1. Introduction
Although daily minimum and maximum temperatures (Tmin and Tmax, respectively) are individually influenced by El Niño–Southern Oscillation (ENSO; Gershunov and Barnett 1998; Lim and Schubert 2011), ENSO’s relationship to daily temperature ranges (DTRs) has not been fully explored. DTR variability may or may not reflect maximum and minimum temperature variability. By definition, DTRs are calculated by subtracting concurrent daily Tmin from Tmax. The relationship between ENSO and these calculated DTRs is not necessarily the same as the combined relationships of individual Tmax and Tmin with ENSO. The annual cycle and trends of DTRs have been examined in several studies because of the heightened awareness of climate change in recent years (Karl et al. 1984; Leathers et al. 1998; Durre and Wallace 2001; Easterling et al. 1997; Saxena et al. 1997), but the natural climate variability of DTRs has received relatively little attention. Mote (1996) showed that monthly averaged DTRs vary with respect to the phase of ENSO during winter months at 88 southeastern U.S. (hereinafter Southeast) stations for the period of 1931–94. Wu (2010) showed broadly that eastern North American monthly mean DTRs are negatively correlated with ENSO. This study looks at daily DTRs over all seasons and the relationships among ENSO phase, humidity, and DTRs. The strong relationship between DTR and humidity allows us to use DTR as a metric at stations for which no long-term humidity data exist. This information can then be used to prepare for possible impacts on agricultural interests such as the cattle industry on the basis of the predicted ENSO phase.
DTRs in the Southeast are examined to determine how they vary in response to ENSO. ENSO is an interannual sea surface temperature (SST) signal in the equatorial Pacific Ocean that has well-documented climate impacts on the Southeast (e.g., Ropelewski and Halpert 1986; Smith et al. 1998, 1999). In the Southeast, the El Niño phase (La Niña phase) is typically associated with cool and wet (warm and dry) winters.
Southeastern DTR values can range from 0°F (0.0°C) to over 50°F (27.8°C) in a 24-h span. The moisture content of the atmosphere, which also varies with ENSO phase, is a contributing factor to how broad the DTR can be in a given day. In humid climates, atmospheric water vapor absorbs Earth’s infrared radiation and radiates a portion back to the surface, which limits overnight surface radiative cooling and increases the minimum temperature (resulting in a decreased DTR). Dai et al. (1999) showed that high values of surface specific humidity are significantly correlated (at 99% confidence) with low DTRs in the Southeast. Low DTRs can affect a variety of agricultural industries in the Southeast, especially beef and dairy cattle. In particular, low DTRs can increase heat stress in cattle. The temperature–humidity index (THI), which is calculated from ambient temperature and the percent of relative humidity (Thom 1959) is the most common index used to measure heat stress in the animal agricultural industry.
The rate of thermal exchange between the animal and the environment depends on the ability of the environment to accept heat and water vapor [see Finch (1986) for a description of radiative, conductive, evaporative, and metabolic processes]. In humid conditions, the efficiency of the animal’s evaporative cooling process decreases, thus inhibiting heat loss, which increases the core body temperature of beef cattle and reduces their appetite (Finch 1986). Small upward shifts in core temperature can also affect tissue and neuroendocrine functions, which reduces fertility, growth, lactation, and ability to work (McDowell 1972). Heat stress on dairy cattle, which can result from a variety of environmental factors (temperature, relative humidity, and radiant heat load), can potentially reduce dry-matter intake and milk yield (West 2003). Although each environmental factor affects the heat load on cattle, it has been shown that, given enough nighttime cooling, cattle can tolerate relatively high daytime temperatures (West 2003). However, if temperatures are high and DTRs are low (an occurrence that is typically associated with moist conditions; see section 5), this nighttime cooling can be inhibited. We examine the relationship between humidity and DTRs briefly in section 3.
To visualize the distributions of DTRs (overall and ENSO separated), temperature distributions are drawn from 290 Cooperative Station Network stations in the Southeast and histograms are created for four seasons. We use a conditional-ratio test to directly compare individual bins from the El Niño and La Niña DTR distributions and to perform spatial and statistical analyses. We limit our focus to comparing DTRs between El Niño and La Niña conditions and exclude neutral-phase DTRs.
Our study is unique in that it examines the relationship between ENSO and daily concurrent Tmax and Tmin values in the Southeast within particular seasons, rather than monthly extreme-temperature composites. Furthermore, our analysis is performed with a spatially larger and more recent dataset than those used in previous studies. Stefanova et al. (2013) examined the moments (mean, standard deviation, and kurtosis) of DTR distribution with respect to ENSO, whereas this study focuses on differences in the individual bins of distributions associated with El Niño and La Niña. In addition, previous studies (e.g., Gershunov and Barnett 1998; Lim and Schubert 2011; Stefanova et al. 2013; Wu 2010) examined ENSO impacts on these extreme temperatures and DTRs for primarily the winter season, whereas this study examines ENSO impacts on DTRs for all seasons.
2. Station selection, data processing, and ENSO-phase separation
The Southeast is defined in this study as Alabama, North Carolina, South Carolina, Florida, and Georgia. Maximum and minimum temperature data are taken from the National Climatic Data Center’s (NCDC) “TD3200” archived station observations (available online from http://rda.ucar.edu/datasets/ds510.0/). The dataset underwent automated and manual quality control at NCDC (Angel et al. 2003). Of the 292 stations available for the Southeast, 229 have records of 365 (or 366) days of observations from 1948 to 2009 after being filled with a regression technique by Smith (2007), as described below. Between 35 and 61 years of data are available for 61 of the remaining 63 stations; all of these 63 stations began recording observations after 1948 and have periods of record ending in 2009. The periods of record for two stations, Allendale, South Carolina, and Spartanburg, South Carolina, are 27 and 16 years, respectively. These periods of record contain a smaller sample of ENSO events and fall below the 30-yr threshold that is considered necessary to determine an accurate long-term “climatology” (World Meteorological Organization 2010). Therefore, these two stations are removed from the study, leaving 290 stations for the analysis (Fig. 1).
