Climatic Controls and Hydrologic Impacts of a Recent Extreme Seasonal Precipitation Reversal in Arizona

By Goodrich, Gregory B Ellis, Andrew W

ABSTRACT The winter (December-February) of 2005/06 ranked as the driest in the instrumental record (since 1895) for nearly all regions of Arizona. The city of Phoenix, Arizona, recorded no precipitation during this time period, which was part of a record dry streak of 143 days without measurable precipitation. More important, the Salt and Verde watersheds, which supply the greater Phoenix area with approximately 50% of its water supply, received less than 3% of normal precipitation. Remarkably, this historically dry winter was preceded by the second wettest winter on record in 2004/05, a winter that filled reservoirs statewide and ameliorated a drought that has persisted since 1996 in some parts of the state. This study begins with a brief overview of the historical context of such reversals of extreme seasonal precipitation in Arizona followed by an analysis of the teleconnective impacts. The authors find that while an extreme reversal such as this has only happened once before in Arizona (1904/05 and 1905/06), there is a trend for increasing variability in winter precipitation from one year to the next in Arizona, especially since the 1960s. Large reversals of winter precipitation are followed by large reversals of the opposite sign in the summer monsoon more than 75% of the time. In general, large dry-to-wet reversals are associated with neutral ENSO-to-neutral ENSO conditions or a neutral ENSO-to-El Nino transition, whereas wet- to-dry reversals are associated with an El Nino-to-La Nina transition or, more commonly, with an El NIno-to-neutral ENSO transition. In addition, changes in the sign of the Atlantic multidecadal oscillation, eastern Pacific oscillation, and Pacific- North American (PNA) pattern are all significantly associated with precipitation reversals. During the seven winters when neutral ENSO and strongly positive PNA coexist, large wet-to-dry reversals occur in every case and nearly all rank among the largest such reversals. It is suggested that small reservoirs are more at risk for increasing climatic volatility than are large reservoirs.

1. Introduction

Arizona has experienced varying levels of drought since 1996 that at times have ranked among the worst since the late nineteenth century (Sheppard et al. 2002). The lengthy drought has coincided with explosive population growth in the metropolitan areas of Phoenix (3.7 million) and Tucson (0.9 million), Arizona, that has stressed water resource management and raised questions of sustainability (Morehouse et al. 2002). The impact of extended drought in Arizona is complicated by the fact that many areas rely on a surface water supply from large reservoirs that are fed from different climate types. A good illustration of this is the largest center of water demand in Arizona, metropolitan Phoenix, with three primary sources of water: reservoirs on upland watersheds, local groundwater, and the Colorado River along the western Arizona border (Fig. 1).

The complex drought challenge facing Arizona has spurred the drafting of a comprehensive drought plan that will guide stakeholders from agriculture, tourism, and watershed management sectors, among others, in decision making and vulnerability assessment (Jacobs et al. 2005). While the drought plan utilizes cutting-edge tools in climate prediction and monitoring, it is doubtful that policymakers were prepared for the extreme climate reversal that occurred during the winters of 2004/05 and 2005/06.

As late as August 2004, the statewide composite Palmer drought severity index (PDSI) value for Arizona ranked as the 8th-lowest August value since 1895, marking 38 consecutive months of severe drought, a period that included 28 months with a PDSI below -3.0. In fact, 80% of the months since the middle of 1995 had a negative PDSI. Water levels at Lake Mead along the Colorado River and at Roosevelt Lake on the Salt River watershed were at historical lows. At the time, the drought of 1996-2004 ranked as the second worst in Arizona history since the drought of 1896-1904 (Goodrich and Ellis 2006). Fortunately, a relatively wet end of the summer monsoon season brought a quick end to the drought, and by October 2004, the statewide PDSI was above zero. A relentless series of storms began in October 2004 and ended in February 2005, producing statewide monthly precipitation totals that ranked in the 93rd percentile for the 5-month period. The statewide winter season [December-February (DJF)] average for precipitation was 232% of the 1971-2000 normal, allowing Roosevelt Lake to fill to capacity for the first time since the winter and spring of 1995. Winter precipitation on the Salt River watershed was so great that it prompted the reluctant release of water from Roosevelt Lake, allowing the Salt River to flow freely through Phoenix for the first time since 1995. It appeared as though the second wettest winter in Arizona history had effectively ended the nine-year drought. Unfortunately, a once-per-century seasonal precipitation reversal was about to plunge Arizona back into severe climatological drought.

