Interannual Variability of Precipitation for the Centro de Lançamento de Alcântara in ENSO-Neutral Years

AbstrAct: The interannual variability of precipitation in El Niño/Southern Oscillation-neutral years was studied for the Centro de Lançamento de Alcântara region. Monthly precipitation, sea surface temperature, wind at 925 hPa and outgoing longwave radiation data from various gridded datasets for the 1951–2010 period (60 years) were used. The data grouping was based on terciles. For the Centro de Lançamento de Alcântara in El Niño/Southern Oscillationneutral years, March is the month of the rainy quarter (March to May) when the interhemispheric gradient of the sea surface temperature anomalies over the Atlantic (GRAD) and the atmospheric circulation at 925 hPa over the Centro de Lançamento de Alcântara were able to best explain the variability of precipitation. In this month, the wind direction at 925 hPa was the factor that explained the highest fraction of precipitation variance (40%), followed by GRAD (30%) and the wind magnitude (20%). For the Centro de Lançamento de Alcântara, in general, above-average precipitation was related to weak north-northeasterly low-level winds and southward GRAD, while below-average precipitation was related to strong east-northeasterly low-level winds and northward GRAD. These features were related to an eastward expansion of the Amazon convection towards the northern Northeast Brazil and might be related to a slight southward displacement of the Intertropical Convergence Zone in above-average precipitation years.


