Integration of remote sensing data into hydrological models for reservoir management

Abstract

The purpose of this paper is to present the methodology set up to derive catchment soil moisture from Earth Observation (EO) data using microwave spaceborne Synthetic Aperture Radar (SAR) images from ERS satellites and to study the improvements brought about by an assimilation of this information into hydrological models. The methodology used to derive EO data is based on the appropriate selection of land cover types for which the radar signal is mainly sensitive to soil moisture variations. Then a hydrological model is chosen, which can take advantage of the new information brought by remote sensing. The assimilation of soil moisture deduced from EO data into hydrological models is based principally on model parameter updating. The main assumption of this method is that the better the model simulates the current hydrological system, the better the following forecast will be. Another methodology used is a sequential one based on Kalman filtering. These methods have been put forward for use in the European AIMWATER project on the Seine catchment upstream of Paris (France) where dams are operated to alleviate floods in the Paris area.Key words soil moisture; remote sensing data (SAR): variational assimilation; sequential assimilation; rainfall-runoff models; reservoir management; Kalman filtering Intégration de données satellitales dans des modèles hydrologiques pour la gestion des barrages réservoirs Résumé Le but de cet article est de présenter une méthodologie mise au point pour déduire d'images ERS/SAR un indice d'humidité des sols pour un bassin versant et d'étudier les améliorations possibles de l'assimilation de cette information dans des modèles hydrologiques. La méthodologie utilisée pour traiter des données satellitales est basée sur la sélection de cultures cibles, pour lesquelles le signal radar est principalement sensible aux variations d'humidité des sols. Ensuite, un modèle hydrologique est choisi, dans le but d'y intégrer l'information sur l'humidité des sols obtenue par télédétection. L'assimilation de cette information dans des modèles hydrologiques est basée tout d'abord sur une méthode de type variationnelle, qui consiste à remettre en cause les valeurs des paramètres du modèle. On fait l'hypothèse que plus le modèle simule de manière correcte la situation en cours, meilleure sera la prévision. Une autre méthode utilisée est une méthode séquentielle qui utilise le filtre de Kalman. Ces méthodes ont été proposées pour être utilisées dans le cadre du projet européen AIMWATER sur le bassin versant de la Seine en amont de Paris (France), où des barrages réservoirs sont utilisés pour diminuer le niveau des crues à Paris.

Integration of remote sensing data into hydrological models for reservoir management

“…CETP/CNRS and Cemagref, with results published at different times. Some references were missing in the paper by Loumagne et al (2001) and readers are referred to these here: Le Hégarat- Mascle et al (2000a);Normand et al (2000); François et al (1999); Quesney et al (2000aQuesney et al ( , 2001and François et al (submitted).…”

“…-It is not clearly stated why soil moisture is derived following the multi-selective method. The reason given in Loumagne et al (2001) that "compared to a single land cover selection method, the soil moisture index is most representative of the global watershed" is correct, but it is not the main reason. The main reason is that, in order to rule out roughness change effects, one has to consider a sufficiently large number of various crops to ensure that agricultural works do not happen on the same date, and that mean roughness effects are about constant throughout the agricultural year.…”

“…Concerning the assimilation of the space estimations of soil water content in hydrological models, Loumagne et al (2001) presented the results obtained using two different assimilation techniques: the variational method and the sequential Kalman filter method. Here also, the results presented were obtained in the framework of the PhD by A. Quesney.…”

“…Unfortunately, the SAR data measurements were too sparse (the temporal resolution is 35 days) to bring any improvement combined with antecedent flows and considering time assimilation windows smaller than 30 days. The results presented in Table 3 of Loumagne et al (2001) were obtained by assimilation of both antecedent flows and in situ soil moisture point measurements acquired every day by a TDR probe on the Orgeval catchment. The assimilation window for the flows was optimized to five days and varies between 10 and 30 days for soil moisture according to the period under consideration.…”

“…In Loumagne et al (2001) concerning the Orgeval watershed and supported by the PNRH, not in the framework of the AIMWATER project. -In Fig.…”