Table 2.
The top three linear mixed-effect models as ranked by AICc scores, and the first model after a ∆AICc > 2 was achieved.
| Variable | Predictor variables | LogLik | AICc | ∆AICc | Weight | R 2 |
|---|---|---|---|---|---|---|
| ST | AT WS RH SR HF --- ---- B | - 375.909 | 770.9 | 0 | 0.177 | 0.80 |
| AT WS RH SR HF --- SA --- --- | −376.437 | 772.0 | 1.06 | 0.104 | 0.81 | |
| AT WS RH SR --- --- SA --- --- | −377.701 | 772.3 | 1.35 | 0.090 | 0.81 | |
| AT WS RH SR HF M ---- B | −375.904 | 773.2 | 2.25 | 0.057 | 0.80 | |
| IRT-handled | AT WS ---- SR ---- --- ---- B MS dHF ---- | −134.392 | 298.3 | 0 | 0.349 | 0.69 |
| AT WS RH SR ---- M ---- --- MS dHF ---- | −133.601 | 299.8 | 1.46 | 0.168 | 0.52 | |
| AT WS ---- SR ---- M ---- --- MS dHF ---- | −135.195 | 299.9 | 1.60 | 0.156 | 0.61 | |
| ---- WS ---- SR ---- M ---- B MS dHF dST | −134.685 | 301.9 | 3.63 | 0.056 | 0.50 | |
| IRT-unhandled | ---- WS ---- SR MS | −596.061 | 1213.5 | 0 | 0.516 | 0.43 |
| ---- WS RH SR MS | −595.844 | 1215.4 | 1.86 | 0.203 | 0.43 | |
| AT WS ---- SR MS | −595.857 | 1215.4 | 1.88 | 0.200 | 0.46 | |
| AT WS RH SR MS | −595.660 | 1217.3 | 3.82 | 0.076 | 0.44 | |
| HF | AT ---- RH ---- ST --- --- --- MS | −869.152 | 1761.9 | 0 | 0.055 | 0.28 |
| AT ---- RH SR ST --- --- --- MS | −868.224 | 1762.4 | 0.46 | 0.043 | 0.30 | |
| AT ---- RH ---- ST M --- B MS | −867.060 | 1762.5 | 0.48 | 0.043 | 0.30 | |
| AT ---- RH ---- ST --- SA --- MS | −869.119 | 1764.2 | 2.25 | 0.018 | 0.28 |
Models for direct-contact ST (°C), IRT determined ST (IRT, °C) and HF (W/m2) are shown. Predictors include the extrinsic factors of ambient temperature (AT, °C), wind speed (WS, m/s), relative humidity (RH, %) horizontal solar radiation (SR, W/m2), and the intrinsic factors of molt status (MS), mass (M, kg), surface area (SA, m2), average blubber thickness (B, cm) and ST (°C) or HF (W/m2). Models used dorsal ST and HF (dST and dHF) in the IRT models of handled animals, which were based on images of animals taken prior to sedation. Models of unhandled IRT were limited to extrinsic factors and molt status, as those animals were not handled later for full physiological measurements. Animal ID was retained as a random effect in all models. Factors whose increase led to an increase in the variable of interest are shown in bold, whereas factors whose decrease led to an increase in the variable of interest are unbolded. All P values were less than 0.0001.