linear multivariate regression

Erodibility, which is determined by the soil's intrinsic features, is one of the most important elements in soil erosion. This factor reflects how sensitive the particles of a particular soil are to separation and transmission by erosion causes, both quantitatively and qualitatively. For measuring soil loss, the Universal Soil Loss Equation (USLE) is very useful. Sources reveal that erodibility is influenced by a variety of physical and chemical features of soil. In several soil erosion and sedimentation models, such as USLE, RUSLE, and MUSLE, one of the essential parameters is erodibility, which is represented as K. Particle size distribution, organic matter, structure, and permeability all have a role. The goal of this research was to quantify the amount of erodibility (K) in dry and semi-arid soils, as well as the physicochemical parameters that influence it. Another purpose of this research is to develop a connection that uses principal component analysis (PCA) and linear multivariate regression to estimate the quantity of soil erodibility based on effective physicochemical parameters.

The research location is 20 kilometers from Yazd city, along the Yazd-Ardakan road, on the edge of the dunes facies, which includes bare, mantled, and covered pediments. Using the stratified random sampling approach, soil samples were gathered to a depth of 10 cm within the facies in this study. The size and form of aggregates, as well as water penetration in the soil, were used to calculate soil structure codes using Wischmeier and Smith's tables. In the desert, soil permeability was assessed using double cylinders based on the ultimate infiltration rate. The hydrometer technique was used to determine the spread of soil granulation. Wet sieving and the Walkley and black methods were used to assess the proportion of extremely fine sand and organic matter, respectively. Lime was calculated by multiplying the volume of the hydrochloric acid neutralization reaction by the quantity of neutralizing agents. Statistical indicators such as mean, minimum, maximum, and standard deviation were derived at this step after computing the soil erodibility index. Principal component analysis was performed using SPSS17.0 software, and the linear multivariate regression model was utilized to predict soil erodibility index. After selecting significant components, linear multivariate regression between these components and soil erodibility was conducted concurrently. The coefficient of determination was used to assess the equation's accuracy in this investigation (R2).

The findings of the physical and chemical features of soil study revealed that the texture of the soil is mostly light sandy to loamy, with low organic content and calcareous. In terms of structural form, the analyzed soils were extremely fine granular and spongy, and their structural code was based on USLE (2 and 1). The permeability of the soil profile was high to extremely high (18.4 cm/h), and it was often in Class 1, 2, and in some instances Class 3 according to USLE. In the three naked, mantled, and covered pediments, the estimated erodibility indexes based on Wischmeier and Smith regression relationships were 0.0385, 0.03, and 0.0199 ton.hr/MJ.mm, respectively. According to the particular values acquired from the parameters and the percentage of variance, the top three components may be picked as the major component using principal component analysis. The first, second, and third components have correlation values of 0.88, -0.04, and 0.41, respectively, with the soil erodibility index. As a result, the first component has a stronger relationship with the soil erodibility index than the second and third ones. The percentage of sand and silt, soil permeability, and percentage of clay have a higher correlation with the soil erodibility index, respectively, and the correlation of other factors (organic matter, gravel, fine sand, and lime) is low in this component, according to the values for the given loading period. The amount of sand in the soil and its permeability are negatively correlated; whereas, the percentage of silt and clay in the soil is positively correlated. The maximum load is connected to the variables of gravel and lime in the second component, and it is related to organic matter and extremely fine sand in the third component. The effect of characteristics on soil erodibility is significant (0.001>p) and its coefficient of determination (R2) is 0.88 percent, according to an investigation of the relationship between soil erodibility and principal component values obtained from PC1, PC2, and PC3 using a linear multivariate regression model.

