PRP-ENSO Webinar No 2. by Michael McPhaden (NOAA/PMEL)

追加: 2026-06-03

[00:00:00]I'm I'm Anna Depailler, I'm part of the Pacific Regional Panel of CLIVAR and we're really excited to be hosting this ENSO webinar series where we are now in the second installment of talks. The first one was given by Jerome Viard on um ENSO theory and today we have Mike McPhaden.

[00:00:20]Um Mike's a senior scientist at NOAA's Pacific Marine Environmental Laboratory or PMEL in Seattle who's been doing research on ocean observations and tropical ocean observations for 40 years and or more than 40 years now. We're really excited to hear him talk about um the TAO array, the Tropical Atmosphere Ocean array, and how it evolved in time and what his insights are and um learn more about the array that helped us learn so much about ENSO. Mike has a number of accolades. I think um I hope I hope Mike you um excuse me for not naming them all, but um we're we're really excited to hear you talk about this.

[00:01:04]Um so please unmute, share your share your screen, and take it away and then we will have uh time for questions for Mike in the end. Thank you everyone for joining us today.

[00:01:17]Um I was asked to talk about ENSO observations which is a bit of a open-ended and daunting subject.

[00:01:24]So um I'm going to give a review and there are two parts to this review.

[00:01:29]Uh the first one is uh a review of the scientific ideas, the technological developments, and the climate extremes that influenced the design and implementation of the system.

[00:01:39]And this will be about uh a third of the talk.

[00:01:42]Um and then I'm going to give a review of some of what we learned in terms of um the scientific discoveries and and advances that we've made using data from this observing system.

[00:01:53]I'm going to focus mainly but not exclusively on the TAO array because um um it's not the only important component of the observing system, but it is a key one and it's the one that I'm most familiar with, so I'm going to build my story around that.

[00:02:08]And then I'll end with a few general thoughts about observing system challenges for the future.

[00:02:14]So, by way of brief introduction, El Niño and the Southern Oscillation or ENSO, this is the strongest year-to-year fluctuation of the climate system on the planet. It originates in the tropical Pacific through interactions between the ocean and the atmosphere mediated by wind and sea surface temperature feedbacks.

[00:02:32]Uh we refer to the warm phase of ENSO as El Niño and the cold phase La Niña.

[00:02:37]Um it's a fascinating scientific problem that has occupied the community for 60 years or more, uh but what makes it a really compelling subject is the uh impacts on patterns of weather variability worldwide. And um it basically through far-field atmospheric teleconnections shifts probability for droughts, floods, heat waves, and other extreme events in in different parts of the globe, and these all have significant socioeconomic consequences, which motivates our need to better understand and predict uh ENSO.

[00:03:13]So, the the big bang for ENSO research came uh from the work of Jacob Bjerknes in the mid-1960s, uh when he opened up a whole new universe of study in the seasonal climate based on these four concepts.

[00:03:28]Uh one is that the ocean and the atmosphere are essential for generating El Niño. He was the first to link El Niño to Walker's Southern Oscillation.

[00:03:37]That it could affect uh patterns of weather variability at higher latitudes through atmospheric teleconnections.

[00:03:43]That it was basin-scale, uh not just confined to the coast of South America as long had been held.

[00:03:50]And then finally, because of the oceans' uh enormous thermal inertia, that it might be predictable on seasonal time scales.

[00:03:58]Now, Bjerknes got interested in the problem of El Niño around the time of the International Geophysical Year in 1957-58.

[00:04:08]And this was a coordinated effort to study the Earth and its environment by 70 different nations. And it's probably best known for the dawn of the space age with the launch of Sputnik 1. But all aspects of the Earth system, oceans, atmosphere, and land, and space were relatively well sampled.

[00:04:29]And serendipitously, it just happened to coincide with the major El Niño in 1957-58.

[00:04:35]And it was this event and the relatively a large amount of data available for it that sparked Bjerknes' interest in the problem and and um sparked his fundamental insights. And in his first paper, yeah, he he highlights data from this event as as kind of sparking his his insights.

[00:04:57]Now, even at this early stage, Bjerknes Bjerknes realized that we would need more data to understand this problem.

[00:05:05]And he even gave us some ideas about how to collect it. And this is from an address in late 1969. It's really very visionary. Uh you know, in the future, we can envision the creation of a worldwide service of monitoring buoys reporting by way of communication satellites data to give electronic computers the right input for long-range predictions. I mean, and this is at the very beginning of ENSO studies. So, really um uh he anticipated what we were going to be doing for the next 40 years.

[00:05:39]Uh the next milestone a few years later was the 1972-73 El Niño, very strong. Uh it caused a collapse of the Peruvian anchovy fishery, which at the time was this largest fishery in the world.

[00:05:53]Anchovy fish meal was used as a feed supplement for cattle and poultry, and so the collapse of this fishery led to widespread uh uh price spikes and meat shortages.

[00:06:05]And it was uh caused by a reduction in upwelling along the coast of Peru and Ecuador that not only warmed the surface waters, but reduced the supply of nutrients to the surface layer and caused the disruption of the entire ecosystem.

[00:06:21]Uh About this time Klaus Wyrtki got interested in the problem of El Ni- of El Nino. He wanted to know why the waters off the coast of Peru warmed during El Nino onset.

[00:06:33]And back then, you know, you couldn't go to the web and pull down your favorite data set. He basically had to make his own.

[00:06:40]So in the 1970s, he started the uh University of Hawaii Sea Level Center.

[00:06:45]Uh and he also collected wind measurements from uh the merchant vessels in the tropical Pacific.

[00:06:52]And he created databases of these variables, and it was really the first uh systematic attempt to study the basin-scale variations relevant to El Nino.

