REGIONALIZATION OF PRECIPITATION ANOMALIES ASSOCIATED WITH EL NIÑO EVENTS IN SOUTHERN SOUTH AMERICA ABSTRACT Alice M. Grimm (1); Moira E. Doyle (2) (1) Department of Physics Federal University of Paraná Caixa Postal 19081 CEP 81531-990 Curitiba Paraná Brasil Phone: 55 41 361-3097 E-mail: grimm@fisica.ufpr.br (2) Department of Atmospheric Sciences - University of Buenos Aires Pabellon 2 Ciudad Universitaria 1428 Nuñez Buenos Aires - Argentina A comprehensive view is given of the precipitation anomalies associated with the various stages of El Niño events all over southern South America. This view comprises the delineation of coherent regions with respect to these anomalies, the identification of the seasons of maximum anomalies and the assessment of their magnitude and consistency. Key-words: Weather and Climate Prediction, El Niño, coherent rainfall regions. RESUMO Uma visão abrangente das anomalias de precipitação associadas aos vários estágios dos eventos El Niño sobre todo o sul da América do Sul é apresentada. Esta visão compreende a delimitação das regiões coerentes em relação a estas anomalias, a identificação dos períodos de máximas anomalias e a avaliação de sua magnitude e consistência. 1. Introduction Several areas in Southern South America (SSA) (which comprises Southern Brazil, Argentina, Chile, Uruguay and Paraguay) have been reported as presenting strong interannual precipitation variability associated with extreme events of the Southern Oscillation (e. g., Ropelewski and Halpert 1987, hereafter referred to as RH87; Aceituno 1988; Pisciottano et al. 1994; Grimm et al. 1998) These previous studies have given valuable insights into several aspects of the El Niño (EN) impact on SSA, as well as detailed descriptions of this impact on certain regions. Unfortunately, shortage of data or data gaps in certain regions and differences between the methodologies hamper the formation of a more general panel of this impact on SSA. In this study we propose, on the basis of an ample data set from 134 stations, to give an overall view of precipitation anomalies associated with the various stages of EN events over SSA. RH87 identified in SSA a region with consistent precipitation anomalies from November of EN year through February of the following year. However, they found a relatively low coherence for this region regarding rainfall anomalies during EN events, which indicates that there are different timings for these anomalies along a EN cycle. The regional differences point out to the importance of a more detailed study with a large database. 2. Data and methodology The precipitation data used in this study are monthly amounts of 134 selected stations from Argentina, Brazil, Chile, Uruguay and Paraguay, for the period 1956-1991. The years of EN episodes included in this study were chosen following the same criteria as in Grimm et al. (1998) and indicate the beginning of the episodes (Table 1).
TABLE 1. List of El Niño episodes included in this study El Niño years 1957, 1963, 1965, 1969, 1972, 1976, 1979, 1982, 1986, 1991. In order to allow a comparison of the results with those of RH87, Pisciottano et al. (1994) and Grimm et al. (1998), the methodology is based on RH87 and is summarized below. Some questions related to this methodology are discussed in Grimm et al. (1998). The first three steps aim the analysis of the spatial structure of the rainfall anomalies associated with EN events, besides giving a first estimation of their magnitudes. 1) At each station, the monthly precipitation data are represented as percentile ranks. EN episodes composites of the percentile ranked precipitation are formed for the 24-month period from July of the year before (July (-)) to June of the year after an episode (June (+)). 2) The first Fourier harmonic of each composite is represented as a vector (amplitude and phase). The phase of the vector refers to the maximum of the first harmonic. 3) Regions of spatially coherent EN-related rainfall anomalies are selected, through the maximization of an index of coherence, given by the ratio between the magnitude of the vector sum and the sum of the magnitudes of the vectors for all stations in a region. In each region, the timing and consistency of the anomalies are examined as follows. 4) Monthly rainfall amounts for each station are transformed into precipitation percentiles based on gamma distributions fit to the data of each calendar month. As the time series for some stations in northern Chile contain several occurrences of zero monthly precipitation amounts, a correction is introduced in the gamma distribution fit to these series. 5) EN composites are formed from the precipitation percentiles for each station for the 36- month period from January (-) to December (+) and averaged to form a EN aggregate composite for each coherent region. 6) Time series of station-averaged precipitation percentiles for the periods identified in step 5 are analyzed in order to assess the statistical significance of the relationship between EN events and rainfall anomalies. The hypothesis is tested that these periods are especially wet or especially dry during these episodes, by using the hypergeometric distribution, as explained in Grimm et al. (1998). 3. Results Figure 1 shows the vectors representing the maximum of the first harmonic fit to the composite percentile precipitation at each station. It suggests the existence of eight regions of coherent behavior as regards precipitation anomalies during EN events. The vectors map confirms the conclusion of Grimm et al. (1998) that in SSA the area of largest impact of EN events on precipitation is South Brazil. The vectors present a very ample range of directions, reflecting the shift of the maximum of the first harmonic from the winter of year (0) to the winter of year (+), with most frequent occurrence during the spring-summer of the year (0). This is why RH87 found a low coherence for the region they determined in SSA. It is worth pointing out that the maximum of the first harmonic does not always indicate the actual period of the maximum anomaly. This can be seen in the aggregate composites for each region (not shown). In spite of the differences in the magnitude and timing of the anomalies between regions, it is possible to summarize a general behavior, from the aggregate composites (not shown) and the consistency of the anomalies. Table 2 shows the statistical significance of the hypothesis test that the indicated seasons are wetter or drier than normal during EN events. It is worth pointing out that not always the largest average anomalies are the most consistent ones. This is explained by the influence of outliers on average anomalies.
