What controls the interannual variation of the wetseason onsets over the Amazon?Lei Yin1,2, Rong Fu1, Yong-Fei Zhang1, Paola A. Arias3, D. Nelun Fernando1,4,5, Wenhong Li6,Katia Fernandes7, and Adam R. Bowerman1

1Department of Geological Sciences, Jackson School of Geosciences, University of Texas at Austin, Austin, Texas, USA,2Department of Statistics and Data Science, College of Natural Sciences, University of Texas at Austin, Austin, Texas, USA, 3Grupode Ingeniería y Gestión Ambiental, Universidad de Antioquia, Medellín, Colombia, 4Visiting Scientist Programs, UniversityCorporation for Atmospheric Research, Boulder, Colorado, USA, 5Surface Water Resources Division, Texas Water DevelopmentBoard, Austin, Texas, USA, 6Division of Earth and Ocean Sciences, Nicholas School of the Environment, Duke University, Durham,North Carolina, USA, 7International Research Institute for Climate and Society, Columbia University, Palisades, New York, USA

Abstract Previous studies have established that sea surface temperature anomalies (SSTAs) in the tropicalPacific and Atlantic are the main forcing of the interannual variation of the wet season onsets in the Amazon.However, this variation appears to be complex and not uniquely determined by SSTAs. What causes such acomplexity and to what extent the interannual variation of the wet season onsets is predictable remainunclear. This study suggests that such a complex relationship is the result of several competing processes,which are nonlinearly related to the SSTAs. In particular, three dry season conditions are crucial for determininginterannual variation of the wet season onset. (i) A poleward shift of the Southern Hemisphere subtropical jet(SHSJ) over the South American sector, initiated from awave train-like structure possibly forced by south centralPacific SST patterns, can prevent cold frontal systems frommoving northward into the Amazon. This delays coldair incursion and results in late wet season onset over the southern Amazon. (ii) An anomalous anticycloniccenter, which enhances westerly wind at 850hPa over the southern Amazon and also the South Americanlow-level jets, leads to moisture export from the southern Amazon to La Plata basin and reduces convectivesystems that provide elevated diabatic heating. (iii) Smaller convective available potential energy (CAPE) limitslocal thermodynamically driven convection. Based on the stepwise and partial least squares regressions, thesethree selected preseasonal conditions (Niño 4, SHSJ, and CAPE) can explain 57% of the total variance of the wetseason onset.

1. Introduction

Interannual variation of the wet season onsets in the Amazon has a strong impact on the ecosystem servicesof rainforests, fire risk, river levels, and small-scale agriculture nearby river banks, and significantly contributesto interannual variations of global atmospheric carbon dioxide concentration [Ackerley et al., 2011; Asner,2009; Chen et al., 2011; Cochrane et al., 1999; Friedlingstein et al., 2006; Lee et al., 2011; Sombroek, 2001]. Mostprevious studies identify the El Niño-Southern Oscillation (ENSO) and Atlantic sea surface temperatureanomalies (SSTAs) as the primary forcing of interannual variability of dry season rainfall and the following wetseason onset over the Amazon [Fernandes et al., 2011; Kousky et al., 1984; Liebmann and Marengo, 2001;Marengo et al., 2001;Marengo et al., 2011; Yoon and Zeng, 2010]. However, the influence of oceanic forcing ononset dates is complex and not sufficiently robust enough for predicting the wet season onsets [Liebmannand Marengo, 2001]. How can we improve our understanding on what controls the interannual variability ofthe wet season onset in the Amazon beyond its linear relationship with ENSO and Atlantic SSTAs? This studyaims to synthetically explore the most critical preseasonal conditions for interannual variability of the wetseason onset over southern Amazonia.

Previous studies have suggested that the wet season onset in the southern Amazon is initiated by increasedevapotranspiration (ET) [Fu et al., 1999; Li and Fu, 2004; Li et al., 2006] as a result of the rainforest’s response toa seasonal increase of solar radiation [Myneni et al., 2007]. Increased ET provides moisture into the atmosphericboundary layer and reduces convective inhibition (CIN). The passage of cold fronts propagated from theextratropics lifts moist and warm surface air over southern Amazonia, triggering deep convection over a largespatial area [Li and Fu, 2006]. The elevated diabatic heating produced by deep convection, together with the

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2314

PUBLICATIONSJournal of Geophysical Research: Atmospheres

RESEARCH ARTICLE10.1002/2013JD021349

Key Points:• The wet season onset over theAmazon is not uniquely deteriminedby SSTAs

• A poleward SHSJ, enhanced U850, andlower CAPE are crucial for delayed onset

• Selected conditions can explain 57%of the total variance

Correspondence to:L. Yin,yin@utexas.edu

Citation:Yin, L., R. Fu, Y.-F. Zhang, P. A. Arias, D. N.Fernando, W. Li, K. Fernandes, and A. R.Bowerman (2014), What controls theinterannual variation of the wet seasononsets over the Amazon?, J. Geophys.Res. Atmos., 119, 2314–2328,doi:10.1002/2013JD021349.

Received 11 DEC 2013Accepted 10 FEB 2014Accepted article online 14 FEB 2014Published online 13 MAR 2014

southwardmigration of the warmest Atlantic SSTs, drives the reversal of the cross-equatorial flow and increasesmoisture transport from the Atlantic Ocean to the Amazon [Wang and Fu, 2002]. The latter accelerates themoistening of the lower troposphere and provides more water vapor for a further increase of rainfall. Large-scale deep convection associated with frequent incursions of extratropical cold frontal systems convertsconvective available potential energy (CAPE) in the lower troposphere to kinetic energy in the upper troposphere,spins up the anticyclonic circulation around the Bolivian High, and results in wet season onsets [Lenters andCook, 1997; Li et al., 2006].

Interannual variation of the wet season onset in the Amazon appears to be influenced by changes in thesefactors. For example, an anomalously high land surface Bowen ratio during the preceding dry seasonwould delaythe subsequent wet season onsets in the southern Amazon [Fu and Li, 2004]. Changes of cross-equatorialmoisture transport could influence convection and thus the wet season onset [Marengo et al., 2001; Rao et al.,1996]. Weak and infrequent extratropical cold front incursions during the transition season also contribute to adelay of the wet season onsets [Li et al., 2006]. However, the relative importance of these changes and themechanisms by which the interactions between these processes contribute to the complexity of the relationshipbetween ENSO and Atlantic SSTAs and the wet season onset over the southern Amazon remain unclear.

How do ENSO and Atlantic SSTAs influence these processes? Previous studies have shown that El Niño, warmerSST in the northern tropical Atlantic or cooler SST in the southern Atlantic, and the Atlantic MultidecadalOscillation (AMO) could reduce moisture transport and increase atmospheric stability over the Amazon [Ariaset al., 2010; Chiang and Sobel, 2002; Marengo, 1992; Marengo et al., 2008; Neelin, 2003; Zeng et al., 2008], andbiases in these SSTAs have been shown to be responsible for dry biases over Amazonia in some global climatemodels [Yin et al., 2012]. However, only some of the El Niño and southern tropical Atlantic cooling events areactually linked to a delay of the wet season onset, whereas other events show no impact on the wet seasononsets [Fu and Li, 2004; Marengo et al., 2010]. How these remote oceanic forcings influence the land surfaceBowen ratio, the atmospheric thermodynamic structure, and incursions of the extratropical frontal systemswithin Amazonia during the transition seasons is still not adequately understood. These processes may respondto SSTA nonlinearly. For example, the Amazon rainforests can draw water from soil layer as deep as 14–18mthrough their deep roots and efficient hydraulic lifting [Nepstad et al., 1994; Oliveira et al., 2005]. Thus,photosynthesis and evapotranspiration (ET) can increase with solar radiation during the early transition season,even duringmoderate droughts induced by ENSO [Juarez et al., 2007]. In this case, the ENSO influence would bemitigated by the deep root system of the rainforest. The frequency and intensity of cold front incursions into theAmazon is not only influenced by ENSO and the Pacific Decadal Oscillation (PDO) [Grassi et al., 2012; Vera et al.,2004; Vera et al., 2002] but may also be influenced by southern Atlantic SSTs, Antarctic Oscillation [Garreaud,2000], radiative forcing by greenhouse gases [Deser et al., 2010], and biomass burning [Zhang et al., 2009].

