Introduction
The Amazon region represents one of the main convective centers in the world
tropics, together with equatorial Africa and the Indian monsoon regions. In
this region, tropical convection is one of the key processes regulating the
climate system, and it plays an important role in the maintaining the water and
energy balance while also helping the development of the zonal and vertical
components of the circulation. The interannual rainfall variability in the
Amazon basin is linked to variations of sea surface temperatures (SSTs) in
both tropical Pacific and Atlantic oceans (Marengo, 1992; Ronchail et al.,
2005; Yoon and Zeng, 2010; Marengo and Espinoza, 2015, and references quoted
therein). Indeed, previous studies have documented that warm conditions in
the equatorial Pacific (e.g., El Niño events) produce a rainfall deficit
in Amazonia, which can cause extreme drought periods, as observed in
1926, 1983, 1997–1998 and 2010 (Williams et al., 2005; Marengo et al., 2008,
2011; Espinoza et al., 2011; Marengo and Espinoza, 2015). In the early
21st century, large-scale extreme seasonal events, such as extreme
droughts in 2005, 2010 and 2015 and floods (2009, 2012, 2014) have affected
the Amazon region. Rainfall anomalies were the consequence of circulation
changes forced by anomalous warming or cooling of the tropical Pacific and/or
tropical North or South Atlantic oceans, as documented by Marengo and
Espinoza, 2015, and references quoted therein).
In sum, while we had several droughts in Amazonia linked to El Niño, as
in 1925, 1983, 1987, 1998 and recently in 2015–2016, some other drought
events were reported in 1963 and 2005, which were not related to El Niño but to
a warmer tropical North Atlantic. When the tropical North Atlantic is warmer
than the tropical South Atlantic, the intertropical convergence zone is
displaced northward, leaving less rainfall in the region. This may happen with
an El Niño (1983, 1998) or without an El Niño (2005). Every drought
in Amazonia is different in terms of spatial coverage.
The occurrence of extreme weather and drought and flood events has changed
people's perception of climate extremes – as happened after those extensive
droughts and flooding. In Amazonia, the population perceives drought as
anomalously low river levels during the peak season May–July, more than as
low rainfall during the peak season in February–April. Drought is an impact
while deficient rainfall is the climatic forcing of this impact. However, in
terms of some ecological impacts or for agriculture, drought may be more
related to deficient rainfall during the peak of the rainy season. Drought
and floods and their impacts on natural and human systems in the region show
the vulnerability of residents and ecosystems to the occurrence of
hydrometeorological extremes in the region.
Rainfall in the Amazon basin is mainly supported by the moisture flux from
the equatorial Atlantic associated with the trade winds (Angelini et al.,
2011). However, the Amazon climate has to be seen as coupled interactive
atmosphere–ocean–land phenomena (Runyan et al., 2012), and current research
(Makaireva et al., 2013) suggests that land surface processes may play a
comparable part, equal in importance to that of the ocean, and have explored
the role of the forest in the recycling of water, suggesting an active role
of the vegetation in the regional water cycle. Previously, Salati and
Vose (1984) calculated this influence to be 50 % for inland Atlantic
moisture and 50 % for local recycling by evapotranspiration and
precipitation, using isotope techniques. An intensification of the
hydrological cycle in Amazonia over the last two decades has been identified
by Gloor et al. (2013), who attributed it to an increase in atmospheric water
vapor coming from a warmer tropical Atlantic. This is consistent with a
positive trend in precipitation in the northwestern Amazon since 1990, and is
also reflected water levels at Manaus and in the Amazon discharges at
Óbidos (Marengo and Espinoza, 2015).
Rainfall anomaly (mm month-1) for 3-month periods from
December 2014–May 2015 to September–November 2015. Data come from GPCC and
anomalies are relative to the 1951–2001 climatology. Black square indicates
the location of Manaus.
![]()
Variability of wet and dry seasons suggests that the onset and demise of the
wet and dry seasons and thus the length of the wet and dry season are
changing with time (Marengo et al., 2012; Marengo and Espinoza, 2015).