Locations of, and lengths of record (see legend) for, 290 stations in the Southeast. The longest lengths of record are from 1948 to 2009. All shorter lengths of record begin later than 1948 and end in 2009. The stars on the map represent the Ocklawaha station (data from 1999 to 2012) from the FAWN network and the Salisbury station (data from 1996 to 2012) from the ECONet system.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
Daily temperature values that were missing from each station’s period of record were filled at the Center for Ocean–Atmospheric Prediction Studies (COAPS) using a multiple linear regression method developed by Smith (2007). For any given day and station, detrended temperatures from surrounding stations [within a 50-mi (~80.5 km) radius] on the same day were used to estimate missing maximum or minimum temperatures. Filling missing data ensured the existence of both a maximum temperature and a minimum temperature for each day and preserved seasonal and spatial signals.
Daily Tmax and Tmin are used to compute DTRs for each of the 290 stations. The DTR is computed by subtracting the minimum temperature from the maximum temperature. These computations are reviewed to ensure that nonphysical negative DTR values are not used in the analysis. No such values exist at the 290 stations.
In addition to the data collected from these stations, hourly meteorological variables are taken from the Florida Automated Weather Network (FAWN; http://fawn.ifas.ufl.edu/) station in Ocklawaha, Florida, located in southeastern Marion County (data from 1999 to 2012) and from a North Carolina Environment and Climate Observing Network (ECONet; http://www.nc-climate.ncsu.edu/econet) station in Salisbury, North Carolina (data from 1996 to 2012). Because of their locations, these two stations are chosen to represent rural/agricultural areas across the study domain.
The Tmax and Tmin are recorded in whole degrees Fahrenheit in the DS3200. Because ranchers find the degree Fahrenheit unit to be more meaningful than degrees Celsius or kelvins, the data and analyses are presented in whole degrees Fahrenheit. Comparable degrees Celsius, with two significant figures, are presented in parentheses when a Fahrenheit value is given in the text.
For each station time series, DTR values are binned in 5° Fahrenheit (2.8°C) increments, beginning with the bin containing 0°–4°F, inclusive (0.0°–2.2°C). These bins are referred to by the median value in the 5° range, called the bin number. For example, bin number 2 and bin number 12 refer to the DTRs bound by 0° and 4°F (0.0°–2.2°C) and by 10° and 14°F (5.6°–7.8°C), respectively. Bin ranges are depicted hereinafter with square brackets enclosing the Fahrenheit values to indicate that the upper and lower bounds are inclusive. DTR values beyond bin number 42, which contains [40°–44°F] (22°–24°C), are binned in bin number >44. The bin count is defined as the total number of occurrences of a particular DTR in the bin range.
Bin-count distributions are separated by season and ENSO phase. Three-month seasons are defined for winter [December, January, and February (DJF)], spring [March, April, and May (MAM)], summer [June, July, and August (JJA)], and autumn [September, October, and November (SON)]. Historical ENSO phases are classified using the Japan Meteorological Agency (JMA) SST index. Previous studies (e.g., Sittel 1994a,b; Green 1996) used 12-month “ENSO years”; however, in this study we assign ENSO phase on a month-by-month basis. The JMA index is the monthly sea surface temperature anomalies averaged over the area from 4°N to 4°S and from 150° to 90°W and then smoothed with a 5-month running mean. For an event to qualify as El Niño (La Niña), the JMA index must be at or above (at or below) 0.5°C (−0.5°C) for five consecutive months. Once these persistence criteria are met, then any and every month of the consecutive string that exceeds the ±0.5°C threshold is classified as being in the appropriate ENSO phase. All months in which the JMA index is between 0.49°C and −0.49°C are classified as the neutral phase. Selection of a specific ENSO index can be problematic since the common indices have varying sensitivities to El Niño and La Niña events (Hanley et al. 2003). The selection of a JMA-based index provides better sensitivity to La Niña events (Hanley et al. 2003), which are known to have more severe impacts on Southeastern agricultural communities (primarily limited water resources due to drought conditions). In addition, the monthly classification scheme used herein matches well with the previous yearly classification in the Northern Hemisphere winter months, when El Niño and La Niña events tend to reach peak strength. There are, however, substantial classification differences in the spring and summer transition months. The monthly classification scheme is more representative of the true state of the ocean and atmosphere during these transition months than are seasonal ENSO indices.
Using this alternate JMA method allows a single season during a given year to contain multiple phases of ENSO. For example, in the 1987/88 DJF season, December 1987 and January 1988 are categorized by JMA as an El Niño whereas February 1988 is categorized by JMA as neutral. Within the period 1948–2009, only a single 3-month span (1973 MAM) has both El Niño– and La Niña–associated months, because a complete transition from a La Niña event to an El Niño event (or a La Niña event to an El Niño event) in 3 months is rare. Of the 744 months in the temporal domain (1948–2009), there are 170 El Niño months, 160 La Niña months, and 414 neutral months (Table 1). Each month has a comparable number of El Niño and La Niña events, excluding March, because El Niño events typically decay in February (i.e., 6 of the 14 events studied), and excluding JJA months, when El Niño events occur more often than La Niña events. Normalization before analysis (section 4) accounts for these differences.
The number of months, out of 744, associated with each ENSO phase per calendar month from 1948 to 2009.
3. Histograms
Bin-count distributions in a histogram format provide a clear representation of DTRs during each ENSO phase. Visual inspection of the overall DTR histograms suggests a Gaussian distribution of DTRs in the Southeast (Fig. 2). Comparison of phase-separated histograms reveals differences in the distributions associated with El Niño and La Niña.