The winter of 2005/06 was as anomalously dry as the winter of 2004/05 was anomalously wet. After an average summer monsoon season and a wetter than average October, statewide monthly precipitation totals ranked in the 6th percentile for the 4-month period ending February 2006. Phoenix was rain-free for a record 143 consecutive days, and the statewide winter season average for precipitation was 7% of normal (Fig. 2). Worse yet, the high-elevation sites that normally accumulate upward of 2000 mm of snowfall and supply the area reservoirs with spring and summer runoff were devoid of snow. Only a wetter than normal March that ultimately raised winter snowfall totals to near one-half of normal mitigated a hydrologie disaster in the high-elevation communities. Incredibly, the drought- ending winter of 2004/05, the second wettest winter in Arizona history, had been immediately followed by the driest winter in the instrumental record. Water resource managers and policymakers must wonder how frequently the drought-sensitive state of Arizona can expect this type of seasonal precipitation reversal.

One of the theorized and modeled outcomes of anthropogenic climate change is an enhanced hydrologie cycle with regional shifts in drought and pluvial patterns, as well as increased seasonal variability with a tendency toward greater extremes (Houghton et al. 2001). Variability in precipitation on seasonal time scales is of importance when engaging in questions regarding climate change or uncertainty and scenarios of drought and the impact on water resources and sustainability. Extremity in seasonal precipitation is likely to produce volatility in physical and social systems of regions that are either unaccustomed to high climatic variability or have adapted to less climatic variability. This is especially the case if patterns in extreme precipitation show a tendency toward short-term reversal. Most research on precipitation extremes from climate change (Karl and Knight 1998; New et al. 2001; Kunkel et al. 2003; Trenberth et al. 2003) has focused on very short-term events (<30 days). Our objective in this preliminary research is to establish a methodology that defines broader extreme seasonal precipitation reversals (ESPR) and identifies them through time. By taking the difference in percentiles of winter precipitation for the seven climate divisions of Arizona, we will be able to examine other large reversals similar to 2005/06 and determine if these reversals are becoming more or less frequent. Then we will illustrate the potential impact of ESPR events on water resources and the implications for watershed managers and smaller stakeholders. Finally, we will then use a number of teleconnection indices with known relationships to winter precipitation in Arizona to examine synoptic controls that lead to large seasonal precipitation reversals.

2. Data

a. Precipitation

Precipitation time series for Arizona climate divisions were obtained from the National Climatic Data Center and include no missing data. Data from each of the seven climate divisions in Arizona represent a simple unweighted average from all representative stations within that division (Guttman and Quayle 1996). State composite averages are then areally weighted from the divisional averages. From 1895 to 1930 a regression technique was used to calculate the divisional averages based on available U.S. Department of Agriculture statewide averages, and this has reduced the temporal variance for these years (Guttman and Quayle 1996). Thus, the pre-1930 data do not show the actual historical spatial variability of the climate divisions. After 1930, the above arithmetic means method was used. (The climate division data can be accessed online at http://www.cdc.noaa.gov/Timeseries/.)

The individual station data for the eight cooperative stations in Fig. 2 were obtained from the Western Regional Climate Center historical summaries (available online at http://www.wrcc.dri.edu/ summary/ Climsmaz.html). Precipitation for each site was collected once daily in a standard 8-in. rain gauge. For both the climate division and individual station data, winter was considered to be December-February, and the year associated with each winter is that of the January-February (JF) period. For example, the winter period from December 2005 to February 2006 will be considered the winter of 2006. Mean values were calculated from 1971-2000 averages. b. Teleconnection

The Nino-3.4 index, which is calculated as the average of monthly sea surface temperature (SST) anoma lies for the area 5[degrees]N- 5[degrees]S, 120[degrees]0 -170[degrees]W, was used to represent ENSO. The National Oceanic and Atmospheric Administration method to determine various ENSO events states that when the 3-month moving average of Nino-3.4 anomalies exceeds +0.5 (-0.5) for 3 consecutive months, an El Nino (La Nina) event is said to occur. All other periods were considered neutral ENSO. Harshburger et al. (2002) found that the strongest lagged teleconnection to the western United States for the DJF period is September-November (SON). Therefore, if the center of the 3-month moving average designating the ENSO event occurred during any part of the September-November period, the following winter was classified accordingly. The dataset used in this study is the Kaplan extended Nino-3.4 dataset (Kaplan et al. 1998) and was obtained from the International Research Institute for Climate and Society data library (available online at http:// iridl.ldeo.columbia.edu/ SOURCES/.Indices/.nino/.EXTENDED/.NINO34/ ).A small adjustment was made to the time series to change the base- period climatology from 1951-1980 to 19712000.