In the year before EN there is a tendency to less than normal precipitation from April (-) to March (0) (with a discontinuity in September (-)), with the exception of region 6, where it rains above normal during July (-) - November (-). There are positive rainfall anomalies during the spring (August (0) November (0)) over all regions but regions 6 and 7. In January (+), the anomalies weaken (or reverse their sign), which begins in December over the coastal regions of Southern Brazil. The behavior is not so coherent during the year (+). While it rains more than normal in Southern Brazil during the winter of the year (+), there are dry or not consistent wet anomalies in other regions. 4. Conclusions The whole SSA, south of 15S, presents precipitation anomalies associated with EN events. The timing of these anomalies changes throughout the region, leading to eight different coherent regions. All of them show a significant response at some part of the cycle. Their spatial coherence index is much higher than that obtained by RH87 because they determined only one coherent region, which encompasses most of our regions 1, 2 and 3. The intensity of the response (approximated by the amplitude of the first harmonic) is also higher in the present study. This may be due to the lower coherence of the vectors in RH87, but also to the earlier beginning of their data series. In fact, the EN influence on SSA seems to have increased around the 60 s. Southern Brazil has the strongest signal in the EN event. The main regions in the EN and La Niña (LN) cases are basically the same, except for some minor differences (Grimm and Doyle, 1998). Furthermore, there is a tendency towards opposite signals in the precipitation anomalies during almost the same months of the EN and LN cycles, indicating a large degree of linearity in the response over SSA to these events The regionalization of rainfall anomalies is a consequence of the processes associated with the circulation anomalies produced by EN events. Each of these processes has largest impact at a particular phase of the EN cycle and on specific regions. These processes are presented in a companion paper (Grimm and Barros, 1998). Acknowledgments. This research has been supported by CNPq (Brazil) and CONICET (Argentina). Alice Grimm also received support from the Federal University of Paraná Foundation (FUNPAR) and Moira Doyle from the University of Buenos Aires. We are grateful to Simone E. T. Ferraz, Andrea de O. Cardoso, Daniel Weingaertner and Rodrigo Siqueira, for their help in processing the data. REFERENCES Aceituno, P, 1988: On the functioning of the Southern Oscillation in the South American sector. Part I: surface climate. Mon. Wea. Rev., 116, 505-524. Grimm, A. M., and V. R. Barros, 1998: Processes leading to precipitation anomalies in Southern South America during El Niño and La Niña cycles. Anais do X Congresso Brasileiro de Meteorologia. Sociedade Brasileira de Meteorologia. Grimm, A. M., and M. E. Doyle, 1998: Regionalization of precipitation anomalies associated with La Niña events in Southern South America. Anais do X Congresso Brasileiro de Meteorologia. Sociedade Brasileira de Meteorologia. Grimm, A M., S. E. T. Ferraz and J. Gomes, 1998: Precipitation anomalies in Southern Brazil associated with El Niño and La Niña events. J. Climate (in press). Pisciotano, G., A. Diaz, G. Cazes and C. R. Mechoso, 1994: El Niño-Southern Oscillation impact on rainfall in Uruguay. J. Climate, 7, 1286-1302. Ropelewski, C. H., and S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115,1606-1626.
