Friday, January 27, 2012

Historical Note on Drought in Climate Models

This morning, I stumbled on a 1999 paper "DETECTABILITY OF SUMMER DRYNESS CAUSED
BY GREENHOUSE WARMING" by Wetherald and Manabe. The paper discusses a single climate model (obviously a by-now very outdated one) which generates very serious drought across much of the world in the second half of the twenty-first century.  The map above gives the general idea.

Much more refined predictions with averages over many far-more sophisticated models have only changed the picture a bit:

The paper's discussion of the physical mechanisms at work is interesting. There are separate discussions of the more northern interior of the US ("CNA1") and the more southerly portions of North America (eg Mexico, Texas - which they call "CNA2").

For CNA1:
In general, the increase of both CO2 and water vapor in the model atmosphere increases the downward flux of longwave radiation absorbed by the continental surface. This results in an early disappearance of snow cover (with a large surface albedo), thereby increasing the solar energy absorbed by the continental surface. Because of the increase in the surface absorption of both longwave and solar radiation, evaporation is enhanced during spring and early summer (Figure 10a), reducing the soil moisture in the CNA1 region. By mid-summer, the soil moisture is reduced to the point where evaporation can no longer increase. Thus, evaporation is decreased and sensible heat increased, reducing the near-surface relative humidity and cloud cover, which increases insolation absorbed by the continental surface and makes more energy available for evaporation. On the other hand, the rate of precipitation hardly increases over the continents in summer because of the low relative humidity in the lower troposphere (Figure 10a). As a matter of fact, the rate of precipitation even decreases slightly after mid-summer when the soil becomes very dry. Therefore, the soil moisture anomaly remains negative throughout the rest of the summer and early fall in ‘CO2 + SUL’

In sharp contrast to the summer situation discussed above, soil moisture in CNA1 increases during winter in ‘CO2 + SUL’ (Figure 9a). Despite the increase in the downward flux of longwave radiation, the rate of evaporation hardly increases over the continental surface in middle and high latitudes where the increase in the downward flux is compensated mainly by the upward fluxes of sensible heat and longwave radiation rather than latent heat flux in winter. On the other hand, the rate of precipitation increases significantly in the CNA1 region where the relative humidity in the lower troposphere is much higher in winter than in summer (Figure 10a). This is quite different from the situation in summer when the relative humidity in the lower troposphere is very low and precipitation hardly increases in this region despite the increase in the rate of evaporation in the surrounding oceans.

The increase in precipitation, together with the failure of evaporation to increase,
accounts for the increase of soil moisture in the CNA1 region during winter in
‘CO2 + SUL’.
By contrast: in the drier regions to the south:
In the CNA2 region, where there is little or no snow cover in spring, the monthly mean soil moisture is considerably below saturation during most of the year. Figure 9b indicates that the CO2-induced change in soil moisture is negative throughout the entire year and the maximum reduction occurs from summer to early winter.

As in CNA1, the reduction of soil moisture in CNA2 is attributable partly to the increased downward flux of terrestrial radiation resulting from the increase of both CO2 and water vapor in the model atmosphere.

The increase in the downward flux of longwave radiation raises the surface temperature, thereby increasing the rate of potential evaporation. As Figure 10b indicates, the increase in potential evaporation increases the evaporation from early winter to May when the soil moisture is relatively large. However, the CO2-induced change in the evaporation rate becomes slightly negative from summer to early fall when soil moisture is relatively small. On the other hand, the CO2-induced change of precipitation rate in the CNA2 region is negative throughout most of the annual cycle (Figure 10b), in sharp contrast to the situation over most of the globe where both evaporation and precipitation increase in response to the increasing atmospheric carbon dioxide. The failure of precipitation to increase is attributable partly to the reduction of near-surface relative humidity in these regions where a major fraction of radiative energy absorbed at the land surface is removed as sensible heat flux rather than through evaporation.

The increase in potential evaporation together with the reduction of precipitation discussed above contribute to the general reduction of soil moisture shown in Figure 9b. The decrease of soil moisture, in turn, reduces the near-surface relative humidity and precipitation rate further throughout most of the year (Figure 10b). The reduced near-surface relative humidity induces a corresponding reduction of low level cloudiness, increasing the insolation reaching the continental surface and further enhancing the drying of the soil. This analysis applies equally well to other semi-arid regions of the world such as central Asia and the area surrounding the Mediterranean Sea.
The paper also predicted it would be a little while until the trend became clear above the natural variability:
Results of the ‘CO2 + SUL’ integration suggest that, over mid-continental regions of middle and high latitudes, the summer reduction and winter increase of soil moisture will not become noticeable until the first half of the 21st century. An analysis of the central North American and southern European regions indicates that the time when the change of soil moisture exceeds one standard deviation about the control integration occurs considerably later than that of surface temperature because the ratio of the forced change to the natural variability is smaller for soil moisture than for temperature.
It does seem to be true that clear emergence of a drying trend has occurred later than clear emergence of a warming trend.  However, now, a decade into the twenty-first century, we do seem to be getting there.


Fred said...

A group in Spain recently published results of a new approach to drought-casting, applied to Spain:

"Extreme droughts could increase by 15 percent in Spain by the middle of the century"

It uses GAMLSS modelling (Generalized Additive Models for Location, Scale and Shape)so is independent of the other approaches, and is intuitively more satisfying because it works for non-stationary distributions (i.e., climate change results in changing underlying meteorological statistics, which does not bother GAMs). The report results are similar to the other models you have been reviewing
(see link)

Unknown said...

This paper was a great find!

Michael Cain said...

The sequence of drought papers has been extremely useful. The rural parts of the US Great Plains are already depopulating except along the major rivers; long-term shifts towards drought can only accelerate that process. It will be interesting to see how things play out for the US as a 500-mile wide stretch down the center of the country becomes a desert, or at least dry enough that agriculture isn't possible.