Determining the terrain characteristics related to the surface expression of subsurface water pressurization in permafrost landscapes using susceptibility modelling

Abstract

Abstract. Warming of the Arctic in recent years has led to changes in the active layer and uppermost permafrost. In particular, thick active layer formation results in more frequent thaw of the ice-rich transient layer. This addition of moisture, as well as infiltration from late season precipitation, results in high pore-water pressures (PWPs) at the base of the active layer and can potentially result in landscape degradation. To predict areas that have the potential for subsurface pressurization, we use susceptibility maps generated using a generalized additive model (GAM). As model response variables, we used active layer detachments (ALDs) and mud ejections (MEs), both formed by high PWP conditions at the Cape Bounty Arctic Watershed Observatory, Melville Island, Canada. As explanatory variables, we used the terrain characteristics elevation, slope, distance to water, topographic position index (TPI), potential incoming solar radiation (PISR), distance to water, normalized difference vegetation index (NDVI; ME model only), geology, and topographic wetness index (TWI). ALDs and MEs were accurately modelled in terms of susceptibility to disturbance across the study area. The susceptibility models demonstrate that ALDs are most probable on hill slopes with gradual to steep slopes and relatively low PISR, whereas MEs are associated with higher elevation areas, lower slope angles, and areas relatively far from water. Based on these results, this method identifies areas that may be sensitive to high PWPs and helps improve our understanding of geomorphic sensitivity to permafrost degradation.

Spatial modelling of gully erosion using evidential belief function, logistic regression, and a new ensemble of evidential belief function–logistic regression algorithm

“…Alternatively, an example can be found in the literature where the variance inflation factor (VIF) is used to recognise the presence of multicollinearity in data (Cama et al, ). Here, we opted for VIF, which is the reciprocal of tolerance (TOL), and calculated as 1 − R 2 , where R 2 is obtained by regressing each variable with respect to the remaining variables in the multivariate regression (Holloway, Rudy, Lamoureux, & Treitz, ). A TOL ≤ 0.1 or a VIF ≥ 10 indicates serious multicollinearity (Du, Zhang, Iqbal, Yang, & Yao, ).…”

Machine Learning-Based Gully Erosion Susceptibility Mapping: A Case Study of Eastern India

“…The multi-collinearity was tested using the tolerance (TOL) and variance inflation factor (VIF). The TOL was calculated using Equation 6, where R2 is obtained by the regression of each variable for the remaining variables in the multivariate regression [55]. In this study, the elevation, slope, rainfall, distance from river, distance from lineament, NDVI, TWI, SPI, and STI were used as numerical variables and reclassified into five sub-categories using the NBM in GIS.…”

“…The multi-collinearity was tested using the tolerance (TOL) and variance inflation factor (VIF). The TOL was calculated using Equation 6, where R2 is obtained by the regression of each variable for the remaining variables in the multivariate regression [55].…”

“…The WofE is an important bivariate statistical method, which calculates the relative importance of effective factors by statistical means using the log-linear form of the Bayesian probability model [57]. In this analysis, the WofE model was used to demonstrate the relationship between gully occurrence and gully conditioning factors [51] obtained using the regression of each variable [55].…”

Multi‐criteria decision analysis integrated with GIS to determine land suitability for cultivation and best harvest time of vine

“…For this purpose, the relationship between all parameters was first investigated using tolerance and VIF statistics. According to Table 6, VIF values are greater than 1 and tolerance is less than 1, showing a linear relationship between the data 61 . Arabameri et al 62 .…”