Other teleconnections known to influence winter precipitation in Arizona were used in this study. Monthly data for the eastern Pacific oscillation (EPO) were obtained from the Climate Diagnostics Center (ftp://ftp.cdc.noaa.gov/Public/gtb/teleconn/ epo.janl948dec2005.asc); monthly data for the Pacific North American (PNA) pattern and the Southern Oscillation index (SOI) were obtained from the Climate Prediction Center (ftp://ftp.cpc.ncep.noaa.gov/ wd52dg/data/ indices/tele_index.nh and http://www.cpc.ncep.noaa. gov/ data/indices/soi, respectively); monthly data for the Pacific decadal oscillation (PDO) were obtained from the Joint Institute for the Study of the Atmosphere and Ocean at the University of Washington (http://jisao. washington.edu/pdo/PDO.latest); and unsmoothed monthly data for the Atlantic multidecadal oscillation (AMO) were obtained from the Earth Science Research Laboratory Physical Sciences Division (http:// www.cdc.noaa.gov/Correlation/ amon.us.data).

c. Streamflow and reservoir storage

Daily streamflow data were obtained from the U. S. Geological Survey’s National Water Information System for one gauge on each of the Salt (identifier = 09498500) and Verde (identifier = 09508500) watersheds (http://waterdata.usgs.gov/az/nwis). The gauge on the Salt River is located just upstream of Roosevelt Lake at an elevation of 663.5 m, and the gauge on the Verde River is located below Tangle Creek just upstream of Horeshoe Dam at an elevation of 618.4 m (Fig. 1). Each gauge represents the first unregulated gauge upstream of the reservoirs on each watershed. Daily data were totaled to monthly values of runoff into the reservoir systems. Monthly storage data for all six reservoirs on the Salt-Verde system were obtained from the Salt River Project and represent storage on the first of each month.

3. ESPR index development

To identify extreme seasonal precipitation reversal events since 1896 we first ranked winter (DJF) climate division data and created percentiles. The time series runs the length of recordkeeping for climate division data by the National Climatic Data Center and consists of the 111 winters from 1896 to 2006. The ESPR was calculated by subtracting the percentile rank of “yr -1″ from the percentile rank of “yr.” Since we are initially interested in identifying past ESPR events as opposed to the direction of the reversal (e.g., wet followed by dry), the absolute value of the ESPR was used. Therefore, the ESPR value for 2006 is the 98th percen tile or simply 98 (absolute value of 1st percentile minus 99th percentile). This methodology for analyzing winter season precipitation variability creates 110 ESPR values from 1897 to 2006 with a theoretical range of 0-99. This new time series can be analyzed for trends. ESPR classifications can be created to identify reversals of similar magnitude (i.e., ESPR80 represents all ESPR events with a percentile difference between successive years of 80% or greater). Since water resource managers and policymakers are most concerned with large hydroclimatic anomalies, understanding the variability and climate controls of years with ESPR values that are relatively large, such as anything greater than 60, would be important.

For the ESPR results for a location to be most useful, it must be shown that the ESPR values are random with minimal lag-1 autocorrelation. In other words, an extremely wet (or dry) winter should have an equal chance of being followed or preceded by a wet, dry, or normal winter. Otherwise, ESPR values will be clustered around zero. Table 1 shows there is minimal autocorrelation from winter to winter (all values are close to zero), which suggests that year-to-year winter precipitation variability is essentially random. With randomness confirmed it is possible to calculate the expected frequency per century and return interval for any number of ESPR values. To do this, we used a random number generator to create 1000 sets of 111 randomly ordered integers that were originally ranked from 1 to 111 to represent the 111 yr (1896-2006) of ranked precipitation. Percentiles were created for each of the 1000 sets of randomly generated numbers and the ESPR methodology was applied to each set as described above. The frequency per century and return interval for ESPR classes ranging from ESPR90 to ESPRlO using the randomly generated numbers can be seen in Table 2. The frequency distribution of randomly generated ESPR values is equal to the square of integers from one to nine. With an idea of how frequent various ESPR events should be, it is interesting to note that for most ESPR classes, the state of Arizona and its seven climate divisions all experience more winter season precipitation variability than would be expected from random chance (Table 2). For example, one would only expect ESPR60 values to occur 16 times per century from random chance, but Arizona has had over 21 ESPR60 events per century since 1896. This higher-than-expected year-to- year winter season precipitation variability is somewhat unexpected for the Southwest’s history of multidecadal drought (Goodrich and Ellis 2006), but recent research has demonstrated that even multiyear drought periods in the West are punctuated by relatively wet years (Meko and Woodhouse 2005). While possible climate controls behind this volatility in winter precipitation will be discussed in a later section, we will now focus on the temporal variability of ESPR events and a noticeable step change in the frequency of ESPR values>60 starting around 1960.

