JLI First Forecasts

Walter Dnes – Click the pic to view at source

Image Credit: Walter Dnes

By Walter Dnes – Edited by Just The Facts

In a recent post I introduced The “January Leading Indicator“. At that time, there were some concerns expressed. The attached spreadsheet has been updated to address items 2 and 3 below.

1) The JLI algorithm is not a “forecast” per se, but rather a “zero skill baseline” that a “real forecast” has to beat in order to show skill. Given that the algorithm is a “rule-of-thumb”, rather than a complex physical model, I have no problem with that description. The whole point of the JLI post was to raise the bar for “zero skill”. Beating 50% is not enough. A “real forecast” has to beat 75% or 80%.

2) The period of data (1979 to 2013) was too short to generate statistically significant results. 1979 marks the start of the satellite data sets (UAH5.6 and RSS) so additional data is not possible for them.

The surface-based data sets go back a lot further. An initial examination shows data back to 1955 to have a reasonable correlation. Sea-surface anomalies count for more than half of the global anomaly. There were significant adjustments applied to sea surface temperatures prior to 1950, due to claimed changes in bucket type, among other reasons. In addition, the complete and accurate collection of global temperature data was not always possible during, and in the immediate aftermath of, a global war.

The updated spreadsheet now has surface data back to 1955.

3) Including the January data in the forecast, after it is available is an unfair advantage. The updated spreadsheet includes 2 additional tabs that compare the current year’s February-through-December anomalies versus the previous year’s February-through-December anomalies. The tab “count_gt11″ does so where a year’s January anomaly is greater (warmer) than the previous year’s anomaly. The tab “count_lt11″ does so where a year’s January anomaly is less (cooler) than the previous year’s anomaly. Here is a comparison between the 12-month (January-December) verification, versus the 11-month (February-December) verification.

1) 12-month (January-December) verification where forecast to be warmer

Data set Had3 Had4 GISS UAH5.6 RSS NOAA
Jan > previous 30 29 32 21 20 29
Ann > previous 24 23 27 18 18 24
Accuracy 0.80 0.79 0.84 0.86 0.90 0.83

2) 12-month (January-December) verification where forecast to be cooler

Data set Had3 Had4 GISS UAH5.6 RSS NOAA
Jan < previous 28 29 26 13 14 29
Ann < previous 19 20 20 11 11 19
Accuracy 0.68 0.69 0.77 0.85 0.79 0.66

3) 11-month (February-December) verification where forecast to be warmer

Data set Had3 Had4 GISS UAH5.6 RSS NOAA
Jan > previous 30 29 32 21 20 29
11 Mo > previous 23 22 25 17 13 23
Accuracy 0.77 0.76 0.78 0.81 0.65 0.79

4) 11-month (February-December) verification where forecast to be cooler

Data set Had3 Had4 GISS UAH5.6 RSS NOAA
Jan < previous 28 29 26 13 14 29
11 Mo < previous 18 18 18 10 10 19
Accuracy 0.64 0.62 0.69 0.77 0.71 0.66

The January data for UAH5.6 and RSS are in. Now for the forecasts. The numbers are from the attached updated spreadsheet.

The Qualitative Forecasts

The January 2014 UAH5.6 monthly anomaly was 0.291 versus 0.497 in January 2013. The JLI algorithm states that the 2014 annual mean should be less than the 2013 annual mean of 0.236. Since we already have the January mean, this is effectively an 11-month forecast. I.e. the mean of the 11 months February 2014 to December 2014 will be less than…

( 12 * 0.236 ) - 0.291
----------------------
11

or less than 0.231.

On the graph at the head of this article the red “X” is the forecast anomaly for UAH.

The January 2014 RSS monthly anomaly was 0.262 versus 0.439 in January 2013. The JLI algorithm states that the 2014 annual mean should be less than the 2013 annual mean of 0.218. Since we already have the January mean, this is effectively an 11-month forecast. I.e. the mean of the 11 months February 2014 to December 2014 will be less than…

( 12 * 0.218 ) - 0.262
----------------------
11

or less than 0.214.

On the graph below the red “X” is the forecast anomaly for RSS:

Walter Dnes – Click the pic to view at source

The Quantitative Forecasts

Tab “jan_and_avg_2″ of the spreadsheet has some statistics in the block P1:V4, comparing the January anomalies with the annual anomalies. The UAH5.6 slope in cell T3 is 0.64062. The intercept in cell T4 is 0.01732. Applying the standard “y = mx + b” equation, we get a predicted 2014 annual anomaly of 0.64062 * 0.291 + 0.01732 = 0.204 with an unknown error margin.

