Faced with the scarcity of surface water accentuated by climate change, particularly in many arid and semi-arid countries, the quality of groundwater used for irrigation is a concern to agronomists and hydrogeologists. When these waters are of deep origin, they may have high mineralization and chemical compositions unsuitable for irrigation; in particular, they may alter soils and crops. It is therefore important to optimize the spatial estimation of the salinity of these waters and contribute to better knowledge of their quality, through an adapted and robust statistical and geostatistical approach. In the case of north-eastern Algeria, the objective of this study is to characterize the quality of deep waters and to test two interpolation methods (Inverse distance weight and ordinary Kriging) of their electrical conductivity (EC) as an indicator of their salinity and of the risk of damaging irrigated soils. 51 groundwater samples were taken in this region where there are many thermal springs, the water of which is used for irrigation and often is highly mineralized (EC between 0.6 and 26.6 dS/m). The geology is composed of karstic rocks crossed by large faults that allow deep water to rise. Based on major elements contents, analysis of the hydrochemical facies of these waters shows that the main facies are hyperchlorinated sodium (38%) and sulfated calcium (32%). The RSC (Residual Sodium Carbonate) and SAR (Irrigation water salt) indexes were used to assess the water quality. The results indicate that the majority of the sampled groundwater present a risk for soils irrigated with these waters (almost 1/3 presents a strong risk). The risk for the soils seems to be explained by the positive value of the residual alkalinity, and the high risks of sodization and alkalinization. The geostatistical analysis reveals strong heterogeneity in electrical conductivity (salinity). The maps based on this analysis allow the identification of risk areas. The comparison of Inverse distance weight and ordinary Kriging methods indicates similar results being obtained through both methods. However, ordinary kriging appears to be more accurate, less biased, and seemingly better represents the organization of the groundwater resources, as NE-SW alignments are visible. With the proposed approach, it is possible to calculate the surface areas of the different salinity thresholds and then optimize the use of deep groundwater for irrigation