scholarly journals Extended Range Arctic Sea Ice Forecast with Convolutional Long-Short Term Memory Networks

2021 ◽  
Author(s):  
Yang Liu ◽  
Laurens Bogaardt ◽  
Jisk Attema ◽  
Wilco Hazeleger
Keyword(s):  
Sea Ice ◽  
Time Scales ◽  
Short Term Memory ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
Learning Approach ◽  
Short Term ◽  
Term Memory ◽  
Arctic Sea ◽  
Long Short Term Memory

AbstractOperational Arctic sea ice forecasts are of crucial importance to science and to society in the Arctic region. Currently, statistical and numerical climate models are widely used to generate the Arctic sea ice forecasts at weather time-scales. Numerical models require near real-time input of relevant environmental conditions consistent with the model equations and they are computationally expensive. In this study, we propose a deep learning approach, namely Convolutional Long Short Term Memory Networks (ConvLSTM), to forecast sea ice in the Barents Sea at weather to sub-seasonal time scales. This is an unsupervised learning approach. It makes use of historical records and it exploits the covariances between different variables, including spatial and temporal relations. With input fields from reanalysis data, we demonstrate that ConvLSTM is able to learn the variability of the Arctic sea ice and can forecast regional sea ice concentration skillfully at weekly to monthly time scales. It preserves the physical consistency between predictors and predictands, and generally outperforms forecasts with climatology, persistence and a statistical model. Based on the known sources of predictability, sensitivity tests with different climate fields as input for learning were performed. The impact of different predictors on the quality of the forecasts are evaluated and we demonstrate that the surface energy budget components have a large impact on the predictability of sea ice at weather time scales. This method is promising to enhance operational Arctic sea ice forecasting in the near future.

scholarly journals Extended Range Arctic Sea Ice Forecast with Convolutional Long-Short Term Memory Networks

2020 ◽  
Author(s):  
Yang Liu ◽  
Laurens Bogaardt ◽  
Jisk Attema ◽  
Wilco Hazeleger
Keyword(s):  
Sea Ice ◽  
Time Scales ◽  
Climate Models ◽  
Short Term Memory ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
Short Term ◽  
Term Memory ◽  
Arctic Sea ◽  
Long Short Term Memory

<p>Operational Arctic sea ice forecasts are of crucial importance to commercial and scientific activities in the Arctic region. Currently, numerical climate models, including General Circulation Models (GCMs) and regional climate models, are widely used to generate the Arctic sea ice predictions at weather time-scales. However, these numerical climate models require near real-time input of weather conditions to assure the quality of the predictions and these are hard to obtain and the simulations are computationally expensive. In this study, we propose a deep learning approach to forecasts of sea ice in the Barents sea at weather time scales. To work with such spatial-temporal sequence problems, Convolutional Long Short Term Memory Networks (ConvLSTM) are useful.  ConvLSTM are LSTM (Long-Short Term Memory) networks with convolutional cells embedded in the LSTM cells. This approach is unsupervised learning and it can make use of enormous amounts of historical records of weather and climate. With input fields from atmospheric (ERA-Interim) and oceanic (ORAS4) reanalysis data sets, we demonstrate that the ConvLSTM is able to learn the variability of the Arctic sea ice within historical records and effectively predict regional sea ice concentration patterns at weekly to monthly time scales. Based on the known sources of predictability, sensitivity tests with different climate fields were also performed. The influences of different predictors on the quality of predictions are evaluated. This method outperforms predictions with climatology and persistence and is promising to act as a fast and cost-efficient operational sea ice forecast system in the future.</p>

scholarly journals Electrocorticography based motor imagery movements classification using long short-term memory (LSTM) based on deep learning approach

2020 ◽  
Vol 2 (2) ◽  
Author(s):  
Mamunur Rashid ◽  
Minarul Islam ◽  
Norizam Sulaiman ◽  
Bifta Sama Bari ◽  
Ripon Kumar Saha ◽  
...  
Keyword(s):  
Deep Learning ◽  
Motor Imagery ◽  
Short Term Memory ◽  
Learning Approach ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory

scholarly journals Can we reconstruct Arctic sea ice back to 1900 with a hybrid approach?

