OceanEmbed - Satellite Embedding-Based Deep Learning Framework for Reconstruction of Subsurface Ocean Temperature from Surface Satellite Observations.
Ministry of Earth Sciences (MoES) · Space Technology · Software
The best-posed machine learning statement in this range — fixed domain, free named datasets and genuinely independent validation — so spend your first days on the regridding pipeline and report skill by depth honestly rather than hiding the deep levels in an average.
What it actually is
Knowing how warm the ocean is below the surface matters for cyclones, fisheries and heat storage, but the only direct measurements come from a scattered handful of drifting floats. Satellites see the whole surface every day and the surface carries indirect traces of what is happening beneath it. The ask is a model that reads those surface traces and reconstructs the temperature all the way down.
What to build
A reconstruction pipeline over the exact domain the statement fixes — the North Indian Ocean from 5 to 30 degrees north and 45 to 105 east, daily at quarter-degree resolution — with a preprocessing stage harmonising and regridding the six named surface inputs of sea surface temperature, salinity, height anomaly, and the two-component surface current and wind fields onto a common grid; an embedding stage compressing that multi-channel surface state into a latent representation using a convolutional, transformer or autoencoder architecture; a reconstruction head predicting temperature at the fifteen standard depth levels the statement lists from the surface down to a thousand metres; and a validation framework scoring correlation, RMSE and bias against independent float observations that were never in training, with a working proof of concept over the Bay of Bengal or Arabian Sea.
Smallest thing that wins the room
Pick a float profile from a held-out date, show your reconstructed temperature curve overlaid on what the float actually measured all the way to a thousand metres, and put the RMSE-by-depth chart beside it.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 66% of the 226 · #77 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: central ministry statements sat below the average.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
4/5Everything is defined, all data is free and named, validation is against genuinely independent observations rather than a held-out slice of your own training set, and physical oceanography attracts very few teams — the caution is that this reconstruction problem has an existing research literature, so you are competing on execution against published baselines rather than on originality.
Feasibility
5/5Every dataset needed is public and the statement names them — the reanalysis product for training targets, gridded float data for independent validation, and all six surface satellite variables are freely distributed — the domain, resolution and depth levels are fixed for you, and the statement even authorises regridding where a product does not match the required resolution.
Innovation scope
3/5The domain, resolution, input variables, output depths, target and validation datasets and evaluation metrics are all fixed, and even the candidate architecture families are listed, so your latitude sits in the embedding design and how you exploit spatial structure rather than in defining the problem.
Clarity
5/5One of the most precisely posed machine learning statements on the portal — it gives geographic bounds, spatial and temporal resolution, the exact input variable list, all fifteen output depth levels in metres, the training target product, the independent validation source and the skill metrics.
Effort
HeavyThe modelling is standard but the data engineering is not — harmonising six satellite products with different native grids, projections, temporal sampling and missing-data conventions onto a common daily quarter-degree grid is where most of the work sits, and teams consistently underestimate it.
Demo-ability
EasyA reconstructed vertical temperature profile laid over what a float actually measured is immediately convincing, and a depth-resolved section of the Bay of Bengal is visually striking in a way tabular results never are.
In its favour
- Green flag: Validation is against independent float observations rather than a held-out portion of the training product, which is a genuinely rigorous setup and lets you make a defensible skill claim instead of a self-referential one
- Green flag: The domain, grid, inputs and output depths are all fixed by the statement, so you cannot lose time on scoping and cannot be criticised for a convenient choice of region or depth range
- Green flag: The physical reasoning is sound and citable — sea surface height responds to thermocline displacement, so there is genuine information about the subsurface in the surface fields and you can explain why the model works rather than only that it does
- Green flag: A vertical section plot of reconstructed ocean temperature is a striking and unusual visual that will stand out in a room of dashboards
Against it
- Red flag: The regridding and harmonisation pipeline is the real project — six products with different native resolutions, land masks and missing-data conventions will consume more time than training the network, so start there rather than with the model
- Red flag: Reanalysis targets are themselves model output rather than observations, so a model that matches the reanalysis perfectly has learned the reanalysis, and the float validation is the only honest measure of skill
- Red flag: Skill falls off sharply with depth because surface signatures carry less information about the deep ocean, so report by depth level and expect the thousand-metre results to be weak — a single averaged RMSE hides exactly the failure a judge will ask about
- Red flag: Subsurface reconstruction from satellite data has an established research literature, so be ready to say what your embedding approach adds over the published regression and neural baselines
What you will be writing
- convolutional autoencoder surface state embedding
- Vision Transformer spatial encoder for gridded fields
- xarray and CDO regridding pipeline
- GLORYS reanalysis training targets
- gridded Argo independent validation
- depth-wise RMSE, bias and correlation skill scoring
- Physical oceanography
- Satellite remote sensing
- Spatiotemporal deep learning
Prior art to read before you start
subsurface field reconstruction from surface observations · latent representation learning for geophysical fields · validation against independent in-situ profiles
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.