Overall DTR histogram for Ocala, showing the total bin counts in each bin number recorded for 1948–2009.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
Overall histograms are created for each station to gather a first impression of the DTR distributions in the Southeast. The historical DTR distribution for the Ocala, Florida, station, located within Florida’s primary ranching region, is presented as an example (Fig. 2). ENSO-phase-separated histograms are then created for each station and season. The dark bars (light bars) in Fig. 3a show bin counts of DTRs for each bin number recorded during a La Niña (El Niño) SON, at the Ocala station.
Seasonal ENSO-phase-separated histograms for Ocala for 1948–2009. Dark bars (light bars) are associated with DTRs recorded during La Niña (El Niño).
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
Histograms for Ocala in Figs. 2 and 3 are representative of the distributions in central Florida and much of the Southeast. The DJF, MAM, and SON La Niña (El Niño) distributions at Ocala tend toward higher (lower) DTRs. While the peaks of each distribution are found in bin numbers 22 or 27, the distributions are either positively skewed (El Niño) or negatively skewed (La Niña). This results in differences between La Niña and El Niño bin counts in the higher and lower bins. For example, DJF bin number 12 contains 141 counts during the El Niño phase and 81 counts during the La Niña phase. Of the 222 occurrences of a DTR of [10°–14°F] (6.1°–8.3°C) during either nonneutral phase of ENSO, 64% (36%) were recorded during an El Niño (La Niña). These percentages indicate the phase in which a DTR is historically more likely to have occurred.
Note that the MAM distribution appears to show the greatest difference in distributions between the warm and cold phases of ENSO whereas DJF shows a moderate difference and SON shows a small difference. The JJA distribution presents two distinctions: 1) much higher kurtosis (and less Gaussian distribution) and 2) a majority of counts related to the El Niño distribution. We hypothesize that this occurs because of increased ambient moisture content in the lower atmosphere that results in a narrow range of commonly occurring DTRs and the greater number of El Niño months in JJA (see section 1), respectively. Many histograms for the Southeastern stations have results similar to the Ocala histogram. Visual inspection of individual station histograms, however, is not an effective method to generalize distribution differences and comparisons. Instead, a quantified measure of distribution differences may reveal a coherent relationship between ENSO and Southeastern DTRs.
4. Conditional ratios
Further quantifying of the distribution shifts is accomplished by developing a ratio between the ENSO-phase-separated histograms. To ascertain whether two samples are drawn from different distributions, the Kolmogorov–Smirnov test could be used. However, here we are interested not in whether the El Niño and La Niña distributions are different in their entirety, but in whether there are significant differences in counts for particular bins away from the mean or median between the two ENSO phases. We use a conditional ratio (CR) to compare the normalized distributions associated with La Niña and El Niño for each specific bin. The histogram results suggest that higher (lower) DTRs are associated with La Niña (El Niño) events more frequently than with El Niño (La Niña) events. A direct comparison implementing the conditional-ratio statistic reveals whether this result is valid on a larger scale. A bootstrapping technique is employed to establish confidence in the results and to evaluate statistical significance.
a. Method
The CR is a metric of the normalized bin counts associated with El Niño and La Niña that allows evaluation of the spatial extent of ENSO influence on DTRs across the Southeast. Normalization of each station’s bin counts is necessary to ensure that the bins are compared relative to the entirety of a given phase’s distribution. For example, during DJF at Ocala (Fig. 3b) there are 1330 total counts in all bins during the La Niña phase. Because 306 of them reside in bin number 22, ~23% of the distribution for the given variables (station: Ocala; season: DJF; phase: La Niña) is located in bin number 22. If, however, the total number of counts were doubled to 2660 while the bin counts in bin number 22 remained at 306, then the percentage of the distribution located in bin number 22 would be one-half of the original percentage, or ~12%.
We define the CR such that values above unity indicate that the La Niña relative frequency is higher whereas values below unity indicate that the El Niño relative frequency is higher. Because the CR is balanced at unity, magnitudes on either side of unity are not intuitively comparable. For example, for a particular bin number, station, and season, if the relative frequency for a La Niña–associated bin is 4 times the relative frequency for an El Niño–associated bin, then the CR will be 4.0. This is defined as a La Niña–dominated CR value. Alternatively, if the relative frequency for an El Niño–associated bin is 4 times the relative frequency for a La Niña–associated bin, then the CR will be 0.25. This is defined as an El Niño–dominated CR value.
Since the CR is not linear across unity, direct comparison of the magnitude of a CR when El Niño dominates the CR signal with the magnitude of a CR when La Niña dominates the CR signal is accomplished by inverting El Niño–dominated CRs [defined as inverted conditional ratios (ICR)]. Because an El Niño–dominated CR value is always less than 1.0, its corresponding ICR value is always greater than 1.0. ICRs can then be directly compared with La Niña–dominated CRs. In section 4b, an El Niño–dominated CR value will often be presented as an ICR to aid comparisons with a La Niña–dominated CR value.
A minimum bin-count requirement is placed on the CR computations. For a given season, station, and bin number, the sum of the counts during La Niña and during El Niño must be at least 20. A sample size of 20 counts is considered to be sufficiently large to maintain confidence in the statistical results. If the total count in these bins fails to meet this minimum count requirement, then the CR is not computed and its value is recorded as missing. Statistical significance testing (see below) ensures that undersampling biases are accounted for.
If, for a given season, station, and bin number, the El Niño (La Niña) relative frequency were near infinity or the La Niña (El Niño) relative frequency were near zero, then the CR would be near zero (infinity). In this study’s results, there are no CR values that pass the minimum count requirement and are also greater than 10.0 or less than 0.1.
