HAL Id: hal-02606248
https://hal.inrae.fr/hal-02606248
Submitted on 16 May 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Reanalysis of the 1893 heat wave in France through offline data assimilation in a downscaled ensemble
meteorological reconstruction
Alexandre Devers, Jean-Philippe Vidal, Claire Lauvernet, B. Graff
To cite this version:
Alexandre Devers, Jean-Philippe Vidal, Claire Lauvernet, B. Graff. Reanalysis of the 1893 heat wave in France through offline data assimilation in a downscaled ensemble meteorological reconstruction.
EGU General Assembly 2017, Apr 2017, Vienna, Austria. 19, pp.1, 2017. �hal-02606248�
Reanalysis of the 1893 heat wave in France through offline data assimilation in a downscaled ensemble meteorological reconstruction
Alexandre Devers
1Jean-Philippe Vidal
1, Claire Lauvernet
2, Benjamin Graff
31
Irstea, UR HHLY, Hydrology-Hydraulics, Villeurbanne, France
2
Irstea, UR MALY, Villeurbanne, France
3
CNR (Compagnie Nationale du Rhˆ one), Lyon, France
[email protected] During the months of August 1893 and August 2003 two heat waves greatly affected France. This study considers “offline” data assimilation method using theEnsemble Kalman Filter to provide a reanalysis of the two heat waves. The method is here applied for reconstructing the 8-24 August 1893 heat wave in France, using all available daily temperature observations from that period. In order to assess the performance of assimilation methods, the method is used to reconstruct the well-known 3-14 August 2003 heat wave by using (1) all available stations, and (2) the same station density as in August 1893. Results show a spatially coherent view of the heat waves at the national scale as well as a reduced uncertainty compared to initial meteorological reconstructions, thus demonstrating the added value of data assimilation.
Context
0 1000 2000
1873 1883 1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993 2003 2013
Number of stations
Observations
Temperature
SCOPE Safran
1873 1883 1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993 2003 2013
Year Reconstruction / Reanalysis
SCOPE Safran
Data available over the 1871-2012 period
I Observations : Temperature over France for the 1871-2012 period with associated measurement errors
I Safran [Vidal et al., 2010] : Determinist reanalysis of meterological variables / Daily on the 1958-2012 period / 8 x 8 km cell over France (8602 cells)
I SCOPE [Caillouet et al., 2016] : Ensemble reconstructions (25 members) of
meteorological variables / Daily on the 1871-2012 period / 8 x 8 km cell over France (8602 cells)
Problem
How to assess the spatio-temporal evolution of the daily mean temperature during the heat waves of August 1893 and August 2003 in France?
→ Through the observations?
●
●
●
● ●
●
● ●
●
●
●
●
● ● ●
● ● ●
●
● ● ●
●
●
●
●
●
●
● ●● ●●
●
●
●
●● ●
● ●
●●
●
●
● ● ●
●
●
●
● ●
●
● ●
●
●●● ●● ●
● ●●● ●●
● ● ●●
● ●
●● ●
● ● ●
●● ●
● ●
● ●● ● ●● ●
● ●● ● ● ●
● ● ● ●● ●●
● ● ● ●
●● ●● ●● ●●● ●● ● ●●● ●● ●● ●
● ● ● ● ●●● ●
● ● ● ● ● ● ● ●
● ● ● ●● ● ●● ● ●● ●
● ●● ● ● ● ● ● ● ●
●● ●● ● ● ●● ● ● ● ● ●
● ●● ●● ●●● ●● ● ● ●● ● ● ● ●
●●●●●●●●●●●● ● ● ● ● ● ●●●●●● ● ● ●●● ●●●● ●●●●● ●● ● ●●●● ●
