Guest Post by Willis Eschenbach
In my previous post, “Does This Analysis Make My Tropics Look Big?“I discussed a paper called “Recent Northern Hemisphere tropical expansion primarily driven by black carbon and tropospheric ozone”, by Robert Allen et al, hereinafter A2012. They use several metrics to measure the width of the tropics—the location of the jet stream (JET), the mean meridional circulation (MMC), the minimum precipitation (PMIN), the cloud cover minimum (CMIN), and the precipitation minus evaporation (P-E) balance. Since writing that post, I’ve looked at what the Argo dataset says about the P-E balance, which is the precipitation minus the evaporation. Figure 1 shows the global Argo results regarding the salt concentration (salinity) in the surface of the ocean, which is a proxy for precipitation minus evaporation.
Figure 1. Global average salinity, as revealed by the Argo floats, in practical salinity units (PSU).
We can use the salinity of the ocean in the tropical and temperate regions as a very good proxy for the balance of precipitation and evaporation. In the deep meteorological tropics, just north of the equator, is the Intertropical Convergence Zone, or the ITCZ. In the ITCZ, the precipitation predominates. As a result, in that area there is more fresh-water rain than evaporation, and that makes the ocean less salty (blue).
On the other hand, at about 30° north and south of the Equator are the descending dry air branches of the Hadley cells. These create the global desert belts. Figure 2 shows a cross-section through the atmosphere illustrating the circulation of the great Hadley cells:
Figure 2. The descending branch of the Hadley cells is dry because the water has been rained out in the deep tropics. In the Northern Hemisphere this dry descending air creates the arid belts of the Sonoran desert in Northern Mexico / Southwest US as well as the Sahara, Middle Eastern, and Gobi regions. In the Southern Hemisphere, it encompasses the Atacama (South America), Kalahari (Africa), and Australian deserts.
In these arid regions, the evaporation is much greater than the precipitation and as a result the ocean is saltier in these areas. And once again this is reflected in the salinity, specifically in the areas of high salinity (red) around thirty degrees north and south of the Equator, as shown in Figure 1.
Now let me refresh people’s memory regarding the claims of the A2012 paper. Figure 3 shows their Northern Hemisphere results discussed in my previous paper:
FIGURE 3. ORIGINAL CAPTION: Figure 2 | Observed and modelled 1979–1999 Northern Hemisphere tropical expansion based on five metrics. a, Annual mean poleward displacement of each metric, as well as the combined ALL metric. … CMIP3 models are grouped into nine that included time-varying black carbon and ozone (red); three that included time-varying ozone only (green); and six that included neither time-varying black carbon nor ozone (blue). Boxes show the mean response within each group (centre line) and its 2s uncertainty. Observations are in black. In the case of one observational data set, trend uncertainty (whiskers) is estimated as the 95% confidence level according to a standard t-test.
I was interested in the P-E record (precipitation minus evaporation). The P-E results in the A2012 paper (Figure 3 above) show a net change of 0.75 degrees of latitude in twenty years in the latitude of the Northern Hemisphere maximum salinity, or about 0.36 degrees per decade. Southern Hemisphere P-E results (not shown) are about half that size, at 0.17 degrees per decade.
So I took a look year by year in the various oceanic basins to examine the natural variability in the salinity, which is our best proxy for P-E. Here are the results for the Pacific Ocean (120°E to 100°W longitude) from the Argo data. Figure 4 shows the variability, along with the decadal observational changes reported in A2012.
Figure 4. Year by year changes in the latitudinal salinity in the Pacific. Circles show the annual Southern Hemisphere peak salinities at about 30°S, the Equatorial lows in salinity, and the annual peaks in the Northern Hemisphere salinity at about 30°N. Distance between the two vertical black lines in the upper right illustrates the amount of the decadal Northern Hemisphere tropical expansion claimed in A2012 (0.38°/decade). Two vertical black lines (so close they appear as one line) in the upper left illustrates the amount of the decadal Southern Hemisphere tropical expansion claimed in A2012 (0.16°/decade). There is insufficient data in the years 2002-2003 to plot the Southern Hemisphere peaks.