In this paper, we investigate the relative importance of Pacific and Atlantic SSTAs, moisture transport,extratropical cold air incursion, convective conditions, and land surface conditions in determining theinterannual variability of the wet season onset over the southern Amazon. In a companion paper, wehighlight the significant change of the wet season onset during the past decades using the same rain gaugedata sets [Fu et al., 2013].

2. Data and Methodology2.1. Data Sets

The rainfall data sets used in this study are the National Oceanic and Atmospheric Administration (NOAA)Climate Prediction Center improved 1° gridded historical daily precipitation product for Brazil from 1948 until2007 (referred to as Silva) [Silva et al., 2007] and the NOAA Climate Diagnostics Center (CDC) 1° gridded dailyprecipitation product for Brazil and some other adjacent countries from 1940 until 2011 (updated andreferred to as SA24) [Liebmann and Allured, 2005]. Prior to 1979, too few stations were available overAmazonia to obtain relatively reliable rainfall climatology. Starting from 1979, there were more than 200 raingauge stations over the Amazon [Liebmann and Allured, 2005]. Thus, we select the period from 1979 to 2007.The study region, southern Amazonia, is defined as a rectangular area of 70°W–50°W and 15°S–5°S.

On the one hand, the SA24 data lack spatial coverage in the southwestern part of the southern Amazon[Liebmann and Allured, 2005], while the Silva data have included rain gauges in the region [Silva et al., 2007].

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2315

On the other hand, the SA24 data have frequentupdates for the number of stations included andcovers a longer period. Therefore, for the timeperiod of 1979 to 2007, we average the two datasets at each time step; we only use SA24 from2008 to 2011. The generated data are referred toas PM. Since the two data sets both cover thetime period of 1979–2007, we do the analysisonly based on this period for PM data.

Figure 1 shows the annual cycle of the pentadprecipitation from the two rain gauge data sets.Generally, SA24 and Silva are consistent witheach other, showing clear rainfall seasonalityover the southern Amazon, with rain amount of8–10mm/d in the wet season and 0–4mm/d inthe dry season.

Outgoing longwave radiation (OLR) is anindicator of cloud amount over the tropics, andit has been used to detect the dry seasonlength [Kousky et al., 1984]. The NOAA

interpolated OLR product provided by the NOAA/Earth System Research Laboratory/Physical SciencesDivision is a daily gridded data set constructed using spatial and temporal interpolation from June 1974 tothe present [Liebmann and Smith, 1996]. Given sparse sampling of the rain gauges and impacts of strongconvection on the daily mean OLR estimates, we do not look at the daily-scale changes of these data. Instead,it provides an alternative way for the spatial pattern of the dry season length in the southern Amazon.

NOAA Extended Reconstructed SST v3b analysis from 1854 to present was derived from the InternationalComprehensive Ocean-Atmosphere Data Set, where missing values are filled in by statistical methods [Smithet al., 2008]. It is used to calculate the tropical Atlantic SSTA gradient (AtlG), defined as the difference betweennorth tropical Atlantic SSTA (NTA) and south tropical Atlantic SSTA (STA). Niño 3 and Niño 4 indices areobtained from http://www.cdc.noaa.gov/data/climateindices/list/. AMO index is from http://www.cdc.noaa.gov/data/timeseries/AMO. PDO index is from http://jisao.washington.edu/pdo/.

Zonal wind, meridional wind, geopotential height, specific humidity, and surface sensible and latent heatfluxes from the National Centers for Environmental Prediction/National Center for Atmospheric Research(NCEP/NCAR) reanalysis [Kalnay et al., 1996] are used. NCEP/NCAR has heating profiles in remarkableagreement with ECMWF 40 years Re-Analysis (ERA40) over tropical South America, which implies thatprecipitation rate from heating diagnosis of the two reanalysis data sets is similar to the station and satellitemeasurements [Chan and Nigam, 2009]. Clarke et al. [2010] also found that NCEP/NCAR and ERA40 doreasonably well in reproducing the rain gauge-observed mean dry season length compared with rain gaugenetwork over the Amazon basin, suggesting that the reanalyses may be suitable for studying interannualvariability. Therefore, using NCEP/NCAR reanalysis to calculate certain dry season factors is appropriate.

Latent heating estimates for precipitation derived from Tropical Rainfall Measurement Mission (TRMM)products available from 1998 to 2010 are also used to determine the dominant rainfall type in the dry seasonover Amazonia and to verify its consistency with anomalous CIN derived from reanalysis. The convectiveprofile is dominated by heating throughout the whole troposphere, whereas the stratiform profile is assumedto be positive above the climatological 0 K level and negative below [Schumacher et al., 2004]. The derivation,description, and interpretation of these data are well documented in previous studies [Schumacher et al.,2004; Shige et al., 2004].

2.2. Methodologies

The method to find the wet season onset dates using the pentad precipitation data sets is adapted from Liand Fu [2004]. We set up three criteria to determine the onset pentad number: (i) at least six out of eightpentads before the date in which rainfall is lower than the climatological mean; (ii) at least six out of eight

Figure 1. The annual cycle of precipitation averaged over thesouthern Amazon (70°W–50°W, 15°S–5°S) during 1979–2011 (SA24)and 1979–2007 (Silva). The dates and the pentad number (inparenthesis) are shown as x axis. The shading areas represent±1σ of the interannual variation of rainfall for each data set.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2316

pentads after the date in which rainfall ishigher than the climatological mean; and (iii)rainfall in the date is higher than or equal tothe climatological mean. We calculate theonset dates for every single year. Thecomparisons of different methods definingthe onsets have also been discussed in Liand Fu [2004].

Since the change of the wet season onsetssignificantly contributes to the change of thedry season length [Fu et al., 2013], we furtherlook at the spatial patterns of the dry seasonlengths (Figure 2) from NOAA OLR product(method from Kousky [1988]) and PM data(method from Li et al. [2006]). Both patternsclearly show the increase of dry season lengthfrom the northwest to southeast Amazon.Since the spatial pattern of the OLR isdominated by those of cirrus and anvil clouds[Mitchell and Finnegan, 2009], it is smootherthan that from the PM data.

The Southern Hemisphere subtropical jet(SHSJ), which forms the southern boundaryof the Hadley cell, is crucial for cold airincursions to the Amazon. Lower troposphereconvergence is introduced by the leadingedge of cold air masses, which triggers thewet season onsets [Li and Fu, 2006]. Theinfluence of cold frontal systems onprecipitation in South America is welldocumented in previous studies [Garreaud,

1999, 2000; Kousky, 1979; Siqueira et al., 2005; Vera and Vigliarolo, 2000; Vera et al., 2002]. The occurrence of coldair incursions is mostly frequent between 5°S and 20°S during the transition season from dry to wet conditions[Machado et al., 2004]. Climatologically, the cross-barrier flow at the crest of the Andes is just under the SHSJ axisand the presence of a SHSJ entrance region coincides with downward (upward) motion on the poleward(equatorward) side of the jet [Garreaud, 1999]. A poleward shift of the SHSJ observed during recent decades [Fuand Lin, 2011; Kang et al., 2011] would reduce the equatorward incursion of cold frontal systems into theAmazon. The latitude where 200hPa zonal winds averaged along the Eastern Pacific-South American sector(100°W–50°W) equal 30m/s on the equator side is defined as the SHSJ position.