● ●● ● ● ●● ●●● ●● ●● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ●●●●●●●●● ● ● ● ● ●
● ●●● ● ●● ● ● ●● ● ●● ●●● ●●● ● ● ● ●
●● ● ●● ●● ● ● ●● ● ● ● ●● ●●●●● ●●● ●●● ● ●● ●●●●●●● ●● ●●●●● ● ●●●● ● ● ●●●●
●● ● ●● ● ● ● ●●●● ● ●●● ●●● ● ● ● ●● ●● ●
● ●● ●● ● ● ● ● ● ●● ●● ●● ● ● ●
●● ●● ● ●● ●●● ● ●● ●● ●●●● ● ●●● ●● ●●● ●● ●● ● ● ●●● ● ●●● ● ●●
● ● ●● ● ●● ● ● ●● ● ●● ● ●
●●●● ● ●● ● ● ●● ● ●● ●● ● ● ●●●
● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●●●●
● ● ● ● ● ● ●●●● ●● ●● ● ●● ● ● ●
● ● ● ●●● ●●● ● ●●● ● ●● ● ●●● ●● ● ●● ●●● ●● ●●● ●●● ● ● ●●
●●● ●● ● ●●●●●● ● ●● ● ● ● ●●● ● ●● ●●●●● ●● ●● ● ●●
●●●●●● ● ●● ● ● ●● ●● ●●● ● ●●● ●●●●● ●● ● ● ● ●
●●● ● ●●● ●●●●●●● ● ●●●●●●●●●● ● ●● ●●●●●●●
● ●● ●● ●● ● ●● ● ● ● ● ●
● ●● ● ●● ● ● ● ●● ●● ●
● ●●● ● ● ●● ●●● ●● ●● ● ● ●●
● ● ● ● ●● ● ●● ● ● ● ● ●●●
●● ● ● ●●● ● ● ● ●●● ●● ●● ● ● ● ● ● ●
● ● ● ● ● ● ●●● ●●● ●● ●●●●●●
● ● ●●● ● ● ● ● ● ● ● ●
●● ● ●●●●●● ● ●● ●● ●●●●● ● ●● ● ●●● ●●●●● ●● ●●●●
● ● ● ● ●●● ● ●● ●●● ●
●● ● ●●●● ● ●● ● ●● ● ●
● ●● ● ● ●● ●● ● ● ●● ● ● ●● ● ● ●
● ● ● ● ● ● ● ●● ●● ● ●
● ●● ● ● ● ● ● ●● ●● ●● ● ●●
● ● ●● ●●● ● ● ● ● ●●● ●● ●● ●●●
● ●● ● ● ● ● ●●●● ●●●●●●●● ●●●● ●● ● ●●● ●●●
● ● ●● ●● ●● ●● ●●● ● ● ●●●●● ●●●●● ●●●●●●●● ● ●●●●●●●● ●●●●●●
● ● ● ●● ● ● ● ● ●● ●●● ●●● ●●●● ● ●●●●
●● ● ● ● ● ● ●● ● ●●● ●●● ●●● ● ●● ●● ●●●
● ●● ●●●●●● ●● ●● ● ●● ●●● ●●●●●●●●●●●●● ●●●● ●● ●●●●●● ●● ●●●●
●●●●●●●● ● ●● ● ● ●●● ●●●● ● ● ●●●● ● ●● ●● ● ●●●● ●●●●●●●●●●●●
● ● ●●●● ●● ● ● ● ●●●●● ● ● ● ●● ●● ●
● ● ● ● ●● ●● ● ● ●●●●●● ●●● ●● ●
●● ● ● ● ●●● ●● ● ● ●● ● ● ● ●●
● ● ● ●● ● ● ●●● ●● ● ● ● ●●●
● ● ● ● ●●● ●●●● ● ●● ●●● ●●● ● ●●● ●● ●● ●●●●●●●●● ●●●● ●●●●●● ●●●●●
●● ●● ● ●●● ●●●●●● ● ●●● ● ●●●● ● ●●● ●●●
●●● ● ●●●● ● ●●●● ●●● ●● ●●●●● ●●●●●●●●●● ● ●●● ●● ●●● ●●●●●●● ● ●●●● ●●●●●
●● ● ●●●● ●● ● ● ● ●● ●● ● ●●●●●● ●
● ●●● ● ● ● ●● ●●● ● ●●●● ● ●● ● ●●●
● ● ●●●●●● ●●● ● ●● ●●●●● ●●●● ●● ●●● ●●● ● ●●●●● ●● ● ●●● ●●
● ● ● ● ● ●● ●● ● ●●●●●●●● ●● ● ●
● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●
●● ●● ● ● ● ● ● ●●●● ● ● ●
● ●●● ● ● ● ●●● ● ● ●●● ● ● ● ●● ● ●
● ●●●● ●●●● ●● ●●● ●●●● ●● ● ● ●● ●● ●●●● ●●● ● ●●●●●●● ●●● ●●●●●● ●
● ●● ● ● ●● ●●● ●●●● ●●●● ● ● ● ●
● ● ●● ● ●● ● ● ●●● ● ● ● ● ●●
● ●● ●●●●●●● ● ●●●● ●● ●
●● ● ● ●● ●●● ●●●●● ● ● ●● ● ●●●●●●● ● ●●●●●●●
●●● ●●● ● ●●● ●● ●● ●●● ●● ●● ●●●●● ● ● ●● ● ●●●
● ●● ● ●● ● ●●● ●● ● ●●● ● ●●●●●
●●● ● ●●●●●●● ● ●● ●● ●●● ●●●●●●●●●● ●●●●●●● ●●● ● ●● ●●● ● ●●●● ● ●●●●●●
●● ●● ● ● ● ●● ● ● ●●●● ●● ● ●● ● ●●● ●
●● ● ●● ● ● ●● ●●● ● ● ●● ●● ●● ●●
● ●●● ●●●● ●● ●●●● ● ●●● ●● ●● ● ●●● ● ●●● ● ●● ● ● ●●
● ●● ● ●● ● ●●●● ●●● ●●● ●
●● ●●● ● ●● ●●● ● ● ● ● ● ●● ●
● ● ●● ● ●● ● ● ●● ●● ●●
● ●●● ● ● ●●● ●●● ●●●● ● ●● ●●●●●●● ●
●● ● ●●● ● ●●● ●●● ●●●●●● ●●● ●●●
●● ● ● ●● ● ●●●● ● ● ●●●● ●●●●●
● ● ● ●● ● ●●●●●●●●●● ●● ●●●
● ●● ●●● ● ● ●●●●●●
● ●● ●
● ●● ●
●
●● ●●
●
●
●
1893 2003
200 600 1000 200 600 1000
1800 2200 2600
X Lambert [km]
Y Lambert [km]
Observations stations (orange circle) and example cell (red square)
1893 : Scarse and sparse observations / Unable to give a global vision of the spatio-temporal dynamic of the heat wave
