Table 1 shows seven factors with eigenvalues greater than 1 which collectively express about 79% of the variance in the original data. It signifies that the 19 measured soil quality indicators were relevant but that seven factors (boldface factors in Table 1) efficiently represent the whole changes in soil condition or quality. Factor1 shows the highest variance (13.01%) in the results while Factor2 to 7 reveal 12.63%, 11.91%, 11.83%, 11.56%, 10.85% and 7.15% of the variance, respectively (Table 1). Factor1 included SOC, OC and C/N with positive loading and other parameters with negative loading. These parameters imply that Factor1 was mainly associated with the highest loadings of the chemical properties SOC (0.881) and OC (0.875). Factor2 contains N (chemical property) and SP (physical property) which reveal the highest loading (0.886) and positive loading (0.758), respectively. The microbial respiration and population were only loaded heavily and positively (0.94) on Factor3. The forth Factor is related to soil texture with positive loading for silt (0.882) and negative loading for sand (-0.864). The most important attributes in Factor5 were in relevance to fertility and include CaCO3 (0.802), CEC (0.740) and K (0.711). Factor6 contains physical properties like AS (0.865) and Clay (0.716). BD (0.768) and EC (0.689) appear in Factor 7. The commonality value demonstrates the relative importance or the residual variance of each soil property upon its contribution to all the extracted factors (Brejda et al. 2000). In addition, the commonality values for most variables are above 0.70, except for pH (0.582), EC (0.628) and ?m (0.516). The vector loading plot illustrates soil variables with the greatest effect on differentiating or grouping treatment combinations (Fig. 2). Different land use made large differences in soil conditions, especially soil microbial communities. The factor analysis, which contained all the soil attributes as input variables, separated chemical and microbial activities more along Factor1, 2 and 3 and less along with other Factors.
Factor analysis and other data reduction techniques such as principal component analysis affect soil parameters in separating land use and affect soil quality in the assessment of the land use. They have been formerly applied successfully at field scale using different soil properties (Sena et al. 2002; Imaz et al. 2010). In addition, another group of helpful techniques to select the most important properties are multivariate statistical methods that consider the highest variability in the total dataset from large available data in order to get more information from soil data (Wander & Bollero, 1999; Shukla et al. 2006; Bastida et al. 2006) and, finally, to specify land use influences on soil conditions (Wander & Bollero 1999; Imaz et al. 2010).
The highly weighted variables under Factor1 (SOC, OC and C/N) were all significantly (P < 0.01) correlated with each other (Table 2). Consequently, only SOC with the highest loading (0.881) was included in the MDS. Under Factor2, soil N and SP were highly correlated (r= 0.70), with N showing the highest factor loading (0.886). Microbial respiration and population on Factor3, which accounted for more than 11% of the total variability (r= 1.0, p < 0.01) (Table 2), were considered as an important indicator of the MDS because of their significant correlation. This significant correlation reflects that microbial respiration and population could be redundant for inclusion in the MDS. Silt with 0.882 positive loading and sand with negative loading of -0.864 on Factor4 had a strong correlation (r= -0.83). Under Factor5 attributes of CaCO3 (0.802), CEC (0.74) and K (0.711) had a correlation of 0.46 and 0.37, respectively. AS (0.865) and Clay (0.716) on Factor6 had a correlation of 0.53 and, finally, under Factor7 BD (0.768) and EC (0.689) were weakly correlated (0.18). Only SOC, N, microbial respiration and population, silt, CaCO3, AS and BD from Factors 1-7 were chosen as the smallest set of potential indicators for coverage in the MDS to create an overall SQI. Hence, it could be deduced that geometric means of soil microbial attributes, which are alike in nature and strongly interconnected, would be suitable to sum up the biological properties of soil for selecting MDS indicators. In addition, there was also a very significant correlation (r = 0.83, p < 0.01) between SOC with a high loading (0.881) and OC with the highest loading (o.875) on Factor1, confirming that OC can also be used as an indicator of soil quality. Also in Factor2 the correlation between N with a high loading (0.886) and SP with a loading of 0.758 was significant (r= 0.7, p< 0.01). Both attributes in Factor3 including microbial respiration and population with the loading of 0.94 had a strong correlation (r= 1, p< 0.01) which show one of them can be used for soil quality when the other is not measured.
