The earth surface temperature is an important parameter in atmospheric phenomena and processes occuring in the surface boundary layer. It involves in the surface energy balance, evaporation, evapotranspiration and desertification processes and is also considered as an indicator of environmental degration and climate change. For this study the data were collected from the micrometeorological tower of 10 meters height at the experimental site located in the semi-arid region of the Institute-basin of Federal University of Campina Grande in the city of São João do Cariri-PB N-E Brazil. The data of velocity, temperature, specific humidity, net radiation, surface temperature and pressure were collected in the interval of 20 minutes from 28-02-2001 to 09-03-2001 and 07-11-2001 to 16-11-2001, the rainy and the dry periods respectively. In this study the inverse theory is applied for the estimation of the surface temperature by solving the surface energy balance equation using the forms used in the limited area regional model for obtaining turbulent heat and momentum fluxes in the surface layer. The Levenberg-Marquardt algorithm is used to retrieve values of radiation parameters such as albedo and emissivity (Paz et al., 2003) and also the parameters associated with the turbulent diffusion as surface roughness length and the wind (considered as a parameter for estimation of bulk Richardson number). It is observed that there are significant differences between the calculated (predicted) and measured values of the surface temperature. For this a sensitivity analysis and the estimation of parameters are used applying the minimization technique of Levenberg-Marquardt (Press et al., 1992). Also the measurements errors of the data acquisition system are considered according to the standard deviation furnished by manufacture. Then it is found that the predicted and the measured values of the surface temperature are quite close to each other for both the dry and rainy seasons.