Mexican drought: natural variability and climate change from the Medieval period to the greenhouse future

The ongoing drought in the western U.S. began in 1998 and has now continued, with some interruptions and despite a quite wet winter 2007/8, for close to 10 years. Less well know is that this same drought has impacted Northern Mexico and even began earlier there. The Mexican drought has coincided with major changes in the Mexican economy and agriculture triggered by the North American Free Trade Agreement and moves to privatize water supply in much of Mexico. The combination of drought and economic change has created serious social impacts in Mexico with impacts on internal and cross-border migration. Both the southwestern United States and Mexico are robustly projected by climate models to dry in the current century intensifying social impacts in Mexico where water resources are already stretched. As such it is important to understand the causes of droughts in Mexico, assess their predictability, examine modern droughts in the context of climate change over the last several centuries and examine the character of projected anthropogenic induced drying as a means to assessing whether it is yet occurring. This is what we try to do in a recent Atmosfera paper (PDF).

In the paper we use instrumental observations to show that Mexican hydroclimate is impacted by both tropical Pacific and tropical Atlantic SST anomalies. During winter half years El Niño events in the Pacific tend to make all of Mexico wetter than normal. El Niño conditions during the summer half year tend to make northern Mexico wet but southern Mexico dry. Warm tropical Atlantic SSTs tend to make northern Mexico dry (mostly in the winter half year) and southern Mexico wet (mostly in the summer half year). These SST-Mexican hydroclimate relations are reasonably well produced in an atmosphere model forced by historical observed SSTs indicating potential predictability (if SSTs can be predicted).


Figure 1. The multiple regression between precipitation and the Tropical Pacific (TP, top) and Tropical North Atlantic (TNA, bottom) SST indices (see text for definition) for the November through April half year using the UNAM precipitation data set and for 1945 to 2002. The regression coefficient in mm/month per standard deviation of the SST index is both contoured and colored but coloring is applied only where the relationship is significant at the level.		Figure 2. Same as figure on left but for the May through October half year.


Figure 3. The regression of modeled precipitation on the tropical Pacific SST index and with tropical Pacific SST forcing alone (POGA, top) and on the tropical Atlantic SST index with tropical Atlantic SST forcing alone (TAGA, bottom) for the November through April half year and the 1945 to 2002 period.		Figure 4. Same as figure on left but for the May through October half year.

Instrumental records of precipitation are sparse over Mexico before the mid 20th Century so we also looked at tree ring records that reconstruct spring-summer Palmer Drought Severity Index. Model simulations of the entire period of ship-based SST observations - 1856 to now - were verified against the tree records. The early to mid 1890s and the 1950s droughts stand out as severe within the last century and a half. The 1890s persistent drought - which caused starvation across northern Mexico - seems to have been forced by a persistent La Niña while the 1950s drought was potentially influenced by both the Pacific and the Atlantic Oceans.

Figure 5. The tree ring-reconstructed PDSI and the modeled soil moisture with (top) global SST forcing and (bottom) tropical Pacific SST forcing alone and a mixed layer ocean elsewhere, averaged over northern Mexico, for the 1856 to 2004 period. The time series are standardized and have been smoothed with a 3 year low pass filter. The shading is the two standard deviation spread of the model ensemble.

A novel modeling method is used to examine the means whereby the tropical Pacific influences Caribbean and Mexican precipitation during the summer season. We use a 100 member ensemble of 100 day integrations and examine the day by day evolution of the atmospheric response to a sudden turn-on of a tropical Pacific SST anomaly. Kelvin waves force vertical motion and precipitation anomalies to the east of the Pacific. The induced atmospheric heating anomalies over the tropical Atlantic excite Rossby waves that transmit a signal to the north and west creating the summer time precipitation signal over Mexico, the Gulf of Mexico and the southern United States.

Figure 6. The anomalies in precipitation (contours) and the contribution of anomalies in circulation to the anomalies of vertically integrated moisture convergence (colors) from the superensemble for the case of summer season El Ni\~no minus climatological conditions. All results are averaged over 100 ensemble members for each of the two SST forcings with the 100 pairs of integrations beginning with the same 100 sets of atmospheric initial conditions. Results are shown at two day intervals from day 1 to day 13 as well as the 100 day average over the length of the integration.

Analysis of tree ring records going back to the 14th Century does not show any multicentennial trends but does reveal a late 16th Century drought that far exceeded in severity any since. However model projections do show a robust drying of Mexico in the the early part of the current century. It is not clear that this anthropogenic drying is already underway since the current drought has a different spatial pattern. Anthropogenic drying is expected to impact all of Mexico whereas naturally occurring droughts have ended to strike either northern or southern Mexico but not all. 'All Mexico' droughts will place a new kind of strain on the country. The current drought has already had a notable impact on societies in northern Mexico and future drying will certainly have important consequences on agriculture and water supply for cites and industries and is likely to be part of a mix of factors that will impact future development, the economic and social well being of the Mexican people and internal and cross border migration.

Figure 7. Various time differences in annual mean hydroclimate fields. The difference between the post 1993 period and the 1945 to 1992 period for UNAM precipitation observations (top left) and in the multimodel ensemble mean of IPCC AR4 model precipitation simulations (bottom left) and PDSI from tree ring records extended with instrumental data (middle left). The difference between the post 1993 period and the 1979 to 2002 period for satellite derived precipitation from CAMS (top right) and the multimodel ensemble mean of IPCC AR4 precipitation simulations (middle right) . The IPCC AR4 multimodel ensemble mean precipitation difference between the 2021 to 2040 period and the 1945 to 2000 period (bottom right). Units are mm/month.