Transferability of geomorphological distribution models: Evaluation using solifluction features in subarctic and Arctic regions

Hjort, Jan; Ujanen, Joonas; Parviainen, Miia; Tolgensbakk, Jon; Etzelmüller, Bernd

doi:10.1016/j.geomorph.2013.08.002

Cited by 23 publications

(47 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The 10 modeling techniques utilized in this study produced notable variations across the evaluations statistics resulting from different statistical approaches (Thuiller et al , ). Even though most of the machine learning and classification methods (RF, GBM, MAXENT and FDA) performed well in terms of model transferability (in agreement with Hjort and Marmion, ; Heikkinen et al ; Hjort et al , ), other sophisticated techniques (e.g. ANN) showed only moderate performance.…”

Section: Discussionsupporting

confidence: 60%

Integrating climate and local factors for geomorphological distribution models

Aalto

Luoto

2014

Earth Surf Processes Landf

View full text Add to dashboard Cite

Earth surface processes (ESPs) drive landscape development and ecosystem processes in high‐latitude regions by creating spatially heterogeneous abiotic and biotic conditions. Ongoing global change may potentially alter the activity of ESPs through feedback on ground conditions, vegetation and the carbon cycle. Consequently, accurate modeling of ESPs is important for improving understanding of the current and future distributions of these processes. The aims of this study were to: (1) integrate climate and multiple local predictors to develop realistic ensemble models for the four key ESPs occurring at high latitudes (slope processes, cryoturbation, nivation and palsa mires) based on the outputs of 10 modern statistical techniques; (2) test whether models of ESPs are improved by incorporating topography, soil and vegetation predictors to climate‐only models; (3) examine the relative importance of these variables in a multivariate setting. Overall, the models showed high transferability with the mean area under curve of a receiver operating characteristics (AUC) ranging from 0.83 to 0.96 and true skill statistics (TSS) from 0.52 to 0.87 for the most complex models. Even though the analyses highlighted the importance of the climate variables as the most influential predictors, three out of four models benefitted from the inclusion of local predictors. We conclude that disregarding local topography and soil conditions in spatial models of ESPs may cause a significant source of error in geomorphological distribution models. Copyright © 2014 John Wiley & Sons, Ltd.

show abstract

Section: Discussionsupporting

confidence: 60%

Integrating climate and local factors for geomorphological distribution models

Aalto

Luoto

2014

Earth Surf Processes Landf

View full text Add to dashboard Cite

show abstract

“…This is a machine learning technique and highly flexible for representing interacting and non-linear relations with few statistical assumptions required (Elith et al, 2008;Hjort et al, 2014). The BRT technique uses regression tree models, but these are combined with boosting which builds and combines a collection of models (Elith et al, 2008).…”

Section: Modelling Methodsmentioning

confidence: 99%

“…The BRT technique uses regression tree models, but these are combined with boosting which builds and combines a collection of models (Elith et al, 2008). It has produced models in physical geography that are more transferable to other regions than with conventional generalised linear models (Hjort et al, 2014) and also successfully applied in Ecology (Pittman et al, 2009).…”

Section: Modelling Methodsmentioning

confidence: 99%

Variation in reach-scale hydraulic conductivity of streambeds

Stewardson

Datry²,

Lamouroux³

et al. 2016

Geomorphology

View full text Add to dashboard Cite

Streambed hydraulic conductivity is an important control on flow within the hyporheic zone, affecting hydrological, ecological, and biogeochemical processes essential to river ecosystem function. Despite many published field measurements, few empirical studies examine the drivers of spatial and temporal variations in streambed hydraulic conductivity. Reach-averaged hydraulic conductivity estimated for 119 surveys in 83 stream reaches across continental France, even of coarse bed streams, are shown to be characteristic of sand and finer sediments. This supports a model where processes leading to the accumulation of finer sediments within streambeds largely control hydraulic conductivity rather than the size of the coarse bed sediment fraction. After describing a conceptual model of relevant processes, we fit an empirical model relating hydraulic conductivity to candidate geomorphic and hydraulic drivers. The fitted model explains 72% of the deviance in hydraulic conductivity (and 30% using an external cross-validation). Reach hydraulic conductivity increases with the amplitude of bedforms within the reach, the bankfull channel width-depth ratio, stream power and upstream catchment erodibility but reduces with time since the last streambed disturbance. The correlation between hydraulic conductivity and time since a streambed mobilisation event is likely a consequence of clogging processes. Streams with a predominantly suspended load and less frequent streambed disturbances are expected to have a lower streambed hydraulic conductivity and reduced hyporheic fluxes. This study suggests a close link between streambed sediment transport dynamics and connectivity between surface water and the hyporheic zone.

show abstract

“…13.19). The relief of the slope, depth of thaw, and water content are significant factors in solifluction rates (Walker, 1986;Hjort et al, 2014).…”

Section: Solifluctionmentioning

confidence: 99%