Under a DOE contract that John Christy and I signed, we’re using satellite data to study the behavior of climate models. One of the issues I was interested in is the impact of El Nino and La Nina (ENSO) on our understanding of man-made climate change. A large number of ENSO records show multidecadal variations of this activity, and it has even shown itself in multi-year runs of a GFDL climate model.
Since El Nino generates global average heat and La Nina generates global average cool, I used our 1D Forcing Feedback Model of Sea Temperatures (published by Spencer & Braswell, 2014) to study how the historical record of ENSO variations can be included. using the co-variations of the TOA (top-of-atmosphere) radiation flux observed by the CERES satellite with ENSO.
I updated this model to match the 20 year CERES data (March 2000 – March 2020). I also have the ENSO dataset up to 1525 with Braganza et al. (2009) Multi-Proxy ENSO Reconstruction Data. I have intercalibrated it with the multivariate ENSO Index (MEI) data to date and expanded it to mid-2021 based on the latest NOAA ENSO forecast. The Cheng et al. The temperature data reconstruction for the 0-2000 m layer is also used to calibrate the adjustable coefficients of the model.
I had been working on an extensive blog post with all the details about how the model worked and how ENSO was represented, which was far too detailed. Instead, after a brief description of the model, I’ll just show you some results.
1D Description of the forcing feedback model
The model assumes an initial state of energy balance and calculates the temperature response to changes in the radiation equilibrium of the global ocean-atmosphere system using the global radiation forces CMIP5 (since 1765) along with our calculations of ENSO-related forces. The model time step is 1 month.
The model has a mixed layer with adjustable depth (50 m resulted in optimal model behavior compared to observations), a second layer with a depth of 2,000 m and a third layer with a global average seabed depth of 3,688 m. The transfer of energy between the ocean layers is proportional to their deviation from equilibrium (zero temperature anomaly). The proportionality constants have the same units as the climate feedback parameters (W m-2 K-1) and are analogous to the heat transfer coefficient. A transfer coefficient of 0.2 W m & supmin; ² K & supmin; ¹ for the bottom layer was 0.01 °. C the net warming of the deep ocean (below 2000 m) in the last few decades, the Cheng et al. There is some limited evidence for mentioned.
The ENSO-related forces are both radiant (short and long wave) and non-radiant (increased energy transfer from the mixed layer to the deep ocean during La Nina and the opposite during El Nino). These are explained in more detail in our 2014 paper. The appropriate coefficients are adjusted to provide the best model fit with the behavior observed by CERES compared to the MEIv2 data (2000-2020), observed SST fluctuations, and observed deep-sea temperature fluctuations. The full 500-year ENSO record is a combination of Braganza et al. (2009) based on monthly interpolated annual data, the MEI-extended, MEI and MEIv2 data, all intercalibrated. The Braganza ENSO dataset has a mean of zero over its entire period (1525-1982).
Results
The following graph shows the global average (60N-60S) mixed layer temperature fluctuations produced by the 1D model after the model has been set to match the observed temperature trend of sea surface temperature (1880-2020) and the deep sea temperature trend of 0-2000 m (Cheng et al., 2017 analysis data).
Fig. 1. Temperature fluctuations of the 1D model for the world’s oceans (60N-60S) up to a depth of 50 m compared to observations.
Note that the net radiation feedback parameter specified in the model corresponds to an equilibrium climate sensitivity of 1.91 degrees. C. When the model was forced to conform to the SST observations between 1979 and 2020, the ECS was 2.3 degrees. C. Deviations from these values also occurred when I used HadSST1 or HadSST4 data to optimize the model parameters.
The ECS result also depends heavily on the accuracy of the 0-2000 meter sea temperature readings shown next.

Fig. 2. 1D model temperature changes for the 0-2000 m layer since 1940 and compared to observations.
The 1D model has only been optimized since 1995 so that it corresponds to the temperature trend from 0 to 2000 m. However, in Figure 2 we can see that the limited data available up to 1940 also show reasonably good agreement.
Finally, here’s what the full 500 year model results look like. Again, the CMIP5 forces don’t begin until 1765 (I assume zero beforehand) while the combined ENSO data set begins in 1525.

Fig. 3. The model results go back to 1525 with the proxy ENSO drives and since 1765 with CMIP5 radiation drives.
discussion
The simple 1D model is intended to explain a large number of temperature-related observations with a physically based model with only a few assumptions. All of these assumptions can of course be flawed in one way or another.
However, the monthly correlation of 0.93 between the model and the observed SST variations from 1979 to 2020 is very good (0.94 for 1940 to 2020) because it is such a simple model. Our main purpose was to re-examine how the observed ENSO activity affects our interpretation of warming trends in terms of human causation.
For example, ENSO can then be deactivated in the model to see how this affects our interpretation (and causes) of temperature trends over different time periods. Or one can investigate the effects of the assumption of a certain imbalance in the climate system at the time the model is initialized.
If nothing else, the results in Figure 3 could give us an idea of the ENSO-related SST variations for 300-400 years before anthropogenic forces became significant, and how those variations affected temperature trends on different time scales. Because if these naturally induced temperature trend fluctuations existed earlier, then they still exist today.