There are a few things of interest here. First, in the Pacific the location of the minimum salinity at the ITCZ just north of the Equator is relatively stable. The location of the Pacific ITCZ doesn’t drift north or south too much. The Southern Hemisphere peak is more variable in latitude, and the Northern Hemisphere is more variable yet. All of them move around much, much more than the amount of the estimated decadal year trend. Bear in mind that according to A2012 this tropical expansion is driven by black carbon and ozone … note also the relative sizes of the expansion claimed in A2012, shown by the vertical black lines. [As an aside, I was surprised by the difference in the widths of the northern and southern Tropics, with the southern Tropics being twice as wide as the northern Tropics. It suggests that the outer edges of the tropics, the areas of peak salinity, are controlled by physical rather than meteorological considerations … but I digress.]
Figure 5 shows the corresponding data for the Indian Ocean (20°E to 120°E longitude). The Indian Ocean doesn’t go far enough north to experience a minimum, so the Southern Hemisphere peak and the low at the ITCZ are shown.
Figure 5. Year by year changes in the latitudinal salinity in the Indian Ocean. Circles show the annual Southern Hemisphere peak salinities at about 30°S, and the Equatorial lows in salinity. Two vertical black lines (so close they appear as one line) in the upper left illustrates the amount of the decadal Southern Hemisphere tropical expansion claimed in A2012 (0.16°/decade).
In the Indian Ocean, the situation is reversed. Unlike in the Pacific, the peak salinity is stable, but the location of the low salinity at the ITCZ is greatly variable. In addition, the ITCZ appears to have moved generally southwards over the decade.
Finally, the Atlantic Ocean takes the tropical variability prize, as shown in Figure 6.
Figure 6. Year by year changes in the latitudinal salinity in the Atlantic. Circles show the annual Southern Hemisphere peak salinities at about 30°S, the Equatorial lows in salinity, and the peaks in the Northern Hemisphere salinity at about 30°N. Distance between the two vertical black lines in the upper right illustrates the amount of the decadal Northern Hemisphere tropical expansion claimed in A2012 (0.38°/decade). Two vertical black lines (so close they appear as one line) in the upper left illustrates the amount of the decadal Southern Hemisphere tropical expansion claimed in A2012 (0.16°/decade).
As mentioned above, the Atlantic results are all over the map, with all areas varying greatly in both salinity and distance from the Equator.
The obvious conclusion that I draw from all of this is that a trend can be significant without being meaningful. The trends in tropical expansion shown in the A2012 paper are tiny compared to the natural variations in the system. The width of the meteorological tropics varies up to eight degrees in a single year. In such a system, a few tenths of a degree of expansion per decade, even if it turns out to be both accurate and statistically significant, is trivially small.
Finally, I wanted to investigate the relationship between temperature rise and precipitation. So I took a look, for each individual Argo float, at the differences in temperature and salinity (again as a proxy for rainfall minus evaporation) in successive cycles of each float. I then plotted the ratio of the changes globally. Figure 7 shows that result:
Figure 7. Change in rainfall with temperature, as indicated by the proxy of the change in salinity with temperature. Blue areas are where the rainfall increases as the temperature increases, and red areas are where the rainfall goes down as temperatures rise.
This was an interesting result, as it shows a more complex and nuanced pattern than the usual mantra of “a warmer world is a wetter world”. It is also interesting in that there is only a small relationship, albeit statistically significant, between salinity and temperature. For each degree of temperature rise, the salinity goes up by only 0.04 PSU, a tiny amount (although the p-value is 2e-16).
Finally, here’s the strange part. Averaged over the entire globe, since salinity goes up with temperature, globally the Argo data says precipitation goes down fractionally with increasing temperature. In the tropics, the relationship is as expected, rainfall increasing with temperature. But globally, it goes the other way, rainfall decreases with increasing temperature … and there is only a minuscule effect. I didn’t expect that at all. [UPDATE—see below for why I didn’t expect it]
That’s the beauty of climate science being settled … there are so many surprising results.
w.
[UPDATE] Global estimates of the water cycle are on the general order of this one:
Thanks to commenters in the thread below, I see now that my expectation of the direction of change in the global oceanic P-E with increasing temperature was mistaken. The basic equation for the ocean mass balance says that precipitation plus additions from the rivers (including ice melt and groundwater extraction) minus evaporation from the ocean gives mass balance change.
Now, the ice melt and groundwater extraction don’t vary much year to year. So if the ocean mass is roughly constant, we can take the basic equation as being evaporation from the ocean equals rain into the ocean plus net rain over land … what goes up must come down.
