Evaluation of Oceanic and Terrestrial Sources of Moisture for the North American Monsoon Using Numerical Models and Precipitation Stable Isotopes

FEBRUARY 2015 H U A N D D O M I N G U E Z 19

HUANCUI HU AND FRANCINA DOMINGUEZ Department of Atmospheric Sciences, The University of Arizona, Tucson, Arizona

(Manuscript received 9 April 2014, in ﬁnal form 25 September 2014)

ABSTRACT

This work evaluates the oceanic and terrestrial moisture sources that contribute to North American monsoon (NAM) precipitation over a 30-yr period using the modified analytical dynamic recycling model. This computationally efficient modeling framework reveals previously overlooked moisture source regions such as Central America and the Caribbean Sea in addition to the well-known Gulf of California and Gulf of Mexico source regions. The results show that terrestrial evapotranspiration is as important as oceanic evaporation for NAM precipitation, and terrestrial sources contribute to approximately 40% of monsoonal moisture. There is a northward progression of terrestrial moisture sources, beginning with Central America during the early season and transitioning north into northern Mexico and the NAM region itself during the peak of the monsoon season. The most intense precipitation occurs toward the end of the season and tends to originate in the Gulf of California and the tropical Pacific, associated with tropical cyclones and gulf surges. Heavy stable isotopes of hydrogen and oxygen in precipitation (dD and d18O) collected for every precipitation event measured in Tucson, Arizona, for the period 1981–2008 complement the numerical results. The analysis shows that precipitation events linked to sources from the Gulf of Mexico and Caribbean Sea are more isotopically enriched than sources from the Gulf of California and tropical Pacific. It is also seen that terrestrial regions that derive their precipitation from the Gulf of Mexico are also more isotopically enriched than moisture sources from the Pacific.

1. Introduction subtropical high help drive water vapor from the Gulf of Mexico and the Caribbean into the NAM region The North American monsoon (NAM) accounts for (Douglas et al. 1993; Schmitz and Mullen 1996; Adams approximately 70% of the total annual precipitation in and Comrie 1997; Higgins et al. 1997; Bosilovich et al. northwestern Mexico and 40%–50% of annual precipi- 2003). However, the elevated topography of the SMO tation in the southwestern United States (Douglas et al. acts as a barrier for lower-level moisture that comes from 1993). The onset of the NAM in early June is character- the Gulf of Mexico and blocks it from entering the NAM ized by a westerly-to-easterly shift in the midtropospheric region. Consequently, the Gulf of California (GOC) and winds over northern Mexico and heavy rainfall over eastern tropical Pacific Ocean are also important sources western Mexico (Douglas et al. 1993; Stensrud et al. 1995; for NAM precipitation (Hales 1972; Brenner 1974; Adams and Comrie 1997; Higgins et al. 1997). As the Carleton 1986). One of the main mechanisms responsible season progresses, the region of heavy rainfall spreads for strong advection of moist tropical Pacific air chan- northward along the western slopes of the Sierra Madre neled up the Gulf of California are ‘‘gulf surges,’’ char- Occidental (SMO) into Arizona and New Mexico (Fig. 1). acterized by strong south-southeast winds, a drop in The Gulf of Mexico (GOM) is an important moisture temperature, and an increase in humidity over southern source for the NAM, as the positioning of the monsoon Arizona (Stensrud et al. 1997; Fuller and Stensrud 2000; ridge and westward extension of the North Atlantic Higgins et al. 2004; Douglas et al. 1993; Higgins et al. 1997). Tropical cyclones (TCs) originating in the eastern tropical Pacific can also contribute to NAM precipitation Corresponding author address: Francina Dominguez, Department of Atmospheric Sciences, The University of Arizona, PAS Building, during the monsoon season (Douglas et al. 1993; Higgins 1118 E. 4th St., Tucson, AZ 85721. et al. 1997). Furthermore, recent research has found that E-mail: [email protected] a large amount of gulf surge events are related to TCs,

DOI: 10.1175/JHM-D-14-0073.1

Ó 2015 American Meteorological Society Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 20 JOURNAL OF HYDROMETEOROLOGY VOLUME 16

FIG. 1. Geography of the NAM region (dotted area), showing locations mentioned in text. indicating a role of TCs in surge initiation (Fuller and of NAM precipitation. The DRM is a two-dimensional Stensrud 2000; Anderson et al. 2000; Douglas and Leal semi-Lagrangian analytical model (Dominguez et al. 2003; Higgins and Shi 2005). Thus, in general, the sources 2006; Dominguez and Kumar 2008) that has recently for NAM precipitation include a major source provided been modified to quantify moisture sources from dif- by the Gulf of California and eastern tropical Pacific at ferent regions (Martinez and Dominguez 2014). Similar lower levels (below 850 mb) and a source provided by the to more sophisticated three-dimensional Lagrangian Gulf of Mexico and the Caribbean Sea at upper levels methods, such as the Lagrangian Flexible Particle Dis- (Carleton 1986; Carleton et al. 1990; Schmitz and Mullen persion Model (FLEXPART; Stohl and James 2004, 1996; Adams and Comrie 1997). 2005; Gimeno et al. 2010; Nieto et al. 2014) and the quasi- While oceanic sources are clearly important for mon- isentropic back-trajectory (QIBT) method (Dirmeyer soon rainfall, significantly less attention has been paid to and Brubaker 1999, 2007; Brubaker et al. 2001), the terrestrial sources. Dominguez et al. (2008) argued that DRM can be used ‘‘offline’’ with different sources of at- precipitation after monsoon onset dramatically increases mospheric data, such as the North American Regional the soil moisture content, and this moisture is later Reanalysis (NARR) used in this study [see Gimeno et al. evapotranspired and can then significantly contribute to (2012) for a review of the different methods to quantify convective precipitation. A similar positive feedback moisture sources]. The DRM is computationally more between surface wetness and monsoon precipitation was efficient than FLEXPART and QIBT and has been identified by Small (2001) in a mesoscale modeling ex- shown to provide similar quantifications of moisture periment. In an effort to explicitly quantify the moisture sources over the South American continent as the QIBT sources for the NAM, Bosilovich et al. (2003) performed method (Martinez and Dominguez 2014). However, as a numerical simulation of the NAM water sources using with any two-dimensional analytical model, there is an three-dimensional tracers incorporated within a general assumption of a well-mixed atmosphere, which could circulation model. Surprisingly, they found that the con- cause incorrect estimates in regions where there is strong tinental sources provide much of the water for pre- vertical wind shear or incomplete mixing (Goessling and cipitation and suggested that precipitation recycling was Reick 2012). Some simple corrections have been in- an important factor contributing to monsoon intensity. corporated into numerical two-dimensional models to However, this study was limited by uncertainties related alleviate this shortcoming (van der Ent et al. 2013). to the global-scale modeling, including that the GCM However, there is currently no analytical solution to the used by Bosilovich et al. (2003) was unable to resolve the DRM that incorporates incomplete mixing, so the results Gulf of California, which could significantly affect the from our analytical model must be interpreted as first- source delineation. order estimates of moisture sources and are meant to In this study, we use the dynamical recycling model guide further efforts that incorporate the full three- (DRM) to delineate and quantify the moisture sources dimensional physics.

