Table 1.

Set of best-ranked models, according to AIC values, examining variations in adult elephant numbers drinking from waterholes in the KNP. Each line of the table represents a model that includes the variables that have estimated values. The estimates of the effect of each variable are averaged across the five models with AIC values differing less than seven units from the best model. The relative importance of each variable is calculated according to Akaike weights (across all possible 52 models). The results of this table correspond to the eight pairs of reservoir + pan; the results of the models including all 21 waterholes are in electronic supplementary material, table S1. CFU, bacteria abundance; COD, chemical oxygen demand; NTU, water turbidity.

	(intercept)	CFU	COD	NTU	elephant abundance	water hardness	CFU × COD	CFU × NTU	d.f.	AIC	ΔAIC	weight	R2	R2 conditional
	1.653	−0.525							4	23.198	—	0.710	0.572	0.771
	1.559	−0.630		0.201					5	27.444	4.246	0.085	0.673	0.779
	1.392	−0.518			0.182				5	28.108	4.910	0.061	0.589	0.784
	1.718	−0.526				−0.027			5	28.627	5.430	0.047	0.546	0.781
	1.730	−0.514	−0.058						5	28.651	5.453	0.046	0.556	0.764
averaged estimates (s.e.)	1.635 (±0.237)	−0.534 (±0.103)	−0.003 (±0.052)	0.018 (±0.069)	0.012 (±0.069)	−0.001 (±0.054)
variable importance		0.981	0.062	0.107	0.073	0.058	0.003	0.01