Spatial Variability of Soil Properties in Palm Groves of the Central Algerian Sahara (Case of Zelfana)

. Monitoring soil quality in irrigated areas is essential for assessing the sustainability of production systems. In this respect, the spatial variability of the properties of irrigated soils is a mean to know the evolution of the latter. This study aims to determine the spatial variability of soil organic matter, salinity, pH and active limestone using a geostatistical approach. The present study was carried out in the region of Zelfana located in the Algerian central Sahara, the samples were collected from a depth of 0 to 30 cm and analysed for organic matter, salinity, pH and active limestone. The analytical results show that the soil is very poor in organic matter, very salty, alkaline to very alkaline and moderately calcareous. The geostatistical analysis revealed various patterns and levels of spatial distribution of the studied properties. The results showed a weak spatial dependence for organic matter, moderate for pH and salinity and strong for active limestone. The variographic analysis showed that the nugget effect is weak for organic matter and pH, moderate for salinity, while active limestone does not show a nugget effect. The range varies from 75 meters for salinity to 299 meters for organic matter, confirming the validity of the adopted sampling and allowing the optimization of future sampling plans. The Arcgis autoKriging function was used to select the best theoretical variogram model from those most commonly used in geostatistics (Gaussian, spherical, exponential and circular). This model was used to produce the spatial variability maps using ordinary kriging. Spatial variability of soil properties is influenced by agricultural intensification, something that must be taken into consideration for integrated and sustainable land management in similar regions.


 C. Statistical Analysis
The results were subjected to a descriptive analysis in which, the various descriptive parameters were calculated. The Principal Component Analysis (PCA) was also established for the quantitative variables using R software (v3.5.3).
The spatial data was interpolated according to the kriging method using the "AUTOKRIGE" function of the "AUTOMAP" package under R v3.5.3. In fact, the "AUTOKRIGE" function aims to choose the best theoretical model of a variogram among those most used in geostatistics (Gaussian, spherical, exponential and circular) [12,13]. The geostatistical models generated were subsequently run on ArcGIS v10.3 to obtain thematic variability maps for each parameter studied. The coefficient of variation (intensity of variability) is interpreted according to the scale of Nolin et al. (1997): CV (%) = 100.σ/m, including: CV<15% low, 15%<CV<35% moderate, 35%<CV<50% high, 50%<CV<100 very high.

III. RESULTS AND DISCUSSIONS
The results related to the study of soil parameters in the palm grove of Zelfana are summarized in the following table.  Table 1, shows that the studied soil has an organic matter content between 0.13 and 0.45 % with an average of 0.23±0.09 %, which reflects a poverty of this soil in organic matter. The pH value is high in the order of 8.27±0.15, indicating a very alkaline reaction of the studied soil. While, electrical conductivity (EC) presents a maximum of 2.61 dS/m which classifies the studied soil as salty to very salty. The average value of active limestone is 6.59 ± 1.8 % indicates that the Zelfana palm grove has a moderately calcareous soil.
The coefficient of variation for the parameters studied shows little spatial variation in organic matter (43.54%), electrical conductivity (20.66%) and active limestone (27.37%). However, the pH values seem homogeneous (1.88%).
In order to analyze the spatial dependence of these parameters, we used the Variogram tool with the "AUTOKRIGE" function for a better choice of the theoretical model. The obtained results are detailed in the table 2 and graphically represented by figure 3. The autoKrige function showed that the most reliable model to represent the organic matter and electrical conductivity data is the Gaussian model. The spherical model is the most reliable model for pH and the exponential model for active limestone.  Based on the best variogram models obtained, the spatial variation maps of different studied parameters are presented in Figure 4.   According to figure 5, the contribution rate of axis 1 is 48.82%, that of axis 2 is 24,75, reflecting a good representation of the variables studied on the factorial plan with 73.57% of total inertia. Axis 1 represents a positive correlation between the parameters MO, EC and negative with pH. Axis 2 shows a negative correlation of the pH factor with the active limestone factor. In general, we found that the two parameters MO and EC are negatively correlated with the other two parameters (pH and active CaCO 3 ).