The local convective conditions, represented by CAPE and CIN, are defined as the accumulated buoyant energyfrom the level of free convection (LFC) to the limit of convection and the accumulated negative buoyant energyfrom parcel start point to the LFC, respectively, as in Fu et al. [1999]. Bowen ratio (BR) is defined as the ratio ofsensible heat flux to latent heat flux and is used to quantify the surface heat conditions [Bowen, 1926].

All the correlation coefficients are computed based on detrended time series, and their significances areestimated based on the effective number of freedom following Bretherton et al. [1999].

2.3. Identifying the Preseasonal Conditions Associated With the Early and Late Wet Season OnsetsOver Southern Amazonia

The events of the late and early wet season onsets are selected based on the time series of the onsets usingPM data (Figure 3a). The dashed lines show the departure from one standard deviation (1σ) of the onset dates.Early onsets are defined when the onset pentad is earlier than 1σ relative to the mean, while late onsets aredefined when the onset pentad is later than 1σ from the mean. Specifically, early onsets are no later than Oct

OLR

25 30

30

35

35

3535

40

40

40

Pentads

70W 65W 60W 55W 50W15S

10S

5S

15

20

25

30

35

40

45

PM

20

25

25

253030

3035

3535

35

354040

40 40

4040

45

Pentads

70W 65W 60W 55W 50W15S

10S

5S

15

20

25

30

35

40

45

Figure 2. The mean dry season lengths in the southern Amazonderived fromNOAAOLR and PM data. The unit is number of pentads.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2317

8–12 (57th pentad) and late onsets are no earlierthan Nov 7–11 (63th pentad). Based on thiscriterion, we select five early-onset years (1981,1985, 1986, 1988, and 2006) and six late-onsetyears (1994, 1995, 1997, 2002, 2004, and 2005).One should notice that rainfall does notsignificantly change over the 1990s compared tothe 1980s (not shown), suggesting that the delayof the onsets does not simply depend on thedrying or decadal variability of the Amazonrainfall under global warming.

Figure 3b shows the composite time series ofpentad rainfall for the whole time period, early-onset years, and late-onset years. The onset dateis consistent with a sudden increase of rainfall.The dip, interestingly seen in the five pentadsprior to the onset of the late-onset years, mayresult from inadequate cold air incursions orstable thermodynamic conditions [Li and Fu,2006]. The similar rainfall amounts in wet and dryseasons for the two composites suggest that theinterannual variation of the onsets is not simplydue to persistence of rainfall anomalies, and dryseason and wet season rainfall does not show aclear in-phase change (not shown).

We use the composite pattern of the difference ofJuly SSTs between late and early onsets in thetropical Pacific to guide our determination of theSST indices (Figure 4). The pattern bearssimilarities with El Niño-type SSTs, and themagnitudes of the differences over the eastern

Year

Pen

tad

Late Onset

Early Onset

(a)

1980 1985 1990 1995 2000 2005

55

60

65

70y = 0.14x−217.77

Figure 3. (a) The time series of thewet season onsets from SA24(red), Silva (blue), and PM data sets (black) averaged over thesouthern Amazon (70°W–50°W, 15°S–5°S). The black line is thesame as the blue line in Fu et al. [2013, Figure 1]. (b) Compositetime series of precipitation from PM data. The vertical dottedline represents the defined initiation of the wet season. Theblack line is for 1979–2007, the green line is for early-onset years,and the magenta line is for late-onset years. The shading areasrepresent ±1σ of the interannual variation

Nino3Nino4

NTA

STA

SSTLate−Early

(Jul)

K

60E 120E 180 120W 60W 0 60E

80S

60S

40S

20S

EQ

20N

40N

60N

80N

−0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8

Figure 4. The composite pattern of SSTs in July between the late and earlywet season onset years. Significant difference basedon one-way analysis of variance test is shown in shading areas. The boxes represent the domains for Niño 3, Niño 4, NTA, and STA.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2318

Pacific are larger than 0.8 K. A cooling of the southeastern Atlantic and a warming of the north tropicalAtlantic and southwest extratropical Atlantic also appears favorably in the late-onset years. The northernAtlantic may influence the southern Amazon through its influence on the large-scale moisture transport [Fuet al., 2001;Marengo et al., 2008;Moura and Shukla, 1981; Zeng et al., 2008]. The SST difference in April shows asimilar pattern to that of July, but it emphasizes more on central Pacific (not shown). Based on the compositepreseasonal SSTs anomalies and their potential connection suggested in literature, we will evaluate Niño3,Niño4, PDO, AMO, and AtlG as potential predicting factors for interannual variation of the wet season onsetover the southern Amazon.

We also evaluate the difference of the wind field and relative vorticity at 850 hPa and 200 hPa in Augustbetween late and early onsets to identify the areas with strongest preseasonal wind anomalies (Figure 5). Atthe 850 hPa level (Figure 5a), a dipole of anomalous cyclonic vorticity over the Altiplano plateau andanticyclonic vorticity over western Amazonia associated with a late wet season onset suggests a southwarddisplacement of synoptic disturbances and, consequently, a weakening of the cold front incursions into thesouthern Amazonia basin. In addition, the South American low-level jet (SALLJ) also appears to bestrengthened, leading to an enhancedmoisture export from southern Amazonia to La Plata basin. The dipoleappears to be part of a series of anomalous relative vorticity dipoles originating from the equatorial CentralPacific (Figure 5b). At 200 hPa (Figure 5b), the differences of the zonal wind (contour) and vorticity (shaded)fields between the late and early wet season onset suggest a poleward displacement of the subtropical jetsand enhanced subtropical anticyclonic vorticity over Southeastern Pacific. Such changes lead to a strongercross-Andes flow over 5°–15°S, in the upstream of the southern Amazonia. The stronger westerly cross-Andesflow north of 20°S is consistent with and physically responsible for the strengthened SALLJ shown inFigure 5a [e.g.,Marengo et al., 2004; Vera et al., 2002;Wang and Fu, 2004]. The latter enhances moisture exportfrom the southern Amazonia to La Plata basin and strengthens the cyclonic activities over the latter region.

3 m/s

VI

850hPaLate−Early (Aug) s−1(a)

Altiplano

La Plata

80W 60W 40W60S

40S

20S

EQ

20N

−1

−0.5

0

0.5

1x 10−5

Figure 5. The composite pattern of August (a) 850 hPa wind (vector) and relative vorticity (shaded) and (b) 200 hPa zonalwind (contour) and relative vorticity (shaded) between the late and early wet season onset years (left) and 200 hPa zonalwind averaged between 100°W and 50°W for early- and late-onset composites (right). The latitude where 200 hPa zonalwind averaged along the Eastern Pacific-South American sector (100°W–50°W) equals 30m/s on the equator side is definedas the SHSJ position. The boxes represent the domains for VI and U850 (the southern Amazon).

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2319

This weather pattern shift can weaken the cold front incursions into the southern Amazonia and contributesto late wet season onset over southern Amazonia [Li and Fu, 2006].

Based on literature review and Figures 4 and 5, we explore the tropical SSTs (July), atmospheric circulation,and thermodynamic and surface conditions (August) as crucial factors listed in Table 1. Five of them are SSTindices, as a domain average of SSTA: Niño 3, Niño 4, PDO, AMO, and AtlG. Three of these factors are capturedfrom wind fields: SHSJ, V Index at 925 hPa (VI, 925 hPa meridional wind averaged over 75°W–65°W, 5°S–5°N),and zonal wind at 850 hPa (U850, averaged over the southern Amazon). The other three factors represent thelocal thermodynamic and surface conditions in the southern Amazon: CAPE, CIN, and BR. Table 1 and boxes inFigures 4 and 5 both illustrate the domains and time for the 11 factors.

