This article in MIT News:
https://news.mit.edu/2025/simpler-models-can-outperform-deep-learning-climate-prediction-0826
“New research shows the natural variability in climate data can cause AI models to struggle at predicting local temperature and rainfall.” … “While deep learning has become increasingly popular for emulation, few studies have explored whether these models perform better than tried-and-true approaches. The MIT researchers performed such a study. They compared a traditional technique called linear pattern scaling (LPS) with a deep-learning model using a common benchmark dataset for evaluating climate emulators. Their results showed that LPS outperformed deep-learning models on predicting nearly all parameters they tested, including temperature and precipitation.“
Machine learning and other AI approaches such as symbolic regression will figure out that natural climate variability can be done using multiple linear regression (MLR) with cross-validation (CV), which is an outgrowth or extension of linear pattern scaling (LPS).
https://pukpr.github.io/results/image_results.html
“When this was initially created on 9/1/2025, there were 3000 CV results on time-series
that averaged around 100 years (~1200 monthly readings/set) so over 3 million data points“
I developed this github model repository to make it easy to compare many different data sets, much better than using an image repository such as ImageShack.
There are about 130 sea-level height monitoring stations in the sites, which is relevant considering how much natural climate variation a la ENSO has an impact on monthly mean SLH measurements. See this paper Observing ENSO-modulated tides from space
“In this paper, we successfully quantify the influences of ENSO on tides from multi-satellite altimeters through a revised harmonic analysis (RHA) model which directly builds ENSO forcing into the basic functions of CHA. To eliminate mathematical artifacts caused by over-fitting, Lasso regularization is applied in the RHA model to replace widely-used ordinary least squares. “
https://www.tide-forecast.com/system/charts-png/Brest-France/tides.png
“Drivers and predictability of extreme still water level trends and interannual variability along the coast of Australia across different time scales”
https://www.sciencedirect.com/science/article/pii/S0378383925000304
https://www.americanwx.com/bb/topic/61644-2025-2026-enso/page/106/
“The Influence of ENSO on the Structure of Internal Tides
in the Xisha Area”
https://agupubs.onlinelibrary.wiley.com/doi/pdfdirect/10.1029/2019JC015405
Another important recent paper is “Evaluation and prediction of the Effects of Planetary Orbital Variations to Earth’s Temperature Changes” https://www.tandfonline.com/doi/epdf/10.1080/17538947.2025.2487058?needAccess=true
“The influence of planetary orbital changes on Earth’s temperature has been poorly quantified and subject to speculation. Here, we delineated the effects of greenhouse gases and planetary orbital changes on Earth’s temperature and forecasted the latter. Our results indicate that Earth’s revolution around the Sun and its rotation explain ∼75.36% and 15.91% of Earth’s temperature variations in one year, while the Moon’s revolution around the Earth and other planet motions account for 8.26% and 0.26%, respectively. Orbital forcings contributed ∼11.5% global warming since 1837 and will continue to warm the Earth by ∼0.13 °C from 2020 to 2027. However, orbital forcings may contribute to ∼0.25 °C cooling of Earth from 2027 to 2050, but this effect remains insufficient to offset the warming caused by CO₂”
So
1. 75.36% is seasonal
2. 15.91% is daily
3. 8.26% is lunar cycles
4. 0.26% is planetary cycles
5. (I calculate 0.21% unaccounted for)
How they come up with the 8.26% for the lunar orbit is fascinating. I think they are finding a pattern match in day-to-day measurements separated by fairly long time spans but it appears to be pinned to vernal equinoxes. The common sense (perhaps overly naive) idea would be to attribute this simply to weather variability, yet they specifically state it’s a lunar cause. ” This regular rising and falling pattern in temperature is accompanied by slight irregular fluctuations. We believe that these slight irregular fluctuations are mainly due to changes in lunar forcing exerted on Earth”
The way they substantiate this is “f1(t) quantifies the influence of Earth’s rotation on its temperature variation in a year. Building on Section 2.2.1 above, the quasi-sinusoidal diurnal variation in Earth’s temperature is mainly due to its rotation. We calculated the temperature difference (D1) of two consecutive hours for each day from 1979 to 2020. A remarkable agreement is found between the D1 values of two consecutive days. For example, the D1 values for January 1 and 2 from 1979 to 2020 are shown in Figure 6. These results indicate that the variation in D1 is mainly determined by the rotation of the Earth. Therefore, the changes in D1 can be used to quantify the impact of Earth’s rotation on the global temperature in consecutive hours, which can be inferred by taking the derivative of the function f1(t). The derivative of f1(t) is fitted to the time series of D1 by using the Levenberg – Marquardt scheme to determine its unknown parameters. f2(t) quantifies the influence of the lunar revolution on Earth’s temperature variations in a year.The changes in CO2 concentration for two consecutive days have almost no effect on Earth’s temperature. Thus, we can determine the change in Earth’s temperature over two consecutive days (due to the motion of the Moon) by using Equations (16) − (18). To eliminate the impact of large variations in CO2 concentration in two successive days on temperature, we utilized only D2 values for days in which the CO2 concentration was relatively stable.”
https://imagizer.imageshack.com/img922/2284/VXFgQ4.png
Someone needs to duplicate these findings.
Pingback: Simpler models … examples