CRs must be interpreted carefully. A CR above (below) unity could be a result of a comparatively small (large) El Niño relative frequency, a comparatively large (small) La Niña relative frequency, or a combination of both. Thus a CR that is different from unity simultaneously represents the movement of one phase’s distribution into a given bin and the movement of another phase’s distribution out of that bin. These movements or shifts can be considered as relative to the overall distribution (i.e., including the neutral-phase observations). For example, for a given bin both El Niño and La Niña could have identical small relative frequencies (their distributions have shifted away from that bin relative to the corresponding neutral-phase frequency), resulting in a CR of perfect unity. This example shows that a direct interpretation of the CR is challenging. Clearly, both a distribution shift away from a bin and a distribution shift toward a bin can affect the CR. To determine which is occurring (shifts toward or away), CRs are computed by comparing El Niño and La Niña relative frequencies with neutral-event relative frequencies. Although we are not directly interested in the relationship between the warm and cold ENSO phases versus the neutral phase (see section 1), it provides a baseline to compare the shifts, and results from these computations (not shown) are used in the CR interpretation.
To assess the significance of the CRs, we use bootstrapping, a nonparametric approach involving sampling with replacement (Efron and Gong 1983). Here, bootstrapping is used to construct 1000-member resampling distributions to determine whether a particular CR is significantly different from unity at 95% confidence. Values that are significantly greater (less) than unity indicate a CR with significantly more La Niña (El Niño) DTR counts than the corresponding El Niño (La Niña) DTR counts.
Bootstrapping results reveal that the greatest number of stations with significant CRs is found in bin numbers 7 or 12 (bin numbers 32 or 37) when El Niño (La Niña) dominates the CR signal (Table 2). Instead of presenting every combination of bin numbers and seasons, we limit our attention to bin numbers 12 and 32 to highlight the most interesting results. Bin numbers 12 and 32 have higher bin counts than do bin numbers 7 and 37 and are less likely to be affected by the 20-count requirement described above. Bin number 22 is used for comparison because 1) it has few significant stations, providing the cleanest cutoff between results in higher and lower bins (Table 2); 2) it is often the median bin for overall bin counts; and 3) it is equally distanced between bin numbers 12 and 32. Not all DTRs in the Southeast that may be affected by ENSO are analyzed using this method (e.g., bin numbers 7 and 37); however, the majority of results relevant to this study are depicted. All seasons are analyzed for bin numbers 12, 22, and 32, allowing comparisons between seasons.
The number of stations with CRs that are significantly different (95% confidence) from unity using the bootstrapping technique is shown in each column, separated by bin (2, 7, 12, …, >44), season, and ENSO phase. “Above unity” rows indicate counts of stations with significant CRs that are La Niña dominated. “Below unity” cells indicate counts of stations with significant CRs that are El Niño dominated.
b. Results and discussion
Spatial plots of CRs in bin numbers 12, 22, and 32 provide a summary of the associations between the ENSO signal and DTRs in the Southeast (Figs. 4–6). The consistent result for DJF, MAM, and SON is that the majority of stations have CRs below the unity line in bin number 12, CRs near the unity line in bin number 22, and CRs above the unity line in bin number 32.
Map of conditional ratio values for the 290 stations during the four seasons for bin number 12. Squares (circles) represent stations at which El Niño (La Niña) dominates the CR value. Filled (open) symbols indicate CRs that are (are not) significantly different from unity. Stations with missing CRs or that fail the minimum count requirement are not plotted.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
As in Fig. 4, but for bin number 22.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
As in Fig. 4, but for bin number 32.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
The results of significance testing for every bin are summarized in Table 2. The significance results for DJF and MAM show a marked shift from nearly no stations with significant CRs above unity (La Niña dominated) in bin numbers 2, 7, 12, and 17 to a number of stations with significant CRs above unity in bin numbers 27, 32, and 37. Similarly, in these months there is a shift from a number of stations with significant CRs below unity (El Niño dominated) in bin numbers 2, 7, 12, and 17 to nearly no stations with significant CRs below unity in bin numbers 27, 32, 37, 42, and >44. SON exhibits a similar separation across the DTR bins, but with fewer stations having significant CRs in each bin as ENSO impacts begin to manifest in the autumn months, and JJA shows very few significant results as ENSO impacts are weak in the summer months. Because ENSO first manifests in autumn months, it appears that CR values with significance begin in the northwestern portion of the Southeast (near the Appalachian Mountains), creeping southward and eastward during DJF and even farther south during MAM before disappearing in JJA.
Bin number 12 shows a clear CR value dominance by El Niño (Figs. 4a–d), especially during DJF and MAM. Table 2 shows that during DJF (MAM) 27% (42%) of stations are significantly below unity. In bin number 12, MAM (Fig. 4c) has the most significant stations, followed by DJF (Fig. 4b), and then SON (Fig. 4a). CRs during JJA (Fig. 4d) have no consistent pattern, with insignificant fluctuations above and below unity in each state.
The map of bin number 12 during DJF (Fig. 4b) shows strong significant results below unity throughout central Florida. The largest ICR values with significance are found between Orlando and Lake Okeechobee, where ICRs are between 2.0 and 4.0. The remaining significant CRs are mainly clustered in southern Georgia and southeast South Carolina, with significant ICRs between 1.1 and 3.0. Only 44 stations have CRs above unity (15%), with only a single significant CR at Fort Lauderdale, Florida.
The MAM map of bin number 12 (Fig. 4c) shows more significant results in Florida than the DJF map does, with spatial coverage extending throughout the state, excluding the Panhandle. There are significant ICR values between 1.1 and >4.0. A distinct swath of significance stretches from southwestern Alabama to northwestern North Carolina, the Appalachian Mountains region, with ICR values between 1.1 and 4.0. Regions of significance are also found in southern Georgia and southeastern South Carolina. The largest ICR values are once again in southern and central Florida. For both MAM and DJF there is only one station in the Florida Panhandle with a significant CR: Wewahitchka, Florida.
Bin number 12 during SON (Fig. 4a) shows that the pattern of significant CRs is focused around the Appalachians in western North Carolina, western South Carolina, and northeastern Georgia. Significant ICR values range from 1.1 to 3.0. Beyond these regions, however, there are few stations with significance.
