Forecast of the October-December 2002 Atmospheric Circulation Using the UCLA-AGCM and the NCEP-forecasted TP-SST, combined with a statistical downscaling to estimate Oct-Dec/02 Precipitation in regions of Uruguay-Rio Grande do Sul (Brazil) contributed by Gabriel Cazes and Gabriel Pisciottano

GDAO ¹, IMFIA², Universidad de la Republica, Montevideo, Uruguay

¹Grupo de Dinamica de la Atmosfera y el Oceano

²Instituto de Mecanica de los Fluidos e Ingenieria Ambiental

Several studies have confirmed a significant influence of the ENSO on the interannual rainfall variability of South-Eastern South America (SESA), with a tendency for above normal rainfall during warm Pacific events, especially in November and December, i.e. austral spring, when the event is near its mature phase (Rapelewski and Halpert, 1987; Aceituno 1988; Pisciottano et al. 1994; Diaz et al. 1998; Montecinos et al., 200; Grimm et al., 2000; Barros et al, 2000). There are spatial gradual variations both, in the annual cycle and, also, in the relationships of the subregional rainfall with the ENSO phenomenon, which include the timing, intensity and persistence (measured e.g. through statistical significance), etc., of the ENSO response. Mechanisms explaining this association have to do with changes in the subtropical southern westerlies by remote forcing of the large scale upper SH circulation (PSA type anomalous patterns) and local interaction with the regional-scale mechanisms (Aceituno et. al. 1989, Grimm et. al. 2000, Robertson et. al. 2001). The location (e.g. N-S position, exit regions), structure, intensity and other features of the anomalous jet stream determine both the baroclinic developments and the probability of convective events in SESA. Cazes et al. (1996, 2000) showed that changes in circulation and precipitation patterns affecting Uruguay, during November of warm Pacific events, can be simulated through the UCLA AGCM and well compared with adequate composites of past events.

These ENSO-regional rainfall relationships have been used to develop statistical prediction techniques of seasonal regional precipitation based on the ENSO state in subregion of SESA (see Pisciottano et al. 1997, 1999, 2000, 2001) which have been evaluated a-posteriori.

Also, regional scale upper troposphere (200 hPa) circulation above the SESA + Atlantic sector statistical patterns have been described (and will be shortly revised here) and also its relationships to rainfall in SESA, at interannual and seasonal time scales. Particularly for spring, an anticyclonic vortex is identified as the most relevant mode of variation of the upper troposphere circulation, and it appears to be associated to increased seasonal rainfall in the UYRS subregion. (Cassarin and Kousky 1986, Mo and Nogues-Peagle 2001, Barros et. al. 2000, Robertson et. al. 2001). This allow us to built a statistical downscaling technique from atmospheric circulation to regional rainfall.

As NCEP CMP 14 Tropical Pacific (TP) SST forecast (Ji et. al. 1998) are available, we use that SST forecast to simulate an ensemble mean of the global atmospheric circulation. Then the downscaling technique allow us to "project" the regional part of this circulation on the statistical patterns and estimate the regional rainfall in UYRS based on statistical relationships between these projections and rainfall. In these way we by-pass the use of model-calculated seasonal rainfall which we know have deficiencies derived from the low resolution of the version of the model used, and from other causes as deficiencies in the tropical South America processes, etc. Then, we expect that, as, for spring, the most relevant part of the ENSO influence on rainfall in SESA is caused via subtropical and extratropical changes of the atmospheric circulation (as PSA type patterns), we can simulate the large scale regional aspects of the atmospheric circulation affecting SESA and then estimate the regional rainfall at regions in SESA (UYRS).

First, we show results for the simulation of the atmospheric circulation of five warm TP cases and its projection in the most relevant statistical pattern, in order to approach to have a qualitative idea of the skill of the NCEP-forecasted TP-SST based UCLA-AGCM ensemble simulation during warm Pacific events.