Similarly, for RSS the slope in cell U3 is 0.64755 and the intercept in cell U4 is 0.03456. The predicted 2014 annual anomaly is 0.64755 * 0.262 + 0.03456 = 0.204 with an unknown error margin.

This is in contrast to the current ENSO forecast for the Nino3.4 region (shown below), which implies an El Nino event, which should result in a warmer year.
NINO 3.4 SST Anomalies Forecast

National Oceanic & Atmospheric Administration (NOAA) – Climate Prediction Center – Click the pic to view at source


Additional Forecasts for Comparative Purposes:

Met Office 2014 Prediction
19 December 2013 – “The global average temperature in 2014 is expected to be between 0.43 C and 0.71 C above the long-term (1961-1990) average of 14.0 C, with a central estimate of 0.57 C, according to the Met Office annual global temperature forecast. Taking into account the range of uncertainty in the forecast, it is likely that 2014 will be one of the warmest ten years in the record which goes back to 1880. The forecast range and central estimate for 2014 are the same as were forecast by the Met Office for 2013.”

“* Observationally based estimates of global average temperature are an average of the three main global temperature datasets, which are compiled by the Met Office and University of East Anglia (HadCRUT4), NOAA National Climatic Data Center (NOAA NCDC) and NASA Goddard Institute of Space Studies (NASA GISS).”

Hansen et al., 2014 prediction:

“So what are the near-term prospects? El Niño depends on fickle wind anomalies for initiation, so predictions are inherently difficult, but conditions are ripe for El Niño initiation in 2014. About half of the climate models catalogued by the International Research Institute predict that the next El Ni ño will begin by summer 2014, with the other half predicting ENSO neutral conditions 21. The mean NCEP forecast 21 issued 13 January has an El Niño beginning in the summer of 2014, although a significant minority of the ensemble members predicts ENSO neutral conditions for 2014.

The strength of an El Niño, too, depends on the fickle wind anomalies at the time of initiation. We speculated 22 that the likelihood of “super El Niños, such as those in 1982 – 3 and 1997 -
8, has increased. Our rationale was that global warming increased SSTs in the Western Pacific, without yet having much 13 effect on the temperature of upwelling deep water in the Eastern Pacific (Fig. 2 above), thus allowing the possibility of a larger swing of Eastern Pacific temperature. Recent paleoclimate 23 and modeling 24 studies find evidence for an increased frequency of extreme El Niños with global warming.

Assuming that an El Niño begins in summer 2014, 2014 is likely to be warmer than 2013 and perhaps the warmest year in the instrumental record. However, given the lag between El Niño initiation and global temperature, 2015 is likely to have a temperature even higher than in 2014.”

About these ads

34 thoughts on “JLI First Forecasts

  1. Note the colors on the ENSO forecasts ensembles. The most recent 8 forecasts are trending down from the earlier ones (as is the current ENSO value if you look at other charts on the WUWT ENSO page).

  2. Werner Brozek says:
    > February 12, 2014 at 4:11 pm
    > Should tables 2 and 4 have “less than” signs “<" for cooler?

    Ouch. Again. That's what I get for being lazy and copy+pasting. Mods, please fix that up

  3. walterdnes says: February 12, 2014 at 4:13 pm

    Werner Brozek says: February 12, 2014 at 4:11 pm

    Should tables 2 and 4 have “less than” signs “<" for cooler?

    Mods, please fix that up

    Corrected

  4. 4) 11-month (January-December) verification where forecast to be cooler

    Should that be February-December as for 3)?

    As well, for 3) and 4) it does not seem correct to use “Jan > or or <” or something like that?

  5. TRG says:
    February 12, 2014 at 4:58 pm
    I’m sorry, but just stop wasting people’s time with this.
    Do you prefer the Met forecasts?

    Met Office 2014 Prediction
    19 December 2013 – “The global average temperature in 2014 is expected to be between 0.43 C and 0.71 C above…

    That is a range of 0.28. To put it another way, the difference in rankings between #1 and #19 is only 0.253 on Hadcrut4. If they want to make the “barn door” that wide, how can they miss? On top of that, they are only 90% confident of this. As well, their forecasts are to be used for:
    “It also has a broad range of potential applications
    in terms of policy making and investment decisions.”