2003 : Dense and abundant observations / Give a good idea of the spatio-temporal dynamic of the heat wave
→ Through SCOPE reconstruction?
High uncertainty / Sometimes difficulties to reproduce extreme events
⇒ By using data assimilation taking into account both the reconstruction (SCOPE) and the observations available
Methods
I Offline data assimilation : No model / Only analysis step when observations are available
Model Analysis step
Analysis Background
(SCOPE)
Observations
tn+1
Analysis step
Analysis Background
(SCOPE)
Observations tn
Offline assimilation scheme
I Spatial data assimilation with Ensemble Kalman Filter [Evensen, 2003]
I The observations are perturbed following the distribution N(0, σobs2 ) with σobs = 1◦C
I Observation error covariance matrix is diagonal
I Background error covariance matrix : defined by the 25 members of SCOPE / constrained by Safran climatology (i.e spatial correlations) to avoid spurious correlations [Houtekamer and Mitchell, 2001]
Case study 1: Reanalysis of the 2003 heat wave with all observations available
CRPS (Safran,Background) CRPS (Safran,Analysis)
200 600 1000 200 600 1000
1800 2200 2600
X Lambert [km]
Y Lambert [km]
0 1 2 3
CRPS (°C)
CRPS value between Safran and the background and between Safran and the analysis
Background Analysis
0 10 20 30 40 0 10 20 30 40
0 10 20 30 40
T (°C) Safran
T (°C)
0 5000 10000 15000
Count
Daily mean temperature for all cells between Safran and the background and between Safran and the analysis
I Improvement on 98% of the cells / Correction of bias
●
● ●
● ●
●
●
●
●
●
●
● ● ●
● ● ●
● ●
●
●
● ● ●
● ●
● ●
●
● ●
●
●
●
● ●
●
15 ●
20 25 30
07−27 08−01 08−06 08−11 08−16 08−21 08−26 08−31
Date
T (°C)
●
●
●
●
●
Analysis Background Obs assimilated Obs not assimilated Safran
Evolution of the daily mean temperature for the example cell. Observations errors are represented by the error bar (+/- σobs)
I Reduction of uncertainty / Recovery of dynamics
Case study 2: Reanalysis of the 2003 heat wave with the density of 1893 observations
CRPS (Safran,Background) CRPS (Safran,Analysis)
200 600 1000 200 600 1000
1800 2200 2600
X Lambert [km]
Y Lambert [km]
0 1 2 3
CRPS (°C)
CRPS value between Safran and the background and between Safran and the analysis
Background Analysis
0 10 20 30 40 0 10 20 30 40
0 10 20 30 40
T (°C) Safran
T (°C)
0 5000 10000 15000
Count
Daily mean temperature for all cells between Safran and the background and between Safran and the analysis
I Improvement on 96% of the cells / Correction of bias / Some overestimation for high temperatures
●
● ●
● ●
●
●
●
●
●
●
● ● ●
● ● ●
● ●
●
●
● ● ●
● ●
● ●
●
● ●
●
●
●
● ●
●
15 ●
20 25 30
07−27 08−01 08−06 08−11 08−16 08−21 08−26 08−31
Date
T (°C)
●
●
●
●
●
Analysis Background Obs assimilated Obs not assimilated Safran
Evolution of the daily mean temperature for the example cell. Observations errors are represented by the error bar (+/- σobs)
I Reduction of uncertainty / Recovery of dynamics
⇒ Assimilation scheme very efficient even with sparse observations → Application on the 1893 heat wave
References
