@Article{Limber_WRR_20250920,
 author		= {Russ Limber and Forrest M. Hoffman and Jon Schwenk and Jitendra Kumar},
 title		= {Long Short-Term Memory Model to Forecast River Ice Breakup Throughout {A}laska {USA}},
 journal	= WRR,
 volume		= 61,
 number		= 9,
 pages		= {e2025WR040635},
 doi		= {10.1029/2025WR040635},
 day		= 20,
 month		= sep,
 year		= 2025,
 abstract	= {The annual breakup of river ice in Arctic regions poses significant risk of ice jam flooding, causing property damage, altering ecosystems, and jeopardizing inhabitants. Predicting the timing of the annual breakup is crucial for residents to prepare for potential flooding and assess the safety of rivers for transportation. This analysis develops a deep learning algorithm using widely available meteorological and geospatial data products to forecast river ice breakup. We selected 33 locations along eight major rivers across Alaska, USA, and Western Canada, leveraging annual breakup dates from the Alaska-Pacific River Forecast Center database. Daily meteorological data from Daymet and static watershed attributes from the pan-Arctic catchment database were used to develop a Long Short-Term Memory (LSTM) model for predicting river ice breakup. Of the 33 locations, 23 were used for tuning, training and testing the LSTM. The model demonstrated high efficacy, predicting the annual breakup date with a mean absolute error (MAE) of 5.40 days, standard deviation of 4.03 days and mean absolute percentage error (MAPE) of 4.37\%. The spatial generalizability of the LSTM was evaluated using the remaining 10 locations as holdouts, with most locations showing MAPE $\lt$8\% over the entire time series (1980--2023). Additionally, we retrieved 51 long-range seasonal forecast ensembles from the Copernicus Climate Data Store and applied the trained model to them to showcase the capability of the LSTM to predict future river ice breakup using operational weather forecasts. LSTM was able to predict the breakup dates within 5--14 days of observed breakup.}
}