[00:07:02]And it was through the analysis of these data that he came up with his his uh great insight that at the onset of El Nino, it wasn't the local winds off the coast of Peru and Ecuador that mattered. It was the collapse of the trade winds thousands of kilometers to the west. And uh this collapse communicated was communicated the eastern Pacific he hypothesized by the excitation and propagation of an eastward uh a downwelling equatorial Kelvin wave.

[00:07:34]Uh The next important milestone was the uh extremely bitter winter in the US in 1976-77.

[00:07:44]And at least in the US, this focused attention on the ocean's role in North American climate. Uh it was so unusual that it made the pages of National Geographic. And there's a scene from Buffalo, New York here, where I was born and raised.

[00:07:59]Um and uh you know, there were competing ideas for what role the ocean played.

[00:08:05]Bjerknes' ideas had been out there for now for a decade, and 1976-77 was a weak El Niño.

[00:08:12]Uh but Jerome Namias uh had different ideas. He He thought it was mainly this uh sea surface temperatures in the North Pacific.

[00:08:21]And uh and there was in fact an a program started in the US called the North Pacific Experiment or NORPAX to test these competing ideas.

[00:08:30]Uh Jerome Namias was at Scripps Institution of Oceanography at the time, and that's where NORPAX was headquartered.

[00:08:36]I was also there as a graduate student.

[00:08:39]Uh and this is This is when I got interested in the problem of El Niño with all this exciting research about uh impacts on North America going on.

[00:08:49]Uh There are many studies from this era that moved the needle in the direction of supporting Bjerknes' hypothesis. I just want to mention two of them.

[00:08:57]One was by Russ Davis, um a professor at Scripps. And he analyzed historical ship data from the North Pacific, and he found that sea level pressure variations and associated wind variations led sea surface temperature on seasonal time scales. In other words, the North Pacific atmosphere was forcing the ocean rather than the other way around. And this was not consistent with uh Namias' hypothesis.

[00:09:23]Uh Russ, by the way, is the guy who 15 years later invented the uh profiling float that made Argo possible.

[00:09:32]Uh The other important study from this era was the identification of the Pacific North American pattern or PNA by Wallace and Gutzler, which said that um you could have these you see this uh sequence of high-low-high pressure centers emanating out of the tropics forced by variations in in tropical deep convection associated with warming and cooling uh in the tropical Pacific. And uh this PNA pattern said you could have pressure variations in the North Pacific that would force changes in sea surface temperature. You could have disturbed weather over North America, but the common cause of both of those was tropical Pacific forcing.

[00:10:11]So, this uh this set the stage for a major 10-year international program uh to study El Niño under the auspices of the newly created World Climate Research Program. And it was felt that um El Niño was a sufficiently mature subject at this time that it could be the uh featured as the first initiative of the WCRP.

[00:10:36]And in the planning process for TOGA, something truly dramatic happened, which was the 1982-83 El Niño.

[00:10:45]This was the strongest El Niño of the 20th century at the time. It was not predicted because we had no forecast models uh for El Niño. It was not even detected until nearly at its peak. And there were a few reasons for this. The one is The first is that um the Mexican volcano El Chichón erupted in early 1982, and it threw a cloud of uh sulfuric aerosols into the stratosphere that effectively blinded uh NOAA's newly launched polar orbiting weather satellites from seeing the sea surface accurately. It biased the satellite retrievals cold by several degrees centigrade.

[00:11:22]Uh this was compounded by the fact that there was no real-time data coming back from the ocean. There were data, but typically it would be months before you could get access to that and start working with it.

[00:11:33]So, we had no checks on what the satellite was measuring.

[00:11:37]And the situation was further muddled by the fact that the early signatures of El Niño warming that we expected did not appear.

[00:11:47]Rasmusson and Carpenter Carpenter in early 1982 published their composite El Niño based on the average of six El Niño events from the '50s to the '70s using volunteer observing ship data across the basin and they described it in terms of these phases with the first phase of warming occurring in March to May off the coast of Peru and Ecuador. And then this signal would and signal in the winds would later propagate westward. Well, that didn't happen.

[00:12:18]So, there was a lot of confusion about what was going on in the system. And as characterized as confusion is characterized by climate class Wickey made at a planning workshop at Princeton University in October of 1982 near the height of this event when he gave his assessment of some sea surface temperature measurements from the islands areas where he was maintaining tide gauges that the temperatures were unusually warm and this could be a big El Niño. And what he said was to call this El Niño would be a case of child abuse.

[00:12:56]Well, this was a clever play on words, um, but it was also highlighting a shocking community-wide failure.

[00:13:05]And they the logjam of information was finally broken when John Toole of Woods Hole uh, radioed back from a cruise on the research vessel Conrad to the eastern equatorial Pacific later in October 1982 that the thermocline was 100 m deeper than normal.

[00:13:25]And so, we now know, of course, that that '82 '83 El Niño was a monster with devastating consequences worldwide, and it really emphasized how little we really knew about El Niño at this point, how we had utterly failed to observe this huge event, and how ill-prepared we were for its consequences.

[00:13:49]And so, uh this hard lesson really uh caused the scientific community to laser focus on three goals for TOGA going forward.

[00:14:03]TOGA lasted uh started 1985, lasted 10 years, and those goals were simply to understand the underlying mechanisms for El Niño and other forms of tropical uh climate variability, to develop forecast models for predicting seasonal climate variations, and to establish a real-time observing system to support research and forecasting.

[00:14:27]So, it turns out that um TOGA achieved some early success with forecasting. This is uh Mark Cane's experimental forecast of the 1986-87 El Niño using a a simple coupled ocean-atmosphere dynamical model, and this shows his uh forecast in the middle panel here. This is a compa- This is the ensemble average of six individual forecasts.

