- Abstract Spatial and temporal variability in soil heat flux ( G ) under sparse/clumped vegetation conditions is significant and has been studied. However, little attention has been devoted to evaluating the variability of G with respect to row crops, particularly with respect to row direction. The variation in G for row crops is related to the effect of differential shading of the soil surface, which is dependent on plant architecture, row spacing, and row direction. This paper reports the effect of row direction and sensor position on G magnitude and variability in an irrigated row crop of cotton. In addition, the effect of errors in water content estimation on the heat storage in the uppermost soil layer is assessed. The research was conducted in the Southern High Plains of the USA, as part of the Bushland Evapotranspiration and Agricultural Remote Sensing Experiment of 2008 (BEAREX08). Measurements were concentrated in two irrigated cotton fields, one with north–south (N–S) and the other with east–west (E–W) row directions, with ten sets of sensors in each field. Row direction had an effect on both the temporal dynamics and the total daily G . Important short-term (15-min average) variability in G at the various positions in the interrow was observed under partial canopy cover conditions for the N–S row direction, while the daily sum of G (Σ G ) in both row directions was similar. In the beginning and the end of the growing season, Σ G was larger in the N–S direction field. In the E-W direction field, strategically located 3-replicate sensor sets (as are often deployed at flux tower installations) were found to adequately describe the 10-sensor average G , with errors as small as 6% and with a transitory maximum error of 12%. In the N-S row direction field, however, no 3-position combination was adequate to represent the 10-sensor average G .