Seasonal Predictability of Tornadic Activity Using Antecedent Soil Moisture Conditions

EarthzineArticles, Earth Observation, Extreme Weather Theme, Original, Sections, Themed Articles

This table shows how the quality of each CREST model input varies at differing spatial resolutions (2m, 30m, 90m, and 2m resampled to 90m). The Drainage Direction Map (DDM), produced with the D8 algorithm, shows flow direction of each cell to the steepest adjacent slope, and the Flow Accumulation Map (FAM) displays surface rainfall from each upslope cell. Data sources included the U.S. Department of Natural Resources, U.S. Geological Survey and NASA.

Theresa K. Andersen and J. Marshall Shepherd

University of Georgia

Department of Geography and Atmospheric Sciences Program

What is the Motivation?

The period of 2010-2011 has included a breathtaking display of nature’s fury. Record-breaking death tolls and hazards from tornadoes occurred in Joplin, Missouri; near Springfield, Massachusetts; and in the southeastern United States (Fig. 1). It is likely that natural weather variability, La Nina, and perhaps even changing climate can be linked to these recent events. Yet other factors also influence extreme storm systems. In this article, we examine tornado climatology from a different perspective by examining effects of antecedent land surface conditions and land surface heterogeneity on tornado days in the southeastern United States, one of the regions most severely affected by the mega-tornadic outbreaks of 2011. The role of the land surface in affecting the atmosphere on climate time scales for prediction is a major area of study, but examination of its role in extremes at smaller spatial and temporal scales has not been as thorough.

Image of tornado tracks from Alabama's 2011 tornado outbreak.

Fig. 1 Storm tracks from the the 2011 Alabama tornado outbreak. These tracks were acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Aqua satellite on April 28, 2011.

If Intergovernmental Panel on Climate Change (IPCC) [1] projections are accurate, the frequency and severity of extreme water cycle events (e.g. droughts, floods) will increase as a consequence of climate change. The IPCC notes that hydrological extremes such as flooding and drought occurrence have increased markedly in the last three decades, with more intense and longer episodes. Large variability in the atmospheric component of the water cycle is directly linked to terrestrial soil moisture distributions. The consensus on soil moisture-atmospheric feedbacks is that surface heat fluxes and moisture gradients can influence convective development (e.g., [2],[3], [4], [5], [6], [7], [8], [9], [10], [11], [12]).

These studies and others have discussed the following connections between soil moisture and convective storm development: (1) Partitioning of surface energy into latent and sensible heating, which in turn affects the atmospheric boundary layer and regional convergence through moisture and energy exchanges; (2) Higher evaporative fraction; (3) High concentrations of entropy (or moist static energy) in the lower atmosphere; and (4) Development of mesoscale (e.g., sea breeze-like) circulations.

However, there is a lack of literature on the extension of soil moisture-atmospheric feedbacks to extreme hazards like tornadoes. At the same time, there is convergence of hydroclimate processes scales and weather-climate model resolutions, such that explicit feedbacks and connectivity between soil moisture and the atmosphere should be resolved [13]. Further, seasonal predictability of tornados has clear societal implications for safety, insurance companies, the construction industry, and even retailers like Home Depot or Lowes.

Previous Perspectives on Soil Moisture and Tornado-Producing Convection

The study by [14] found a small trend for high-precipitation years to have more tornadoes than low-precipitation years. Of three study areas (southeastern United States, Great Plains, and Great Lakes), the Southeast had the greatest correlation between departure from normal precipitation and tornado occurrence. [15] found that dry soil conditions, through mixing and surface heating, are important to the formation of the elevated mixed layer or ‰ÛÏlid‰Û in the dryline regions of the Great Plains. Finally, they indicated that drought conditions displace areas of moisture convergence and thereby shift the location of potential convective triggers. [16] used a Goddard mesoscale model to reveal that realistic soil moisture (from antecedent precipitation observations) and vegetation (from NDVI) produced more realistic and enhanced dynamic and thermodynamic structure along a stationary front that triggered convection. [17] and [18] found similar results.

ap of climate division boundaries for Alabama, Georgia, and South Carolina.

Fig. 2. Map of climate division boundaries for Alabama, Georgia, and South Carolina. The study area is outlined in red and overlaid with a terrain map from Google.

