The Pioneering Works of Professor Duzheng YE on Atmospheric Dispersion, Tibetan Plateau Meteorology, and Air-Sea Interaction

Throughout his distinguished career, Professor Duzheng YE has made many fundamental contributions to the atmospheric sciences. Particularly noteworthy are his works on atmospheric energy dispersion, the role of the Tibetan Plateau (TP) in the evolution of the monsoonal circulation over East Asia, and changes in the atmosphere-ocean system associated with El Niño episodes. His investigations on these topics have exerted a strong influence on the research paths of many individuals, including the present author. On this occasion of the 100th anniversary of Professor YE's birth, this paper provides a brief review of the linkages between his original ideas and subsequent investigations into various aspects of atmospheric circulation. The studies drawn upon to exemplify the foundations laid by Professor YE were mostly conducted by the present author or his associates over the past several decades. However, reference will also be made to other works of particular relevance to the topics at hand. Results based on observational data as well as outputs from general circulation model (GCM) experiments will be considered.

In his seminal work on atmospheric dispersion (Yeh, 1949), Professor YE laid a firm theoretical grounding for energy transmission by group velocity, which is distinctly faster than the phase propagation speed of individual waves embedded in the wave group (or "wave packet"). Evidence of such dispersive behavior has been reported by various investigators through diagnosing the spatiotemporal evolution of atmospheric fluctuations associated with cold surges and prominent teleconnection patterns.

The observed atmospheric changes accompanying strong cold surges over Korea were examined by (Joung and Hitchman, 1982). The development of geopotential height perturbations at various pressure levels, as obtained by compositing over 16 cases of severe polar outbreaks, is illustrated in the time-longitude sections in Fig. 1. Day 8 on the time-axis of these diagrams corresponds to the date of occurrence of the outbreaks in Korea. Strong positive and negative height anomalies indicate the presence of prominent ridges and troughs, respectively. The evolution of the individual ridges and troughs is highlighted by the thick dashed lines, which represent the axes of height anomalies at 300 hPa.

Inspection of Fig. 1 reveals that the cold surges over East Asia (130^°-140^°E) on Day 8 are coincident with an amplifying trough at 300 hPa(labeled as T3). This trough is situated between strong ridges to the west (labeled as R2) and to the east (R3). The individual trough and ridges (T3, R2, R3) travel eastwards, with a phase speed of about 10 m s^-1 (as may be deduced from the slope of the pertinent dashed lines). Comparison between the individual panels of Figs. 1a-c indicates that the height changes in the East Asian sector exhibit a distinctly baroclinic character, with the peak anomalies at 850 hPa being displaced to the east of the corresponding features at 300 hPa.

It is further noted that the R2-T3-R3 complex forms a "wave packet", and that it is preceded 6-7 days earlier by an analogous packet (T1-R1 complex) originating from the North Atlantic sector west of the Greenwich Meridian. Additional diagnoses by (Joung and Hitchman, 1982) supported the interpretation that the T1-R1 packet transmits energy eastwards, thus leading to successive downstream development of troughs and ridges across the Eurasian land mass. The group speed of this energy transfer, as estimated from division of the distance between the T1-R1 and R2-T3-R3 packets by the elapsed time between the appearance of these packets, is about 30 m s^-1, which is much faster than the phase speed of individual troughs and ridges quoted in the previous paragraph.

Another demonstration of successive downstream development of atmospheric fluctuations in association with cold air outbreaks was given by (Lau and Lau, 1984). By retaining fluctuations with time scales longer than five days, and compositing such time-filtered observations for 11 outbreak cases over central and southern China in the winter of 1978-79, the authors examined the evolution of slowly-varying geopotential height perturbations near the day of occurrence of the cold surge episodes (defined as Day 0). Their results, shown in Fig. 2, indicated the presence of a prominent trough in the vicinity of the East Asian seaboard. A feature of specific interest in the present context is the rather abrupt appearance on Day 3 of a wave train over the central subtropical Pacific. The latter signal is again suggestive of the effect of energy transfer from the wave packet over the cold surge region.

The evidence put forth in the two case studies mentioned above is illustrative of dispersive behavior in atmospheric circulation. The prominence of this characteristic in cold surge events indicates that the concept of atmospheric energy dispersion may find practical applications in extended-range forecasting, since notable circulation changes may be linked to development of wave packets situated upstream several days in advance.

