scholarly journals Improving Surface Melt Estimation over Antarctica Using Deep Learning: A Proof-of-Concept over the Larsen Ice Shelf

2021 ◽  
Author(s):  
Zhongyang Hu ◽  
Peter Kuipers Munneke ◽  
Stef Lhermitte ◽  
Maaike Izeboud ◽  
Michiel van den Broeke
Keyword(s):  
Deep Learning ◽  
Climate Model ◽  
High Surface ◽  
Absolute Error ◽  
Coefficient Of Determination ◽  
Climate Modelling ◽  
Ice Shelf ◽  
Surface Melt ◽  
Mlp Model ◽  
Larsen Ice Shelf

Abstract. Accurately estimating surface melt volume of the Antarctic Ice Sheet is challenging, and has hitherto relied on climate modelling, or on observations from satellite remote sensing. Each of these methods has its limitations, especially in regions with high surface melt. This study aims to demonstrate the potential of improving surface melt simulations by deploying a deep learning model. A deep-learning-based framework has been developed to correct surface melt from the regional atmospheric climate model version 2.3p2 (RACMO2), using meteorological observations from automatic weather stations (AWSs), and surface albedo from satellite imagery. The framework includes three steps: (1) training a deep multilayer perceptron (MLP) model using AWS observations; (2) correcting moderate resolution imaging spectroradiometer (MODIS) albedo observations, and (3) using these two to correct the RACMO2 surface melt simulations. Using observations from three AWSs at the Larsen B and C Ice Shelves, Antarctica, cross-validation shows a high accuracy (root mean square error = 0.95 mm w.e. per day, mean absolute error = 0.42 mm w.e. per day, and coefficient of determination = 0.95). Moreover, the deep MLP model outperforms conventional machine learning models (e.g., random forest regression, XGBoost) and a shallow MLP model. When applying the trained deep MLP model over the entire Larsen Ice Shelf, the resulting, corrected RACMO2 surface melt shows a better correlation with the AWS observations for two out of three AWSs. However, for one location (AWS 18) the deep MLP model does not show improved agreement with AWS observations, likely due to the heterogeneous drivers of melt within the corresponding coarse resolution model pixels. Our study demonstrates the opportunity to improve surface melt simulations using deep learning combined with satellite albedo observations. On the other hand, more work is required to refine the method, especially for complicated and heterogeneous terrains.

scholarly journals Improving surface melt estimation over the Antarctic Ice Sheet using deep learning: a proof of concept over the Larsen Ice Shelf

2021 ◽  
Vol 15 (12) ◽  
pp. 5639-5658
Author(s):  
Zhongyang Hu ◽  
Peter Kuipers Munneke ◽  
Stef Lhermitte ◽  
Maaike Izeboud ◽  
Michiel van den Broeke
Keyword(s):  
Deep Learning ◽  
Climate Model ◽  
Ice Sheet ◽  
Katabatic Wind ◽  
Ice Shelf ◽  
Antarctic Ice Sheet ◽  
Surface Melt ◽  
Mlp Model ◽  
The Antarctic ◽  
Larsen Ice Shelf

Abstract. Accurately estimating the surface melt volume of the Antarctic Ice Sheet is challenging and has hitherto relied on climate modeling or observations from satellite remote sensing. Each of these methods has its limitations, especially in regions with high surface melt. This study aims to demonstrate the potential of improving surface melt simulations with a regional climate model by deploying a deep learning model. A deep-learning-based framework has been developed to correct surface melt from the regional atmospheric climate model version 2.3p2 (RACMO2), using meteorological observations from automatic weather stations (AWSs) and surface albedo from satellite imagery. The framework includes three steps: (1) training a deep multilayer perceptron (MLP) model using AWS observations, (2) correcting Moderate Resolution Imaging Spectroradiometer (MODIS) albedo observations, and (3) using these two to correct the RACMO2 surface melt simulations. Using observations from three AWSs at the Larsen B and C ice shelves, Antarctica, cross-validation shows a high accuracy (root-mean-square error of 0.95 mm w.e. d−1, mean absolute error of 0.42 mm w.e. d−1, and a coefficient of determination of 0.95). Moreover, the deep MLP model outperforms conventional machine learning models and a shallow MLP model. When applying the trained deep MLP model over the entire Larsen Ice Shelf, the resulting corrected RACMO2 surface melt shows a better correlation with the AWS observations for two out of three AWSs. However, for one location (AWS 18), the deep MLP model does not show improved agreement with AWS observations; this is likely because surface melt is largely driven by factors (e.g., air temperature, topography, katabatic wind) other than albedo within the corresponding coarse-resolution model pixels. Our study demonstrates the opportunity to improve surface melt simulations using deep learning combined with satellite albedo observations. However, more work is required to refine the method, especially for complicated and heterogeneous terrains.