[00:14:53]This is the verification. Um not a perfect forecast, uh but uh he over-predicted the amplitude and under-predicted the duration, but he got the sign right, which was something no one had ever done before.

[00:15:07]Uh and then, you know, the question was what accounts for the predictability of El Niño?

[00:15:14]And in that same paper, he focused on heat content, which he inferred from both theoretical considerations and his numerical model results, but he also gave a nod to Wyrtki who the year before had come to the same conclusion uh that predictability of ENSO was related to upper ocean heat content uh based on analysis of his tide gauge measurements.

[00:15:37]And so, this At this point, we knew that the three key variables that we needed to measure were surface winds, sea surface temperatures, and upper ocean temperatures to give us a measure of heat content.

[00:15:52]So, how did we go about doing that?

[00:15:55]Well, this was the first step.

[00:15:58]Uh we have an expression in English which is to throw spaghetti at a wall to see what sticks.

[00:16:06]And and this is a metaphor for the creative process which can be messy, um but more specifically, it says if you don't really if you got a thorny problem and you don't really know the answer, you try everything you possibly can and uh see if and see if anything makes sense at the end of it all.

[00:16:24]So, uh this was the first attempt at designing a an observing system for ENSO. And and basically, it's it's everything we had at the time. It's the uh volunteer observing ships along these tracks every couple of months making measurements.

[00:16:41]Uh there were a handful of drifting thermistor chains, moored thermistor chains, some current meter moorings, uh Klaus's tide gauges.

[00:16:49]Uh the problem is that um none of these components were really ideal. Uh it needed to be basin scale, but not all of these were basin scale.

[00:16:59]Uh not all of them could deliver data in real time, and not all of them could deliver data with daily resolution.

[00:17:06]And why was daily resolution important?

[00:17:08]Well, here's an ex post facto example from a TAO mooring that we later deployed showing the zonal winds at 0170 West. Tremendous amounts of high frequency variability, which and this is also reflected in the ocean. If you don't resolve this variability, it's going to alias into the low frequency signals that you're interested in. That was the main problem with the the XBT measurements.

[00:17:33]So, uh the breakthrough came with uh the Atlas mooring designed at PMEL.

[00:17:42]This was the brainchild of Stan Hayes and Hugh Milburn, uh the chief engineer at PMEL. Originally, the Atlas mooring was just at the mister chain. In fact, the name Atlas is an acronym for autonomous temperature line acquisition system.

[00:17:57]But, they later added surface winds and bingo, uh this was a winner. Now, they had a platform that could measure the three key variables, surface winds, sea surface temperature, and upper ocean temperature every day and deliver the data in real time to shore.

[00:18:15]Uh they also made this mooring low cost, and one of the important ways they did that was to remove current meters.

[00:18:23]And one of the preferred ways to measure ocean circulation at the time was from current meter moorings.

[00:18:29]Um the problem was that in this era uh the current meters of this era were big, heavy, clunky, and very expensive. If you did not include them on the mooring, you saved a lot of uh cost.

[00:18:41]The other thing is that more these current meters had a lot of moving parts, propellers, rotors, and veins that could easily biofoul.

[00:18:50]So, if you eliminate the current meter moorings, you not only save cost, but you can double the lifetime of the mooring from 6 months to 12 months.

[00:18:58]So, Stan took this idea one step further, and he said, "Hey, we got a low cost mooring, does everything we want it to.

[00:19:07]Let's string them out along the equator across the entire basin. Let's put 70 of these things out.

[00:19:14]And he called this idea the tropical atmosphere ocean array or TAO.

[00:19:18]And the scientific community at the time immediately embraced this idea. It was Goldilocks. It checked all the boxes in terms of meeting requirements. It was economically feasible.

[00:19:31]The challenge was to build it.

[00:19:33]That's a third of the circumference of the globe along the equator.

[00:19:38]Unfortunately, Stan passed away at a rather early age. So, he never lived to see this dream come true.

[00:19:46]I was had worked with him for several years. In fact, I was responsible for maintaining the current meter moorings embedded in in the in this array.

[00:19:55]So, the director of PMEL at the time asked me to step in and carry on.

[00:20:00]Which I did. And we we formed an implementation panel of partners, five partner nations interested in contributing resources. And then we built the array.

[00:20:11]Here's the timeline towards completion during the TOGA period 1985 to 94. And you see this big push during the second half of TOGA. It took the full 10 years to build.

[00:20:22]So, this is what the array looked like at the end of TOGA, the last month, December 1994. There is the array in red, the ship of opportunity program for XBTs in blue, a drifter network set up by this time, and Klaus's tide gauges.

[00:20:39]And all of these components were greatly expanded from the beginning of the program.

[00:20:44]But the most substantial change in the measurement strategy was the TAO array.

[00:20:49]We did have satellite measurements. The only one that really spanned the entire program was NOAA's polar orbiting satellites for SST.

[00:20:57]At the end of TOGA, TOPEX, the high-precision altimeter, TOPEX/Poseidon was launched.

[00:21:03]There were other satellite measurements, but these saw had limited utility uh because of uh problems with calibration or peculiarities in their sampling. So, we really did depend on the in situ measurements during this period.

[00:21:20]Okay, so that's the first part of the presentation, a little historical overview. Now, I want to kind of show you some of the data from the array.

[00:21:29]And we'll warm up with a few um visualizations and we'll compare December 1997 to December 1998. 97-98 El Niño was the strongest El Niño of the 20th century. It beat out 82-83.

[00:21:44]And when we're looking at these, keep in mind how much how different our our world has become because of the available availability of these data compared to 82-83 when we were flying totally blind.

[00:21:58]So, here's the uh peak of the 97-98 El Niño. It was 28-29° water filling the entire basin, collapse of the trade winds in the western Pacific. Just 1 year later, uh you can see this uh really intense upwelling in the eastern Pacific associated with a very strong trade winds.

