Abstract 232

Abstract ID: 232

The Predictability of the South Asian Summer Monsoon Intra-Seasonal Variability (Active-Break Cycle) from re-forecasts of the ECMWF ensemble prediction system at three resolutions.

Lead Author: David Martin Straus
George Mason University, United States of America

Keywords: Monsoon, Active-Break Cycle, Circulation Regimes, Diabatic Heating, Regime Transitions

Abstract: The South Asian summer Monsoon intra-seasonal variability has been diagnosed with minimal time-filtering in terms of transition paths between large-scale Circulation Regimes based on the circulation fields of winds and vertical motion from ERA5 reanalyses. This approach captures clearly both the active-break cycle and the broad tropical Boral Summer Intra-Seasonal Oscillation in a unified manner, as verified in the circulation fields, diabatic heating and precipitation. Cases of the expected northward propagation are clearly identified, as well as unusual periods in which this propagation is not observed.
Intra-Seasonal circulation regimes are identified from a cluster analysis base on the principal components (PCs) of the pentad anomaly fields of 850 hPa horizontal winds from the ERA-Interim reanalysis for the 120 days starting 01June for the years 1979 – 2018 over the broad Indian region (5 E-100 E; 5 S – 35 N). The anomalies are formed with respect to a parabolic (in time) seasonal cycle computed separately for each year, thus filtering out periods of greater than 240 days. The k-means clustering method yields clusters that are significant compared to synthetic data sets for any value of k > 3, where k is the number of clusters.
The transition matrices (giving the number of transitions between regimes) establish that the system is most likely to stay in the same cluster from one pentad to the next, but that the significant transitions (with 95% confidence level using a modified bootstrap method) form a cycle. The similarity between the cycle as depicted from 4 or 5 clusters is established by composites of 850 hPa winds, 200 hPa divergence, 500 hPa vorticity and vertical pressure velocity: Strong convection (inferred from rainfall, vertical pressure velocity, divergence and vorticity) over the subtropical Indian Ocean, moves to the central Bay of Bengal and over central India, then subsequently to the northern Bay of Bengal and west Bengal, and then further north into the Himalayas. (The Indian rainfall, composited over the periods of the regimes, show a similar cycle.) The phases in which strong convection is seen over central and northern India are seen for about 60% of the time for both k=4 and k=5 analyses.
The number of complete cycles (including a return to the starting cluster) found in the 40 years of data is 7 in the 4-cluster analysis, while the number of times the system undergoes four (three) consecutive legs of the cycle is 16 (31). Fewer instances of complete cycles are found for 5 clusters (only 3), but sequences of five, four and three consecutive legs occur 10, 11 and 28 times respectively. The transition paths between circulation regimes can be seen in greater detail in trajectories of the system in low-dimensional PC-space.
We analyze the predictability of this subseasonal evolution of circulation in extensive sets of boreal summer forecasts of the coupled ECMWF Integrated Forecast System at three different horizontal resolutions: 60 km, 18 km and 9 km. For each resolution, we analyze forecasts of 60 days from initial conditions of 01Jun, 01Jul and 01Aug for 30 years (1986-2015) with 15 ensemble members. The forecast anomaly circulation fields in the broad Indian region are projected onto the Empirical Orthogonal Functions of ERA5 to obtain pseudo-PCs that characterize the forecast in the same phase space as ERA5. The predictability of the individual circulation regimes, and the transition paths comprising the intra-seasonal oscillation, are evaluated by comparing the growth of errors between the PC trajectories (from ERA5) and the trajectories of the pseudo-PCs (from the forecasts).