3. Links Between Dry Season Conditions and the Wet Season Onsets3.1. Correlations Between Preseasonal Conditions and Change of the Wet Season Onsets

Figure 6 shows the relationships between the onsets and the 11 factors. Based on the confidence level of95%, two of the five SST indices are significantly correlated with the onsets. Niño 3 has a correlation

Table 1. A List of Selected Dry Season Factors (July for SSTA and August for Others)

Category Factor Variables Domain

Oceanic variability Niño 3 Sea surface temperature 150°W–90°W, 5°S–5°NNiño 4 160°E–150°W, 5°S–5°NAMO Oceanic area, 78°W–10°E, 0°–70°NPDO Leading PC, North Pacific, poleward of 20°NAtlG 60°W–30°W, 5°N–25°N and 30°W–0°, 25°S–5°S

Atmospheric circulation SHSJ 200 hPa zonal wind Eastern Pacific-South American sector, 100°W–50°WVI 925hPa meridional wind 75°W–65°W, 5°S–5°N

U850 850 hPa zonal wind 70°W–50°W, 15°S–5°SThermodynamic condition CAPE & CIN Temperature, specific humidity, geopotential height,

topography, surface pressure70°W–50°W, 15°S–5°S

Surface condition BR Latent and sensible heat fluxes 70°W–50°W, 15°S–5°S

R=0.48,p=0.01

Nin

o3

55 60 65

−2

0

2

R=0.56,p=0

Nin

o4

55 60 65−1

0

1R=0.05,p=0.8

PD

O

55 60 65−2

0

2

4R=−0.19,p=0.33

AM

O

55 60 65

55 60 65 55 60 65 55 60 65

55 60 65 55 60 65 55 60 65

55 60 65

−0.4

−0.2

0

0.2

0.4

R=0.12,p=0.54

Atl

G

−1

0

1R= −0.42,p=0.02

SH

SJ

−40

−30

−20

R=0,p=1

VI

0

1

2

R=0.34,p=0.07

U85

0

−6

−4

−2

R= −0.37,p=0.05

CA

PE

0.5

1

R=0.07,p=0.71

CIN

0.04

0.06

0.08

R=0,p=0.99

BR

0

1

2

3

Figure 6. The scatter plots of the wet season onsets and the 11 factors considered. The regions for these variables aresummarized in Table 1.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2320

coefficient (R, p value = 0.01) equal to 0.48, and Niño 4 has R= 0.56, p< 0.01. PDO, AMO, and AtlG do notappear to be correlated with the onsets at the interannual time scale. The linkage between the preseasonalSSTA in the tropical Pacific and wet season onsets over the southern Amazon is consistent with previousstudies [Marengo et al., 2001]. The SHSJ is negatively correlated with the onsets with R=�0.42, p=0.02, andU850 has R= 0.34, p= 0.07. The former is consistent with an earlier finding that more poleward SHSJ wouldprevent the cold frontal system from entering the southern Amazon, whereas the latter is presumablyassociated with less-frequent deeply penetrated and vigorous continental type of convective systemsassociated with weakened easterly winds at 850 hPa as observed during both field campaigns and satelliteobservations [Petersen et al., 2006, 2002; Rickenbach, 2002; Silva Dias, 2002]. Such deep convective systemswould increase the buoyance of the lower troposphere and destabilize the atmosphere. As a consequence, theelevated diabatic heating can trigger intense rainfall along with the cold fronts passing by each time, whichfavors the wet season onsets over the southern Amazon [Li and Fu, 2006]. Although the p value for U850 issomewhat larger than 0.05, it provides some evidence for the importance of low-level zonal winds. Among thelocal effects, CAPE is negatively correlated with the onsets with R=�0.37 (p=0.05), while the correlationbetween CIN and the wet season onset is not statistically significant (R=�0.07, p=0.71). This is also consistentwith the previous finding that smaller CAPE during the dry season inhibits local thermodynamically drivenconvection that is central in initiating the transition from dry to wet season [Li and Fu, 2004].

3.2. Composite Analysis for the Robustness as Important Factor

The correlations presented in section 3.1 show the important preseasonal conditions. In order to statisticallydemonstrate that these conditions are crucial for late-onset years, we need to check whether theseconditions have robust anomalies (>1σ). In Figure 7, we compare the differences in the SHSJ, U850, andCAPE/CIN between the late and early wet season onsets. The error bars represent ±1 standard error, and thesample size is 6 for late onsets and 5 for early onsets. Given such small sample sizes, it is difficult to determinewhether the true differences are significant or not. However, both of them are different beyond 2 standarderrors between the late and early-onset years. The SHSJ shows a poleward positioning during the late-onsetyears. U850 shows stronger westerly wind anomalies over the Amazon during the late-onset years, associatedwith the stronger SALLJs observed in Figure 5a.

We also compare the differences in atmospheric thermodynamic structure and land surface conditionsbetween the late and early wet season onsets (Figure 7). During late-onset years, there are larger CIN and BRvalues, associated with smaller CAPE values. CAPE is clearly different beyond 2 standard errors between late

Figure 7. The domain-averaged and composite time-averaged SHSJ, U850, CAPE, CIN, and BR in August between the lateand early wet season onsets. Error bars represent ±1 standard error.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2321

and early-onset years based on the error bars (1 standard error), as it is an important thermodynamiccondition for the dry-to-wet transition season [Fu et al., 1999]. The about 2 standard deviation differences forCIN and BR between late and early wet season onsets imply some influence of CIN and BR. However, oneshould notice that the correlation between the onsets and CIN or BR is insignificant (Figure 6). Thus, theymight be important in extreme cases, e.g., the much larger CIN in the 2005 drought year (Table 2). Thepotential uncertainties in surface fluxes modeled by NCEP/NCAR reanalysis could contribute to the poorrelationship between BR and the onsets.

Latent heating estimates from TRMM show that the maximal heating over the Amazon occurs at the height of2–4 km and the profiles are all positive throughout the troposphere (Figure 8), indicating that shallowconvective precipitation dominates in the dry season of Amazonia. In the three late-onset years (2002, 2004, and2005), latent heating ismuch smaller than the other years throughout thewhole troposphere from the availabletime period of 1998–2010. This coincides with the limiting thermodynamic condition for convection indicatedby low CAPE and high CIN, enhancing our confidence in interpreting the results from reanalysis data.

In Table, 2, we compare the anomalous preconditions associated with each late wet season onsets. Theweaker easterly winds, associate with late wet season onset in 1994 and 1995, are generally larger than onestandard deviation (1σ, Table 2). By contrast, CAPE anomalies were smaller than 1σ prior to the late wetseason onset in 1997. In 2002, the strongest delay of the wet season onsets throughout the whole period wasled by a strong poleward shift of SHSJ and strongly anomalous CAPE/CIN. The late onset in 2005 followseither smaller easterly winds or a stronger CIN.

3.3. Links Between the Preseasonal Conditions and External Planetary-Scale Climate Variability

In Figure 9, we explore the links between the planetary-scale climate variability modes and regionallyanomalous dry season conditions. Results show that the SHSJ in August over the South American sector is

mainly correlated with PDO index in July (p< 0.1).U850 is mostly correlated with Niño 4 and VI(p< 0.1). Such a correlation may be a result of twopossible mechanisms: (i) the warming of thetropical eastern Pacific results in variations intropical Atlantic SSTA and weakens the cross-equatorial flow and the easterly winds from thetropical Atlantic to tropical South America [Cartonet al., 1996;Mo and Häkkinen, 2001], and (ii) El Niñoevents and SST anomalies over the subtropicalsouth central Pacific could also induce thepoleward displacement of cold fronts [Barros andSilvestri, 2002; Vera and Vigliarolo, 2000; Veraet al., 2002].