    How useful is that?

    http://www.metoffice.gov.uk/media/pdf/1/8/decadal_forecast_2014-2018_jan2014.pdf

  6. Sorry! My previous copy and paste did not copy and paste well. It should have said:

    As well, for 3) and 4) it does not seem correct to use “Jan > or or <” or something like that?

  7. In this post I find:
    This is in contrast to the current ENSO forecast for the Nino3.4 region (shown below), which implies an El Nino event, which should result in a warmer year.

    In the NOAA [cpc.ncep.noaa.gov] report of 10 Feb 2014, I find:
    ENSO-neutral is expected to continue through the Northern Hemisphere spring
    2014.

    This implies that spring 2014 might be followed by El Niño starting in Apr-June (AMJ) 2014. A “warmer year” might also follow, but “should” in this case is 2 iffy steps away from “will.”

  8. The lag between ENSO values and global temperatures is about 3-6 months. Given the current negative value that should hit in spring or early summer. However, it could also be the stimulus to fuel an El Nino in the fall. Never a dull moment. ;)

  9. Remember what happened after Hansen’s “Super El Nino” prediction in 2006? And NOAA’s ENSO forecasts are a joke too.

  10. So, is the JLI forecasting (or projecting, or whatever) that 2014 won’t be one of the ten warmest years? That would be nice. It would nibble away at one warmist talking point.

  11. I think JLI is just a variant of the idea that, if you don’t have anything better, the best forecast for this year’s average is last year’s average. If you cut down the sampling period to, say, last December, then you may slightly improve the mean of the forecasts, because it is more current, but you’ll greatly increase the variance. January would be much the same, but with some (unfair) arithmetic advantage, in that Jan is already included in the average being predicted.

    People trying to predict stock prices often find that the best predictor of tomorrow’s price is today’s price. You can try a fancier formula including rate of change etc. It may be on average slightly better, but will be more variable.

  12. rogerknights says:
    February 12, 2014 at 9:35 pm
    So, is the JLI forecasting (or projecting, or whatever) that 2014 won’t be one of the ten warmest years?

    Unfortunately, WFT only has UAH version 5.5 and there the ranking was 7th last year and this year should be lower. However I have not seen the January 2014 number yet and this article only talks about version 5.6. According to the value of 0.204, it would be tied for 7th on version 5.6. As for RSS, the value of 0.204 would put it in 11th place for 2014. So the pause or whatever you wish to call it will be extended another year in 12 months from now if projections hold.

  13. Just the facts
    You can analyze this as a statistical sampling exercise. Each year is a whole population N = 12. Your sample must be included in the population (January). The sample size is n = 1. Is this enough to represent the whole population accurately? It depends on the variability of the populations. You have two populations: forecast year and previous year. This explains why different years have different forecast accuracy.

    The question of statistical significance is not really related to the no. of years. What is being tested is the probability of the observed results happening purely by chance. You can do this by Monte Carlo simulation and fitting a normal curve to find the deviation from the mean. My guess is you will get between one to two sigma, which correspond to P = 0.32 to 0.05. Not strictly statistically significant but far better than random guesses. BTW you can increase the accuracy by increasing the sample size.

  14. Sorry mate, there ain’t no significant CO2-AGW and the real AGW, from Asian aerosols reducing cloud albedo, has ceased changing so no more SW energy excess in the oceans.

    We are into global cooling as shown by the split in the Polar Vortex and the warming of the mid-winter Arctic by the major Atlantic storms.

  15. This is in contrast to the current ENSO forecast for the Nino3.4 region (shown below), which implies an El Nino event, which should result in a warmer year.

    That ENSO forecast ensemble has no skill. It has been wrong for over a year. There is nothing in the current trade wind anomalies that would indicate any El Nino condition and the Western Pacific Warm Pool is cool. Even if the trades slacked, there is no pool of warm water to slosh east and generate a strong El Nino condition.

    I have no idea who is running those models but they are useless.

  16. To elaborate on my previous comment: First go to the ENSO page on this site here: http://wattsupwiththat.com/reference-pages/climatic-phenomena-pages/enso/

    Scroll down to the graphic labeled UNISYS Current Sea Surface Temperature Anomaly Plot and have a gander at the area around Indochina. You will see the sea surface there is anomalously cool, not warm. Now also notice the rather pronounced cold tongue sticking out into the equatorial Pacific from South America. Now scroll down to Equatorial Pacific Sea Surface Temperature & Anomaly With Wind – Current and you see the wind anomalies. Notice that the only eastward anomaly is in the far western part of the graphic.

