The Madden-Julian Oscillation (MJO) is one of the most important meteorological phenomena in the tropics. With a timescale of 30–90 days it bridges the gap between weather and climate (Zhang, 2013), potentially providing predictability over several weeks. It consists of large-scale envelopes of, alternately, organized convection and clear skies, propagating slowly (∼5 m s-1) eastwards from the Indian Ocean, through the Maritime Continent, to the Pacific; and an associated planetary-scale circulation. The MJO interacts with several other phenomena, including the El Niño–Southern Oscillation (Tang and Yu, 2008), monsoons (Lavender and Matthews, 2009; Singh et al., 2017), tropical cyclones (Klotzbach, 2014) and the diurnal cycle (Peatman et al., 2014).
In 1971, Roland Madden and Paul Julian published a study of 10 years of radiosonde data from Canton Island, Kiribati (Madden and Julian, 1971), noting a “very pronounced maximum” in the co-spectrum of zonal wind at 850 and 150 hPa, with period 41–53 days (Figure 1). They hypothesized the cause to be a “large circulation cell oriented in zonal planes and centred in the mid-Pacific”, thousands of kilometres in scale and located near the equator. The following year a further study (Madden and Julian, 1972) analysed data from 25 stations throughout the tropics (and beyond) and determined that the oscillation was indeed a zonally-oriented planetary-scale circulation, confined to 10°N–10°S, which propagates from the Indian Ocean to the Pacific (Figure 2).
Figure 1: Co-spectra of u850 and u150 (dashed, left axis); and surface pressure and u850 (solid, right axis) from radiosonde data from Canton Island, Kiribati. From Madden and Julian (1971).
Figure 2: Schematic diagram of the MJO circulation and convection, with eastward propagation indicated between successive panels. Tropopause height and pressure anomaly are sketched at the top and bottom respectively of each panel. From Madden and Julian (1972).
However, it has recently emerged that although Madden and Julian were unaware of this oscillation before their 1971 paper, they were not the first to have demonstrated its existence. As described by Li et al. (in press), a study published in Chinese eight years earlier (Xie et al., 1963) found a 40–50 day oscillation in radiosonde data from several tropical locations. The Xie study, devoted to typhoon genesis, noted “[t]here is a quite definite relationship between the time, location and frequency of typhoon genesis and the location and strength of the basic flow in the low latitudes”, thus not only discovering the intraseasonal oscillation but also its relation to tropical cyclogenesis in one paper! They further suggested the “oscillation might be helpful for the extended-range forecast of initiation and development of typhoons”, a fact which is still to be fully exploited by forecasters today.
Xie et al. (1963) plotted time series of u700 from three weather stations, shown from east to west down the page in Figure 3: Thiruvananthapuram (Kerala, India), Ho Chi Minh City (Vietnam) and Zamboanga (Philippines). These were hand-drawn and had no temporal filtering applied to the data. They noted “[t]here is a consistent phase change of the zonal wind from Station 43371 [Thiruvananthapuram] to Station 98836 [Zamboanga]. When the westerlies intensified in India, they also intensified in south-east Asia, with a slight temporal delay… The change of zonal wind with time at these stations exhibited a wave-like oscillatory characteristic, with an average oscillatory period of around one-and-a-half months.” We can now recognize this as the MJO circulation, with intraseasonal timescale and eastward propagation.
Figure 3: Time series of u700 against month from three stations (rows; longitudes overlaid in blue — see main text for details) during 1958–1960 (columns). Black dots denote typhoons. From Xie et al. (1963). Red arrows added by Li et al. (in press) to highlight intraseasonal periods.
The Real-Time Multivariate MJO (RMM) indices (Wheeler and Hendon, 2004), computed from OLR and zonal wind, are widely used to plot MJO propagation but date to 1974 only. However, Oliver and Thompson (2012) reconstructed the indices back to 1905 by regressing against surface pressure data. Using the reconstructed indices we can plot the propagation events documented by Xie et al. (1963; Figure 4). There is reasonable agreement – e.g. anomalous low-level westerlies in July 1958 (Figure 3), as seen in Figure 2A, roughly correspond to phases 5–6 (Figure 4b); while anomalous low-level easterlies in late July–August, as seen in Figures 2C–E, roughly correspond to phases 8–2.
The Xie paper is not widely known, mainly because it was published in Chinese. Who knows how many other important papers may exist in non-English journals, unknown to the wider academic world? The analysis of the intraseasonal oscillation by Xie et al (1963) was less detailed than that by Madden and Julian (1971, 1972), and it was unarguably the latter brace of papers which brought the oscillation to the attention of the tropical community. However, if it should come to be known as the XMJO, in recognition of Xie et al.’s pioneering study of 1963, I for one would not complain.
Figure 4: (a) MJO diagram showing the location of active convection in each phase. (b-d, below) MJO events for the times shown in Figure 3, using the reconstructed RMM indices of Oliver and Thompson (2012).