Generally, the current study shows that SOC, N, microbial respiration and population, silt, CaCO3, AS and BD are crucial soil properties that should be counted as soil quality indicators under different land uses. In fact, the MDS contains overall microbial activity for the management goal of crop productivity, environmental quality and the regulation of greenhouse gases such as C sequestration. Based on the key role that SOM (C and N) and its labile fraction plays in defining the quality of soil (Doran & Parkin 1994; Gregorichet et al. 1997), these parameters were included in the MDS due to high loading coefficients (0.875 and 0.886 respectively) which were within about 13% and 12% of the highest loading in Factors 1 and 2 (Table 1). Soil ?m, pH and P as physical and chemical properties were less sensitive to land use (Table 1) and, thus, were not included in the MDS.
Although all five indicators of SOC, N, AS, microbial respiration and population were used as “more is better” functions due to their positive influence on crop productivity (i.e., the management goal), therefore on soil quality, BD was used as "less is better" function, because of negative effect on soil properties. Moreover, silt and CaCO3 had both positive and negative effects on soil quality and, based on their “optimum” threshold value, the “more is better” curve type was applied to them.
A closer inspection of the contribution of each MDS indicator to the linear and non-linear SQI confirms that microbial population had the highest contribution to the linear SQI (17.84% in the crop land and 16.93% in the planted area) with a concomitant increase of contribution toward the non-linear SQI by 18.16% and 17.33% in the crop and planted lands, respectively (Figures 3 & 4). The next significant contribution in linear SQI belongs to SOC which shows the highest percentage of variance in the SQI (13%) followed by N (12.6%) and microbial respiration and population (11.9%) (Table 3). This clearly shows the effect of the weighting factors attributed through factor analysis. A high weighting for SOC (0.130) indicates that this variable had the highest variance in the minimum data set (Table 1). The L-SQI shows that physical attributes had significantly higher values than chemical and biological attributes and varied from the maximum value of 35.17% in the farm and 38.74% in the garden (Fig. 5). Low values of SQI were found in a biological attribute in both crop (31.61%) and plant (26.17%) land uses (Fig. 5). All of the attributes in L-SQI had a significant difference in both land uses.
The NL-SQI shows that biological SQI had the highest contribution in the crop (73.94%) and plant (64.15%) land uses and low values of SQI were related to physical attributes (Fig. 5). In both L-SQI and NL-SQI, the biological attributes in crop land were higher than in planted land, which is because of tillage, other activities and an increase in substrate availability for microbial utilization as well as better soil conditions for microbial growth and activity. In contrast, physical and chemical attributes were higher in planted lands than in crop lands (Fig. 5).
As the physical and chemical properties alter only when soil undergoes a drastic long-term change, they are of less use than biological and biochemical parameters (Filip 2002). Contrarily, biological and biochemical properties are more sensitive to small changes that the soils can undergo in the presence of any disturbing action and soil management (Franchini et al. 2007). Furthermore, the relationship between soil quality and crop production is largely controlled by the temporal and spatial interaction of several other factors such as climate, soil type, crop varieties and other soil attributes not considered in this study. Also, Erkossa et al. (2007) and Armenise et al. (2013) did not report a correlation between SQI and crop performance under different management and agricultural practices. Therefore, a functional relationship between plant production and SQI needs to be fixed over a longer period and, in particular, by using other soil quality attributes that might have a stronger influence on crop yield.
4. Summary and conclusions
In this study, land use (crops and plants) had remarkable effects on the population and function of the microbial community of soil. SOC, N and Microbial activities were the most significant soil variables that efficiently differentiated the impacts of land use on soil quality in the study area and confirmed our hypothesis. All in all, the most sensitive indicators of land use were the microbial and biochemical attributes of soil which made a great contribution to SQI. In both linear and non-linear SQI, a higher biological quality of the soil was observed in crop land in comparison with planted land, which confirms that implementation of tillage practices could improve the microbial activity and quality of the soil. The microbial activities of soil were more efficient in NL-SQI and consistent indicators of land use changes in soil quality than other soil variables; hence, they should be taken into account when assessing the effect of land use on soil quality in the study area. Accordingly, a functional relationship between plant performance and SQI should be considered in future studies. Evaluation of soil quality is a promising tool for observing and differentiating land use to improve soil management practices and increase the sustainability of dominant soil in the northwest of Iran. In addition, assessing soil quality would be useful for evaluating the impacts of different land use on soil performance and functions at a local scale. However, quantifying the value of changes in soil quality and crop production at a regional scale is serious for perceiving how land use can be better managed to improve soil quality and increase the long-term productivity of the soil.