This study was carried out to predict some surface soil properties and mapping units of some bare desert soil throught the analysis of satellite digital data. Spatial variability of some surface soil properties was studied using coefficients of variation and multivariate analysis were used to predict all the studied surface soil properties. Geographical information system processes were used to map the original and predicted soil attributes variability and mapping units.     The image processing showed that the area can be classified to bare soil (86.4 ٪) and vegetated area (13.6 ٪). The vegetated area has three levels of natural vegetation density: high, moderate and low. The supervised and unsupervised classifications using spectral signature techniques showed that the bare soil area can be subdivided to five classes with the same shapes and different attitudes. The statistical analysis have been done for seven different combination of data sets depending on the number of the surface samples of augers, profiles and sectors with presence and absent of natural vegetation. The descriptive statistical analysis showed that the highly variable soil properties (CV > 60 ٪) on the area are CaCO₃ ٪ and EC and the rest soil properties has medium effect (CV between 10 and 60 ٪) for 80 and 68 data sets. While, for 20 and 17 data sets, the very high soil variability caused by CaCO₃ ٪, infiltration rate, EC, gravel ٪ and hydraulic conductively. In the same time, the grain size distribution has medium effect. All color components has medium effect even under dry or wet conditions with CV values between 10 and 60%. The digital numbers for all data sets has very low effect (CV<10 ٪) on the soil variability except for bands 2, 3, and 7 for 80 surface samples and for all bands for 20 sample size, which has CV between 10 and 20٪       The multivariate regression analysis have been successfully applied to all soil properties using extracted image digital numbers. The R values for CaCO₃ ٪ and EC ranged between 0.705 to 0.867 and 0.748 to 0.781, respectively. For sample size 80 and 68, the R values for all predicted available nutrients, cation exchange capacity and organic matter were more than 0.953.  The predicted color components under dry and wet conditions had R values more than 0.910 for 20 sectors, and 20 and 17 profiles data sets as well as sand, silt and clay, while for gravel % was 0.803. Infiltration rate has the least R values ( 0.757) and hydraulic conductivity had more than 0.904 value. The Spatial variability of the original and predicted values of CaCO₃ ٪  and EC from image digital numbers have been mapped for comparison. The average accuracy for CaCO₃ ٪  and EC were 91.45٪ and 88.70 ٪, respectively, while their mapping units has 87.88 % average accuracy. This results showed the high potentials of using digital numbers of satellite image for predicting soil properties and mapping units.