IV. DISCUSSION
Soil analyses of the palm grove of Zelfana (Central Algeria) revealed that the soil is very poor in organic matter with an average content of 0.23±0.09%. Several authors have reported that the Algerian Sahara soils are known to be, as a whole, poor in organic matter which is often less than 0.1% [14,15,6,16].
This poverty in OM can be justified, on one hand, by the extreme climatic conditions of the Saharan regions, in particular temperature and rainfall [17,18]. On the other hand, agricultural intensification including irrigation can be considered as an incubator of optimal conditions (humidity and temperature) for the degradation of soil organic matter in Mediterranean regions [19,20].
The results of soil pH analysis in the study area show that the soil is alkaline to very alkaline with an average pH value of about 8.27±0. 15. The values recorded are within the range for soils in arid regions, from a pH slightly below 7 to a pH of around 9 [21][22][23][24]6,25,4,26]. Inside the soil, when roots take up mineral elements in the state of anions, such as nitrate NO -3 they releases OHwhich induces an alkalinization of the soil [27].
The results of the multiple correlation study showed no significant correlation between the different parameters studied (P> 0.05). Nevertheless, the most important positive correlation was noted between organic matter (OM) and electrical conductivity (EC) (R=0.43). The electrical conductivity increases with the contribution of organic matter, the organic matter increased the salinity of the soil through the mineralization of these organic compounds [39].
Other negative correlations were similarly recorded between pH and active limestone (R=-0.45), on one hand, and between pH and electrical conductivity on the other (R=-0.3751). The pH increases when the calcium concentration in the solution decreases [40]. The decrease in pH is related to the presence of limestone in the soil [41]. For the electrical conductivity, this result is close to the work of [42], the study of the relationship between pH and EC shows a negative correlation.
The variographic analysis revealed a low nugget effect, which means that the variation of organic matter, pH and salinity at distances less than the sampling step (150 m), the nugget effect of active limestone is (0) which signified the reliability of sampling step. Indeed, the nugget effect can be defined as an indicator of continuity at close distances [43]. A significant nugget effect requires additional sampling of properties at smaller distances in order to detect spatial dependence [44]. Spatial dependence is determined by the ratio between the nugget and the bearing and is expressed as a percentage [45]. The spatial dependence of organic matter is low with a value of 100%, salinity and pH are moderately spatially dependent, the nugget to step ratio values are 44% and 66.67% respectively. Active limestone is in the class (<25%), which is consistent with strong spatial dependence ( Table 2).
After selecting the best performing variogram models, spatial variability maps were made by ordinary kriging. These maps show that the highest values are recorded as follows: organic matter in the east of the study area, pH in the southeast, salinity in the west and active limestone in the northeast.

CONCLUSION
The study of spatial variability of soil properties in the Zelfana palm grove (central Algerian Sahara) was carried out by random sampling of fifteen (15) profiles. The results obtained revealed that the studied soil is very poor in organic matter and alkaline to very alkaline. We also recorded that the soil is very salty and moderately calcareous. The study of multiple correlations showed no significant correlation between the different parameters studied (P> 0.05). Nevertheless, the most important positive correlation was noted between OM and EC and other negative correlations were similarly recorded between pH and active CaCO 3 . The spatial variation of organic matter is high with a CV of 43.54%, moderate for electrical conductivity and active limestone with a CV of 20.66% and 27.37% respectively. While that of pH is low with a CV of 1.88%. The most reliable variogramm models of the studied parameters are the Gaussian model for organic matter and electrical conductivity, the spherical model for pH and the exponential model for active limestone. Organic matter is weakly space dependent, salinity and pH are moderately space dependent, and active limestone has strong space dependence. The study on spatial variability is interesting for a good understanding of the current situation of agricultural soils in order to better manage, maintain and improve their productivity.