My mistake was in thinking that the actual value of the oceanic P-E overall was positive. It is not. From the data given in the table above, we can see that P-E is about – 35,000 cubic kilometres per year.
And that means that if we increase the speed of the hydrological cycle by say 10%, and we assume that all of the proportions remain the same, the value of P-E becomes more negative, not more positive as I had assumed.
In other words, I should have expected that if the temperature increased the value of P-E should go down, not up as I thought. In that regard, it appears that my finding, that oceanic P-E decreases with increasing temperature, is consonant with expectations.
Always more to learn …
timetochooseagain says: May 25, 2012 at 8:48 pm
I ran the numbers adding the NH and abs(SH) together. Here are the decade trends:
1911-2010 equatorward 0.050 +/- 0.038
1951-2010 poleward 0.078 +/- 0.073
Richard, I think my non science background caused me to misuse the concept of error bands.
Nonetheless there are observed changes in the positioning which appear to fit changes in historical climate shifts as per my post at May 25, 2012 at 11:59 pm
Philip,
Yes it is proposed that GHGs raise the temperature of the ocean skin thereby reducing the temperature differential from ocean bulk to ocean skin which is supposed to reduce the energy flow from ocean to air through the skin thereby making the bulk oceans warmer and ultimately heating up the air as well.
However I have elsewhere spent a lot of effort explaining why that proposition doesn’t work due to the net cooling effect of the evaporative process.
Try this:
http://climaterealists.com/attachments/ftp/TheSettingAndMaintainingOfEarth.pdf
AJ,
In the period 1911 to 2010 you have a warming spell followed by cooling, then warming then a short period of cooling again so the net overall would be very small.
Could you run figures for the following periods ? :
1911 to 1940
1940 to 1970
1970 to 2000
2000 to 2010.
AJ-Thank you. It looks like the uncertainty is larger. Hm, it might help with understanding why the trend is sensitive to endpoints if one took running averages. Eleven year centered averages should smooth over all the inter-annual noise. Then you could probably see when the trends change signs. Stephen Wilde seems to be hypothesizing that the changes should roughly coincide with the known shifts from warming to cooling trends and vice versa in the temperature record.
I’ll try to run some more numbers in the next couple of days. I believe the uncertainty is larger when I consider the uncertainty in the source dataset, which is also interpolated.
“Stephen Wilde seems to be hypothesizing that the changes should roughly coincide with the known shifts from warming to cooling trends and vice versa in the temperature record.”
Correct. If not then either I would have to do some serious rethinking or, possibly, current data is inadequate for purpose.
So:
1911 to 1940 should show a poleward drift.
!940 to 1970 should show an equatorward drift.
1970 to 2000 should show a poleward drift
2000 to 2010 might be too short a period to show an equatorward drift but it should be at least neutral.
Go to it AJ.
Well, I found some free time. Here are the requested decadal trends for the NH:
1911 to 1940 -0.30 +/- 0.14
1940 to 1970 0.11 +/- 0.15
1970 to 2000 -0.04 +/- 0.15
1999 to 2009 0.07 +/- 0.42
2010 was an outlier which resulted in a negative trend, so I swapped it for 1999.
Thanks AJ.
You confirm that 1911 to 1940 and 1970 to 2000 are both negative and both were warming periods so I assume that negative means poleward. Is that right ?
Stephen Wilde says:
May 25, 2012 at 11:24 pm
Many thanks, Stephen, and also timetochooseagain. The oceans may not be a good enough diagnostic indicator, and it may well be that timetochooseagain is right that over the planet there is a “positive relationship with temperature”.
My point is simple—for 80% of the rainfall, falling on 70% of the planet, the trend is slightly negative, not positive at all. That may be offset over the land, we don’t know.
But for it to just get back to zero, the P-E balance over the land would have to be a) in the reverse sense and b) four times as large as it is over the ocean …
I don’t know if that’s the case, but it would strike me as odd if it were to be true.
All the best,
w.
“P-E balance over the land would have to be a) in the reverse sense and b) four times as large as it is over the ocean …”
Well, since there is very little evaporation over land and lots of orographic rainfall over high ground I don’t think that is hard to believe at all 🙂
SW… no, in the NH negative numbers are equatorward.