Figure 1.  Time-longitude distributions of geopotential height anomalies at (a) 300, (b) 500 and (c) 850 hPa, as obtained by compositing over 16 cold surge events in Korea occurring on Day 8. Contour intervals for the 300, 500 and 850 hPa patterns are 60, 60 and 30 m, respectively. Solid and dotted contours indicate positive and negative values, respectively. The zero contour is omitted. Thick dashed lines correspond to principal trough and ridge axes at 300 hPa, and are labeled as T1, T2, T3, R1, R2 and R3. [Reprinted from (Joung and Hitchman, 1982). @ American Meteorological Society. Used with permission.]

Figure 2.  Spatial distributions of low-pass filtered geopotential height anomalies at 500 hPa for the period from Day (-1) to Day (+3), as obtained by compositing over 11 cold surge events in East and South China occurring on Day (0). Contour interval: 10 m. Shading indicates negative values. The zero contour is omitted. [Reprinted from (Lau and Lau, 1984). @ American Meteorological Society. Used with permission.]

Beyond cold surge events, the general spatiotemporal characteristics of the development of geopotential height fluctuations associated with various well-documented atmospheric teleconnection patterns (e.g., Wallace and Gutzler, 1981) were systematically studied by (Blackmon et al., 1984). Using low-pass filtered observational analysis products for the wintertime geopotential height field at 500 hPa, they computed the correlation coefficients between fluctuations at selected reference points (RPs) with time scales of 10-30 days and those at all other grid points in the data domain. The RPs correspond to the nodes and antinodes (or "centers of action") of various teleconnection patterns. The development of the teleconnection patterns was then inferred from the spatial distributions of these correlation statistics (hereafter referred to as "one-point teleconnection maps") at various time lags between the RP and other grid points.

Figure 3.  Spatial distributions of temporal correlation coefficients between low-pass filtered 500 hPa height fluctuations at various grid points and the corresponding fluctuations at the reference point (RP) located at (a, b) (55^°N, 75^°E), (c, d) (55^°N, 20^°W), and (e, f) (45^°N, 165^°W). Correlation patterns for fluctuations at various grid points leading (lagging) those at the RP by three days are displayed in the left-hand (right-hand) panels. [Reprinted from (Blackmon et al., 1984). @ American Meteorological Society. Used with permission.]

Figure 4.  Spatial distributions of the phase difference and coherence between height fluctuations at various grid points and the corresponding fluctuations at the reference point (RP), for the spectral band with period centered at 20 days. The left-hand (right-hand) panels are obtained using 500 (515) hPa data for the observational (GCM) data. The RP (see solid dots) is located at or near (a, b) (60^°N, 60^°E), (c, d) (55^°N, 30^°W) and (e, f) (50^°N, 170^°W). The phase difference is depicted by the orientation of the vectors, which point due north for zero phase difference, and rotate clockwise (counterclockwise) by one degree per degree of phase lag (lead). Phase differences are shown only at those grid points that are coherent with the RP at or above the 80% significance level. Shading indicates squared coherences above the 95% significance level. Various coherent regions are labeled using letters. [Reprinted from (Lau and Nath, 1987). @ American Meteorological Society. Used with permission.]

Figure 3 separately shows the one-point teleconnection maps for RPs located at (55^°N, 75^°E), (55^°N, 20^°W) and (45^°N, 165^°W). These RPs correspond to the centers of action for the Eurasian (EU), East Atlantic (EA) and Pacific-North American (PNA) teleconnection patterns, respectively. Charts in the left (right) column are computed with the variations at other grid points leading (lagging) those at the RP by three days. The charts in Fig. 3 illustrate that, for all three teleconnection patterns considered here, the centers situated upstream (typically west or north) of the RP exhibit stronger correlations when the local fluctuations lead those at the RP (see left-hand panels). Conversely, the centers positioned downstream (east or south) of the RP are more correlated with the RP when the local fluctuations lag those at the RP (right-hand panels) by several days. Hence, the results in Fig. 3 indicate that atmospheric variations with 10-30-day periods are characterized by mobile wave trains travelling from midlatitudes to the tropics through successive downstream development. (Blackmon et al., 1984) interpreted this mode of temporal behavior in terms of the two-dimensional dispersion of Rossby waves.

