ABSTRACT. Agricultural production in semi-arid regions is limited due to water availability. In addition, the water quality available for irrigation is often compromised due to the high salt content present. Millet is a forage species considered tolerant to water deficit and moderately salt tolerant. In view of the above, the objective was to evaluate the growth of millet under water and saline stress associates. The experiment was carried out in a randomized block design, in a 4x3 factorial scheme, composed of four levels of water replacement, based on crop evapotranspiration (ETc): 25%.ETc, 50%.ETc, 75%.ETc and 100%.ETc and three levels of water salinity (0.03, 2.0 and 4.0 dS m-1). With 25%.ETc independent of salinity, all morphological characteristics of millet were affected, occurring death of plants in the initial growth phase. In the absence of salt and greater availability of water, greater plant growth occurred. With respect to salinity, there was a reduction in the increment of all variables evaluated, with the highest reduction at the highest saline level (4.0 dS m-1). Water and salt stresses, when associated, reduce the growth of millet, since concentrations above 2.0 dS m-1 and less than 50%.ETc compromise its full development, providing declines in yield.

ABSTRACT. Leaf area measurements are of the main parameters used in agronomic studies to evaluate plant growth. The current study used a non-destructive method based on linear leaf dimensions (length and width) to select the regression model to estimate millet (Pennisetum glaucum) leaf area. For two millet genotype (IPA BULK 1 BF and ADR 300) 128 randomly-chosen leaves were measured at different vegetative growth stages. Measures of length and width of each leaf were made using digital calipers. Leaf area was measured using a gravimetric method. The best-fit leaf area estimation model was selected via linear, potential and gamma regression models. Leaf area values varied from 3.02 to 209.21 cm2. The average value was 95.31 cm2. The potential regression model exhibited lower residual sum of squares and Akaike's information criterion and similar determination coefficient and Willmott index. Thus, potential regression was more efficient in explaining the leaf area of millet, independent of the genotype, when compared to other models evaluated in this research. Length (L) and width (W) could be used in the following potential regression model Y ̂ = 0.879 L W 0.971to estimate millet leaf blade.