Investigating the Impact of Feature Reduction for Deep Learning-based Seasonal Sea Ice Forecasting

Lars Uebbing, Harald Lykke Joakimsen, Luigi Tommaso Luppino, Iver Martinsen, Andrew McDonald, Kristoffer Knutsen Wickstrøm, Sébastien Lefèvre, Arnt B. Salberg, Scott Hosking, Robert Jenssen
Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL), PMLR 265:245-254, 2025.

Abstract

With the state-of-the-art IceNet model, deep learning has contributed to an important aspect of climate research by leveraging a range of climate inputs to provide accurate forecasts of Arctic sea ice concentration (SIC). The deep learning subfield of eXplainable AI (XAI) has gained enormous attention in order to gauge feature importance of neural networks, for instance by leveraging network gradients. In recent work, an XAI study of the IceNet was conducted, using gradient saliency maps to interrogate its feature importance. A majority of XAI studies provide information about feature importance as revealed by the XAI method, but rarely provide thorough analysis of effects from reducing the number of input variables. In this paper, we train versions of the IceNet with drastically reduced numbers of input features according to results of XAI and investigate the effects on the sea ice predictions, on average and with respect to specific events. Our results provide evidence that the model generally performs better when less features are used, but in case of anomalous events, a larger number of features is beneficial. We believe our thorough study of the IceNet in terms of feature importance revealed by XAI may give inspiration for other deep learning-based problem scenarios and application domains.

Cite this Paper

BibTeX
@InProceedings{pmlr-v265-uebbing25a, title = {Investigating the Impact of Feature Reduction for Deep Learning-based Seasonal Sea Ice Forecasting}, author = {Uebbing, Lars and Joakimsen, Harald Lykke and Luppino, Luigi Tommaso and Martinsen, Iver and McDonald, Andrew and Wickstr{\o}m, Kristoffer Knutsen and Lef{\`e}vre, S{\'e}bastien and Salberg, Arnt B. and Hosking, Scott and Jenssen, Robert}, booktitle = {Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL)}, pages = {245--254}, year = {2025}, editor = {Lutchyn, Tetiana and Ramírez Rivera, Adín and Ricaud, Benjamin}, volume = {265}, series = {Proceedings of Machine Learning Research}, month = {07--09 Jan}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v265/main/assets/uebbing25a/uebbing25a.pdf}, url = {https://proceedings.mlr.press/v265/uebbing25a.html}, abstract = {With the state-of-the-art IceNet model, deep learning has contributed to an important aspect of climate research by leveraging a range of climate inputs to provide accurate forecasts of Arctic sea ice concentration (SIC). The deep learning subfield of eXplainable AI (XAI) has gained enormous attention in order to gauge feature importance of neural networks, for instance by leveraging network gradients. In recent work, an XAI study of the IceNet was conducted, using gradient saliency maps to interrogate its feature importance. A majority of XAI studies provide information about feature importance as revealed by the XAI method, but rarely provide thorough analysis of effects from reducing the number of input variables. In this paper, we train versions of the IceNet with drastically reduced numbers of input features according to results of XAI and investigate the effects on the sea ice predictions, on average and with respect to specific events. Our results provide evidence that the model generally performs better when less features are used, but in case of anomalous events, a larger number of features is beneficial. We believe our thorough study of the IceNet in terms of feature importance revealed by XAI may give inspiration for other deep learning-based problem scenarios and application domains.} }
Endnote
%0 Conference Paper %T Investigating the Impact of Feature Reduction for Deep Learning-based Seasonal Sea Ice Forecasting %A Lars Uebbing %A Harald Lykke Joakimsen %A Luigi Tommaso Luppino %A Iver Martinsen %A Andrew McDonald %A Kristoffer Knutsen Wickstrøm %A Sébastien Lefèvre %A Arnt B. Salberg %A Scott Hosking %A Robert Jenssen %B Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL) %C Proceedings of Machine Learning Research %D 2025 %E Tetiana Lutchyn %E Adín Ramírez Rivera %E Benjamin Ricaud %F pmlr-v265-uebbing25a %I PMLR %P 245--254 %U https://proceedings.mlr.press/v265/uebbing25a.html %V 265 %X With the state-of-the-art IceNet model, deep learning has contributed to an important aspect of climate research by leveraging a range of climate inputs to provide accurate forecasts of Arctic sea ice concentration (SIC). The deep learning subfield of eXplainable AI (XAI) has gained enormous attention in order to gauge feature importance of neural networks, for instance by leveraging network gradients. In recent work, an XAI study of the IceNet was conducted, using gradient saliency maps to interrogate its feature importance. A majority of XAI studies provide information about feature importance as revealed by the XAI method, but rarely provide thorough analysis of effects from reducing the number of input variables. In this paper, we train versions of the IceNet with drastically reduced numbers of input features according to results of XAI and investigate the effects on the sea ice predictions, on average and with respect to specific events. Our results provide evidence that the model generally performs better when less features are used, but in case of anomalous events, a larger number of features is beneficial. We believe our thorough study of the IceNet in terms of feature importance revealed by XAI may give inspiration for other deep learning-based problem scenarios and application domains.
APA
Uebbing, L., Joakimsen, H.L., Luppino, L.T., Martinsen, I., McDonald, A., Wickstrøm, K.K., Lefèvre, S., Salberg, A.B., Hosking, S. & Jenssen, R.. (2025). Investigating the Impact of Feature Reduction for Deep Learning-based Seasonal Sea Ice Forecasting. Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL), in Proceedings of Machine Learning Research 265:245-254 Available from https://proceedings.mlr.press/v265/uebbing25a.html.

Related Material