Datasets for forecasts

[nicl-nno]

Thank you for your interesting work - this is a very impressive contribution!

I would like to kindly ask whether it's possible to obtain the NetCDF files corresponding to the forecasts used to evaluate the metrics reported in Table 3.

I'm interested in evaluating additional metrics for ice forecasts and comparing it with my own model' results. However, re-running all SOTA models and retraining FH-Mamba would be quite challenging.

Thank you in advance for your help!

[jialiangZ]

In fact, our indicator calculation is derived by comparing the predicted results with the actual values. We did not save the predicted results as a netcdf file.

[nicl-nno]

Thank you for the answer! If so, is it possible to share the weights of trained models used to obtain the predictions? Just to make our comparison more correct.

[ghost-him]

Thank you for your interest in our work!

Unfortunately, due to subsequent experiments and updates to the codebase, I am no longer able to locate the original weight files for the AMSR2 and OSI-450-a1 datasets as reported in the paper.

However, I have managed to find the weights from an ablation study previously conducted on the OSI-450-a1 dataset. I have packaged them and verified that they run correctly on my local machine. You can find them here: https://github.com/oucailab/FH-Mamba/releases/tag/data. Please note that this version has a smaller parameter count, and its accuracy may be slightly lower than the results shown in the paper.

Additionally, for the AMSR2 dataset preprocessing, you can refer to this toolbox: https://github.com/ghost-him/AMSR2-Sea-Ice-Toolbox. It employs a superior interpolation algorithm compared to the one used in my original experiments, which helps achieve better prediction accuracy by handling missing values more effectively.

I hope these resources are helpful for your research!

[nicl-nno]

Thank you, this data be very helpful for us!

And we will try to use toolbox for AMSR interpolation too.