Estimating Surface Melt on the Larsen Ice Shelf Using a Deep Neural Network: Opportunities and Challenges

2021 ◽  
Author(s):  
Zhongyang Hu ◽  
Peter Kuipers Munneke ◽  
Stef Lhermitte ◽  
Maaike Izeboud ◽  
Michiel van den Broeke
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Deep Neural Network ◽  
Climate Model ◽  
Climate Modeling ◽  
Coefficient Of Determination ◽  
Ice Shelf ◽  
Heterogeneous Terrain ◽  
Surface Melt ◽  
Larsen Ice Shelf

<p>Presently, surface melt over Antarctica is estimated using climate modeling or remote sensing. However, accurately estimating surface melt remains challenging. Both climate modeling and remote sensing have limitations, particularly in the most crucial areas with intense surface melt.  The motivation of our study is to investigate the opportunities and challenges in improving the accuracy of surface melt estimation using a deep neural network. The trained deep neural network uses meteorological observations from automatic weather stations (AWS) and surface albedo observations from satellite imagery to improve surface melt simulations from the regional atmospheric climate model version 2.3p2 (RACMO2). Based on observations from three AWS at the Larsen B and C Ice Shelves, cross-validation shows a high accuracy (root mean square error = 0.898 mm.w.e.d<sup>−1</sup>, mean absolute error = 0.429 mm.w.e.d<sup>−1</sup>, and coefficient of determination = 0.958). The deep neural network also outperforms conventional machine learning models (e.g., random forest regression, XGBoost) and a shallow neural network. To compute surface melt for the entire Larsen Ice Shelf, the deep neural network is applied to RACMO2 simulations. The resulting, corrected surface melt shows a better correlation with the AWS observations in AWS 14 and 17, but not in AWS 18. Also, the spatial pattern of the surface melt is improved compared to the original RACMO2 simulation. A possible explanation for the mismatch at AWS 18 is its complex geophysical setting. Even though our study shows an opportunity to improve surface melt simulations using a deep neural network, further study is needed to refine the method, especially for complicated, heterogeneous terrain.</p>

Evolution of supraglacial lakes on Shackleton Ice Shelf, East Antarctica

2020 ◽  
Author(s):  
Jennifer Arthur ◽  
Chris Stokes ◽  
Stewart Jamieson ◽  
Rachel Carr ◽  
Amber Leeson
Keyword(s):  
Climate Model ◽  
High Surface ◽  
Surface Melting ◽  
East Antarctica ◽  
Ice Shelf ◽  
Ice Shelves ◽  
Shelf Stability ◽  
Supraglacial Lakes ◽  
Surface Melt ◽  
The Impact