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 21

One of the main limitations of many numerically or box models, numerical water vapor tracers, and based moisture source studies of NAM precipitation is physical water vapor tracers (Gimeno et al. 2012). the lack of observational validation of the results. The The DRM is a two-dimensional analytical model heavy stable isotopes of hydrogen and oxygen in pre- initially designed for the study of local precipita- cipitation [dD (deuterium) and d18O] are ideal obser- tion recycling (Dominguez et al. 2006). Much like vational tracers for sources of moisture because the other analytical models, which are mathematically isotopic composition depends in part on the tempera- derived from the equation of conservation of atmo- ture and humidity characteristics of the source region spheric moisture, this model assumes that the atmo- where the moisture originated, as well as on the changes sphere is vertically well mixed. However, a significant of phase that the water within the air mass experiences difference is that the DRM takes the temporal change as it moves over the continent and rains out (Gat and of moisture storage into account, while other sim- Carmi 1970). Within the United States, Welker (2000) ilar models assume it to be negligible (Budyko 1974; shows that the average d18O values of precipitation Brubaker et al. 1993; Eltahir and Bras 1994)and originating from the Pacific Northwest are depleted in hence are only valid at monthly or longer time heavy isotopes (i.e., more negative) as compared to scales. By applying a semi-Lagrangian scheme, the precipitation originating from the Gulf of Mexico. In the recycling ratio computed by the DRM can be ex- southwestern United States, measurements of dDin pressed as water vapor during one premonsoon period show that ð t « x j t most of the isotopic variability can be associated with x j t 5 2 2 ( , , ) ›t0 R( , , ) 1 exp v x j t , (1) changes in the advection of atmospheric moisture from 0 ( , , ) different source regions (Strong et al. 2007). In their study, Strong et al. (2007) show that periods of high dD where R is the recycling ratio; « is the ET; and v is the are linked to precipitation originating from the Gulf of precipitable water in the semi-Lagrangian coordinate Mexico, while low dD values originate from the Pacific. system (x, j, t), representing x, y, and t dimensions. A Similarly, in an analysis of isotopic composition of mon- more detailed derivation of this equation can be found soonal precipitation in Tucson, Arizona, Wright et al. in Dominguez et al. (2006). The DRM is not limited to (2001) show that isotopically depleted years are cooler the study of local recycling, as it can also be used to and are associated with atmospheric circulation patterns quantify the relative contributions from different sour- that entrain more tropical moisture from the Pacific, ces to the atmospheric moisture over a given sink region while warmer, isotopically enriched years entrain less (Martinez and Dominguez 2014). In the modified DRM Pacific tropical moisture. It is important to note that the the ‘‘recycling ratio’’ over a large domain can be de- isotopic composition of water vapor originating through composed into moisture contributions from several evapotranspiration (ET) from terrestrial regions reflects subregions (Martinez and Dominguez 2014). The mod- the characteristics of the precipitation that falls over land ifiedDRM was successfully used in an application over and is subsequently evaporated and transpired. These are South America (Martinez and Dominguez 2014). In a complex mixture of fractionating and nonfractionating the current paper, the sink is the NAM region, while processes (Yakir and Sternberg 2000). In this work, we the potential source regions encompass North Amer- use observations of isotopic composition of precipitation ica and the eastern Pacific and western Atlantic basins events measured over a 28-yr period in Tucson, Arizona, (Fig. 1). The core NAM region is defined based on to complement the modified DRM model. both the North American Monsoon Experiment There are three main questions that we address in this (NAME) Science and Implementation Plan and study. 1) What are the main sources of moisture for NAME precipitation zones, as defined by Castro et al. NAM precipitation and their seasonal cycle? 2) What (2012).TheNAMregion(Fig. 1)includessouthern are the dominant modes of variability in source regions Arizona in the United States and Sonora and Sinaloa and associated synoptic patterns? 3) Can we use isotope in Mexico. information to provide information about source re- Throughout the manuscript we use the term ‘‘ratio gions and complement the numerical results? contribution,’’ which refers to the averaged ratio of precipitation contributed by each source to the total 2. Data and methods precipitation over the NAM region. As the ratios are computed based on values for each grid cell, they are a. Model description weighted by the precipitation rate of each grid and av- There are three primary methods used to quantify eraged over the NAM region. We also use the term source–sink regions of atmospheric moisture: analytical ‘‘amount contribution,’’ which is the product of the ratio

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 22 JOURNAL OF HYDROMETEOROLOGY VOLUME 16 contribution and area-averaged precipitation over the NAM region. The combination of the ratio and amount contribution provides a comprehensive picture of the moisture that originates from different sources in terms of precipitation intensity. b. Input data We use 30 years of NARR data (1979–2008) to drive the DRM (Mesinger et al. 2006). We limit our analysis to the summer monsoon period (from 16 June to 28 September) and do not use data after 2009 because several variables over Mexico had spatial incon- sistencies, partly due to changes in the assimilation procedure (W. Ebisuzaki 2014, personal communica- tion). Daily averaged precipitable water, zonal and FIG. 2. Isotopic composition (&) of all summer precipitation meridional winds, daily accumulated precipitation, and events measured in Tucson, AZ, for the 1981–2008 period, against ET are used as input to the DRM. The 3-hourly verti- the global meteoric water line. cally integrated moisture fluxes are divided by 3-hourly precipitable water to give the zonal and meridional wind vertically weighted by the specific humidity at each d 5 2 1000(Rx/Rs 1), (2) level, to provide the back trajectories. In contrast to the other variables, the estimates of ET are subject to error where R denotes the ratio of the heavy to light isotope because of a lack of observations (Nigam and Ruiz- and Rx and Rs are the ratios in the sample and standard Barradas 2006), and this makes ET the most uncertain (SMOW), respectively. Wright et al. (2001) used this variable in our moisture source analysis (Dominguez same dataset (17 years in their case) to calculate sea- and Kumar 2008). Furthermore, even though pre- sonally averaged (July–August) precipitation isotopic cipitation is assimilated in the NARR, the quality of composition. In this work, we will use individual events. observations dramatically decreases south of the U.S.– Mexico border, so there are quality concerns about d. EOF analysis the NARR-derived ET. We compared NARR pre- For each of the 3150 days in the 30-yr period (105 days 3 cipitation with that of the University of Delaware 30 years), the DRM provides a ratio contribution for precipitation product (Legates and Willmott 1990)and each grid cell. This contribution is the ratio of pre- found that the spatial distribution is similarly repre- cipitation contributed by each particular grid to the sented, but NARR tends to slightly underestimate NAM precipitation, as explained in section 2a, and precipitation, particularly in the states of Hidalgo, varies from 0 to 1. In constructing a ratio matrix (a Michoacán, Guerrero, and Chiapas in Mexico (not matrix of dimensions t 3 y, where t 5 3150 events and shown). NARR ET follows the spatial distribution of y 5 96 673, which is equal to 349 longitude 3 277 latitude NARR precipitation but is smaller in magnitude during grid points), we subtract the time mean of each grid box July and August. This gives us some confidence in the to indicate contribution anomalies at each time step; realistic representation of ET. thus, the magnitude of the contribution of each grid point is retained with its relatively positive or negative c. Isotopic data contribution at each time step. In a manner similar to The Department of Geosciences at the University of Wei et al. (2012), we reduced the dimensionality of the Arizona has been analyzing the heavy stable isotopes of data by performing an empirical orthogonal function oxygen and hydrogen (d18O and dD) in liquid water for (EOF) analysis on the ratio matrix. This provides the individual precipitation events over the Tucson area dominant modes of variability (EOFs) and their asso- since 1981. Figure 2 shows the isotopic composition for ciated principal component (PC) time series. The PC all available summer (June–September) events. The time series can be used to determine days that are strong stable isotope compositions are normally reported as d representatives of a particular mode. We determined values in units of parts per thousand (&) relative to the that if a particular day has PC above (below) one stan- standard mean ocean water (SMOW) composition; d dard deviation, it is representative of the positive (neg- values are calculated as ative) phase of the mode. This criterion is used to

