MSTWeatherGen

01 Historical data

Dataset

Examples contain 6 years, 9 sites and 3 variables.

Import my observations

Dataset name Observations CSV Stations CSV (optional if coordinates are included)

UTF-8, comma or semicolon delimiter, 60 MiB maximum. Daily dates without gaps; no imputation or unit conversion is performed.

02 Fit the model

First year Last year

Variables to simulate

Load a dataset first.

Seasonality Calibration seed

Standardize the annual cycleRequires at least 6 complete years.

Estimation mode

Guarded seasonal validates DJF before MAM, then JJA and SON. The native mode runs the official multi-season estimator in one call for comparison.

Scientific parameters & troubleshooting

Optimizer iterations Estimation lag tmax Reference points n1 Neighbour points n2 Stop after

min

The engine automatically selects up to 6 weather types using the selected variables. Low iteration counts are useful for technical tests, not scientific validation.

Season diagnostic

Runs only one season, writes a detailed diagnostic JSON, and does not create a simulation-ready model.

03 Generate realizations

Saved calibration

Fit a complete model before simulating.

Start date Number of days Realizations Autoregressive order First seed Wet-day thresholdDiagnostic: precipitation > threshold, in source units.

Simulation budget

Stop after (minutes)

Consecutive seeds are used for realizations. Simulation dates define a calendar, not a weather forecast or climate-change scenario.

History—observed days

Domain—geographical sites

Variables—simulated jointly

Last run—execution time

WEATHER GENERATOR WORKBENCH

Start with historical observations

Load one of the examples or import observations to inspect the data, calibrate a model and generate stochastic realizations.

From history to plausible weather sequences

Weather types → marginal transformations → space-time Gaussian field → multivariate daily weather.

All computation stays on your machine. No external simulation service is used.

VariableSiteRealization

Daily series

Monthly cycle

Spatial pattern

Longitude / latitude. Select a point to change site.

Empirical distribution functions

The point mass at zero remains visible for precipitation.

Observed / simulated quantiles

The diagonal indicates equal quantiles, from 1% to 99%.

Statistics at selected site

Statistic	Observed	Simulated

Wet/dry frequencies and run lengths use the threshold stored with the simulation.

LOCAL PERSISTENCE

Run history

Preserved across restarts.

No jobs yet.

Start with historical observations

Daily series

Monthly cycle

Spatial pattern

Empirical distribution functions

Observed / simulated quantiles

Statistics at selected site

Temporal autocorrelation

Observations

Simulation

Saved realizations

Run history