<p>Supraglacial lakes (SGLs) enhance surface melting and their development and subsequent drainage can flex and fracture ice shelves, leading to their disintegration. However, the seasonal evolution of SGLs and their potential influence on ice shelf stability in East Antarctica remains poorly understood, despite a number of potentially vulnerable ice shelves. Using optical satellite imagery, climate reanalysis data and surface melt predicted by a regional climate model, we provide the first multi-year analysis (1974-2019) of seasonal SGL evolution on Shackleton Ice Shelf in Queen Mary Land, which is Antarctica’s northernmost remaining ice shelf. We mapped >43,000 lakes on the ice shelf and >5,000 lakes on grounded ice over the 45-year analysis period, some of which developed up to 12 km inland from the grounding line. Lakes clustered around the ice shelf grounding zone are strongly linked to the presence of blue ice and exposed rock, associated with an albedo-lowering melt-enhancing feedback. Lakes either drain supraglacially, refreeze at the end of the melt season, or shrink in-situ. Furthermore, we observe some relatively rapid (≤ 7 days) lake drainage events and infer that some lakes may be draining by hydrofracture. Our observations suggest that enhanced surface meltwater could increase the vulnerability of East Antarctic ice shelves already preconditioned for hydrofracture, namely those experiencing high surface melt rates, firn air depletion, and extensional stress regimes with minimum topographic confinement. Our results could be used to constrain simulations of current melt conditions on the ice shelf and to investigate the impact of increased surface melting on future ice shelf stability.</p>

Determination of number of check dams by artificial neural networks in arid regions of Iran

2015 ◽  
Vol 72 (6) ◽  
pp. 952-959 ◽  
Author(s):  
Seyed Ali Asghar Hashemi ◽  
Hamed Kashi
Keyword(s):  
Curve Number ◽  
Absolute Error ◽  
Coefficient Of Determination ◽  
Ann Model ◽  
Effective Parameters ◽  
Check Dams ◽  
Artificial Neural ◽  
Artificial Neural Network Ann ◽  
Mlp Model

An artificial neural network (ANN) model with six hydrological factors including time of concentration (TC), curve number, slope, imperviousness, area and input discharge as input parameters and number of check dams (NCD) as output parameters was developed and created using GIS and field surveys. The performance of this model was assessed by the coefficient of determination R2, root mean square error (RMSE), values account and mean absolute error (MAE). The results showed that the computed values of NCD using ANN with a multi-layer perceptron (MLP) model regarding RMSE, MAE, values adjustment factor (VAF), and R2 (1.75, 1.25, 90.74, and 0.97) for training, (1.34, 0.89, 97.52, and 0.99) for validation and (0.53, 0.8, 98.32, and 0.99) for test stage, respectively, were in close agreement with their respective values in the watershed. Finally, the sensitivity analysis showed that the area, TC and curve number were the most effective parameters in estimating the number of check dams.

scholarly journals Deep Learning-Based Optimal Smart Shoes Sensor Selection for Energy Expenditure and Heart Rate Estimation

Sensors ◽  
2021 ◽  
Vol 21 (21) ◽  
pp. 7058
Author(s):  
Heesang Eom ◽  
Jongryun Roh ◽  
Yuli Sun Hariyani ◽  
Suwhan Baek ◽  
Sukho Lee ◽  
...  
Keyword(s):  
Heart Rate ◽  
Deep Learning ◽  
Energy Expenditure ◽  
Mean Squared Error ◽  
Pressure Sensors ◽  
Absolute Error ◽  
Coefficient Of Determination ◽  
Proposed Model ◽  
Point Pressure

Wearable technologies are known to improve our quality of life. Among the various wearable devices, shoes are non-intrusive, lightweight, and can be used for outdoor activities. In this study, we estimated the energy consumption and heart rate in an environment (i.e., running on a treadmill) using smart shoes equipped with triaxial acceleration, triaxial gyroscope, and four-point pressure sensors. The proposed model uses the latest deep learning architecture which does not require any separate preprocessing. Moreover, it is possible to select the optimal sensor using a channel-wise attention mechanism to weigh the sensors depending on their contributions to the estimation of energy expenditure (EE) and heart rate (HR). The performance of the proposed model was evaluated using the root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2). Moreover, the RMSE was 1.05 ± 0.15, MAE 0.83 ± 0.12 and R2 0.922 ± 0.005 in EE estimation. On the other hand, and RMSE was 7.87 ± 1.12, MAE 6.21 ± 0.86, and R2 0.897 ± 0.017 in HR estimation. In both estimations, the most effective sensor was the z axis of the accelerometer and gyroscope sensors. Through these results, it is demonstrated that the proposed model could contribute to the improvement of the performance of both EE and HR estimations by effectively selecting the optimal sensors during the active movements of participants.

