Configure a crop-yield model comparison
Description
Configure a crop-yield model comparison
Usage
crop_model_config(
methods = crop_model_methods(),
weather_variables = c("SLR", "PCP", "MAX", "MIN", "HMD", "DIF", "TAS"),
initial_fraction = 0.35,
tuning_fraction = 0.3,
grids = NULL,
interval_level = 0.8,
robust = TRUE,
stress_temperature = NULL,
stress_precipitation = NULL,
calibration_years = 0L
)
Arguments
methods |
Character vector. 'persistence' is the last-observation benchmark. 'ensemble' is learned from the requested base methods. If it is requested alone, all base methods are fitted. |
weather_variables |
Monthly weather variables used when raw files are read. Drought indices are not included by default. |
initial_fraction |
Fraction of observed years assigned to the initial history before rolling hyperparameter tuning. |
tuning_fraction |
Fraction of observed years assigned to rolling tuning. Remaining years form optional calibration and later evaluation block. |
grids |
Named list of hyperparameter grids. 'NULL' uses the documented package defaults. |
interval_level |
Coverage level for empirical prediction intervals. Currently uses 0.80 so interval columns and diagnostics remain unambiguous. |
robust |
Whether to use robust iteratively reweighted fitting where supported. |
stress_temperature, stress_precipitation |
Optional weather-variable labels for compound hot-dry terms in 'stress_lag'. Specify both or neither. Matching calendar months and seasonal positions are required. Thresholds are standardized anomalies, not physiological temperature thresholds. |
calibration_years |
Number of years reserved after tuning and before evaluation for interval calibration. Zero preserves tuning-error calibration. A positive value separates interval calibration from parameter/method selection. |
Value
An object of class 'crop_model_config'.
Examples
crop_model_config(methods = c("trend", "hierarchical"))