Compute ToE

[eguil]

• compute ToE as last date for which signal = diff hist-histNat > noise = mult * stddev_histNat
  • per model
  • compute S:
  • hist-histNat diff for each hist member [rho,lat,time,Nh]
    • use histNat_mean for ref
  • Ed H. uses another method from piControl (see 2012 GRL)
  • compute N as std dev of histNat: stddev_histNat [rho,lat]
  • use max of Nhn Stddev
  • S/N is (diff hist-histNat) / (mult * stddev_histNat) : ToE[rho,lat,Nh]
  • ToE = last date for which S/N > 1 (toe1) or 2 (toe2) (we then get a PDF and a mean)
  • ToE may be where model agreement on S/N is over 66%
  • How do these two computations compare on the MME based ToE ?
• do calculations in mme_ave_msk

[eguil]

use median rather than mean for MME TOE (some have not emerged)