基于LSTM的北極海冰范圍多步預(yù)測策略研究

作者：王漫漫1 2 3 4  鄒斌2 3  石立堅2 3  曾韜2 3  張穎2 3  路敦旺1 2 3

單位：1. 國家海洋環(huán)境預(yù)報中心, 北京 100081; 2. 國家衛(wèi)星海洋應(yīng)用中心, 北京 100081; 3. 自然資源部 空間海洋遙感與應(yīng)用研究重點實驗室, 北京 100081; 4. 國家海洋環(huán)境預(yù)報中心 自然資源部海洋災(zāi)害預(yù)報技術(shù)重點實驗室, 北京 100081

關(guān)鍵詞：北極海冰范圍 長短期記憶網(wǎng)絡(luò) 多步預(yù)測策略 Seq2Seq策略

分類號：P731.32

出版年·卷·期（頁碼）：2025·42·第一期（11-22）

摘要：

已有研究對北極海冰范圍開展單步預(yù)測，而多步預(yù)測及其策略研究有待進一步探索。使用1979—2022年的北極月平均海冰范圍數(shù)據(jù)，采用長短期記憶網(wǎng)絡(luò)（LSTM）深度學(xué)習(xí)方法，結(jié)合遞歸（Recursive）、直接（Direct）、多輸入多輸出（MIMO）和Seq2Seq策略實現(xiàn)對未來12個月北極海冰范圍的多步預(yù)測。結(jié)果表明：24個月為模型的最佳輸入長度；與另外3種基本的多步預(yù)測策略相比，Seq2Seq策略對12個月北極海冰范圍預(yù)測的準(zhǔn)確性更好，均方根誤差為3.30×105 km2。

While previous researches have primarily focused on single-step prediction of Arctic sea ice extent, multi-step prediction and strategy are yet to be explored. This study utilizes monthly average Arctic sea ice extent data spanning from 1978 to 2022 and employs Long Short-Term Memory to implement multi-step predictions of Arctic sea ice extent for the next 12 months using four strategies: Recursive, Direct, Multi-input Multi-output, and Seq2Seq. The results show that a model input length of 24 months performs optimally. When compared to the other three basic multi-step prediction strategies, the Seq2Seq strategy demonstrates superior accuracy in forecasting Arctic sea ice extent over the next 12 months, with an root mean square error of 0.33 million square kilometers.

