Global Decadal Hydroclimate Predictability, Variability and Change: A Data-Enriched Modeling Study (GloDecH)
Lamont-Doherty Earth Observatory of Columbia University 61 Route 9W Palisades, NY 10964

Did active dunes and dust storms intensify the North American Medieval megadroughts?

Ben Cook (NASA GISS and Lamont Doherty Earth Observatory)
Richard Seager (Lamont Doherty Earth Observatory)
Ron Miller (NASA GISS)
Joseph Mason (University of Wisconsin)

sand hills

Nebraska Sandhills, Hooker County, Nebraska, south of the Dismal River. Source: Wikipedia Commons
Linear dunes superimposed on a larger dune form are clearly visible in the photo (the ridges running left to right across the larger hill in the background). Every linear dune of that type dated by Mason's group was active in the Medieval period, and many of them seem to have been constructed at that time.

The Plains and Southwest were struck be a series of severe droughts of multidecadal duration during the Medieval period from about 800 AD. to 1500 A.D. This is well known from tree ring reconstructions of past hydroclimate. These past megadroughts lasted much longer than any that have occurred in the period of European settlement and instrumental records. Given the water scarcity in the region the causes of the megadoughts are a matter of great concern. After all, knowing what caused them would guide an assessment of how likely they are to return. A multidecadal megadrought equivalent, say, to the one which occurred in the mid 12th Century would place such severe stress on the regional water supply that serious disruption of agriculture, urban water supply and the economy would be inevitable. Alas, the causes of the Medieval megadorughts remain unknown. One theory is that they were caused by multidecadal length and frequent changes in tropical sea surface temperatures (SSTs) - cold tropical Pacific waters and/or warm subtropical North Atlantic waters. This theory is based on sparse proxy reconstructions of SSTs in the two oceans but the data is open to challenge.

In this work (Cook et al. 2013) we chose to take at face value that the tropical Pacific ocean was La Niña-like - i.e. colder than the current normal - during the Medieval period and then examine the roles of coupling between the atmosphere-land surface-vegetation and dust. This was motivated by geomorphological evidence of widespread blowing sand dunes across the Plains from the Gulf Coast into Canada during the Medieval period. The dunes are now fossilized and stable with grass covering and little or no blowing dust. The only time since the Medieval period when there was blowing dust on a large scale was during the 1930s Dust Bowl drought. In that case crop failure during drought combined with poor agricultural practices to cause tremendous soil erosion and dust storms. In prior work we have 1) identified the unusual spatial pattern of the Dust Bowl drought which was centered in the central Plains, well north of typical SST-forced droughts, 2) that the dust storms interacting with solar radiation and atmospheric circulation are required to correctly reproduce this spatial pattern and 3) that the only other droughts that were centered on the central Plains occurred in the Medieval period, also a time of active wind erosion and dust storms (was the Dust Bowl predictable? did dust storms make the Dust Bowl drought worse). Hence we decided to examine the impact on the climate system of the active dunes during the Medieval period.

This work was an extension of earlier work that tested the impact of tropical Pacific SSTs on Medieval megadroughts. Coral records from a single location in the tropical Pacific (Palmyra) were used to reconstruct tropical Pacific SSTs for the period of 1320-1462 A.D. that the coral covered. These SSTs were used to force an ensemble of atmosphere GCM runs. The coral records indicated cold, La Niña-like, SSTs during this late Medieval period and, as expected, the model produced extended droughts over the Southwest and Plains. In this work we used the GISS atmosphere model, which has an active dust module, and imposed under it the same history of tropical Pacific SSTs for 1320-1462. We also used geomorphological evidence of active dunes to indicate where the Plains were devegetated and dust sources were present. We then ran three ensembles: 1) SST forcing only (SST-only), 2) SST forcing plus devegetated, bare soil surface where the dunes were active (SST+BSOIL) and 3) SST forcing, bare soil and interactive dust (SST+BSOIL+DUST). The modeled dust emissions were actually on the low end of estimates based on geomorphological evidence.

The results of these model experiments are:

1. SST forcing alone produces drying and warming over the Plains but cannot represent the persistence, year to year, of the megadroughts (as reconstructed from tree ring records).

2. The replacement of grasses with bare soil in the observed region of active Plains dunes causes warming (as evapotranspiration decreases and the surface energy balance has to be met via sensible and radiative heat loss, requiring higher surface temperatures) but does not strongly impact precipitation.

3. The addition of active dust sources led to dust emissions and significant aerosol loading that further decreased precipitation relative to the case with SST and bare soil forcing alone. Active dust emission from the dunes also increased the persistence of the megadroughts making them more in agreement with tree ring evidence of the character of the megadroughts. In the model dust aerosols suspended in the atmosphere reflect solar radiation reducing energy available at the surface increasing atmospheric stability and reducing upward motion and convection. This suppresses precipitation. These impacts are shown in the two figures here.

Ensemble mean monthly surface temperature (JJA, K) (a,c,e) and precipitation (June, mm day-1) (b,d,f) responses over North American in our Medieval model runs. The central Plains region (105W-95W, 32N-44N) is outlined in the black dashes. Cells with insignificant (p <0.05) differences between the Medieval and modern runs have been masked out.
fig11 Ensemble average summer season (JJA) PDSI calculated from our Medieval runs and averaged over the Central Plains region (105W-95W, 32N-44N): (a) SSTOnly, (b) SST+BSOIL, and (c) SST+BSOIL+DUST. PDSI anomalies for all Medieval runs were normalized relative to 1857-2005 C.E. from the modern ensemble.

This work presents the fist evidence that the devegetation of dunes, their mobilization and resulting dust storms and aerosol loading were important drivers of the well-observed Medieval megadroughts. Ultimately it remains the case that the droughts were probably initiated by persistent tropical SST anomalies (which occurred for unknown reasons). However the unusual persistence of these, much longer than in the recent period of instrumental observations, caused drying over the Plains that was persistent enough to cause grasses to die and allow dunes to be formed and become active. The blowing dust then caused even more drying in a climate-land surface-dust self sustaining feedback loop. The work suggest that it is this interaction that explains the severity and persistence of the Medieval megadroughts.

However some uncertainties remain. We ignored the possibility that Atlantic SST anomalies could have contributed to the megadroughts, largely because of the paucity of high temporal resolution SST recosntructions from the Atlantic. The University of Nebraska group led by Robert Oglesby have argued, on the basis of what evidence does exist, that warm Atlantic SSTs contributed to the megadroughts (Oglesby et al., 2012). Also the tropical Pacific SSTs we use for 1320-1462 are highly uncertain. Much more work needs to be done to reconstruct tropical SSTs over the last millennium to reduce uncertainties in the ocean forcing. Of course we also need to explain why the SSTs varied as they did over the past millennium. Also dust-climate interactions are known to be model sensitive and repeating these experiments with other atmosphere models would be a good idea.