Most CRs in bin number 22 are clustered around unity (Figs. 5a–d). An exception is the pattern in the western peninsula of Florida during MAM and a small number of other El Niño–dominated significant CRs scattered across the Southeast during MAM and DJF. The likely explanation for this is that the La Niña distribution shifts toward higher DTRs away from bin number 22 with more magnitude than El Niño’s shift toward lower DTRs. Results from La Niña versus neutral CRs and El Niño versus neutral CRs confirm this (not shown). The neutral relative frequencies significantly dominate in comparison with La Niña relative frequencies at the same stations where regular CRs are significantly below unity in bin number 22, whereas El Niño versus neutral CRs are near unity.
The signal in bin number 32 contrasts strongly with the signal in bin number 12. Many stations in bin number 32 have CRs above the unity line (Figs. 6a–d). This La Niña dominance of the CR values is shown across each of the states, and a slightly weaker magnitude is associated with the Alabama stations.
The DJF map of bin number 32 (Fig. 6b) shows 256 stations, or 88%, have CRs above unity. Only two stations, Vero Beach, Florida, and Talladega, Alabama, have ICRs that are greater than 1.1, and neither of these CRs is significant. In Florida, DJF bin number 32 has significant CRs that are somewhat similar to the DJF bin number 12 signal but that have less CR magnitude and a less coherent signal. Florida stations with significant CRs are not clustered in the center of the state but instead are spread from Lake Okeechobee to northeastern Florida and the eastern edge of the Panhandle, with many nonsignificant CRs between the stations with significant CRs. The significant CRs have values between 1.1 and 2.0. Southern Georgia has a pattern of significant CRs with values also between 1.1 and 2.0. Stations in southeastern South Carolina, eastern North Carolina, and western North Carolina all have consistently significant CRs above unity, with CR values between 1.1 and 3.0.
The MAM map of bin number 32 (Fig. 6c) shows a markedly different spatial pattern. Significance in southern and central Florida is much more prominent than in any of the other spatial patterns previously discussed, with 36 of the 62 Florida stations, or 58%, having significant CRs with values between 1.5 and 4.0. A CR value of 4.26 at Fort Myers Page Field, Florida, is the largest CR computed in this study and is significant. Stations in southern North Carolina are clustered with significant CRs between 1.1 and 2.0. Other stations with CRs significantly above unity are scattered throughout the Southeast, but there is no recognizable spatial pattern.
SON stations (Fig. 6a) show a number of significant CRs above unity in the Appalachian Mountains, similar to the pattern in bin number 12, in which CRs are below unity. The values of bin number 32 CRs range from 1.1 to 2.0. Some significant stations in eastern Florida and Georgia have CR values between 1.5 and 3.0, but this pattern does not appear to be consistent.
In summary, La Niña dominates the high-DTR bin number 32 and El Niño dominates the low-DTR bin number 12. MAM significance and CR magnitude are generally greater than DJF and SON significance and CR magnitude. DJF significant results are clustered farther southeast in bin number 32 and are more widespread in bin number 12. MAM results are widespread across the Southeast and are arranged in southwest-to-northeast-oriented swaths in both bin number 12 and bin number 32. Significant CRs during SON are clustered farther northwest than significant CRs during DJF. Although there are fewer significant stations during SON than during DJF, the CR magnitudes during both seasons are similar. JJA shows very little significance.
5. Physical discussion
The goal of this section is to show that the DTR findings are physically consistent with present understanding of ENSO impacts on the climate of the Southeast. A relationship between cloud cover/precipitation and DTRs has been established (Karl et al. 1987; Leathers et al. 1998): increased (decreased) cloud cover and daily precipitation frequency correspond to decreased (increased) DTRs because diurnal radiation is suppressed (enhanced). We hypothesize that our results are explained by variations in these climatic variables during the warm and cold phases of ENSO. Results are interpreted and understood in light of previous work identifying mean climate patterns during ENSO for winter and spring (e.g., Ropelewski and Halpert 1986; Smith et al. 1998, 1999). Autumn DTR results are less understood because little emphasis has been placed on researching autumn ENSO impacts.
Winter El Niño events are typically cooler and wetter than neutral events in the Southeast. El Niño events generally have a southwesterly upper-level jet (Smith et al. 1998), which supports more frequent midlatitude storm tracks in the Southeast (Eichler and Higgins 2006). At the same time, the Bermuda high shifts eastward over the Atlantic Ocean. Low-level southwesterly flow transports moisture from the Gulf of Mexico inland along the Gulf Coast, generating increased cloudiness and precipitation in the Southeast (Mote 1996; Smith et al. 1998) through increased low-level convergence, increased uplift from upper-level divergence, and likely isentropic lift from warm-air advection crossing surface stationary or warm-frontal boundaries. These moist conditions are accompanied by seasonal mean temperatures that are cooler than those in the neutral ENSO phase (e.g., Ropelewski and Halpert 1986; Smith et al. 1998), decreased occurrence of extreme (10th and 90th percentiles) daily mean temperature relative to ENSO cold phases (Higgins et al. 2002), and less-frequent Tmax extremes (90th percentile) as compared with cold phases (Lim and Schubert 2011).
La Niña winters are typically warmer and drier (especially in Florida) than neutral or El Niño winters. The upper-level jet stream is oriented zonally over the mid-Atlantic region of the U.S. East Coast and is shifted poleward of the jet stream during an El Niño (Smith et al. 1998). At the surface, the Bermuda high expands westward to cover the Southeast and is accompanied by southerly low-level flow over Alabama, Mississippi, and Louisiana. Southerly flow from the Gulf on the western periphery of the Bermuda high combines with upper-level positive vorticity advection and low-level convergence, resulting in increased precipitation from central Mississippi through northern Alabama and Georgia (Smith et al. 1998), whereas subsidence and dry conditions exist over the eastern and southern regions of the Southeast. In addition, during the westward extension of the Bermuda high, surface cyclones track more frequently over Mississippi and northern Alabama and Georgia (Eichler and Higgins 2006). The dry areas experience seasonal mean temperatures that are warmer than those in neutral ENSO phases (e.g., Ropelewski and Halpert 1986; Smith et al. 1998), increased occurrence of extreme (10th and 90th percentiles) daily mean temperature relative to ENSO warm phases (Higgins et al. 2002), and more-frequent Tmax extremes (90th percentile) as compared with warm phases (Lim and Schubert 2011).
