AquaCropUC Davis Global Environmental Change LabView code

One flexible framework

Calibration at the scale your question demands.

Bring the best available observations into one transparent workflow—from detailed field experiments and farm records to remote-sensing products, regional statistics, and long subnational yield records. The framework calibrates AquaCrop at the spatial scale required by the research question while maintaining a consistent, reproducible treatment of parameters, observations, and uncertainty.

Local

Field-scale data

Calibrate phenology, canopy, biomass, soil moisture, and yield against observations from trials and farms.

  • In situ observations
  • Management records
  • Multiple target variables

Regional

Remote sensing + counties

Merge satellite observations with county yield records to constrain crop dynamics across heterogeneous landscapes.

  • Canopy and soil-moisture signals
  • County-level yield
  • Spatially explicit calibration

Continental to global

Subnational yield data

Use long yield records and representative locations to build robust crop parameterizations across climate zones.

  • Global sensitivity analysis
  • Parallel calibration
  • Gridded simulation

Our framework

Screen. Calibrate.
Evaluate.

A transparent three-stage workflow selects climate-representative rainfed and irrigated locations, uses global sensitivity analysis to identify the most influential parameters, and applies Shuffled Complex Evolution to calibrate them efficiently. Evaluation beyond the calibration locations shows that this approach improves the model accuracy across climates and management systems.

Global sensitivity analysis

Calibrate what matters.

Sobol analysis screens the full model and retains parameters with meaningful total-order influence across climate, soils, and management.

Canopy decline
Water productivity
Stomatal control
Root depth

Calibration engine

Explore a difficult parameter space.

Shuffled Complex Evolution evolves candidate solutions in parallel and regularly shares information—combining broad exploration with focused refinement while resisting local optima.

Sample→Evolve→Shuffle→Converge

U.S. maize evaluation

Seventeen counties. Continental reach.

Twenty years of observations from nine rainfed and eight irrigated counties improved performance across U.S. maize-producing regions.

Mean absolute bias28% → 16%
Normalized RMSE47% → 26%
Temporal correlation0.32 → 0.42

Wu et al. (2026), manuscript in preparation

Open science

Research code,
ready to extend.

The UC Davis implementation reproduces the official FAO model while opening the workflow to Python's data, visualization, calibration, and machine-learning ecosystem.

AquaCrop Python Website repository
aquacrop · python
from aquacrop import AquaCropModel

model = AquaCropModel(
  weather=climate,
  soil=soil_profile,
  crop="Maize",
  management=scenario
)

results = model.run()