This paper describes a new method to reliably monitor and assess drought impact on coffee yields just before the beginning of the maturation growth stage. A good model requires the evaluation of the sensitive indices and test for a specific variety and region. Phenological and yield data for Mundo Novo cultivar were taken from experiments conducted at the Instituto Agronômico Experimental Station located in Mococa (lat. 21°28'S, long. 47°01'W, alt. 665 m), São Paulo State, Brazil, during 1967 to 1974. The models tested are based in the penalization of the crop yield potential according the previous yield and the ER/EP relation (actual/potential evapotranspiration), derived by 10-day soil water balance occurred during different growth stages. These ratios were weighted by derivation of crop phase yield-response coeficients (ky values) in additive and multiplicative type models. Based on the model's statistics, the additive type model presented better performance than the multiplicative one. The additive model which considers the crop phases combination D1: bud dormancy/ flowering beginning (Jun/Jul/Aug), flowering/grain formation (Sep/Oct/Nov) and grain formation/maturation (Dec/Jan/Feb) presented the best performance. An analysis of the ky values shows that this combination D1 gives higher weigth to the water relations during the flowering and grain formation (fruit set), with R2 of 0.93 and d-index agreement of 0.98.
coffee; soil water balance; crop phase; agrometeorological models; coffee yield