Second, we use August/2002 calculated NCEP TP-SST forecast for Oct-Dec/2002, to simulate the Oct-Dec/02 atmospheric circulation, and then we project this (regional) circulation on the principal mode in order to infere the regional rainfall for this season by using the downscaling technique, in four regions in UYRS.

Simulation of past warm events.

We use monthly NCEP CMP 14 TP SST validation forecasts provided by the NCEP, for the season September to January, with initial conditions measured during August, for the years 1982, 1986, 1991, 1994 and 1997 (warm episodes). We simulate the atmospheric circulation with the UCLA AGCM. The UCLA AGCM is a finite difference model, with state of the art parameterization of the physical processes. Its description and recent developments can be found at the site www.ucla.atmos.edu. In this work we use the 6.95 version, in low resolution, that has an horizontal resolution of 4º of latitude by 5º of longitude, and 16 layers in the vertical direction, which extends from the earth surface to the level of 1 hPa. For each warm case to be simulated, an ensemble of four simulations is calculated, by prescribing the global SST: in the tropical Pacific, (between 14S and 18N, and between 135E and the Americas), the NCEP CMP 14 SST forecast is prescribed. The adjacent regions that extend between 25S -14S, 120E-135E, and 18N- 30N are used as transitions, where the NCEP SST forecast is linearly relaxed to the climatological SST, which is prescribed elsewhere. The climatological SST used here is derived from the GISST dataset (Rayner et. al. 1994), and it is computed from 1961 to 1990. Each individual simulation extends from September 25 to December 31. The only difference among these simulations, in each warm event case, is found in the initial conditions; the initial zonal and meridional wind, the moisture and the temperature 3-D fields have in each point of the grid a random perturbation, specific for each run. The random perturbations never exceed the 2% of the respective absolute value of those magnitudes. We substract from the average of these runs the average of other four runs ("control" or "model climatology"), for which global climatological SST is prescribed throughout all the ocean. The difference of these averages represent the model forecast of the anomalous circulation.

In order to access the skill of the AGCM for our purposes (prediction of the circulation directly affecting SESA during a TP warm event) when is run using forecasted SST as boundary condition, we compare the AGCM results with the respective NCEP Reanalysis circulation (as in similar previous studies, Cazes and Pisciottano 2000, which were bases in observed SST anomalies). We found that the 6.95 version of the model, in low resolution, with the NCEO CMP 14 forecasted TP SST has some skill for forecasting the 200 hPa anomalous circulation in the South Pacific and South America regions, during the season October-December corresponing to warm episodes. Fig. 1-a shows the anomalous 200 hPa circulation for October-December averaged over the years 1982, 1986, 1991, 1994 and 1997, derived from the NCEP Reanalysis, (Kalnay et. al. 1996). Fig. 1-b shows the average of the respective forecasts. We compute the pattern correlation of these fields, as in Farrara et. al. 2000; we focus on a region directly related to SESA: 46S-14S, 70W-20W, whish is shown with a dashed line in Figs. 1-a and 1-b. The pattern correlation coefficient is 0.58, similar to those obtained in other studies but focused on other regions and periods of interest.

Qualitatively, we note that the composite of the "forecasted" five warm events shows that a clear Walker cell type response is obtained to the west of the TP warm anomaly, a clear Hadley cell and subtropical South Pacific response appears, and the typical extratropical response, "Pacific South America" (PSA) teleconnection pattern, affecting subtropical and mid-latitude South America, is also obtained (Fig. 1-b), and these are also the most relevant features observed (Fig. 1-a). There is not agreement in the tropical sector of the Southamerican continent, an issue already found in previous studies (Bernardi 2002). However, in the box or region of our focus, the qualitative agreement can be observed, the anticyclonic anomaly, and it is quuantified by the pattern correlation coefficient.

Downscaling from circulation to precipitation.