Observational studies in southern Amazonia suggest that the dry season has
increased in length by about 1 month since the 1970s (Marengo et al., 2011;
Fu et al., 2013). Furthermore, the length of the dry season also exhibits
interannual and decadal-scale variations linked either to natural climate
variability, or, as suggested by Wang et al. (2011), due to the results of
the influence of changes in land use in the region. While it is important to
know the future characteristics of the total rainy-season rainfall, it is
also important to highlight the urgency of improving our understanding and
capability to detect and predict the rainy season onset and demise, as well
as the variability of the wet and dry seasons, not just through model
experiments but also through observational analysis. The droughts of 2005 and
2010 and their impacts on humans and on the tropical forest have been
characterized by late onsets of the rainy season and/or longer dry seasons
(Marengo et al., 2011; Marengo and Espinoza, 2015). During the recent El
Niño in 2015–2016, rainfall over central-northern Amazonia has been
below normal (by approximately 200–300 mm for the wet season), and this
contributed to an extensive drought and subsequent problems in the hydrology
of the region, as well as an increased number of fires there (CPTEC –
www.cptec.inpe.br). The length of the dry season has a large temporal
and spatial variability, and is strongly associated with dry conditions that
may impact the occurrence of fires and their release of carbon and aerosols,
as well as affecting human systems (Aragão et al., 2014; Martin et al.,
2016).
Various studies have discussed observational and modeled aspects of the
onset and demise of the rainy season in Amazonia using a variety of climatic
indicators, such as rainfall, outgoing long-wave radiation or dynamic fields
(e.g., Kousky, 1988; Marengo et al., 2001, 2012; Fu et al., 1999; Liebmann
and Marengo, 2001; Gan et al., 2004; Wang and Fu, 2002;
Silva and Carvalho, 2007; Raia and Cavalcanti, 2008; Marengo and
Espinoza, 2015). However, modeling work still shows uncertainties in the
representation of the onset of the rainy season. This may be due to the poor
representation of clouds and land surface–atmosphere interactions or due to
the role of aerosols and other particles, which are still not well represented in
models.
Moisture transport across the Equator and its variations could influence
convection and thus the wet-season onset (Rao et al., 1996; Marengo et al.,
2001; Wang and Fu, 2002; Alves, 2016). Li and Fu (2006) showed that weak and
infrequent extratropical cold front penetrations during the transition season
also contribute to a delay of the wet-season onset. However, the complexity
of the relationship between ENSO, Atlantic SSTs and the onset of the wet
season over the southern and central Amazon remains unclear. Butt et
al. (2011) identify significant differences in the onset of the rainy season
in Rondonia between 1970 and 2000, due to land-use changes in the region.
However, the uncertainties of the attribution of these extremes and their
variations to natural and human influences are still great. This highlights
the urgency of understanding the underlying causes of the onset and demise of
the rainy season and our ability to predict them. Furthermore, evidence of
the possible role of human influences (deforestation, increase of greenhouse
gases and aerosol released due to biomass burning or urban pollution) on
rainfall and river variability has started to appear in the literature
recently (Alves, 2016; Spracklen and Garcia-Carreras,
2015; Magrin et al., 2014; Zhang et al., 2009).
Seasonal SST anomalies (∘C), SLP anomalies (hPa) and
anomalies of vertically integrated moisture transport (vectors) from the
surface to 500 hPa in South America from December 2014 to May 2015. SST and
circulation anomalies correspond to the 1961–2012 long-term mean. The bar at
the bottom of the panel shows the scale of the SST anomalies. The vector at
the bottom of the panel shows the scale of the moisture transport
(kg-1 m s-1). Black full lines show SLP anomalies (hPa).
![]()
Pressure–longitude section (80∘ W–100∘ E) of the
mean (a) and
anomalies (b) of divergence (contour interval is
1 × 10-6 s-1) and divergent circulation between
5∘ N and 5∘ S. The divergent circulation is represented by
vectors of combined vertical velocity and the divergent component of the
zonal wind. Red shading and solid contours denote divergence (a) and
anomalous divergence (b). Blue shading denotes
convergence (a) and anomalous convergence (b). Anomalies
are departures from the 1981–2010 long-term mean.
