A Statistical-Empirical Forecast of March-July 1999 Precipitation

in Uruguay Based on the ENSO State

contributed by Gabriel Pisciottano, Gabriel Cazes, Alvaro Diaz and Jose L. Genta.

GDAO ¹IMFIA ² ,Universidad de la Republica, Montevideo, Uruguay

¹GDAO Grupo de Dinamica de la Atmosfera y el Oceano

²IMFIA Instituto de Mecanica de los Fluidos e Ingenieria Ambiental

Rainfall in Uruguay (30-35S, 54-58W, see Fig. 1) occurs during all 12 months of the year. The complex interaction of several rainfall mechanisms produces a wet climate with a "flat" annual cycle of precipitation. This is the case especially in the southern part of Uruguay, which is influenced by the Rio de la Plata and the Atlantic Ocean. In the northern part of Uruguay, fall and spring average rainfall exceeds that of winter and summer (Pisciottano et al. 1994, 1997).

Various studies have confirmed a significant influence of ENSO on interannual rainfall variability in Uruguay, with a tendency for below normal rainfall in association to cold events in the Pacific Ocean, especially between October and December (Ropelewski and Halpert 1987; Aceituno 1988; Pisciottano et al. 1994, 1997; Diaz et al. 1998). The rainfall is also below normal during the March-July period of the year following a year with a high Southern Oscillation Index (HSOI).

Pisciottano et al.(1998) presented a forecast calling for a drier than normal October-December period in 1998 in all regions of Uruguay, with probabilities of below median precipitation ranging from 89% to almost 100%.

Table 1 shows that the predicted negative anomalies in the northern region were consistent with the observations. In the southwestern region the observed value was slightly above the climatological median and the forecast can be considered "slightly wrong". In the southeast the observed rainfall (336 mm) was similar to the 75th percentile of the climatological distribution (330 mm), far above from the forecast value (196 mm) and from the forecast interval (163 mm - 240 mm). The forecast can be considered "wrong" for this region.

For diagnostic studies and for the results presented here, we use the 1950-1997 time series of monthly values of the Niño 3.4 Index (N3.4) of SST anomalies in the central tropical Pacific (C *100), available from NOAA. In previous studies (see Pisciottano et al. 1997) we used the Index of Wright. The diagnostic studies of the relationships between the Index of Wright and precipitation in Uruguay, and between the Niño3.4 Index and precipitation in Uruguay appear to support the idea that a substantial part of these links is related to the SST field in the tropical central Pacific. Accordingly, it seems to be sufficient to select the index related to Niño 3.4 region to "forecast" precipitation anomalies in Uruguay, even when the available time series is considerably shorter than the Index of Wright. Using a shorter record has an impact on the statistical significance. We also note that the rainfall climatology used in this study is defined by values from the last 48 years so that decadal variations cannot be accounted for.

Diagnostic studies have revealed that there are highly significant negative precipitation anomalies between March and July of the years following a HSOI year in all regions in Uruguay (Pisciottano et al. 1994). HSOI years tend to agree with strong negative anomalies of N3.4 or Wright SST indices.

Our precipitation data are monthly values (in mm) from 13 rainfall stations in Uruguay (Fig. 1). The data come from the Direccion Nacional de Meteorologia (DNM) - Uruguay. For diagnostic and prediction purposes we have grouped the stations into the same four regions defined in Pisciottano et al.(1997) and (1998). For each region, we calculated a monthly regional precipitation value for each month in the common period spanned by the records of stations included in the respective region. There are no gaps for these periods, and no "filling-blanks" procedures were used. Time series for March-July are formed from the monthly values for each region.

We defined fixed equiprobable categories (quartiles) of the predicted variable. Changes in the probabilities given a specific climatic event (e.g. an ENSO event) are calculated for each category and used to form a probabilistic prediction. First the statistical distributional parameters (first quartile, median, and third quartile-called qcl, mcl and Qcl, respectively) for the whole record period are determined for March-July regional precipitation. These statistics characterize the rainfall climatology (cl) of the respective region. In the absence of any climate-determining information, a value "around mcl" is expected and the interval (qcl, Qcl) has a large (50%) chance of containing the real observed value. In this case, the "forecast" would be mcl and an "error bar" would be (qcl, Qcl).