[00:22:17]You can subtract out the climatology and here are the anomalies. Uh anomalies approaching 5° C in the eastern Pacific warm, a year later about 3 and 1/2° C cool uh during the La Niña. And you can see here two aspects of ENSO asymmetry, uh which is that um the cold anomalies associated with strong La Niñas tend to be located further west than the warm anomalies associated with strong El Niños. And the other aspect of asymmetry here is that the big El Niños tend to be bigger than the big La Niñas.

[00:22:53]Uh we can look below the surface as well.

[00:22:56]Uh this is the temperature in the upper 500 m along the equator, December 1997 with the collapse of the trade winds, the thermocline is actually sloping down to the east along the equator.

[00:23:08]A year later, a strong slope down to the west is very intense equatorial upwelling in the east.

[00:23:16]Now, aside from making these pretty pictures, uh we can learn something about uh the physics of the ENSO cycle.

[00:23:26]And uh we'll compare we'll use these data to validate the delayed oscillator theory. This is the Schopf and Suarez and Battisti and Hurst idea.

[00:23:36]Um and what happens at the onset of El Niño, according to these theories, is the trade winds weaken, generates an eastward propagating uh downwelling Kelvin wave that warms the cold tongue region.

[00:23:47]Uh this sets in uh motion the large-scale Bjerknes feedback that continue to amplify the anomalies.

[00:23:54]At the same time, uh this weakened trade wind system generates upwelling uh westward propagating Rossby waves that slowly uh propagate westward, and when they reach the western boundary, they bounce off and come back as an upwelling Kelvin wave many months after the initial weakening of the trades to shut the event down.

[00:24:16]So, we can see this sequence of events in the TAO data.

[00:24:21]And so, what I'm showing here is 5-day analyses. Now, so we have really fine temporal resolution, which is it's uh as I mentioned earlier, was critical.

[00:24:30]Time is running down. This is zonal wind anomalies, sea surface temperature anomalies, and thermocline depth anomalies as measured by 20°C.

[00:24:38]And um you can see uh the at the beginning of the El Niño, the trade winds relax, but they do so in a very episodic way. These are the westerly wind bursts, uh a lot of high-frequency variability, and these wind bursts have three important impacts.

[00:24:56]The first is they cool the western Pacific warm pool through enhanced air-sea flux, mainly evaporative flux because of the high wind speed.

[00:25:04]They generate very strong eastward currents because there's no rotational constraints at the equator, and those eastward currents advect the leading edge of the warm pool as shown here by the 29° isotherm into the central Pacific, so we get warming in the central Pacific by zonal advection.

[00:25:19]And then they generate this sequence of downwelling Kelvin waves that take a couple of months to cross the basin, and each successive one uh pushes the thermocline down deeper and uh reduces the efficiency of upwelling to cool the surface, so we get warming in the eastern Pacific.

[00:25:37]But you also notice at the trailing edge of this downwelling signal is an upwelling signal, very slow eastward propagation of upwelling. And when that upwelling uh lifts the thermocline shallow enough, it initiates cooling in the eastern Pacific and the onset of La Niña, in this case very abruptly.

[00:25:58]We can see this in the TAO data.

[00:26:01]Let's look at 5° north 20° isotherm depth.

[00:26:05]And so time is running down here.

[00:26:08]Uh we're looking at monthly averages now from 1995 to 2001, so several years.

[00:26:13]You can see the annual Rossby wave propagating westward each of these years. This is an annual signal, but some years it's different. And you can see here in uh And And these are the anomalies, uh subtracting the anomaly the uh climatology from the means, you see the anomalies.

[00:26:32]Big upwelling signal in 1997 and early 1998.

[00:26:38]So, let's leave that anomaly uh time series here for 5° north and put the anomaly time series for the equator here, 2° north to 2° south average.

[00:26:48]Okay, so now we're looking at monthly data. Here is that initial deepening of the uh thermocline along the equator during the onset and development phase of El Nino, but you can see this eastward progression of an upwelling signal.

[00:27:03]So, I'm going to do a little trick here.

[00:27:05]I'm going to flip the axis of uh the 5 North anomaly. I'm going to put west on the right and east on the left.

[00:27:13]And when you do that, the the um junction of the right and left panels is the western boundary. And what you can see is the upwelling Rossby wave comes into the western boundary and it comes back out as an upwelling Kelvin wave to shut the event down.

[00:27:34]So, the 1997-98 El Nino really uh spotlighted the great success of TOGA uh in terms of revolutionizing our ability to not only observe and understand the ENSO cycle, but also to predict. Uh as I mentioned, the the data are made available in real time from the TAO array and other components of the observing system. They can be fed into forecast models as they were in different institutions around the world to predict not only the evolution of Pacific sea surface temperatures, but also the far-field impacts like this precipitation uh forecast for precipitation uh in the winter of 97-98 in the US.

[00:28:24]So, um so we were uh uh in some sense lucky to have observed such a large signal after completing the array in 1995.

[00:28:35]Uh the there was so much focus on you know, worldwide on this tremendous event in 97-98 that the next event to come along uh struck everyone as startlingly different. The 2002-2003 El Niño uh it was weaker for one thing. Um the very little warming in the eastern Pacific and the largest warming was found in the central Pacific.

[00:29:01]And this mattered because um the associated patterns of deep convection, the far-field teleconnections, and the societal impacts were different.

[00:29:10]In fact, there were stories of uh farmers in Ecuador trying to sue the government for a busted forecast of extreme rainfall.

[00:29:18]Uh cuz they were thinking in terms of '97-'98, when in fact it was a drought in 2002-2003.

[00:29:26]And in today's parlance, we would call '97-'98 an eastern Pacific El Niño and 2002-2003 a central Pacific El Niño.

