The objective of the thesis is the optimal combination of two meteorological datasets from different sources covering the entire 20th century (observations and reconstructions based on atmospheric reanalysis), taking into account their respective uncertainties.
The methodology developed heavily relies on assimilation methods, providing a framework suited to this type of work. The improvement of meteorological forcings led to coherent streamflow reconstructions over 140 years.
The obtained hydro-meteorological reconstructions have the advantage of broader spatial and temporal coverage compared to simple observations, and a reduced uncertainty when compared to reconstructions using only atmospheric reanalysis information.

Climate and hydrology have a strong impact on our society, and adaptating to climate change requires a better knowledge of the historical evolution of hydrometeorological variables. In France, the low density of meteorological and hydrometric observations before the 1970s made it difficult to have a clear vision of their long-term evolution. To fill this gap, various modeling chains have recently been developed to create hydrometeorological reconstructions that are independent of the available historical observations. This thesis work combines these two sources of information - observations and reconstructions - and their respective uncertainties through data assimilation methods, in order to better understand the evolution of climate and hydrology in France from the late nineteenth century onwards.This work is based on the SCOPE Climate high-resolution reconstructions, obtained from the 20CR global reanalysis through an analogue downscaling method and covering the period 1871-2012. These reconstructions are used as a background for assimilating historical precipitation and temperature observations using an ensemble Kalman filter approach. The assimilation method applied is an offline ensemble Kalman filter, that allows taking advantages of the different sources of information without necessarily resorting to a numerical model, and leading to a low computational cost. The assimilation is carried out at both daily and annual time steps to capture the day-to-day variations and extremes as well as the multidecadal climate variability. Applying the method over the period 1871-2012 led to FYRE Climate (French hYdrometeorological Reanalysis Climate), a daily meteorological reanalysis composed of 25 members on an 8 km grid covering the whole of France.FYRE Climate is then used as forcings to a conceptual hydrological model to reconstruct streamflow time series from 1871 onwards over 661 near-natural catchments. In particular, these reconstructions take into account the uncertainties on the input and hydrometric observations during the model calibration, but also the model uncertainty through a post-processing approach. The few hydrometric observations available over the 20th century also made it possible to constrain these hydrological reconstructions using an offline deterministic Kalman filter. This approach led to the FYRE Hydro daily streamflow reanalysis, also composed of 25-member ensemble time series for each of the 661 catchments.Applying an ensemble Kalman filter at a local scale, with a daily time step and on non-Gaussian variables - such as precipitation and discharge - is particularly innovative in the context of a long-term hydroclimatic reanalysis. It required implementing an anamorphosis transformation and an anisotropic localization approach. It also required to carefully define the errors on observations, on the climate background, and on the hydrological background, but also on the hydrological modeling itself. Beyond the reduction of uncertainty induced by the integration of additional information, the multidecadal variability is well reproduced in both FYRE Climate and FYRE Hydro. These two reanalyses also allow analyzing local extreme events over the whole 20th century, thanks to the integration of in-situ meteorological and hydrometric information. Several high-quality data sets - the FYRE Climate and FYRE Hydro reanalyses, but also hydrological reconstructions with additional variables like the snow water equivalent - are made available for studying the climatic and hydrological evolution over the 1871-2012 period.

Funding

50% bourse Irstea, 50% CNR (2016-2019)

