Calibração de dispositivo TDR (Time Domain Reflectometry) para estimativa de umidade volumétrica em latossolo

Abstract

Soil mineralogy strongly influences the determination of soil moisture with the use of the TDR (Time Domain Reflectometry) technique. This study sought to establish ma- thematical regression equations between the known (real) moisture and the dielectric constant (Ka) and to compare them with models considered "universal", as well as to evaluate its efficiency for a Hapludox. The soil samples were collected in depth from 0 to 30 cm, sieved in 2 mm mesh and arranged in PVC cylinders 30 cm high and 10 cm in diameter. Nine samples were saturated by capillary rise, one left in the moisture condition of TFSA and one cylinder containing only water (boundary condition). The readings were performed when the samples reached saturation, followed by the sam- ples reaching the permanent wilting point. For the others (water only and TFSA), only one reading was performed. The readings were performed with 3-stem probet each reading the samples were weighed, obtaining the soil mass moist. After the readings were finalized, the soil contained in the cylinders was greenhouse dried, obtaining the mass of dry soil. Linear, polynomial and logarithmic models were adjusted between real moisture and dielectric constant for three situations, not using and using the boun- dary condition and in specific real humidity range (<0.4 cm 3 cm -3 ). The precision of the models was evaluated by the coefficient of determination (R2), by line 1: 1, by the standard error of the estimate, the mean square error, the mean error and the root mean square error. The models considered "universal" were not accurate enough to estimate the soil moisture content. The adoption of the contour condition only provided better accuracy for water content> 0.40 cm 3 cm -3 . The logarithmic equation, for the specific moisture range (<0.40 cm 3 cm -3 ), showed a high accuracy. The linear CC, linear SC, and logarithmic equations showed better efficiency than the others. The li- near CC, linear SC, and logarithmic equations showed better efficiency than the others. The equation of determination of θ E that presented the best accuracy and optimal pre- cision was: f(x) = 0.1681 Ln (x) -0.1629, when Ka <= 34.7.