![]()
Increasing aerosol concentrations can have substantial impacts on spatial and
temporal rainfall patterns in the Amazon (e.g., Martins et al., 2009a;
Reutter et al., 2009; Pöhlker et al., 2016). The initial work by Andreae
et al. (2004) suggest that aerosols from biomass burning in Amazonia may
delay the onset of the rainy season in southern Amazonia, but not much is
known on the possible role of aerosols from urban areas in rainfall and
the water cycle in the Amazon region. Previous studies have found that cloud
microphysical properties, cloud cover, precipitation, lightning and regional
climate over the Amazon basin can be significantly affected by aerosol
particles (Li et al., 2006; Altaratz et al., 2010; Koren et al., 2012; Gonçalves et al.,
2015; Wang et al., 2016).
In this paper, we use data from the GoAmazon and CHUVA experiments during
2014–2015 as well as global reanalyses to investigate regional and
large-scale circulation and rainfall patterns during the onset and demise of
the rainy season. Emphasis is on the identification of large-scale patterns
leading to the onset and demise of the rainy season in the Manaus region in
central Brazilian Amazonia in both years. We have taken advantage of the high
resolution of surface meteorological data collected during these two field
experiments as part of GoAmazon in 2014–2015, and also the presence of El
Niño during summer of 2015, to investigate daily and diurnal rainfall
variability. We also investigated the large-scale and regional circulation
patterns linked to rainfall variability in those regions. With the GoAmazon
data we have investigated the onset and demise of the rainy season of
2014–2015, as well as convection and the planetary boundary layer (PBL)
heights in some sites near Manaus, where thermodynamic indices were
calculated to identify the transition regimes before and after both the onset and demise of the rainy season.
Methods
The data used for this study come from the GoAmazon Project (Martin et al.,
2016), designed to study some of the characteristics of the rainy season in
Amazonia, such as onset and demise of dry and wet seasons, and we have used
some of the CHUVA and GoAmazon 2014/5 rainfall and surface and upper-air meteorological data available from 2014 to 2015. The expression green
ocean (Go) was introduced by Williams et al. (2005) due to the similarities
in aerosol particle concentrations and cloud microphysics between the Amazon
basin and remote oceanic regions during clean periods of the wet season.
Observations and data from modeling work from the GoAmazon Experiment
(GoAmazon 2014/5 – http://campaign.arm.gov/goamazon2014/) were
collected in the central region of Amazonia near Manaus from 1 January 2014
through 31 December 2015. More details on the nature and objectives of GoAmazon 2014/5 can be found in Martin et al. (2016).
As described by Machado et al. (2014), the CHUVA project (CHUVA meaning
“rain” in Portuguese) is the acronym for the Cloud Processes of the Main
Precipitation Systems in Brazil: A Contribution to Cloud-Resolving Modelling
and to Global Precipitation Measurement (GPM). It began in 2010 and has
conducted five field campaigns with the last experiment having been held in Manaus
as part of the GoAmazon 2014/5 experiment. CHUVA's main scientific
motivation is to contribute to the understanding of cloud processes, which
represent one of the least-understood components of the climate system. Field
data from the CHUVA and GoAmazon 2014/5 campaigns are
used to improve our understanding of the dynamics of the onset of the rainy
season and the characteristics of the dry season during that year. The data
turned out to be useful since we now had the possibility of identifying the onset the
rainy season over central Amazonia using high-resolution meteorological data
collected near Manaus since the end of 2013. In addition, we analyze
ground-based remotely sensed data (ceilometer) to identify changes in the
PBL during the onset and demise of 2014–2015
rainy season. Wind, outgoing long-wave radiation (OLR) and sea surface
temperature data from the National Center for Environmental Prediction
(NCEP) Climate Prediction Center (CPC) are used to examine the influence of
atmospheric and oceanic conditions on the onset and demise of the rainy
seasons.
Rainfall data (at diurnal and daily levels) come from the Brazilian
Meteorological Service (INMET) station at Manaus (3.11∘ S, 59.95∘ W) and at Manacapuru from the University of the State of
Amazonas UEA (70 km upwind of Manaus: 03.05∘ S, 60.00∘ W). The pentad of the onset and end (or demise) of the rainy
season in Amazonia was calculated using the criterion of rainfall
accumulation data, defined by Liebmann and Marengo (2001) and adapted by
Bombardi and Carvalho (2009) using gridded rainfall data. This criterion was
applied to data from the Manaus and Manacapuru (nearby Manaus) rainfall
stations from the GoAmazon-CHUVA network during 2014–2015, by averaging all
available data for a given day from those two stations onto a 1.0∘
grid. The availability of data from a large number of stations in a grid box
as allows a more regional view compared to single station data.
