nsvb

National Scale Volume and Biomass Estimators — a Python implementation of Westfall et al. (2024), A national-scale tree volume, biomass, and carbon modeling system for the United States, USDA Forest Service GTR WO-104 (doi:10.2737/WO-GTR-104).

Every estimator is validated against the four worked examples in the GTR's "Examples of Tree-Level Calculations" section, to the 12 decimal places the publication prints.

Installation

pip install nsvb

Quick start

from nsvb import estimators

# A live Douglas-fir, 20 in DBH, 110 ft tall, in the Marine ecodivision.
estimators.total_aboveground_biomass(202, 20.0, 110.0, "240")   # 3154.55 lb
estimators.carbon_content(202, 20.0, 110.0, "240")              # 1626.47 lb C

The first four arguments are always spcd, dia, ht, division:

Everything else is optional and describes deductions:

cull is a percentage and cr is a fraction — a 30 percent crown ratio is cr=0.30, not cr=30. Passing cr=30 raises.

Dead trees and decay class

Dead-tree reduction is applied through decaycd. Following GTR table 1, the stem wood is scaled by a wood density proportion, bark and branches by further proportions, and foliage is zero for any dead tree. Dead trees also use a different carbon fraction (table S10b, by decay class) than live trees (table S10a, by species).

# A dead tanoak, decay class 2, 10% cull, broken top at 21 ft.
estimators.agb_predicted_reduced(
    631, 11.3, 28.0, "M240", cull=10.0, ah=21.0, decaycd=2, province="M242",
)                                                               # 241.97 lb
estimators.carbon_content(
    631, 11.3, 28.0, "M240", cull=10.0, ah=21.0, decaycd=2, province="M242",
)                                                               # 114.45 lb C

FIADB populates DECAYCD only for standing dead trees, so live trees arrive as NULL. decaycd=None and decaycd=NaN are both treated as live.

Broken tops and province

For a broken-top tree the remaining branch and foliage mass depends on the crown ratio. If you observed one, pass cr. If you did not, the estimator falls back to the Table S11 regional mean, which is keyed by ecoprovince (e.g. "M242"), not ecodivision ("M240"). Passing a division where a province is expected still returns a number, but a different one — up to about 27 percent on branch biomass — so the package warns rather than resolving it silently. Omitting province falls back to a national mean and warns too.

Gross vs. reduced

Bare names are the gross model prediction with no deductions; the _reduced and _sound suffixes carry the cull, dead-tree and broken-top reductions. The gross forms are not just a convenience — agb_reduce_factor needs them undeducted for its denominator.

estimators.total_stem_wood_dry_weight(...)          # Wtotib   (gross)
estimators.total_stem_wood_dry_weight_reduced(...)  # Wtotibred

What's available

Roughly following the GTR's numbered steps:

Not implemented

Belowground coarse-root biomass. The GTR describes it in a single sentence (p. 11) — the approach of Heath et al. (2009), modified to use NSVB merchantable stem wood volume and the table 1 density proportions for standing dead trees — but publishes no equation, no coefficients (there is no belowground table among S1–S20), and no worked example. Implementing it needs the Heath et al. (2009) method, and substituting a different published root model would not be the NSVB quantity. Note that FIADB REF_SPECIES ships JENKINS_ROOT_RATIO_B1/B2, but those are the Jenkins et al. (2003) ratio applied to aboveground biomass — a different method from the one the GTR describes.

Arrays

Every estimator accepts scalars or array-likes and returns the matching shape, so a whole inventory can go through in one call:

import numpy as np

estimators.carbon_content(
    np.array([202, 631]),
    np.array([20.0, 11.3]),
    np.array([110.0, 28.0]),
    np.array(["240", "M240"]),
    np.array([0.0, 10.0]),          # cull
    np.array([np.nan, 21.0]),       # ah — NaN means an intact top
    np.array([0, 2]),               # decaycd
    np.array([np.nan, np.nan]),     # cr — NaN means use the Table S11 default
    np.array(["", "M242"]),         # province
)

Missing measurements propagate as NaN rather than raising, so a batch with unrecorded heights yields NaN for those rows and real numbers everywhere else. Out-of-domain values (negative diameter, cull > 100, ah > ht, an unknown species code) raise ValueError — the underlying models are unguarded power functions that would otherwise return complex numbers or silently wrong results.

Accuracy notes

Two quantities cannot be reproduced from the published supplementary CSVs to the precision of the GTR's printed examples:

• Ten rows in tables S2a/S6a/S7a store the a coefficient in Excel scientific notation truncated to three significant figures (e.g. 3.19E-05 where the GTR text uses 0.000031886237), which drifts bark quantities by up to ~4e-4 relative.
• The GTR's printed merchantable and sawlog heights are under-converged iterates — the residual of equation 7 at the published values is ~1e-6. This package solves to ~1e-14, so the two differ by ~1e-6 relative.

Both are publication-precision limits rather than implementation differences, and the test suite documents each one where it applies.

References

Westfall, James A.; Coulston, John W.; Gray, Andrew N.; Shaw, John D.; Radtke, Philip J.; Walker, David M.; Weiskittel, Aaron R.; MacFarlane, David W.; Affleck, David L.R.; Zhao, Dehai; Temesgen, Hailemariam; Poudel, Krishna P.; Frank, Jereme M.; Prisley, Stephen P.; Wang, Yingfang; Sánchez Meador, Andrew J.; Auty, David; Domke, Grant M. 2024. A national-scale tree volume, biomass, and carbon modeling system for the United States. Gen. Tech. Rep. WO-104. Washington, DC: U.S. Department of Agriculture, Forest Service. 37 p. https://doi.org/10.2737/WO-GTR-104