New Metrics for MJO Prediction Skill Based on Large-Scale Precipitation Tracking
Abstract
The Madden-Julian Oscillation (MJO) is the dominant mode of intraseasonal variability in the tropics, and it influences the global weather and climate via teleconnections. As such, the MJO is regarded as a potential source of subseasonal-to-seasonal (S2S) predictability. Conventionally, EOF based MJO indices such as the RMM Index are used to assess how well global models reproduce and predict the MJO. The method of large-scale precipitation tracking (LPT) is more directly representative of the MJO precipitation and convective heating. In the LPT method, the MJO is identified as a large-scale (> 1000 km) precipitation feature which tracks eastward near the equator for at least a week. In this study, new metrics based on LPT are introduced and applied to operational medium-range (45 day) Climate Forecast System (CFS) model forecasts. The first metric is MJO occurrence to measure the degree to which the model predicts an MJO when the MJO was observed. Additional metrics are also derived for MJO initiation location, centroid location, duration, and intensity. These metrics are assessed in weekly lead time intervals (week 1 6), by region (Indian Ocean, Maritime Continent, and Western Pacific), and by season. The CFS model is found to be skillful (Heidke Skill Score > 0.3) out to 6 weeks in Boreal winter (December February), but only 1 to 2 weeks in Boreal summer (June August). Prediction skill is lowest in the Maritime Continent. False alarms are a major impediment to the prediction skill, though this can be partially mitigated by tuning the rainfall tracking threshold. The LPT method and near real-time verification of the MJO has been experimentally implemented at the NOAA Climate Prediction Center.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.A43H..05K