    There is nothing to give any indication of any impending El Nino conditions. The trade winds aren’t right, and even if they slackened, there would be no pool of warm water in the Western Pacific.

  17. wbrozek says:
    February 12, 2014 at 9:55 pm

    According to the value of 0.204, it would be tied for 7th on version 5.6. As for RSS, the value of 0.204 would put it in 11th place for 2014.

    My magic eight-ball says it will be the 13th warmest.

  18. Find myself agreeing with Nick Stokes on this one. How do other months perform for the subsequent 11 months (after adjusting for the length of the month)?

  19. michael hart says:
    > February 13, 2014 at 4:23 am

    > Find myself agreeing with Nick Stokes on this one.
    > How do other months perform for the subsequent 11
    > months (after adjusting for the length of the month)?

    There is no adjustment for length-of-month, but see table 3 versus table 1 and table 4 versus table 2. Tables 3 and 4 show how often the 11-month period Feb-Dec is higher or lower than previous year, compared with January as the leading indicator. I’ve concentrated on the 12-month period, because it’s the “headline number” that’s always trumpeted in the media, including the blogosphere.

  20. What the spaghetti graph is showing is that the models all agree on one thing. Natural variability is high. Otherwise the forecasts would not be all over the place.

  21. Just The Facts – you need to get a global idea of who you are talking to – it is already high summer where I live – so has El Nino already started?

    (just some advice – not criticism :-) )

  22. Douglas Jones says: @ February 13, 2014 at 6:47 am
    …. so has El Nino already started?
    >>>>>>>>>>>>>>>>>>>
    No. WUWT ENSO METER

    NOAA on ENSO

    El Niño and La Niña episodes typically last nine to 12 months, but some prolonged events may last for years. They often begin to form between June and August, reach peak strength between December and April, and then decay between May and July of the following year. While their periodicity can be quite irregular, El Niño and La Niña events occur about every three to five years. Typically, El Niño occurs more frequently than La Niña.

    El Niño
    El Niño means The Little Boy, or Christ Child in Spanish. El Niño was originally recognized by fishermen off the coast of South America in the 1600s, with the appearance of unusually warm water in the Pacific Ocean. The name was chosen based on the time of year (around December) during which these warm waters events tended to occur.

    The term El Niño refers to the large-scale ocean-atmosphere climate interaction linked to a periodic warming in sea surface temperatures across the central and east-central Equatorial Pacific.

    Typical El Niño effects are likely to develop over North America during the upcoming winter… season….

    Walter Dnes is trying to determine whether there are indications in January as to what will happen BEFORE El Niño and La Niña episodes begin to form between June and August.

    Northern vs Southern Hemispheres have nothing to do with it. He is looking for a ‘Leading Indicator’

    If my animals start growing heavy winter coats in mid summer can that indicate a harsh winter to come? Maybe – if there is a link between certain wavebands of sunlight and animal behavior and similar wavebands of sunlight and ozone/cloud formation/ ENSO.

  23. The Hadley Centre Central England Temperature Record (which, per my reply to Dnes’ first post, showed a post 1979 prediction accuracy of 53% for higher temperatures and 59% for lower temperatures) shows a January 2014 average of 5.7 C (up from 3.5 C last year) so there is a 53% probability that the average for 2014 will be higher than last year’s 12 month average of 9.56 C.

  24. Richard Mallett says:
    > February 13, 2014 at 12:10 pm
    > The Hadley Centre Central England Temperature Record
    > (which, per my reply to Dnes’ first post, showed a post
    > 1979 prediction accuracy of 53% for higher temperatures
    > and 59% for lower temperatures) shows a January 2014
    > average of 5.7 C (up from 3.5 C last year) so there is a 53%
    > probability that the average for 2014 will be higher than
    > last year’s 12 month average of 9.56 C.

    If I had only found 53% or even 59% correlation, I wouldn’t have bothered with an article. Correlations of 66% to 90% are more impressive. BTW, I’ve plotted CET, and it’s graph bears no relation to global anomalies that I can figure out. I can easily accept it going up whilst global anomalies go down. Another item is that it has zero or negative slope from September 1987 (Yes, nineteen eighty-seven) through January 2014.

Comments are closed.