Contribution of dynamic vegetation phenology to decadal climate predictability

• Published in: Journal of Climate
• Main Authors: Weiss, Martina, Miller, Paul A., Van den Hurk, Bart J., Van Noije, Twan P., Stefanescu, Simona E., Haarsma, Reindert J., Van Ulft, Lambertus H., Hazeleger, Wilco, Le Sager, Philippe, Smith, Benjamin (R19508), Schurgers, Guy
• Other Authors: Hawkesbury Institute for the Environment (Host institution)
• Format: Article in Journal/Newspaper
• Language: English
• Published: U.S., American Meteorological Society 2014
• Subjects: XXXXXX - Unknown; plants; atmospheric temperature; climatology; Sea ice
• Online Access: https://doi.org/10.1175/JCLI-D-13-00684.1; http://handle.westernsydney.edu.au:8081/1959.7/uws:48486

In this study, the impact of coupling and initializing the leaf area index from the dynamic vegetation model Lund–Potsdam–Jena General Ecosystem Simulator (LPJ-GUESS) is analyzed on skill of decadal predictions in the fully coupled atmosphere–land–ocean–sea ice model, the European Consortium Earth System Model (EC-Earth). Similar to the impact of initializing the model with the observed oceanic state, initializing the leaf area index (LAI) fields obtained from an offline LPJ-GUESS simulation forced by the observed atmospheric state leads to a systematic drift.A different treatment of the water and soil moisture budget in LPJ-GUESS is a likely cause of this drift. The coupled system reduces the cold bias of the reference model over land by reducing LAI (and the associated evaporative cooling), particularly outside the growing season. The coupling with the interactive vegetation module implies more degrees of freedom in the coupled model, which generates more noise that can mask a portion of the extra signal that is generated. The forecast reliability improves marginally, particularly early in the forecast. Ranked probability skill scores are also improved slightly in most areas analyzed, but the signal is not fully coherent over the forecast interval because of the relatively low number of ensemble members. Methods to remove the LAI drift and allow coupling of other variables probably need to be implemented before significant forecast skill can be expected.