To incorporate the knowledge provided by the ENSO-rainfall relationships, we compare the statistical parameters of the "whole population" (all the values of the March-July total precipitation through the years) with those of the "subpopulation" obtained by selecting only the March-July total precipitation values for the years whose previous September-October-November mean N3.4 index was lower than a "critical value" (N3.4cr). In agreement with former studies, we find that in the so defined "subpopulation" there is a tendency for lower rainfall. We test this statistical tendency with a hypergeometric distribution using the frequencies of above and below median rainfall in the population versus the frequencies for the "subpopulation". We note a balance between the "intensity" of a required event (as prescribed by N3.4cr) and the statistical significance of the rainfall differences (in the sense of Ropelewski and Halpert 1987; Pisciottano et al. 1994; Cazes et al. 1994). A very low N3.4cr produces a small number of events, yielding a lower significance than a moderate N3.4cr. It is found that N3.4cr=-60 (i.e., SSTA(N3.4)=-0.60 C) results in significance values higher than 95% for the 4 regions. Using this critical value, we estimate the statistical parameters q*, m*, and Q* for the "subpopulation" (Sep-Oct-Nov N3.4 < -60) for March-July for the 4 regions. The procedure is similar to that used by Ropelewski and Halpert (1996). The statistical parameters (qcl, mcl, Qcl) and (q*, m* and Q*) for the 4 regions are shown in Fig. 2.

NOTE: Similar studies using periods from April to July (i.e. excluding March) yield similar conclusions to an expected drier than normal season during 1999, with the exception of the NW region for which the inclusion of March produces a lower than normal rainfall for the March-July period 1999 and a no-clear tendency for the April-July 99 period.

Table 2 illustrates the rainfall distribution shift associated to years having Sep-Oct-Nov N3.4 < -60 through frequencies of occurrence of each of the four quartiles for each of the four regions.

These results enable us to "predict" the expected precipitation as the median of the "subpopulation" (m*) rather than the climatological median (mcl). The interval (q*, Q*) from the subpopulation is the "error bar" of the prediction. This technique circumvents problems associated with the nonlinearities of the SST-rainfall relationships and skewness of the rainfall distributions, and has been used during the past several years to experimentally predict rainfall in Uruguay. For example, in 1996 and 1997, a version of this technique predicted rainfall in three regions of Uruguay for November 1996-February 1997 and for September-December 1997 reasonably well. These forecasts were used by rice growers for water resource management (Pisciottano and Diaz, 1997; Cabral and Pisciottano 1997).

Based on the current values of the Pacific SST anomalies and the distributional shifts discussed above, we issue the following rainfall predictions for the March-July 1999 period for the 4 regions covering Uruguay.

Region I-NW Uruguay: A value approximately ("~") 409 mm of rainfall is expected, with 50% chance of between 316 and 490 mm. The probabilities of rainfall in the 1st and the 2nd quartile of the climatological distribution are 30% and 50%, respectively; and for rainfall below the median (497 mm) the probability is 80%.

Region II-NE Uruguay: ~400 mm is expected, with 50% chance of between 360 and 524 mm. The climatological 1st and 2nd quartiles have probabilities of 60% and 20%, respectively; and for rainfall below the median (585 mm) the probability is 80%.

Region III-SW Uruguay: ~320 mm is expected, with 50% chance of between 277 and 385 mm. The climatological 1st and 2nd quartiles have probabilities of 50% and 30%, respectively; and for rainfall below the median (426 mm) the probability is 80%.

Region IV-SE Uruguay: ~389 mm is expected, with 50% chance of between 319 and 437 mm. The climatological 1st and 2nd quartiles have probabilities of around 40% and 40%, respectively; and for rainfall below the median (443 mm) the probability is 80%.