[19] suggest that moisture fluxes from the prairie agro-ecosystem may be associated with the seasonal pattern of tornado days.[20] likewise show that soil moisture in the Canadian prairie could be a good predictor of severe convective weather during the warm season. A recent study, spanning the 1950 to the early 2000s, by [21] suggests that tornado activity in the spring is suppressed over North Georgia when meteorological drought (i.e. precipitation departures occurred the previous fall and winter). Similar results are seen in the preliminary analysis conducted over the Midwest by Trapp et al. [personal communication]. The aforementioned studies are still limited in geographic extent and physical underpinnings. It is essential to advance the research using newer data and different approaches to yield more robust results. Our goal was to try to establish some physical underpinnings for the results in [21] and in new findings with new soil moisture data.

Physical Insight on Drought-Tornado Day Relationships

Our recent work, funded by NASA’s water and energy cycle program, has documented an asymmetric relationship between antecedent (fall/winter) soil moisture and spring tornadic days (Fig. 1) in the southeastern United States. Spring is defined as March-May, the three months with the highest number of tornadoes within the study area totaled over 1980-2006 (Table 1). We note a secondary peak in November but do not focus on that period at this time.

The study area (Fig. 2) is defined here as 32å¡N-35å¡N and 82å¡W-88å¡W. The area is chosen to include a representative portion of the southeastern region and to exclude the areas of Georgia and Alabama that might experience coastal convection and associated landspout tornadoes. The time period for this study spans 1980-2006 because of available data and reliability of the tornado reports as compared to pre-1980 records. Past tornado reporting was unreliable and accuracy in the records is limited. Tornado reporting relies on local documentation and inherently has issues related to discontinuities, changes in reporting style, population changes, storm spotter networking, and new technology ([22], [23], [24]). For these reasons, we use a tornado day metric (i.e., a tornado happened on that day in the study domain) rather than total tornado counts.

This research utilizes tornado records from the Storm Prediction Center (SPC). Soil moisture, and geopotential height values are from the North American Regional Reanalysis model (NARR). NARR uses the high resolution (32km/45 layer) NCEP Eta model in conjunction with the Regional Data Assimilation System. It is run eight times per day and incorporates the Noah Land-Surface Model [25]. The data are available for download from Earth System Research Laboratory (ESRL) Physical Sciences Division (PSD).

NARR contains four soil layers (0-10, 10-40, 40-100, 100-200 cm), with three prognostic soil states available for each layer. Total soil moisture (sum of liquid plus frozen moisture) is utilized here. Monthly mean soil moisture is obtained for the specified gridpoints for the 6-month antecedent period to the spring season and also spring of each year. The soil moisture content variable is valid at the midpoint of the 0-100 cm soil layer for monthly means and is reported as a volumetric fraction. NARR soil moisture has been shown to be a viable product and an improvement over predecessor reanalysis products [26]. Soil moisture content, as opposed to precipitation data or drought indices, is selected because future satellite (SMAP mission), in-situ, and reanalysis datasets (NARR, MERRA) will provide similar spatio-temporal measurements, which facilitates a more meaningful comparison. However, future research might consider repeating this methodology with more traditional drought indicators as well.

Table showing monthly count of Fujita categorized tornadoes in the southeastern United States, 1980-2006.

Table 1. Monthly count of Fujita categorized tornadoes in the southeastern United States, 1980-2006.

Drought periods are calculated from the soil moisture content data using a modified z-score. The average (m) and standard deviation (ìÄ) of the soil moisture values for the antecedent period (i.e., the previous fall and winter months) are calculated, and the average (åµ) over all of the years (for the respective antecedent period) is calculated in order to find the z-score (z=(x-m)/ìÄ). The calculations are repeated for the climate divisions. Once a z-score is assigned to each division-averaged period, a logical test (e.g. an IF-THEN statement) is performed to identify which periods had z-scores below -0.75 standard deviations and below -1.0 standard deviations. Both standard deviations are used to determine the correct threshold for categorizing drought conditions by comparing them to the Palmer Z-index drought periods. Although the current normal period is defined between 1970-2000, the NARR data are not available until 1979, so years 1980-2006 are used to normalize the data. Based on the six climate zones, it is determined that normalized soil moisture values below -1.0 indicate drought periods most appropriately and are used as the drought threshold.

Table showing Attributes of drought and non-drought years.

