Parameter Sweeps

The parameter_sweep() function runs the model over the Cartesian product of any supported input parameters and returns a pandas DataFrame. This is useful for sensitivity analysis, look-up table generation, and exploring how albedo responds to changing conditions.

Basic usage

from biosnicar.drivers.sweep import parameter_sweep
 
df = parameter_sweep(
    params={
        "solzen": [30, 40, 50, 60, 70],
        "rds": [100, 200, 500, 1000],
    }
)
print(df[["solzen", "rds", "BBA"]])

Impurity sweeps

df = parameter_sweep(
    params={"black_carbon": [0, 1000, 10000]}
)
print(df[["black_carbon", "BBA"]])

Including spectral output

By default, the sweep returns broadband values. To include the full 480-band spectral albedo:

df = parameter_sweep(
    params={"rds": [500, 1000, 2000]},
    return_spectral=True,
)

Chaining to satellite bands

The sweep result supports .to_platform() to append band and index columns:

df = parameter_sweep(
    params={"rds": [500, 1000], "solzen": [50, 60]},
).to_platform("sentinel2")
 
print(df[["rds", "solzen", "BBA", "B3", "NDSI"]])

Multi-platform comparison:

df = parameter_sweep(
    params={"rds": [500, 1000]},
).to_platform("sentinel2", "landsat8", "modis")
 
print(df[["rds", "sentinel2_NDSI", "landsat8_NDSI", "modis_NDSI"]])

Pivot tables for analysis

df = parameter_sweep(
    params={
        "solzen": [30, 40, 50, 60, 70],
        "rds": [100, 200, 500, 1000],
    }
)
df.pivot_table(values="BBA", index="solzen", columns="rds").plot()

Options

Parameter	Type	Default	Description
`params`	dict	required	`{name: [values]}` for each parameter to sweep
`solver`	str	`"adding-doubling"`	`"adding-doubling"` or `"toon"`
`return_spectral`	bool	False	Include 480-band spectral albedo in output
`progress`	bool	True	Show tqdm progress bar

Supported parameter keys: solzen, direct, incoming, rds, rho, dz, layer_type, and impurity names.

Working with Impurities Plotting