L'apport de la teledetection a un modele de neige applique a un systeme d'aide a la gestion des barrages dans le sud du Quebec.

摘要

The Centre d'expertise hydrique du Quebec (CEHQ) operates a distributed hydrological model (MOHYSE), which integrates a snow model (SPH-AV), for the management of dams in the south of Quebec. It appears that the estimation of the water quantity of snowmelt in spring remains a variable with a large uncertainty. This research aims to evaluate the potential of remote sensing data for the characterization of snow and ultimately to develop methods of integration of satellite data in the snow model for the improvement of the simulations of spring floods.;The study also shows the potential of QSCAT data for the characterization of snow cover. Overall accuracies around 90% are obtained for the detection of melt during the month of April from 2001 to 2007 on both studied watersheds. The relation between the rise of the backscatter coefficient and the snow depth surveys shows good correlation for the 2004 to 2006 years for the Lachute and St-Jerome stations (0.64 to 0.93), but less interesting results for the St-Hippolyte station (0.29 to 0.73). QSCAT products considering only the descendant orbit give best results.;The integration of remote sensing albedo product did not allow improvement in the simulations because of holes in the temporal series caused by cloud cover. Also, the relation between fractional snow cover and snow depth did not show interesting results in an operational context.;The study shows the interest to create new remote sensing SCA products more precise on the studied region. Future works should also evaluate the possibility to adapt the seuilEEN method for a Kalman filter approach. A more spatially extensive study and a better comprehension of the backscatter response in microwaves of the different elements might eventually permit to obtain useful results with QSCAT data.;Remote sensing snow cover area (SCA) products (MODISSCN & IMS) are compared with snow depth surveys at Environment Canada stations and initial simulations of the models. Thru these comparisons, an effective method of integration (seuilEEN) of remote sensing SCA products, based on the hypothesis that satellites can not identify small amount of snow because snow become "dirty" and discontinuous, was developed. The improvement of the Nash coefficient and the root mean square error for spring 2004 to 2007 for the simulations with the approach developed compared with streamflow simulated without remote sensing is 0.11 and 21% on the optimized watershed (du Nord) and 0.13 and 22% on the verification watershed (aux Ecorces). The method also relies to improve peaks identification as much as 36% on the du Nord watershed and 19% on the aux Ecorces watershed.;Keywords: Remote sensing SCA; MODIS; QSCAT; Hydrological model; Snow model; Me Hood.