scholarly journals The 32-year record-high surface melt in 2019/2020 on the northern George VI Ice Shelf, Antarctic Peninsula

2021 ◽  
Vol 15 (2) ◽  
pp. 909-925
Author(s):  
Alison F. Banwell ◽  
Rajashree Tri Datta ◽  
Rebecca L. Dell ◽  
Mahsa Moussavi ◽  
Ludovic Brucker ◽  
...  
Keyword(s):  
High Surface ◽  
Microwave Radiometer ◽  
Austral Summer ◽  
Landsat 8 ◽  
Ice Shelf ◽  
Near Surface ◽  
Ice Shelves ◽  
Air Content ◽  
Air Temperatures ◽  
Surface Melt

Abstract. In the 2019/2020 austral summer, the surface melt duration and extent on the northern George VI Ice Shelf (GVIIS) was exceptional compared to the 31 previous summers of distinctly lower melt. This finding is based on analysis of near-continuous 41-year satellite microwave radiometer and scatterometer data, which are sensitive to meltwater on the ice shelf surface and in the near-surface snow. Using optical satellite imagery from Landsat 8 (2013 to 2020) and Sentinel-2 (2017 to 2020), record volumes of surface meltwater ponding were also observed on the northern GVIIS in 2019/2020, with 23 % of the surface area covered by 0.62 km3 of ponded meltwater on 19 January. These exceptional melt and surface ponding conditions in 2019/2020 were driven by sustained air temperatures ≥0 ∘C for anomalously long periods (55 to 90 h) from late November onwards, which limited meltwater refreezing. The sustained warm periods were likely driven by warm, low-speed (≤7.5 m s−1) northwesterly and northeasterly winds and not by foehn wind conditions, which were only present for 9 h total in the 2019/2020 melt season. Increased surface ponding on ice shelves may threaten their stability through increased potential for hydrofracture initiation; a risk that may increase due to firn air content depletion in response to near-surface melting.

scholarly journals Automated mapping of the seasonal evolution of surface meltwater and its links to climate on the Amery Ice Shelf, Antarctica

2021 ◽  
Author(s):  
Peter A. Tuckett ◽  
Jeremy C. Ely ◽  
Andrew J. Sole ◽  
James M. Lea ◽  
Stephen J. Livingstone ◽  
...  
Keyword(s):  
Time Series ◽  
Spatial Distribution ◽  
Climate Model ◽  
Google Earth ◽  
Katabatic Wind ◽  
Lake Area ◽  
Ice Shelf ◽  
Amery Ice Shelf ◽  
Surface Melt ◽  
The Antarctic

Abstract. Surface meltwater is widespread around the margin of the Antarctic Ice Sheet and has the potential to influence ice-shelf stability, ice-dynamic processes and ice-albedo feedbacks. Whilst the general spatial distribution of surface meltwater across the Antarctic continent is now relatively well known, our understanding of the seasonal and multi-year evolution of surface meltwater is limited. Attempts to generate robust time series of melt cover have largely been constrained by computational expense or limited ice surface visibility associated with mapping from optical satellite imagery. Here, we implement an existing meltwater detection method alongside a novel method for calculating visibility metrics within Google Earth Engine. This enables us to quantify uncertainty induced by cloud cover and variable image data coverage, allowing us to automatically generate time series of surface melt area over large spatial and temporal scales. We demonstrate our method on the Amery Ice Shelf region of East Antarctica, analysing 4,164 Landsat 7 and 8 optical images between 2005 and 2020. Results show high interannual variability in surface meltwater cover, with mapped cumulative lake area totals ranging from 384 km2 to 3898 km2 per melt season. However, by incorporating image visibility assessments into our results, we estimate that cumulative total lake areas are on average 42 % higher than minimum mapped values, highlighting the importance of accounting for variations in image visibility when mapping lake areas. In a typical melt season, total lake area remains low throughout November and early December, before increasing, on average, by an order of magnitude during the second half of December. Peak lake area most commonly occurs during January, before decreasing during February as lakes freeze over. We show that modelled melt predictions from a regional climate model provides a good indication of lake cover in the Amery region, and that annual lake coverage is strongly associated with phases of the Southern Annular Mode (SAM); surface melt is typically highest in years with a negative austral summer SAM index. Furthermore, we suggest that melt-albedo feedbacks modulate the spatial distribution of meltwater in the region, with the exposure of blue ice from persistent katabatic wind scouring influencing the susceptibility of melt ponding. Results demonstrate how our method could be scaled up to generate a multi-year time series record of surface water extent from optical imagery at a continent-wide scale.