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 23

FIG. 3. The 30-yr-averaged ratio contributions from each grid point in the NARR to the NAM (dotted area). analyze the synoptic characteristics and cases associated subregions is shown in Figs. 4b–d, and the spatial pattern with each mode of variability. of contributions from representative subregions is shown in Fig. 5. 3. Results The seasonal variation of moisture contribution from the Gulf of Mexico and the Caribbean Sea is primarily a. Climatology of intraseasonal NAM moisture driven by the climatological development of the mon- sources soon ridge and the displacement of the North Atlantic The 30-yr-averaged map of gridded ratio contribution subtropical high (Fig. 5a). The monsoon seasonal- during the monsoon season is shown in Fig. 3. Large averaged ratio contribution to NAM precipitation contributions from the SMO (both western and eastern from the Gulf of Mexico and Caribbean Sea is similar sides) and the Gulf of California are evident. Central with an average value of about 10% (Fig. 4b). By the end America, the eastern tropical Pacific, the western Gulf of June (26–30 June pentad), precipitation originating of Mexico, and the Caribbean Sea are also important from both sources is primarily located in southern source regions. The NARR domain is subdivided into Mexico (Fig. 5a). As the monsoon develops, easterly 15 representative subregions (Fig. 4a) that are part of moisture flux from the Caribbean Sea and, to a lesser three main zones: the Pacific (PAC), which includes the extent, the Gulf of Mexico crosses the SMO and sig- eastern tropical, lower, middle, and high Pacific and the nificantly contributes to NAM precipitation (21–25 July Gulf of California; the Atlantic (ATL), which includes pentad and 15–19 August pentad). During the late the Caribbean, Gulf of Mexico, and eastern Atlantic; monsoon season (9–13 September pentad), the dynam- and the terrestrial component (LAND), which includes ically favorable conditions for moisture advection into Central America, Baja California, mountains (eastern the NAM region diminish as the North Atlantic sub- slopes of the SMO), the western and eastern United tropical high retreats to its original position and the States, and the NAM region. Note that the region de- monsoon ridge dissipates. On the other hand, the fined as ‘‘eastern United States’’ includes a part of moisture contributed by the Gulf of Mexico to precip- Mexico. The 30-yr climatology of the area-averaged itation along the southeastern coast of the United States amount of precipitation contributed by each of the and over the SMO increases dramatically.

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 24 JOURNAL OF HYDROMETEOROLOGY VOLUME 16

FIG. 4. (a) The 15 subregions based on NARR data coverage. The sources are as follows: low-latitude Pacific (L-PAC), eastern tropical Pacific (E-PAC), Central America (C-AMR), CaribbeanSea(CARS),midlatitudePacific(M-PAC),BajaPeninsula (BAJA), GOC, NAM, MNT, eastern United States (E-US), GOM, high-latitude Pacific (H-PAC), western United States (W-US), Canada (CAN), and eastern Atlantic Ocean (E-ATL). Amount contributions to the NAM precipitation throughout the season (from 16 June to 28 September) from subregions of the (b) Atlantic, (c) Pacific, and (d) land. The pentads are marked by their starting dates. The values in each legend suggest the seasonal averaged ratio contributions from each subregion.

The monsoon ridge also drives the seasonal vari- largely driven by the Caribbean low-level jet and begins ability of moisture from the Pacific zone (Fig. 5b). The in early April, with a maximum in mid-June (Amador eastern tropical Pacific and Gulf of California have 1998; Wang and Lee 2007; Durán Quesada et al. 2010). similar contributions to NAM precipitation (;5%– The increase in soil moisture associated with pre- 7%), with a larger contribution per unit area from the cipitation over Central America is then reevaporated Gulf of California given its much smaller spatial ex- and can contribute to precipitation northward driven tent. Moisture advected from the eastern tropical by east-southeasterly wind. Consequently, early in Pacific contributes more to the total moisture than the NAM, there is a large contribution from Central that from the Gulf of California during early July, America(from1to20JulyinFig. 4d). The contri- especially to western Mexico (Fig. 4c). This situation bution from the NAM region is small early in the is quickly reversed near the end of July because of season(from16Juneto15JulyinFig. 4d), but in- the northward transport of moisture along the Gulf of creases continually because of the rapid increase in soil California into the NAM region. Gulf surge events moisture over the western flank of the SMO. The re- are likely responsible for much of this transport. cycling contribution then becomes as large as the These two sources contribute precipitation to a large contribution from Central America. In addition to area within the NAM region during September Central America and the NAM region, there are two (Fig. 5b), and this is likely related to TC activities, other important terrestrial sources: the eastern including some that potentially ‘‘trigger’’ gulf surge mountains of the SMO (MNT) and the eastern United events (Higgins and Shi 2005). The likelihood of States, which contribute ;8% and 6% of monsoon TC landfall in the region is much greater during precipitation each (Fig. 4d). September and October when the monsoon ridge Figure 6 summarizes the contributions (ratio and weakens and retreats southward, allowing the cy- amount) from the PAC, LAND, and ATL regions. The clones to recurve and affect terrestrial regions results highlight the importance of terrestrial moisture (Ritchie et al. 2011). This conclusion is supported by sources for monsoonal rainfall, particularly during the Fig. 5b, which shows the large contribution from mature stages of the monsoon in July and August, when western oceanic sources late in the monsoon, resulting roughly 40% of the precipitation is of terrestrial origin. in intense precipitation. Initially, the Pacific zone is the most important oceanic Terrestrial sources show a clear south-to-north pro- source, but the Atlantic zone quickly replaces it in early gression. Intense precipitation over Central America is July, corresponding to the well-known transition to

