Skip to content

Commit 34832c1

Browse files
bnubaldbnubald
authored
Merge pull request #20 from bnubald/post_icenet_ramadda_available
Add blog post on IceNet forecasts on RAMADDA
2 parents cd5a349 + dfe1b0e commit 34832c1

File tree

3 files changed

+136
-37
lines changed

3 files changed

+136
-37
lines changed

‎src/data/post/oct25-icenet-forecasts-published.mdx‎

Lines changed: 79 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,79 @@
1+
---
2+
publishDate: 2025-10-07T00:00:00Z
3+
author: Bryn Noel Ubald
4+
title: Operational IceNet Forecasts Now Publicly Accessible!
5+
excerpt: IceNet forecasts are now available on the PDC's RAMADDA platform.
6+
image: https://images.unsplash.com/photo-1598439210625-5067c578f3f6?q=80&w=2072&auto=format&fit=crop&ixlib=rb-4.1.0&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D
7+
category: Article
8+
tags:
9+
- news
10+
metadata:
11+
canonical: https://astrowind.vercel.app/astrowind-template-in-depth
12+
---
13+
14+
import DListItem from '~/components/ui/DListItem.astro';
15+
import ToggleTheme from '~/components/common/ToggleTheme.astro';
16+
17+
We are thrilled to announce that the Polar Data Centre's [RAMADDA data repository](https://ramadda.data.bas.ac.uk/repository/a/icenet-daily-sea-ice-forecasts) is now home to daily IceNet forecasts! This platform now enables a wide range of end-users to access these forecasts. With automated inclusion of latest forecasts as they are generated on a daily basis.
18+
19+
Use-cases could involve:
20+
21+
- Empowering researchers/students in research
22+
- Informing policymakers
23+
- Navigation of ships in polar regions
24+
- Environmental monitoring
25+
26+
**Please note that IceNet forecasts are highly experimental and is a research codebase**
27+
28+
<a href="https://ramadda.data.bas.ac.uk/repository/a/icenet-daily-sea-ice-forecasts" target="_blank">Explore IceNet on RAMADDA</a>
29+
30+
### Why This Matters
31+
Sea ice plays a critical role in global climate systems, influencing ocean currents, wildlife habitats, and human activities like shipping. IceNet forecasts can enable users to visualise and analyse daily forecasts of sea-ice.
32+
33+
### Join Us in Exploring the Polar Regions
34+
Whether you're a scientist, student, or simply curious about Earth's changing environment, we invite you to dive into IceNet forecasts via RAMADDA.
35+
36+
37+
## Model definition
38+
39+
The current release of the forecasts contain daily sea ice forecasts across the northern and southern hemispheres (via two separate models) derived from OSI-SAF 25km<sup>2</sup> grid used as ground truth. They are labelled as `exp23_north` and `exp23_south`, and forecast up to 93 days ahead, with a roughly 5-7 day delay depending on ERA5 data release cycle.
40+
41+
These models were trained using [icenet v0.2.4_dev](https://pypi.org/project/icenet/0.2.4/) on the British Antarctic Survey's internal HPC (BAS HPC) with the following input variables:
42+
43+
* ERA5
44+
* psl
45+
* ta500
46+
* tas
47+
* tos
48+
* uas
49+
* vas
50+
* zg250
51+
* zg500
52+
* OSI-SAF
53+
* siconca (also, the target variable)
54+
55+
### Train splits
56+
The date ranges used for training are as follows:
57+
* 1994-01-01 to 1995-12-31
58+
* 2006-01-01 to 2008-12-31
59+
* 2011-01-01 to 2013-12-31
60+
61+
### Validation splits
62+
The date ranges used for validation are as follows:
63+
* 2009-07-01 to 2010-06-30
64+
65+
### Prediction
66+
67+
The predictions are generated using [icenet v0.2.9](https://pypi.org/project/icenet/0.2.9/), also on BAS HPC.
68+
69+
### License
70+
71+
Unless otherwise stated, all content is owned by British Antarctic Survey and The Alan Turing Institute 2025, and made available via the [Open Government License](https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/) which is compatible with the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/).
72+
73+
## Future
74+
75+
In the future, RAMADDA will also host IceNet forecasts from different models:
76+
- AMSR2 sea-ice based predictions.
77+
- Monthly predictions, up to 6 months ahead.
78+
- Fine-tuned models combining OSI-SAF and AMSR2 training.
79+

‎src/navigation.ts‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -61,7 +61,7 @@ export const headerData = {
6161
],
6262
},
6363
],
64-
actions: [{ text: 'Documentation', href: 'docs/', target: '_blank' }],
64+
actions: [{ text: 'Documentation', href: '/docs/', target: '_blank' }],
6565
};
6666