[00:29:35]And you know, with so much data available now on uh for understanding and and describing ENSO cycle variations, uh we were able to think in terms of not just what was common between El Niño and La Niña events, like say with the Rasmussen's or Carpenter composite, but we could begin to think about their differences and importantly, why they were different. And so, it's about this time that the uh study of ENSO diversity and complexity really took off.

[00:30:10]Um Okay.

[00:30:14]In 1997, Fei-Fei Jin published his recharge oscillator theory. This is an elegant mathematical representation of the concepts that Birky and Cane discussed a decade earlier in terms of the importance of heat content as a predictor of ENSO.

[00:30:28]And a recharge oscillator is consistent with the laid oscillator. It incorporates the effects of the waves without the details of the equatorial waves.

[00:30:39]And this whole schematic shows how it works. Uh you know, so the the solid black line here is a representation of thermocline depth. So, a build-up of excess heat content along the equator is a necessary precondition for El Nino to occur. It can trigger the onset of El Nino, which during the El Nino excess excess heat is purged. It causes the thermocline to shoal and in some cases overshoot, which sets up the conditions favorable for La Nina. And then La Nina causes a build-up of heat content along the equator and the cycle repeats.

[00:31:14]Well, we were able to validate this recharge oscillator theory using TAO data.

[00:31:21]Uh doing a EOF decomposition of the 20° isotherm depth using about 20 data 20 years of data.

[00:31:29]And the two dominant EOFs are shown here. This is the first EOF. It's a tilt mode, east-west tilt.

[00:31:35]And the second mode is what we call the recharge mode. It's basically a uniform heat along the equator up and down.

[00:31:43]And these are the principal components.

[00:31:45]The blue line for the tilt mode and the red line for the recharge mode.

[00:31:49]Now, if you over plot on these principal components, the Nino 3.4 in green and the wind stress in the central Pacific in black, you can see that wind stress Nino 3.4 and the tilt mode are all in phase, just like this little picture.

[00:32:11]And you can see that the red line PC1 PC2 for the recharge mode is in quadrature and leads the tilt mode, just like in this little picture.

[00:32:22]And so, with these data, we were able to validate this really um beautiful theory of Faye Faye's, which has turned out to be a powerful diagnostic for ENSO variability.

[00:32:35]Um so, we've come to appreciate that ENSO evolution is go- governed by multi-timescale processes. Um, they at the one end we have these deterministic seasonal timescale uh, dynamics that are incorporated into recharge oscillator and delayed oscillator.

[00:32:55]And at the higher end we have this high-frequency weather noise forcing, uh, fluctuations on uh, timescales of days to weeks. These are the westerly wind bursts.

[00:33:05]And these wind bursts introduce irregularity into the ENSO cycle in terms of timing, duration, and amplitude.

[00:33:13]And yeah, so, this noise we know affects ENSO variability, but it turns out that ENSO, the phase of ENSO, affects the statist- the statistics of the noise.

[00:33:27]And we refer to this as state-dependent or multiplicative noise forcing. And that's illustrated here.

[00:33:34]Again, we'll use data from the '97-'98 El Niño.

[00:33:38]And on the right, no, that's the left, I'm sorry. On the left, we have the total sea surface temperatures.

[00:33:45]Uh, there's the 29° isotherm overlain on that. Time is running down again.

[00:33:50]In the middle panel, we have outgoing longwave radiation. Uh, it's a measure of high cloudiness. You can see lots and lots of intraseasonal variability.

[00:33:58]All contained behind this 29° isotherm.

[00:34:02]Um, this intraseasonal variability in this particular instance is largely related to the Madden-Julian oscillation. These these uh, convective cells filling the entire tropos- sphere generated over the Indian Ocean and propagating eastward into the Pacific.

[00:34:19]And the surface manifestation of the Madden-Julian oscillation is these westerly wind bursts. Again, all this high-frequency variability contained over the warm water where can- where which is favorable for developing these convective cells.

[00:34:33]Uh over the cold water, as a surface manifestation of these westerly wind bursts is greatly diminished. And so, uh this state-dependent noise forcing is important uh as a it's important in the non-linearity of the ENSO cycle, and it's one of the reasons why El Niños are stronger than La Niñas because the noise forcing is amplified during warm events as compared to cold events.

[00:35:01]Okay, so we revisited Fei-Fei's recharge oscillator uh in the early 2010s with another decade of data.

[00:35:14]And lo and behold, we discovered decadal variability in the chara- characteristics, dynamics, and predictability of ENSO.

[00:35:25]And so, what I'm showing here is El Niño 3.4 in blue.

[00:35:30]And in the red, uh this is an index for the recharge mode. It's basically heat content uh based on an average across the basin between 5 north and 5 south of uh temperature anomalies in the upper 300 m of the water column.

[00:35:47]And you can see across year 2000 um a big change.

[00:35:52]And there's a whole list of changes actually here, and I'll walk through these quickly. The first one is the cycle becomes weaker and higher frequency after 2000 compared to before 2000.

[00:36:04]The lead time between heat content and sea surface temperature weakened uh shortens to one season, whereas it was two to three seasons in the '80s and '90s.

[00:36:14]A one-season lead greatly diminishes the effectiveness of heat content as a predictor because it actually performs no better than SST persistence at one season lead.

[00:36:25]Um across 2000, there was a shift also from strong a dominance of strong Eastern Pacific El Niños in the '80s and '90s to weaker Central Pacific El Niños in the 2000s.

[00:36:40]Um the thermocline feedback, which is most active in the equatorial cold tongue region, that's how the thermocline affects sea surface temperature. The thermocline feedback weakened, which is why we did not have big sea surface temperature anomalies in the Eastern Pacific.