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 25

21 FIG. 5. The 30-yr-averaged amount of precipitation (mm day ) contributed from important sources of the ATL, PAC, and LAND during different pentads: (a) from CARS and GOM; (b) from L-PAC, E-PAC, and M-PAC; and (c) from C-AMR, the NAM, MNT, E-US, and W-US. The 30-yr-averaged 500-mb geopotential heights (m) are shown as dashed contours. The NAM region is denoted by dotted area. easterly winds that characterizes the monsoon. Toward greatly affect the NAM region during late September. the end of the season, however, the Paciﬁc again be- During the period of peak precipitation (mid-August), comes more important than the Atlantic, as the mon- terrestrial and Atlantic sources play a dominant role soon signal diminishes and Paciﬁc tropical cyclones (Fig. 6).

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 26 JOURNAL OF HYDROMETEOROLOGY VOLUME 16

FIG. 6. Oceanic and terrestrial contributions to the NAM precipitation throughout the season (from 16 June to 28 September): (a) the ratio contribution from the Pacific Ocean, land area, and the Atlantic Ocean throughout the 21 pentads (the pentads are marked by their starting dates) and (b) the amount contribution from these three regions together with the averaged total amount of precipitation in NAM region. b. Dominant modes of moisture source variability intense precipitation events in the 30-yr period (Fig. 8a). and associated synoptic features Events from the east are more frequent during the first half of the monsoon period; however, they are Using EOF analysis, the spatial patterns of the dom- on average less intense than those of Pacific origin inant modes of moisture source variability and their [Figs. 8a(1), 8a(2)]. The negative to positive phase associated PC time series are evaluated (Fig. 7). transition during the monsoon season is also illustrated by Fig. 7b(2). 1) EAST–WEST SOURCES 2) NORTH–SOUTH SOURCES The first EOF [Fig. 7a(1)] separates moisture sources from the east (primarily the Gulf of Mexico and the The second mode separates sources that originate Caribbean Sea) from those of the west (eastern tropical north and south of approximately 258N[Fig. 7a(2)]. Pacific, low and midlatitude Pacific, and the Gulf of Southerly sources dominate the beginning and end of California). The PC time series shows that sources from the monsoon season and tend to be associated with the west progressively become more important toward heavier precipitation events during the end [Figs. 7c(2), the end of the season [Fig. 7c(1)]. All events that score 8b(1), 8b(2)]. Northerly sources (north of about 258N) above (below) one standard deviation in the PC time are frequent in the middle of the season, when southerly series were classified as westerly (easterly) events sources drop dramatically. This can be considered a re- (Fig. 8a). Westerly sources also account for the most iteration of a northward transition of moisture sources to

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 27

FIG. 7. The first three EOFs with their PC time series: [a(1)] the first EOF mode; [a(2)] the second EOF mode; [a(3)] the third EOF mode; [b(1)] the PC of the first EOF mode smoothed by 5-day averages; [b(2)] the seasonal cycle of PC1, averaged over 30 years; [c(1)] the PC of the second EOF mode smoothed by 5-day averages; [c(2)] the seasonal cycle of PC2, averaged over 30 years; [d(1)] the PC of the third EOF mode smoothed by 5-day averages; and [d(2)] the seasonal cycle of PC3, averaged over 30 years. The NAM region is suggested by the dotted area in [a(1)], [a(2)], and [a(3)].

NAM precipitation during the monsoon and a frequent [Fig. 7a(3)]. The positive phase of this mode is associated impact of southern oceanic sources (probably the Pacific) with gulf surge patterns, as 69% of the major surges during the late season, as stated in section 3a. identified by Stutler (2013) occur during the positive phase of this mode. Precipitation of mountainous origin 3) SMO–GOC SOURCES is, on average, less intense than that of GOC origin during The third EOF differentiates the moisture of moun- the monsoon season [Fig. 8c(2)], as the western oceanic tainous origin around the SMO from those of the Gulf of sources account for the most intense events toward the California and adjacent regions. Negative EOF values end of the monsoon (as was previously shown). How- are associated with sources over both eastern and west- ever, precipitation originating as ET from the SMO ern slopes of the SMO, which includes the southern becomes more frequent than that of oceanic origin NAM region (local recycling). Positive values cover the during the peak monsoon period of August and early Gulf of California, the northern NAM, and the Pacific September [Fig. 8c(1)].

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 28 JOURNAL OF HYDROMETEOROLOGY VOLUME 16

FIG. 8. The number of events and daily precipitation associated with each phase of the ﬁrst three EOF modes, grouped with the same pentads as Fig. 4: [a(1)] the numbers of westerly–easterly source events, [a(2)] the daily precipitation of northerly–southerly source events, [b(1)] the numbers of northerly–southerly source events, [b(2)] the daily precipitation of northerly–southerly source events, [c(1)] the numbers of SMO–GOC source events, and [c(2)] the daily precipitation of SMO–GOC source events. The ‘‘mn’’ indicates the averaged daily precipitation for each phase of the modes.

