Zhang’s ICM model a success among failures

Last November, NOAA’s Climate Prediction Center ran a routine “Sanity Check,” checking the progress of model forecasts for sea surface temperature (SST) conditions associated with El Niño and La Niña events in the tropical Pacific. In a widely circulated e-mail, Dr. Martin Hoerling of the NOAA Earth System Research Laboratory brought to attention the failed forecast of Niño 3.4 SST from the June initial condition, particularly in depicting what’s called a “double dip evolution” of cold SST anomalies in the middle-late 2011 by “almost all models.”*

The failure was true for the prediction of the 2011 cold SST conditions in the tropical Pacific, with the exception of Dr. Rong-Hau Zhang’s model called “ESSIC ICM.” Zhang, Senior Research Scientist at ESSIC and Affiliate Professor at Department of Atmospheric and Oceanic Science at the University of Maryland, developed a model that delivered the only correct predictions from June until November of this past year.

Zhang’s success begins at the International Research Institute for Climate and Society (IRI) at Columbia University, where he worked as an Associate Research Scientist from 1999 to 2003. There, he focused on developing and improving El Nino models that could be routinely used to predict climate for society as a whole. Many of these El Niño predictions were made after the development of the first ENSO dynamic model at Columbia University in the mid-1980s.

In the decades that followed, members of the scientific community developed various models in order to improve the early predictions, said Zhang.

In 2002, scientists Keenlyside and Kleeman developed a new intermediate ocean model (IOM) as an extension of a previous, less accurate, baroclinic modal model. The new dynamical ocean model enabled scientists to take into account new methods of collecting data and creating predictions, including spatially varying stratification and partial nonlinear effects.

According to a paper published by Zhang in 2003, this meant that the new ocean model could “well simulate the observed mean current structure over the tropical Pacific Ocean, including the equatorial undercurrent (EUC) and the seasonal variations of surface equatorial currents. In particular, the model realistically captures the phase and amplitude of the seasonal cycle of zonal currents, as well as the westward propagation of seasonal maxima in the eastern basin.”

Based on this intermediate ocean model, Zhang began to develop a new intermediate coupled ocean-atmosphere model that he hoped would better simulate and predict El Niño.

“Without a good model you can’t predict something in the future; that’s why we need to develop realistic models for the coupled system of the atmosphere and the ocean in the tropical Pacific basin,” said Zhang.

In 2003, he presented a new intermediate coupled model (ICM) used to realistically simulate and predict sea surface temperature (SST) variability in the central and eastern tropical Pacific.

“This intermediate coupled ocean-atmosphere model is a simplified model that is computationally efficient and can be used for many, many calculations,” Zhang said.

One of the more important calculations to take in to account is vertical mixing; which is not adequately represented in some models, said Zhang.

This essentially means that his model takes in to account water temperatures not only at the surface but just below it as well, developing optimized parameterization for intermit temperatures that may be mixed in at a later time, said Zhang.

“Coupled with an atmospheric model, this is very realistic,” he said. “This feature of the model is the basis for future predictions. Without the model you cannot predict this, that’s why we develop realistic models for the atmosphere and the ocean.”

While the ESSIC ICM is the only one to recently produce accurate predictions, there are still many other good models out there, though real-time forecasts of sea surface temperature evolution in the tropical Pacific is not easy to predict longer than six months in advance.

“This is a very huge topic, even when there are very good models there needs to be a procedure of initialization,” he said. “ A consistency between the model, the forcing fields and the initial conditions is critically important for EL Niño/La Niña forecasts at a long  lead time, say six months or longer, using coupled models.”

Along with Zhang’s successful model, there are more than 20 other models for real time ENSO prediction, including eight statistical models and 15 dynamic models across the country and internationally.

The success of each of these models is measured against the real-time observations.  The ESSIC ICM model can predict evolutionary features of the sub-surface temperatures in the tropical Pacific, said Zhang.

A better representation of subsurface thermal structure and evolution in the tropical Pacific and then their connections with SST on the equator are most likely a contributor to the model’s success, he said.

Zhang has been sending his model predictions to the IRI every month since 2003, and actually had no idea that only his model was producing accurate results for the 2011 cold conditions in the central and eastern tropical Pacific.

Since being notified of his success by other scientists, Zhang has been asked for a detailed analysis of his model and to explore why it performed better than the others.

“It’s a very difficult question,” said Zhang. “I have to analyze lots of data together to see the entire process.”


*El Niño-Southern Oscillation (ENSO), a natural climate phenomenon centered in the tropical Pacific, has significant impacts on climate variability and predictability all over the world. In the past two decades, extensive ENSO studies have made remarkable progress. In particular, reasonable predictions can be now made six months or longer in advance (e.g., see the summary of model ENSO forecasts at the International Research Institute for Climate and Society (IRI) website:http://iri.columbia.edu/climate/ENSO/currentinfo/update.html). Still, large uncertainties do exist in its simulation and prediction, as represented by large discrepancies in real-time forecasts of ENSO.

-A Q&A on the failed Niño 3.4 SST forecasts and reasonin
g behind the problems can be found here as well as attached below.-


{arisexylightbox activeType=”rcontent” width=”75%” height=”75%” url=”http://essic.umd.edu/joom2/media/com_form2content/documents/c8/a597/f46/Zhang ICM Model Image.pdf” link=”Download Attachment” title=”” }{/arisexylightbox}