Fig. 1-a and b show a composite of observed and simulated circulation on our focus box. Interannual variability of the 200 hPa October-December circulation on the box can be approached via calculation of the EOF patterns. We compute EOFs to the (u,v) 200 hPa observed anomalous field, over the 46S-14S, 70W-20W domain. For this purpose we take the anomalies of the October-December 200 hPa zonal and meridional winds between 1968 to 2000, derived from the NCEP reanalysis. For each year, the u and v fields are combined into a single vector, as in Robertson and Mechoso 2000, and EOFs are computed from the corresponding covariance matrix. The grid of the observed winds is interpolated to the same grid of the model before doing this computations. The first three PCs, account for a 67% of the added variance of all the PCs. Table I shows the variance of each PC as a percentage of the total added variance of all the PCs.

Figure 2 shows the EOF1 in terms of linear regression of the (u,v) field with the standardized time series of the respective PC, as in Robertson and Mechoso 2000. This EOF, again, shows an anticyclonic vortex. Figure 3 shows the respective PC1, that accounts for a 36% of the total variance. Note that many of the years when PC1 peaks (positive or negative) are strongly related to ENSO extremes. Correlation between this PC and the Niño 3.4 index is 0.60.

For studying the precipitation we focus on (four) subregions in Uruguay and Rio Grande do Sul (in southern Brazil), herein UYRS, (Fig. 4). For each region we compute regional rainfall time series as an average of the raingauge data in the region, for each October-December from 1968 to 1998. The location of the raingauges is also shown in Fig. 4. These data was supplied by DINAMET (Uruguay) and FEPAGRO and INMET (RS-Brazil). Table 2 show the correlation of the regional precipitation time series with the first three Pcs of the regional 200 hPa circulation. It is found that the correlation with PC1 is statistically significant at a level above 99%.

Fig. 6 shows the statistical relationship between PC1 and the regional precipitation, for each region at UYRS. This is the key piece in the downscaling procedure we will use to estimate the October-December regional precipitation. This plot works as a statistical predictor using the PC1 (observed or forecasted) as input, and the output will be an interval of regional precipitation associated to a statistically representative number of cases (* in the plots). Anomalous forecasted anomalous circulation for the box, by the UCLA AGCM ensembles, based on NCEP CMP 14 TP SST forecasts, can be projected on the first pattern (EOF1), as the observed circulation, obtaining a forecasted PC1 value for each warm event. Fog. 3 also shows these forecasted values (by *s) for the years 1982, 1986, 1991, 1994, 1997, which shows in an additional way, the skill of the UCLA AGCM ensemble simulations to well simulate the regional upper troposphere circulation on SESA, associated to those warm events during spring. It also appears that the simulations show less variations from event to event than the reanalysis based calculated PC1 ("underestimation of positive extremes").

The October-December 2002 forecast.

Fig. 5 shows the forecasted October-December 2002 200 hPa anomalous circulation, based on the NCEP CMP 14 TP SST forecast, and the same version and resolution of the UCLA AGCM than in the simulation of the past event. It is observed, again, the characteristic anticyclonic anomalous circulation. Fig. 3 also shows the forecasted PC1 for October-December 2002. By using this forecast of the PC1 (2002), which is 16.2, we select an interval around this value, from -4 to 36, in order to have enough cases (years with similar values of the PC coefficient) to allow a reasonable robust estimation of parameters (median m* and quartils q* and Q*) of the associated regional precipitation conditional distribution (subpopulation) as it is shown in the plots in Fig. 6.

These results enable us to predict the expected regional 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.

Based on this study, we issue the following rainfall predictions for the period October-December 2002 for these 4 regions of Uruguay-Rio Grande do Sul (UYRS, see Fig. 4):

Region 1- Southern and Southeastern Uruguay. A value around 266 mm of rainfall is expected (close to mcl, which is 269 mm) with 50% chance of between 199 and 368 mm.

Region 2- Northeastern Uruguay and Southeastern Rio Grande do Sul. A value around 346 mm of rainfall is expected (instead of mcl, which is 325 mm) with 50% chance of between 244 and 467 mm.

