Welcome to BioSNICAR

BioSNICAR is an open-source Python framework for computing, emulating, and inverting spectral albedo of snow and glacier ice containing light-absorbing particles.

from biosnicar import run_model
 
# Predict spectral albedo for dirty glacier ice
outputs = run_model(solzen=50, rds=1000, glacier_algae=50000)
print(outputs.BBA)  # broadband albedo

Spectral albedo curve

Why BioSNICAR?

Snow and ice surfaces reflect sunlight back to space — albedo is one of the most important controls on how fast ice melts. Light-absorbing particles (black carbon, mineral dust, algae) darken the surface and accelerate melt, but quantifying their effect requires solving the radiative transfer equations across 480 spectral bands.

BioSNICAR does this, and more:

Forward model — predict spectral albedo for any combination of ice properties, impurities, and illumination. Two radiative transfer solvers, multi-layer support, five grain shapes, and a bio-optical module for algal pigments.
50,000x faster emulator — a neural network surrogate that reproduces the forward model in ~1 microsecond, making optimisation and MCMC practical. Just three matrix multiplications and a ReLU — no GPU, no PyTorch, no TensorFlow.
Inverse retrieval — estimate ice properties (specific surface area, impurity concentrations) from field spectrometer or satellite observations, with uncertainty quantification.
Native satellite and climate model bands — map 480-band output onto Sentinel-2, Sentinel-3, Landsat 8, MODIS, CESM, MAR, and HadCM3 with a single method call.
Subsurface light field — depth-resolved spectral fluxes, PAR at depth, and radiative heating rates from the adding-doubling solver.

The integrated workflow

Each capability is useful on its own - many people will find value in the forward model alone - but together, they enable a powerful workflow: take satellite spectra, retrieve SSA and impurity concentration with uncertainties, and predict the subsurface photic environment — all in under a second.

Four-step integrated workflow pipeline

from biosnicar.emulator import Emulator
from biosnicar.inverse import retrieve
import numpy as np
 
# Load emulator and retrieve ice properties from satellite bands
emu = Emulator.load("data/emulators/glacier_ice_8_param_default.npz")
 
result = retrieve(
    observed=np.array([0.82, 0.78, 0.75, 0.45, 0.03]),
    parameters=["ssa", "black_carbon", "glacier_algae"],
    emulator=emu,
    platform="sentinel2",
    observed_band_names=["B2", "B3", "B4", "B8", "B11"],
    fixed_params={"direct": 1, "solzen": 50, "dust": 1000, "snow_algae": 0},
)
print(result.summary())

Get started

Quick Start — install BioSNICAR and run your first simulation in 5 minutes
Fundamentals — understand the model structure, inputs, and outputs
Remote Sensing — satellite band convolution and parameter retrieval from observations
Subsurface Light — PAR at depth and spectral heating rates
Fast Emulator — build and use the neural network surrogate
API Reference — complete reference for every function and class
Examples — worked code examples covering the full workflow

Citation

If you use BioSNICAR in a publication, please cite:

Cook, J. et al. (2020): Glacier algae accelerate melt rates on the western Greenland Ice Sheet, The Cryosphere, doi:10.5194/tc-14-309-2020

Flanner, M. et al. (2007): Present-day climate forcing and response from black carbon in snow, J. Geophys. Res., 112, D11202

BioSNICAR is open source on GitHub under the MIT licence. Contributions are welcomed — new issues and pull requests are the best way to get involved.

Quick Start