INTRODUCTION
The interannual variability of precipitation in the northern Northeast Brazil (NNEB) has been studied extensively over the last decades.Since sea surface temperature anomalies (SSTAs) play a key role in the climatic predictability of tropical regions (Shukla 1998), numerous papers have focused on the relation between the precipitation variability in the NNEB and the SSTAs in the tropical Pacific and Atlantic oceans (review in Kayano and Andreoli 2009).For the tropical Pacific, the influence would occur through the El Niño/Southern Oscillation (ENSO) phenomenon; for the tropical Atlantic, through the interhemispheric SSTA gradient (GRAD).
The main features of ENSO are SSTAs in the central and eastern equatorial Pacific linked to the Southern Oscillation, which is an east-west seesaw pattern in surface pressure over the tropical Pacific (Trenberth and Hoar 1996;Kane 1997).In the ENSO-warm (cold) phase, also called El Niño (La Niña), positive (negative) SSTAs are found in the central and eastern equatorial Pacific.The general relation between the ENSO phases and the precipitation variability in the NNEB -warm (cold) phase associated to below-(above-) average precipitation -is well known (e.g.Ropelewski and Halpert 1987).
In the ENSO-neutral phase, i.e. without the occurrence of El Niño or La Niña, the ascending (descending) branch of the Walker cell is located over the warmer (colder) surface waters in western (eastern) tropical Pacific.These two branches are connected by easterly trade winds over the tropical Pacific.Cold waters from the ocean upwelling along the west coast of South America spread westward towards central Pacific.In the Marques RFC, Oyama MD ENSO-warm (cold) phase, there is weakening (intensification) of the trade winds and the coastal upwelling, leading to positive (negative) SSTAs in eastern Pacific and to an anomalous ascending (descending) motion over this region, which is partially balanced by an anomalous descending (ascending) motion over the NNEB.Since anomalous descending (ascending) motion means less (more) favorable conditions to deep convection, precipitation over the NNEB decreases (increases).These processes explain how the ENSO phases could influence the precipitation in the NNEB.However, in general, this influence is clearer only for the more intense El Niño and La Niña events (Kane 1997;Andreoli et al. 2004;Lucena et al. 2011;Marques and Fortes 2012).
For the tropical Atlantic, numerical simulations (Moura and Shukla 1981) and observational analysis (Hastenrath and Heller 1977;Nobre and Shukla 1996) indicated that precipitation anomalies in the NNEB would be associated with a meridional dipole of SSTAs, called the Atlantic dipole.It was regarded as an ocean-atmosphere coupled mode of variability on decadal timescales (Chang et al. 1997), and it would affect the precipitation in NNEB through anomalies in the position and intensity of the Intertropical Convergence Zone (ITCZ).The concept of Atlantic dipole evolved to the GRAD (Enfield et al. 1999;Andreoli et al. 2004).During the NNEB rainy season, a southward GRAD -due to positive SSTAs in the tropical South Atlantic and/or negative in the tropical North Atlantic -would be related to lower (higher) surface pressure in the tropical South Atlantic (North Atlantic), weaker (stronger) southeast (northeast) trade winds, southward displacement of the trade winds confluence axis and above average precipitation in the NNEB.For a northward GRAD, the opposite features would occur, ending up with below-average precipitation in the NNEB.
Due to the importance of both tropical oceans, numerous papers have focused on the interaction between ENSO and the variability in tropical Atlantic (Saravanan and Chang 2000;Andreoli and Kayano 2007;Münnich and Neelin 2005;Giannini et al. 2004).This interaction does exist, but there is also considerable independence between them, which justifies considering the ENSO phases (warm/cold) and the GRAD direction (northward/southward) as distinct variables related to the precipitation anomalies in the NNEB.Which one is more important?Although higher correlation is found between the precipitation anomalies and the GRAD direction (Moura et al. 2009), Andreoli and Kayano (2007) point out that the ENSO phase and the GRAD direction act to reinforce or inhibit the precipitation anomalies (e.g. for El Niño conditions, northward GRAD would intensify the negative precipitation anomalies in the NNEB rainy season).
In the present study, we focus on the variability of precipitation for a specific region in the NNEB: the Centro de Lançamento de Alcântara (CLA) region, located at the northern coast of Brazil.From the CLA, the space vehicles developed at the Instituto de Aeronáutica e Espaço, such as sounding rockets (VS-30, VS-40, VSB-30, among others) and the satellite launch vehicle (VLS), are launched.Since these space vehicles are not designed to withstand adverse meteorological conditions, such as rain and/or lightning occurrence, the knowledge of the interannual variability of precipitation is important for the planning of rocket launching missions (Marques and Fisch 2005).
In the CLA, the rainy season extends from February to mid-June, when the ITCZ reaches its southernmost position and the rainiest months are March and April, when precipitation is on average higher than 300 mm/month (Marques and Baungartner 2008).The rainy season features -onset and demise day, as well as total precipitation amount -show a pronounced interannual variability.Previous studies suggest that this variability would be related to oceanic factors, such as the ENSO phases and the GRAD direction, and to dynamical factors, such as the meridional position of the Atlantic ITCZ and the low-level circulation over the CLA.For instance, Pinheiro and Oyama (2013) showed that anomalies in the rainy season onset, for specific years, would be related to the ENSO phase, GRAD direction and meridional position of the Atlantic ITCZ.Marques and Fortes (2012) showed that the relation between ENSO-warm (cold) phase and below-(above-) average precipitation in the CLA would hold only in intense El Niño or La Niña events occurring in the rainiest months.
In ENSO-neutral years, preliminary studies indicate that the variability of precipitation in the CLA, comparing contrasting years (few above-and below-average years), would depend on the combined action of GRAD direction and meridional position of the Atlantic ITCZ (Marques et al. 2013;Pereira and Marques 2014).Marques and Correa (2014) showed that, independently from the GRAD direction, the low-level circulation over the CLA is related to the precipitation anomalies: weak (strong) northeasterly (easterly) winds would favor (inhibit) precipitation occurrence.
This research aims to study the factors related to the variability of precipitation in the CLA for the ENSO-neutral phase.Two factors are addressed here: the GRAD direction (oceanic factor) and the low-level circulation over the CLA (dynamical factor).The variability of the meridional position of the Atlantic ITCZ is not considered, because it is largely explained by the variability of the GRAD direction (Carvalho 2011).Differently from the previous studies (Marques et al. 2013;Pereira and Marques 2014;Marques and Correa 2014), the present study uses longer datasets as well as standardizes the method to classify the anomalies and obtain more comprehensive and detailed results.