Additional datasets for regional rainfall analyses during those two rainy
seasons come from the NCEP–NOAA CPC (www.ncep.noaa.gov) and from the
Global Precipitation Climatology Center (GPCC) gauge-based gridded
precipitation dataset, available for the global land surface only. The GPCC datasets are available at spatial resolutions of
1.0∘ latitude × 1.0∘ longitude as mean monthly precipitation
totals and anomalies relative to the 1951–2000 mean.
Once the pentad of the onset was identified for Manaus and Manacapuru,
various analyses are performed for the periods before and after the onset, in
order to identify and understand possible shifts in atmospheric circulation
and energy fluxes and characteristics of the PBL and thermodynamic indices
convective available potential energy (CAPE) and convective inhibition energy
(CIN) that would favor the establishment of the onset of the rainy season of
2014–2015. CAPE and CIN values were computed using the original variables
for the 500 m thick mixing-layer parcel. The height of the PBL was derived
from a ceilometer installed at the T3 site in Manacapuru and its hourly
values were computed for pentads and post-onset and post-demise of the rainy
season in 2014–2015. The latent (LE) and sensible (H) turbulent heat fluxes
as well as the Bowen ratio were derived from these fluxes at the T3 site and the
EMBRAPA Flux for 2014–2015 in order to quantify their values before, during
and after the onset and end (or demise) of the rainy season, so we could
investigate the energy partition and convective processes that accompany the
evolution of the rainy season during the days of the GoAmazon campaign. More
details about the instrumentation used can be found in Machado et al. (2014),
Martin et al. (2016) and Wang et al. (2016).
Results
Characteristics of the 2014–2015 rainy season<?xmltex \hack{\break}?> in Amazonia
The mean climatic features of the Manaus region, where the peak of the rainy season occurs around March–May, are described elsewhere
(Greco et al., 1990; Cohen et al., 1995; Machado et al., 2014; Martin et al.,
2016). The GPCC
rainfall (Fig. 1a–d) shows rainfall anomalies from
December 2014–February 2015 (representing the wet season) to
September–November 2015. Over central and eastern Amazonia in the wet
season, rainfall was about 80–90 mm month-1 below normal, while over western Amazonia rainfall was
about 50–90 mm month-1 above normal. For the September–October–November period the rainfall
was well below normal (by around 90 mm month-1), in almost all of
Amazonia. This is a signal of the impacts of El Niño 2015–2016, which was
under development in the tropical Pacific since the middle of 2015 (Fig. 2).
Warm surface waters (1.5–2.5 ∘C) were detected along the equatorial
Pacific in the period from March to May 2015. This 2015–2016 drought caused
the longest fire season of the 21st century, with 5 months exceeding
10 000 fire detections and the largest number of active fire occurrences per
square kilometer of deforested land (Aragão et al., 2014). This combination
of a longer dry season, more frequent extreme droughts and an increased risk
of fire could play a critical role in a future Amazon rainforest dieback in
spite of the increased resilience of tropical forests in an elevated
atmospheric CO2 environment (Huntingford et al., 2013).
