Band Convolution

BioSNICAR outputs 480-band spectral albedo at 10 nm resolution. Satellites and climate models observe the surface through coarser spectral windows. The .to_platform() method maps the high-resolution spectrum onto platform-specific bands, enabling direct comparison between simulated and observed albedos.

Quick start

.to_platform() chains directly onto run_model():

from biosnicar import run_model
 
s2 = run_model(solzen=50, rds=1000).to_platform("sentinel2")
print(s2.B3)     # green-band albedo
print(s2.NDSI)   # Normalized Difference Snow Index
 
cesm = run_model(solzen=50, rds=1000).to_platform("cesm2band")
print(cesm.vis)  # VIS broadband albedo
print(cesm.nir)  # NIR broadband albedo

Supported platforms

Spectral indices

Spectral indices are computed automatically where platform bands permit:

s2 = run_model(solzen=50, rds=1000, glacier_algae=50000).to_platform("sentinel2")
print(f"NDSI: {s2.NDSI:.3f}")
print(f"Impurity Index: {s2.II:.3f}")

Multi-platform comparison

A single Outputs object can be convolved onto multiple platforms:

outputs = run_model(solzen=50, rds=1000)
for plat in ["sentinel2", "landsat8", "modis", "cesm2band"]:
    r = outputs.to_platform(plat)
    print(f"{plat:12s}  bands: {len(r.band_names):2d}  indices: {r.index_names}")

With parameter sweeps

.to_platform() also works on sweep results, appending band columns to the DataFrame:

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

For multiple platforms, columns are prefixed:

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

GCM parameterisation

Generate band albedos for climate model radiation schemes:

cesm = run_model(solzen=60, rds=1000).to_platform("cesm2band")
print(f"CESM VIS: {cesm.vis:.4f}, NIR: {cesm.nir:.4f}")
 
mar = run_model(solzen=60, rds=1000).to_platform("mar")
print(f"MAR sw1: {mar.sw1:.4f}, sw2: {mar.sw2:.4f}")

Convolution methods

Two methods are used depending on the platform:

SRF convolution (satellite platforms)

Used by Sentinel-2, Sentinel-3, and Landsat 8. Each band has a Spectral Response Function (SRF) describing its wavelength sensitivity:

band_albedo = sum(albedo × SRF × flx_slr) / sum(SRF × flx_slr)

SRFs are stored as CSV files in data/band_srfs/ and use tophat approximations by default. These can be replaced with official manufacturer SRFs for higher fidelity.

Flux-weighted interval averaging (GCMs and MODIS)

Used by CESM, MAR, HadCM3, and MODIS. The band albedo is the flux-weighted mean over all model grid points within the wavelength interval:

band_albedo = sum(albedo[lo:hi] × flx_slr[lo:hi]) / sum(flx_slr[lo:hi])

Adding new platforms

New platforms can be added by providing a SRF CSV file in data/band_srfs/ with 480 rows (one per model wavelength) and one column per band, without modifying the core code.