This project used an observational dataset to quantify the trends since 1930 in the characteristics of summertime EHEs. The dataset was a combination of the USHCN v2 monthly and daily datasets; similarly to how Hamlet and Lettenmaier and Maurer et al (2002) did. The 1930-1970 and 1970-2010 periods were quantified seperately and compared with eachother and the 1930-2010 trends. EHEs with requirements on only the daily minimum temperature, only the daily maximum temperature and both daily extreme temperatures were quantified (i.e. three EHE types) and the trends compared. Results indicated that a decrease in EHEs occured during the 1930-1970 period, an increase during the 1970-2010 period and during the 1930-2010 period it depended on EHE type. Maps created indicated substantial spatial structure over the CONUS with general cooling/lack of warming experienced in the geographic center of the CONUS.

This study was devided into three periods (1930-1970, 1970-2010 and 1930-2010); all of which used a different set of stations. The stations for each period were required to meet strict requirements. A few examples are, a) all 5 summer months (M,J,J,A,S) must have Tmin and Tmax data present for each year considered "available" , b) does not have more than 6 consecutive missing years of available data during the period , c) must have available data for at least 80% of the years in the period , d) must have available data for the first or second and last or second-to-last years in the period , e) not be of lowest site-quality ranking according to the surfacestations.org project and f) at least 80% of the available months in the period must be original values (i.e. not estimated)

There is no universal EHE definition, however the definition should be specific to the study goals. Most studies require the daily high to be elevated but rarely do studies require both the daily high and daily low to be elevated. Many past EHEs with intense mortality impacts (France 2003, St. Louis 1966, Chicago 1995, Russia 2010) had elevated daily highs and lows. This study separately quantified EHEs requiring a) only daily minimum compliance (Tmin type), b) only daily maximum compliance (Tmax type) and c) both minimum and maximum compliance (Tmin&Tmax type). In this study, EHE type refers to one of these three types of EHEs.

The general definition was two consecutive dates with the percentile exceeding the 92.5th percentile and subsequent maintenance of an EHE-average percentile above the 92.5th percentile. Once the EHE-average percentile dropped below 92.5 the EHE was over. Which daily extreme depends on EHE type, as mentioned above.

This study needed a robust description of how EHEs changed over the past 81 years. Thus the trends in eight different characteristics specifically relevant to end user's were investigated. The table above lists seven of the eight EHE characteristics as well as the station sample means and standard deviations of those temporal means.

Preliminary investigation led us to characterizing the signal to noise ratio and perctange of stations with specific types of trends. No gridding or spatial averaging was needed to do this however the results are still telling. The above table exhibits the EHE characteristic averaged values however the non average values can be found here and the time-period averaged values can be found here. The EHE characteristic averaged values suggest a few things: 1. The mean-to-spatial variability ratio of the Tmax (Tmin) EHE type trends was larger for all variables in the 1930-1970 (1970-2010) period and the Tmin&Tmax type EHE trends in the 1930-2010 period had especially low values. 2. All EHE types had more stations with significant negative trends during the 1930-1970 period, and more stations with significant positive trends during the 1970-2010 period. During the 1930-2010 period the EHE type determined the prevailing sign.3. The Tmin&Tmax EHE type during the 1930-2010 period had more evenhanded numbers of positive and negative trends than any other EHE type during a specific period.

The time-period averaged table highlights some differences between the EHE characteristic trends. 1. The mean onset date often behaves differently than the other metrics and generally had less stations with significant trends. 2. The two EHE duration characteristics have smaller signal to noise ratios and less stations with significant trends - albeit the significance of our results is a large question.

The CONUS spatial average trend results are shown in the above graph. Regardless of time period or metric, the trends were different for all three types of EHEs. For all EHE types, there was a general decrease in the 1930-1970 period and increase in the 1970-2010 period. For the Tmin (Tmax) type EHE, the increase during the 1970-2010 period was larger (smaller) than the 1930-1970 decrease. For the Tmin&Tmax type of EHE, the two time periods had similar increasing and decreasing trend magnitudes.

There existed variability from station-to-station that manifested itself as spatial noise in each map, but there also existed regional scale variability. The regional spatial structure between different EHE types was notable but not as large as the difference between time periods. The regional spatial structure was hardly different between different EHE characteristic metrics with the same period and EHE type, and so only the EHE characteristic-wide general spatial patterns are described below. The geographical regions referenced are simply the East (north and south), Central (north and south) and West (north and south). Lastly, the terms “decrease” and “negative” will refer to a lessening, weakening or shortening of the metrics, “increase” or “positive” will refer to an increasing, strengthening or elongating of the metrics and “neutral” will refer to a zero overall magnitude trend of the metrics.

The first result elaborated upon is the Tmin type EHEs during the 1930-1970 period where the majority of the Central and East regions show significant negative trends. However, small areas near Texas, Louisiana, Florida, Rhode Island and North Carolina were positive. The south West region exhibited significant trends of both signs and the north West region was the most consistently positive region.

The next result, the Tmin type EHE during the 1970-2010 period, exhibited overwhelming warming in the southern half of the CONUS that took on a loose horseshoe shape. The non-positive region had lack of significance and trends of mixed sign. However there did exist two smallish areas which displayed stations with significant negativity in some of the EHE characteristics that were centered on Idaho and Minnesota.

In the 1930-2010 period the Tmin type EHEs displayed a general horseshoe shape of significant positive trends, with the area from north Texas to Minnesota (occasionally the area to the east of that region as well) being excluded. Often this area would be predominantly negative with the most negativity in the area focused around Iowa. Also the positive trends in the East region were usually larger than the trends in the West region.