Figure 3 shows the time series of ESPR values along with a linear trend line and a 20-yr moving average. There are three distinct time periods of seasonal precipitation variability seen in Fig. 3 as well as in Table 3. There is a period of moderate variability from 1900 to 1939 that includes the 1905 event, a time of low variability from 1940 to 1959, and a time of increased variability that began in 1960 and continues to this day. In fact, when the 110-yr ESPR time series is split by the year 1960 (Table 4), the shift in year-to-year variability becomes very noticeable and is highly significant (rho = 0.01). Most notable is that both ESPR60 and ESPR70 events occur more than three times as often after 1960, during a time of explosive population growth. Because of this step-change around 1960, the decadal trend in average ESPR values for every climate division in Arizona is increasing (Table 2), with the highest and most significant rates of increase (rho [asymptotically =] 0.10) in northwestern Arizona (climate divisions 1 and 3). The uptick in seasonal variability during the past few decades is again demonstrated here by the fact that 10 of the 12 years that an ESPR80 event occurred in at least 1 climate division in Arizona happened after 1959.

Finally, we considered the trends in the time series of ESPR if the absolute value of the difference in percentiles is not taken (Fig. 4). Positive ESPR values represent dry winters followed by wet winters (dry to wet) while negative ESPR values represent wet winters followed by dry winters (wet to dry). While the overall trend over the past century is flat, it is interesting to note that the average ESPR value for wet-to-dry winters (negative ESPR) is decreasing (becoming more negative) at a rate of -2.5 percentiles per decade (rho = 0.07). In fact, six of the seven ESPR70 events that have occurred since 1960 have been of the wet-to-dry variety that occurred in 2005/06. Since wet-to-dry reversals are becoming more likely than dry-to-wet reversals in recent decades, we will now suggest why this should concern water resource managers in Arizona.

4. Hydrologie impacts

a. Reservoir size

Extreme reversals in a critical hydrologie season from one year to the next could have a dramatic impact on water reservoirs such as the soil, aquifers, and lakes. Winter in central Arizona provides a good example of such a critical hydrologie season. The primary water resource for Phoenix is runoff on the Salt and Verde watersheds to the north, accounting for approximately 50% of the water supply. Delivering more than 1 million acre feet to users in Phoenix, the Salt-Verde system has a capacity of approximately 2.3 million acre feet, which makes it a significantly more variable source of water for Phoenix than the regional Colorado River. Winter precipitation, especially snowfall, is the critical factor in the variability of water supply to the SaltVerde system, as the climatologically wet period from December to March and subsequent snowmelt in spring produces high mean monthly runoff from January to April. Runoff on the two watersheds has shown high covariability through their common period of record (r = 0.92; Fig. 5a). Applying the ESPR methodology to the seasonal (January-April) runoff values reveals a trend toward increasing variability in runoff, or greater absolute differences in runoff percentiles, from one year to the next on both the Salt and Verde watersheds (Figs. 5b,c). On the Salt watershed (n = 92) the increase is at a rate of 1.45 percentiles per decade (rho = 0.11), while it is 1.41 percentiles per decade (rho = 0.41) on the Verde (n = 60). Of the year-to-year differences in runoff greater than 60 percentiles, 70% occurred over the second half of the period of record on the Salt watershed (1960-2006) and 62.5% occurred over the second half of the record on the Verde watershed (1976-2006). This corresponds with the post-1960 increase in the ESPR across Arizona, indicating an expected link between extreme seasonal reversals in runoff and precipitation. The extreme reversal in runoff from the wet winter of 2004/05 to the dry winter of 2005/06 ranked second on the Salt watershed and first on the Verde watershed (Figs. 5b,c).

Relatively small reservoirs should exhibit a quick response to an ESPR event, while large reservoirs may only respond significantly to a positive ESPR event (dry to wet). Large-volume reservoirs should respond minimally to negative ESPR events (wet to dry), which is the impetus for constructing large reservoirs in the western United States, such as the artificially created surface water reservoirs on the Salt and Verde watersheds. High runoff on the Salt and Verde watersheds during January through April leads to an average annual peak water storage in the systems’ reservoirs in May. From the drought of the 1950s to the wet period of the 1970s and 1980s, May storage on the Salt-Verde system generally increased (Fig. 6). However, the recent dry period from the mid-1990s to 2004 produced a dramatic decline in storage within the Salt system until the obvious impact of the record-breaking wet winter of 2004/05.

