In this study, we have developed a new approach to define features of wet and dry seasons, using the linear splines (LSI) model. A set of formulas have been elaborated including our "gamma-index". This last, oscillates sensibly between 0 and 1 depending on observational data of stations. The wettest stations have the highest indices while the arid lands have the lowest ones. Moreover, two distinct types of dry seasons are described in this work. The first one is the most conventional type induced by high temperature (HDS: Hot-dry season). However, the second one, is mainly linked with a sustained lack of liquid water because of low temperatures (CDS: Cold-dry season), inducing a physiological dryness of species. Using machine learning algorithms, the LSI model also was able to accurately (90%) predict the different climate classes defined by Ko center dot ppen. The obtained results were very coherent to reality and to Ko center dot ppen-Geiger climate classifications. Using this index, we were also able to map out the most arid areas in Algeria and those that are less. In cold regions, the calculated gamma-indices reveal low values, very similar to those recorded in arid regions. The results have shown also a gradual decrease when approaching poles or tops of high mountains revealing, thus, increasing dryness.