2008 ◽  
Vol 4 (4) ◽  
pp. 955-979 ◽  
Author(s):  
S. Brönnimann ◽  
T. Lehmann ◽  
T. Griesser ◽  
T. Ewen ◽  
A. N. Grant ◽  
...  
Keyword(s):  
Sea Ice ◽  
Time Scales ◽  
Historical Data ◽  
Arctic Sea Ice ◽  
Coupled Processes ◽  
The Arctic ◽  
Statistical Reconstruction ◽  
Historical Climatology ◽  
Arctic Sea ◽  
The Future

Abstract. The variability and trend of Arctic sea ice since the mid 1970s is well documented and linked to rising temperatures. However, much less is known for the first half of the 20th century, when the Arctic also underwent a period of strong warming. For studying this period in atmospheric models, gridded sea ice data are needed as boundary conditions. Current data sets (e.g., HadISST) provide a historical climatology, but may not be suitable when interannual-to-decadal variability is important, as they are interpolated and relaxed towards a (historical) climatology to fill in gaps, particularly in winter. Regional historical sea ice information exhibits considerable variability on interannnual-to-decadal scales, but is only available for summer and not in gridded form. Combining the advantages of both types of information could be used to constrain model simulations in a more realistic way. Here we discuss the feasibility of reconstructing year-round gridded Arctic sea ice from 1900 to 1953 from historical information and a coupled climate model. We decompose sea ice variability into centennial (due to climate forcings), decadal (coupled processes in the ocean-sea ice system) and interannual time scales (atmospheric circulation). The three time scales are represented by a historical climatology from HadISST (centennial), a closest analogue approach using the coupled control run of the CCSM-3.0 model (decadal), and a statistical reconstruction based on high-pass filtered data (interannual variability), respectively. Results show that differences in the model climatology, the length of the control run, and inconsistent historical data strongly limit the quality of the product. However, with more realistic and longer simulations becoming available in the future as well as with improved historical data, useful reconstructions are possible. We suggest that hybrid approaches, using both statistical reconstruction methods and numerical models, may find wider applications in the future.

Assessment of the Regional Arctic System Model Intra-Annual Ensemble Predictions of Arctic Sea Ice

2020 ◽  
Author(s):  
Wieslaw Maslowski ◽  
Younjoo Lee ◽  
Anthony Craig ◽  
Mark Seefeldt ◽  
Robert Osinski ◽  
...  
Keyword(s):  
Sea Ice ◽  
Time Scales ◽  
Weather Prediction ◽  
Arctic Sea Ice ◽  
System Model ◽  
The Arctic ◽  
Ensemble Prediction ◽  
Ensemble Forecasts ◽  
The Past ◽  
Arctic Sea

<p>The Regional Arctic System Model (RASM) has been developed and used to investigate the past to present evolution of the Arctic climate system and to address increasing demands for Arctic forecasts beyond synoptic time scales. RASM is a fully coupled ice-ocean-atmosphere-land hydrology model configured over the pan-Arctic domain with horizontal resolution of 50 km or 25 km for the atmosphere and land and 9.3 km or 2.4 km for the ocean and sea ice components. As a regional model, RASM requires boundary conditions along its lateral boundaries and in the upper atmosphere, which for simulations of the past to present are derived from global atmospheric reanalyses, such as the National Center for Environmental Predictions (NCEP) Coupled Forecast System version 2 and Reanalysis (CFSv2/CFSR). This dynamical downscaling approach allows comparison of RASM results with observations, in place and time, to diagnose and reduce model biases. This in turn allows a unique capability not available in global weather prediction and Earth system models to produce realistic and physically consistent initial conditions for prediction without data assimilation.</p><p>More recently, we have developed a new capability for an intra-annual (up to 6 months) ensemble prediction of the Arctic sea ice and climate using RASM forced with the routinely produced (every 6 hours) NCEP CFSv2 global 9-month forecasts. RASM intra-annual ensemble forecasts have been initialized on the 1<sup>st</sup> of each month starting in 2019 with forcing for each ensemble member derived from CSFv2 forecasts, 24-hr apart from the month preceding the initial forecast date.  Several key processes and feedbacks will be discussed with regard to their impact on model physics, the representation of initial state and ensemble prediction skill of Arctic sea ice variability at time scales from synoptic to decadal. The skill of RASM ensemble forecasts will be assessed against available satellite observations with reference to reanalysis as well as hindcast data using several metrics, including the standard deviation, root mean square difference, Taylor diagrams and integrated ice-edge error.</p>

Revisiting the Linkages between the Variability of Atmospheric Circulations and Arctic Melt-Season Sea Ice Cover at Multiple Time Scales

2020 ◽  
Author(s):  
Lejiang Yu ◽  
Sharon Zhong
Keyword(s):  
Sea Ice ◽  
Arctic Ocean ◽  
Time Scales ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
Multiple Time Scales ◽  
Multiple Time ◽  
Atmospheric Circulations ◽  
The Pacific ◽  
Arctic Sea

