# BAS EO/DT abstract - UK Earth Obs conference

## Authors
Ben
Martin
James
Scott
Maria
Andrew

## Themes
* Earth obs - near-real-time and archive
* SDA DT
* Env. DT
* Wildlife from space

## Possible sessions
High cadence EO: Viewing landscape processes through a different lens
Towards the Development of Earth System Digital Twins
Complex environmental monitoring with EO and In-situ sensor fusion
Fusion of Satellite Data and Models for Understanding Environmental Change

### Session structure

I (JB) propose we start with a broad overview of EO @ BAS (AF/MAGIC), the move w/AI Lab towards building Digital Twins (JB), then deep dive into how EO is supporting the development of DT components in specific use-cases (BE/...)

_If this sounds good, I can further flesh out the below easily enough and then you chaps can gut it/add the specifics in!_

4 min DT ambitions and overview
* Motivations (try to keep science-heavy - SDA as means of achieving sustainable env sci)
* Framework / Schematic
* Architectures and collaborations
* R&D driven by use-cases to produce data assets in appropriate way to be int

4 min EO methods and products
Sea ice
Lead detection
IB detection
IB tracking
Icenet

4m on incorporating with DTs
- SDA integration
- Get maria on board for a route planner slide
- Andrew for an Ant DT use case argument
-
-
## Title
Developing Digital Twins in Antarctica

## Abstract
Digital Twins underpin the digital strategy for delivering effective and sustainable environmental science and meeting net-zero obligations. The data driven approaches for developing Digital Twins, in areas studied by the British Antarctic Survey, have comparatively low coverage of data points when compared with other geographical domains that benefit from developed infrastructure and accessibility. As such, Earth Observation (EO) data and subsequent processing techniques are crucial to providing access to real-world data that can benefit the decision-making capabilities of environmental Digital Twins.

Here we describe some example methods being developed to exploit high-cadence EO sources in support of our digital twinning ambitions. We use a convolutional neural network (CNN) to map sea ice extent and density in Sentinel 1 synthetic aperture radar (SAR) imagery, allowing for increased spatio-temporal resolution compared to existing sea ice products. This increases opportunities to understand dynamics, including drivers of the dramatic minimum ice extent extent in 2016. We use an unsupervised non-parametric mixture model to identify icebergs in the same SAR imagery, providing improvements over existing detection methods while also being adaptive to scene variability. This provides the basis for a scalable operational monitoring system facilitating estimation of ice fluxes from the continent, meltwater distributions and navigational hazards. Finally we present Icenet, a CNN trained using microwave and ERA5 reanalysis data to predict sea ice concentrations and distributions.
We illustrate how these methods and data products integrate with each other through the DT framework, enabling diverse scientific and operational insight before highlighting priorities for future development.

## blurb

Antarctic sea ice forms a barriers between the atmosphere and the ocean, modulating the exchanges of heat, momentum, fresh-water, and gases such as CO2 (Bitz et al. 2006). Sea ice influences regional and global climate: its melt reduces regional surface albedo and results in the injection of freshwater into the ocean, contributing towards ocean stratification (Shephard et al. 2018). Sea ice also provides an important habitat for keystone Antarctic species including penguins, seals and sea lions.
Bitz, C.M., Gent, P.R., Woodgate, R.A., Holland, M.M. and Lindsay, R., 2006. The influence of sea ice on ocean heat uptake in response to increasing CO2. Journal of Climate, 19(11), pp.2437-2450.
Shepherd, A., Fricker, H.A. and Farrell, S.L., 2018. Trends and connections across the Antarctic cryosphere. Nature, 558(7709), pp.223-232.
1:39
probably only need a sentence or two from this, but hope it helps gives contextual detail