DATA AND METHODOLOGY
dAtA Monthly precipitation, sea surface temperature, wind at 925 hPa and outgoing longwave radiation (OLR) data from various gridded datasets (Table 1) were used.The datasets were provided by Earth System Research Laboratory/National Oceanic and Atmospheric Administration (ESRL/NOAA): • For the wind data, the 925 hPa level is used because it is a mandatory level representative of the lowest atmospheric layer.

•
For the CLA, time series of precipitation and wind at 925 hPa are obtained from the gridded data.Precipitation values are computed as the area average over the four grid points that surround the CLA (Pinheiro and Oyama 2013), whereas wind values refer to the (45°W; 2.5°S) grid-point values.The computing procedures are not the same because the precipitation and wind data are given on different grids (Table 1), and the idea is to locate the CLA approximately at the center of the area for which the data is representative.

•
The SSTAs are obtained considering 1951-2010 as the base period.The Niño 3 index is computed as the average SSTA over the 150°W-90°W; 6°S-6°N area.GRAD is defined as TNA minus TSA, where TNA (tropical North Atlantic) is the average SSTA over the 60°W-20°W; Equator-20°N area and TSA (tropical South Atlantic), over the 30°W-10°E; 20°S-Equator area.Northward (southward) GRAD refers to positive (negative) difference.The definition of the indices follows Andreoli and Kayano (2007).The precipitation data from the Global Precipitation Climatology Centre (GPCC) have been used in climate studies (Schneider et al. 2014) and also in studies focusing on the South America precipitation variability (Kayano et al. 2011).For the CLA, GPCC data are compared to the Climate Prediction Center (CPC)/NOAA data, which were used by Pinheiro and Oyama (2013) to obtain the rainy season features for the CLA.For the total precipitation in the rainy quarter (March, April and May), there is close agreement between the time series in the 1979-2010 period, which is the common period for both datasets (Fig. 1).The differences are small (< 10%), except in few specific years (e.g. 1991 and 2007), and the linear correlation between the two datasets is high (Pearson correlation coefficient @ 0.97).Therefore, the use of GPCC data for the CLA is appropriate.

MethodoloGy
The data grouping was based on terciles.For a given month, the terciles for each variable -precipitation and wind (direction and magnitude at 925 hPa) for the CLA, as well as oceanic indices (Niño 3 and GRAD) -are computed.For instance, for March, considering the data in ascending order, the first (P 33 ) and second terciles of precipitation (P 67 ) are 320 and 402 mm, respectively (Table 2).Then, the 60-years data  are grouped in 3 categories: • 20 years when precipitation is lower than P 33 (belowaverage years).
• 20 years when precipitation is higher than P 33 and lower than P 67 (normal years).
• 20 years when precipitation is higher than P 67 (aboveaverage years).Table 2 shows the terciles and categories for all the variables.From the visual inspection of the results (Table 3), it is possible to find out the categories related to below-and above-average precipitation categories (pink and green cells in Table 2, respectively).
The procedure of grouping precipitation data using terciles was done by Kayano and Andreoli (2006).Here, the procedure is extended to the all variables to standardize the method to classify the anomalies.For the oceanic indices, the tercile-based procedure may lead to differences, relative to the literature, in the classification of the categories (ENSO phase and/or GRAD direction) for a given month:

•
For the Niño 3 index, the classification of the ENSO phases in March from the tercile-based procedure follows a simple rule: warm phase when the index is above +0.10°C; cold phase, below -0.25 °C (Table 2).This procedure differs from the usual methods to identify the ENSO phases.For instance, Kayano and Andreoli (2006) used the Niño 3 index to identify the ENSO phases following the criterion proposed by Trenberth (1997): warm phase when the detrended and smoothed (5-month running mean) values remain, for at least 6 consecutive months, above +0.5 °C; cold phase, below -0.5 °C.The tercile-based procedure has the advantage of not including the short warm/cold periods (< 6 consecutive months) in the neutral phase category, but has the limitation of not distinguishing between the intraseasonal and interannual variabilities.

