reproducing the prediction results of the aurora 0.1-degree model

Hello! I am currently reproducing the prediction results of the aurora 0.1-degree model. I have noticed that the predicted results of the aurora 0.1-degree model exhibit unreasonable and significant fluctuations in variables at near-surface pressure levels (e.g., u 1000hPa, v 1000hPa, temp 1000hPa). In a single forecast, there are notably unreasonable fluctuations in prediction accuracy—sometimes good, sometimes poor. This phenomenon of inconsistent performance is quite common. Meanwhile, other atmospheric variables at different pressure levels, as well as surface variables, demonstrate relatively good accuracy and stable prediction results.

Initially, I suspected that this might be caused by the inability of the official aurora 0.1-degree static variables to simulate static variables under complex topographic conditions, as I compared the provided 0.1-degree static file with the ECMWF HRES analysis static variable data and found significant differences in values, corresponding locations, and scales—particularly in topographic height (z).

However, even in plain areas (where the terrain is relatively simple), the same issue persists. I am quite puzzled by this. Is this an inherent issue with the aurora 0.1-degree model itself? I hope you can clarify this for me. I have carefully reviewed the aurora package provided by you and my aurora 0.1-degree prediction script but have been unable to identify the source of the problem. If this issue arises from errors in my reproduction process, I would greatly appreciate it if you could provide reasonable guesses or suggestions. Thank you very much!

Start by reviewing the Aurora package and the prediction script used for the 0.1-degree reproduction. Re-run a forecast and compare the near-surface u, v, and temperature variables with other pressure levels and surface variables, including the provided static file and ECMWF HRES static data. Done means determining whether the fluctuations come from the reproduction process or the model, with evidence to support the conclusion.