Mixed effects regression for snow distribution modelling in the central Yukon

Statistics;Snow;Snow water equivalent;Remote sensing;Mixed effects model;yukon;mountain;boreal;regression;model;snow distribution;net accumulation;Geography

Kasurak, Andrew

University of Waterloo

关键词: Statistics;    Snow;    Snow water equivalent;    Remote sensing;    Mixed effects model;    yukon;    mountain;    boreal;    regression;    model;    snow distribution;    net accumulation;    Geography;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/4896/1/kasurak_andrew.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

To date, remote sensing estimates of snow water equivalent (SWE) in mountainous areas are very uncertain. To test passive microwave algorithm estimations of SWE, a validation data set must exist for a broad geographic area. This study aims to build a data set through field measurements and statistical techniques, as part of the Canadian IPY observations theme to help develop an improved algorithm. Field measurements are performed at, GIS based, pre-selected sites in the Central Yukon. At each location a transect was taken, with sites measuring snow depth (SD), density, and structure. A mixed effects multiple regression was chosen to analyze and then predict these field measurements over the study area. This modelling strategy is best capable of handling the hierarchical structure of the field campaign.A regression model was developed to predict SD from elevation derived variables, and transformed Landsat data. The final model is: SD = horizontal curvature + cos( aspect) + log10(elevation range, 270m) + tassel cap: greenness, brightness (from Landsat imagery) + interaction of elevation and landcover.This model is used to predict over the study area. A second, simpler regression links SD with density giving the desired SWE measurements. The Root Mean Squared Error (RMSE) of this SD estimation is 25 cm over a domain of 200 x 200 km.This instantaneous end of season, peak accumulation, snow map will enable the vali- dation of satellite remote sensing observations, such as passive microwave (AMSR-E), in a generally inaccessible area.