Soil surface temperature, measured on a 36-by-64 regular grid in a 6-hectare field, is modeled using two-dimensional spectral analysis and a signal-detection procedure. Temperature is found to behave randomly in one direction of the field (width), whereas the distribution in the other direction can be described as a superposition of four cyclic functions.

Investigation of the soil temperature’s cyclic behavior revealed that the salt content of the water applied earlier was related spatially to temperature. Different salinities of water had been applied to the field in a cyclic fashion, and co-spectral analysis yielded high correlations at certain frequencies, with higher salinities corresponding to lower temperatures and vice versa.