參考文獻：

[1] RANTANEN M, KARPECHKO A Y, LIPPONEN A, et al. The Arctic has warmed nearly four times faster than the globe since 1979[J]. Communications Earth & Environment, 2022, 3(1): 168. [2] DAVY R, GRIEWANK P. Arctic amplification has already peaked [J]. Environmental Research Letters, 2023, 18(8): 084003. [3] DELHAYE S, MASSONNET F, FICHEFET T, et al. Dominant role of early winter Barents-Kara sea ice extent anomalies in subsequent atmospheric circulation changes in CMIP6 models[J]. Climate Dynamics, 2024, 62(4): 2755-2778. [4] 孫曉宇, 孫啟振, 沈輝, 等. 1979—2022年北極海冰范圍變化特征及趨勢分析[J]. 海洋預(yù)報, 2023, 40(5): 49-55. SUN X Y, SUN Q Z, SHEN H, et al. Analysis of characteristics and trends of Arctic sea ice extent changes from 1979 to 2022[J]. Marine Forecasts, 2023, 40(5): 49-55. [5] SUN X Y, LV T T, SUN Q Z, et al. Analysis of spatiotemporal variations and influencing factors of sea ice extent in the Arctic and Antarctic[J]. Remote Sensing, 2023, 15(23): 5563. [6] JAHN A, HOLLAND M M, KAY J E. Projections of an ice-free Arctic Ocean[J]. Nature Reviews Earth & Environment, 2024, 5(3): 164-176. [7] GAO Y Q, SUN J Q, LI F, et al. Arctic sea ice and Eurasian climate: a review[J]. Advances in Atmospheric Sciences, 2015, 32(1): 92-114. [8] CHRIPKO S, MSADEK R, SANCHEZ-GOMEZ E, et al. Impact of reduced Arctic sea ice on Northern Hemisphere climate and weather in autumn and winter[J]. Journal of Climate, 2021, 34(14): 5847-5867. [9] SÉVELLEC F, FEDOROV A V, LIU W. Arctic sea-ice decline weakens the Atlantic Meridional Overturning Circulation[J]. Nature Climate Change, 2017, 7(8): 604-610. [10] ESAU I, PETTERSSON L H, CANCET M, et al. The Arctic amplification and its impact: a synthesis through satellite observations[J]. Remote Sensing, 2023, 15(5): 1354. [11] 劉泉宏, 張韌, 汪楊駿, 等. 深度學(xué)習(xí)方法在北極海冰預(yù)報中的應(yīng)用[J]. 大氣科學(xué)學(xué)報, 2022, 45(1): 14-21. LIU Q H, ZHANG R, WANG Y J, et al. Application of deep learning methods to Arctic sea ice prediction[J]. Transactions of Atmospheric Sciences, 2022, 45(1): 14-21. [12] 劉驥平, 雷瑞波, 宋米榮, 等. 適應(yīng)極地快速變化海冰模式的研發(fā)與挑戰(zhàn)[J]. 大氣科學(xué)學(xué)報, 2021, 44(1): 12-25. LIU J P, LEI R B, SONG M R, et al. Development and challenge of sea ice model adapting to rapid polar sea ice changes[J]. Transactions of Atmospheric Sciences, 2021, 44(1): 12-25. [13] WANG L, YUAN X J, TING M F, et al. Predicting summer Arctic sea ice concentration intraseasonal variability using a vector autoregressive model[J]. Journal of Climate, 2016, 29(4): 1529-1543. [14] CHI J, KIM H C. Prediction of Arctic sea ice concentration using a fully data driven deep neural network[J]. Remote Sensing, 2017, 9(12): 1305. [15] ALI S, HUANG Y Y, HUANG X, et al. Sea ice forecasting using attention-based ensemble LSTM[EB/OL]. arXiv: 2108.00853, (2022-02-08)[2023-10-11]. http://arxiv.org/abs/2108.00853. [16] KIM E, KRUSE P, LAMA S, et al. Multi-task deep learning based spatiotemporal Arctic sea ice forecasting[C]//2021 IEEE International Conference on Big Data (Big Data). Orlando: IEEE, 2021: 1847-1857. [17] National Snow and Ice Data Center. Sea ice concentrations from nimbus-7 SMMR and DMSP SSM/I-SSMIS passive microwave data, Version 2[DS/OL]. [2023-10-25]. https://nsidc.org/data/nsidc-0051/versions/2. [18] Climate Data Store. Seasonal forecast monthly statistics on single levels[DS/OL]. 2018[2024-03-26]. https://cds.climate.copernicus.eu/doi/10.24381/cds.68dd14c3. [19] HOCHREITER S, SCHMIDHUBER J. Long short-term memory [J]. Neural Computation, 1997, 9(8): 1735-1780. [20] Understanding LSTM networks[EB/OL]. Colah’s Blog, 2015[2024-12-26]. https://colah.github.io/posts/2015-08-Understanding-LSTMs/. [21] TAIEB S B, BONTEMPI G, ATIYA A F, et al. A review and comparison of strategies for multi-step ahead time series forecasting based on the NN5 forecasting competition[J]. Expert Systems with Applications, 2012, 39(8): 7067-7083. [22] SORJAMAA A, HAO J, REYHANI N, et al. Methodology for long-term prediction of time series[J]. Neurocomputing, 2007, 70(16-18): 2861-2869. [23] HAMZAÇEBI C, AKAY D, KUTAY F. Comparison of direct and iterative artificial neural network forecast approaches in multiperiodic time series forecasting[J]. Expert Systems with Applications, 2009, 36(2): 3839-3844. [24] TAIEB S B, SORJAMAA A, BONTEMPI G. Multiple-output modeling for multi-step-ahead time series forecasting[J]. Neurocomputing, 2010, 73(10-12): 1950-1957. [25] BONTEMPI G. Long term time series prediction with multi-input multi-output local learning[C]//Proceedings of the 2nd European Symposium on Time Series Prediction (TSP). Helsinki: Helsinki University of Technology, Department of Information and Computer Science, 2008: 145-154. [26] SUTSKEVER I, VINYALS O, LE Q V. Sequence to sequence learning with neural networks[EB/OL]. arXiv: 1409.3215, (2014-12-14)[2023-10-11]. http://arxiv.org/abs/1409.3215. [27] 王濤, 張偉, 賈宇欣, 等. 基于Seq2Seq模型的港口進出口貨物量預(yù)測[J]. 計算機系統(tǒng)應(yīng)用, 2020, 29(3): 132-139. WANG T, ZHANG W, JIA Y X, et al. Prediction on import and export goods volume of ports based on Seq2Seq model[J]. Computer Systems & Applications, 2020, 29(3): 132-139. [28] KIM Y J, KIM H C, HAN D, et al. Prediction of monthly Arctic sea ice concentrations using satellite and reanalysis data based on convolutional neural networks[J]. The Cryosphere, 2020, 14(3): 1083-1104. [29] CHI J, BAE J, KWON Y J. Two-stream convolutional long-and short-term memory model using perceptual loss for sequence-tosequence Arctic sea ice prediction[J]. Remote Sensing, 2021, 13(17): 3413. [30] Sea Ice Prediction Network. Sea ice outlook[EB/OL]. [2024-12-26]. https://www.arcus.org/sipn/sea-ice-outlook.

服務(wù)與反饋：

【文章下載】【發(fā)表評論】【查看評論】【加入收藏】