newest input age = 1 day + added delay
The paper shifts the previous-day rainfall predictors by the chosen added delay. Information age changes the prediction problem.
Stale-input experiment in P3Choose how to read
The full view retains the technical content, figures and sources. The plain-language explanation introduces the ideas without changing the findings.
Learning when rainfall records arrive late or are scarce
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
How do rainfall models behave when their inputs are old, their probabilities are adjusted using an earlier period, or few local examples are available for training?
How the comparison works
Three experiments use daily CHIRPS records at selected Caribbean cells. They delay the inputs, test a probability adjustment on later dates, and compare local training with training that also uses other locations.
What the result means
At the Jamaica average, seven extra days of delay remove the neural advantage over seasonal climatology. The tested probability adjustment worsens later neural wet-day scores at all seven cells. Sharing training examples helps local neural models, while a simpler local ridge model still wins at Kingston.
Where the evidence stops
The delays are imposed scenarios rather than recovered publication histories. Sharing data also changes the amount of training, and the selected cells do not represent entire countries. Independent rain gauges were not used to verify these results.
Use the full view for the complete technical discussion, figures, interactive elements and sources.
P3 · INTERACTIVE 3D COMPANION
Test delayed rainfall inputs
Watch a synthetic rain field while the model receives older information. Then change how much the model borrows from pooled records.
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 useful is a rain history when the newest available observation is already old? Watch the current reference field and the information lag behind it.
- The method The browser fits local and pooled ridge regressions to invented past records, using only rainfall old enough to be available. Increase input delay, then compare a local estimate with one that gives more weight to the pooled fit.
- Check the result Check whether the prediction follows the reference’s changing wet areas. The baseline and difference layers expose errors that a plausible-looking rain field can conceal.
Added delay in this scenario, not a reconstructed product release history.
Assumed persistence of wet spells in the invented rainfall history. Changing it regenerates the same seeded scenario with a different memory.
Weight given to a pooled ridge fit relative to a local ridge fit, both trained on synthetic records.
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 useful is a rain history when the newest available observation is already old? Watch the current reference field and the information lag behind it.
Synthetic rainfall histories illustrate delay and pooling; displayed errors are not the paper’s CHIRPS measurements.
Data, views and methods
Calculated values
What this browser model represents
This viewer focuses on delay and pooling using constructed histories. Recalibration on a later period is covered in the paper’s separate experiment section.
No CHIRPS observations, trained neural checkpoint, ridge fit or isotonic calibration from the paper is run here.
Rainfall memory changes the persistence of the generated wet spells. Pooling blends separately fitted local and pooled regressions. These controls are not estimates of the paper’s fitted effects.
Method in the paper
compare pooled learning with local controls
Pooling is evaluated against both local neural fits and simpler local models. More data alone does not isolate a model-family effect.
Regional pooling experiment in P3Geography 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
Daily rainfall learning at Caribbean CHIRPS cells under three data constraints: stale inputs, post hoc recalibration and scarce local labels
Terms explained
Select a term to read its meaning in this study.
CHIRPS
Climate Hazards Group InfraRed Precipitation with Station data. A rainfall dataset that combines satellite information, station observations and a long-term rainfall pattern. A value represents a grid cell rather than a direct measurement at every place inside it.
Calibration
For probabilities, agreement between predicted chances and how often events occur. A well-calibrated forecast of 20% should be followed by the event about one time in five across comparable cases.
Brier score
The average squared difference between an event probability and its outcome, recorded as one for an event and zero otherwise. Lower is better on the same cases.
Ridge regression
A fitted linear model that discourages very large coefficients. It is a simple statistical comparison, not a neural network.
Climatology
A summary of past conditions for a season or date. In these experiments it provides a simple comparison using the training record.