Stephen Wilde says:
May 26, 2012 at 12:44 pm
Thanks as always for your thoughts, Stephen. However, most of the precipitation is not falling on the high mountains, but in the wet tropics, where there is lots and lots of evaporation. Take a look at the figure I showed above. When it is area adjusted, it shows that over half of the rainfall on land (by volume) is in the tropics, between ± 25° N/S.
It strikes me that we should be able to compare the average rainfall over land (which is about 200 mm) with the total flow of all of the rivers to get a first-order estimate of the P-E balance over the land … too many ideas, too little time, but it’s the weekend, I’ll see what that looks like.
w.
Willis Eschenbach-“My point is simple—for 80% of the rainfall, falling on 70% of the planet, the trend is slightly negative, not positive at all.”
I don’t think you understood my point. The salinity is not a proxy for the rainfall, but the Precipitation minus the Evaporation. What I am saying is that globally, over the oceans, you have found that the precipitation minus evaporation response to temperature is about zero. I am saying that this is not the same thing as there being no positive relationship between precipitation and temperature change over the oceans. I am saying that precipitation and evaporation probably globally increase by the same amounts with an increase in temperature, thus the P-E relationship with temperature is indeed about zero, but it is not necessarily true, as you seem to imply, that there is no relationship with precipitation and evaporation separately. One needs data that separates the precipitation and evaporation effects to test if this is true. Salinity combines both effects.
Also, Steven, I’m talking not about the direct P-E balance, but the CHANGE in the P-E balance with temperature. Over the ocean the net precipitation minus evaporation goes down slightly with temperature.
So it would have to the other way around on the land, we’d have to get more precip and less evaporation as temperatures rise … that’s what I was saying was doubtful.
Regards,
w.
timetochooseagain says:
May 26, 2012 at 2:00 pm
Sorry for my lack of clarity, timetochooseagain. I am of course referring to P-E. However, your point is well taken.
What you say is not exactly true however. You say that I “have found that the precipitation minus evaporation response to temperature is about zero.” In fact, I have found a small but statistically very significant trend in salinity (as a proxy for P-E) with respect to temperature. As the earth warms the salinity of the oceans increases, and as the earth cools the salinity of the oceans decreases. The trend is small, to be sure, but not zero.
Of more interest to me is the geographical distribution of the changes. In the tropics, precipitation outpaces evaporation as the earth warms, but in the extra-tropics, the relationship is reversed.
w.
I can’t let this topic slide by without mentioning Joe D’Aleo posted a neat article on the Weatherbell site about recent harsh winters in South America, including a snippet about a blast of Antarctic air that ducked behind the Andes Mountains and roared north right through the Amazon and past the northern tip of Peru, to a point past the equator. Joe D’Aleo wrote, ” Temperature even fell in Roraima, where the state capital Boa Vista recorded 20C (normal lows are 25C) and the wind were blowing from the South. Boa Vista is located at 2º North of latitude, so the influence of the Antarctic cold blast crossed the Equator line and reached towns in the Northern Hemisphere.”
How wide are the tropics, you ask? Not very wide, that day.
“SW… no, in the NH negative numbers are equatorward.”
Are you sure ?
So what are the global numbers ?
Do they support warmer/poleward and cooler/equatorward or not ?
Even in the NH we saw more poleward / zonal jets during the late 20th century and during the MWP as compared to the LIA so I’m puzzled that your numbers seem to show equatorward during warmer periods in the NH.
And previously you said this:
“In the Northern Hemisphere for the years 1979-1999 I found a poleward 0.18 decade rate”
Yet you now say that for the years 1970 to 1999 it was equatorward -0.04.
I suppose that it could be that cooling continued until around 1979 and converted the sign but that would still leave us with the odd proposition that the mid century cooling period gave a poleward shift in the NH according to you but we can all see that the mid latitude jets have been more equatorward recently than during the late 20th century. Also, recent equatorward jets are more akin to what we saw during the mid century cooling period.
It may be that the fineness of the system response makes it essential to determine precisely when the inflection points in the temperature record actually occurred and then work from them rather than on broad decadal periods which do not reflect the actual observations as regards temperature trends..
I think we are onto something here because I have been long decrying the absence of good enough data about latitudinal climate zone shifts to make adequate comparisons with changes in tropospheric temperature trends.
At least you are making a start even if we are having problems refining it.
