An anatomy of the forecast errors in a seasonal prediction system with EC-Earth
Initialization is a key step when performing climate predictions, and for this the use of the latest high-quality observations and their assimilation in the model realm is of paramount importance. Less attention has been paid to other essential aspects of initialization that are equally important. F...
Main Authors: | , , , , |
---|---|
Format: | Still Image |
Language: | unknown |
Published: |
Zenodo
2019
|
Subjects: | |
Online Access: | https://dx.doi.org/10.5281/zenodo.3567552 https://zenodo.org/record/3567552 |
id |
ftdatacite:10.5281/zenodo.3567552 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.5281/zenodo.3567552 2023-05-15T15:04:54+02:00 An anatomy of the forecast errors in a seasonal prediction system with EC-Earth Cruz-García, Rubén Ortega, Pablo Acosta Navarro, Juan C. Massonnet, François Doblas-Reyes, Francisco J. 2019 https://dx.doi.org/10.5281/zenodo.3567552 https://zenodo.org/record/3567552 unknown Zenodo https://zenodo.org/communities/applicate https://dx.doi.org/10.5281/zenodo.3567551 https://zenodo.org/communities/applicate Open Access Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 info:eu-repo/semantics/openAccess CC-BY Arctic Sea ice Forecast Shock Initialization Bias Text Poster article-journal ScholarlyArticle 2019 ftdatacite https://doi.org/10.5281/zenodo.3567552 https://doi.org/10.5281/zenodo.3567551 2021-11-05T12:55:41Z Initialization is a key step when performing climate predictions, and for this the use of the latest high-quality observations and their assimilation in the model realm is of paramount importance. Less attention has been paid to other essential aspects of initialization that are equally important. For example, inconsistencies between the initial conditions (ICs) used for the different model components can cause important initialization shocks, hindering the prediction capacity during the first weeks of the forecast. In this study we investigate this and other different contributions to the forecast error in a seasonal prediction system with the EC-Earth general circulation model where sea ice is initialized via Ensemble Kalman filter assimilation of European Space Agency (ESA) derived sea ice concentrations. Large initial forecast errors in Arctic sea ice appear in regions of high observational uncertainty and little model spread, a combination that brings the assimilation, and in turn the ICs, close to the model attractor and far from the observations. We also investigated the development of the model drift during the first forecast month, and how it competes with the initial shock due to the inconsistency in ICs. After 24 (19) days the drift, as characterized by the systematic model error, becomes the largest contributor to the forecast error for the May (November) initialized forecasts, while the initial inconsistency dominates in the previous days. However, there are regions like the Greenland Sea for which the impact of the ICs inconsistency is still present after one month. Moreover, the development of both types of errors is sensitive to the month of initialization: the shock is more pronounced in November than in May. The main differences between both months relate to the systematic error, which is much higher in November, as well as to the direction of the shock with respect to the seasonal trend. In both cases the shock leads to sea ice melting, but, unlike in May, in November it happens in a context of sea ice expansion. The results stress that this opposing effect during November might be enhancing the generation of the drift. Our findings also highlight the importance of looking at high frequency data to disentangle the evolution of errors within the first forecast month, whose effects are harder to detect with the monthly averages. Still Image Arctic Greenland Greenland Sea Sea ice DataCite Metadata Store (German National Library of Science and Technology) Arctic Greenland |
institution |
Open Polar |
collection |
DataCite Metadata Store (German National Library of Science and Technology) |
op_collection_id |
ftdatacite |
language |
unknown |
topic |
Arctic Sea ice Forecast Shock Initialization Bias |
spellingShingle |
Arctic Sea ice Forecast Shock Initialization Bias Cruz-García, Rubén Ortega, Pablo Acosta Navarro, Juan C. Massonnet, François Doblas-Reyes, Francisco J. An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
topic_facet |
Arctic Sea ice Forecast Shock Initialization Bias |
description |
Initialization is a key step when performing climate predictions, and for this the use of the latest high-quality observations and their assimilation in the model realm is of paramount importance. Less attention has been paid to other essential aspects of initialization that are equally important. For example, inconsistencies between the initial conditions (ICs) used for the different model components can cause important initialization shocks, hindering the prediction capacity during the first weeks of the forecast. In this study we investigate this and other different contributions to the forecast error in a seasonal prediction system with the EC-Earth general circulation model where sea ice is initialized via Ensemble Kalman filter assimilation of European Space Agency (ESA) derived sea ice concentrations. Large initial forecast errors in Arctic sea ice appear in regions of high observational uncertainty and little model spread, a combination that brings the assimilation, and in turn the ICs, close to the model attractor and far from the observations. We also investigated the development of the model drift during the first forecast month, and how it competes with the initial shock due to the inconsistency in ICs. After 24 (19) days the drift, as characterized by the systematic model error, becomes the largest contributor to the forecast error for the May (November) initialized forecasts, while the initial inconsistency dominates in the previous days. However, there are regions like the Greenland Sea for which the impact of the ICs inconsistency is still present after one month. Moreover, the development of both types of errors is sensitive to the month of initialization: the shock is more pronounced in November than in May. The main differences between both months relate to the systematic error, which is much higher in November, as well as to the direction of the shock with respect to the seasonal trend. In both cases the shock leads to sea ice melting, but, unlike in May, in November it happens in a context of sea ice expansion. The results stress that this opposing effect during November might be enhancing the generation of the drift. Our findings also highlight the importance of looking at high frequency data to disentangle the evolution of errors within the first forecast month, whose effects are harder to detect with the monthly averages. |
format |
Still Image |
author |
Cruz-García, Rubén Ortega, Pablo Acosta Navarro, Juan C. Massonnet, François Doblas-Reyes, Francisco J. |
author_facet |
Cruz-García, Rubén Ortega, Pablo Acosta Navarro, Juan C. Massonnet, François Doblas-Reyes, Francisco J. |
author_sort |
Cruz-García, Rubén |
title |
An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
title_short |
An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
title_full |
An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
title_fullStr |
An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
title_full_unstemmed |
An anatomy of the forecast errors in a seasonal prediction system with EC-Earth |
title_sort |
anatomy of the forecast errors in a seasonal prediction system with ec-earth |
publisher |
Zenodo |
publishDate |
2019 |
url |
https://dx.doi.org/10.5281/zenodo.3567552 https://zenodo.org/record/3567552 |
geographic |
Arctic Greenland |
geographic_facet |
Arctic Greenland |
genre |
Arctic Greenland Greenland Sea Sea ice |
genre_facet |
Arctic Greenland Greenland Sea Sea ice |
op_relation |
https://zenodo.org/communities/applicate https://dx.doi.org/10.5281/zenodo.3567551 https://zenodo.org/communities/applicate |
op_rights |
Open Access Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode cc-by-4.0 info:eu-repo/semantics/openAccess |
op_rightsnorm |
CC-BY |
op_doi |
https://doi.org/10.5281/zenodo.3567552 https://doi.org/10.5281/zenodo.3567551 |
_version_ |
1766336647984578560 |