The spatiotemporal evolution of the prevalent teleconnection patterns was examined in further detail by (Lau and Nath, 1987). By applying cross-spectral analyses, they computed the phase differences between mid-tropospheric geopotential height variations at selected RPs and those at other grid points in the extratropical Northern Hemisphere. Similar to the study of (Blackmon et al., 1984), the RPs were chosen to correspond with the centers of action of the teleconnection patterns in the EU, EA and PNA sectors. In addition to observational analysis products, the computations were also performed on the 12-year output from a low-resolution GCM developed at the Geophysical Fluid Dynamics Laboratory (GFDL).

Figure 4 displays the spatial distributions of phase differences between the wintertime height fluctuations at selected RPs and those at the other grid points, for the spectral band centered at a period of 20 days, with bandwidth of 16-27 days. Results based on observational and model-simulated data are presented in the left- and right-hand panels, respectively. The RPs chosen for the observational data are located at (60^°N, 60^°E) (Fig. 4a), (55^°N, 30^°W) (Fig. 4c), and (50^°N, 170^°W) (Fig. 4e). These sites are situated in close proximity to those used for the correlation analysis in Fig. 3, and for diagnosing the model output (Figs. 4b, d and f).

In Fig. 4, the phase difference between the height variations at a given location and those at the RP (denoted by a solid dot) is indicated by the orientation of the arrow plotted at that location. An arrow pointing due north indicates zero phase difference. The arrow rotates clockwise (counterclockwise) by one degree for each degree of phase lag (lead) at the grid point in question relative to the RP. Phase differences are shown only for coherences exceeding the 80% significance level. Coherences surpassing the 95% significance level are highlighted by stippling.

The phase relationships shown in Fig. 4 indicate that, for periods centered at 20 days, the height fluctuations in the immediate vicinity of the RP (i.e., Regions L, F and B for the EU, EA and PNA patterns, respectively) are mostly in phase with those at the RP itself. However, the height variations with polarity opposite to that at the RP are seen to arrive at the region further downstream (i.e., M, E/G and C) with a delay of about 1/16-1/8 of a cycle (i.e., 1-2 days), as can be inferred from the southwestward or south-southwestward orientation of the phase arrows in the latter regions. For the EU pattern (Figs. 4a and b), there is also evidence that signals with polarity opposite to those in Region L appear in the upstream region (K) several days in advance. The successive development of alternating troughs and ridges from upstream to downstream of the RPs is again indicative of atmospheric dispersive behavior. The considerable resemblance between the left- and right-hand panels in Fig. 4 indicates that this phenomenon is captured well in the GCM simulation.

By constructing composites of one-point teleconnection statistics based on observations and the output from a hierarchy of numerical models, (Lee and Held, 1993) documented the evolution of coherent baroclinic wave packets in the time-longitude domain. For all of the datasets examined in that study, the group velocity of the wave packets was noticeably higher than the phase speed of the individual troughs and ridges, thus further confirming the prevalence of energy dispersion and downstream development of midlatitude disturbances in both observed and simulated atmospheres. The authors noted that this behavior is particularly evident in the Southern Hemisphere during the summer season.

In addition to the diagnostic studies cited above, much progress has been made in the theoretical aspects of energy dispersion in a spherical atmosphere (e.g., Hoskins et al., 1977; Hoskins and Karoly, 1981). Many atmospheric response patterns to sources and sinks of heat or vorticity, such as those induced by perturbed sea surface temperature (SST) conditions at the lower boundary, have been interpreted in terms of Rossby wave dispersion. The concept of energy dispersion has therefore played a central role throughout recent decades in advancing our understanding of the dynamical origin of atmospheric fluctuations on different time scales. There remains fertile ground for further research on the dynamical interactions between the dispersive wave patterns and other modes of atmospheric variability, such as those associated with baroclinically unstable waves.