scholarly journals 32-year record-high surface melt in 2019/2020 on north George VI Ice Shelf, Antarctic Peninsula

2020 ◽  
Author(s):  
Alison F. Banwell ◽  
Rajashree Tri Datta ◽  
Rebecca L. Dell ◽  
Mahsa Moussavi ◽  
Ludovic Brucker ◽  
...  
Keyword(s):  
High Surface ◽  
Microwave Radiometer ◽  
Austral Summer ◽  
Landsat 8 ◽  
Ice Shelf ◽  
Near Surface ◽  
Ice Shelves ◽  
Air Content ◽  
Air Temperatures ◽  
Surface Melt

Abstract. In the 2019/2020 austral summer, the surface melt duration and extent on the northern George VI Ice Shelf (GVIIS) was exceptional compared to the 31 previous summers of dramatically lower melt. This finding is based on analysis of near-continuous 41-year satellite microwave radiometer (and scatterometer) data, which are sensitive to meltwater on the ice-shelf surface and in the near-surface snow. Using optical satellite imagery from Landsat 8 (since 2013) and Sentinel-2 (since 2017), record volumes of surface meltwater ponding are also observed on north GVIIS in 2019/2020, with 23 % of the surface area covered by 0.62 km3 of meltwater on January 19. These exceptional melt and surface ponding conditions in 2019/2020 were driven by sustained air temperatures ≥ 0 °C for anomalously long periods (55–90 hours) from late November onwards, likely driven by warmer northwesterly and northeasterly low-speed winds. Increased surface ponding on ice shelves may threaten their stability through increased potential for hydrofracture initiation; a risk that may increase due to firn air content depletion in response to near-surface melting.

scholarly journals Central tropical Pacific convection drives extreme high temperatures and surface melt on the Larsen ice shelf

2021 ◽  
Author(s):  
Kyle Clem ◽  
Deniz Bozkurt ◽  
Daemon Kennett ◽  
John King ◽  
John Turner
Keyword(s):  
High Temperatures ◽  
Deep Convection ◽  
Extreme Temperature ◽  
Drake Passage ◽  
Tropical Pacific ◽  
Ice Shelf ◽  
Surface Mass ◽  
Surface Warming ◽  
Surface Melt ◽  
Larsen Ice Shelf

Abstract The northern parts of the Larsen ice shelf, eastern Antarctic Peninsula (AP) have experienced dramatic break-up since the early 1990s as a result of strong summertime surface melt, which has been linked to stronger circumpolar westerly winds. Here we show extreme summertime surface melt and high temperatures over the eastern AP and Larsen C ice shelf occur because of enhanced deep convection in the central tropical Pacific, which produces cyclonic conditions across the middle and high-latitude South Pacific, and a strong high pressure anomaly over Drake Passage. Together these transport extreme heat and moisture from low latitudes to the AP, at times in the form of "atmospheric rivers", producing strong surface warming and melt on the Larsen ice shelf by the Foehn effect. Therefore, variability in central tropical Pacific convection is crucial for interpreting past and projecting future AP surface mass balance and extreme temperature events.

Sign in / Sign up

Export Citation Format

Share Document