Interannual variability of the occurrence of MJO at different

20 In the present study, we investigate the interannual variability of the occurrence of the Madden Julian 21 Oscillation (MJO) at different Real-time Multivariate MJO (RMM) phase regions (MJO frequency) and its 22 association with the El Niño Southern Oscillation (ENSO). Evaluating the all-season data, we identify the 23 dominant zonal patterns of MJO frequency exhibiting prominent interannual variability. Using Principal 24 Component Analysis Biplot (PCA Biplot) technique, we demonstrate that the MJO frequency has two 25 distinct modes of variability related to RMM1 and RMM2 spatial patterns. The first spatial mode of MJO 26 frequency related to RMM1 is associated with a higher frequency of MJO active days over the Maritime 27 Continent and a lower frequency over the central Pacific Ocean and the western Indian Ocean, or vice versa. 28 The second mode related to RMM2 is associated with a higher frequency of MJO active days over the 29 eastern Indian Ocean and a lower frequency over the western Pacific, or vice versa. We find that these two 30 types of MJO frequency patterns are associated with the central Pacific and eastern Pacific ENSO modes, 31 respectively. These MJO frequency patterns are the lag response of the underlying ocean state. 32


132
The arrows which group together by having the same direction in the two-dimensional plane,

141
We obtained correlation coefficients and cosine square, two statistical quantities, to quantify the 142 goodness of representation of the MJO frequency variation through the leading EOFs. The cosine square 143 parameter represents the percentage of the variance of MJO frequency over a location expressed through a 144 certain EOF (Fig. 1e). The sum of cosine square values for all the EOFs is equal to 1. We also measured 145 the correlation coefficient between MJO frequency time series at eight phase locations and the EOF time 146 series to obtain the goodness of representation of MJO frequency data by the two leading EOFs. In Table 1 147 the correlation coefficient and cosine square values are represented.

170
To examine the mean state of the tropical ocean associated with the MJO frequency modes, we 171 obtained the correlated seasonal mean SST anomalies of EOF1 and EOF2 time series (Fig. 2e

283
In the case of canonical EP-type ENSO phases, the main descending branch of the Walker 284 circulation is situated over the maritime continent and western Pacific at the east of 120° E (Fig. 5j). The 285 mean moisture distribution (specific humidity) over the maritime continent and west Pacific shows the 286 negative anomalies at the east of 120° E following the zonal circulation pattern (Fig. 5l). Therefore, the 287 intraseasonal MSE tendency term is negative over the maritime continent and west Pacific, restricting the 288 MJO propagation over these regions (Fig. 5h). The intra-seasonal MSE tendency over the eastern Indian

310
propagations between the warm and cold ENSO phases.

312
The relationship between the MJO and ENSO has been studied rigorously in the past few decades which

353
Hence, we tried to understand the interannual variability of the MJO frequency in terms of the two 354 leading EOFs of MJO frequency through the biplot technique ( Fig. 1,3). Basically, this representation is

368
We conducted the composite analyses of different atmospheric variables during CP and EP-type 369 ENSO phases to identify the reason behind MJO frequency spatial patterns. We find that the mean Walker

398
The interannual variability of the MJO is investigated in terms of the frequency of occurrence of the MJO 399 phases in boreal winter by numerous studies 18,19,26 . We have adopted the same definition for the frequency

410
MJO frequency possesses seasonal characteristics which we discussed in the introduction section. We 411 removed the seasonality from the MJO frequency data by standardizing each specific season's data by that 412 particular season's climatology (e.g. MJO frequency in SON is standardized by SON climatology of MJO 413 frequency) and thus we obtained normalized MJO frequency anomaly data at each phase location (Fig. 1a).

414
This normalized MJO frequency anomaly is independent to seasonal characteristics of MJO which are 415 evident through more MJO frequencies in boreal winter and spring than in boreal summer and autumn. The

416
MJO frequency anomaly, therefore, represents the interannual variability of MJO frequency excluding the 417 seasonal cycle of the MJO.

418
The derived MJO frequency anomaly data is a multivariate dataset which has eight variables 419 representing eight RMM phase regions (m=8) and 156 cases (n=156) representing 156 seasons (Fig. 1a).

510
We used daily NOAA interpolated OLR data with 2.5 ∘ x 2.5 ∘ resolution in our present study 37 . We 511 obtained the 2.5 ∘ x 2.5 ∘ pentad CPC Merged Analysis of Precipitation (CMAP) data and interpolated the 512 data to daily 38 . We used the zonal wind (u), meridional wind (v), omega (ω ), specific humidity (q), air 513 temperature (T) and geopotential height (φ ) from NCEP/NCAR reanalysis 1 dataset 39 . We also used monthly