This study focuses on two aspects of the ENSO impact on the agricultural sector of southeastern South America (SESA). The first topic is a statistical exploration of ENSO signals on the regional climate. We highlight the importance of decadal-scale climate trends for (a) quantifying the ENSO signal, and (b) formulating risk scenarios for wetter and drier decades. Low-frequency precipitation flu--ctuations were most noticeable: differences of about 180 mm for October-April (growing season for summer crops) occurred between the 1950s (drier) and the 1980s (wetter). After removing low-frequency trends, we investigated ENSO effects on agronomically important climate variables. An ENSO signal on rainfall, maximum temperature and radiation is present in October-March, and is more pronounced in November-January. Higher (lower) rainfall and lower (higher) maximum temperature and radiation were observed during warm (cold) events. ENSO effects on minimum temperature occur in two periods: June-August and March-April. Lower (higher) minimum temperatures were observed during cold (warm) events. Solar radiation anomalies during cold events could account for a 15% increase in sunflower potential yield, while rainfall anomalies could result in a 20% decrease in maize yields.
Statistical analyses of ENSO signals are limited by the availability of long historical records. Therefore, the second part of this study focuses on the use of stochastic precipitation generators for ENSO-related risk assessment and management. Precipitation generators can produce long, synthetic daily weather series with statistical characteristics similar to those of historical data. We have developed precipitation generators conditioned on ENSO: model parameters are estimated separately for each ENSO phase. We tested various parameterization schemes in six agriculturally important SESA locations for October-March. Conditional precipitation models (for occurrence, intensity, or both) were superior to simple models in 24 of the 36 location/month combinations analyzed. Both occurrence and intensity conditional models were comparably supported, suggesting the ENSO precipitation signal is associated with both the number of rain days and the rainfall amounts on wet days. Simple occurrence models overestimate (underestimate) the probability of observing dry spells of a certain length during warm (cold) events. Conditional model capture differences in the number and persistence of wet days among ENSO phases. Simple intensity models underestimate (overestimate) daily precipitation amounts during warm (cold) events, whereas conditional models improved the agreement between theoretical and empirical distributions of daily rainfall. Simulated distributions of monthly precipitation totals separated clearly by ENSO phase, reproducing the behavior of observed series. In contrast, unconditional models under-represented both the low and high precipitation totals.