Reservoir storage is regulated and does not reflect natural variability or unregulated runoff into the reservoirs. The larger reservoir system on the Salt watershed (capacity of 2 million acre feet) promotes year-to-year consistency, as the May storage in any given year is highly correlated with the May storage from the previous year (r = 0.57, rho = 0.00). This is not the case with the smaller reservoir system of the Verde watershed (capacity of 0.2 million acre feet), for which May storage in any given year is not significantly correlated with May storage from the previous year (r = 0.18, rho = 0.17), indicating greater year-to-year variability. The year-to-year difference in storage percentile is highly correlated with the year-to-year difference in runoff percentile on both the Salt (r = 0.46, rho = 0.00) and the Verde (r = 0.59, rho = 0.00) watersheds, and more so for the years of a positive difference in runoff on successive years (dry to wet) (Salt: r = 0.62; Verde: r = 0.67). For those years of a negative difference in runoff on successive years (wet to dry), the Verde continues to exhibit a significant correlation between the difference in runoff and the difference in storage from one year to the next (r = 0.51, rho = 0.01) while the larger Salt does not (r = 0.18, rho = 0.27). This suggests that large and small reservoir systems will be dramatically impacted by positive ESPR events (dry to wet), to the extent that the reservoirs have a significant available water capacity (e.g., after a dry period). During years of a positive seasonal reversal in runoff (dry to wet) of 60 percentiles or greater, the mean change in storage was +70.9 percentiles on the smaller Verde watershed (n = 4) and +46.2 percentiles on the Salt watershed (n = 9). However, smaller systems, like the Verde watershed, can also be significantly impacted by negative ESPR events (wet to dry). During years of a negative seasonal reversal in runoff (wet to dry) of 60 percentiles or greater, the mean decrease in storage was 45.1 percentiles on the smaller Verde watershed (n = 4) and only 19:6 percentiles on the Salt watershed (n = 10). On the Verde, the contrast between the two directions of change (positive and negative seasonal reversals) totals nearly 116 percentiles (+70.9 and -45.1), while it totals only about one-half that (66 percentiles) on the Salt (+46.2 and – 19.6). For example, in May 2005, storage on the Verde watershed increased 31 percentiles over storage in May 2004, but this was matched by a decrease of 32 percentiles in May 2006. The wet winter of 2005 and the dry winter of 2006 cancelled each other in terms of storage on the Verde watershed. On the Salt watershed, storage in May increased over 63 percentiles in 2005 from 2004, but decreased only 21 percentiles in May 2006, for a net increase of 42 percentiles. This illustrates the likelihood of greater response to seasonal climatic volatility within smaller systems.

b. Summer monsoon

While precipitation during the winter season is the primary contributor to spring runoff on the Salt and Verde watersheds, precipitation during the summer monsoon season (July-September) is a major contributor to annual precipitation in Arizona and can influence storage on both reservoir systems during the autumn in years when monsoon precipitation is particularly high (Douglas et al. 1993). The relationship between winter precipitation and monsoonal precipitation in the southwestern United States is well known, with dry (wet) winters generally followed by wet (dry) summers (Gutzler and Preston 1997; Small 2001). Since monsoonal precipitation is a nonnegligible part of Salt and Verde discharge, we were interested in determining the relationship between reversals of winter precipitation and reversals of the opposite sign of monsoonal precipitation during the following summer.

After using the ESPR methodology to develop a summer monsoon ESPR index using July-September precipitation, we found that winter and summer ESPR were negatively correlated (r = -0.264, rho = 0.01) as expected. After the 15 most extreme negative ESPR winters (wet to dry), the following monsoon season was wetter than the previous summer 13 of 15 times by an average of 31 percentiles (random chance would suggest a wetter summer only 7 or 8 of 15 times). The relationship between the 15 most extreme positive ESPR winters (dry to wet) and the occurrence of a drier than normal monsoon season was also evident, though not as consistent (10 of 15 times by an average of 43 percentiles). We were also interested in whether large monsoon reversals are consistently preceded by a large winter ESPR event of the opposite sign. However, we found that the 15 largest monsoon reversals, whether positive or negative, were not necessarily preceded by a large winter ESPR event or even one of the opposing sign. While beyond the scope of this article, our findings concerning the relationship between summer and winter ESPR may lead to improved forecast methods for the summer monsoon once the ESPR value for the preceding winter is known.

