SHELTER AND REMOTELY SENSED NIGHT TEMPERATURES IN ORANGE GROVES

In previous papers we have used a linear regression approach for determining nocturnal air temperature in orange groves from satellite thermal data. However, this procedure has a poor precision (almost-equal-to 2-degrees-C) for applications such as frost forecasting. For this reason a theoretical method has been proposed, which is based on the following assumptions: (1) the air temperature (T(a)) is the result of the convective heat exchange between ground and air, and between air and orange trees, and (2) the remotely-sensed temperature (T) can be expressed as a function of ground (T(g)) and orange tree (T(s)) temperatures. So the relationship T = T(a) + (a(g) - alpha) (T(g) - T(s)) has been derived, where a = (1 + h2-pi-R/h1L)-1 and a(g) = (epsilon-g/epsilon) [P(g) + (1 - epsilon-0)G'P(s)]; h1 is the convective heat transfer coefficient between ground and air, h2 is the convective heat transfer coefficient between air and orange tree, R is the orange tree radius, L is the distance between two orange tree trunks, epsilon-g and epsilon-0 are the emissivities of the ground and of the orange tree, epsilon is the effective emissivity, P(g) and P(s) are the proportions of ground and side of the orange tree observed by the sensor, and G' is the shape factor ground-side. Two experiments were carried out in order to validate this model, in which air temperature was measured by means of a mercury thermometer positioned at 1.5 m above the ground and in the middle of two orange tree rows. The temperature of the orange tree and the ground was measured with a radiometer, and the temperature of the orange grove was obtained by means of a manual scanning system. Ground and orange tree emissivities were measured using the box method. We have analyzed the dependence of the T - T(a) relationship on weather conditions, field architecture and viewing angle, and we can conclude that if these parameters are known, the shelter temperature can be obtained from satellite thermal data with a precision of 0.8-degrees-C.