•
For GRAD, the thresholds adopted here for March (-0.21 °C and +0.25 °C; Table 2) are close to those used by Andreoli and Kayano (2007) for the December to February quarter (± 0.20 °C).

•
The differences may also result from the use of distinct base periods.Composites of SSTA, circulation at 925 hPa and OLR for below-and above-average precipitation years in the ENSOneutral phase (Table 3) are obtained.For a given variable, the statistical significance of the composite differences is evaluated by applying the Student's t-test.The confidence level is 90%.For the wind, the statistical significance is evaluated for the components (zonal and meridional wind) separately.
Given two variables, y and x, the coefficient of determination (R 2 ) measures the maximum fraction of the variance of y which could be explained by a linear model in x (y = ax + b, a and b constants) (Costa Neto 1977).Here, R 2 is used to single out the factors (among wind direction, wind magnitude and GRAD) that would be more important to explain the variability of precipitation in the CLA for the ENSO-neutral phase.

RESULTS AND DISCUSSION
Before focusing on the ENSO-neutral phase, the general relation between the ENSO phases and the variability of  P 33 e P 67 are the first and second terciles, respectively, and their values are found in parentheses in the second and fourth columns.Green/pink color indicates the categories related to above -/below -average precipitation in March.
precipitation for the CLA is examined.Figure 2 shows the probability of above or below-average precipitation in the rainy quarter (March to May) for the three ENSO phases.

•
The probability of above-average precipitation is similar for La Niña and ENSO-neutral years (35 -50%), and clearly lower (15%) for El Niño years (Fig. 2b).Therefore, in the rainy quarter, low probability -less than 20% -of below-(above-) average precipitation is found for La Niña (El Niño) years.For the ENSO-neutral years, there is no clear preference for one of the categories (below-or aboveaverage precipitation).
The average and standard deviation of the monthly precipitation are not substantially affected when, instead of using all years to compute the statistics, only the ENSO-neutral years are used (Fig. 3).This is an intriguing result, because by removing one factor of interannual variability (the ENSO phase), a lower variability (i.e.lower standard deviation) would be expected.So, what are the factors that induce (or are related to) the high variability of precipitation for the ENSO-neutral phase?Cell color refers to the categories shown in Table 2.As mentioned earlier, the influence of three factors -GRAD, low-level wind magnitude and direction -is analyzed.

