E N D
Point of view: New Reanalysis makes sense when the guess-field-producing-model has improved over the model used in previous Reanalysis.Also:-) Input data may have improved some. More complete. Errors repaired.-) Known execution errors will be corrected-) Can’t wait too longSo have the forecasts improved???
Has the model improved?? #1 Let’s check the scores of 6hr forecasts against radio-sondes.CDAS (1995;2001) vs GDAS (2001)
How about any improvements to be expected in 1950???, I.e. in a limited data environment.Anecdotal evidence for 1953 gale.
Forecasts valid for Feb, 1, 1953, 3Z • Reanalysis model T62L28 (vintage 1995)
Does a newer model help???T254(2003) vs T62(1995) In addition to resolution, there are many changes in physics. In this 1953 case the analysis is the same.
Other evidence of model improvement:1) Atmospheric tides have hugely improved
Other topics:1) Soil moisture conditions underneath a global Reanalysis. Specify them?, Reanl2 strategy, or a new effort?Thanks to Yun Fan we have global ½ degree monthly analysis 1948-present.2) Playfull asides: Reanalysis without a model (and thus without model biases)For instance one can use EWP or CA as propagator to make the 6hr guess fields.Van den Dool&Anderson3) Reanalysis back to ‘1850’ for monthly conditions using EOT, i.e. orthogonal functions (derived from modern data) tethered to points where we have observations back to ‘1850’.This is how Smith&Reynolds have done the latest SST Reanalysis 1850-present.