Table 2. Summary of the best-performing ordinary least squares (OLS) model.

Variable	Year	Estimate	SE	t	P	Signif.
(Intercept)		−16760	9687	−1.73	0.087	.
NDVI (early summer)	1	13.96	5.02	2.78	0.007	**
NDVI (late summer)	4	7.77	5.15	1.51	0.135
NDVI (early summer)	0	6.85	4.81	1.43	0.158
Air temperature	4	1.03	1.66	6.21	<0.001	***
NDVI (late summer)	1	−1.53	4.96	−3.08	0.003	**
Precipitation	0	0.70	0.34	2.05	0.043	*
Year		8.51	4.84	1.76	0.082	.
Air temperature	0	−55.19	22.81	−2.42	0.017	*
Mean annual precip.		−0.58	0.20	−2.90	0.005	**
NDVI (annual)	1	−14.43	6.31	−2.29	0.025	*
NDVI (fall)	0	9.21	3.35	2.75	0.007	**

Potential parameters of the best OLS model (RMSE = 156.9 g C m⁻² yr⁻¹ on 94 d.f., adjusted R² = 0.61, P<0.001) were selected by the CI-RF algorithm before OLS was performed (see Methods and Table 1). Columns include variable included in OLS regression, year of data stream (0 = current year, 1 = previous year, etc.); OLS estimate and standard error (SE); t-value; P-value; and significance (“.” <0.1; “*” <0.05; “**” <0.01; “***” <0.001).