[00:36:56]But according to Fei-Fei Jin's recharge oscillate and be according to his BJ index uh diagnostic the thermocline feedback is the strongest of the positive feedbacks that energizes the ENSO cycle. So, when it weakens, it means that the ENSO cycle becomes more damped and stable.

[00:37:15]A more damped and stable ENSO cycle means that noise forcing become becomes more prominent in energizing ENSO variations.

[00:37:24]And because noise is so prominent operational forecast skill declined from the '80s and '90s to the 2000s. So, we've got this welter of changes uh on decadal time scales across the 2000.

[00:37:40]Now these changes in variability and predictability uh are linked to a long-term La Niña-like cooling trend over the past 40 years in the tropical Pacific basin.

[00:37:53]What accounts for this trend is an unresolved but very important problem.

[00:37:58]Uh it could be uh unresolved decadal variability.

[00:38:02]Uh the PDO, Pacific Decadal Oscillation, shifted sign from the 1980s and '90s to the 2000s. So, it could be unresolved decadal variability, but some uh theories suggest that this is how the tropical ocean should respond to greenhouse gas forcing, elevated greenhouse gas forcing.

[00:38:21]Um you know, the answer probably lies somewhere in between, but um this is still a very hotly debated topic.

[00:38:28]In any case, we can see these multi-decadal trends in both um observations and in oceanic and atmospheric reanalyses constrained by those observations.

[00:38:39]And this is one reanalysis, it's a GODAS reanalysis, and these are 40-year trends. Uh it's this forced by NCEP winds. You can see the stronger trades uh over this the trend towards stronger trades, trend towards cooler sea surface temperatures in the east, warmer in the west, and a more steeply sloping thermocline down to the west.

[00:39:01]Now, um you may not be a fan of reanalysis because they have known issues, but these trends are also apparent in the raw data themselves, and we'll look at that and how these trends affect ocean circulation.

[00:39:17]So, these are TAO wind data uh looking at the zonal component of the winds, 2000s minus the 1990s. So, you can see in the central and western Pacific, the trade winds have intensified by 1 to 2 m/s. This is basically purely from the TAO wind data.

[00:39:35]The stronger trade winds pile up more water in the western Pacific, so we have a higher sea level in the west. This is from altimetry. Lower sea level in the east.

[00:39:45]Uh thermocline depth is a mirror for sea level. So, if we look at the TAO data below the surface at the mooring sites along the equator, in 2022, the thermocline was shallower in the east and deeper in the west than it was in 1993 by uh several meters, 10 m or so.

[00:40:05]Uh and we have these long-term current meter moorings along the equator. We can see the 40-year trends in the undercurrent.

[00:40:12]Undercurrent flows upward. Uh it flows eastward in the thermocline, and because the winds are stronger, the sea level is higher, the thermocline is is more tilted down to the west, uh the pressure gradient that forces the undercurrent is stronger.

[00:40:28]And so, what we find is the undercurrent has accelerated over the last 40 years in the thermocline, and it's it's deep uh it's shallower in the east in 2022, and it's deeper in the west in 2022 than it was in 1993.

[00:40:46]The other thing we can see is that um from the drifter buoy data, now we got 40 years of that in the tropical Pacific, the westward South Equatorial Current has become stronger, the eastward North Equatorial Countercurrent has become stronger.

[00:41:03]That means the shear, the lateral shear between these two currents has become stronger, and we can see from the long-term mooring data at 0140 West, that the tropical instability waves have become stronger. There There's more energy to extract from the the mean circulation, and so this shows a trend towards strengthening um eddy kinetic eddy kinetic energy in the tropical instability wave band over the last 40 years.

[00:41:34]So, these trends are real. Uh they're apparent in reanalyses, and they're apparent in the raw data that we've collected over the last 40 years from the tropical Pacific observing system.

[00:41:46]Uh I just want to say a few words about forecasting.

[00:41:51]Um you know, from that first experimental forecast that Mark Cane published 40 years ago now, 1986, there are about 20 groups worldwide that routinely issue ENSO forecasts, and this is the most recent collection for March 2026 initial conditions, um predicting uh this is El Nino 3.4 anomaly predicted from these various groups. Uh, you can see that we're currently in a coming out of a a weak La Nina, but all these models, most of these models are predicting a warming uh, developing later this year. Some of them quite strong. Uh, now we have to be careful because we're in that season of the spring predictability barrier where there's a lot of uncertainty. Um, but this is the the current state of the forecast.

[00:42:39]Now, we've talked about how uh, initializing these models with upper ocean measurements is important.

[00:42:47]Uh, we can quantify what that impact is by looking at the various generations of the ECMWF forecast model system. Uh, of which there's five. The first was implemented in 1997 at and the most recent in 2017. And this is from uh, chapter three of our recently published ENSO book, uh, on observations.

[00:43:13]And the metric here is the lead time for which the anomaly correlate correlation coefficient between the observed and the predicted Nino 3.4 stays above 0.9. So, very high correlation. And you can see with each successive generation of the ECMWF forecast system, uh, this lead time is extending. And for system five, it's it's uh, four and a half months.

[00:43:40]Now, if you use system five to reforecast the last 30 years of El Nino events, but with no subsurface data in the initial conditions. This is any This is all data. This is XBTs, TAO, Argo, and it's even the uh, altimeter uh, data which is projected downward. If you exclude all that subsurface information, you find that the uh, the the lead time relaxes to only three months, the same level that it was in 2002. So, with no subsurface information assimilated into the initial conditions, the drop in forecast skill for the ECMWF system is equivalent to 20 years of progress in forecast model development.

[00:44:22]So, with that uh I'll end with some concluding comments.

[00:44:32]Uh to be clear, the global ocean observing system as it exists today is much more comprehensive than I've described in this presentation.

[00:44:40]There's a vast array of in situ observing components uh complemented by a diverse constellation of Earth observing satellites.

