Table 1 Performance values for the top-ranked generalised least-squares models.

From: Climate-human interaction associated with southeast Australian megafauna extinction patterns

Model

Area

Lag(ka)

k

DE(%)

AICc

wAICc

LogLik

Extt ~ Pa + EminPa + DFa + (Pa × EminPa) + (DFa × EminPa)

n/coexist

0

8

99.4

634.4

0.6

−306.7

Extt ~ Pa + EminPa + DFa

n/coexist

1

6

99.9

645.8

0.4

−315.5

Extt ~ Pa + EminPa + DFa

n/coexist

2

6

48.8

643.7

0.3

−314.5

Extt ~ Pa + EminPa + DFa

n/coexist

3

6

10.1

644.8

0

−315.1

Extt ~ none

n/coexist

4

–

–

–

–

–

Extt ~ none

n/coexist

5

–

–

–

–

–

Extt ~ none

coexist

0,1,2,3,4,5

–

–

–

–

–

Extb ~ NPPb

n/coexist

0

4

98.8

444.9

0.3

−217.8

Extb ~ Tb + Pb

n/coexist

1

6

74.7

1875.9

0.6

−931.7

Extb ~ Tb + NPPb + (Tb × Pb)

n/coexist

2

7

73.1

1865.5

0.9

−925.4

Extb ~ Tb + NPPb + (Tb × Pb)

n/coexist

3

7

70.4

1866.9

0.8

−926.1

Extb ~ Tb + NPPb

n/coexist

4

5

66.6

1873.1

0.6

−931.3

Extb ~ Tb + NPPb

n/coexist

5

5

68.9

1876.9

0.9

−933.3

Extb ~ Hb + Pb + EminPb

coexist

0

6

81.2

1886.9

0.4

−937.2

Extb ~ Hb + NPPb

coexist

1

5

99.9

422.7

0.8

−205.4

Extb ~ Hb + NPPb

coexist

2

5

99.9

422.9

0.7

−205.5

Extb ~ Hb + NPPb

coexist

3

5

99.9

423.1

0.7

−205.6

Extb ~ Hb + NPPb

coexist

4

5

99.9

424.2

0.7

−206.2

Extb ~ Hb + NPPb

coexist

5

5

99.9

424.6

0.7

−206.4

  1. These top-ranked generalised least-squares models contain climate (T = mean annual temperature, P = mean annual precipitation, EminP = mean annual freshwater availability, NPP = mean annual net primary production and DF = fraction of desert within the grid cell), and human predictors (H) to describe (i) the timing of megafauna extirpation (Extt) in human-megafauna non-coexistence areas (n/coexist) and (ii) in human-megafauna coexistence areas (coexist), (iii) the bearing of timing of megafauna extirpation (Extb) in human-megafauna non-coexistence areas (n/coexist) and (iv) in areas with coexistence (coexist). For each of these four scenarios we included five temporal lags (Lag) between the climate from 0 to 5 ka (at a 1 ka-year time step, with ka = 1000 years) for the period earlier than the estimated timing of megafauna extirpation in each grid cell. Predictor variables subscripted a (Pa, EminPa, DFa) indicate that we used the mean annual anomaly relative to the period 50–30 ka for these variables, whereas predictor variables subscripted b (Tb, Pb, NPPb, EminPb, Hb) indicate that we used the directional bearing. Shown are the number of parameters (k), metric of the model’s structural goodness of fit (%DE), minimised negative log-likelihood (LogLik), weight scaled to a sum of 1 (wAICc) and the Akaike’s information criterion corrected for small sample sizes (AICc) for the highest-ranking model. See details of the full list of generalised least-squares models describing all the combinations among variable predictors in Supplementary Data 2
Back to article page