Calibration of sea ice drift forecasts using random forest algorithms

Palerme, Cyril; Müller, Malte

doi:10.5194/tc-15-3989-2021

Cited by 12 publications

(14 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has a 10 km spatial resolution and one-day temporal resolution from daily data at 12:00 UTC. Although there are inconsistencies with buoy SIM data [14], SAR SIM data are still the optimum for random forest model training, mainly due to the most abundant near-shore SIM data.…”

Section: The Target Variablementioning

confidence: 99%

“…However, such SIM data cannot accurately reflect the dynamics of sea ice under the impact of internal friction, ocean currents, and coastal boundaries [3,8]. In recent years, machine learning (ML) methods have been shown to have great potential for sea ice motion estimation and prediction [13][14][15][16]. For example, Palerme and Müller (2021) [14] reproduced the result that the wind field is the most important factor affecting the Arctic ice speed, and the influence of parameters related to sea ice and shore boundaries cannot be ignored since they accounted for 20-30% of the variable importances (for the concept of importance in RF models, see [17]).…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, machine learning (ML) methods have been shown to have great potential for sea ice motion estimation and prediction [13][14][15][16]. For example, Palerme and Müller (2021) [14] reproduced the result that the wind field is the most important factor affecting the Arctic ice speed, and the influence of parameters related to sea ice and shore boundaries cannot be ignored since they accounted for 20-30% of the variable importances (for the concept of importance in RF models, see [17]). in addition, Hoffman et al (2023) [16] compared the performance of persistence (PS), linear regression (LR), and a convolutional neural network (CNN) model, and found that different machine learning methods have better SIM estimation and prediction quality in areas far from the coast than in near-shore areas.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Estimating Winter Arctic Sea Ice Motion Based on Random Forest Models

Zhang,

Shi,

Leppäranta

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

Sea ice motion (SIM) plays a crucial role in setting the distribution of the ice cover in the Arctic. Limited by images’ spatial resolution and tracking algorithms, challenges exist in obtaining coastal sea ice motion (SIM) based on passive microwave satellite sensors. In this study, we developed a method based on random forest (RF) models to obtain Arctic SIM in winter by incorporating wind field and coastal geographic location information. These random forest models were trained using Synthetic Aperture Radar (SAR) SIM data. Our results show good consistency with SIM data retrieved from satellite imagery and buoy observations. With respect to the SAR data, compared with SIM estimated with RF model training using reanalysis surface wind, the results by additional coastal information input had a lower root mean square error (RMSE) and a higher correlation coefficient by 31% and 14% relative improvement, respectively. The latter SIM result also showed a better performance for magnitude, especially within 100 km of the coastline in the north of the Canadian Arctic Archipelago. In addition, the influence of coastline on SIM is quantified through variable importance calculation, at 22% and 28% importance of all RF variables for east and north SIM components, respectively. These results indicate the great potential of RF models for estimating SIM over the whole Arctic Ocean in winter.

show abstract

Section: The Target Variablementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Estimating Winter Arctic Sea Ice Motion Based on Random Forest Models

Zhang,

Shi,

Leppäranta

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…In contrast, the skill and resolution of current sea-ice forecast systems are still not fulfilling the requirements of users for tactical decision-making 24,25 . However, the combination of manual ice charts provided by the national ice services, the latest high-resolution satellite images, as well as machine learning-based calibration of sea-ice forecast information is under development and can be expected to be in operation in the coming years 26,27 .…”

Section: Efficient Use Of Existing Information Servicesmentioning

confidence: 99%

Arctic shipping trends during hazardous weather and sea-ice conditions and the Polar Code's effectiveness.

Müller

Knol-Kauffman

Jeuring

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

The Arctic’s extreme environmental conditions and remoteness make it a complex and dynamic environment for maritime operators. We find that Arctic shipping has grown by 7% per year over the past decade, despite the hazardous weather and sea-ice conditions that pose risks to vessels operating in the region. As a result of a strong increase in winter sailing, the time ships operate in these extreme conditions has tripled in the past decade. To mitigate maritime risks, the Polar Code has been introduced. Among other things, it regulates Arctic shipping by specifying hazardous conditions with a sea-ice classification scheme and design temperature threshold. However, we argue that the Polar Code needs refinement through the integration of maritime warning systems and a broader description of hazardous conditions. This is supported by an analysis of shipping activity patterns in severe sea-spray icing conditions and a discussion of a recent sea-ice induced incident along the Northern Sea Route.

show abstract

“…For the marginal ice zone in the Southern Ocean, de Vos et al (2021) find seasonal differences for short-term forecast accuracy (skill lower in winter than in spring) using buoy observations. Palerme and Müller (2021) show that the mean absolute error in operational 10 d ice drift forecasts from TOPAZ4 can be reduced using newly developed calibration methods based on supervised machine learning, and Andersson et al (2021) present a probabilistic deep learning forecasting system producing more accurate seasonal forecasts of the summer ice state than SEAS5 (Johnson et al, 2019), especially for extreme sea ice events.…”

Section: Introductionmentioning

confidence: 99%

Predictability of Arctic sea ice drift in coupled climate models

Reifenberg

Goessling

2022

The Cryosphere

View full text Add to dashboard Cite

Abstract. Skillful sea ice drift forecasts are crucial for scientific mission planning and marine safety. Wind is the dominant driver of ice motion variability, but more slowly varying components of the climate system, in particular ice thickness and ocean currents, bear the potential to render ice drift more predictable than the wind. In this study, we provide the first assessment of Arctic sea ice drift predictability in four coupled general circulation models (GCMs), using a suite of “perfect-model” ensemble simulations. We find the position vector from Lagrangian trajectories of virtual buoys to remain predictable for at least a 90 (45) d lead time for initializations in January (July), reaching about 80 % of the position uncertainty of a climatological reference forecast. In contrast, the uncertainty in Eulerian drift vector predictions reaches the level of the climatological uncertainty within 4 weeks. Spatial patterns of uncertainty, varying with season and across models, develop in all investigated GCMs. For two models providing near-surface wind data (AWI-CM1 and HadGEM1.2), we find spatial patterns and large fractions of the variance to be explained by wind vector uncertainty. The latter implies that sea ice drift is only marginally more predictable than wind. Nevertheless, particularly one of the four models (GFDL-CM3) shows a significant correlation of up to −0.85 between initial ice thickness and target position uncertainty in large parts of the Arctic. Our results provide a first assessment of the inherent predictability of ice motion in coupled climate models; they can be used to put current real-world forecast skill into perspective and highlight the model diversity of sea ice drift predictability.

show abstract

Calibration of sea ice drift forecasts using random forest algorithms

Cited by 12 publications

References 44 publications

Estimating Winter Arctic Sea Ice Motion Based on Random Forest Models

Estimating Winter Arctic Sea Ice Motion Based on Random Forest Models

Arctic shipping trends during hazardous weather and sea-ice conditions and the Polar Code's effectiveness.

Predictability of Arctic sea ice drift in coupled climate models

Contact Info

Product

Resources

About