[00:44:48]And in the past 25 years, we've witnessed the advent of the Argo revolution.

[00:44:52]Uh we've expanded the TAO array into the Atlantic and Indian Oceans, and we've introduced new technologies like gliders and sail drones.

[00:45:00]TAO itself has undergone evolution. It was rechristened TAO Triton in 2000 when JAMSTEC introduced Triton moorings that were designed to be functionally equivalent to ATLAS moorings into the western Pacific.

[00:45:14]Uh in the mid-2000s, NOAA transferred the management of the TAO array from PMEL to the National Weather Service, who unfortunately let the array collapse in 2012 to 2014.

[00:45:27]And um that's when they did not fund the mid-life refit of the ship that serviced the array.

[00:45:32]This prompted JAMSTEC to withdraw its Triton moorings from the western Pacific, a region critical for understanding and predicting ENSO.

[00:45:40]Despite these setbacks, NOAA resuscitated TAO under the guidance of the TPOS project, and it continues to serve a vital function in GOOS.

[00:45:51]And uh with regard to the Triton hole here, these moorings that we now greatly miss, um we're fortunate that that there's redundancy in the observing system.

[00:46:01]Argo and satellite measurements can provide some of the um some of the coverage now in this region.

[00:46:07]But it's not the same as having long time series of high quality synchronous collocated multivariate hourly data of key oceanic and atmospheric parameters at fixed locations in the western Pacific warm pool. So at some point it would be good if we can to reestablish at least some of these sites.

[00:46:28]Um So, beyond this though, I see at least two grand observational challenges related to ENSO dynamics and its impacts.

[00:46:41]And the first is regards to climate change.

[00:46:46]Uh the instrumental record, paleo proxies, and climate change models all show that ENSO cycle the ENSO cycle has become more energetic in the latter half of the 20th century as shown down here.

[00:46:59]Uh there's also evidence to suggest that ENSO impacts are becoming associated with more extreme weather.

[00:47:06]And you know, these are related to the uh rises of greenhouse gas concentrations in the atmosphere. Uh rises that will likely continue into the future.

[00:47:16]So, this requires that we continue to sustain the long climate records we've established in the tropics to ensure we have an accurate baseline to quantitatively assess not only changes in ENSO characteristics but also changes in the large-scale background conditions on which they occur.

[00:47:35]Uh we also need a better understanding of the key physical processes underlying cycle ENSO cycle dynamics in order to know why changes occur to ensure that dynamical climate forecast models properly represent these uh processes. So, uh these process-oriented field studies, like illustrated here the cold tongue and uh warm pool studies, uh we need to continue uh improving our process understanding by going out in the field and and and learning more deeply about processes that are important in ENSO cycle dynamics.

[00:48:12]The second important challenge I see regarding ENSO dynamics is that understanding better the uh two-way interbasin interactions that affect ENSO variability and predictability. Um currently, there appears to be a practical limit to ENSO forecast skill of about 9 months due largely to the spring predictability barrier.

[00:48:33]Rapid advances in artificial intelligence and the recent development of deep learning forecast models are pushing these limits.

[00:48:40]Uh but it's clear that at longer lead times, there are significant sources of predictability that lie outside the tropical Pacific and at higher latitudes.

[00:48:49]As shown in this uh schematic here.

[00:48:53]Uh in fact, uh and and we know these two-way interactions affect ENSO diversity, asymmetry, and predictability. And this is an example of how forecast skill can be enhanced out to 15 months using an extended recharge oscillator model configured for forecasting, uh one that includes the impacts of the global tropics as these different regions here.

[00:49:21]Such a model out-performs significantly out-performs a recharge osc- a recharge oscillator model running prediction mode that includes only the equatorial Pacific, particularly at the long lead times.

[00:49:35]And so, our view of ENSO observation should expand to incorpor- incorporate a wider geographical focus as proposed in this Frontiers article recently by um Greg Foltz et al. So, we can better characterize these interbasin interactions and and properly incorporate them into seasonal forecast models.

[00:49:55]Um it's also important to reduce the dynamical model systematic errors that limit predictability in all three ocean basins.

[00:50:02]And observations are valuable for observations are valuable for assessing these area errors and for guiding model improvements to reduce them.

[00:50:12]And so with that, I'm going to end because I think I'm out of time. And thank you for your attention.

[00:50:18]And uh I'll take any questions.

[00:50:24]Thank you very much, Mike. Um that was a great overview um of the beginning of ENSO observations towards the end and where we are now what we now now know and what we need. Um really appreciate that. Thank you so much. Um I can see a hand raised. So, for everyone who's joining, uh feel free to yeah, re- raise your hands or and put your questions in the chat and then you're um you're welcome to pose your question.

[00:50:55]And I can see that uh Alexi, I think was the first to um raise his hand. So, if you want to ask your question, feel free to unmute.

[00:51:06]Um thank you, Mike. Uh can you hear me?

[00:51:10]Again. Yes. Thank you.

[00:51:11]>> Can you hear me? Yes.

[00:51:13]>> Yeah, okay. Uh great talk. Great talk. Uh one question, what is the status of array now and perspective for the next couple of years and the future?

[00:51:28]Right. Um you know, that is a question that I mean there's a this TPOS, Tropical Pacific Ocean Observing System project that's been running for about 10 years and they have been involved in um well, they were involved in resuscitating the array after the weather service let it collapse, but they've also been redesigning it. Um and you know, that question is better uh left for somebody who's involved in TEPOS and and maybe there'll be a presentation in the seminar series about TEPOS, which I think would be a nice compliment to what I've talked about. I mean, I was kind of looking uh more at the development and and some of the bigger scientific issues, but the details of what is going on in TEPOS are very interesting.

[00:52:11]Thank you.