As to regional-scale circulation features, the December 2014 to January–February 2015 season did not
show signals of El Niño in the tropical Pacific, but the warm surface
waters were already present during March– April–May (MAM) 2015 (Fig. 2). This warming increased
continuously until March 2016, indicating the intensification of El Niño
in 2015–2016, with warm surface water (3–4 ∘C above normal) over
the equatorial eastern Pacific by the austral summer of 2016
(www.cptec.inpe.br). The low-level circulation anomaly vectors show
small southerly wind anomalies, suggesting weakened northerly flow over the
tropical North Atlantic and Amazon sectors (Fig. 2). Along the Equator,
Fig. 2 shows that there are westerly anomalies, but these are away from the
coast (looking at December–January–February, DJF). Along the Atlantic coast and north of the Equator,
however, the anomalies are near zero. There are huge positive transport
anomalies from the Equator into the southern Amazon, which are consistent
with above-normal precipitation to the west (south of the Equator), since
there appears to be anomalous convergence of moisture there (Fig. 2a). An
analysis of the near-surface and upper-air circulation discussed previously
can provide a better idea of the regional east–west circulation of the region
during December 2014 and January 2015. The east–west vertical cross-section
along the Equator (5∘ N–5∘ S) in Fig. 3 shows upward-motion anomalies over western Amazonia during both summer months while
reduced convection and downward motion with subsidence is found over central
Amazonia, eastern Amazonia and northeastern Brazil. The latter region has been
experiencing a record drought since 2011 (Marengo et al., 2016). These
circulation anomalies are consistent with negative rainfall anomalies over
central Amazonia near the Manaus region. Therefore, interannual variations of
the wet-season onset in the Amazon appear to be influenced by changes in
large-scale and regional circulation over the tropical Atlantic and Pacific
sectors.
Pentad rainfall (mm pentad-1) at Manaus (INMET site) from
January 2014 to December 2015. The red bar shows the climatological occurrence of
the onset and end of the rainy season, while the blue bar indicates the onset and
end for 2014–2015. The grey line is the climatological annual cycle of pentad
mean precipitation.
![]()
Daily rainfall variation and characterization of the onset and
demise of the rainy season during 2014–2015
In order to verify the variability in rainy-season characteristics, the onset
and demise of the rainy season over Manaus has been defined following
Liebmann and Marengo (2001) using daily precipitation data from Manaus
station for the period 1961–2015. Using this methodology, the climatological
onset is detected around pentad 64 (12–16 November) and the demise of the rainy season around pentad 29
(21–25 May). Figure 4 plots the pentad rainfall recorded during 2014–2015
for Manaus station. It also plots the annual precipitation climatology of the
pentad cycle and the onset and demise of the rainy season. The data showed
anomalously late onset (around the pentad 69: 7–11 December) in 2014–2015.
The demise of the rainy season occurs around pentad 29 (21–25 May), which
agrees with the climatological date, giving a shorter-than-normal rainy season. In
addition, the data showed that the Manaus station experienced reduced
rainfall totals compared with climatology, with some dry spells.
Hovmoller diagram of 30–60-day filtered OLR anomalies (W m-2)
along the Equator (5∘ N–5∘ S) during the period of June
2014 to May 2015.
![]()
Daily variation of relative humidity (%), 850 hPa air
temperature (∘C) and Bowen ratio at the T3 Manacapuru site, from
January 2014 to July 2015. The blue bar indicates the onset and end of the rainy
season for 2014–2015. No Bowen ratio data were available during some days
between January 2014 and July 2015.
![]()
As seen in the previous section, despite the oceanic and atmospheric
conditions in the equatorial Pacific (El Niño 1 + 2 and 3) which show
ENSO-neutral conditions during the period before the onset, it is noted that
the patterns of regional precipitation distribution over central and eastern
Amazonia were consistent with a distinct phenomenon that caused changes in
weather conditions in the region, for example, the Madden–Julian Oscillation
(MJO; Madden and Julian, 1994; Liebmann et al., 1999; De Souza and Ambrizzi,
2006; Alvarez et al., 2015). This is shown in Fig. 5 in the time diagram of
OLR anomalies between 5∘ N and 5∘ S over the globe in 2014 and
2015.
It is observed that from July through late October 2014 the intraseasonal
signal became less coherent, with a weaker anomaly field, and is inconsistent
with a canonical El Niño signal. The pattern became more organized during
late November as the MJO strengthened, as indicated by eastward propagation of
alternating anomalies into January 2015. At this time, the MJO may have
contributed to enhanced rainfall and the onset of the rainy season for portions
of Amazonia, as indicated by negative OLR anomalies (blue shading) favoring
conditions for precipitation around 60∘ W.
This fact was reflected in the temporal distribution of rainfall (Fig. 4),
which depicts the regional frequency of precipitation in this period. In
summary, the phases of MJO-associated OLR anomalies were evident throughout
the equatorial region, and in particular over the central and eastern Amazon.
This suggests that the negative MJO phase in mid-January may have contributed
to favouring conditions to enable the convection and the onset of the rainy
season from pentad 6. This is consistent with other atmospheric mechanisms on
a local scale, for example, the near-surface local circulation and
thermodynamic patterns.