Although some studies cited in the preceding paragraphs define boreal winter as January–March rather than DJF, the patterns should not be disparate. DTR results discussed in sections 3 and 4 agree with these known winter ENSO patterns. Drier conditions during La Niña allow solar radiation to efficiently heat the surface and overlying air, resulting in higher afternoon temperatures (higher Tmax), and rapid thermal radiation loss from the surface at night, yielding lower morning temperatures (lower Tmin; Zhang et al. 2011), than would occur under more humid El Niño conditions, which in turn results in higher DTRs. Bin number 32 shows more significantly large CRs during La Niña in Florida, southeastern Georgia, eastern South Carolina, and eastern North Carolina where dry conditions (low relative humidity) prevail. Conversely, during El Niño cloud cover limits DTRs to smaller values (significant CRs in DJF bin number 12) in the Southeast, especially in the peninsula of Florida, southeastern Georgia, and southeastern South Carolina. In contrast, the Florida Panhandle, northwestern Georgia, northwestern South Carolina, and northwestern North Carolina do not have significance patterns (DTRs are not as low during El Niño), likely because storm tracks in those areas (and associated cloud cover) are no more frequent during El Niño than during La Niña (Eichler and Higgins 2006).
We hypothesize that cloud cover and radiative cooling/warming are similarly driving MAM results, although future work must be conducted to determine the validity of this hypothesis. It can be surmised that the MAM significant CRs in bin number 12 (bin number 32) are related to increased (decreased) cloud cover and decreased (increased) radiative cooling/warming during El Niño (La Niña) events. In bin number 12, the increased spread of the significant CRs is likely related to spatially variable storm tracks during El Niño spring seasons whereas the significance pattern for bin number 32, which is focused in southern Florida, may result from the storm track activity being north of the state—focused in northern Georgia—during La Niña spring seasons (T. Eichler 2012, personal communication). We cannot hypothesize as to why the magnitudes are stronger for MAM bin numbers 12 and 32; future work could evaluate the physical reasons for these results.
6. Agricultural application
Understanding the relationships between potential heat stress, DTRs, and ENSO enables agricultural stakeholders to assess risk associated with the effects of low or high DTRs and to protect their economic interests. To examine the relationship between DTR values and humidity in depth at high resolution, hourly temperature and humidity data from a 14-yr period of record (1999–2012) from the FAWN station and data from a 17-yr period of record (1996–2012) from an ECONet station are put through the same analysis conducted on the NCDC data. Daily DTR values are binned in a similar manner.
a. Temperature–humidity index
Histogram plots of DTR bins and average RH values from the (a) FAWN station at Ocklawaha and (b) ECONet station at Salisbury. The bars represent the number of DTR occurrences within that bin over the station’s period of record. The line represents the trend in average RH across the DTR bins.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
Since THI depends on RH, and because DTR exhibits a strong relationship to RH and ENSO phase, the authors anticipate that the maximum observed daily THI value will also exhibit a relationship to DTR. Further examination of normalized occurrences of THI that are based on ENSO phase (Table 3) from the Ocklawaha station shows that values of THI of ≥72 are more likely to occur in DTR bins 12–22 during an El Niño spring than during a La Niña. In bin 17, an El Niño spring has 5.2 occurrences, as compared with the 1.3 occurrences during a La Niña. No instances of THI over 72 occur during a La Niña at the Salisbury ECONet station during the 17-yr period of record. At the same station, occurrences of THI of ≥72 are observed in most DTR bins during an El Niño (Table 3). Knowing this relationship and the predicted phase of ENSO can help those in the cattle industry to take appropriate actions (such as providing shade and access to more water) to prevent heat-related stress in their stock.
The number of normalized occurrences of THI ≥ 72 during MAM for both El Niño and La Niña, separated by DTR bins for the periods of record of 1999–2012 at Ocklawaha and 1996–2012 at Salisbury.
b. Exceedance probability bins
Although the CR clearly shows the relationship between DTRs and ENSO for each bin, it is a complex metric for stakeholders to apply specifically. A simpler, more direct metric, the probability of exceeding a DTR threshold, is produced to allow stakeholders to address their specific risk tolerance. Exceedance probabilities are calculated to determine the probability of reaching DTR thresholds and are defined as the historical probability of reaching the lowest DTR accounted for in a given bin.
Exceedance-probability graphs for Ocala during the (a) DJF and (b) MAM seasons. The dashed and dotted lines represent the exceedance probabilities for the El Niño and La Niña distributions, respectively.
Citation: Journal of Applied Meteorology and Climatology 52, 11; 10.1175/JAMC-D-12-0273.1
The probability of every DTR in the study reaching bin number 2 is 1.0. This is a result of the probability definition, in which the sum of the counts in each bin from j = bin number 2 through bin number >44 [the numerator of Eq. (4)] is identical to the total number of counts in all bins [the denominator of Eq. (4)], producing a quotient of 1.0. Intuitively, the probability of a DTR being at least 0°F (0.0°C) is 1.0.
Exceedance probabilities should have results similar to the histogram and CR results because they are calculated from the same distributions. In exceedance-probability graphs, cumulative effects of ENSO on DTRs can be examined and historical probabilities can be determined. Reaching a higher DTR threshold should be easier during a La Niña event than during an El Niño event, because La Niña is generally associated with higher DTRs and El Niño is generally associated with lower DTRs (section 4). This distribution difference can lead to a “cascading effect” in La Niña exceedance-probability graphs, in which the greater number of counts in higher DTR bins drives the probability of each lower DTR bin upward. The cascading effect alters the comparison between ENSO-separated exceedance probabilities. Careful interpretation of the exceedance probabilities necessitates accounting for this caveat.