In summary, a drier than normal March-July period is expected this year in all regions of Uruguay with 80% probabilities for below median precipitation for each region.

References

Cabral, A. and G. Pisciottano, 1997: Elementos para entender el fenomeno 'El Niño' y sus aplicaciones en la agricultura. Revista de la Asociacion de Cultivadores de Arroz del Uruguay. Junio 1997, 32-42.

Cazes, G., J. L. Genta and G. Pisciottano, 1994: Generacion de informacion hidrologicamente relevante a partir de informacion y diagnostico climatico. Aplicacion en Uruguay. Memorias XVI Congreso Latino-americano de Hidraulica, Santiago, Chile, 3, 121-127.

Diaz, A., C. Studzinski and C. R. Mechoso, 1998: Relationships between precipitation anomalies in Uruguay and southern Brazil and sea surface temperature in the Pacific and Atlantic Oceans. Journal of Climate, 11, 251-271.

Pisciottano, G., G. Cazes, A. Diaz and J. L. Genta, September 1998: A statistical-empirical forecast of October-December 1998 precipitation in Uruguay Based on the ENSO state. COLA - Experimental Long-Lead Forecast Bulletin, Vol 7, N 3, 81-85.

Pisciottano, G., A. Diaz, G. Cazes and C.R. Mechoso, 1994: El Niño-Southern Oscillation impact on rainfall in Uruguay. J. Climate, 7, 1286-1302.

Pisciottano, G. and A. Diaz, 1997: Diagnostico y prediccion climatica en Uruguay. Memorias del Taller "Variabilidad climatica interanual: metodos de pronostico e impactos asociados". Santiago, Chile.

Ropelewski, C.F., and M.S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606-1626.

Ropelewski, C.F., and M.S. Halpert, 1996: Quantifying Southern Oscillation-precipitation relationships. J. Climate, 9, 1043-1959.

Wright, P. 1989: Homogenized long-period Southern Oscillation Indices. Int. J. of Climatol., 9, 34-54.

Table 1. Climatology median, predicted and observed precipitation for October-December 1998 for each of the four regions in Uruguay (in mm). An interval of precipitation with 50% probability is indicated together with a central value.

Region	I-NW	II-NE	III-SW	IV-SE
Clim. median	322	334	276	247
Forecast Oct-Dec 1998	255 (213-291)	216 (174-267)	173 (148-235)	196 (163-240)
Observed	262	225	318	336

Table 2. Frequencies of occurrence of March-July rainfall in each of the four quartiles of the climatological distribution, given a year having the previous Sep-Oct-Nov N3.4 < -60 for the four regions in Uruguay. The quartiles are sorted according to increasing precipitation, i.e., q.1 is the lowest quartile and q.4 is the highest quartile.

	I-NW	II-NE	III-SW	IV-SE
cases	10	10	10	10
q.1	3	6	5	4
q.2	5	2	3	4
q.3	1	2	2	2
q.4	1	0	0	0

Fig.1. The 13 rainfall stations used in this study and the 4 regions in Uruguay. Stations: 1 Artigas; 2 Salto; 3 Paysandu; 4 Rivera; 5 Tacuarembo; 6 Melo; 7 Colonia; 8 Mercedes; 9 Trinidad; 10 San Jose; 11 Treinta y Tres; 12 Rocha; 13 Montevideo. Regions: I-NW, Northwest; II-NE, Northeast; III-SW, Southwest; IV-SE, Southeast.

Fig. 2. a) Medians and quartiles of the distribution of precipitation for the period March-July, in the region I-Northwest, for all the cases (Climatology) and for the subpopulation of cases when the Niño 3.4 index averaged in the previous September-October-November period was below -60 (N3.4 < -60). The data record spans from 1951 to 1997. b) Same as a), for the region II-Northeast. c) Same as a), for the region III-Southwest. d) Same as a), for the region IV, Southeast.