The convectively active part of the Madden-Julian Oscillation (MJO) propagates eastward through the warm pool, from the Indian Ocean through the Maritime Continent (the Indonesian archipelago) to the western Pacific. The Maritime Continent's complex topography means the exact nature of the MJO propagation through this region is unclear. Model simulations of the MJO are often poor over the region, leading to local errors in latent heat release and global errors in medium-range weather prediction and climate simulation.Using 14 northern winters of TRMM satellite data it is shown that, where the mean diurnal cycle of precipitation is strong, 80% of the MJO precipitation signal in the Maritime Continent is accounted for by changes in the amplitude of the diurnal cycle. Additionally, the relationship between outgoing long-wave radiation (OLR) and precipitation is weakened here, such that OLR is no longer a reliable proxy for precipitation. The canonical view of the MJO as the smooth eastward propagation of a large-scale precipitation envelope also breaks down over the islands of the Maritime Continent. Instead, a vanguard of precipitation (anomalies of 2.5 mm day −1 over 10 6 km 2 ) jumps ahead of the main body by approximately 6 days or 2000 km. Hence, there can be enhanced precipitation over Sumatra, Borneo or New Guinea when the large-scale MJO envelope over the surrounding ocean is one of suppressed precipitation.This behaviour can be accommodated into existing MJO theories. Frictional and topographic moisture convergence and relatively clear skies ahead of the main convective envelope combine with the low thermal inertia of the islands, to allow a rapid response in the diurnal cycle which rectifies onto the lower-frequency MJO. Hence, accurate representations of the diurnal cycle and its scale interaction appear to be necessary for models to simulate the MJO successfully.

Early in EarthWe thus conclude that a higher nitrogen level probably helped warm the early Earth, and suggest that the effects of N 2 should be considered in assessing the habitable zone for terrestrial planets.

Conventional analyses of the MJO tend to produce a repeating cycle, such that any particular feature cannot be unambiguously attributed to the current or previous event. We take advantage of the sporadic nature of the MJO and classify each observed Madden-Julian (MJ) event as either primary, with no immediately preceding MJ event, or successive, which does immediately follow a preceding event. 40% of MJ events are primary events. Precursor features of the primary events can be unambiguously attributed to that event. A suppressed convective anomaly grows and decays in situ over the Indian Ocean, prior to the start of most primary MJ events. An associated mid-tropospheric temperature anomaly destabilises the atmosphere, leading to the generation of the active MJ event. Hence, primary MJ events appear to be thermodynamically triggered by a previous dry period, although stochastic forcing may also be important. Other theories predict that boundary-layer convergence, humidity, propagation of dynamical structures around the Equator, sea surface temperatures, and lateral forcing by extratropical transients may all be important in triggering an event. Although precursor signals from these mechanisms are diagnosed from reanalysis and satellite observational data in the successive MJ events, they are all absent in the primary MJ events. Hence, it appears that these apparent precursor signals are part of the MJO once it is established, but do not play a role in the spontaneous generation of the MJO. The most frequent starting location of the primary events is the Indian Ocean, but over half of them start elsewhere, from the maritime continent to the western Pacific.

The ability of 15 atmospheric GCM models (AGCM) to simulate the tropical intraseasonal oscillation has been studied as part of AMIP. Time series of the daily upper tropospheric velocity potential and zonal wind, averaged over the equatorial belt, were provided from each AGCM simulation. These data were analyzed using a variety of techniques such as time filtering and space-time spectral analysis to identify eastward and westward propagating waves. The results have been compared with an identical assessment of ECMWF analyses for the period 1982-1991. The models display a wide range of skill in simulating the intraseasonal oscillation. Most models show evidence of an eastward propagating anomaly in the velocity potential field, although in some models there is a greater tendency for a standing oscillation, and in one or two the field is rather chaotic with no preferred direction of propagation. Where a model has a clear eastward propagating signal, typical periodicities seem quite reasonable although there is a tendency for the models to simulate shorter periods than in the ECMWF analyses, where it is near 50 days. The results of the space-time spectral analysis have shown that no model has captured the dominance of the intraseasonal oscillation found in the analyses. Several models have peaks at intraseasonal time scales, but nearly all have relatively more power at higher frequencies

SUMMARYA life cycle of the Madden-Julian oscillation (MJO) was constructed, based on 21 years of outgoing long-wave radiation data. Regression maps of NCEP-NCAR reanalysis data for the northern winter show statistically significant upper-tropospheric equatorial wave patterns linked to the tropical convection anomalies, and extratropical wave patterns over the North Pacific, North America, the Atlantic, the Southern Ocean and South America. To assess the cause of the circulation anomalies, a global primitive-equation model was initialized with the observed three-dimensional (3D) winter climatological mean flow and forced with a time-dependent heat source derived from the observed MJO anomalies. A model MJO cycle was constructed from the global response to the heating, and both the tropical and extratropical circulation anomalies generally matched the observations well. The equatorial wave patterns are established in a few days, while it takes approximately two weeks for the extratropical patterns to appear. The model response is robust and insensitive to realistic changes in damping and basic state. The model tropical anomalies are consistent with a forced equatorial Rossby-Kelvin wave response to the tropical MJO heating, although it is shifted westward by approximately 20 • longitude relative to observations. This may be due to a lack of damping processes (cumulus friction) in the regions of convective heating. Once this shift is accounted for, the extratropical response is consistent with theories of Rossby wave forcing and dispersion on the climatological flow, and the pattern correlation between the observed and modelled extratropical flow is up to 0.85. The observed tropical and extratropical wave patterns account for a significant fraction of the intraseasonal circulation variance, and this reproducibility as a response to tropical MJO convection has implications for global medium-range weather prediction.