The differences in exceedance probabilities between DJF (Fig. 8a) and MAM (Fig. 8b) and between ENSO phases are shown for Ocala, providing a representative example for Florida’s ranching region. The shift of probabilities between the El Niño and La Niña phases is similar in both seasons, but with a larger separation during MAM. For example, for MAM, bin number 22 has a probability of 0.89 during a La Niña event in contrast to a probability of 0.78 during an El Niño event. Because the probability is higher during La Niña than during El Niño, DTRs are historically more likely to reach the threshold of bin number 22, at least 20°F (11.1°C), during a La Niña event. In bin number 22, 284 stations (98%) have larger exceedance probabilities during La Niña than during El Niño. The exceedance probability of bin number 32 is 0.32 during La Niña and 0.18 during El Niño.
In summary, during El Niño the distribution is “front loaded” with higher probabilities of reaching lower bin numbers. Conversely, during La Niña the distribution is “back loaded” with higher probabilities of reaching higher bin numbers. Given the higher probability of small DTRs during El Niño springs, which are associated with increased humidity, the likelihood of reaching a THI above the critical threshold of 72 is increased.
When the THI results (Table 3) are compared with the findings yielded by the probability-of-exceedance analysis (Fig. 8), a dichotomy is revealed. Although the likelihood is higher (~10%) of having a DTR at or above bin 22 in a La Niña than in an El Niño, the THI results do not show an increase in instances of THI > 72 in La Niña. Winter and spring La Niña events are typically characterized by drier-than-normal conditions (see section 5), which lead to an increase in the DTR but not to an increase in the relative humidity; therefore, THI > 72 is less common at bin 22. Although the probability of DTRs greater than or equal to bin 22 is lower in El Niño than in La Niña, the threshold of THI > 72 is met at lower DTR thresholds in El Niño, largely because of the increase in atmospheric moisture (see section 5). Focusing future research more on the THI than on the probability of exceedance would be beneficial to the ranching community.
7. Conclusions
DTR distributions are shown to shift depending on the phase of ENSO. Conditional ratios are created and analyzed for stations throughout the Southeast, revealing with statistical confidence that El Niño is associated with lower DTRs and that La Niña is associated with higher DTRs. These results are placed into physical context on the basis of synoptic patterns during ENSO phases and are examined to show the relationship of DTRs to RH and the subsequent impact on THI, which is a critical measure of heat stress in cattle.
During DJF and MAM 34% of all stations have significantly more DTRs with ranges of [10°–14°F] (5.6°–7.8°C) in El Niño than in La Niña, whereas during DJF and MAM 25% of all stations have significantly more DTRs with ranges of [30°–34°F] (16.6°–18.8°C) in La Niña than in El Niño. There are 12% more significant stations during MAM than during DJF for DTRs of [10°–14°F] (5.6°–7.8°C). In addition, the significance pattern for DTRs of [10°–14°F] (5.6°–7.8°C) has a wider spread across the Southeast in MAM than in DJF. Although the number of significant stations is similar between DJF and MAM when DTRs are [30°–34°F] (16.6°–18.8°C), the magnitude of CRs, the measure of the comparable difference between the normalized ENSO bin counts, is about 2 times as large for MAM as for DJF. DTRs with a range of [20°–24°F] (11.1°–13.3°C) have few associated significance patterns in any season. In each season the Florida peninsula has more significant stations than does the Florida Panhandle.
These results may be explained by increased (decreased) cloud cover and precipitation associated with midlatitude cyclones and surface fronts, which lead to lower (higher) DTRs. The physical mechanisms of El Niño and La Niña are not entirely opposite; thus, DTR distribution signals between the two phases are not equal and opposite. Significant CRs are typically found in regions where ENSO effects are known to consistently increase or decrease cloud cover and precipitation. However, results can only partially be explained by the influence that ENSO phases have on cloud cover and associated precipitation. The synoptic patterns associated with ENSO phases in the Southeast likely contribute to the increase or decrease in the amount of near-surface relative humidity (RH) associated with El Niño and La Niña events, respectively. Increases in RH and lower DTRs during El Niño events lead to a higher likelihood of THI ≥ 72, particularly in the livestock region of north-central Florida.
Other relationships between DTRs and natural climate variability, such as the Atlantic multidecadal oscillation, the Pacific decadal oscillation, the Arctic Oscillation, the North Atlantic Oscillation and so on, likely exist but are not explored in this study. It is noteworthy that significant relationships between Tmax variability and the Arctic Oscillation have been determined (Lim and Schubert 2011), and future work should explore the role of these well-known climate signals on DTRs in the Southeast.
The results discussed herein are pertinent to stakeholders needing to protect livestock from the effects of heat stress associated with ENSO phase. An Internet-based tool that is based on the exceedance probability is currently being designed at COAPS for use on AgroClimate (http://agroclimate.org/), which is a service of the Southeast Climate Consortium. Farmers and ranchers will be able to quickly access graphs displaying the historical probability of exceedance for particular DTR values depending on the current or upcoming phase of ENSO.
Acknowledgments
The lead author thanks Dr. Lydia Stefanova for helpful discussion about physical results and Dr. Timothy Eichler for providing insight into spring storm tracks. Thanks are given to Mr. David Zierden for the use of the monthly ENSO index and to Ms. Kathy Fearon for her invaluable reviews and editing suggestions. The authors also thank the FAWN and ECONet database and project managers for sharing the data from their agricultural stations. This research was supported by the NOAA Applied Research Center Grant NA06OAR4310070 and USDA National Institute of Food and Agriculture (NIFA) Grant 2011-67003-30347.
