Abstract
This study examines the possible relationship between predictions of weekly and biweekly averages of 10-m winds at 3-week lead time and interannual variability over the western Pacific and Indian Ocean (WP-IO) using Climate Forecast System version 2 (CFSv2) reforecasts for period 1979–2008. There is a large temporal correlation between forecasts and reanalyses for zonal, meridional, and total wind magnitudes at 10 m over most of WP-IO for the average of weeks 1 and 2 (W1 and W2) in reforecasts initialized in January (JIR) and May (MIR). The model has some correlations that exceed 95% confidence in some portions of WP-IO in week 3 (W3) but no skill in week 4 (W4) over most of the region. The model depicts prediction skill in the 14-day average of weeks 3–4 (W3–4) over portions of WP-IO, similar to the level of skill in W3. The amplitude of interannual variability (IAV) for 10-m winds in W1 of JIR and MIR is close to that in reanalyses. As lead time increases, the amplitude of IAV of 10-m winds gradually decreases over WP-IO in reforecasts, in contrast to behavior in reanalyses. The amplitude of IAV of predicted 10-m winds in W3–4 over WP-IO is equivalent to that in W3 and W4 in reforecasts. In contrast, the amplitude of IAV in W3–4 in January and May of the reanalysis is much smaller than IAV of W3 and W4. Therefore, one of the possible causes for prediction skill in W3–4 over subregions of WP-IO is due to a reduction of IAV bias in W3–4 in comparison to IAV bias in W3 and W4.
SIGNIFICANCE STATEMENT
Reliable prediction at the subseasonal time scale using a coupled land–atmospheric–ocean model is useful for making management decisions in agriculture and water management. This study explores a relationship between prediction skill of weekly average surface winds at lead times of 1–4 weeks and interannual variability over the western Pacific and Indian Ocean using the Climate Forecast System version 2 reforecasts during 1979–2008. The model has prediction skill in some subregions that exceeds 95% confidence in week 3 but no skill in week 4. Taking 14-day averages of weeks 3 and 4 produces forecasts whose skill is similar to that in week 3. There is a concurrent reduction in interannual variability during weeks 3 and 4 in reforecasts. A hypothesis is put forward that the subseasonal skill is related to the diminution of variability in the model.
Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/WAF-D-20-0181.s1.
© 2021 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
