AI-Driven Spatio-Temporal Tracking of Extreme Weather Anomalies in Medium-Range Forecasts
Ministry of Earth Sciences (MoES) · Disaster Management · Software
A genuinely well-argued research pipeline that is simply too large — the diffusion downscaler needs compute and data you will not have, so if you take this, scope openly to the tracking stage rather than promising the half you cannot deliver.
What it actually is
Finding where an extreme weather event will actually hit inside a huge global forecast dataset is still largely done by eye, and the coarse grid these models run on smears out exactly the sharp peaks of rain and wind that matter. The ask is a system that automatically finds and tracks the anomaly across a three-to-ten-day window, then sharpens it to a finer grid without flattening the extremes.
What to build
The two-stage pipeline the statement lays out: a graph neural network operating on an icosahedral mesh rather than a flat pixel grid, computing an extreme forecast index against a long reanalysis climatology to isolate anomalies and draw a moving bounding box around each one across the forecast window; and a conditional generative diffusion model taking that cropped region and downscaling it from the coarse ensemble grid to a finer one, trained to generate physically plausible high-amplitude fields rather than to minimise mean error, since minimising mean error is precisely what blurs the peaks — with physics-informed penalties in the loss for states that violate conservation, plus a visualisation layer and an alerting interface that drops a coordinate at the anomaly centre with graded severity.
Smallest thing that wins the room
Take a documented historical cyclone, show the tracker's bounding box following it across the forecast window, then put your downscaled field beside a conventional regression downscaling of the same slice and show the peak intensities surviving in yours and disappearing in theirs.
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 42% of the 226 · #131 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
2/5The compute and data volumes put this beyond what a student team can execute, the architecture is fully dictated so there is nothing to invent, and most submissions will implement the tracking stage adequately while the generative downscaler — the part that solves the stated problem — remains a slide.
Feasibility
2/5This is a research-laboratory pipeline written as a problem statement — training a conditional diffusion model on four-dimensional atmospheric fields against decades of reanalysis is a serious GPU budget and a multi-terabyte data problem, and access to historical global ensemble archives from the sponsoring centre is not something a student team can arrange in the time available.
Innovation scope
1/5Every design decision is already made in the statement — the icosahedral mesh, the message-passing graph network, the extreme forecast index against a reanalysis baseline, the conditional denoising diffusion downscaler, the physics-informed loss penalties and even the software libraries are all specified.
Clarity
5/5Exceptionally detailed, correctly identifying the spectral smoothing problem that motivates the whole approach, specifying both architecture stages, naming the frameworks, the baseline and forecast datasets, the resolutions and all four deliverables.
Effort
MassiveA graph network on a spherical mesh, a conditional diffusion model with custom physics-guided losses, a multi-terabyte four-dimensional data pipeline, a visualisation layer and an alerting API is a research programme, and either stage alone would be a full project.
Demo-ability
MediumThe side-by-side showing preserved peak amplitudes against a blurred conventional downscaling is a genuinely good comparative visual and makes the whole argument in one frame, but it requires the diffusion stage to actually work, which is the least likely part to.
In its favour
- Green flag: The statement correctly identifies a real and well-known failure of conventional downscaling — that optimising mean error blurs exactly the peaks forecasters need — and that framing gives you a sharp, defensible argument and an obvious comparative demo
- Green flag: The tracking stage alone is genuinely achievable on a reduced domain and is a complete, useful contribution if you scope to it honestly
- Green flag: Reanalysis climatology for computing an extreme forecast index is freely available, so the anomaly detection half has no data barrier
- Green flag: Everything is named down to the libraries, so no time is lost on tooling decisions
Against it
- Red flag: Training a diffusion model on atmospheric fields is a research-lab undertaking in both compute and data volume, and this is the stage that actually solves the stated problem — a submission with a working tracker and a conceptual downscaler has answered half the statement
- Red flag: Historical global ensemble archives from the sponsoring centre are not openly distributed, so the forecast inputs the pipeline is designed around may simply be unavailable to you
- Red flag: Innovation scope is effectively zero because the architecture is fully written, so every responsive team builds the same pipeline and only execution differs
- Red flag: Generative downscaling produces plausible high-amplitude fields, which is not the same as correct ones — be clear that the diffusion stage generates a physically consistent realisation rather than a deterministic prediction, because conflating those is the error a forecaster will catch
What you will be writing
- message-passing GNN on icosahedral mesh
- conditional denoising diffusion downscaling
- Extreme Forecast Index against ERA5 climatology
- physics-informed conservation loss terms
- Xarray and Dask on 4D NetCDF/GRIB2 arrays
- amplitude-preserving verification against regression baseline
- Numerical weather prediction post-processing
- Generative modelling for geosciences
- Extreme event tracking
Prior art to read before you start
statistical downscaling preserving extreme amplitudes · graph neural networks on spherical grids · ensemble anomaly detection and tracking
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.