Time series of observed Ap and best-fits to Ap using Pα estimates, both using annual averaging intervals. The fitted series use linear regressions and are denoted by a prime: tests for homoskedasticity, normal distribution of residuals, linearity are all readily passed because correlation coefficients are so high. (a) and (b) are for 1996–2016, (c) and (d) for 1964–2016. Panels (b) and (d) show the data availability, giving the fraction f of hours for which an hourly mean value of Pα is available. In (a) and (c) the green line gives the means of all the Ap data, 〈Ap〉_τ=1yr and the blue line gives the means of only the Ap data that are coincident with valid Pα data 〈Ap_c〉_τ=1yr. (Note that in (a) the blue line is essentially identical to the green and so is completely overwritten by it). In (a) the black line uses annual means (τ = 1 yr) of Pα values computed from 5-minute means of interplanetary data, 〈Pα〉′_τ=1yr (i.e., the data are combined and then averaged) and is based on the correlation A in Table 2. The red line uses Pα_ann which is compiled using annual means of all available N_SW, V_SW, B and sin⁴(θ/2) data (i.e., the data are averaged and then combined) and is based on correlation B in Table 2. In panel (d) the black line uses 〈Pα〉_τ=1yr and 〈Ap_c〉_τ=1yr and is based on correlation E in Table 2 whereas the red line uses Pα_ann and 〈Ap_c〉_τ=1yr and is based on correlation D in Table 2.