References

• Bhend, J., Franke, J., Folini, D., Wild, M., and Brönnimann, S.: An ensemble-based approach to climate reconstructions, Clim. Past, 8, 963-976, doi: 10.5194/cp-8-963-2012, 2012
• Caillouet, L., Vidal, J.-P., Sauquet, E., Devers, A., and Graff, B.: Ensemble reconstruction of spatio-temporal extreme low-flow events in France since 1871, Hydrol. Earth Syst. Sci., 21, 2923-2951, doi:10.5194/hess-21-2923-2017, 2017.
• Caillouet, L., Vidal, J.-P., Sauquet, E., Graff, B., and Soubeyroux, J.-M.: SCOPE Climate: a 142-year daily high-resolution ensemble meteorological reconstruction dataset over France, Earth Syst. Sci. Data Discuss., doi:10.5194/essd-2018-79, in review, 2018.
• Devers, A., Vidal, J.-P., Lauvernet, C. and Graff, B.: A 140-year high-resolution meteorological reanalysis over France through offline data assimilation in an ensemble of downscaled reconstructions from 20CR, In Keller, J. (Ed.): International Symposium on Regional Renalysis – Scientific Program, p. 8-9, 2018.
• Evensen, G. (2003).The Ensemble Kalman Filter: theoretical formulation and practical implementation. Ocean Dynamics, 53(4):343--367. doi:10.1007/s10236-003-0036-9.

Publications and communications :

Paper

• DEVERS, A., VIDAL, J.P., LAUVERNET, C., GRAFF, B., VANNIER, O.- 2019. A framework for high-resolution meteorological surface reanalysis through offline data assimilation in an ensemble of downscaled reconstructions. Quarterly Journal of the Royal Meteorological Society.

Oral communications

• DELUS, C., GRELIER, B., FRANCOIS, D., DROGUE, G., VIDAL, J.P., CAILLOUET, L., DEVERS, A., SAUQUET, E., LAUVERNET, C.- 2019. Comparaison de séries climatiques historiques dans les bassins versants de la Meuse et de la Moselle. Bases de données et modèles climatiques. Séminaire de la commission "Changements climatiques et territoires" du CNFG 21/03/2019-22/03/2019, Dijon, FRA. 23 p.
• DEVERS, A., VIDAL, J.P., LAUVERNET, C., VANNIER, O.- 2019. Une reconstruction hydrologique ensembliste depuis 1871 sur 662 bassins versants en France. Ateliers de Modélisation de l'Atmosphère 2019 11/03/2019-13/03/2019, Toulouse, FRA. 39 p.

Posters

• DEVERS, A., VIDAL, J.P., LAUVERNET, C., VANNIER, O.- 2019. A 140-year ensemble streamflow reconstruction of over 662 catchments in France. European Geosciences Union 2019 07/04/2019-12/04/2019, Vienna, AUT. 1 p.
• DEVERS, A., VIDAL, J.P., LAUVERNET, C., GRAFF, B.- 2018. High-resolution precipitation reanalysis over France through offline data assimilation in a downscalled ensemble meteorological reconstruction. EGU General Assembly 2018 08/04/2018-13/04/2018, Vienna, AUT. 1 p.
• DEVERS, A., VIDAL, J.P., LAUVERNET, C., GRAFF, B.- 2017. Reanalysis of the 1893 heat wave in France through offline data assimilation in a downscaled ensemble meteorological reconstruction. EGU General Assembly 2017 23/04/2017-28/04/2017, Vienna, AUT. 1 p.
• Communication scientifique sans actes
• DEVERS, A., VIDAL, J.P., LAUVERNET, C., GRAFF, B- 2018. A 140-year high-resolution meteorological reanalysis over France through offline data assimilation in an ensemble of downscaled reconstructions. Fisrt International Symposium on Regional Reanalysis 17/07/2018-19/07/2018, Bonn, DEU. 28 p.
• Communication scientifique sans actes
• DEVERS, A., PRINTEMPS, A., VIDAL, J.P., LAUVERNET, C., VANNIER, O.- 2018. Vers une réanalyse hydrologique par assimilation de débits sur plus de 600 bassins versants en France. Colloque National d'Assimilation de données 28/09/2018, Rennes, FRA. 46 p.

Cite the thesis

Alexandre Devers. Vers une réanalyse hydrométéorologique à l'échelle de la France sur les 150 dernières années par assimilation de données dans des reconstructions ensemblistes. Hydrologie. Université Grenoble Alpes, 2019. Français. ⟨NNT : 2019GREAU029⟩. ⟨tel-02506254⟩

Access manuscript on HAL thèses