Professor YE devoted considerable attention to the study of energy sources and sinks over the TP, and their influences on the evolution of various circulation features related to the Asian summer monsoon (e.g., Yeh and Gao, 1979; Ye, 1981). In particular, his voluminous works on the atmospheric energy balance over different parts of the TP motivated the present author to conduct a similar analysis using the output from a 20-year simulation based on a GCM with a spatial resolution of about 50 km. In our model study (Lau and Ploshay, 2009), different contributions to the atmospheric energy budget were evaluated separately for various subregions where the South or East Asian monsoons prevail. These subdivisions include the western and eastern portions, as well as the southern edge of the TP (Region 1, 2 and 3, respectively), the northern part of the Bay of Bengal and the surrounding land areas (Region 4), the Indochina Peninsula (Region 5), and southeastern China (Region 6). For every consecutive 10-day period from April to September, climatological averages of the condensational heating, surface sensible heating and radiative cooling terms were computed for each of the subregions. The results are summarized in Fig. 5.

The data presented in Fig. 5 clearly illustrate the dominance of condensational heating in the energy budget for Regions 3, 4 and 5 after the monsoon onset over these sites. Surface sensible heating plays a considerable role in the heat balance over Regions 3, 4 and 5 in the April-May period, prior to the arrival of the summer monsoon rains. This sensible heating term continues to make substantial contributions to the heat balance over the TP (Regions 1 and 2) over much of the warm season. In Region 6, latent heat release remains the primary energy input throughout the April-September period. Radiative processes lead to a net cooling over all six regions during the warm season.

A detailed comparison was made by (Lau and Ploshay, 2009) between the model estimates of various energy input terms, as reported in Fig. 5, with the observed counterparts presented by Professor YE and collaborators, and by other investigators based on various modern datasets. The observational and model-simulated values are in broad agreement. The most noteworthy discrepancy is the overestimated latent heat release in Region 4, most likely due to excessive Bay of Bengal rainfall in the model.

The rapid establishment of the energy source in the May-June period, especially over Regions 4 and 5 (see Fig. 5) bears a strong relationship with the transition of the large-scale circulation over South and East Asia. An outstanding example is the development of the 200 mb height field. Figure 6 shows the sequence of climatological charts of this field at 10-day intervals from early May through to the end of July. These patterns are based on the simulation from the same high-resolution (50-km) GCM that yielded the data for Fig. 5.

The map sequence in Fig. 6 starts with a relatively weak ridge axis over the Indochina-South China Sea sector in early May (Fig. 6a). This feature strengthens noticeably and moves northwestwards in mid- and late-May (Figs. 6b and c). It continues on its northwestward path and arrives over the TP in mid-June (Figs. 6e). This anticyclone then remains in an almost stationary state thereafter, but undergoes further intensification and attains maximum strength in July. In concert with the northward movement of the high-pressure center, the westerly belt situated to the north of this center is also seen to leap polewards by about 10^° of latitude within a time span of less than one month.

Another indicator of abrupt circulation changes accompanying the spatial reconfiguration of energy sources is the evolution of intense precipitation systems in the China-Japan-Korea sector in late spring and early summer. These heavy rain bands, popularly referred to as mei-yu (or "plum rain"), account for a substantial fraction of the total annual rainfall in those regions. The structural and propagational properties of this phenomenon have long attracted the attention of many meteorologists in that area. Of particular interest is the tendency for the precipitation zones to advance polewards in a series of jumps, and the presence of both tropical and extratropical characteristics in the circulation features associated with these rainbands.

Figure 5.  Temporal variations of principal terms of the atmospheric energy budget over six zones of the Asian summer monsoon region, as obtained based on the output from a high-resolution GCM developed at GFDL. Climatological 10-day means of latent heat release, surface sensible heating and total radiative heating [see color code in panel (a)] are displayed in a cumulative format. The sum of these three terms is shown using orange dots connected by orange line segments. The domains of the six zones used for computing the areal means of the heating terms are illustrated in the inset map between panels (a) and (b). Units: W m^-2. [Reprinted from (Lau and Ploshay, 2009). @ American Meteorological Society. Used with permission.]

Figure 6.  Spatial distribution of consecutive climatological 10-day means of 200 hPa geopotential height in the May-July period, as obtained from the output of a high-resolution GCM developed at GFDL. Units of scale bar at bottom: m.

