Understanding evaporation #4: climate change

One of a series; also see editions: #1, #2, #3

A warmer climate must mean more evaporation, a drier continent and more drought? Well, yes, that’s pretty much the prediction, but unfortunately the observed response of evaporation to climate change has long been a bit of hole in the story — a real one, unlike all that manufactured tosh.

In lots of places recorded pan evaporation has actually been decreasing as the climate warms. The data is fairly robust, but extreme care is warranted. First there’s that old bird-guard thing, which could make our pre-1975 data appear higher than it actually is¹. Also, unlike a thermometer in a Stevenson screen, an evaporation pan really is highly sensitive to site conditions. Trees growing slowly in the far distance to block just a few minutes of morning or afternoon sun – or a little bit of prevailing wind – really will spuriously reduce the reading. But folks have looked hard and it seems that’s not it; the effect is probably real:



That is post-1975 annual pan evaporation averaged across the whole of Australia, from the Bureau of Meteorology. The downtrend is tiny — about 1% a decade — but it is statistically significant (at the 5% level² — the likelihood of seeing such a trend by chance in a sample of data with no underlying trend is less than 5%).  Similar and larger trends have been observed in a range of places around the world.  The thing even has a name: the pan evaporation paradox.

So what’s going on?  Before proceeding it’s worth revisiting a couple of points.  An evaporation pan measures pan evaporation.  That is not meteorological evaporation; not agronomic evaporation, not evaporation from a forest or field; not even evaporation from a dam or lake.  It’s evaporation from a fancy tin can in a paddock.  It’s a guide to some of those practically important evaporation rates, but that’s all.  So when we notice an unexpected trend, that is definitely interesting and we ought to find out why, but we shouldn’t automatically assume that it’s of relevance to the real world.  We certainly shouldn’t assume that it means that average global evaporation or even terrestrial evaporation are decreasing and will decrease more.


Pan evaporation paradox

The first explanation folks proffered, a couple of decades back, was that old global dimming thing — particulate pollution reduces sunlight penetration to ground level, so there’s less energy input and less evaporation.  The trouble with that is that “dimming” was probably never global (how much in Australia?) and in any case it doesn’t seem to have increased much since about 1990, while pan evaporation has continued to decrease.

Then an influential 1998 paper in Nature pointed out that pan evaporation could be decreasing as result of the effect that also reduces areal potential evapotranspiration compared with point potential.  An overall increase in the intensity of the hydrological cycle as the world warms (through more actual evaporation and precipitation) should operate to reduce pan evaporation (and point potential evapotranspiration) due to the altered meteorology.  It didn’t stop there.  A later paper by Australian geographer Mike Roderick³ in rival tier-1 Science claimed instead that the effect was due to dimming plus increased cloudiness … which, if you think about it, isn’t necessarily all that contradictory.

A lot more has been said since, but if you accept Roderick’s latest assessment (see references below), the right answer is pretty much “all of the above” … plus, umm, reduced average wind speed, which also seems to be widespread.

The long and tortuous path to this rather unsatisfactory answer illustrates the complexity of real-world evaporative processes.  Changing evaporation matters to the outcome, but our level of understanding is so weak that we’re still not even sure what the sign is.


Where to next?

For a start, reviewers on serious journals could reject any paper that references Thornthwaite, Penman, Monteith, Priestley & Taylor, Morton, Bouchet or Budyko.  Over-the-top maybe, but surely it’s time to move on from twentieth century fudge formulas and unsupported hypotheses.  Evaporation is a dynamic feature of a dynamic system with important variations and granularity in all four dimensions — especially the time dimension.  Handling all that in a sophisticated way just wasn’t possible last century, but it is now.  Measurement, modelling and understanding all need to operate in that league.

The measurement people have made a start with Fluxnet — instrumented towers around the world that record the micro-meteorology around real-world transpiring plants.  Integrating that gives a direct measure of actual evapotranspiration (and lots else).  [There’s a Fluxnet tower at Dargo High Plains south of Mt Hotham, and there used to be another out from Falls Creek, near Mt Nelse.]

Modelling needs to move to finer granularity too, particularly in the time dimension.



1. There’s a standard correction factor for that (-7%), but Bill Weeks showed many years ago that it can be way low; at least it was for his site in central Queensland.

2. There’s a little bit of (probably spurious) autocorrelation, but the trend significance persists with the data lumped into 5-year averages to counteract it.

3. Roderick is one of the most influential contributors on this topic globally.  I don’t agree with a fair bit of what he’s written over the years, but I think he and others now have a workable explanation for the paradox.  But it still lacks detail.



  • Brutsaert, W., and M. B. Parlange. “Hydrologic cycle explains the evaporation paradox” Nature 396, no. 6706, 1998: 30-30.