Jan Feb Mar Apr May
The fraction of precipitation variance explained by each factor (measured by the coefficient of determination, R 2 ) is shown in Fig. 4. For a given factor, there is considerable variation of the (a) (b) fraction over the months, but two general features are noticeable: for all factors, the highest fraction occurs in March, followed by a marked decrease in April.Therefore, in the rainy quarter, March is the month when all factors are able to best explain the variability of precipitation.Moreover, another interesting and novel result is that, for March, the low-level wind direction is the factor that explains the highest fraction of precipitation variance (R 2 ~ 40%), followed by GRAD (30%) and the wind magnitude (20%).
The values of precipitation, GRAD, wind magnitude and direction at 925 hPa in March for ENSO-neutral years are shown in Table 3.In general, above-average precipitation (green shaded cells) is related to weak north-northeasterly (NNE) low-level winds and southward (negative) GRAD; below-average precipitation (pink shaded cells), to strong east-northeasterly (ENE) lowlevel winds and northward (positive) GRAD.Exceptions to this general relationship do occur.For instance, in 1972For instance, in (1978)), below-(above-) average precipitation occurs, but GRAD is southward (northward), and the precipitation anomaly sign is consistent only with the wind magnitude and direction anomalies.Next, the composites (average fields) of SSTA, circulation at 925 hPa and OLR in March for below-and above-average precipitation years in the ENSO-neutral phase are shown to provide a broader view of the general relationship obtained for the CLA.
The composites of SSTA are shown in Fig. 5.A clear statistically significant dipole pattern between TNA and TSA is found in both composites, showing the usefulness of the meridional dipole concept (Moura and Shukla 1981) for the purposes of the present study.When precipitation is above (below) the average, TNA is negative (positive) and TSA, positive (negative); therefore, GRAD is southward (northward).This result is consistent with the literature (Giannini et al. 2004;Kayano and Andreoli 2006;Andreoli and Kayano 2007).In both composites, however, substantial negative SSTAs are found in the Niño 1+2 region, close to the west coast of equatorial South America.It means that the ENSO-neutral phase, defined here considering the Niño 3 index, could be including numerous cases of La Niña onset, and it could also explain the smaller differences between the ENSO-neutral and cold phases in Fig. 2 (particularly for the probability of above-average precipitation).
The composites of atmospheric circulation at 925 hPa are shown in Fig. 6.The large-scale circulation shows an easterly ITCZ-related confluence zone located between the north and south subtropical highs that turns counter-clockwise and becomes northeasterly when crossing the NNEB coast.Over the NNEB, for the below-average years, there is smooth confluence of the streamlines towards the central South America (Fig. 6a), and ENE winds are found over the CLA; for the above-average years, the counter-clockwise turning of the atmospheric fl ow is increased, the streamlines converge (Fig. 6b) and NNE winds are found over the CLA.These circulation diff erences are statistically signifi cant (for both wind components, zonal and meridional) over the NNEB coast.Th e additional counter-clockwise turning of the wind direction over the CLA for above-average precipitation years is related to the pronounced convergence anomaly that comes out over the NNEB (Fig. 6c).Th is convergence anomaly extends zonally towards the Atlantic, and divergence anomaly is found immediately to the north of it; this dipole pattern Figure 6.Composites of circulation at 925 hPa for (a) below-and (b) above-average precipitation years, as well as the difference between them (c), in the ENSO-neutral phase.Shading refers to atmospheric convergence (10 -6 /s).The zonal and meridional wind differences are statistically signifi cant at the 90% confi dence level within the areas enclosed by the thick blue and red contours, respectively (c).The CLA location is indicated by the small box at the northern coast of Brazil.
suggests that the ITCZ is slightly displaced to the south in above-average precipitation years, which is consistent with the more intense north Atlantic subtropical high and also to the dipole pattern related to the southward GRAD (negative SSTAs over TNA, and positive over TSA; Fig. 5b).
Th e composites of OLR are shown in Fig. 7. Deep convection is found over South America and the ITCZ, while dry zones are found over the subtropical highs.Th e convergence anomaly over the NNEB for the above-average years is related to the eastward expansion of the Amazon convection towards the NNEB (Figs. 7a  and 7b), leading to a large statistically signifi cant negative OLR anomaly over the NNEB (Fig. 7c).Th e slight southward displacement of the ITCZ, suggested by the convergence/ divergence anomalies at 925 hPa (Fig. 6c) and the southward  GRAD (Fig. 5b), is also consistent with the symmetric dipole pattern of the OLR anomalies over the equatorial Atlantic (although this pattern lacks statistical significance).The OLR difference field also shows a statistically significant reduction in deep convection over the northwestern and southern South America.
Regarding the predictability, considering GRAD and wind direction at 925 hPa in February as predictors of the precipitation in March, the use of linear multiple regression leads to R 2 ~ 46% (adjusted R 2 ~ 40%; Fig. 8).This is a preliminary but promising result.Previously, it was showed that R 2 for each factor separately at lag-0 ranged from 20 to 40% (Fig. 4).At lag-1, a decrease in R 2 would be expected; however, considering the GRAD and wind direction jointly, about half of the precipitation variance would continue to be explained.Therefore, for the ENSO-neutral phase, precipitation in March could be reasonably predicted one month ahead from the GRAD and low-level wind direction in February.only these neutral years, the interannual variability is pronounced (and close to the total variability).In the rainy quarter (March to May), March is the month when GRAD direction and low-level circulation are able to best explain the variability of precipitation.In this month, the wind direction at 925 hPa is the factor that explains the highest fraction of precipitation variance (40%), followed by GRAD (30%) and the wind magnitude (20%).For the CLA, in general, above average precipitation is related to weak NNE low-level winds and southward (negative) GRAD; belowaverage precipitation, to strong ENE low-level winds and northward (positive) GRAD.
The circulation at 925 hPa in March shows a northeasterly atmospheric flow crossing the NNEB coast from the equatorial Atlantic.For the below-average years, there is smooth confluence of the streamlines from the NNEB towards the central South America, and ENE winds are found over the CLA.For the above-average years, the streamlines turn counter-clockwise and converge over the NNEB, leading to NNE winds over the CLA.These circulation differences are statistically significant over the NNEB coast.The convergence anomaly over the NNEB is related to the eastward expansion of the Amazon convection, and this expansion explains the occurrence of above-average precipitation over the CLA.This is also explained, on the other hand, by the formation of a statistically significant meridional dipole pattern of negative (positive) SSTAs in TNA (TSA), which might be related to a slight southward displacement of the ITCZ.This integrated picture expands and synthesizes the results of previous studies (Marques et al. 2013;Pereira and Marques 2014;Marques and Correa 2014) and shows how the dynamical and oceanic factors shape the variability of precipitation in ENSO-neutral years.Moreover, for these neutral years, a preliminary but promising result showed that precipitation in March could be reasonably predicted one month ahead from the GRAD and low-level wind direction in February.