Daily variation of CAPE and CIN (J kg-1) from 1 January to
25 February 2015 (around the onset of the rainy season) and from 11 April to
30 June (around the demise of the rainy season) at Manaus. Pentads of the
onset and demise are identified with blue vertical bars.
![]()
Broadly speaking, the influence of the MJO on precipitation over the tropics
occurs by eastward propagation of Rossby wave trains from the tropical
Pacific Ocean (Muza et al., 2009). Previous observational and modeling
studies generally indicated that the MJO and ENSO have a decadal variation
and seasonal dependence (Hendon et al., 2007); however,
these have not been well identified due to their nonlinear nature. These
studies also show significantly lagged correlations between MJO and ENSO
indices. Despite this, Shimizu et al. (2016) recently examined the regional
relationship between ENSO and MJO phases in climatological patterns of
precipitation over South America. The results indicated that combined
responses showed that precipitation is strongly influenced by the MJO phases
rather than by ENSO conditions, especially during the austral summer. Thus,
our results corroborate those of Shimizu et al. (2016), who observed the highest
percentages of days with active MJO and increased rainfall occurring during
El Niño and neutral year.
Figure 6 shows the daily variation of air temperature at 850 hPa, relative
humidity and Bowen ratio at the T3 Manacapuru site. Wet conditions were
consistent with higher relative humidity and lower temperatures, while dry
spells occur with lower relative humidity and higher air temperatures. Before
the onset we noticed a reduction in temperature and humidity, while it is
hard to see any tendency of the Bowen ratio before the onset or after the
demise of the rainy season. The latent heat fluxes over the land surface are
important sources of atmospheric humidity during the initial stages of the
transition season between dry and wet periods (Fu and Li, 2004). Together
with changes in the onset of the rainy season, changes of dry season length
may be key in favouring the present risk of fire. A high land-surface Bowen
ratio during the preceding dry season would delay the subsequent wet-season onset in the southern Amazon (Fu and Li, 2004), and this may have been the
case in the onset rainfall in 2014–2015.
Diurnal cycle of the height of boundary (m) layer (average from
1 January up to 30 June 2015) (a), from January 2015 (b)
and June 2015 (c).
![]()
Thermodynamic indicators and PBL behavior during the onset and
demise of the rainy season<?xmltex \hack{\break}?> in 2014–2015
Using the T3 radiosonde data, 1-day averages of CAPE and CIN (Fig. 7) were
calculated for each profile and these in turn were averaged over pentads. One
pentad was selected to encompass the onset of the rainy season and another
its demise. The figure shows that CAPE and CIN are very noisy; hence, a
two-sample t test for daily average CAPE and CIN values during the rainy season
(26 January 2015–10 May 2015) and before and after the rainy season
(1 November 2014–25 January 2015 and 11 May–1 December 2015) was
performed. Results suggest that both CAPE and CIN values change between the two
periods (during the rainy season and before and after the rainy season),
indicating a difference significant t the 5 % level in CAPE and CIN
between the days with and without deep convection.
The diurnal cycle of the heights of the PBL was computed with the ceilometer
installed at the T3 site. Figure 8a shows the composite PBL diurnal cycle for
the wet season 2014–2015. The PBL height remained stationary at around
300–400 m during nighttime (nocturnal boundary layer), then increased
during daytime (convective boundary layer), and reached its maximum
(1100–1200 m) in the early afternoon (14:00 LT). This pattern is
consistent with the previous values obtained for Amazonia for the wet season
(Fisch et al., 2004). There is no signal of the anomalous wet season in the
PBL heights. Individual diurnal cycles for other months (Fig. 8b, c) also did
not present a significant difference between the onset and demise of the wet
season of 2015.
A moistening of the planetary boundary layer and a lowering of the
temperature at its top may reduce CIN and control the conditioning of the
large-scale thermodynamics prior to onset (Fu et al., 1999). In addition, Li
and Fu (2006) found that the main increase in CAPE and reduction in CIN occur
prior to the rainy season onset, although in the tropical atmosphere a
decrease CAPE often exists in the absence of deep convection (Williams and
Renno, 1993).