I observed an abnormal behavior in the simulated streamflow (CHRTOUT_DOMAIN1):
- Streamflow suddenly spikes at the very beginning of the simulation (around Jan 2, 08–09)
- After that, discharge remains unrealistically high (around ~1200–1700 m³/s) even during no-rain periods
- This behavior persists regardless of the simulation start date (e.g., starting from Jan 1, Jan 3, or even May)

To investigate the issue, I compared key forcing variables (PSFC, T2D, Q2D, U2D, V2D, SWDOWN, LWDOWN, RAINRATE) between GLDAS and ensemble data around the spike period. But there were no obvious anomalies found.

I performed several tests to isolate the cause: Changing simulation start date, Turning off groundwater (GWBASESWCRT = 0), or Warm start using repeated forcing (treadmill spin-up). However, the same initial spike still appears.

At this point, I am unsure whether the issue comes from. I would greatly appreciate any insights; anyone experienced similar initial streamflow spikes when switching forcing datasets (especially ensemble forcing datasets?)

Any suggestions, similar experiences, or troubleshooting ideas would be greatly appreciated.

Thank you in advance for your help!