CONCLUSION
The interannual variability of precipitation in ENSOneutral years was studied for the CLA region.Even considering

Figure 1 .
Figure 1.Accumulated precipitation over the CLA for the rainy quarter (March to May) from 1979 to 2010.Values obtained from the GPCC dataset (black line); and from the CPC/NOAA dataset (green line).

Figure 2 .
Figure 2. Probability of (a) below-or (b) above-average precipitation from January to June for the three ENSO phases.

Figure 4 .
Figure 4. Fraction of precipitation variance explained by each factor: direction (GRAD), wind magnitude (MAG) and direction (DIR) at 925 hPa -from January to June in the ENSO-neutral phase.

Figure 3 .
Figure 3. Average (AVG, solid line) and standard deviation (STD, dashed line) of the monthly precipitation (mm) from January to June.For a given month, the AVG and STD were computed using all the 60 values (from 1951 to 2010; green) or only the 20 values for the ENSO-neutral phase (black).

Figure 5 .
Figure 5. Composites of SSTA (°C) for (a) below-and (b) above-average precipitation years in the ENSO-neutral phase.The SSTA differences between the two panels are statistically significant at the 90% confidence level within the areas enclosed by the thick blue contours.The CLA location is indicated by the small box at the northern coast of Brazil.
Precipitation for the Centro de Lançamento de Alcântara in ENSO-Neutral Years

Figure 7 .
Figure 7. Composites of OLR (W/m 2 ) for (a) below-and (b) above-average precipitation years, as well as the difference between them (c), in the ENSO-neutral phase.The OLR differences are statistically signifi cant at the 90% confi dence level within the areas enclosed by the thick blue contours.The CLA location is indicated by the small box at the northern coast of Brazil.

Figure 8 .
Figure 8.Comparison between the estimated and observed precipitation in March for the ENSO-neutral phase.

Table 1 .
Brief description of the datasets used in this study.

Table 2 .
Categories for each variable.

Table 3 .
Monthly values for March in the ENSO-neutral phase.