<p>The sharp decline of Arctic sea ice in recent decades has captured the attention of the climate science<br>community. A majority of climate analyses performed to date have used monthly or seasonal data. Here,<br>however, we analyze daily sea ice data for 1979–2016 using the self-organizing map (SOM) method to further<br>examine and quantify the contributions of atmospheric circulation changes to the melt-season Arctic sea ice<br>variability. Our results reveal two main variability modes: the Pacific sector mode and the Barents and Kara<br>Seas mode, which together explain about two-thirds of the melt-season Arctic sea ice variability and more<br>than 40% of its trend for the study period. The change in the frequencies of the two modes appears to be<br>associated with the phase shift of the Pacific decadal oscillation (PDO) and the Atlantic multidecadal oscillation<br>(AMO). The PDO and AMO trigger anomalous atmospheric circulations, in particular, the<br>Greenland high and the North Atlantic Oscillation and anomalous warm and cold air advections into the<br>Arctic Ocean. The changes in surface air temperature, lower-atmosphere moisture, and downwelling longwave<br>radiation associated with the advection are consistent with the melt-season sea ice anomalies observed<br>in various regions of the Arctic Ocean. These results help better understand the predictability of Arctic sea ice<br>on multiple (synoptic, intraseasonal, and interannual) time scales.</p>

scholarly journals Arctic Energy Budget in Relation to Sea Ice Variability on Monthly-to-Annual Time Scales

2015 ◽  
Vol 28 (16) ◽  
pp. 6335-6350 ◽  
Author(s):  
F. Krikken ◽  
W. Hazeleger
Keyword(s):  
Sea Ice ◽  
Time Scales ◽  
Energy Budget ◽  
Climate Models ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
Albedo Feedback ◽  
Arctic Sea ◽  
Cloud Ice

Abstract The large decrease in Arctic sea ice in recent years has triggered a strong interest in Arctic sea ice predictions on seasonal-to-decadal time scales. Hence, it is important to understand physical processes that provide enhanced predictability beyond persistence of sea ice anomalies. This study analyzes the natural variability of Arctic sea ice from an energy budget perspective, using 15 climate models from phase 5 of CMIP (CMIP5), and compares these results to reanalysis data. The authors quantify the persistence of sea ice anomalies and the cross correlation with the surface and top-of-atmosphere energy budget components. The Arctic energy balance components primarily indicate the important role of the seasonal ice–albedo feedback, through which sea ice anomalies in the melt season reemerge in the growth season. This is a robust anomaly reemergence mechanism among all 15 climate models. The role of the ocean lies mainly in storing heat content anomalies in spring and releasing them in autumn. Ocean heat flux variations play only a minor role. Confirming a previous (observational) study, the authors demonstrate that there is no direct atmospheric response of clouds to spring sea ice anomalies, but a delayed response is evident in autumn. Hence, there is no cloud–ice feedback in late spring and summer, but there is a cloud–ice feedback in autumn, which strengthens the ice–albedo feedback. Anomalies in insolation are positively correlated with sea ice variability. This is primarily a result of reduced multiple reflection of insolation due to an albedo decrease. This effect counteracts the ice-albedo effect up to 50%. ERA-Interim and Ocean Reanalysis System 4 (ORAS4) confirm the main findings from the climate models.

scholarly journals Two-Stream Convolutional Long- and Short-Term Memory Model Using Perceptual Loss for Sequence-to-Sequence Arctic Sea Ice Prediction

2021 ◽  
Vol 13 (17) ◽  
pp. 3413
Author(s):  
Junhwa Chi ◽  
Jihyun Bae ◽  
Young-Joo Kwon
Keyword(s):  
Sea Ice ◽  
Loss Function ◽  
Prediction Models ◽  
Short Term Memory ◽  
Arctic Sea Ice ◽  
Feature Maps ◽  
Short Term ◽  
Arctic Sea ◽  
Input Variables ◽  
Recent Attention

Arctic sea ice plays a significant role in climate systems, and its prediction is important for coping with global warming. Artificial intelligence (AI) has gained recent attention in various disciplines with the increasing use of big data. In recent years, the use of AI-based sea ice prediction, along with conventional prediction models, has drawn attention. This study proposes a new deep learning (DL)-based Arctic sea ice prediction model with a new perceptual loss function to improve both statistical and visual accuracy. The proposed DL model learned spatiotemporal characteristics of Arctic sea ice for sequence-to-sequence predictions. The convolutional neural network-based perceptual loss function successfully captured unique sea ice patterns, and the widely used loss functions could not use various feature maps. Furthermore, the input variables that are essential to accurately predict Arctic sea ice using various combinations of input variables were identified. The proposed approaches produced statistical outcomes with better accuracy and qualitative agreements with the observed data.

Sign in / Sign up

Export Citation Format

Share Document