Brazil is the world’s leading soybean producer, so early, credible yield signals derived from climate variability are strategically important for growers, traders, and policymakers. Here we provide a systematic, multi-scale applied climatology assessment of whether the El Niño–Southern Oscillation (ENSO), used on its own, offers operationally useful early indications of Brazilian soybean yields. We analyse 2000–2020 crop-years at municipal, state, and national scales, remove long-run technological trends to isolate interannual variability, and quantify ENSO–yield relationships using linear correlation, phase-based contrasts, and multiple-testing control via the Benjamini–Hochberg false discovery rate. At the state level, associations are consistently weak and not robust to multiple-testing adjustment: across ten major producing states, the median absolute correlation is ≈ 0.17 (interquartile range 0.07–0.27), slopes are small (median − 53 kg ha⁻¹ per °C), and no state remains significant after false-discovery-rate control at q = 0.10. Phase contrasts likewise show strong overlap between neutral, El Niño, and La Niña years (only Bahia exhibits a borderline three-phase difference), and El Niño versus La Niña pairwise tests are uniformly non-significant, with Cliff’s δ typically small-to-medium and negative. Municipal correlation maps reveal a fine-grained patchwork of mixed signs that cancels under aggregation, explaining the absence of state- and national-scale signals. We conclude that ENSO-scale proxies alone are insufficient for operational yield forecasting or climate services for soybean in Brazil; investment should prioritise spatiotemporally varying, high-resolution weather-based models, with ENSO signals—at most—serving as weak background context.
Mello et al. (Sat,) studied this question.