Table 2. Attributes of drought and non-drought years.

To analyze drought periods in relation to tornado days, the soil moisture values are averaged over the study area for each 6-month antecedent period, and a z-score is calculated for each period. The percent of normal tornado days is found by averaging the tornado days (per spring season) over the 27-year period, then dividing each seasonal value by the total average and multiplying by 100%. The soil moisture departures are correlated with percent of normal tornado days using linear regression analysis and tested for significance with a t-test. Fig. 3 clearly reveals that significantly negative antecedent soil moisture departures (e.g., drought) are almost always associated with below normal tornado days during spring. Table 2 summarizes the attributes of all drought and all non-drought years. The drought years are all associated with equal to or below normal tornado days. Drought years have an average of 4.17 tornado days per tornado season, while non-drought years have an average of 5.19 tornado days per tornado season (Table 4). The results presented herein are extremely consistent with results reported in [21] although different methods and data (i.e., precipitation-driven soil moisture rather than precipitation departure) over a larger area were applied. There seems to be a consistent signal when very dry or drought conditions are present. These results are also consistent with others that indicate an asymmetry in the influence of soil moisture. [27], using NCAR CCM3, reported that dry anomalies had larger impacts on the precipitation feedback than wet anomalies in the central United States. [28] shows the moisture recycling ratio in the central United States was above average in the dry summer of 1988 and below average in the wet summer of 1993, suggesting soil moisture exerts more influence on the atmosphere during droughts. The consistency of the results herein and those of [21] suggest that at least for the southeastern United States, there may be seasonal predictive capacity for tornado days when drought is present the previous season.

Table showing Relationship between normalized tornado days and 6-month antecedent soil moisture 1980-2006.

Fig. 3. Relationship between normalized tornado days and 6-month antecedent soil moisture 1980-2006. The vertical line is normal tornado days (per tornado season) while the horizontal line is normal soil moisture.

In an unpublished analysis, co-author Shepherd recently evaluated the lack of tornadic activity during season 1 of the NSF field experiment VORTEX-2 and found significantly low antecedent soil moisture in the southern part of the VORTEX study area. Season 2 was more active and was associated with above normal antecedent soil moisture in the same study region. These results hint at possible soil moisture gradient-convective relationships, but this is certainly speculative at this stage. We are not trying to suggest a first order effect of the soil moisture on VORTEX conditions, as there were certainly other major factors.

Is Soil Moisture Memory The Reason?

As stated, some studies have shown that soil moisture conditions create ‰ÛÏsoil moisture memory‰Û that affects future convective activity ([29], [30], [31]). Due to the large heat capacity of the ocean, anomalies in ocean characteristics change slowly. Similarly, land surface anomalies change slowly relative to the atmosphere, but are not well understood [32].

Our initial analysis shows the asymmetric relationship between tornado days and antecedent soil moisture is explained by soil moisture memory arguments. Six-month antecedent drought periods correspond to the driest spring soil moisture regimes, while dry to normal years correspond to either dry or wet springs. The wettest antecedent periods are associated with wet springs (Fig. 4).

Figure showing a a scatterplot of 6-month antecedent soil moisture (x-axis) and spring soil moisture (y-axis).  Drought years are denoted by red boxes. Quadrants are defined by the bold lines:  Top left (I), top right (II), bottom left (III), and bottom right (IV). The dashed red line represents the threshold for antecedent drought as defined in the study.

Fig. 4. Scatterplot of 6-month antecedent soil moisture (x-axis) and spring soil moisture (y-axis). Drought years are denoted by red boxes. Quadrants are defined by the bold lines: Top left (I), top right (II), bottom left (III), and bottom right (IV). The dashed red line represents the threshold for antecedent drought as defined in the study.