CIN in the southern Amazon is significantlycorrelated with the PDO and BR. Physically, CIN isdetermined by land surface dryness, as indicated byBR, and the thermodynamic stability of the lowertroposphere, which is controlled by subsidenceabove the boundary layer. The correlation between

Table 2. The Values of the Contributing Factors in Late-Onset Years

Late-Onset Years Onset Date (Pentad) SHSJ (°) U850 (m/s) CAPE/CIN (kJ/kg)

1994 65 �24.7 �3.40a 0.76/0.0451995 63 �20.9 �3.52a 0.77/0.0541997 64 �19.7 �4.61 0.71a/0.0552002 68 �27.0a �4.42 0.68a/0.060a

2004 63 �20.2 �3.87 0.72/0.0582005 64 �24.3 �3.68a 0.89/0.066a

aValues are greater than the mean ±1σ.

Latent Heating (K day−1)

Hei

gh

t (k

m)

−0.1 0 0.1 0.2 0.3 0.40

2

4

6

8

10

12

14

16

LateOthers

Figure 8. The latent heating profile for TRMM precipitationaveraged over the southern Amazon in August of late-onsetyears’ and other years’climatology (1998–2003 and 2006–2010).

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2322

AtlG and BRmay be caused by tropical Atlantic’sinfluence on Amazon rainfall during late australfall, which in turn influences the BR during thesubsequent dry and transition season.

VI is also correlated with CAPE and BR but notwith ENSO and Atlantic SSTAs. Thus, itsinterannual variation presumably results fromthe variation of convection over the southernAmazon, which in turn produces elevateddiabatic heating that enhances northerly(negative) VI [Li and Fu, 2004].

4. Interannual Variations of theWetSeason Onsets Explained by theAnomalous Preseasonal Conditions

To explore how much the dry season factorscan explain the interannual variation of wet

season onsets, we need to minimize the multicollinearity between SSTAs over the Pacific and Atlantic oceansas well as the atmospheric circulation and local conditions discussed above. Two statistical methods,stepwise regression and partial least squares (PLS) regression, are employed for this purpose. Stepwiseregression selects the most important variables that are uncorrelated with each other as predicting factors byan automatic procedure and some selection criteria, while PLS regression uses all the variables and projectsthem to a new space based on the PLS algorithm. The PLS regression bears some similarities with principalcomponents analysis. More details of the two methods are documented in previous studies [Hocking, 1976;Tenenhaus et al., 2005].

One important part of this study is to understand how much interannual variability of the onsets can beexplained by these factors other than Niño indices. Thus, we set up two groups of factors from the 11 ones.One only has Niño 3 and Niño 4 (referred to as Niño-only), while the other one also includes SHSJ, U850, andCAPE (referred to as Niño+). For the Niño-only group, the correlation coefficient between the predicted

onsets using stepwise regression andobserved onsets is 0.60, while thatbetween the predicted onsets usingPLS regression and observed onsets is0.62 (Figure 10a). The automaticselection procedure of stepwisemethod selects Niño 4, which explains36% of the total variance, andindicates that Niño 3 and Niño 4 arecovariates. The PLS regression showsthat for the Niño-only group thereconstructed wet season onsetchange explains about 38% of thetotal variance (Figure 10b). From thesame calculations for Niño+ group,we can see that the selected factors bystepwise procedure, Niño 4, SHSJ, andCAPE, can explain 57% of the totalvariance of the wet season onset. Thedifference between the two groupshighlights the importance ofmonitoring of the SHSJ, U850, andCAPE, which are not significantly

Nino3

Nino4

PDOAM

OAtlG

SHSJ VI

U850

CAPECIN BR

BR

CIN

CAPE

U850

VI

SHSJ

AtlG

AMO

PDO

Nino4

Nino3

−0.6

−0.4

−0.2

0

0.2

0.4

0.6

Figure 9. The correlation matrix between the 11 factors consid-ered. Stars and solid circles indicate that p value is less than 0.05and 0.1, respectively.

Pen

tad

(a)

1980 1985 1990 1995 2000 200550

55

60

65

70

PM

PLS (Nino+)Stepwise (Nino+)PLS (Nino)Stepwise (Nino)

1 2 3 4 520

30

40

50

60

70

# of PLS

Exp

lain

ed P

erce

nt

Var

ian

ce (

%)

(b)

Nino+Nino

Figure 10. (a) Time series of the PM and predicted onsets from factorsusing the two methods considered. Blue lines represent the reconstructedonset dates using Niño 3, Niño 4, SHSJ, U850, and CAPE (referred to asNiño+), while red lines represent those using only Niño 3 and Niño 4(referred to as Niño). (b) The corresponding explained percent variance forthe number of PLS components. Blue is for Niño+, and red is for Niño.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2323

correlated with Pacific and Atlantic variability modes (Figure 9), for the predictability of the following onsetsince they can explain nearly 22% more of the total variance, in addition to that explained by the Niño 3 andNiño 4.

Figure 10 suggests that these three anomalous dry season conditions within the Amazon basin could provideadditional information in determining the rainfall variations in the dry season, which are crucial for thelikelihood of interannual variation of the wet season onset.

5. Conclusions and Discussion

This study explores characteristics of anomalous dry season conditions associated with the late onset ofthe subsequent wet season over the southern Amazon, in addition to the external oceanic influences dueto ENSO.

The results reported in sections 3 and 4 suggest that the interannual variation of the wet season onset overthe southern Amazon is correlated with dry season conditions in the Pacific and South America. This isbecause these anomalous dry season conditions directly influence the mechanisms critical for the transitionfrom dry to wet season over the Amazon. In particular, (i) a weaker CAPE/stronger CIN delays the initiation ofthis dry-to-wet season transition by limiting local thermodynamically driven convection. (ii) A strongerwesterly U850 and SALLJs can reduce the occurrence of deep and more vigorous convective systems thatprovide elevated diabatic heating. The SALLJs play a critical role in converting convective available potentialenergy into rotational kinetic energy in the upper troposphere, which is needed to overturn the monsooncirculation [Krishnamurti et al., 1998; Li and Fu, 2004]. (iii) A poleward displacement of the SHSJ over SouthAmerica would weaken the incursions of cold fronts into the southern Amazon region, which are the mainatmospheric dynamic trigger for the wet season onset over the southern Amazon. Although previous studieshave established the importance of these mechanisms for the climatological wet season onset, this studymore quantitatively demonstrates their importance in the interannual variability of the wet season onsetsover the southern Amazon.

These anomalous dry season conditions are largely forced by ENSO. However, how these SST variabilitymodes influence the large-scale and land surface conditions, which are key to the wet season onsets, has notbeenwell studied. This study suggests that the land surface feedback, the enhanced SALLJs, and lee cyclogenesisover extratropical South America associated with a poleward shift of the SHSJ can either dampen or amplifythe influences of Pacific and Atlantic SSTA on SHSJ, U850, and CAPE anomalies. These factors add nonlinearityto the relationships between ENSO, AMO-related tropical Atlantic SSTA, and changes in wet season onset overthe southern Amazon.

Although southern Amazon CIN is not correlated with the onset, its influence on wet season onsetchanges appears to be nonlinear since it might be responsible for the delayed onsets (Figure 7). While astrong CIN increase (>1σ) is important for delaying wet season onsets (Table 2), CIN anomalies overallexplain only a negligible percentage of the total variance of interannual wet season onsets as suggestedby NCEP reanalysis.