5. Teleconnective relationships with ESPR

a. ENSO

It has long been known that one of the primary controls of winter season precipitation in the Southwest is the El Nino-Southern Oscillation and that wet winters are associated with El Nino while dry winters are associated with La Nina (Ropelewski and Halpert 1986; Redmond and Koch 1991). Thus it is logical to suspect that changes in average ESPR values over time may be related to ENSO. A time series of sea surface temperatures from the Nino-3.4 region averaged over the months of September-November for the period of record was selected to best represent the ENSO teleconnection to Arizona (Harshburger et al. 2002). The ENSO values were placed into percentiles and the ESPR methodology was applied to find large relative changes from year to year in ENSO anomalies (Fig. 7). Table 3 shows that similar to average ESPR values, the year-to-year variability in autumn ENSO conditions increased sharply from 1960 to 1979, although the results when comparing the pre- and post-1960 averages are not significant (rho [asymptotically =] 0.25). Also similar to ESPR, the majority of large swings in year-to-year ENSO conditions (Nino-3.4 percentile difference>60) occurred after 1959 (13 of 21). In fact, 6 of the 7 occurrences of a Nino-3.4 percentile difference >80 occurred since 1964. Even with these recent large swings in annual ENSO conditions, the overall trend since 1900 is constant. That ENSO displays multidecadal variability has been well documented in the literature (Zhang et al. 1997; Diaz et al. 2001; among others), but in this case does not seem to correlate with Arizona ESPR values.

While the 1904/05 precipitation reversal was associated with a weak La Nina followed by a moderate El Nino, many of the other ESPR events >60 (22 of 24) were not associated with large swings in ENSO values (Nino-3.4 percentile difference >60). In fact, the majority of ESPR60 events (17 of 24) do not involve a swing from El Nino to La Nina or vice versa. Nearly half of the ESPR60 events are associated with neutral ENSO conditions that follow neutral ENSO conditions. The three times that ESPR60 events occurred in succession (1960-62,1984-86,1994-96), which entail large swings of either wet-dry-wet-dry or dry-wet-dry-wet, were almost entirely associated with neutral ENSO conditions. Since reversals from El Nino to La Nina and vice versa do not seem to be a strong predictor of ESPR events, we decided to investigate other teleconnections with a known relationship to winter precipitation in Arizona. b. Other teleconnections

The Atlantic multidecadal oscillation and Pacific decadal oscillation in addition to ENSO represent low-frequency variability in SSTs that correlates with winter precipitation in Arizona. Autumn values (SeptemberNovember) of the AMO, PDO, and Nino-3.4 indices were used to represent the antecedent state of the oceans prior to each winter. The EPO, PNA, and SOI represent low-frequency variability in the atmosphere and all have known relationships to winter precipitation in Arizona. Winter values (December-February) of the EPO, PNA, and SOI were used to represent the concurrent state of the atmosphere during each winter, since atmospheric teleconnections mostly correlate highly to precipitation without a lag. Since DJF values of two of the atmospheric teleconnections are only available since 1951, this section of the study is limited to 1952-2006.

The ESPR methodology was applied to all six teleconnections to develop a reversal index for each teleconnection, which was subsequently correlated with statewide ESPR values. Like ESPR, five of the six teleconnections displayed minimal lag-1 autocorrelation, with the AMO being the only exception. Table 5 shows that the directions (but not necessarily the values) of the AMO, SOI, EPO, and Nino-3.4 are all significantly correlated with ESPR. The results show that, broadly speaking, negative ESPR events (wet to dry) are likely to occur when El Nino is followed by La Nina and values of AMO and EPO increase. Positive ESPR events (dry to wet) are associated with the opposite reversals. However, during the historic 2005/06 ESPR event, neutral ENSO conditions followed El Nino and there was no change in the AMO. While a negative EPO in 2005 was followed by a positive EPO in 2006, this was not enough to explain why 2006 was such a strong ESPR event.

Since so many large ESPR events were associated with neutral ENSO, we examined only the subset of winters that occurred during a neutral ENSO event. Using the 1952-2006 data, we counted 27 winters of neutral ENSO. We then performed a number of statistical tests to determine the relationships of the remaining teleconnections (EPO, PNA, AMO, and PDO) to ESPR when ENSO was neutral. During the 27 winters of neutral ENSO, the relationship between the reversal index for AMO and PNA became much stronger (Table 5). Therefore, negative ESPR (wet to dry) events during winters of neutral ENSO are associated with both the AMO and PNA increasing in value. In addition, the actual values of each teleconnection were sorted from the subset of 27 neutral ENSO winters to allow for means testing of ESPR values for the highestlowest quartiles of the AMO, EPO, PNA, and PDO. The relationship between high and low quartiles of PNA values was impressive. The average ESPR value during the high quartile (7 winters) of PNA values (PNA > 0.5) was -63 while the average ESPR value during the low quartile (7 winters) of PNA values (PNA < -0.5) was +23 for a t value of 6.6 (p 0.5) coexisted all had strongly negative ESPR values (all <-32), and 6 of the 7 winters accounted for 6 of the 11 lowest ESPR values since 1952. All seven winters were preceded by either neutral ENSO or El Nino. For comparison purposes, during the 10 winters that El Nino and positive PNA (PNA > 0.5) coexisted, only 4 of the 10 winters even had negative ESPR values and only 1 of the 10 was among the 11 lowest ESPR values since 1952.