Region 3- Central and Northwestern Uruguay and Southwestern (inland) Rio Grande do Sul. A value around 418 mm of rainfall is expected (instead of mcl, which is 367 mm) with 50% chance of between 348 and 455 mm.

Region 4- Northern Rio Grande do Sul. A value around 485 mm of rainfall is expected (instead of mcl, which is 426 mm) with 50% chance of between 403 and 593 mm.

In summary, a wetter than normal October-December/2002 period is expected in Uruguay-Rio Grande do Sul. This tendency to a wet period is stronger in the inland region close to the Uruguay-Brazil boundary. In Southern Uruguay the difference from the climatological distribution is weak.

Acknowledgments: To Dr. A. Kumar and Dr. W. Wang for providing the historical CMP 14 TP SST forecasts, to Prof. Moacir A. Berlato from UFRGS-Br, for the dataset from RS stations and Dirección Nacional de Meteorología-Uruguay for the UY dataset. This work was supported by the Instituto Nacional de Investigaciones Agropecuarias (INIA), Uruguay, project FTPA Nº 81.

References

Aceituno 1988: On the functioning of the Southern Oscillation in South America, Part I: Surface Climate. Monthly Weather Review, 116, 505-524.

Aceituno 1989:On the functioning of the Southern Oscillationin South America, Part II: Upper air circulation. Journal of Climate, 2, 341-355

Barros, V., M. Gonzalez, B. Liebmann, and I. Camilloni, 2000: Influence of the South Atlantic convergence zone and South Atlantic Sea surface temperature on interannual summer rainfall variability in Southeastern South America. Theoretical and Applied Climatology, 67, 123-133.

Bernardi R. 2002: Influencia de las TSM sobre la circulación en el Sudeste de Sud América en Invierno y relación con las Precipitaciones regionales. Simulación del caso 1972 con un Modelo de Circulación General. X Jornada de Jovenes Investigadores de la Asociación de Universidades Grupo de Montevideo.

Cassarin D. P. and Kousky 1986: Precipitation anomalies in the southern part of Brazil and variations of the atmospheric circulation. Revista Brasileira de Meteorologia, 1, 83-90.

Cazes G., G. Pisciottano and R. Terra, 1996:Variabilidad climática en el sudeste de Sud America y patrones anómalos en Noviembre. Simulación con un modelo General de Circulación de Atmósfera. VII Congreso Iberoamericano de Meteorología, Setiembre 1996, Buenos Aires, p 336-336.

Cazes G. and G. Pisciottano, 2000:Climate variability on Southeastern South America related to ENSO. A numerical study. Proccedings of the VI International Conference on Southern Hemisphere Oceanography and meteorology, Santiago, Chile, 2000.

Díaz A., C. Studinski and C. R. Mechoso, 1998: Relationships between precipiation anomalies in Uruguay and Southern Brazil and Sea Surface Temperature in the Pacific and Atlantic Oceans, Journal of Climate, 11, 251-271

Farrara, J. D., C. R. Mechoso, and A. W. Robertson, 2000: Ensembles of AGCM two-tier predictions and Simulations of the Circulation Anomalies during Winter 1997-1998. Mon. Wea. Rev., 128, 3589-3604.

Grimm, A. M, V. R. Barros, and M. E. Doyle, 2000: Climate Variability in Southern South America Associated with El Niño and La Niña Events. Journal of Climate, 13, 35-58.

Ji M. A., D. W. Behringer and A. Leetma 1998: An improved copupled modelo for ENSO prediction and Implications for Ocean initialization, part II: the coupled model. Monthly Weather Review, 126, 1022-1034.

Mo K. C. and J. Nogués-Peagle, 2001: The Pacific-South American Modes and their downstream effects. International Journal of Climatology, 21, 1211-1229.

Montecinos A., A. Díaz and P. Aceituno 2000:Seasonal Diagnostic and predictability of Rainfall in Subtropical South America Based in Tropical Pacific SST. Journal of Climate, 13, 746-758.