 The quantity of erodibility in dry and semi-arid soils, as well as the physicochemical parameters that impact it, were investigated in this research. Using principal component analysis and linear multivariate regression, a link was found to estimate the quantity of soil erodibility based on the effective physicochemical parameters. Because of the high amount of sand in the region's soils, these soils are readily separated due to poor adhesion, but because they contain bigger particles, they resist runoff and hence create less sediment. This barrier to transfer reduces as the quantity of clay and silt in the soil increases, and consequently more sediment is transported. Furthermore, a considerable quantity of sand improves soil permeability and reduces runoff. However, when the amount of silt and clay in the soil increases as a result of surface sealing, permeability reduces and greater runoff occurs. Soil erodibility is additionally influenced by organic content, lime, gravel, and permeability. Lime has a negligible influence on soil erodibility since it contains calcium cation, which increases particle homogeneity and hence increases soil resilience to rain drops. Organic matter has a negative relationship with soil erodibility as well. The breakdown of aggregates is slowed by increasing the quantity of organic matter in the soil. As a result, as organic matter levels rise, the rate of aggregate decomposition in a particular soil falls by one-third. Similar research studies in other semi-arid and arid soils in Iran are required to provide a more reliable connection for forecasting erodibility of soils in semi-arid and arid locations.

Landslide as a process of change in the stress-strain state of a slope occurs under the influence of natural and human parameters leading mass separation and its movement to down slopes. However, the relationship between the sliding mass and the slope remains constant. Accordingly, the mechanism of formation and development of a landslide is a systematic sequence of changes in the stress-tension state of a slope influenced by natural and anthropogenic parameters. This event is destroying human settlements and infrastructures and causing financial losses and many deaths around the world annually. The rapid population growth in the last half-century, the expansion of settlements towards steep mountainous areas on the one hand, and the false human being intervention in the destruction and changes of slopes, on the other hand, increased the frequency of landslides and this has led to an increase in damages. Iran has favorable natural conditions for a wide range of landslides with mainly mountainous topography, high tectonic and seismic activity as well as diverse climatic and geological conditions. Therefore, landslide studies on understanding factors and parameters affecting it, and identifying high risk and vulnerable areas in the world as well as in Iran have received serious attention. This research mainly aims to investigate the parameters affecting the landslide in the Vahregan catchment which located in the Sanandaj-Sirjan construction zone. Where metamorphic rocks, marl and shale, as well as wide area of quaternary sediments, have provided very favorable conditions for landslide occurrence.

Multiple linear regression method was used to perform this research. Thus, the scatter map of the landslides of the region as dependent variable and twelve factors includes elevation, slope, slope direction, lithology, fault, precipitation, drainage network, road,
land use, vegetation, TWI and SPI as independent variables were considered. To prepare landslide distribution map of the study area, aerial photographs of 1994 with a scale of 1: 40,000 were used and interpreted. Accordingly, the landslides area and their location in Google Earth software were determined. Then, 138 landslides occurred in the Vahregan catchment were determined with field studies, with the help of available maps and information, and the use of GPS system. It was then mapped using GIS software. After converting all the factors to information layers in GIS, these layers were adapted to the scattering map of the landslides of the region and were calculated the percentage of region located within the landslide area for all factors.

The results showed that the most effective factors in Vahregan catchment landslides based on multivariate regression method are distance from road, lithology, precipitation, land use, slope direction, distance from drainages, distance from faults, SPI drought, elevation, slope and TWI, with coefficient of 0.851, respectively. Their coefficient of R is 0.851 which is acceptable. The results showed that although natural factors can alone cause landslides, human factors are currently the most important parameters in causing landslides in the study area. Accordingly, most new landslides occur in close proximity to roads. In other words, it can be said that the downstream cutting of slopes by human being has increased the frequency and magnitude of landslides. Therefore, results showed that the road with 0.411 standard coefficient was the most important factor in creating landslide so that much of the landslide has occurred within less than 3 km of roads. Then, the natural factors includes lithology and precipitation with a standard coefficient of 0.362 and 0.299 and land use with a standard coefficient of 0.286 played the most role. However, vegetation factor and the TWI index with a standard coefficient of 0.103 and 0.127, played the lowest role in the landslides of the Vahregan catchment.  According to the final landslide zoning map, more than 50% of the area has located in a high risk area.

The study area has great potential for landslides in terms of natural features such as lithology, precipitation, elevation, Permanent River, and slope. The landslide map with 382 landslides indicates this. However, in the last two to three decades, environmental changes such as drought and consequently changes in vegetation covers on the one hand, and false human intervention, including the construction of multiple roads and the geometrical change of slope, on the other hand, have increased the frequency and magnitude of landslides in the studied area. The results of the final mapping showed that more than 50% of the basin is in high and very high risk areas. Accordingly, special attention should be paid to the extent of landslide risk and its threat in all human activities, especially environmental planning and management.