Furthermore, our result in Figure 5a shows enhanced northerly low-level wind, presumably associated withSALLJs, during the dry season prior to the late wet season onset. SALLJs are maintained by strong zonalpressure gradients due to lee cyclogenesis [Vera et al., 2002; Wang and Fu, 2004]. The poleward shift of theSHSJ (Figures 5b and 6) may promote lee cyclogenesis in the extratropical South America, enhancingextratropical segment of the SALLJs. The latter, in turn, would export more moisture from a closedcontinent of the Subtropical Atlantic High to the La Plata basin, feeding more water vapor and latentenergy to strengthen extratropical synoptic systems in that region. Thus, the enhanced northerly low-levelwind over southern Amazonia associated with poleward relationship implies that a potential positivefeedback to the poleward displacement of cold fronts during El Niño events [Robinson, 2002; Seager et al.,2003] may contribute to the nonlinearity between the wet season onset over the southern Amazonand ENSO.

Ding et al. [2011] suggest that the subtropical jet over South America is mainly influence by Rossby waveresponse to south central Pacific. Although the wave train pattern is comparable to the Pacific-South Americapattern associated with ENSO forcing, it is not significantly correlated with ENSO indices. Vera et al. [2002]

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2324

have shown that the SHSJ variation over South America can differ among subsequent ENSO events due tochanges of the SSTA in the extratropical Pacific, which are affected by the variations of the Southern AnnualMode (SAM) [Ciasto and Thompson, 2008]. Although SAM is partially related to El Niño [L’Heureux and Thompson,2006], a favorable configuration of the two may cause the SHSJ to move extremely poleward, which in turnweakens cold air incursions over the southern Amazon [Garreaud, 2000].

Li and Fu [2006] indicate that cold air incursions trigger intense rainfall in austral spring to the east of the Andeswhen the thermodynamic condition becomes unstable. The low-level atmospheric humidity and buoyancyincreases and high-level anticyclonic flow develops as the cold frontal systems pass by every time. As a result,moisture convergence intensifies and rain areas expand southeastward, resulting in the wet season onset.

Chen et al. [2011] demonstrate that evapotranspiration in late dry season is largely impacted by the terrestrialwater storage accumulated in previous months, which is linked to wintertime Atlantic and Pacific SSTsthrough their influence on precipitation. This mechanism enables potential fire risk prediction in the Amazonsimply based on SSTs. A late wet season onset is also followed by an increase of fire counts [Fu et al., 2013] andseems to be part of this cascade of processes.

Figure 11 summarizes the links between adjacent oceanic forcing and the key regional preseasonal conditionsaffecting the wet season onset over southern Amazonia and their underlying mechanisms suggested by thiswork and previous studies. The major external oceanic or planetary-scale circulation anomalies—mainly ElNiños and the positive phases of the AMO and SAM—could dry the land surface, weaken the moisturetransport from tropical Atlantic ocean to Amazonia, and induce a poleward displacement of the SHSJ in the dryseason. These changes together could consequently weaken the convective instability and the probability ofcold front incursions. They are not a linear function of the external forcing but can, in turn, weaken the diabaticheating in the middle and upper tropospheres and delay dry-to-wet season transition over Amazonia.

The variability of the wet season onset over southern Amazonia is likely linked to both preseasonal conditionsand also random atmospheric and oceanic variability during the dry-to-wet transition. Our analysis suggeststhat the anomalous climate conditions within Amazonia during the previous dry season could provide moreinformation for the subsequent wet season onset variability than those of oceanic forcing alone, even thoughthe former is nonlinearly related to the latter. This study suggests that a clear understanding of suchnonlinear relationship between the atmospheric anomalous thermodynamic and dynamic conditions andthe oceanic and planetary-scale atmospheric variabilities may provide a key for improving our interpretationof the interannual variations of the wet season onsets over Amazonia.

Figure 11. A schematic diagram showing the primary processes affecting the wet season onset over the southern Amazon.Dark boxes and arrows indicate the preseasonal conditions and their links to planetary-scale climate variability established inthis study. Thin boxes indicate the underlying processes of the links suggested in previous studies. Two-way arrows indicatepositive feedbacks between the two processes.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2325

Appendix A: List of Acronyms and Abbreviations

AMO Atlantic Multidecadal OscillationAtlG tropical Atlantic SSTA gradientBR Bowen ratioCAPE convective available potential energyCIN convective inhibitionENSO El Niño-Southern OscillationET evapotranspirationNTA north tropical Atlantic SSTAPDO Pacific Decadal OscillationPLS partial least squaresOLR outgoing longwave radiationSALLJ South America low-level jetsSHSJ Southern Hemisphere subtropical jetSSTA sea surface temperature anomalySTA south tropical Atlantic SSTAU850 zonal wind at 850 hPaVI V Index

ReferencesAckerley, D., B. B. B. Booth, S. H. E. Knight, E. J. Highwood, D. J. Frame, M. R. Allen, and D. P. Rowell (2011), Sensitivity of twentieth-century

Sahel rainfall to sulfate aerosol and CO2 forcing, J. Clim., 24(19), 4999–5014, doi:10.1175/jcli-d-11-00019.1.Arias, P. A., R. Fu, C. D. Hoyos, W. Li, and L. Zhou (2010), Changes in cloudiness over the Amazon rainforests during the last two decades:

Diagnostic and potential causes, Clim. Dyn., 37(5-6), 1151–1164, doi:10.1007/s00382-010-0903-2.Asner, G. P. (2009), Tropical forest carbon assessment: Integrating satellite and airborne mapping approaches, Environ. Res. Lett., 4(3), 034009,

doi:10.1088/1748-9326/4/3/034009.Barros, V. R., and G. E. Silvestri (2002), The relation between sea surface temperature at the subtropical south-central Pacific and precipitation

in southeastern South America, J. Clim., 15, 251–267.Bowen, I. S. (1926), The ratio of heat losses by conduction and by evaporation from any water surface, Phys. Rev., 27(6), 779–787.Bretherton, C. S., M. Widmann, V. P. Dymnikov, J. M. Wallace, and I. Blade (1999), The effective number of spatial degrees of freedom of a

time-varying field, J. Clim., 12, 1990–2009.Carton, J. A., X. Cao, B. S. Giese, and A. M. Da Silva (1996), Decadal and interannual SST variability in the tropical Atlantic Ocean, J. Phys.

Oceanogr., 26(7), 1165–1175, doi:10.1175/1520-0485(1996)026<1165:daisvi>2.0.co;2.Chan, S. C., and S. Nigam (2009), Residual diagnosis of diabatic heating from ERA-40 and NCEP reanalyses: Intercomparisons with TRMM,

J. Clim., 22(2), 414–428, doi:10.1175/2008JCLI2417.1.Chen, Y., J. T. Randerson, D. C. Morton, R. S. DeFries, G. J. Collatz, P. S. Kasibhatla, L. Giglio, Y. Jin, and M. E. Marlier (2011), Forecasting fire

season severity in South America using sea surface temperature anomalies, Science, 334(6057), 787–791, doi:10.1126/science.1209472.Chiang, J. C. H., and A. H. Sobel (2002), Tropical tropospheric temperature variations caused by ENSO and their influence on the remote

tropical climate, J. Clim., 15(18), 2616–2631, doi:10.1175/1520-0442(2002)015<2616:tttvcb>2.0.co;2.Ciasto, L. M., and D. W. J. Thompson (2008), Observations of large-scale ocean-atmosphere interaction in the Southern Hemisphere, J. Clim.,

21(6), 1244–1259, doi:10.1175/2007JCLI1809.1.Clarke, R. T., C. A. Bulhoes Mendes, and D. Costa Buarque (2010), A comparison of extreme rainfall characteristics in the Brazilian Amazon

derived from two gridded data sets and a national rain gauge network, J. Geophys. Res., 115, D13104, doi:10.1029/2009JD013217.Cochrane, M. A., A. Alencar, M. D. Schulze, C. M. Souza Jr., D. C. Nepstad, P. Lefebvre, and E. A. Davidson (1999), Positive feedbacks in the fire

dynamic of closed canopy tropical forests, Science, 284(5421), 1832–1835.Deser, C., A. Phillips, V. Bourdette, and H. Teng (2010), Uncertainty in climate change projections: The role of internal variability, Clim. Dyn.,

38(3-4), 527–546, doi:10.1007/s00382-010-0977-x.Ding, Q., E. J. Steig, D. S. Battisti, and M. Küttel (2011), Winter warming in West Antarctica caused by central tropical Pacific warming, Nat.