REFERENCES
Angel, W. E., M. L. Urzen, S. A. Del Greco, and M. W. Bodosky, 2003: Automated validation for summary of the day temperature data. Preprints, 19th Conf. on Interactive Information and Processing Systems, Long Beach, CA, Amer. Meteor. Soc., 15.3. [Available online at https://ams.confex.com/ams/pdfpapers/57274.pdf.]
Dai, A., K. E. Trenberth, and T. R. Karl, 1999: Effects of clouds, soil moisture, precipitation, and water vapor on diurnal temperature range. J. Climate, 12, 2451–2473.
Durre, I., and J. M. Wallace, 2001: Factors influencing the cold-season diurnal temperature range in the United States. J. Climate, 14, 3263–3278.
Easterling, D. R., and Coauthors, 1997: Maximum and minimum temperature trends for the globe. Science, 277, 364–367.
Efron, B., and G. Gong, 1983: A leisurely look at the bootstrap, the jackknife, and cross-validation. Amer. Stat., 37, 36–48.
Eichler, T., and W. Higgins, 2006: Climatology and ENSO-related variability of North American extratropical cyclone activity. J. Climate, 19, 2076–2093.
Finch, V. A., 1986: Body temperature in beef cattle: Its control and relevance to production in the tropics. J. Anim. Sci., 62, 531–542.
Gershunov, A., and T. P. Barnett, 1998: ENSO influence on intraseasonal extreme rainfall and temperature frequencies in the contiguous United States: Observations and model results. J. Climate, 11, 1575–1586.
Green, P. M., 1996: Regional analysis of Canadian, Alaskan, and Mexican precipitation and temperature anomalies for ENSO impact. COAPS Tech. Rep. 96-6, 104 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.]
Hanley, D. E., M. A. Bourassa, J. J. O’Brien, S. R. Smith, and E. R. Spade, 2003: A quantitative evaluation of ENSO indices. J. Climate, 16, 1249–1258.
Higgins, R. W., A. Leetmaa, and V. E. Kousky, 2002: Relationships between climate variability and winter temperature extremes in the United States. J. Climate, 15, 1555–1572.
Karl, T. R., G. Kukla, and J. Gavin, 1984: Decreasing diurnal temperature range in the United States and Canada from 1941 through 1980. J. Climate Appl. Meteor., 23, 1489–1504.
Karl, T. R., G. Kukla, and J. Gavin, 1987: Recent temperature changes during overcast and clear skies in the United States. J. Climate Appl. Meteor., 26, 698–711.
Leathers, D. J., M. A. Palecki, D. A. Robinson, and K. F. Dewey, 1998: Climatology of the daily temperature range annual cycle in the United States. Climate Res., 9, 197–211.
Lim, Y.-K., and S. D. Schubert, 2011: The impact of ENSO and the Arctic Oscillation on winter temperature extremes in the southeast United States. Geophys. Res. Lett., 38, L15706, doi:10.1029/2011GL048283.
McDowell, R. E., 1972: Improvement of Livestock Production in Warm Climates. W. H. Freeman, 711 pp.
Moran, J., 2005: Appendix 1: Temperature humidity index. Tropical Dairy Feeding: Feeding Management for Small Holder Daily Farmers in the Humid Tropics, Landlinks Press, 275.
Mote, T. L., 1996: Influence of ENSO on maximum, minimum, and mean temperatures in the southeast United States. Phys. Geogr., 17, 497–512.
Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation (ENSO). Mon. Wea. Rev., 114, 2352–2362.
Saxena, V. K., S. Yu, and J. Anderson, 1997: Impact of stratospheric aerosols on climate: Evidence for aerosol shortwave and longwave forcing in the southeastern U.S. Atmos. Environ., 31, 4211–4221.
Sittel, M. C., 1994a: Differences in the means of ENSO extremes for maximum temperature and precipitation in the United States. COAPS Tech. Rep. 94-2, 50 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.]
Sittel, M. C., 1994b: Marginal probabilities of the extremes of ENSO events for temperature and precipitation in the southeastern United States. COAPS Tech. Rep. 94-1, 156 pp. [Available from Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840.]
Smith, R. A., 2007: Trends in maximum and minimum temperature deciles in select regions of the United States. M.S. thesis, Dept. of Meteorology, The Florida State University, 67 pp.
Smith, S. R., P. M. Green, A. P. Leonardi, and J. J. O’Brien, 1998: Role of multiple-level tropospheric circulations in forcing ENSO winter precipitation anomalies. Mon. Wea. Rev., 126, 3102–3115.
Smith, S. R., D. M. Legler, M. J. Remigio, and J. J. O’Brien, 1999: Comparison of 1997–98 U.S. temperature and precipitation anomalies to historical ENSO warm phases. J. Climate, 12, 3507–3515.
Stefanova, L., P. Sura, and M. Griffin, 2013: Quantifying the non-Gaussianity of wintertime daily maximum and minimum temperatures in the southeast United States. J. Climate, 26, 838–850.
Thom, E. C., 1959: The discomfort index. Weatherwise, 12 (2), 57–60.
West, J. W., 2003: Effects of heat-stress on production in dairy cattle. J. Dairy Sci., 86, 2131–2144.
World Meteorological Organization, 2010: Guide to climatological practices. 3rd ed. WMO 100, 117 pp. [Available online at http://www.wmo.int/pages/prog/wcp/ccl/guide/guide_climat_practices.html.]
Wu, Q., 2010: Associations of diurnal temperature range change with the leading climate variability modes during the Northern Hemisphere wintertime and their implication on the detection of regional climate trends. J. Geophys. Res., 115, D19101, doi:10.1029/2010JD014026.
Zhang, T., M. P. Hoerling, J. Perlwitz, D.-Z. Sun, and D. Murray, 2011: Physics of U.S. surface temperature response to ENSO. J. Climate, 24, 4874–4887.