Sample cases. We select two representative cases of like the events depicted in Higgins and Shi (2005), there EOF 1 (based on the value of their PC score) to analyze is another cyclone over the Gulf of Mexico off the east their associated synoptic patterns. The first case repre- coast of Mexico [see Fig. 3 in Higgins and Shi (2005)]. sents an anomalously high contribution from the Pacific This corresponds to Hurricane Earl, which affected zone (case 1), and the second case is for the Atlantic the region between 31 August and 3 September 1998 zone (case 2). (www.nhc.noaa.gov/data/tracks/tracks-at-1998.png). The ratio contribution maps from the eastern tropical 4) THE PACIFIC ZONE AS THE MAIN SOURCE Pacific, during the period of 2–4 September (Fig. 9b), show how the moisture from the south is able to On 4 September 1998, 68% of the area-averaged penetrate northward into the NAM region. Moving precipitation over the NAM region originated from northward along the cyclonic circulation, moisture the Pacific zone (27% from the Gulf of California, 30% from the tropical Pacific reached southern Arizona from the eastern tropical Pacific, and 20% from the low- on 3 September and eventually penetrated into Cal- latitude Pacific). This moisture contribution can be ifornia and Utah. linked to TC activity in the region, as Hurricane Isis moved up the Gulf of California into the Mexican state 5) THE ATLANTIC OCEAN AS THE MAIN SOURCE of Sonora during the period of 1–3 September 1998 (www.nhc.noaa.gov/data/tracks/tracks-ep-1998b.png). The Atlantic zone was the dominant source of pre- The 850-mb geopotential heights and integrated mois- cipitation over the NAM region on 25 August 1996. ture fluxes 3 days prior to the event show the evolution During this day, 60% of the total precipitation was of of a cyclone that originated in the tropical Pacific Atlantic origin. Much like the Pacific case, this event was (Fig. 9a). The cyclone moved north, bringing moisture linked to TC Dolly, which made landfall over the from the tropical oceans into the NAM region, and en- Yucatan Peninsula, as depicted by the 850-mb geo- training evaporation from the Gulf of California along potential height and vertically integrated moisture flux its path. Thus, a gulf surge was identified in the Yuma field (Fig. 9c). The westward movement of the TC cre- station as an anomalous moisture flux lasting 43 h ated enhanced southeasterly flow and brought eastern (Stutler 2013). This event is TC related, according to the oceanic moisture into Mexico and the central plains. The definition of Higgins and Shi (2005). In addition, much easterly transport channeled the moisture from the Gulf

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 29

FIG. 9. (a),(b) An example of a predominant source from the Pacific zone during 1–4 Sep 1998: [a(1)]–[a(3)] the 3-day evolution of vertically integrated moisture fluxes (streamlines) and 850-mb geopotential height (low- and high-pressure areas shaded by blue and red) and [b(1)]–[b(3)] the ratio contributions to NAM precipitation from the eastern tropical Pacific. (c),(d) An example of a predominant source from Atlantic Ocean during 21–26 Aug 1996: [c(1)]–[c(3)] the 5-day evolution of vertically integrated moisture fluxes (streamlines) and 850-mb geopotential height (low- and high-pressure areas shaded by blue and red) and [d(1)]–[d(3)] the ratio contributions to NAM precipitation from CARS. The NAM region is suggested by the dotted area. of Mexico into the southwestern United States across c. Analysis of stable water isotopes Texas (Fig. 9d). A maximum ratio contribution of 33% from the Gulf of Mexico to precipitation was 2 days in As shown before, for each summer day in the 1979– advance of a maximum contribution from the Caribbean 2008 period, we can delineate the regions that provide Sea of 34% on 25 August. The lag is mainly attributed to moisture for NAM precipitation. In addition, we have the larger distance of NAM from the Caribbean Sea. the d18O and dD composition of liquid water for all

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 30 JOURNAL OF HYDROMETEOROLOGY VOLUME 16

18 storm events in the Tucson area since 1981. Conse- TABLE 1. Correlation coefficients between d O and dD and the quently, for each monsoon storm event, we can evaluate PCs of the three dominant EOFs. The asterisks represent values the dominant sources of precipitation and how the iso- that are statistically significant at the 99% level for a two-tailed test. topic composition is related to the source region. For d18O dD example, during the event of 4 September 1998— PC1 20.304* 20.276* described above as a day where a TC influenced the PC2 20.023 20.086 monsoon region—the measured d18O was 214.2& and PC3 20.171* 20.15* dD was 296.5&. During this day, as mentioned above, we see that the moisture came predominantly from the tropical Pacific and the Gulf of California. This analysis composition for each bin. While the heaviest events are can be done for all summer events that had measure- clearly more negative, we see that at all levels of in- ments of isotopic composition in Tucson since 1981 (432 tensity, the easterly sources are more positive in d18Oin total events). EOF 1 and the SMO sources are more enriched in EOF If we calculate the correlation coefficient between the 3. EOF 2 shows no clear signature, as was expected from isotopic compositions and the principal components for the initial correlation analysis. As stated before, it is the three leading EOFs, we see statistically significant difficult to interpret the isotopic composition of terres- negative correlations between the d18O and dD values trial ET, but in the case of EOF 3, ET from the SMO is and EOF 1 and, to a smaller degree, for EOF 3 (see more enriched than moisture coming from the Gulf of Table 1). This indicates that moisture coming pre- California. Precipitation on the eastern side of the SMO dominantly from the east will likely be more isotopically largely originates from the Gulf of Mexico, which is enriched than average, while moisture from the lower- isotopically more enriched than average. This moisture latitude Pacific, Baja, Gulf of California, or NAM region is then evaporated by the soil and transpired by vege- (sources from the west) will be more depleted than aver- tation and then transported to the NAM region. Tran- age. The statistically significant correlation for EOF 3 in- spiration at steady state is a nonfractionating process dicates that moisture coming from the Gulf of California (Farquhar and Cernusak 2005), so transpired water from tends to be more isotopically depleted than moisture the SMO will have a very similar isotopic signature than originating from terrestrial ET from the SMO, although that of the Gulf of Mexico. the relationship is weaker than for EOF 1. We cannot use isotopes to distinguish the north–south sources that are 4. Discussion and conclusions depicted in EOF 2, as there is no statistical relationship. We can also use the EOF analysis presented above In this work, we use the modified DRM to evaluate and evaluate the isotopic composition for individual the moisture sources that contribute to summer NAM days that are representative of a strong positive or precipitation over a 30-yr period using the DRM and negative phase for each mode (Fig. 10, left). Not sur- water stable isotopes. While most of the literature has prisingly, we find that sources from the east are more focused on the relative role of the oceanic sources of isotopically enriched than sources from the west monsoonal moisture—with particular emphasis on the (EOF 1), and sources from the SMO are more enriched relative influence of the Gulf of Mexico versus the Gulf than those from the Gulf of California (EOF 3). It is of California—our study finds that moisture from ter- important to keep in mind that there is an observed restrial ET is in fact as important as oceanic evapora- ‘‘amount effect’’ where heavier precipitation events tion for NAM precipitation. Much like the results from tend to have more negative isotopic compositions than Bosilovich et al. (2003) using water vapor tracers in lighter ones during the summertime over midlatitudes a GCM, we find that terrestrial sources contribute to (Dansgaard 1964). This relationship is due to several approximately 40% of monsoonal moisture and are processes involved in the isotopic variations of convec- particularly important during the peak monsoon season. tive rainfall, but particularly important for arid and We find that during the early season (mid-June) Central semiarid regions is the fact that, during lighter pre- America is an important source of moisture. As the cipitation events, raindrops evaporate below cloud base season progresses, we find a northward progression of and attain a more positive isotopic composition; this terrestrial sources, and by midseason, local recycling and effect is reduced in very high intensity events moisture from the SMO are also important sources of (Dansgaard 1964). Consequently, we must evaluate moisture. Dominguez et al. (2008) and Bosilovich et al. whether the more negative isotopic compositions are (2003) also found that local recycling peaks midseason. simply due to heavier precipitation. To do this, we bin Interestingly, terrestrial sources have also been found to the storm events by intensity and evaluate the isotopic be very important contributors to precipitation during

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 31

18 FIG. 10. The d O grouped by (left) pentads (every other pentad marked by their starting dates) and (right) precipitation bins associated with three EOFs: (a) westerly and easterly sources, (b) northerly and southerly sources, and (c) the SMO and GOC sources. The ‘‘mn’’ indicates the pentad-averaged d18O for each phase of the modes.

