It has recently been argued that single-model seasonal forecast ensembles are overdispersive, implying that the real world is more predictable than indicated by estimates of so-called perfect model predictability, particularly over the North Atlantic. However, such estimates are based on relatively short forecast data sets comprising just 20 years of seasonal predictions. Here we study longer 40 year seasonal forecast data sets from multimodel seasonal forecast ensemble projects and show that sampling uncertainty due to the length of the hindcast periods is large. The skill of forecasting the North Atlantic Oscillation during winter varies within the 40 year data sets with high levels of skill found for some subperiods. It is demonstrated that while 20 year estimates of seasonal reliability can show evidence of overdispersive behavior, the 40 year estimates are more stable and show no evidence of overdispersion. Instead, the predominant feature on these longer time scales is underdispersion, particularly in the tropics.
(private-note)The competition, in the ongoing Palmer vs Scaife epic.
They say that GloSea5 estimates of skill -- and the underconfidence of the model -- are wrong because the 20-year hindcast is too short.
However, the thorough Bayesian analysis of Siegert et al (2015+) (http://arxiv.org/abs/1504.01933 p18) says that their conclusions are flat out wrong -- GloSea is genuinely underconfident.
%0 Journal Article
%1 Shi2015Impact
%A Shi, W.
%A Schaller, N.
%A MacLeod, D.
%A Palmer, T. N.
%A Weisheimer, A.
%D 2015
%J Geophysical Research Letters
%K verification skill seasonal nao
%N 5
%P 1554--1559
%R 10.1002/2014gl062829
%T Impact of hindcast length on estimates of seasonal climate predictability
%U http://dx.doi.org/10.1002/2014gl062829
%V 42
%X It has recently been argued that single-model seasonal forecast ensembles are overdispersive, implying that the real world is more predictable than indicated by estimates of so-called perfect model predictability, particularly over the North Atlantic. However, such estimates are based on relatively short forecast data sets comprising just 20 years of seasonal predictions. Here we study longer 40 year seasonal forecast data sets from multimodel seasonal forecast ensemble projects and show that sampling uncertainty due to the length of the hindcast periods is large. The skill of forecasting the North Atlantic Oscillation during winter varies within the 40 year data sets with high levels of skill found for some subperiods. It is demonstrated that while 20 year estimates of seasonal reliability can show evidence of overdispersive behavior, the 40 year estimates are more stable and show no evidence of overdispersion. Instead, the predominant feature on these longer time scales is underdispersion, particularly in the tropics.
@article{Shi2015Impact,
abstract = {It has recently been argued that single-model seasonal forecast ensembles are overdispersive, implying that the real world is more predictable than indicated by estimates of so-called perfect model predictability, particularly over the North Atlantic. However, such estimates are based on relatively short forecast data sets comprising just 20 years of seasonal predictions. Here we study longer 40 year seasonal forecast data sets from multimodel seasonal forecast ensemble projects and show that sampling uncertainty due to the length of the hindcast periods is large. The skill of forecasting the North Atlantic Oscillation during winter varies within the 40 year data sets with high levels of skill found for some subperiods. It is demonstrated that while 20 year estimates of seasonal reliability can show evidence of overdispersive behavior, the 40 year estimates are more stable and show no evidence of overdispersion. Instead, the predominant feature on these longer time scales is underdispersion, particularly in the tropics.},
added-at = {2018-06-18T21:23:34.000+0200},
author = {Shi, W. and Schaller, N. and MacLeod, D. and Palmer, T. N. and Weisheimer, A.},
biburl = {https://www.bibsonomy.org/bibtex/27f88aeece848de8078423878b0dba6c8/pbett},
citeulike-article-id = {13539529},
citeulike-linkout-0 = {http://dx.doi.org/10.1002/2014gl062829},
comment = {(private-note)The competition, in the ongoing Palmer vs Scaife epic.
They say that GloSea5 estimates of skill -- and the underconfidence of the model -- are wrong because the 20-year hindcast is too short.
However, the thorough Bayesian analysis of Siegert et al (2015+) (http://arxiv.org/abs/1504.01933 p18) says that their conclusions are flat out wrong -- GloSea is genuinely underconfident.},
day = 16,
doi = {10.1002/2014gl062829},
interhash = {b9fb31d60f5d45f023d30ec284347bb5},
intrahash = {7f88aeece848de8078423878b0dba6c8},
issn = {0094-8276},
journal = {Geophysical Research Letters},
keywords = {verification skill seasonal nao},
month = mar,
number = 5,
pages = {1554--1559},
posted-at = {2015-08-05 17:52:05},
priority = {2},
timestamp = {2018-06-22T18:35:16.000+0200},
title = {Impact of hindcast length on estimates of seasonal climate predictability},
url = {http://dx.doi.org/10.1002/2014gl062829},
volume = 42,
year = 2015
}