A new method to combine atmospheric reanalysis and observations to study the multi-decadal variability of the Seine river

Size: px
Start display at page:

Download "A new method to combine atmospheric reanalysis and observations to study the multi-decadal variability of the Seine river"

Transcription

1 A new method to combine atmospheric reanalysis and observations to study the multi-decadal variability of the Seine river Rémy Bonnet¹, Julien Boé¹ ¹CECI-CERFACS, Toulouse, France

2 River flows observations of the Seine at Paris Low-pass filtered annual series of river flows anomalies of the Seine at Paris Observations 1 Observations 2 40% Average 315m3/s Map of the main French watershed ( - Strong multi-decadal variations of the Seine river flows at Paris - Important differences between observations 2

3 River flows observations of the Seine at Paris Low-pass filtered annual series of river flows anomalies of the Seine at Paris Observations 1 Observations 2 40% Average 315m3/s Map of the main French watershed ( What are the mechanisms at the origin of this variability? 3

4 Data issues River flows - Only few long-term observations available à large uncertainties - Temporal homogeneity problems? - Non-climatic anthropogenic influence? (e.g. dam or pumping) Other variables (evapotranspiration, soil moisture ) - No long-term observations à Development of hydro-meteorological reconstructions 4

5 Safran-Isba-Modcou - Hydro-meteorological system simulating the water and energy fluxes at surface, the river flows evolution and the main aquifers over France - 3 components: - Safran: hourly meteorological analysis (8*8km), available over France from 1958 to present - Surfex v Isba: land-surface model - Modcou: hydrogeological model 5

6 Safran-Isba-Modcou - Interests: - The only calibration in MODCOU is the concentration time ( daily time scale) - Anthropogenic non-climatic effect (dams, pumping) not taken into account à Independance between observed and simulated river flows - All variables of the hydrological cycle available at high resolutions - Main limitation: Safran only available from 1958 to present 6

7 Statistical downscaling method Atmospheric reanalysis E.g. 20CRv2c, ERA20C; available on long-term periods, but with a coarse resolution Analog method (statistical downscaling method) Long-term Meteorological forcing Isba-Modcou River flows + hydrological cycle variables 7

8 Statistical downscaling method Atmospheric reanalysis Analog method (statistical downscaling method) Long-term Meteorological forcing Isba-Modcou Hypothesis: large scale climate state of day d1 large scale climate state of day dn Local variables of day d1 Local variables of day dn Predictor: Pr, Tas, PSL and HUS850 Data: - Large scale: 20CRv2c - Local scale: Safran analysis River flows + hydrological cycle variables 8

9 Main limitation mm/d Observations 20CRv2c reanalysis Analog Low-pass filtered annual series of precipitation anomaly over the Seine watershed (21 years) Years à Unrealistic trends can be present in reanalyses à The reconstructions quality depends of reanalyses 9

10 Observationaly constrained downscaling Idea: to use observations to correct unrealistic trends and improve the temporal variability Steps: - Creation of a set of possible trajectories - Analog method = stochastic method à the k best analogs can be selected each day à 50 best analogs x 56 members = 2800 trajectories - Selection of the trajectory the closest to the observations Bonnet et al., 2017, WRR 10

11 Observationaly constrained downscaling Idea: to use observations to correct unrealistic trends and improve the temporal variability Steps: - Creation of a set of possible trajectories - Analog method = stochastic method à the k best analogs can be selected each day à 50 best analogs x 56 members = 2800 trajectories - Selection of the trajectory the closest to the observations Daily constraint: - Aim: to improve the daily variability - Based on daily observations of precipitation and temperature (SQR4) Monthly constraint: - Aim: to improve the trend and multi-decadal variability - Based on long-term monthly homogenized observations of precipitation and temperature (Météo-France) Bonnet et al., 2017, WRR 11

12 Impact on precipitation reconstructions mm/d Low-pass filtered annual series of precipitation anomaly over the Seine watershed (21 years) Years Observations 20CRv2c reanalysis Analog Analog + Obs constraint à Better representation of trends and temporal variations 12

13 River flows evaluation Monthly and daily correlations of the river flows between the observations and the reconstructions (the series have been deseasonalysed) ( ) Daily SIM (reference) Analog Analog + Obs constraint Seine at Paris Monthly à The constraint method improves significantly the daily and monthly river flows variability over the Seine watershed 13

14 River flows evaluation Low-pass filtered annual series of the Seine river flow anomaly at Paris (21 years) Observations 1 correlation = 0.90 Observations 2 correlation = 0.96 Analog + Obs Constraint 14

15 Annual maximum of daily river flows of the Seine at Paris Observations Analog + Obs constraint Annual series of the standardized annual maximum of daily river flows (reconstruction) and the standardized annual maximum height (observations) (Standardization: ) Correlation = 0.80 ( ) Correlation = 0.70 ( ) - Good representation of the variability through the whole period - A multi-decadal variability is also present in the maximum annual river flows 15

16 piezometric height at Toury Observations Safran-Isba-Modcou (SIM) Analog + Obs Constraint Annual series of the piezometric height anomaly at the Toury station Correlation = 0.77 ( ) - Good correlation with the observations - Underestimation of the variability of the reconstruction due to SIM - A multi-decadal variability is also present in piezometric height 16

17 Conclusion - We developed a new method constraining the results of a statistical downscaling applied to atmospheric reanalysis by long-term observations - Reduction of irrealistic trends/low frequency variations induced by the reanalysis - Improvement of the daily variability - The hydro-meteorological reconstruction developped over the Seine watershed = valuable dataset to study the multi-decadal variability of the Seine hydrological cycle - Multi-decadal variations observed in the Seine river are mainly climate driven and not limited to the 20th century - With a suitable methodology, long-term atmospheric reanalyses provide a great opportunity to improve our understanding of the multi-decadal variations of the hydrological cycle - More details: R.Bonnet et al., 2017, WRR 17