[00:52:13]Thank you very much. And then we had another hand raised from a DTD.

[00:52:17]So, I can see if you want to unmute, please. Um yeah, thank you. Am I audible?

[00:52:22]Yes.

[00:52:23]Uh all right. Uh uh great talk, Mike. Uh I have actually a two-part question. One is refers to the part where you said that the ENSO evolution is essentially a multi uh scale interaction process. And now with climate change and ocean warming, uh I would like to know if we have an understanding of whether these multi-scale interactions are also changing and what way they would affect ENSO evolution. That is one part. And the second part is we also know that with warmer oceans, our global circulation is getting affected. So, in what way do you think that would affect ENSO behavior or ENSO evolution? Thanks.

[00:53:09]Yeah. Yeah, I I think the the longer time scales are affecting ENSO variability. I mean, this this trend towards cooling uh you know, the the the 2000s, the cycle has generally been weaker.

[00:53:21]It seems to be picking up now. Uh so, there may be some natural decadal variability in there.

[00:53:26]But, you know, there was a period of time for about 15 years at the turn of the 20th 21st century where noise forcing was really uh took on greater prominence in affecting ENSO cycle variability. So, so that's an example of where the you know, there was a shift in the multi-time scale nature of ENSO towards a more noise-driven cycle as opposed to one that was you know, perhaps closer to neutral stability and and a little bit more freely oscillating.

[00:53:54]Yeah, I mean this um the time scales extend beyond seasonal to decadal and and centennial in fact.

[00:54:02]So, sorting that out is is really important.

[00:54:05]Um you know, there are influences of AMOC on the tropical Pacific on even longer time scales. So, I mean you know, what happens in the Atlantic relative to you know, the AMOC circulation in the longer time scale in the longer time frame is is something to be concerned about as well.

[00:54:24]Yeah, I mean you know, it's this is ENSO's like an onion, you know, you peel back a layer and there's always new things to find. And you know, these connections to other parts of the climate system are really fascinating.

[00:54:38]Thanks Mike. Um we have one more question in the chat. Um Uh this as model resolution is increasing, we need higher resolution observations in the ocean. With help of all past data and model simulations reanalysis for past periods, can AI or machine learning be used to generate ocean data in real time?

[00:54:59]If this allows you to comment on uh Uh wait a minute. Now you're saying generate fake data?

[00:55:06]Yeah, I yeah I I don't know. We don't want fake data. I mean, wasn't Henry Stommel said data is the lifeblood of oceanography?

[00:55:14]Uh you know, we do not want fake data.

[00:55:17]Fake news, I don't know, but not fake data.

[00:55:21]Yeah, we might get a fake fake news about an El Niño coming or not coming, but yeah, we do want the the real deal.

[00:55:32]Okay, so then there's been a couple of questions about how people can access this presentation. So, this presentation is recorded and will be uploaded, I believe, to YouTube under the CLIVAR webinar series, so you will be able to rewatch it if you've missed any of the details.

[00:55:46]>> Yeah.

[00:55:49]Yeah, I get I can make the slides available, too, if anyone wants them.

[00:55:53]Thank you very much. We might include them then in the wherever the presentation is. That would be wonderful.

[00:56:01]But, thank you very much. Um I don't see any more hands or questions.

[00:56:07]So, in the interest of time, oh, one more question. Okay. Go. You can unmute yourself.

[00:56:14]Yeah, hello. This is Isabel from Bolivia. Great talk.

[00:56:19]Um I wanted to ask you what about this relative uh only system now? The only predictions just removing the anomaly of the sea surface.

[00:56:31]Right.

[00:56:32]>> Are they being used now for the forecast? Or Uh NCEP has gone to using it as a as their operational index now.

[00:56:42]Um you know, and ONI is really a manifestation of the interbasin interactions that are affecting ENSO.

[00:56:50]Because the what it's doing is it's affecting the local convective response to sea surface temperature anomalies in the Pacific. The warming trends in the Atlantic and the Indian Oceans are affecting the local convective response to sea surface temperature in the Pacific. That's why we need this relative index. That's an example of the interbasin interactions that we need to be paying attention to to improve our um understanding and ability to forecast ENSO.

[00:57:18]Yeah, so we're talking about the Oceanic Niño index in case any >> Yeah. catch >> Yeah, the relative Niño index. Exactly.

[00:57:27]Okay.

[00:57:28]We have one more A in the chat asking whether there's still value in um going out and getting new observations. I I feel like you made that point. Um Yeah, I think, you know, I mean if you make a map and you put all the different measurement types on the map, it looks really dense.

[00:57:47]But you realize um you know, that the ocean is huge. It's It's a huge expanse and those little symbols on a on a small map look big.

[00:57:57]But um you know, there are scales that we don't uh there are scales that we have not measured well, not extensively.

[00:58:05]There are scale interactions we don't understand. Um and so there is, you know, there's always a need for new information about the key processes that affect ENSO cycle variability. And then, you know, down the line are those important enough to try to build into forecast models.

[00:58:22]Exactly. Yeah, Mike, I I um I just wanted to I appreciate your point that you made earlier that maybe we should have one of the talks about the deep ocean and the TPEX um upcoming experiments going towards this region and measuring more things.

[00:58:36]So, seems that people are interested in that, which is great. Um so, yeah, stay tuned. Um with that, we are going to wrap up. So, I'm going to thank Mike again. Thank you so much for this uh presentation. And everyone, thank you for joining for whatever time at whatever time it is for you. If it's not Europe, then I think I think it was either early or late for you. So, thank you so much for being here at really great numbers.

[00:58:59]And um yeah, we hope to see you again for the next one. Keep your eyes peeled for the next announcement of the next installment of this webinar series.

[00:59:06]Thank you so much, Mike.

[00:59:08]Thank you.