6767
export const footerData = {

‎src/pages/index.astro‎

Lines changed: 56 additions & 36 deletions
Original file line numberDiff line numberDiff line change
@@ -86,43 +86,63 @@ const images = import.meta.glob<{ default: ImageMetadata }>('/src/assets/images/
8686
</Fragment>
8787

8888
<Fragment slot="subtitle">
89-
<p>
90-
IceNet is a deep learning sea ice forecasting system developed by an <a
91-
class="hyperlink"
92-
href="https://www.bas.ac.uk/media-post/artificial-intelligence-to-help-predict-arctic-sea-ice-loss/"
93-
>international team and led by the British Antarctic Survey and The Alan Turing Institute</a
94-
>. The original IceNet research model, published in <a
95-
class="hyperlink"
96-
href="https://www.nature.com/articles/s41467-021-25257-4"><b>Nature Communications</b></a
97-
>, was trained on climate simulations and observational data to forecast the next 6 months of monthly-averaged
98-
sea ice concentration maps. This version advanced the range of accurate sea ice forecasts, outperforming a
99-
state-of-the-art dynamical model (ECMWF SEAS5) in seasonal forecasts of summer sea ice, particularly for extreme
100-
sea ice events.
101-
</p><br />
102-
<p>
103-
Since then, the IceNet team has focused on building an operational version of the model which forecasts on a
104-
daily resolution. The <a class="hyperlink" href="https://www.github.com/tom-andersson/icenet-paper"
105-
>original research code</a
106-
> was refactored into <code>icenet</code> - <a class="hyperlink" href="https://github.com/icenet-ai/icenet"
107-
>a library for operational forecasting</a
108-
>. The <code>icenet</code> library can support further research efforts into AI-based operational sea ice forecasting.
109-
</p><br />
110-
<p>
111-
In addition, several use cases and an ecosystem of service components are contained within this organization,
112-
supporting execution and downstream analysis.
113-
</p>
114-
<p>
115-
<br />
116-
For further information about the team involved, please look at the <a
117-
class="hyperlink"
118-
href="https://www.bas.ac.uk/project/icenet/">project pages at BAS</a
119-
> or <a
120-
class="hyperlink"
121-
href="https://www.turing.ac.uk/news/artificial-intelligence-help-predict-arctic-sea-ice-loss"
122-
>The Alan Turing Institute</a
123-
>.
124-
</p>
89+
<div class="intersect-once intersect-quarter motion-safe:md:opacity-0 motion-safe:md:intersect:animate-fade">
90+
<p>
91+
IceNet is a deep learning sea ice forecasting system developed by an
92+
<a
93+
class="hyperlink"
94+
href="https://www.bas.ac.uk/media-post/artificial-intelligence-to-help-predict-arctic-sea-ice-loss/"
95+
>
96+
international team and led by the British Antarctic Survey and The Alan Turing Institute
97+
</a>.
98+
The original IceNet research model, published in
99+
<a
100+
class="hyperlink"
101+
href="https://www.nature.com/articles/s41467-021-25257-4"
102+
>
103+
<b>Nature Communications</b>
104+
</a>,
105+
was trained on climate simulations and observational data to forecast the next 6 months of
106+
monthly-averaged sea ice concentration maps. This version advanced the range of accurate sea
107+
ice forecasts, outperforming a state-of-the-art dynamical model (ECMWF SEAS5) in seasonal
108+
forecasts of summer sea ice, particularly for extreme sea ice events. <br><br>
109+
</p>
110+
111+
<p>
112+
Since then, the IceNet team has focused on building an operational version of the model which
113+
forecasts on a daily resolution. The
114+
<a class="hyperlink" href="https://www.github.com/tom-andersson/icenet-paper">
115+
original research code
116+
</a>
117+
was refactored into <code>icenet</code> –
118+
<a class="hyperlink" href="https://github.com/icenet-ai/icenet">
119+
a library for operational forecasting
120+
</a>.
121+
The <code>icenet</code> library can support further research efforts into AI-based operational
122+
sea ice forecasting. <br><br>
123+
</p>
124+
125+
<p>
126+
In addition, several use cases and an ecosystem of service components are contained within
127+
this organisation, supporting execution and downstream analysis.
128+
</p>
129+
130+
<p>
131+
For further information about the team involved, please look at the
132+
<a class="hyperlink" href="https://www.bas.ac.uk/project/icenet/">
133+
project pages at BAS
134+
</a>
135+
or
136+
<a
137+
class="hyperlink"
138+
href="https://www.turing.ac.uk/news/artificial-intelligence-help-predict-arctic-sea-ice-loss"
139+
>
140+
The Alan Turing Institute
141+
</a>.
142+
</p>
143+
</div>
125144
</Fragment>
145+
126146
</Hero>
127147

128148
<!-- HighlightedPosts Widget ******* -->

0 commit comments

Comments
 (0)