A closer look at the soil moisture memories reveals some very interesting results. A soil moisture persistence scatterplot is evaluated to see if drought conditions in fall/winter are a good predictor of drier soil moisture conditions the following spring during the 27-year study period (Fig. 4). Six-month antecedent drought values are outlined in red. An important result from this analysis is when the antecedent fall/winter soil moisture is under drought, such conditions are extremely likely to persist into the spring. Since our previous results associate drought with reduced tornado days, this suggests that extremely dry antecedent drought fall/winter periods are very good predictors of below normal tornado days the following spring. Guidelines for predicting spring soil moisture from antecedent conditions are summarized in the table below figure 4. The values tend to cluster in quadrants I, II, and III. The driest antecedent periods (e.g., drought) correspond to the driest spring values, while marginally dry to normal years correspond to either dry or wet springs. The wettest antecedent periods are almost always associated with wet springs. The dry soil results agree with previous literature, however the wet soil results are more surprising. [29], [27], and [33] found that dry soil has a more lasting memory than wet soil. It would be expected that wet antecedent soil does not have a predictable spring soil moisture anomaly as suggested by the tornado day analysis.

Concluding Thoughts

Using soil moisture memory and persistence arguments, the results have provided some insight on why drought in the previous fall/winter appears to be a predictor of below normal tornado days. In other words, the extremely dry antecedent conditions are very likely to linger into the following season whereas marginally dry conditions are not. This analysis indicates the persistence of soil moisture conditions for antecedent drought and wet soil cases, respectively. The hypothesis is anchored to the notion of precipitation and moisture recycling. Precipitation recycling is an understanding of the water cycle from an atmospheric viewpoint. On a regional scale, there is probability that a water molecule evaporated at the surface will precipitate within the same region [32]. This notion is extended to assume that soil moisture persistence can be represented by the same framework for convective storms.

Factors that promote tornadic activity include southwesterly wind shear, dry transients (intrusion) at the mid-troposphere, moist transients at low levels, and an increase in Convective Available Potential Energy (CAPE) [34]. The results presented herein should not be misinterpreted to mean that any given tornadic storm is associated with soil moisture. Certainly, larger scale dynamic and meteorological conditions are major factors. Additionally, the reader is reminded that we looked at a tornado day metric rather than counts. From a climatological perspective, there may be some guidance in seasonal predictability, in the same manner that forecasters predict hurricane season activity based on west African rainfall, sea surface temperatures, and other variables.

Image of NASA's Soil Moisture Active Passive (SMAP) Mission. Source: NASA

Fig. 5. NASA’s Soil Moisture Active Passive (SMAP) Mission. Source: NASA

This research suggests that soil moisture content provides an adequate measure of drought conditions in the southeastern United States and may be utilized in future studies. Use of soil moisture alone to assess drought is spatially and temporally more compatible with current and future satellite and in-situ datasets (Fig. 5).

The main weaknesses of this study are that convective potential is affected by both surface-based and mid-level processes, surface-forced convection is not the only scenario for tornadic storms to form. Additionally, we did not rigorously evaluate large scale dynamic conditions. However, our composite analysis of 500-hPa heights from NARR (not shown) did show more ridging during the drought years of the study.

Our ongoing studies and future work will include a more robust assessment of the role of El Nino and La Nina on dry/wet conditions, analysis in other geographic regions (i.e., traditional ‰ÛÏtornado alleys‰Û), and consideration of tornado counts, if temporal and observational biases can be properly considered. Even with the limitations, the potential predictability of spring tornado season intensity (i.e., tornado days) based on fall/winter drought conditions is worth continued pursuit.

Acknowledgements. The authors acknowledge the support of the NASA Water and Energy Cycle program for funding that partially supported this research. We also thank one anonymous reviewer for very useful insight and comments.

References Cited

[1] Trenberth K.E., et al., ‰ÛÏObservations: Surface and atmospheric climate change. In: Climate Change 2007: The Physical Science Basis‰Û Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2007.

[2] Pielke, R. A., ‰ÛÏInfluence of the spatial distribution of vegetation and soils on the pre-diction of cumulus convective rainfall,‰Û Rev. Geophys., 39, 2001, 151-177.

[3] Basara, J. B., and K. C. Crawford, ‰ÛÏLinear relationships between root-zone soil moisture and atmospheric processes in the planetary boundary layer,‰Û J. Geophys. Res., 107, 2002, 1-18.

[4] Koster, R. D., M. J. Suarez, R. W. Higgins, and H. van den Dool, ‰ÛÏObservational evidence that soil moisture variations affect precipitation,‰Û Geophys. Res. Lett., 30, 1241, 2003, doi:10.1029/2002GL016571.

[5] Findell, K. L., and E. A. B. Eltahir, ‰ÛÏAtmospheric controls on soil moisture-boundary layer interactions. Part I: Framework development,‰Û J. Hydrometeor., 4, 2003, 552-569.

