Soil heat flux (G) is an important component of energy balance by constraining the available amount of latent heat and sensible heat. There are many methods and formulations in the literature to estimate G accurately. In this study, widely used G estimation models are chosen to test. The models are based on Spectral Vegetation Indices (SVIs) namely, Normalized Difference Vegetation Index (NDVI), and Soil Adjusted Vegetation Index (SAVI) together with leaf area index (LAI), and crop height. Two successive growing periods of winter wheat (Triticum Aestivum L.), sunflower (Helianthus annuus L.), and maize (Zea mays L.) fields, located in the northwest part of Turkey, are used. Midday values (average of 09:30- 13:30) of G and net radiation (R-n) used in order to capture the time period, when G is proven to be much dominant. According to the results, overall the best relation obtained with an exponential NDVI model with a determination coefficient value of 0.83 and a root mean square (RMS) error value of 20.28 Wm(-2) for maize. For winter wheat, G is predicted the best with SAVI based model (r(2) =0.74), and for sunflower, LAI based model worked best with 0.75 r(2) value. Crop height (CH) based nonlinear regression G model that suggested in this study worked better than linear models suggested in the literature with a better determination coefficient (r(2) =0.70) and a lower RMS error value (10.8 Wm(-2)).