Prediction and predictability of land and atmosphere initialized CCSM4 climate forecasts over North America. Issue 21 (8th November 2016)

• Record Type: Journal Article
• Authors: Infanti, Johnna M.; Kirtman, Ben P.
• Is Part Of: Journal of geophysical research. Volume 121:Issue 21(2016)
• Page Start: 12, 690
• Page End: 12, 701
• Publication Date: 2016-11-08
• Subjects: climate prediction -- precipitation -- North America -- land/atmosphere initialization -- predictability; Atmospheric physics -- Periodicals; Geophysics -- Periodicals; 551.5
• DOI: 10.1002/2016JD024932
• Languages: English
• ISSNs: 2169-897X

Abstract

Abstract: Subseasonal‐to‐seasonal prediction is influenced by slowly varying surface fields such as sea surface temperature (SST) and soil moisture. Fully coupled hindcasts were recently completed in the Community Climate System Model version 4.0 (CCSM4) as part of the North American Multi‐Model Ensemble project. Using similar land and atmosphere initialization strategies, but with prescribed climatological SSTs, we attempt to determine the isolated impact of combined observed atmosphere and land initialization and of observed atmosphere initialization on monthly precipitation and 2 m temperature prediction‐estimated skill (i.e., skill assessed without SST variability) and predictability on monthly time scales. CCSM4 has been cited as having low land‐atmosphere coupling, and while combined land and atmosphere initialization significantly increases the estimated skill of precipitation and temperature in the first month after initialization (lead 0), land initialization influence is weak, consistent with low land‐atmosphere coupling in CCSM4. In contrast, atmosphere initialization is a stronger contributor to prediction skill and predictability. We find stronger influence of land and atmosphere initialization on precipitation in CCSM4 versus results from CCSM3. Predictability results show that there is potential skill to be gained for both precipitation and temperature should model errors, atmosphere or land initial state errors, and/or land‐atmosphere coupling improve. Key Points: Observed land and atmosphere initial states increase skill of both precipitation and 2 m temperature in CCSM4 hindcasts Perfect model predictability indicates skill to be gained should model errors, initial errors, or land‐atmosphere coupling strength improve