The increased demand for water use has caused environmental problems and conflicts between users. The management of water resources requires proper quantification of its availability. Concrete evidence of global climate change mandates the predictive study of water scarcity in basins along with preparatory actions to mitigate potential losses. In this context, the hydrological modeling allows a better understanding of the hydrological behavior in the basin. The use of hydrological models has great potential to characterize water availability in terms of climate change or land use. The objective of this work was development of a model, from hydroclimatic data, that estimates the minimum flow for seven consecutive days and return period of 10 years (Q7,10) and to assess the trend of variation of Q 7,10 in nine sections of the Paracatu River until the end of the 21st century. To conduct the study, monthly data of precipitation and temperature, and daily flow data for the period 1980 to 2000 were used. The model for estimating the minimum flow was developed based on the recession curve of underground flow, which has as parameters the recession coefficient (α) and the flow corresponding to the beginning of the recession period (Q0). The Q0 was estimated based on water balance, which included the monthly values of precipitation (P) and real evapotranspiration (ETR), obtained by the climatic water balance developed by Thornthwaite and Mather. The α value was obtained by analysis of annual hydrographs of groundwater flow, and its average value was used for each season. A predictive equation for Q 0 and a value of α was obtained for each station. With the daily flow rates obtained by the model, Q 7 was obtained (average of seven smaller flows) for each year (1980 to 2000) and with these values Q7,10 were estimated. Was also calculated for each station fluviometric, the Q7, 10 from the streamflow time series for the period 1980 to 2000. The model evaluation was performed using the coefficient Willmott (d), the relative percentage error (ER), and angular and linear coefficients of the regression equation. To assess the decline of Q7,10 due to climate change, data of monthly precipitation and temperature were simulated for the period 2011 to 2100, using the regional climate model ETA, A1B IPCC scenario, Member 2, in which the response to temperature change is low. The values of P and ETR were used in the model, thereby obtaining daily flow. With these, the Q7 for each year (2011 to 2100) and season, and Q 7,10 for each of the nine decades were calculated. Using the values of Q7,10 for each decade, a linear regression was adjusted for each fluviometric station. The model based on data hydroclimatic data, satisfactorily estimated values of minimum flows for seven consecutive days and return period of 10 years. It was estimated a tendency of reduction in water balance and q 7,10 until the end of the century, with four of the nine stations studied showed null values at the lower limit of the confidence interval with a confidence level of 99% for the parameter q7,10, in the decade 2091-2099, which indicates that these four sections, at least once every 10 years, the river probably will dry for seven consecutive days.
