The height growth model structure is identical to the diameter
growth model with the exception that total height (tht) replaces
diameter at breast height (dbh) for the variables associated with
parameters b1 and b2. A suite of models for each species was constructed
to include the range of potential parameter values (e.g. for
TEMP, mean annual or seasonal maximum or minimum or number
of degree days were included). Final model structure for each species
was selected using the Akaike Information Criterion and Residual
Maximum Likelihood Estimation (West et al., 2007) to evaluate
explanatory power and parsimony. Final model parameters varied
by species and height versus diameter growth (Supplementary
Tables S2–S11). As an example, winter precipitation provided the
most explanatory power for ponderosa pine diameter growth and
winter, spring and summer precipitation provided the best fit for
height growth. For sugar pine, winter precipitation was the only
precipitation explanatory variable for diameter growth, while
spring and summer precipitation provided the most explanatory
power for height growth. Additionally, the sign of each coefficient
varied by species (Supplementary Tables S12–S21). The growth
models were then incorporated into the FVS model code, forming
the climate sensitive variant. Model validation was conducted
using tree growth data from different sites within the geographic
region where the parameterization data were gathered. Modeled
height and diameter were compared to empirical data for each of
the four sites. These sites were located in northern, central, and
southern Sierra Nevada and from the coastal mountains due west
across the Central Valley. The existing FVS mortality models were
retained and therefore the direct influence of climate on mortality
is unaccounted. Furthermore, this approach did not account for
species-specific physiological responses to changes in climate
and atmospheric CO2 concentration (see Section 4). For an in-depth
description of model development, parameterization, and validation,
see Robards (2009).