Trying Out a Planetary Ring System for ENSO Prediction

If Earth has a ring system (as other planets do), it could be the driver of ENSO, as the changing configuration of a likely ring system would drive global cycles of Earth shading on ENSO-like time frames. Further, ring dust that must fall to Earth continually would favor locations cyclically. Cold and electrically charged, it would drive local cycles in wind speed and direction, precipitation intensity, frequency of lightning strokes, sea surface temperature, and nutrient availability.

The purpose of this effort is to derive and present an ENSO prediction from that hypothesis, for comparison with the climate of the next few years to 2020.

O'Keefe (1980) first proposed that Earth has a ring-driven climate. The key challenge has been the consensus opinion that the Moon is dead. This is an obstacle because for dynamical reasons only the Moon can be attributed as the source of material for a persistent Earth ring system. Recently, however, the Dead Moon consensus is weakening following analysis of Lunar Reconnaissance Orbiter data by Watters et al. in 2012, and Braden et al. in 2014.

Some have asked, why has an Earth ring system not been observed? After all, if a ring system shades Earth significantly then it is interacting with visible light. This is a very valuable consideration but not insuperable. Available data are ambiguous and do not rule out a ring system. We will not address that challenge here except to say that we have reviewed CWOP solar radiation data, a half-billion observations contributed by thousands of volunteer weather stations from 2009 forward in a global distribution. The solar radiation dataset is not inconsistent with the ring hypothesis (Hancock and Chadwick 2016, https://ams.confex.com/ams/96Annual/webprogram/Session38896.html).

In this presentation we hypothesize that Earth has a two-ring system: a ring in the plane of the Moon's orbit (an accretion disk mediating mass exchange) and a ring in the plane of Earth's equator (a derivative structure whose development would be favored by dynamical considerations). The proposed system is the most general possibility in the sense that all known planetary rings occupy an analogue of one of these two orientations.

We further hypothesize that ENSO comprises all the global effects due to the cycle in configuration as the two rings swing in and out of phase. Two rings like those hypothesized would cycle in and out of phase at the period of the precession of lunar nodes (18.6 years). Of note, they would have been in phase in 1997 and 2015, out of phase in 2007. The first order effect of this cycle would be a cycle in Earth shading: in phase, the system would be least effective at deflecting solar radiation; out of phase, most effective. A visualization of Earth wrapped in such a ring system is presented at 1997 and for comparison at 2007 (from the point of view of the Sun). These schematics illustrate the difference in shading.

This cycle would be modulated by the varying optical depth of the ring system, which would be modulated by solar activity because solar radiation pressure is important for small particles.

To test this hypothesis, a predictive model of the multivariate ENSO index (https://www.esrl.noaa.gov/psd/enso/mei/) is developed from time series of (a) the phase of the lunar precession of nodes, (b) phase of the solar year, (c) phase of the eclipse year, (d) sunspot number as a proxy for solar activity that is expected to thin the entire structure when strong; and (e) MEI of two months prior, introduced as a proxy for climate persistence. The model is developed using naive Bayes techniques. It predicts which of four bands MEI will fall in at each monthly timepoint.

A typical train/test run of the model (80%/20%) over the 800+ monthly values in these time series 1950 - 2016 leads to the following typical truth table for the test data.

	True category of MEI
Predicted category of MEI	(-5,-1]	(-1,0]	(0,1]	(1,5]
(-5,-1]	24	6	0	0
(-1,0]	2	42	6	0
(0,1]	0	10	44	7
(1,5]	0	0	5	18

The predictive model appears to have skill. Can it indeed forecast? Adapting the predictive model for forecasting requires filling in somehow for the two-months-prior MEI at each time point. To be precise, we require to know in which of four value bands the prior MEI fell. Because we do not know this, we undertake a kind of “ensemble” forecast: we forecast each point four times, once for each of the four possible band values of the two-months-prior MEI. The Figure below presents this "ensemble" forecast to end 2019 obtained this way. Each time point has four forecasts attached to it:

• a red circle presents the forecast obtained using for the two-months-prior MEI a band value between 1 and 5 (an El Nino signal);
• a pink circle presents the forecast obtained using for the two-months-prior MEI a band value 1 and 0 (a warm-neutral signal),
• a green circle presents the forecast based on two-months-prior MEI between 0 and -1 (cool-neutral),
• a blue circle presents the forecast for two-months-prior MEI less than -1 (a La Nina signal).

In this figure the predicted value of MEI is given at the vertical axis. As there are only four possible predicted values, the predictions are grouped in four horizontal bands. The uppermost band comprises all predictions of high MEI values (El Nino events would populate this); the lowest band is predictions of low MEI values, and the two between are warm and cold neutral.

Statistical analysis of these results model would present a number of challenges that we are not going to take up here. What we can say is that If the calculated truth table is representative, then this “ensemble” forecast suggests that a strong persistent El Nino is unlikely until late 2019.

The presentation will update the forecast through 2020 and take the hindcast back to 1750 for comparison with past events.

The code will be available at https://github.com/lohancock/enso-forecast