The spatiotemporal behavior of the plum-rain system was revisited by (Lau and Ploshay, 2009) using modern observations collected by the Tropical Rainfall Measurement Mission (TRMM), which makes use of radar equipment mounted on satellites. These results are used to assess the fidelity of the simulation by the high-resolution GCM examined earlier in this section. The time-latitude distributions of the rainfall climatology averaged over the 110^°-130^°E longitudinal span in successive five-day intervals are separately displayed in Fig. 7 for the GCM simulation and TRMM observations. The corresponding distribution of the climatological horizontal wind at 850 hPa is shown using arrows.

The most notable feature in the results based on observational data (Fig. 7b) is the precipitation maximum originating from ∼20^°N in early June, and migrating polewards to ∼35^°N in late June. The timing of this northward advance matches with that of the movement of the 200 hPa anticyclone over South China (see Fig. 6). The rainfall signal is also coincident with the northward shift of the belt of strong southwesterly flow (see wind vectors in Fig. 7b). As illustrated in more detail by (Lau and Ploshay, 2009), this displacement of the southwesterly wind belt is associated with the northward advance and westward extension of the axis of the high pressure ridge over the subtropical western Pacific.

To a certain extent, the northward migration of the precipitation center over East Asia in June, and its association with strengthened southwesterly flow, are reproduced in the model atmosphere (Fig. 7b). However, the simulated rainfall amounts during this plum-rain regime are noticeably lower than the observed values. The model pattern indicates that the precipitation centers advance to as far north as 40^°N. It also suggests that this northward movement is achieved by several distinct shifts, with the jump from 30^°N on 10 June to 40^°N 15 days later being the most remarkable. These abrupt leaps are much less evident in the observations.

Figure 7.  Latitude-time distributions of climatological five-day means of precipitation (shading; see scale bar at bottom) and 850 hPa horizontal wind (vectors; see scale in upper right of each panel), as obtained by averaging the data within the 110^°-130^°E zone. Patterns are shown for (a) the output from a high-resolution GCM at GFDL and (b) observational reanalysis data. [Reprinted from: (Lau and Ploshay, 2009). @ American Meteorological Society. Used with permission.]

In recent decades, much effort has been made in utilizing measurements from a myriad of remote sensing platforms to produce more accurate estimates of the spatiotemporal characteristics of various heat sources and sinks over the TP. These efforts have been augmented by ground-based observations as well as data assimilation techniques based on high-resolution climate models. Such improved databases facilitate our continuing efforts to assess the effects of the TP on the atmospheric circulation over East Asia. Recent advances in our understanding of the impacts of the TP on the seasonal evolution of the Asian monsoon, as well as updated estimates of heat sources associated with the TP, were reviewed by (Yanai and Wu, 2006) and (Wu and Liu, 2016). The realistic simulation of the monsoonal features over East and South Asia remains a stiff challenge to the modeling community, particularly with regards to the intensity, evolution and spatial distribution of the principal precipitation systems. Further advances in such efforts will entail the incorporation of a better treatment of precipitation processes, as well as the complex terrain over the TP, and perhaps still higher spatial resolutions in model atmospheres.

Professor YE maintained a strong interest in the effects of El Niño events on various atmospheric circulation systems, especially those with strong impacts on weather and climate over East Asia. Traditionally, most attention has been paid to El Niño-related atmospheric features in the Southern Hemisphere, as exemplified in the popular usage of the term "Southern Oscillation", which highlights the out-of-phase sea level pressure (SLP) variations at Darwin and Tahiti, with both sites being located south of the Equator. Professor YE noted to the present author that analogous signals in the Northern Hemisphere should not be overlooked. Specifically, he mentioned the equally strong anticorrelation between SLP variations at Manila and Honolulu, and suggested more widespread consideration of the concept of "Northern Oscillation" in El Niño research.

In order to study the interactions between El Niño episodes in the tropical eastern and central Pacific and the global air-sea coupled system, a suite of specially designed GCM experiments have been performed at the GFDL during recent decades. A summary of this long-term project and some of the essential results is provided in (Lau, 2016). Selected findings with relevance to air-sea interactions in the East Asian sector are described in this section.