The reason for this powerful relationship between dry winters and neutral ENSO-positive PNA was puzzling because, in general, DJF PNA is poorly correlated with winter precipitation in Arizona for the entire time series (1952-2006: r = -0.13; rho = 0.33). For this reason, PNA is often not considered to be an important seasonal forecast tool for Arizona, even though the PNA is positively correlated with Nino-3.4 (r = 0.48; rho = 0.00) and most winters of El Nino-positive PNA result in above-normal winter precipitation. However, during winters of neutral ENSO, PNA has a negative correlation with precipitation (r = -0.48; rho = 0.01) that is highly significant. This reversal of the correlation between PNA and precipitation depending on the phase of ENSO is what reduces the correlation between PNA and precipitation for the entire time series. When the newfound relationship among PNA, AMO, and winter precipitation during years of neutral ENSO is put into a regression equation, the resulting equation (r = 0.65; rho < 0.01) explains 41% of the variance of ESPR. Figure 8 shows the difference in the location of 500-hPa height anomalies between winters of El Nino- positive PNA and winters of neutral ENSO-positive PNA. The average 35-gpm height rise over the southwestern states during winters of neutral ENSO-positive PNA produces subsidence over the region and blocks the westerly flow of storms in the subtropical jet stream, which leads to the decrease in winter precipitation during those winters.

Not to be lost in the surprising finding that all seven winters of neutral ENSO-positive PNA are associated with large negative ESPR values is the fact that for large negative ESPR values to occur in the first place, the previous year had to have been anomalously wet. The 7 preceding winters averaged precipitation in the 84th percentile and included 3 of the 10 wettest winters in the time series. By contrast, the 7 winters of neutral ENSO-positive PNA averaged precipitation in the 21st percentile and included the driest winter of the time series (2006). Of the seven winters that preceded the seven winters of neutral ENSO-positive PNA, four were also neutral ENSO while three were El Nino. In the 4 neutral ENSO- to-neutral ENSO cases, the PNA increased by an average of 1.0 from the previous winter. For the three El Nino-to-neutral ENSO cases, the PNA actually decreased from the previous winter in two of the three cases, but still remained positive. The sequence from 2005/06 was the lone exception as an El Nino with a PNA of 0.14 was followed by neutral ENSO conditions with a PNA of 0.57 in 2006.

6. Conclusions

The state of Arizona recently experienced the second wettest winter (2005) and the driest winter (2006) over the past 111 yr in back-to-back years. To determine the historical precedence of this event, a methodology for calculating extreme seasonal precipitation reversal (ESPR) was developed. By calculating the absolute difference in percentiles of winter precipitation from year to year, we determined that the 2006 ESPR event was one of two similar events in the past 110 yr (1905). By comparing the frequency of ESPR events to that of a random distribution, we determined that winter precipitation in Arizona experiences more variability from year to year than would be expected from chance and that variability has increased significantly since 1960, especially in events that involve a reversal of more than 60 percentiles. Using data from the primary reservoirs that supply water to the greater Phoenix area, we demonstrated that while highly positive ESPR events (very dry winter followed by very wet winter) can significantly impact large reservoirs, highly negative ESPR events (very wet winter followed by very dry winter) likely have a similar impact only on small reservoirs. We also found that large positive and negative ESPR events can impact the amount of precipitation during the following summer monsoon season.

The most surprising finding of the study was that while ENSO was related to large ESPR events, other teleconnections such as the AMO and EPO were more strongly correlated to ESPR than anomalies of Nino- 3.4 SSTs. More than half of all large ESPR events, whether positive or negative, were associated with winters of neutral ENSO rather than the expected winters of El Nino or La Nina. During winters when neutral ENSO conditions coexisted with positive PNA, large negative ESPR events occurred every time and account for the majority of all ESPR values <0 -50 since 1952.