El Niño Period\Region 1 2 3 4 5 6 7 8 Jan(-) Mar(-) 43.58 d 68.91 w 98.13 w 34.27 w 34.28 d 54.61 w 28.46 d 74.91 d Feb(-) Apr(-) 43.58 w 61.14 w 92.92 w 48.95 d 39.81 w 40.43 d 36.54 d 37.28 w Mar(-) May(-) 43.58 d 43.58 d 75.84 w 64.62 d 92.17 d 85.02 d 88.46 d 92.88 d Apr(-) Jun(-) 81.82 d 71.89 d 78.55 d 84.85 d 80.34 d 87.68 d 99.50 d 57.95 d May(-) Jul(-) 98.18 d 60.23 d 60.23 d 89.01 d 97.67 d 86.86 d 96.70 d 74.91 d Jun(-) Aug(-) 86.79 d 47.26 d 98.18 d 64.62 d 99.43 d 43.21 d 48.90 w 70.66 d Jul(-) Sep(-) 98.18 d 68.91 w 78.55 d 68.93 d 81.99 d 70.30 w 76.92 w 79.64 w Aug(-) Oct(-) 98.18 d 39.77 w 43.58 d 54.55 d 63.75 d 51.55 w 66.47 w 34.36 d Sep(-) Nov(-) 78.55 d 78.55 d 64.44 d 68.93 d 72.46 d 90.82 w 48.90 w 57.87 d Oct(-) Dec(-) 90.76 d 78.55 d 64.44 d 48.95 d 70.27 d 73.74 w 50.00 d 66.79 d Nov(-) Jan(0) 89.08 d 94.20 d 73.09 d 45.24 d 43.72 w 40.45 w 85.00 d 94.33 d Dec(-) Feb(0) 84.50 d 87.36 d 84.50 d 45.24 w 55.44 w 68.79 d 38.24 d 41.14 d Jan(0) Mar(0) 43.58 d 99.43 d 99.43 d 82.05 d 34.28 d 77.63 d 71.54 w 54.29 w Feb(0) Apr(0) 56.42 d 68.25 d 93.81 d 58.74 d 74.36 w 73.91 d 60.77 d 62.72 d Mar(0) May(0) 43.58 d 43.58 d 48.09 w 65.73 w 59.46 w 85.02 d 48.90 d 81.25 w Apr(0) Jun(0) 71.89 w 95.30 w 92.92 w 87.88 w 49.89 w 37.58 w 87.41 w 73.01 w May(0) Jul(0) 56.42 w 68.91 w 68.91 w 60.44 w 39.81 w 41.08 w 84.62 w 81.25 w Jun(0) Aug(0) 64.44 d 52.74 w 56.42 w 82.38 w 75.77 w 65.02 w 51.10 d 59.52 w Jul(0) Sep(0) 56.42 w 39.77 w 86.79 w 87.41 w 75.77 w 70.30 w 72.53 d 50.90 w Aug(0) Oct(0) 81.82 w 68.91 w 81.82 w 37.76 d 98.95 w 51.55 w 33.53 d 94.01 w Sep(0) Nov(0) 92.92 w 92.92 w 92.92 w 87.41 w 98.23 w 41.12 d 51.10 d 84.34 w Oct(0) Dec(0) 99.35 w 92.92 w 97.00 w 90.21 w 92.72 w 94.00 d 69.00 d 75.77 w Nov(0) Jan(+) 92.75 w 96.87 w 93.41 w 83.33 w 83.29 w 42.57 d 36.54 d 78.34 w Dec(0) Feb(+) 54.01 w 80.56 w 80.83 w 33.33 d 79.74 d 77.00 w 38.24 w 58.86 w Jan(+) Mar(+) 42.26 w 96.45 w 83.48 w 93.01 w 70.27 d 38.49 w 85.90 d 54.29 w Feb(+) Apr(+) 57.74 w 78.23 w 87.80 w 80.42 w 60.19 d 40.43 d 36.54 d 67.61 w Mar(+) May(+) 71.66 w 71.66 w 64.74 w 53.85 w 59.46 w 93.93 w 34.62 w 45.71 d Apr(+) Jun(+) 42.26 d 42.26 w 65.23 d 27.27 w 50.11 d 96.71 w 54.55 d 57.95 d May(+) Jul(+) 71.66 w 42.75 d 72.25 d 27.47 w 39.81 w 56.79 w 27.47 w 81.25 w Jun(+) Aug(+) 93.94 w 58.62 d 42.26 w 46.15 d 45.37 w 90.87 d 60.84 d 85.13 w Jul(+) Sep(+) 50.00 w 42.75 d 87.71 d 49.45 d 54.63 d 73.92 d 34.27 w 50.90 w Aug(+) Oct(+) 57.74 d 91.34 d 57.74 d 39.56 d 97.55 d 79.22 d 75.34 d 54.63 d Sep(+) Nov(+) 87.71 d 87.71 d 50.00 d 49.45 d 92.17 d 40.82 w 34.62 d 41.14 d Oct(+) Dec(+) 71.66 w 35.26 d 50.00 d 61.54 d 34.28 d 82.80 w 28.82 d 54.63 d TABLE 2. Level of significance of the hypothesis test that the indicated periods are wetter (w) or drier (d) than normal during El Niño events. Only the highest values are included and values above 90% are highlighted