Geosci., 4(6), 398–403, doi:10.1038/ngeo1129.Fernandes, K., et al. (2011), North tropical Atlantic influence on western Amazon fire season variability, Geophys. Res. Lett., 38, L12701,

doi:10.1029/2011GL047392.Friedlingstein, P., et al. (2006), Climate-carbon cycle feedback analysis: Results from the C4MIP model intercomparison, J. Clim., 19(14),

3337–3353, doi:10.1175/JCLI3800.1.Fu, Q., and P. Lin (2011), Poleward shift of Subtropical Jets inferred from satellite-observed lower-stratospheric temperatures, J. Clim., 24(21),

5597–5603, doi:10.1175/jcli-d-11-00027.1.Fu, R., and W. Li (2004), The influence of the land surface on the transition from dry to wet season in Amazonia, Theor. Appl. Climatol., 78,

97–110, doi:10.1007/s00704-004-0046-7.Fu, R., B. Zhu, and R. E. Dickinson (1999), How do atmosphere and land surface influence seasonal changes of convection in the tropical

Amazon?, J. Clim., 12, 1306–1321.Fu, R., R. E. Dickinson, M. Chen, and H. Wang (2001), How do tropical sea surface temperature influence the seasonal distribution of preci-

pitation in the equatorial Amazon?, J. Clim., 14, 4003–4026.Fu, R., et al. (2013), Increased dry-season length over southern Amazonia in recent decades and its implication for future climate projection,

Proc. Natl. Acad. Sci. U. S. A., 110(45), 18,110–18,115, doi:10.1073/pnas.1302584110.

AcknowledgmentsThis study was funded by the NationalScience Foundation (NSF) AGS-0937400,the NOAA Climate Program OfficeModeling, Analysis, Prediction andProjection (MAPP) Program throughgrant NA10OAAR4310157, and theNASA Earth and Space ScienceFellowship. Paola Arias was partiallysupported by the Jackson School ofGeosciences through startup fund toRong Fu.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2326

Garreaud, R. D. (1999), Cold air incursions over subtropical and tropical South America: A numerical case study, Mon. Weather Rev., 127(12),2823–2853, doi:10.1175/1520-0493(1999)127<2823:caiosa>2.0.co;2.

Garreaud, R. D. (2000), Cold air incursions over subtropical South America: Mean structure and dynamics, Mon. Weather Rev., 128, 2544–2559.Grassi, B., G. Redaelli, P. O. Canziani, and G. Visconti (2012), Effects of the PDO phase on the tropical belt width, J. Clim., 25(9), 3282–3290,

doi:10.1175/jcli-d-11-00244.1.Hocking, R. R. (1976), The analysis and selection of variables in linear regression, Biometrics, 32(1), 49.Juarez, R. I. N., M. G. Hodnett, R. Fu, M. L. Goulden, and C. von Randow (2007), Control of dry season evapotranspiration over the Amazonian

forest as inferred from observations at a southern Amazon forest site, J. Clim., 20(12), 2827–2839, doi:10.1175/JCLI4184.1.Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77(3), 437–471, doi:10.1175/1520-0477(1996)

077<0437:tnyrp>2.0.co;2.Kang, S. M., L. M. Polvani, J. C. Fyfe, and M. Sigmond (2011), Impact of polar ozone depletion on subtropical precipitation, Science, 332(6032),

951–954, doi:10.1126/science.1202131.Kousky, V. E. (1979), Frontal influences on northeast Brazil, Mon. Weather Rev., 107(9), 1140–1153, doi:10.1175/1520-0493(1979)107<1140:

fionb>2.0.co;2.Kousky, V. E. (1988), Pentad outgoing longwave radiation climatology for the South American sector, Rev. Bras. Meteorol., 3, 15.Kousky, V. E., M. T. Kagano, and I. F. A. Cavalcanti (1984), A review of the Southern Oscillation: Oceanic-atmospheric circulation changes and

related rainfall anomalies, Tellus, 36(5), 490–504, doi:10.1111/j.1600-0870.1984.tb00264.x.Krishnamurti, T. N., H. S. Bedi, and W. Han (1998), Organization of convection and monsoon forecasts, Meteorol. Atmos. Phys., 67(1-4), 117–134,

doi:10.1007/BF01277505.Lee, J.-E., B. R. Lintner, C. K. Boyce, and P. J. Lawrence (2011), Land use change exacerbates tropical South American drought by sea surface

temperature variability, Geophys. Res. Lett., 38, L19706, doi:10.1029/2011GL049066.Lenters, J. D., and K. H. Cook (1997), On the origin of the Bolivian High and related circulation features of the South American climate,

J. Atmos. Sci., 54, 22.L’Heureux, M. L., and D. W. J. Thompson (2006), Observed relationships between the El Niño–Southern Oscillation and the extratropical

zonal-mean circulation, J. Clim., 19, doi:10.1175/JCLI3617.1.Li, W., and R. Fu (2004), Transition of the large-scale atmospheric and land surface conditions from the dry to the wet season over Amazonia

as diagnosed by the ECMWF re-analysis, J. Clim., 17, 2637–2651.Li, W., and R. Fu (2006), Influence of cold air intrusions on the wet season onset over Amazonia, J. Clim., 19, 257–275, doi:10.1175/JCLI3614.1.Li, W., R. Fu, and R. E. Dickinson (2006), Rainfall and its seasonality over the Amazon in the 21st century as assessed by the coupled models for

the IPCC AR4, J. Geophys. Res., 111, D02111, doi:10.1029/2005JD006355.Liebmann, B., and D. Allured (2005), Daily precipitation grids for South America, Bull. Am. Meteorol. Soc., 86(11), 1567–1570, doi:10.1175/

bams-86-11-1567.Liebmann, B., and J. A. Marengo (2001), Interannual variability of the rainy season and rainfall in the Brazilian Amazon basin, J. Clim., 14,

4308–4318.Liebmann, B., and C. A. Smith (1996), Description of a complete (interpolated) outgoing longwave radiation dataset, Bull. Am. Meteorol. Soc.,

77, 1275–1277.Machado, L. A. T., H. Laurent, N. Dessay, and I. Miranda (2004), Seasonal and diurnal variability of convection over the Amazonia: A

comparison of different vegetation types and large scale forcing, Theor. Appl. Climatol., 78(1), 61–77, doi:10.1007/s00704-004-0044-9.Marengo, J. A. (1992), Interannual variability of surface climate in the Amazon basin, Int. J. Climatol., 12(8), 853–863, doi:10.1002/

joc.3370120808.Marengo, J. A., B. Liebmann, V. E. Kousky, N. P. Filizola, and I. C. Wainer (2001), Onset and end of the rainy season in the Brazilian Amazon

basin, J. Clim., 14, 833–852.Marengo, J. A., W. R. Soares, C. Saulo, and M. Nicolini (2004), Climatology of the low-level jet east of the Andes as derived from the NCEP-

NCAR reanalyses: Characteristics and temporal variability, J. Clim., 17(12), 2261–2280, doi:10.1175/1520-0442(2004)017<2261:Cotlje>2.0.Co;2.

Marengo, J. A., C. A. Nobre, J. Tomasella, M. D. Oyama, G. Sampaio de Oliveira, R. de Oliveira, H. Camargo, L. M. Alves, and I. F. Brown (2008),The drought of Amazonia in 2005, J. Clim., 21(3), 495–516, doi:10.1175/2007JCLI1600.1.

