Fig. 1: Correlation skill of the MJO-related precipitation index from the DL model. | npj Climate and Atmospheric Science

Fig. 1: Correlation skill of the MJO-related precipitation index from the DL model.

From: Exploring dominant processes for multi-month predictability of western Pacific precipitation using deep learning

Fig. 1

The correlation skill of the MJO-related precipitation index is shown as a function of the forecast lead month from the DL model for all seasons. The precipitation index is calculated as the average pentad precipitation over the western Pacific region (120°–150° E, 10° S–10° N). The DL model utilizes two predictors: one is the fixed input of the precipitation anomaly map, and the other is selected from various weather variables based on existing MJO theories. The DL model was trained using reanalysis data from 1997 to 2010, and the forecast results from 2012 to 2017 were used for validation. The selected weather variables include the equivalent potential temperature (represented as “theta”), boundary layer moisture convergence (“BLMC”), convective instability (“CIN”), and diabatic heat (“diah”) calculated following Yanai et al. (1972). The zonal (meridional) moisture advection is represented by “udqdqx” (“vdqdy”). Each line in the figure represents the mean of 10 ensemble members.

Back to article page