In the present discussion, attention is focused on a suite of model integrations in which the SST conditions in the eastern and central tropical Pacific (TPAC) were constrained to follow the month-to-month variations as observed through the experiment, which typically spanned over a period of several decades. Outside the TPAC region, the model-simulated energy fluxes across the air-sea interface were used to drive static, one-dimensional mixed-layer models (MLMs) inserted at individual oceanic grid points, so that the SST variations in the world's oceans accompanying prescribed anomalous episodes in the TPAC could be computed. This setup (henceforth referred to as the MLM experiment) allowed for direct forcing of the global atmospheric circulation by the oceanic forcing in the TPAC, as well as simplified two-way interactions between the ensuing atmospheric anomalies outside of the TPAC and the underlying oceanic sites. The model runs were conducted using an ensemble approach, by using independent initial atmospheric conditions to obtain multiple realizations of the model responses to the same forcing scenario.

Figure 8 shows the distributions of the composites of anomalous SST and surface wind (Figs. 8a and b), and precipitation and SLP (Figs. 8c and d), as computed using observational reanalysis data (left-hand panels) and the output of the MLM experiment (right-hand panels). The patterns were computed by subtracting the data for outstanding cold La Niña events in the 1950-99 period from the outstanding warm El Niño events in the same period, and were based on the northern winter season between the year of initiation of these events [hereafter referred to as Year (0)] and the succeeding year [Year (+1)]. The left- and right-hand panels of Fig. 8 are in general agreement, thus confirming that the MLM experiment is able to capture many of the El Niño-related variations in the western Pacific-Australian sector.

Figure 8.  Spatial distributions of (a, b) sea surface temperature (shading; see scale bar at bottom of these panels) and near-surface horizontal wind (vectors; see scale in the upper right of the figure), and (c, d) precipitation (shading; see scale bar at bottom of these panels) and sea level pressure (contours; interval: 0.5 hPa), as obtained by subtracting the composite over 10 La Niña events from the corresponding composite over 10 El Niño events. This composite procedure is applied to the data for the northern winter season between Year (0) and Year (+1) (see definition in the text). Results based on observational reanalysis data and the MLM simulation are displayed in the left- and right-hand panels, respectively. [Reprinted from (Lau and Wang, 2006). With Permission of Springer.]

Figure 9.  Time-longitude distributions of sea surface temperature (shading; see scale bar at the bottom of the figure) and sea level pressure (contours; interval: 0.2 hPa), for averages of monthly mean data within the 10^°-20^°N zone. The time axis extends upwards from June of the year of initiation of El Niño/La Niña events [Year (0)] to August of the following year [Year (+1)]. Results are obtained by subtracting composites over 10 La Niña events from composites over 10 El Niño events, for (a) observational reanalysis data and (b) the MLM simulation. [Reprinted from: (Lau and Wang, 2006). With Permission of Springer.]

The patterns in the lower panels of Fig. 8 are dominated by positive SLP anomalies and generally dry conditions over the western Pacific-Australian sector. Particularly noteworthy are the SLP and precipitation extrema over the Philippine Sea and the vicinity of northern Australia. These anomaly centers straddle the Equator, and may be interpreted as the Rossby-wave response to suppressed convection and condensational heating in the equatorial western Pacific during typical El Niño events. It is also evident that the magnitude and spatial extent of the anomalies over the Philippines region exceed those of the features in the vicinity of northern Australia. The prominence of the anomalies in the Philippine Sea and South China Sea supports Professor YE's call for giving due attention to El Niño signals in the Northern Hemisphere.

The upper panels of Fig. 8 illustrate that the positive SLP changes over the Philippines region are accompanied by an anticyclonic surface circulation pattern. The southerly or southwesterly anomalies over southern China are directed against the climatological northerly or northeasterly winter monsoon flow in that region. The implied weakening of the cold and dry winter monsoon over the South China Sea during El Niño is consistent with the wet condition over southern China (see shading in Figs. 8c and d), and warm SST anomalies over the South China Sea (see shading in Figs. 8a and b, and refer to the following discussion of Fig. 9). Conversely, over the western Pacific zone lying to the east of the positive SLP anomaly, intensification of the northeasterly winter monsoon flow is coincident with below-normal SST conditions.

