Table 5.

Hidden Markov Model results for tree ring indices, with spring temperature from the CRU record added to the model as a covariate. Model results are included for the Cloquet Forestry Center (CFC) and Marcell Experimental Forest (MEF) for three species, Pinus resinosa (PIRE), Picea mariana (PIMA), and Larix laricina (LALA). Model performance for the one-state, two-state, and two-state + covariate models were evaluated by the change in Akaike Information Criterion (Δ AIC and Δ BIC). Model log-likelihood is also indicated for all models and is compared among models using a log-likelihood ratio test (• = p < 0.1, *p < 0.05, **p ≤ 0.01). The superior model is indicated in bold

Models with no covariates								Models with CRU temperature covariate
	n state = 1			n state = 2			Model selection	n state = 2 + temp			Model selection
Record	AIC	BIC	Log L	AIC	BIC	Log L	Δ AIC	AIC	BIC	Log L	Δ AIC
CFC PIMA	−90.73	−85.25	47.36	−122.58	−103.43	68.29**	−31.85	−121.69	−97.07	69.85	0.89
MEF PIMA	−32.27	−26.8	18.14	−56.49	−37.33	35.24	−24.22	−59.3	−34.67	38.65*	−2.81
CFC PIRE	−88.66	−83.19	46.33	−115.22	−96.07	64.61**	−26.56	−113.75	−89.12	65.87	1.47
MEF PIRE	−15.65	−11.24	9.83	−20.17	−4.74	17.09**	−4.52	−16.25	3.59	17.13	3.92
MEF LALA	113.29	118.44	−54.65	82.91	100.93	−34.45	−30.38	77.57	100.74	-29.79**	−23.36