Observations clearly support a significant latitudinal shift from MWP to LIA and LIA to date with poleward for the MWP and equatorward for the LIA which is opposite to what you are now saying.
“on the land, we’d have to get more precip and less evaporation as temperatures rise ”
Warmer oceanic winds carrying more water vapour would do just that on reaching land. The amounts of rainfall dumped on mountain barriers are huge.More clouds would reduce evaporation from the land too but evaporation from dry continental interiors is minimal anyway.
Willis Eschenbach says: “In fact, I have found a small but statistically very significant trend in salinity (as a proxy for P-E) with respect to temperature.”
I am sorry, I thought I had read you saying it was statistically insignificant in the post, I guess I misread that sentence, since ‘albeit statistically signficant’ is a bit of a strange phrase, IMAO.
I regard it as an interesting question to ask what the relationship between precipitation itself with temperature changes is. Do you think RSS’s SSM/I data could be used to examine that question?
http://www.remss.com/ssmi/ssmi_browse.html
Willis said:
“In the tropics, precipitation outpaces evaporation as the earth warms, but in the extra-tropics, the relationship is reversed.”
That must be right because the water vapour fuelling the tropical uplift flows in from the extratropical regions either side of the ITCZ and I think if the extratropics is measured correctly (land plus the higher latitudes) there will be a balance. After all, that is the essence of your Thermostat Hypothesis which is fine as far as it goes and there are indications that global humidity remains approximtely stable whether the system is cooling or warming so there can’t be much of a warming induced drying process overall over time.
One could propose that tropical rainfall increases, subtropical rainfall decreases and higher latitude rainfall stays much the same. That would also fit your Fig 7 would it not ? There are hints of blue around the poleward perimeters and that could be understated by current data.
My main suggestion is that your hypothesis needs to be extended globally and needs to incorporate latitudinal climate zone shifting especially as regards the size of the subtropical high pressure cells which would be highly sensitive to increased equatorial uplift and would have a knock on effect on higher latitudes.
Another feature you need to add is top down solar effects on the extent of the polar air masses but that is another story.
Stephen, in the link you provide above, the first reference to evaporation is,
It cannot be determined by the energy content of the air because under an open sky warm air
above cool water just increases evaporation for a net cooling effect which cancels out the
extra warmth in the air.
You have got this the wrong way round. Increased air temps decrease ocean evaporation, as you have previously agree with me in this thread.
The problem here may be that for people who live in temperate zones, warm daytime temperatures result from cloudless days with low humidity. And obviously lower air humidity increases evaporation. But this is a humidity effect not a temperature effect.
I should amend that last paragraph.
The problem here may be that for people who live in temperate zones, warm daytime temperatures in summer result from cloudless days with low humidity. And obviously lower air humidity increases evaporation. But this is a humidity effect not an air temperature effect. And solar insolation is much higher on cloudless days warming the ocean relative to the air which will also increase evaporation. Again not an air temperature effect.
So how much natural variability is there in the degrees between +23 1/2 Lat and -23 1/2 Lat anyway, I always thought it was 47 degrees right on the button.
“”””” Philip Bradley says:
May 26, 2012 at 3:28 pm
Stephen, in the link you provide above, the first reference to evaporation is,
It cannot be determined by the energy content of the air because under an open sky warm air
above cool water just increases evaporation for a net cooling effect which cancels out the
extra warmth in the air.
You have got this the wrong way round. Increased air temps decrease ocean evaporation, as you have previously agree with me in this thread. “””””
Well I believe evaporation depends on the Temperature of the liquid (ocean); not the Temperature of the air. The water molecules in the ocean have no idea what the atmospheric Temperature is. Now the atmospheric Temperature will affect the rate of precipitation; but it has no effect on the rate of evaporation.
SW, Yep, I’m sure that negative numbers are equatorward. It’s a simple linear regression with the latitude being dependent on the year. The negative coefficient indicates that the tropical border is moving south. The reason the 1979-1999 coefficient was positive and the 1970-1999 coefficient was negative was that the values in the 1970’s were relatively high. You can see them in this plot:
Given the uncertainties, your theory might still be correct. Alternatively, maybe the relationship is opposite of what was expected. Think of El Ninos. I believe they are characterized by a slow down in ocean upwelling. Given the ocean/atmospheric coupling, perhaps there is a slow down in the meridional winds?