internal exchange sums to zero
Symmetric exchanges cancel in the unweighted anomaly sum under the paper’s equal implicit capacities. This is not a measured atmospheric energy budget.
Local cores and symmetric exchange in P2Choose how to read
The full view retains the technical content, figures and sources. The plain-language explanation introduces the ideas without changing the findings.
Testing extra information for temperature forecasts
How to use this companion
Use the controls to inspect the relationship described by the paper. The interactive display is explanatory; it does not rerun the archived training experiment or supply independent evidence of forecast skill. Read the linked full manuscript for the executed comparison.
What the study asks
Does adding satellite information, neighbouring cells or a simplified physical rule improve seven-day temperature forecasts at three Jamaican grid cells?
How the comparison works
The study compares local statistical and neural models with frozen pretrained models on the same temperature targets. Matched comparisons change one component at a time. Scores are calculated against historical reanalysis estimates, not thermometers at every named location.
What the result means
The physical rule improves its matched network in the 2026 subset, but the longer-period uncertainty interval does not establish a consistent gain. Extra satellite predictors and more neighbouring cells do not reliably help. TimesFM remains a strong reference.
Where the evidence stops
Some evaluation years had been inspected earlier, and finalised inputs do not reproduce what was available when a real forecast would have been issued. The paper therefore does not establish operational forecasting skill or local street-level temperatures.
Use the full view for the complete technical discussion, figures, interactive elements and sources.
P2 · INTERACTIVE 3D COMPANION
Compare temperature forecasts and heat exchange
Watch temperature move through connected cells. Change heat input, exchange and observation noise, then compare the resulting fields.
The question
Real geographic contextCalculated synthetic fields
ReferenceEstimate
Loading the 3D scene…
The WebGL 3D view is unavailable
A 2D fallback may appear in its place. You can also follow the study with these steps and the full paper.
- The question How far can a reduced heat-exchange rule take a temperature forecast? The markers locate six constructed temperature reservoirs; the surface joins their values for viewing.
- The method A reduced model combines exchange between cells with an approximate heat input held over a forecast interval. Watch heat move between the cells. Change one parameter at a time to keep the comparison interpretable.
- Check the result Compare prediction with the constructed reference and the baseline. A visually smooth or consistently exchanged field can still miss the reference. Inspect the error after changing heat input or sensor noise.
Synthetic heating multiplier; not a fitted radiation coefficient.
Assumed coupling in the six-reservoir teaching model.
Noise added to the synthetic observation markers.
Drag to rotate. Focus the scene to zoom with the wheel; arrow keys rotate and + / − zoom. Click a cell to inspect its transfers.
The question
How far can a reduced heat-exchange rule take a temperature forecast? The markers locate six constructed temperature reservoirs; the surface joins their values for viewing.
Synthetic temperature forecasts use transparent stand-ins; no TimesFM or other foundation-model checkpoint runs in this browser.
Data, views and methods
Calculated values
What this browser model represents
The browser runs six equal-capacity reservoirs with external forcing. An approximate heat input drives a seven-day synthetic forecast, compared with persistence. It does not execute TimesFM, Chronos, WeatherNext or ECMWF IFS.
The three-dimensional field interpolates the small model for viewing. It is not a resolved three-dimensional atmospheric simulation.
P2’s temperature targets are ERA5-derived grid cells. The paper’s numerical results and uncertainty intervals remain in the manuscript; live browser values do not replace them.
Method in the paper
compare matched inputs, architecture and seeds
The paper isolates the reduced closure by comparing matched graph networks with and without its penalty.
Reduced closure and matched controls in P2Geography and figure design
Terrain is a sampled Mapzen/Tilezen elevation mesh, with a 62.5 m display spacing over an 8 km square. Source tiles have approximately 36 m pixel spacing at this latitude; display spacing does not establish survey accuracy. Colours show the calculated field, while terrain heights provide location context. WebGL displays a 129 × 129 terrain grid; the CPU compatibility view resamples the same elevations to 33 × 33 for responsiveness. Neither display resolution is an accuracy claim. The location selector does not change the numerical model. P6 soil columns and P7 touching heat cells use diagram dimensions. Changing the cube count changes the synthetic lattice and replays the elapsed scenario; it does not refine an atmospheric forecast. P8 markers use requested pilot coordinates, with illustrative model values. P1 column heights repeat rainfall, or absolute error in Difference view, at diagram scale; they are not millimetres on the terrain. P8 column heights repeat the probability or absolute error; their signs come from the colour scale and inspector. P6 shows zero in the reference column when displaying error. Coastline extrusion, sensor shapes, heat-volume sizes and particle speeds are explanatory symbols; they are not surveyed buildings, physical parcel sizes or measured motion.
Sources: Mapzen/Tilezen Terrain Tiles; USGS and NOAA terrain attribution; Natural Earth, public domain. Regional outlines are at 1:10 million, displayed with an equirectangular projection at 17°N. Terrain uses a local equirectangular display. Elevation exaggeration is shown above the scene.
Sequential field colours follow viridis; signed errors use a centred blue–orange scale. The colour limits stay fixed as the example runs; values outside them are counted in the legend. The plan view removes perspective. Figure design follows Crameri et al. (2020); pause and reduced motion follow WCAG motion guidance. The browser export retains a source caption. No atmospheric model skill follows from this animation.
Asset versions and checksums · Coordinate definitions · Numerical model source
Source sections
Satellite predictors, a reduced physical closure and frozen foundation models for seven-day temperature forecasts at Jamaican grid cells
Terms explained
Select a term to read its meaning in this study.
Reanalysis
A historical estimate made by combining observations with a physical model. It is useful reference information, but it is not an independent measurement at every grid cell.
Closure
An approximate rule for processes that a model does not represent explicitly. The rule can be useful and still miss important physical effects.
Ablation
A comparison that removes or changes one model component to examine what that component contributes.
MAE
Mean absolute error. Add the sizes of the errors, ignoring whether they are positive or negative, and divide by the number of predictions.
Holdout
Data kept separate from model fitting. If those data have already influenced design choices, they are no longer a completely untouched test.