Caillouet, L., Vidal, J.-P., Sauquet, E., and Graff, B. (2016). Probabilistic precipitation and temperature downscaling of the twentieth century reanalysis over france. Climate of the Past, 12(3):635–662. doi:10.5194/cp-12-635-2016
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
Houtekamer, P. L. and Mitchell, H. L. (2001). A sequential ensemble kalman filter for atmospheric data assimilation. Monthly Weather Review, 129(1):123–137. doi:10.1175/1520-0493(2001)129¡0123:asekff¿2.0.co;2
Vidal, J.-P., Martin, E., Franchist´eguy, L., Baillon, M., and Soubeyroux,J.-M. (2010).A 50-year high-resolution atmospheric reanalysis over france with the safran system. International Journal of Climatology, 30(11):1627–1644.
doi:10.1002/joc.2003
Case study 3: Reanalysis of the 1893 heat wave with all observations available
15 20 25 30
07−23 07−28 08−02 08−07 08−12 08−17 08−22 08−27 09−01
Date
T (°C)
●
●
●
●
Analysis Background Obs assimilated Obs not assimilated
Evolution of the daily mean temperature for the example cell.
I Reduction of uncertainty / Important correction for specific days Dynamics and intensity of the two heat waves
I In 1893 the resolution of available information is sparse (few observations, no Safran) but the analysis addresses this information gap.
Evolution of the daily mean temperature for each cell / event / data type
I Description of the two heat waves through the analysis :
1893 : The temperatures increase slowly during almost ten day / Stay at the same levels for 4-5 days / And exceed 25◦C on 70% of the cells for at least one day but stay under 30◦C / Then temperatures
drop rapidly
2003 : The temperatures increase very fast, in 2/3 days the maximum temperatures all over France
appear / Then the temperatures reach a plateau for a few day / 95% of the cells exceed 25◦C and 45%
exceed 30◦C for at least one day / The decrease is almost as fast as the increase
⇒ Even if the two heat waves have similarities (same time of the year), the duration, intensity and
dyanmics are different. The 2003 heat wave is short and intense (less than 15 days) with temperatures reaching 30◦C while the 1893 heat wave the temperature increase slowly and continued for almost 25 days but stay under 30◦C
Conclusion / Future work
I Conclusion
◦ Ensemble Kalman Filter: Reduction of uncertainty / Correction of bias/ Even with sparse observations
◦ Comparison of the two heat waves: Same time of the year but different dynamics and intensity
I Future work
◦ Comparison of results with the use of other filters
◦ Improvement of the error statistics definition for observations
◦ Test of the same data assimilation scheme on precipitation / Issue with skewness, zeros, etc...
◦ Complete meteorological reanalysis of the 1871-2012 period for both temperature and precipitation
◦ Hydrological modeling over France for the 1871-2012 period using the reanalysis produced for improving current hydrological reconstruction
For more information on hydrometeorological reconstructions : Downscaling and hydrological uncertainties in 20th century hydrometeorological reconstructions over France by Jean-Philippe Vidal et al. / Hall A at board number A.160 / Wednesday, 26 Apr 2017, 17:30-19:00
EGU General Assembly 2017, Vienna, Autriche, 23-28 April 2017
