Upland rice breeding led to increased drought sensitivity in Brazil

International audienceThe upland rice (UR) crop system located in the Brazilian savannas (states of Mato Grosso, Rondonia, Tocantins and Goias) is the main upland rice growing area of Brazil and the largest rainfed rice growing area in Latin America. It plays an important social and economic role in central Brazil. UR cropped area, however, has decreased by 70% in the last two decades, partly due to high drought risk. Here, we hypothesize that the UR breeding program's focus on direct grain selection for wide adaptation has led to an increase in yield potential but a reduction in the capacity of modem cultivars to maintain yield under drought stress. Using the ORYZAv3 rice crop model, we analyzed changes in relative and absolute drought impact for cultivars representative of three decades of breeding: 1980s (1986-1992), 1990s (1993-2002), and 2000s (2003-2013). We found a mean increase in relative drought impact of 12% (0.35% per year) between 1980s and 2000s. Most of the increase occurred during 1980s and 1990s (7%), where a major shift from landraces (i.e. Douradao) to modem cultivars (i.e. cv. BRS Primavera) occurred in the breeding program. Drought adaptation traits such as greater root length density, shorter cycle, and lower stomatal sensitivity to drought have been systematically bred out of the UR material in the period 1980-2013. We recommend the UR breeding program to be adjusted by better targeting to specific drought environments, and by taking into account drought stress profiles and environmental co-variables in statistical analyses for genotype selection. Moreover, physiological studies on tradeoffs between yield potential and drought tolerance should be performed for each target environment to dissect plant traits that confer both high yield potential and low drought sensitivity. These strategies will ultimately ensure that newly released cultivars are adapted to the environmental conditions of the study region, therefore reducing agroclimatic risk for UR farmers