In the last few years, the availability of new earth observation satellites, such as Sentinel 2, has increased the interest in their potential for precision agriculture. In a non-continuous crop, like grapevines, the influence of interrow space, and vine geometry can disturb the assessment of vegetation properties using satellites. In particular, vine height and row orientation can project shadows differently, according to the relative solar and sensor position. In this study, sun shading and sensor off-nadir effects were estimated for images acquired by three constellations of satellite (Sentinel 2, RapidEye and Pleiades) on 48 vineyards. Shading and off-nadir effects were calculated using satellite image metadata and vineyard row orientation. Then, multiple linear regression models, based on satellite information and shadows, were estimated using pure vine data extracted from images acquired by a multispectral camera carried by unmanned aerial vehicle as dependent variable. Multiple regression models were compared to simple regression models using different model's performance parameters (ρ Spearman, R2 adjusted, root mean square error and Akaike's Information Criteria). The results showed significant effects of sun and off-nadir sensors' effects on pure vine variability estimation using satellite images, with a stronger effect of sensor orientation than solar shading. In multiple regression models, R2 increased of 5.8, 6.6, and 13.4 % in Pleiades, RapidEye, and Sentinel 2, respectively, compared to simple linear regression models based just on the satellite sensor band data. Moreover, these multiple linear regression models have a better fitting and lower estimation errors than simple linear regressions.