For the Tmax type EHEs during the 1930-1970 period, the north Central and East regions exhibited consistent negative significant trends. There was an area between Iowa/Nebraska and Indiana that had particularly strong negative trend magnitudes. The West region (e.g. west of Kansas) and south Central region had mixed sign trends; the north West region was the most positive and south Central the most negative.

During the 1970-2010 period a weak horseshoe shape was exhibited, with positive trends in the West, Southwest, South, Southeast and Northeast areas. The area of negative trends encompassed the eastern portion of the West North Central, the western portion of the East North Central and Illinois. More significant negative trends were seen in this area and the area extended more to the west and south than in the Tmin type EHEs during the same period. The Northwest and Northeast regions had the most inconsistent trends and the Southwest region displayed the most predominantly positive significant trends.

During the 1930-2010 period the Tmax EHE characteristic trends generally were positive in the West and negative the Central and East regions. The area from Minnesota east to Pennsylvania and south through Tennessee and Oklahoma exhibited systematic and significant negative trends. Small areas focused on New England, Florida and even Louisiana indicated positive significant trended areas in the Central and East. The West region however was vaguely mixed with a few stations with significant negative trends amidst many positive trends.

The Tmin&Tmax type EHEs during the 1930-1970 period displayed organized significant negative trends in the East and north Central regions with particularly large magnitudes from Indiana to Nebraska. Good portions of Florida and New England lacked significant trends, however. The area southwest of a hypothetical line between Louisiana and Washington was well mixed in sign, with the most consistently positive trends nearer to Washington.

During the 1970-2010 period the Tmin&Tmax type EHEs again displayed a loose horseshoe shape. An area of negative trends existed in the north Central region, however this area had less significant trends, magnitude and spatial extent than the corresponding area in the Tmax type EHE. The West, south Central and Eastern regions were very consistently positive with larger magnitudes in small areas around Louisiana, Arizona/New Mexico and Georgia.

For the 1930-2010 period there was a general increase in the East and West and decrease in the Central. Specifically, an area from Minnesota to Oklahoma/Texas to Ohio had coherent significant negative trends. The area centered on Iowa had particularly large magnitudes and there did not exist any significant positive trends in south Texas, but the stations in northern Wisconsin/Michigan often had significant positive trends. The West region was consistently positive, as was the majority of the East region (including Louisiana).

In the table above are displayed three values representing the three time periods (1930-1970, 1970-2010, 1930-2010), for each metric and between all three different EHE types. Regular font values indicate the means are statistically not equal at the 0.10 significance level, and bold font values signify a failure to prove they are not equal at that significance level.

Student’s t-test results showed that the means were statistically different more often than not, but the number of instances varied with EHE type. The instances are most numerous of statistically different means between Tmin and Tmax EHE characteristics. The calculated correlation coefficients between the three EHE types, for each period and EHE characteristic are also given in the above table. Results showed that the correlation coefficients between the EHE characteristic metrics of the Tmin and Tmax EHE types are small and sometimes negative. The Tmin/Tmin&Tmax and Tmax/Tmin&Tmax correlation coefficients are much stronger but still vary considerably with EHE metric and time period. The results, of both the Student’s t-tests and correlation coefficients, suggest that the Tmin has a slightly closer relationship to Tmin&Tmax than Tmax does.

Discussion

This study answered a few specific questions that it set out to answer, and those are discussed here. Relevant trends did exist since 1930 in CONUS average for Tmax and Tmin type EHEs, but not Tmin&Tmax type EHEs. The spatial structure of those trends were substantial, with regions of positive and negative trends for nearly all EHE types and periods. Between different time periods there were considerable differences in average trends and spatial patterns. Lastly, a great deal of disparity exists between decadal trends of Tmin and Tmax type EHEs and those of Tmin&Tmax type EHE trends and there is virtually no relationship between the Tmin and Tmax type EHE trends.

Interesting to the climate discussion was the EHE decrease in the central parts of the U.S. being similar to the hydrologically driven “warming hole” analyzed in the Pan et al. (2004) study. Specifically, it was expected that the Tmax EHE type would decease but we found the Tmin and Tmin&Tmax type EHEs decreased as well. Future studies focusing on the numerous linkages between the long-term hydrological cycle and EHEs trends could be invaluable. Also the trends in Tmin&Tmax type EHEs seemed to be limited by the decrease in Tmax type EHEs. Studies addressing how the Tmin&Tmax type EHEs would change without a decreasing EHE type potentially limiting it could be insightful. Lastly, our results similarity to the studies describing single day percentile exceedences suggested the trends in EHEs and percentile exceedences may not be dissimilar. If studies were to explore the relationship between extreme percentile exceedences and EHE characteristic trends, then the results of the many percentile exceedence studies could be used with more confidence in the EHE trend discussion.

Of interest to the urban planning and public health end users discussion, the trends in Tmin, Tmax and Tmin&Tmax EHE trends are not particularly the same in spatial structure or magnitude. Thus studies used as guidance with only some of these EHE types should be recognized as not fully representative of the trends in the other EHE types. Also of importance, was the spatial structure of EHE trends, for all types and periods, varies heavily across the CONUS. We suspect the relationship between mortality and each type of EHE vary on a regional basis as well. Thus it is expected in both the past and future, the amount EHEs impact population mortality varies widely across the CONUS. Last of relevance was how the similarities between the EHE characteristics gives validity to the assumption that using guidance studies with only one of two of the EHE trends could be representative of the other trends.