Marengo, J. A., et al. (2010), Recent developments on the South American monsoon system, Int. J. Climatol., 32(1), 1–21, doi:10.1002/joc.2254.Marengo, J. A., J. Tomasella, L. M. Alves, W. R. Soares, and D. A. Rodriguez (2011), The drought of 2010 in the context of historical droughts in

the Amazon region, Geophys. Res. Lett., 38, L12703, doi:10.1029/2011GL047436.Mitchell, D. L., and W. Finnegan (2009), Modification of cirrus clouds to reduce global warming, Environ. Res. Lett., 4(4), 045102, doi:10.1088/

1748-9326/4/4/045102.Mo, K. C., and S. Häkkinen (2001), Interannual variability in the tropical Atlantic and linkages to the Pacific, J. Clim., 14(12), 2740–2762,

doi:10.1175/1520-0442(2001)014<2740:ivitta>2.0.co;2.Moura, A. D., and J. Shukla (1981), On the dynamics of droughts in northeast Brazil: Observations, theory and numerical experiments with a

general circulation model, J. Atmos. Sci., 38(12), 2653–2675, doi:10.1175/1520-0469(1981)038<2653:otdodi>2.0.co;2.Myneni, R. B., et al. (2007), Large seasonal swings in leaf area of Amazon rainforests, Proc. Natl. Acad. Sci. U. S. A., 104(12), 4,820–4,823,

doi:10.1073/pnas.0611338104.Neelin, J. D. (2003), Tropical drought regions in global warming and El Niño teleconnections, Geophys. Res. Lett., 30(24), 2275 doi:10.1029/

2003GL018625.Nepstad, D. C., C. R. de Carvalho, E. A. Davidson, P. H. Jipp, P. A. Lefebvre, G. H. Negreiros, E. D. da Silva, T. A. Stone, S. E. Trumbore, and S. Vieira

(1994), The role of deep roots in the hydrological and carbon cycles of Amazonian forests and pastures, Nature, 372(6507), 666–669.Oliveira, R. S., T. E. Dawson, S. S. Burgess, and D. C. Nepstad (2005), Hydraulic redistribution in three Amazonian trees, Oecologia, 145(3),

354–363, doi:10.1007/s00442-005-0108-2.Petersen, W. A., S. W. Nesbitt, R. J. Blakeslee, R. Cifelli, P. Hein, and S. A. Rutledge (2002), TRMM observations of intraseasonal variability in

convective regimes over the Amazon, J. Clim., 15(11), 1278–1294, doi:10.1175/1520-0442(2002)015<1278:tooivi>2.0.co;2.Petersen, W. A., R. Fu, M. Chen, and R. Blakeslee (2006), Intraseasonal forcing of convective and lightning activity in the southern Amazon as a

function of cross-equatorial flow, J. Clim., 19, 3180–3196.Rao, V. B., I. F. A. Cavalcanti, and K. Hada (1996), Annual variation of rainfall over Brazil and water vapor characteristics over South America,

J. Geophys. Res., 101(D21), 26,539–26,551, doi:10.1029/96JD01936.Rickenbach, T. M. (2002), Modulation of convection in the southwestern Amazon basin by extratropical stationary fronts, J. Geophys. Res.,

107(D20), 8040, doi:10.1029/2000JD000263.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2327

Robinson, W. A. (2002), On the midlatitude thermal response to tropical warmth, Geophys. Res. Lett., 29(8), 1190, doi:10.1029/2001GL014158.Schumacher, C., R. A. Houze Jr., and I. Kraucunas (2004), The tropical dynamical response to latent heating estimates derived from the TRMM

precipitation radar, J. Atmos. Sci., 61, 1341–1358.Seager, R., N. Harnik, Y. Kushnir, W. Robinson, and J. Miller (2003), Mechanisms of hemispherically symmetric climate variability, J. Clim., 16,

2960–2978.Shige, S., Y. N. Takayabu, W. Tao, and D. E. Johnson (2004), Spectral retrieval of latent heating profiles from TRMM PR data. Part I:

Development of a model-based algorithm, J. Appl. Meteorol., 43, 1095–1112.Silva Dias, M. A. F. (2002), Cloud and rain processes in a biosphere-atmosphere interaction context in the Amazon region, J. Geophys. Res.,

107(D20), 8072, doi:10.1029/2001JD000335.Silva, V. B. S., V. E. Kousky, W. Shi, and R. W. Higgins (2007), An improved gridded historical daily precipitation analysis for Brazil,

J. Hydrometeorol., 8(4), 847–861, doi:10.1175/JHM598.1.Siqueira, J. R., W. B. Rossow, L. A. T. Machado, and C. Pearl (2005), Structural characteristics of convective systems over South America related

to cold-frontal incursions, Mon. Weather Rev., 133(5), 1045–1064, doi:10.1175/MWR2888.1.Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore (2008), Improvements to NOAA’s historical merged land–ocean surface

temperature analysis (1880–2006), J. Clim., 21(10), 2283–2296, doi:10.1175/2007JCLI2100.1.Sombroek, W. (2001), Spatial and temporal patterns of Amazon rainfall, AMBIO: A J. Hum. Environ., 30(7), 388–396, doi:10.1579/0044-7447-

30.7.388.Tenenhaus, M., V. E. Vinzi, Y.-M. Chatelin, and C. Lauro (2005), PLS path modeling, Comput. Stat. Data Anal., 48(1), 159–205, doi:10.1016/j.

csda.2004.03.005.Vera, C. S., and P. K. Vigliarolo (2000), A diagnostic study of cold-air outbreaks over South America, Mon. Weather Rev., 128(1), 3–24,

doi:10.1175/1520-0493(2000)128<0003:adsoca>2.0.co;2.Vera, C. S., P. K. Vigliarolo, and E. H. Berbery (2002), Cold season synoptic-scale waves over subtropical South America,Mon. Weather Rev., 130, 16.Vera, C., G. Silvestri, V. Barros, and A. Carril (2004), Differences in El Nino response over the Southern Hemisphere, J. Clim., 17(9), 1741–1753,

doi:10.1175/1520-0442(2004)017<1741:dienro>2.0.co;2.Wang, H., and R. Fu (2002), Cross-equatorial flow and seasonal cycle of precipitation over South America, J. Clim., 15, 1591–1608.Wang, H., and R. Fu (2004), Influence of cross-Andes flow on the South American low-level jet, J. Clim., 17, 1247–1262.Yin, L., R. Fu, E. Shevliakova, and R. Dickinson (2012), How well can CMIP5 simulate precipitation and its controlling processes over tropical

South America?, Clim. Dyn., 1-17, doi:10.1007/s00382-012-1582-y.Yoon, J.-H., and N. Zeng (2010), An Atlantic influence on Amazon rainfall, Clim. Dyn., 34(2-3), 249–264, doi:10.1007/s00382-009-0551-6.Zeng, N., J.-H. Yoon, J. A. Marengo, A. Subramaniam, C. A. Nobre, A. Mariotti, and J. D. Neelin (2008), Causes and impacts of the 2005 Amazon

drought, Environ. Res. Lett., 3(1), 014002, doi:10.1088/1748-9326/3/1/014002.Zhang, Y., R. Fu, H. Yu, Y. Qian, R. Dickinson, M. A. F. Silva Dias, P. L. da Silva Dias, and K. Fernandes (2009), Impact of biomass burning aerosol

on the monsoon circulation transition over Amazonia, Geophys. Res. Lett., 36, L10814, doi:10.1029/2009GL037180.

Journal of Geophysical Research: Atmospheres 10.1002/2013JD021349

YIN ET AL. ©2014. American Geophysical Union. All Rights Reserved. 2328