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 32 JOURNAL OF HYDROMETEOROLOGY VOLUME 16 the Southeast Asian summer monsoon (Wei et al. 2012; relatively small compared to the oceanic counterpart Drumond et al. 2011). However, the large contribution and terrestrial sources are in general not associated with from terrestrial moisture does not necessarily lead to an the most intense precipitation events. This in part is due added source of predictability (Wei et al. 2012). Impli- to the fact that most terrestrial moisture comes from cations for predictability for the NAM region will be NAM and the SMO region and must cross the SMO explored in future work. before entering the NAM region. We examine the variability of moisture sources using We use observations of heavy stable isotopes of hy- EOF analysis. Each EOF mode suggests a ‘‘preferred drogen and oxygen (dD and d18O) collected for every direction of moisture sources.’’ Using this type of anal- precipitation event measured in Tucson, Arizona, for ysis, we find that the most intense NAM precipitation the period 1981–2008. We compare these measurements events occur toward the end of the season and tend to with our analysis of moisture sources. This very long originate in the oceans west and southwest of the NAM storm-scale observational dataset, as opposed to the region (Gulf of California and tropical Pacific). An ex- more usual monthly or seasonal data, provides an in- ample of this type of event occurred on 1–4 September valuable resource to compare with our numerical re- 1998 (Figs. 9a,b), and it can be directly related to TC sults. We find that there is a clear isotopic distinction activity over the eastern tropical Pacific. Higgins and Shi between ‘‘easterly’’ and ‘‘westerly’’ sources of moisture. (2005) show that gulf surge events (measured in Yuma, Precipitation events linked to sources from the Gulf of Arizona) that are directly related to TCs are associated Mexico and Caribbean Sea are more isotopically en- with stronger and deeper plumes of tropical moisture riched than sources from the Gulf of California and and wetter conditions over the core monsoon region. In tropical Pacific. We also see that terrestrial regions that fact, Ritchie et al. (2011) find that, for locations in the derive their precipitation from the Gulf of Mexico are southwestern United States (Phoenix, Tucson, and more isotopically enriched than moisture from the Gulf Albuquerque), TC remnants account for up to 30% of of California. This result is consistent with previous the annual rainfall. In contrast to the westerly sources, work using water vapor observations during a pre- precipitation events of easterly origin (primarily Gulf of monsoon season (Strong et al. 2007) and with work using Mexico and Caribbean Sea) predominate during the seasonal precipitation observations during the monsoon peak of the season. However, TC activity in the western period (Wright et al. 2001). However, our work is dis- Atlantic basin can also lead to enhanced precipitation in tinct because we have performed a climatological anal- the NAM region. This was the case during 21–26 August ysis linking storm-scale isotopic observations with their 1996 (Figs. 9c,d), where a TC that affected the Yucatan source regions. We hypothesize that the distinct isotopic Peninsula brought moisture that was later advected into signature between eastern and western sources is related the NAM. In this case, the dominant moisture sources to to the rainout processes that the parcels undergo as they the NAM were from the Gulf of Mexico and Caribbean move into the NAM. We also find that precipitation of Sea. The impact of TCs over the western Atlantic basin terrestrial origin from within the SMO region tends to for monsoonal precipitation has not been previously be isotopically more enriched than that of the Gulf of studied in the literature, although Higgins and Shi California. Our hypothesis is that this reflects the in- (2005) allude to their possible influence. fluence of the Gulf of Mexico because precipitation Our analysis finds a general south-to-north pro- within the SMO is primarily from the Gulf of Mexico gression in the sources of monsoonal rainfall. Southerly and the transpiration process maintains the same iso- sources (low-latitude Pacific, Central American, and topic signature of antecedent precipitation. However, Caribbean) dominate the early season, but as the season a more detailed model of isotopic moisture using iso- progresses, source regions begin to shift north. The lo- topic general circulation models (Jouzel et al. 2013) cation of the anticyclonic circulation associated with the would be needed to test both of these hypotheses. extension of the North Atlantic subtropical high is the Finally, as stated before, our simple analytical DRM most important driving factor for the relative contribu- has the strong assumption of a well-mixed atmosphere. tion of northerly versus southerly sources. A south- In any particular region, the degree of mixing in the eastern location of the high tends to result in moisture atmosphere depends on the degree of vertical hetero- coming preferentially from the south, while a north- geneities that arise because of changes in the direction of western location results in moisture originating north of the horizontal winds in the vertical direction (wind 258N. We also see that while, as stated above, oceanic shear) and the strength of vertical mixing due to con- sources are important in the early and late monsoon vection, and in fact, very rarely is the atmosphere well season, terrestrial sources dominate the midseason. mixed (Goessling and Reick 2012). Including the effect However, the terrestrial horizontal flux of moisture is of shear in simple numerical models has been shown to

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 33

FIG. 11. The area-averaged vertical distribution of water vapor transport within the NAM region for the westerly and easterly cases. The vapor transport amounts are scaled to show the relative magnitude. significantly improve the performance of these simple DRM’s well-mixed assumption, we calculated the ver- models (van der Ent et al. 2013). The NAM region is tical profile of zonal and meridional moisture fluxes indeed a region of strong vertical shear, as low-level averaged for selected days that were representative of wind is generally southerly while winds above roughly the positive and negative phases of the first EOF. We 700 mb are predominantly easterly (Stensrud et al. 1997; chose this EOF because it clearly separates westerly Douglas and Leal 2003; Higgins et al. 2004). This is the from easterly origin (when shear is more likely to occur; reason why the general understanding has been that Fig. 11). We see that, unlike the westerly events, easterly moisture at upper levels (above 700 mb) predominantly events show strong shear, with south-southwesterly flow comes from the Gulf of Mexico, while moisture at low below 750 mb and east-southeasterly flow above 750 mb. levels comes from the Gulf of California (Schmitz and On average, the southeasterly flow dominates so the Mullen 1996). To evaluate the potential impacts of the integral over the column would likely result in some

