Preliminary Assessment of Climate Factors Contributing to the Extreme 2011 Tornadoes

Draft Assessment - Updated: 08-July-2011

Disclaimer: This draft is an evolving research assessment and not a final report. Comments are welcome. For further information, contact Dr. Martin Hoerling (
Overview | El Niño/La Niña | Climate Change | Q & A | References

Q&A Regarding Analysis of Tornado-Climate Links

As indicated in our preliminary and ongoing assessment of the tornado-climate link, various data sets are important in assessing if and how evolving climate conditions may impact the frequency of tornado occurrences. These include the statistics of tornadoes themselves, yet as we initially emphasized, time series of tornado reports represent a blend of climate impacts and observing practice changes. Research efforts in the severe storm and climate science communities have involved inferring tornado frequency from proxy information of the larger scale atmospheric conditions. There is a body of science that has identified several favorable large-scale factors in the atmosphere (e.g., vertical wind shear in the low troposphere, atmospheric instability) which contribute to a physical basis for exploring a climate link with tornadoes. There may thus be value in deriving a proxy for tornadoes based upon using an index of environmental conditions, and indeed such knowledge forms part of the scientific basis for NOAA risk outlooks of tornado outbreaks that are issued 12 to 72 hours in advance. Nonetheless, a climate theory for tornadoes is in a relative infant state. Among the challenges to infer a change in tornadic conditions associated with evolving climate is the relatively brief record of the large scale environmental factors. The best available data sets combine remote satellite observations with in situ measurements from rawinsondes in a consistent manner through time, but such analyses (and re-analyses) are only available for recent decades.

Several questions are thus relevant to our current capability for describing and understanding the tornado-climate link.

Q: We use a 30-year period to possible tornado-climate relationships. Is this long enough for detecting change? Why have we chosen the most recent 30-year period?

A: Longer records would in fact be very desirable, if data quality were adequate. Extensive work has been done and is published in the peer-reviewed literature on the challenges of determining trends in the tornado record, although the best available evidence suggest that there are not significant physical trends in higher-intensity tornadoes for which the record is somewhat more reliable. These issues are discussed in our preliminary report, along with a specific example illustrating the point. Factors that relate to a non-physical trend include changed instrumentation (e.g., the implementation of the WSR-88D Doppler radar after 1990), increases in population, and public awareness via enhanced spotter networks that may have contributed to more reporting. A careful appraisal of historical tornado outbreaks has led the scientific community to recognize that temporal trends in tornado occurrences before 1970 are likely more influenced by non-meteorological factors, although the quality of time series even after 1970 may also suffer from inhomogeneities such as described above. An important finding of our preliminary draft report, consistent with extensive published literature and assessment from NOAA experts at the National Severe Storms Laboratory and the NOAA Storm Prediction Center, is that the historical tornado data are not of a quality to permit a rigorous appraisal of long-term trends.

Time series of various climate parameters, such as water vapor, atmospheric stability, and vertical wind shear were examined that are widely known and generally accepted in the severe storm research community and in operational practices as being relevant for evaluating large-scale conditions conducive to severe storm outbreaks. The relevant research indicates that conditions for tornadoes may depend more strongly on the magnitude of wind shear than other classes of severe storms. The data sets that were used begin near 1979 and are considered to be of very high quality over the continental U.S. This was another important consideration in selecting this 30-year period for a preliminary analysis of possible climate-tornado relationships for this case.

It also happens that the period from 1979 to the present is the period in which Earth's mean surface temperature has experienced the most rapid warming. It's therefore reasonable to assume that if a relationship exists between global warming and those three variables - CAPE, wind shear and precipitable water vapor - and tornado outbreaks, it could be most evident during that same time period.

Q: Extreme weather events are by definition at the extremes of the probability distribution. Can one infer changes in extreme events from examining changes in averages?

A: Optimally, higher temporal resolution is required, although this may obscure rather than illuminate some possible climate links, because of the much higher variability that is inherent in high-temporal resolution data. Ideally, we would like to estimate the full distribution and how a condition or factor relevant for tornadoes changes in time; in practice, limitations of data, as discussed in the answer to the previous question, make this a very difficult challenge. Physically, it is reasonable to anticipate that any signal of anthropogenic influences is likely to manifest itself in significant part in signals that are present on longer time scales, consistent with the first question, and so should not change too much over a monthly averaging period. However, we state explicitly in the preliminary draft report: "An implicit assumption, which should be verified in future research, is that a change in the monthly average conditions would indicate similar changes in the tails of the daily distribution, i.e., a slight increase in the mean monthly CAPE would result in more individual days with CAPE sufficiently high to support tornadic activity. Similarly, future research would explore the frequency and change over time of the joint instantaneous occurrence of the high CAPE and high shear, and possibly moderate CIN that are conducive for tornado development." Because the means and variability are extracted from the same time periods, shifts in the means will in most cases also be manifested in changes in extremes. Indeed, this appears to be the primary explanation for changes in frequency (and perhaps intensity) of heat waves and perhaps other phenomena in a warming climate; it is effectively an elevation of the baseline temperature. To take an example more directly related to tornadoes, to achieve a wind shear anomaly of 30 m s-1 as might favor an extensive tornado outbreak, a monthly mean anomaly of 15 m s-1 would only require an additional variation of 15 m s-1, rather than a much larger perturbation. So, while we recognize that analyzing a monthly mean is non-optimal and fully agree on the need to estimate the full distribution (to the extent possible), considering the monthly mean values is an important first step in helping to detect possible tornado-climate relationships.

Q: Do the large scale climate parameters (e.g. April averages of water vapor, stability, and vertical ,wind shear) have some predictive power for the number of tornadoes?

A: The parameters used, as stated above, are of primary importance in short-term predictions of tornado outbreaks, while climate signals are more detectable on longer time scales. As stated above, monthly mean variations implicitly include information on extremes. There is relatively more limited work in this area, although such information has been used on seasonal-to-interannual time scales in associated with the ENSO. There are also several other pieces of evidence that time-mean conditions are indeed important for determining the frequency of severe storms. One is the climatological seasonal cycle from which it is well observed that the statistical frequency of tornadoes increases markedly from winter to spring in many regions of the United States. This is principally the result of the significant increase in atmospheric instability due to mean surface temperature rises and moisture increases between those seasons. Similarly, tornado frequency decreases in many areas from Spring to Summer principally because the mean vertical wind shear declines, even though thermodynamic instability remains high. One can also infer that time-mean conditions matter when examining the diurnal statistics of tornadoes. It is reasonable to examine the changes in mean wind shear and mean stability over time as a potential indicator for both possible influences of climate and changes in tornado frequency, though that relationship will undoubtedly be complex. The data for April 2011 clearly show monthly-mean anomalies that are consistent with the shorter-term relationships. In particular, this April was characterized by unusually large shear (in a monthly-average sense) in the region affected by the tornado outbreak, as well as during the days of the outbreaks themselves (this is an essential physical factor, without which tornadoes will not form).

Overview | El Niño/La Niña | Climate Change | Q & A | References