[6] Findell, K. L., and E. A. B. Eltahir, ‰ÛÏAtmospheric controls on soil moisture-boundary layer interactions. Part II: Feedbacks within the continental United States,‰Û J. Hydrometeor., 4, 2003, 570-583.

[7] Dirmeyer, P. A., ‰ÛÏThe hydrologic feedback pathway for land-climate coupling,‰Û J. Hydrometeor., 7, 2006, 857-867.

[8] Niyogi, D., T. Holt, S. Zhong, P. C. Pyle, and J. Basara, ‰ÛÏUrban and land surface effects on the 30 July 2003 mesoscale convective system event observed in the Southern Great Plains,‰Û J. Geophys. Res., 111, 2006, D19107, doi:10.1029/2005JD006746.

[9] Taylor, C. M., and R. J. Ellis, ‰ÛÏSatellite detection of soil moisture impacts on convection at the mesoscale,‰Û Geophys. Res. Lett., 33, 2006, L03404. 10.1029/2005GL025252.

[10] Siqueira, M., G. Katul, and A. Porporato, ‰ÛÏSoil moisture feedbacks on convection triggers: The role of soil-plant hydrodynamics, ‰ÛÏJ. Hydrometeor., 10, 2009, 96-112.

[11] Frye, J.D., and T. L. Mote, ‰ÛÏConvection Initiation along Soil Moisture Boundaries in the Southern Great Plains,‰Û Mon. Wea. Rev., 138, 2010, 1140‰ÛÒ1151.

[12] Hauck, C., B., L. Krauss, and N. Kalthoff, ‰ÛÏSoil moisture variability and its influence on convective precipitation over complex terrain,‰Û Quart. J. Roy. Meteorol. Soc., 137, 2011, 42‰ÛÒ56. doi: 10.1002/qj.766.

[13] Hohenegger, C., P. Brockhaus, C.S. Bretherton, and C. SchÌ_r, ‰ÛÏThe Soil Moisture‰ÛÒPrecipitation Feedback in Simulations with Explicit and Parameterized Convection,‰Û J. Climate, 22, 2009, 5003‰ÛÒ5020.

[14] Galway, J.G., ‰ÛÏRelationship between precipitation and tornado activity,‰Û Water Resources Res., 15, 1979, 961-964.

[15] Lanicci, J.M., T. Carlson, T. Warner, ‰ÛÏSensitivity of the Great Plains severe-storm environment to soil-moisture distribution,‰Û Mon Wea Rev, 115, 1987, 2660-2673.

[16] Chang, J.-T., and P. J. Wetzel, ‰ÛÏEffects of spatial variations of soil moisture and vegetation on the evolution of a prestorm environment: A numerical case study,‰Û Mon. Wea. Rev., 119, 1991, 1368-1390.

[17] Fast, J. D., and Michael D. McCorcle, ‰ÛÏThe effect of heterogeneous soil moisture on a summer baroclinic circulation in the central United States,‰Û Mon. Wea. Rev., 119, 1991, 2140-2167.

[18] Holt T., D. Niyogi , F. Chen, M. A. LeMone, K. Manning, A. L. Qureshi, ‰ÛÏEffect of Land – Atmosphere Interactions on the IHOP 24-25 May 2002 Convection Case,‰Û Mon. Wea. Rev., 134, 2006, 113-133.

[19] Raddatz R.L. and J.D. Cummine, ‰ÛÏInter-annual variability of moisture flux from the prairie agro-ecosystem: impact of crop phenology on the seasonal pattern of tornado days,‰Û Bound.-Layer Meteor., 106, 2003, 283‰ÛÒ95.

[20] Hanesiak J., A. Tat and R. Raddatz, ‰ÛÏInitial soil moisture as a predictor of subsequent severe summer weather in the cropped grassland of the Canadian Prairie provinces,‰Û Int. J. Climatol., 29, 2009, 899‰ÛÒ909.

[21] Shepherd, M., D. Niyogi, and T. Mote, ‰ÛÏA seasonal-scale climatological analysis correlating spring tornado activity with antecedent fall-winter drought in the Southeastern United States,‰Û Env. Res. Lett., 4, 2009, 24012, doi:10.1088/1748-9326/4/2/024012.

