Effects of scale and data source in periglacial distribution modelling in a high arctic environment, western Svalbard

Abstract

The effects of scale (modelling resolution) and sources of data were explored in relation to periglacial distribution modelling for an area on western Svalbard in the High Arctic. To assess the effects of scale on predictive performance, the distributions of sorted circles and solifluction lobes were modelled at two resolutions (20 × 20 m and 200 × 200 m) using a boosted regression tree, a novel statistical ensemble method. To analyse the effects of sources of data on periglacial distribution modelling, a generalised linear model and a variation partitioning method were used. The explanatory variables were topographic parameters computed from a digital elevation model, vegetation and soil moisture indices derived from a Landsat TM 5 scene, and field survey‐based information on surficial materials. Firstly, similar levels of success were achieved in predicting the periglacial feature distributions at the local (20 m) and landscape (200 m) scales. Secondly, results indicated the potential for modelling to replace labour‐intensive field observations. The importance of topographic parameters for predicting the distribution of periglacial features in the sparsely vegetated High Arctic environment was also evident. Methodologically, novel statistical techniques and earth observation data provided an efficient combination for analysing periglacial landforms and processes in this remote region. Copyright © 2010 John Wiley & Sons, Ltd.

Spatio-temporal variability in remotely sensed land surface temperature, and its relationship with physiographic variables in the Russian Altay Mountains

“…Furthermore, collinearity effects were examined by calculating the Spearman's rank order correlation coefficient (e.g. Hjort et al, 2010) which showed no sign of an unacceptably high level of intercorrelation between the independent variables (all values <0.6). Finally, also the potential problem with spatial autocorrelation (i.e.…”

Integrating climate and local factors for geomorphological distribution models

“…Brenning et al , ; Hjort and Luoto, ), thus failing to consider the potentially important factors operating concurrently or the optimal resolution of the input data (e.g. Hjort et al , ; Etzelmüller, ; Potter, ).…”

Supervised classification of landforms in Arctic mountains

“…13,[15][16][17] These automated classification methods allow for larger areas to be mapped more quickly while reducing human error (although introducing machine error), and facilitating comparable results and model transferability. 13,18 At regional scales, topographic parameters derived from digital elevation models (DEMs) are shown to be one of the primary predictors of landforms in periglacial environments, 13,19,20 particularly in high-Arctic environments because of the low abundance of vegetation. 19 Statistical analyses of remotely sensed topographic, optical and/or climate data landform classifications 21 use a range of multivariate and simple statistical techniques including generalized linear methods such as linear discriminant analysis (LDA), 13,22 logistic regression, 17 and artificial neural networks.…”

“…13,18 At regional scales, topographic parameters derived from digital elevation models (DEMs) are shown to be one of the primary predictors of landforms in periglacial environments, 13,19,20 particularly in high-Arctic environments because of the low abundance of vegetation. 19 Statistical analyses of remotely sensed topographic, optical and/or climate data landform classifications 21 use a range of multivariate and simple statistical techniques including generalized linear methods such as linear discriminant analysis (LDA), 13,22 logistic regression, 17 and artificial neural networks. 13,23 Reported comparisons of different suites of statistical modeling suggest that simple models such as LDA and logistic regression perform equally when compared to more complex machine-learning techniques.…”