TY - JOUR
T1 - A neural network method to determine the presence or absence of permafrost near Mayo, Yukon Territory, Canada
AU - Leverington, David W.
AU - Duguay, Claude R.
N1 - Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.
PY - 1997
Y1 - 1997
N2 - A neural network was used to predict the presence or absence of the permafrost table within 1.5 m below the ground surface, over two study areas near Mayo, Yukon Territory. Input sources used in neural network classifications included land cover (derived from Landsat Thematic Mapper (TM) imagery), equivalent latitude, aspect, and TM band 6 (thermal infrared imagery). For the first study area, maximum median agreement between predicted and field-measured permafrost-table conditions, produced using land cover and equivalent latitude data as input to the neural network, was over 90%. The agreement percentage produced by classification of the second study area, using land cover and equivalent latitude, and using correlative permafrost-surface relations from the first study area, was 60%. Training data, the portability of which is critical in region-wide predictions of active-layer conditions, cannot be transferred between the two study areas examined here.
AB - A neural network was used to predict the presence or absence of the permafrost table within 1.5 m below the ground surface, over two study areas near Mayo, Yukon Territory. Input sources used in neural network classifications included land cover (derived from Landsat Thematic Mapper (TM) imagery), equivalent latitude, aspect, and TM band 6 (thermal infrared imagery). For the first study area, maximum median agreement between predicted and field-measured permafrost-table conditions, produced using land cover and equivalent latitude data as input to the neural network, was over 90%. The agreement percentage produced by classification of the second study area, using land cover and equivalent latitude, and using correlative permafrost-surface relations from the first study area, was 60%. Training data, the portability of which is critical in region-wide predictions of active-layer conditions, cannot be transferred between the two study areas examined here.
KW - Discontinuous permafrost
KW - Neural network
KW - Remote sensing
UR - http://www.scopus.com/inward/record.url?scp=0001586697&partnerID=8YFLogxK
U2 - 10.1002/(sici)1099-1530(199732)8:2<205::aid-ppp252>3.0.co;2-5
DO - 10.1002/(sici)1099-1530(199732)8:2<205::aid-ppp252>3.0.co;2-5
M3 - Article
AN - SCOPUS:0001586697
VL - 8
SP - 205
EP - 215
JO - Permafrost and Periglacial Processes
JF - Permafrost and Periglacial Processes
SN - 1045-6740
IS - 2
ER -