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC 34 JOURNAL OF HYDROMETEOROLOGY VOLUME 16 overestimation of the contribution from the area Budyko, M. I., 1974: Climate and Life. Academic, 508 pp. southeast of NAM and an underestimation of that from Carleton, A. M., 1986: Synoptic-dynamic character of ‘bursts’ the Gulf of California and eastern tropical Pacific. Un- and ‘breaks’ in the South-West U.S. summer precipitation singularity. J. Climatol., 6, 605–623, doi:10.1002/ fortunately, the DRM is an analytical model, and there joc.3370060604. is currently no analytical solution that accounts for ——, D. A. Carpenter, and P. J. Weser, 1990: Mechanisms of in- vertical heterogeneities within the DRM framework. In terannual variability of the southwest United States sum- our current and future work, we are including the three- mer rainfall maximum. J. Climate, 3, 999–1015, doi:10.1175/ , . dimensional aspect of moisture transport using water 1520-0442(1990)003 0999:MOIVOT 2.0.CO;2. Castro, C. L., H. I. Chang, and F. Dominguez, 2012: Can a regional vapor tracers embedded within a high-resolution at- climate model improve the ability to forecast the North mospheric model. American monsoon? J. Climate, 25, 8212–8237, doi:10.1175/ Our results are meant to be first-order estimates of JCLI-D-11-00441.1. moisture source regions for the NAM that can later be Dansgaard, W., 1964: Stable isotopes in precipitation. Tellus, 16A, analyzed with other more detailed methods. One of the 436–468, doi:10.1111/j.2153-3490.1964.tb00181.x. Dirmeyer, P. A., and K. L. Brubaker, 1999: Contrasting evapora- most important contributions to our work is that it shifts tive moisture sources during the drought of 1988 and the flood the attention of moisture sources from the traditional of 1993. J. Geophys. Res., 104, 19 383–19 397, doi:10.1029/ ‘‘Gulf of California’’ and ‘‘Gulf of Mexico’’ oceanic 1999JD900222. sources to other regions that have previously been ——, and ——, 2007: Characterization of the global hydrologic overlooked, including the mountainous regions of the cycle from a back-trajectory analysis of atmospheric water vapor. J. Hydrometeor., 8, 20–37, doi:10.1175/JHM557.1. SMO and Central America. Dominguez, F., and P. Kumar, 2008: Precipitation recycling variability and ecoclimatological stability—A study using NARR Acknowledgments. Support for this study has been data. Part I: Central U.S. Plains ecoregion. J. Climate, 21, provided in part by the National Science Foundation 5165–5186, doi:10.1175/2008JCLI1756.1. (NSF) Grant EF 1065790. The work of P.I. Dominguez ——, ——, X. Z. Liang, and M. Ting, 2006: Impact of atmospheric on isotope analysis was partially funded by NSF Grant moisture storage on precipitation recycling. J. Climate, 19, 1513–1530, doi:10.1175/JCLI3691.1. AGS 1045260. Any opinions, findings, and conclusions ——, ——, and E. R. Vivoni, 2008: Precipitation recycling vari- or recommendations expressed in this publication are ability and ecoclimatological stability—A study using NARR those of the authors and do not necessarily reflect the data. Part II: North American monsoon region. J. Climate, 21, views of NSF. We thank Yolande Serra for useful 5187–5203, doi:10.1175/2008JCLI1760.1. comments on this manuscript. Douglas, M. W., and J. C. Leal, 2003: Summertime surges over the Gulf of California: Aspects of their climatology, mean structure, and evolution from radiosonde, NCEP reanalysis, and rainfall data. Wea. Forecasting, 18, 55–74, doi:10.1175/ REFERENCES 1520-0434(2003)018,0055:SSOTGO.2.0.CO;2. Adams, D. K., and A. C. Comrie, 1997: The North American ——, R. A. Maddox, K. Howard, and S. Reyes, 1993: The monsoon. Bull. Amer. Meteor. Soc., 78, 2197–2213, doi:10.1175/ Mexican monsoon. J. Climate, 6, 1665–1677, doi:10.1175/ 1520-0477(1997)078,2197:TNAM.2.0.CO;2. 1520-0442(1993)006,1665:TMM.2.0.CO;2. Amador, J. A., 1998: A climatic feature of the tropical Americas: Drumond, A., R. Nieto, and L. Gimeno, 2011: Sources of moisture The trade wind easterly jet. Top. Meteor. Oceanogr., 5, 91–102. for China and their variations during drier and wetter condi- Anderson, B. T., J. O. Roads, and S. C. Chen, 2000: Large-scale tions in 2000–2004: A Lagrangian approach. Climate Res., 50, forcing of summertime monsoon surges over the Gulf of 215–225, doi:10.3354/cr01043. California and the southwestern United States. J. Geophys. Durán Quesada, A. M., L. Gimeno, J. A. Amador, and R. Nieto, Res., 105, 24 455–24 467, doi:10.1029/2000JD900337. 2010: Moisture sources for Central America: Identification of Bosilovich, M. G., Y. C. Sud, S. D. Schubert, and G. K. Walker, moisture sources using a Lagrangian analysis technique. 2003: Numerical simulation of the large-scale North American J. Geophys. Res., 115, D05103, doi:10.1029/2009JD012455. monsoon water sources. J. Geophys. Res., 108, 8614, Eltahir, E., and R. L. Bras, 1994: Precipitation recycling in the doi:10.1029/2002JD003095. Amazon basin. Quart. J. Roy. Meteor. Soc., 120, 861–880, Brenner, I. S., 1974: A surge of maritime tropical air—Gulf of doi:10.1002/qj.49712051806. California to the southwestern United States. Mon. Wea. Farquhar, G. D., and L. A. Cernusak, 2005: On the isotopic com- Rev., 102, 375–389, doi:10.1175/1520-0493(1974)102,0375: position of leaf water in the non-steady state. Funct. Plant ASOMTA.2.0.CO;2. Biol., 32, 293–303, doi:10.1071/FP04232. Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, 1993: Estima- Fuller, R. D., and D. J. Stensrud, 2000: The relationship between tion of continental precipitation recycling. J. Climate, 6, 1077– tropical easterly waves and surges over the Gulf of California 1089, doi:10.1175/1520-0442(1993)006,1077:EOCPR.2.0.CO;2. during the North American monsoon. Mon. Wea. Rev., ——, P. A. Dirmeyer, and A. Sudradjat, 2001: A 36-yr cli- 128, 2983–2989, doi:10.1175/1520-0493(2000)128,2983: matological description of the evaporative sources of TRBTEW.2.0.CO;2. warm-season precipitation in the Mississippi River basin. Gat, J. R., and I. Carmi, 1970: Evolution of isotopic composition of J. Hydrometeor., 2, 537–557, doi:10.1175/1525-7541(2001)002,0537: atmospheric waters in Mediterranean Sea area. J. Geophys. AYCDOT.2.0.CO;2. Res., 75, 3039–3048, doi:10.1029/JC075i015p03039.