[22] Doswell III C.A, and D.W. Burgess, ‰ÛÏOn some issues of United States tornado climatology,‰Û

Mon Wea Rev, 116, 1987, 495-501.

[23] Grazulis, T.,‰ÛSignificant Tornadoes, 1680-1991,‰Û Environmental Films, 1993, 1-1326

[24] Ashley, W.S., ‰ÛÏSpatial and temporal analysis of tornado fatalities in the United States: 1880-2005,‰Û Wea Forecasting, 22, 2007, 1214-1228.

[25] Ek MB, K.E, Mitchell, Y. Lin, E. Rogers, P. Grummann, V. Koren, G. Gayno, and J.D. Tarpley, ‰ÛÏImplementation of Noah land-surface model advances in the NCEP operational mesoscale Eta model,‰Û J Geophys Res., 2003, 108:8851. doi:10.1029/2002JD003296.

[26] Miguez-Macho G, H. Li, and Fan Y, ‰ÛÏSimulated water table and soil moisture climatology over North America.‰Û Bull Am Meteorol Soc., 5, 2008, 663‰ÛÒ672.

[27] Oglesby R.J., S. Marshall, D. Erickson, J. Roads , F. Robertson, ‰ÛÏThresholds in atmosphere-soil moisture interactions: Results from climate model studies,‰Û J Geophys Res., 107, 2002, 4224. doi:10.1029/2001JD001045.

[28] Bosilovich, M., and S. Schubert, ‰ÛÏPrecipitation Recycling over the Central United States Diagnosed from the GEOS-1 Data Assimilation System,‰Û J. Hydromet., 2, 2001, 26-35.

[29] Rowntree, P. R., and J. A. Bolton, ‰ÛÏSimulation of the atmospheric response to soil moisture anomalies over Europe,‰Û Quart. J. Roy. Meteorol. Soc., 109, 1983, 501‰ÛÒ526. doi: 10.1002/qj.49710946105.

[30] Pal, J.S., and E.A.B. Eltahir, ‰ÛÏPathways relating soil moisture conditions to future summer rainfall within a model of the land-atmosphere system,‰Û J. Climate, 14, 2001, 1227-1242.

[31] Wu, W., and R.E. Dickinson, ‰ÛÏTime scales of layered soil moisture memory in the con-text of land atmosphere interactions‰Û J. Climate, 17, 2004, 2752-2764.

[32] Dirmeyer, P.A., C.A. Schlosser, and K.L. Brubaker, ‰ÛÏPrecipitation, recycling and land memory: An integrated analysis,‰Û COLA Tech. Rep. 257, 2008, 24 pp.

[33] Meng, L., ‰ÛÏExamining the relationship between antecedent soil moisture and summer precipitation in the U.S. Great Plains,‰Û Ph.D dissertation, Texas A&M University, 2009, 130 pp.

[34] Munoz E, and D. Enfield, ‰ÛÏThe boreal spring variability of the Intra-Americas low-level jet and its

relation with precipitation and tornadoes in the eastern United States,‰Û Clim Dyn, 36, 2011, 247-259. doi:10.1007/s00382-009-0688-3.

Ms. Theresa Andersen is a doctoral student at the University of Georgia, where she also received an MS degree in Geography (Climatology sub-discipline). She received her BS degree from Iowa State University.

Dr. J. Marshall Shepherd is a full professor of geography/atmospheric sciences at the University of Georgia. He conducts research, advising, and teaching in atmospheric sciences, climatology, water cycle processes and urban climate systems. Prior to joining the UGA faculty, Dr. Shepherd spent 12 years as a research meteorologist at NASA. Dr. Shepherd also was Deputy Project Scientist for the Global Precipitation Measurement (GPM) mission. For his work on urban climate, Dr. Shepherd was honored in 2004 at the White House with the Presidential Early Career Award (PECASE) for pioneering scientific research. Dr. Shepherd is a Fellow of the American Meteorological Society and has over 60 publications (referred articles, book chapters, reports). He is an editor for the Journal of Applied Meteorology and Climatology and co-section editor (climatology) for the journal Geography Compass. Dr. Shepherd is also the author of the forthcoming textbook ‰ÛÏThe Urban Climate System.‰Û Dr. Shepherd received his B.S., M.S. and PhD in physical meteorology from Florida State University.