Global Patterns of the Risk of Seasonal Climate Extremes Related to ENSO
Robert S. Webb, Jon K. Eischeid, Henry F. Diaz, Klaus E. Wolter, Catherine A. Smith, and Randall M. Dole
Background
The emerging El Niño conditions in the equatorial Pacific [Figure 1]
lends itself to the question of how and what will be the spatial and temporal signature of climate impacts from the anticipated
changes in the sea surface temperatures in the coming months.
Figure 1. Most Recent weekly SSTs and anomalies from NOAA-CPC ENSO Advisory [http://www.cpc.noaa.gov/products/analysis_monitoring/enso_advisory/] click on figure to enlarge |
Many expectations for upcoming climate extremes and weather events over the 2002 Fall to 2003 Winter and Spring are based on the current memory of the 1997-98 El Niño and other recent El Niños. Alternatively, numerous rigorous studies [e.g., Ropelewski and Halpert, 1987; Kaladis and Diaz, 1989] and WWW based compilations and interperative composite maps have provided global to regional assessments of the average changes in precipitation and temperature associated with El Niños and La Niñas. Unfortunately the development of an objective seasonal prediction system for the global patterns of temperature and precipitation extremes under the potential of evolving El Niño conditions remains an ongoing research task. In lieu of such a prediction capability, we use the instrumental record to make quantitative assessment of global patterns of the risk of ENSO related seasonal climate extremes [often defined as the upper and lower 20th percentiles of the climatological range]. In the following study we have followed Wolter et al [1999] approach developed to charactize the risk of climate [temperature and precipitation] extremes associated with El Niños and La Niñas for the coterminous United States. We expand on and slightly modify the Wolter et al [1999] approach to make a assessment of the gobal patterns of the risk of seasonal temperature and precipitation extremes related to El Niño and La Niña.
Extreme climate conditions strongly impact [both positively and negatively] the natural environment and society. Mearns et al. [1984] highlighted the potential large sensitivity of extreme events to relatively small changes in the mean conditions under climate change. The natural environment and society have been,and will continue to be, strongly impacted by extreme climate events associated with ENSO variability. Our climate extremes focus is based on the recognition that minor changes in the mean can result in rather significant change in the tails of a population distribution. For example, an mean climate shift equal to a 1/2 standard deviation decrease will double the likelihood of dry[cold] extreme events expected under normal conditions while halving the likelihood of wet[warm] extreme events expected under normal conditions (Figure 2). Likewise, a mean climate shift equal to 1/2 standard deviation increase would double the likelihood of wet[warm] while halving the likelihood of dry[cold] extreme events expected under normal conditions.
Data and Methods
To generate that global maps of the risk of seasonal
temperature and precipitation extremes related to El Niño and La Niña, we used temperature and precipitation data from
close to 8000 terrestrial stations with data spanning 1890 to 2000 [Vose et al., 1992].
The monthly station data were converted into 3-month seasonal anomalies and then gridded to the
NCAR PaleoCSM atmosphere 3.75°x°3.75 grid [48 latitudinal and 96 longitudinal grids] for each of the twelve 3-month seasonal averages [JFM, FMA, ....., DJF]
The existence of El Niños and La Niñas conditions for the risk analyses were determined using the monthly Bivariate ENSO Timeseries [ 'BEST' Index; Smith, C.A. and P. Sardeshmukh, 2000]. Given the persistence of sea surface temperatures and the lagged impact on the atmosphere and resulting climate conditions in the 110 year climate record analyzed, if the lead month of the three month seasonal average was determined using the BEST indes to be either an El Niño or La Niña then within the risk analysis the season was considered to be associated with an El Niño or La Niña.
To define ENSO climate extremes analyses were made at each grid box for each of twelve 3-month seasonal averages if data existed for a least 82 [75%] of the 110 years record. The 20 and 80 percentile values for each of the 3-month seasonal averages were used to define the climatological extreme threshold. The risk associated with El Niño or La Niña was calculated as the ratio of the 20 percent expected for a given month for both tails of the distribution versus the actual number of years for each 3-month seasonal average that exceeded the 20 and 80 percentile climatological extreme threshold.
For example, in the 109 year record of seasonal NDJ precipitation along the east coast Australia, 9 of the 11 years classified as La Niñas, were in the upper 20 percent of the dataset (Figure 3). Since by chance one would expect only two of the La Niña years to be extreme (11 years multipled by 0.2) then the change in risk was 4.5 (9 La Niña extreme years divided by the 2 expected extreme years).
We used a boot-strap resampling with replacement test for significance with a sample size of 10,000 and only included results that were significant at >95% confidence interval. in our global maps of the risk of seasonal temperature and precipitation extremes related to El Niño and La Niña,
Results
The resulting global patterns of
Seasonal Precipitation and Temperature Extremes Related to El Niño and La Niña can be accessed below.
Note that in many places there is not always a symmetric response in increase or decrease risk for wet/warm or dry/cold
extremes under El Niño or under La Niña conditions. Likwise many places there is not always a symmetric response in increase or decrease risk for wet/warm or dry/cold
extremes between El Niño and La Niña conditions.
To illustrate how a change in risk associated with El Niño or La Niña relates to shifts in the mean and extreme temperature and precipitation values, we selected a subset of cases with exceptional increases in the risk for extreme conditions. For these cases we generated empirical probability density functions [PDFs] for the complete temperature or precipitation records [all years] and for the subset of years under El Niño or La Niña conditions. In many cases the shift [increase/decrease] in the risk of climate extremes is associated with large shifts in mean climate [e.g., the east coast of Australia ], however in some cases [e.g., along the Pacific coast of South America ] significant increases or decreases in the risk of climate extremes can occur with only minor changes in the mean value. Figures illustrating differences in the tail of the probabilities [extremes] associated with El Niño and La Niña years can be accessed below.
Concluding Remarks
Although past performance is no guarantee of future performance, an examination of the instumental record of the
risk for seasonal temperature and precipitation climate extremes associated El Niños and La Niñas
can provide insights into the range of possible future climate
impacts. Extreme climate conditions associated with El Niños and La Niñas have and will continue to strongly impact
the natural environment and society. The results presented in this study provide an objective quantitative assessment of global patterns
of the risk of ENSO-related seasonal climate extremes. The maps showing the global patterns of the risk of seasonal
precipitation and temperature extremes related to ENSO can be interpreted to provide guidance in terms of where one might expect an increased or decreased
risk of flooding, drought, heatwave, or coldwave under El Niño and La Niña conditions; however, these associations should not be
misinterpreted as deterministic. As illustrated in the empirical probability density function analyses, in almost all cases the increases in risk of extreme wet/warm climate conditions
do not exclude the possibilty of extreme dry/cold or mean climate conditions. In other words, an increase or decrease in risk of climate
extremes does not preclude the opposite from happening, just that the probability of the opposite occuring is significantly reduced.
References