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

Pisciottano G., G. Cazes and A. Díaz 1997:A statistical-empirical forecast for November-December precipitation in Uruguay based on ENSO state, Experimental Long Lead Forecast Bulletin, NOAA, Sept. 1997.

Pisciottano G., G. Cazes and A. Díaz 1998:A statistical-empirical forecast for April-July 1998 precipitation in Uruguay based on ENSO state, Experimental Long Lead Forecast Bulletin, COLA, University of Maryland, March 1998.

Pisciottano G., G. Cazes and A. Díaz 1998:A statistical-empirical forecast for October-December precipitation in Uruguay based on ENSO state, Experimental Long Lead Forecast Bulletin, COLA, University of Maryland, September 1998.

Pisciottano G., G. Cazes and A. Díaz 1999:A statistical-empirical forecast for March-July 1999 precipitation in Uruguay based on ENSO state, Experimental Long Lead Forecast Bulletin, COLA, University of Maryland, Sept. 1999.

Pisciottano G., G. Cazes, A. Díaz and J. L. Genta 2000:A revision of the IMFIA-UR seasonal rainfall forecast method based on the ENSO state, Proccedings of the VI International Conference on Southern Hemisphere Oceanography and Meteorology, Santiago, Chile, 2000.

Robertson A. and C. R. Mechoso 2000: Interannual and Interdecadal variability of the South Atlantic Convergence Zone. Mon. Wea. Rev., 128, 2947-2957.

Robertson, C. R. Mechoso y G. Cazes, 2001: Interannual and Interdecadal variability of South American Monsoon, Eight Scientific Assembly of the International Association of Atmospheric Sciences, Insbruck, Austria

Ropelewski C. and M. Halpert, 1987: Global and regional scale precipitation patterns associated with El Niño Southern Oscillation- Monthly Weather Review, 115, 1606-1626

Tables

Table I: Percentage of the total added variance for the first three Pcs

PC1	PC2	PC3
36%	20%	11%

Table II: Correlation coefficient of the precipitation time series of each region with the first three PCs

	PC1	PC2	PC3
REGION 1	0.41	0.27	0.12
REGION 2	0.57	0.08	0.08
REGION 3	0.67	0.10	0.06
REGION 4	0.60	0.22	0.08

Figures captions

Figure 1: a. Composite of the anomalous 200 hPa October-December circulation, derived from the NCEP Reanalysis, for the years 1982, 1986, 1991, 1994 and 1997. Wind vectors are shown only in the grid-points were either u or v anomalies are statistically significant at the 95% level, with a t-Student double tail test. Scale shown in the bottom left corner. The dashed box indicates the region between 46S-14S, 70W-20W, used in the downscaling. b. Same as a, but derived from the UCLA-AGCM forecasts with the NCEP CMP 14 TP SST forecast.

Figure 2: First EOF for the October-December anomalous circulation, for the years 1968-2000, in the domain 46S-14S, 70W-20W. The data is on a grid of 4º of latitude by 5º of longitude. The zonal and meridional components of each arrow are the linear regression coefficients of the local u and v time series with the standardized PC1 time series. Arrows are only shown in those points were either u or v has a statistically significant correlation coefficient with the PC1, at a 95% level.

Figure 3. Temporal series of the PC1. Stars show the projection of the UCLA-AGCM forecasts (with CMP 14 TP SST forecast) on the first EOF.

Figure 4. Subregions in UYRS for computation of averaged rainfall data considered in this study. Locations of the raingauges are shown with *. Isolines show the climatological mean of the precipitation in October-December (mm).

Figure 5. UCLA-AGCM forecast (with the NCEP CMP 14 TP SST forecast) of the October-December 200 hPa anomalous circulation. Indications about statistical significance and scale are the same as Fig. 1.

Figure 6. PC1 vs. rainfall in region 1, 2, 3 and 4, for the period 1968-1998. The horizontal line shows the median of the precipitation for the whole population. Vertical bars show the interval of values of PC1 between -4 and 36.