The water infrastructure of dams, reservoirs, and canals that has provided the Phoenix region with a reliable source of water for almost 100 yr now faces the challenge of the effect of climate change on regional water supplies. Greater year-to-year volatility in the natural supply of water to the reservoirs would place added stress on water management and could potentially lead to an increase in wildfires since wet years promote growth of vegetation that could become fuel for fires during dry years. This issue is magnified for municipalities without access to large reservoirs, which includes much of the eastern United States, where there could be a distinct challenge to combat social complacency that develops during the first year of a negative ESPR event (wet to dry). Given this and the fact that extreme negative events seem to be increasing in Arizona, our future research will apply the ESPR methodology to the entire United States to identify regions that may be particularly at risk and to see if the teleconnective relationships found in Arizona can lead to tools that may enable the seasonal prediction of large ESPR events.

Acknowledgments. This research was supported in part by the Kentucky EPSCoR Research Startup Fund Grant RSF-021-04.

REFERENCES

Diaz, H. F., M. P. Hoerling, and J. K. Eischeid, 2001: ENSO variability, teleconnections and climate change. Int. J. Climatol., 21, 1845-1862.

Douglas, M. W., R. A. Maddox, and K. Howard, 1993: The Mexican monsoon. J. Climate, 6, 1665-1677.

Goodrich, G. B., and A. W. ElHs, 2006: Climatological drought in Arizona: An analysis of indicators for guiding the governor’s drought task force. Prof. Geogr., 58, 460-469.

Guttman, N. B., and R. G. Quayle, 1996: A historical perspective of U.S. climate divisions. Bull. Amer. Meteor. Soc., 77, 293303. Gutzler, D., and J. Preston, 1997: Evidence for a relationship between spring snow cover in North America and summer rainfall in New Mexico. Geophys. Res. Lett., 24, 2207-2210.

Harshburger, B., H. Ye, and J. Dzialoski, 2002: Observational evidence of the influence of Pacific SSTs on winter precipitation and spring stream discharge in Idaho. J. Hydrol., 264, 157-169.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp.

Jacobs, K. L., G. M. Garfin, and B. J. Morehouse, 2005: Climate science and drought planning: The Arizona experience. J. Amer. Water Resour. Assoc., 41, 437-445.

Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. BIumenthal, and B. Rajagopalan, 1998: Analyses of global sea surface temperature 1856-1991. J. Geophys. Res., 103, 18 567-18 589.

Karl, T. R., and R. W. Knight, 1998: Secular trends of precipitation amount, frequency, and intensity in the United States. Bull. Amer. Meteor. Soc., 79, 231-241.

Kunkel, K. E., D. R. Easterling, K. Redmond, and K. Hubbard, 2003: Temporal variations of extreme precipitation events in the United States: 1895-2000. Geophys. Res. Lett., 30, 1900, doi:10.1029/ 2003GL018052.

Meko, D. M., and C. A. Woodhouse, 2005: Tree-ring footprint of joint hydrologie drought in Sacramento and Upper Colorado river basins, western USA. J Hydrol., 308, 196-213.

Morehouse, B. J., R. H. Carter, and P. Tschakert, 2002: Sensitivity of urban water resources in Phoenix, Tucson, and Sierra Vista, Arizona, to severe drought. Climate Res., 21, 283-297.

New, M., M. Todd, M. Hulme, and P. Jones, 2001: Precipitation measurements and trends in the twentieth century. Int. J. Climatol., 21, 1899-1922.

Redmond, K. T., and R. W. Koch, 1991: Surface climate and streamflow variability in the western United States and their relationship to large-scale circulation indexes. Water Resour. Res., 27, 2381-2399.

Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Nino/ Southern Oscillation (ENSO). Mon. Wea. Rev., 114, 2352-2362.

Sheppard, P. R., A. C. Comrie, G. D. Packin, K. Angersbach, and M. K. Hughes, 2002: The climate of the U.S. Southwest. Climate Res., 21, 219-238.

Small, E. E., 2001: The influence of soil moisture anomalies on variability of the North American monsoon system. Geophys. Res. Lett., 28, 139-142.

Trenberth, K. E., A. G. Dai, R. M. Rasmussen, and D. B. Parsons, 2003: The changing character of precipitation. Bull. Amer. Meteor. Soc., 84, 1205-1217.

Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900-93. J. Climate, 10, 1004-1020.

GREGORY B. GOODRICH

Department of Geography and Geology, Western Kentucky University, Bowling Green, Kentucky

ANDREW W. ELLIS

School of Geographical Sciences, Arizona State University, Tempe, Arizona

(Manuscript received 30 October 2006, in final form 13 June 2007)

Corresponding author address: Gregory B. Goodrich, Department of Geography and Geology, Western Kentucky University, 1906 College Heights Blvd., Bowling Green, KY 42101.

E-mail: gregory.goodrich@wku.edu

Copyright American Meteorological Society Feb 2008

(c) 2008 Journal of Applied Meteorology. Provided by ProQuest Information and Learning. All rights Reserved.