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC FEBRUARY 2015 H U A N D D O M I N G U E Z 35

Gimeno, L., A. Drumond, R. Nieto, R. M. Trigo, and A. Stohl, Small, E. E., 2001: The influence of soil moisture anomalies on 2010: On the origin of continental precipitation. Geophys. Res. variability of the North American monsoon system. Geophys. Lett., 37, L13804, doi:10.1029/2010GL043712. Res. Lett., 28, 139–142, doi:10.1029/2000GL011652. ——, and Coauthors, 2012: Oceanic and terrestrial sources of Stensrud, D. J., R. L. Gall, S. L. Mullen, and K. W. Howard, continental precipitation. Rev. Geophys., 50, RG4003, 1995: Model climatology of the Mexican monsoon. J. Cli- doi:10.1029/2012RG000389. mate, 8, 1775–1794, doi:10.1175/1520-0442(1995)008,1775: Goessling, H. F., and C. H. Reick, 2012: Atmospheric water vapour MCOTMM.2.0.CO;2. tracers and the significance of the vertical dimension. At- ——, ——, and M. K. Nordquist, 1997: Surges over the Gulf mos. Chem. Phys. Discuss., 12, 30 119–30 176, doi:10.5194/ of California during the Mexican monsoon. Mon. Wea. Rev., acpd-12-30119-2012. 125, 417–437, doi:10.1175/1520-0493(1997)125,0417: Hales, J. E., Jr., 1972: Surges of maritime tropical air northward SOTGOC.2.0.CO;2. over the Gulf of California. Mon. Wea. Rev., 100, 298–306, Stohl, A., and P. James, 2004: A Lagrangian analysis of the atmo- doi:10.1175/1520-0493(1972)100,0298:SOMTAN.2.3.CO;2. spheric branch of the global water cycle. Part I: Method de- Higgins, R. W., and W. Shi, 2005: Relationships between Gulf of scription, validation, and demonstration for the August 2002 California moisture surges and tropical cyclones in the eastern flooding in central Europe. J. Hydrometeor., 5, 656–678, Pacific basin. J. Climate, 18, 4601–4620, doi:10.1175/JCLI3551.1. doi:10.1175/1525-7541(2004)005,0656:ALAOTA.2.0.CO;2. ——, X. L. Wang, and Y. Yao, 1997: Influence of the North ——, and ——, 2005: A Lagrangian analysis of the atmospheric American monsoon system on the U.S. summer precipita- branch of the global water cycle. Part II: Moisture transports tion regime. J. Climate, 10, 2600–2622, doi:10.1175/ between Earth’s ocean basins and river catchments. J. Hy- 1520-0442(1997)010,2600:IOTNAM.2.0.CO;2. drometeor., 6, 961–984, doi:10.1175/JHM470.1. ——, W. Shi, and C. Hain, 2004: Relationships between Gulf of Strong, M., Z. D. Sharp, and D. S. Gutzler, 2007: Diagnosing California moisture surges and precipitation in the south- moisture transport using D/H ratios of water vapor. Geophys. western United States. J. Climate, 17, 2983–2997, doi:10.1175/ Res. Lett., 34, L03404, doi:10.1029/2006GL028307. 1520-0442(2004)017,2983:RBGOCM.2.0.CO;2. Stutler, J., 2013: The robustness of operational weather forecast Jouzel, J., G. Delaygue, A. Landais, V. Masson Delmotte, C. Risi, criteria to define Gulf of California moisture surges. M.S. and F. Vimeux, 2013: Water isotopes as tools to document thesis, Dept. of Atmospheric Sciences, The University of oceanic sources of precipitation. Water Resour. Res., 49, 7469– Arizona, 64 pp. 7486, doi:10.1002/2013WR013508. van der Ent, R. J., O. A. Tuinenburg, H. R. Knoche, H. Kunstmann, Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial and H. H. G. Savenije, 2013: Should we use a simple or com- variability in gauge-corrected, global precipitation. Int. plex model for moisture recycling and atmospheric moisture J. Climatol., 10, 111–127, doi:10.1002/joc.3370100202. tracking? Hydrol. Earth Syst. Sci., 17, 4869–4884, doi:10.5194/ Martinez, J. A., and F. Dominguez, 2014: Sources of atmospheric hess-17-4869-2013. moisture for the La Plata River basin. J. Climate, 27, 6737– Wang, C., and S.-K. Lee, 2007: Atlantic warm pool, Caribbean low- 6753, doi:10.1175/JCLI-D-14-00022.1. level jet, and their potential impact on Atlantic hurricanes. Mesinger, F., and Coauthors, 2006: North American Regional Geophys. Res. Lett., 34, L02703, doi:10.1029/2006GL028579. Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/ Wei, J., P. A. Dirmeyer, M. G. Bosilovich, and R. Wu, 2012: Water BAMS-87-3-343. vapor sources for Yangtze River Valley rainfall: Climatology, Nieto, R., R. Castillo, A. Drumond, and L. Gimeno, 2014: A cat- variability, and implications for rainfall forecasting. J. Geo- alog of moisture sources for continental climatic regions. Water phys. Res., 117, D05126, doi:10.1029/2011JD016902. Resour. Res., 50, 5322–5328, doi:10.1002/2013WR013901. Welker, J. M., 2000: Isotopic (d18O) characteristics of weekly Nigam, S., and A. Ruiz-Barradas, 2006: Seasonal hydroclimate precipitation collected across the USA: An initial analysis variability over North America in global and regional re- with application to water source studies. Hydrol. Processes, analyses and AMIP simulations: Varied representation. J. 14, 1449–1464, doi:10.1002/1099-1085(20000615)14:8,1449:: Climate, 19, 815–837, doi:10.1175/JCLI3635.1. AID-HYP993.3.0.CO;2-7. Ritchie, E. A., K. M. Wood, D. S. Gutzler, and S. R. White, 2011: Wright, W. E., A. Long, A. C. Comrie, S. W. Leavitt, T. Cavazos, The influence of eastern Pacific tropical cyclone remnants on and C. Eastoe, 2001: Monsoonal moisture sources revealed the southwestern United States. Mon. Wea. Rev., 139, 192– using temperature, precipitation, and precipitation stable iso- 210, doi:10.1175/2010MWR3389.1. tope timeseries. Geophys. Res. Lett., 28, 787–790, doi:10.1029/ Schmitz, J. T., and S. L. Mullen, 1996: Water vapor transport as- 2000GL012094. sociated with the summertime North American monsoon as Yakir, D., and L. Sternberg, 2000: The use of stable isotopes to depicted by ECMWF analyses. J. Climate, 9, 1621–1634, study ecosystem gas exchange. Oecologia, 123, 297–311, doi:10.1175/1520-0442(1996)009,1621:WVTAWT.2.0.CO;2. doi:10.1007/s004420051016.

Unauthenticated | Downloaded 10/08/21 11:29 AM UTC