In the process of high temperature field reconstruction, it still need through a plurality of single point temperature measurement combined with spatial interpolation methods appropriate to reconstruct the temperature distribution of the whole temperature field. Taking welding temperature field as an example, the spatial variability and distribution of temperature field were analyzed and forecasted by sample points through OK, IDW and spline methods, respectively. The results show that different interpolation methods do not significantly influence on reconstruction accuracy and the amount of sample points significantly affect the prediction accuracy of the spatial distribution of temperature field. In the temperature field evaluation, 25 sample points are more suitable for the sample number. Comparing with the sample number of our results with the formula, it was found that calculation sample number by optimum formula is significantly lower than our results. It show that if we do not consider the spatial variation and only use the formula will lead to uncertainty in temperature field reconstruction.