The nature of the air-sea interaction over the South China Sea and subtropical western North Pacific during El Niño and La Niña episodes is further explored in Fig. 9, which shows the time-longitude distributions of anomalies of the SST and SLP fields, as obtained using the same composite procedure for Fig. 8. The data displayed here are based on averages over the 10^°-20^°N zonal belt. Results are presented for the observations (Fig. 9a) and output from the MLM experiment (Fig. 9b).

Both the observed and simulated contour patterns in Fig. 9 indicate that the anomalous high pressure ridge over the Philippine Sea/western Pacific typically emerges in September-October of Year (0). It gains strength and extends eastwards in the ensuing months, attaining peak amplitude in January of Year (+1). To the west of this positive SLP feature, the weakening of the climatological winter monsoon flow due to the anomalous southerly or southwesterly winds (see previous discussion of Figs. 8a and b) is expected to reduce the heat loss from the underlying ocean, thereby leading to positive SST anomalies. Analogously, strengthening of the winter monsoon by the northerly or northeasterly wind changes to the east of the SLP anomaly would enhance oceanic heat loss and lead to negative SST perturbations. The above inferences have previously been made by (Wang et al., 2000), and are supported by the results presented in Fig. 9. In particular, the model pattern clearly illustrates that the SLP anomaly is straddled by positive SST changes to the west, and negative SST changes to the east. Both the SLP and SST features in the model pattern exhibit a stronger tendency to migrate eastwards with time, as compared to the observations.

Additional model experiments have been conducted with the goal of assessing the relative impacts on East Asian monsoon variability by remote SST forcing from the TPAC, and by more local SST fluctuations in the western Pacific/South China Sea sector (e.g., Lau and Nath, 2009). Such results suggest that the interplay between the atmospheric responses to these remote and local oceanic anomalies is dependent on the detailed evolution of El Niño and La Niña events through the summer season. The modulation of El Niño/La Niña behavior by global climate change in the coming decades, and the implications for the variability of East Asian climate, are prominent issues that warrant further research efforts.

Professor YE's endeavors in atmospheric research across a broad spectrum of topics have left an indelible imprint in the works of numerous investigators of the succeeding generation, and will undoubtedly continue to do so for the generations to come. Among the many contributions that he has made during the past several decades, his insights on dispersion, monsoon meteorology related to thermal forcing of the TP, and large-scale atmosphere-ocean coupling associated with El Niño, are highlighted in this article. The effects of these dynamical and physical processes are illustrated using examples of pertinent phenomena in both observed and GCM-simulated atmospheres. It should be noted that, in addition to the research topics mentioned above, Professor YE also engaged in pursuits in many other disciplines, including the impacts of ground hydrological processes on the climate system, global climate change and sustainable development, and interactions between orderly human activities and the environment.

The ample use of model results in this review (see right-hand panels of Figs. 4, 8 and 9, as well as Figs. 5, 6 and 7a) demonstrates the ascending role of GCMs as a research tool for diagnosing atmospheric circulation systems. Model-based results offer independent supporting evidence for the occurrence of specific modes of atmospheric variability, such as that caused by energy dispersion (see Fig. 4). Comparison between findings derived from model simulations and from observations (see Figs. 4 and 7-9) serves to assess the capability of GCMs to reproduce specific phenomena. Moreover, the model runs provide estimates of certain quantities (e.g., various heating terms in the Asian monsoon region; see Fig. 5) that are difficult to obtain from observational measurements. With the appropriate experimental design, the GCMs can be used effectively to explore the role of specific mechanisms in generating various phenomena of interest. An obvious example is the application of the MLM experiment to the identification of the processes linking prescribed SST forcing in the TPAC to air-sea coupling in the Philippine Sea and South China Sea region (Figs. 8 and 9).

The present author has had the good fortune of knowing Professor YE since 1981, when he came to GFDL in Princeton for a seven-month stay. We have maintained contact in the ensuing years through correspondence and my visits to Beijing. I have looked upon him as a scientific mentor who is extremely generous in sharing his broad and deep insights in various academic issues. Equally important, I have also regarded him as a personal role model, and hold the highest admiration for his kindness, humility and genuine concern for the common good. Through his personal example and rich experiences, he has taught me many valuable lessons in life, for which I shall remain eternally grateful.