Hydrometry data rescue , a stake for the future

The knowledge of past events and long-term series is a base to understand phenomena, analyze climate change and forecast future events. To increase long term series and their accuracy, NUNIEAU software, sort of


The Beginning in Mediterranean Sea
The first rescue was done after 2004 for sea level series in the Gulf of Lion (French western part of Mediterranean sea).Only a dozen of numerical years was available in this area before this work.This rescue had two objectives.
The first one was to capitalize knowledge of past events and try to give indication about maximum level and statistic levels [3,4].About 150 cumulate years of data were rescued in sea and lagoons areas.Technical reports were provided to give an overview of data quality.We estimate if series are available for analyze of: x historical extreme events, x sea level statistics x sea level change related to climate change.The second objective was to correlate these series with long-term climate change as weather regimes [5].

Enlargement for other French data type
Data rescue is needed not only in Mediterranean Sea but also in the other coasts of French areas.The French national hydrographic institute (SHOM) was interested to use this software instead of being dependant of a digitizer table.
Some improvements were made to take account of macro-tidal sea particularity (Figure 2).As in the Mediterranean Sea, the objective of this rescue was the capitalization of old data to know extreme events and study effects of climate change [6].But NUNIEAU offers the possibility to work very precisely depending on paper size and scan precision.It allows other uses.For example, SHOM was involved in a tsunami study project after the Sumatra 2004 tsunami.NUNIEAU enabled to rescue tsunami waves along French coasts.Objectives were to correlate seisms and height waves [7,8].
At the same time, SCHAPI (French national service for hydrometeorology and support to flood forecasting) ordered the development of a NUNEAU river module.
There were some benchmarks in some forecasting service.But few services spent time to do the digitalization and this module was not well tested.The main improvement was the refolding effect (Figure 3) which occurs during extreme flows.A test of rainfall module was run with French Meteorological Institute (Météo France).They have also developed their own tool named DIGITALISE and based on digitizer table.Some conclusions of Météo France in one of their ANR EXTRAFLO reports are: x Results are near the same between the two approaches x Differences are visible for very short time step (6 or 15 minutes) x DIGITALISE can lower peaks and NUNIEAU can increase peaks x DIGITALISE is quicker for rain analysis but need to be used by confirmed users with a good knowledge of rainfall measurement x NUNIEAU, more complex in parameters, is more generic (a lot of record can be managed).It could be easily deployed, without material, particularly for tenders.
In this ANR EXTRAFLO project, one objective was the best assessment of extreme rain events [9] and a part of the project was the rainfall data rescue.

Foreign experience
Data rescue is anything but simple.It needs time despite the fact that NUNIEAU or other types of tools are available.Moreover, data rescue is often less considered than working on numerical models even if data are the base for a good calibration and understanding.
Few people work on these data .Even if a lot of archives are known, few institutions decide to launch huge rescue works.
For NUNIEAU software, Survey of India (SOI) wanted to use it.SOI translate a part of the software in English for its internal needs.Some trainees used NUNIEAU rain module in Burkina Faso and Chile in industrial conditions.This means NUNIEAU can be used with few support, by people not involved in the development and for producing long data series in different types of chart format.The two projects aims to produce rainfall Intensity-Duration-Frequency [10,11].NUNIEAU had to be adapted to Chile rain gauge type (Figure 5).Chile rainfall chart use an orthogonal axes with no curved Yaxis.It was also an opportunity to compare with other tools [11], and the conclusions were: x Software exist but are not available x Software are linked to a type of rainfall chart x Software are not free x Software are program prototypes Some software [12,13] seem to use more complex image treatment to digitalize rainfall charts.Nevertheless, in some case, they have some problem in recognition.DIGITALISE and NUNIEAU have the same type of problem for rainfall particularly for extreme rainfall.A lot of time, problems come from record bugs on the chart and not on software recognition (come back of the stylus).One of the particularities of NUNIEAU, coming from the sea application is to provide at each time a value of minimum, maximum and barycenter of all the graphs.This particularity is really important to manage a lot of waves for sea and also river.For rainfall chart, barycenter is used and at each refolding, we manage minimum and maximum to conserve rain.
Even if some problems exist in the use of NUNIEAU [14], the work done by Cerema, other French and foreign LQVWLWXWLRQV HQDEOH WR FRQVLGHU 181,($8 DV ³WKH RQO\ publically available software for digitizing tide gauge FKDUW´ >@ :LWK WKH UDLQIDOO FRQFOXVLRQ >@ WKLV sentence could be extended to sea, lagoon, river and rainfall gauge.

Figure 1 .
Figure 1.NUNIEAU process in Port de Bouc station (Mediterranean Sea) (a) Scan chart: in this case one week and a second little part surrounded in blue; (b) Referencing Chart in blue circle and recognition of signal in green pixels; (c) Result visualization referencing file by referencing file, the signal surrounded in blue in 1.a comes after in 1.c.

Figure 2 .
Figure 2. Port-Tudy Harbors in Atlantic Ocean, there were 10 days on the same image.The duration in x is one day.Magenta polyline is created with predicted tide to recover only day by day.

Figure 3 . 0704021 Figure 4 .
Figure 3. Aude River in Marseillette station during 1981 floods.There are two signals, each about one week.We can notice there are refolding effects on the two signals.The plot below shows the treatment of the original data.Regional service of French Guyana wanted to improve their knowledge of rainfall and we helped them to manage rainfall chart record in NUNIEAU.Technical results were promising.But only developers used rain module (Figure4).

Figure 5 .
Figure 5. Rainfall Chile example, grids are Cartesian axes instead of most of chart as Figure 3.a.
(2012).Refined numerical modeling of the 1979 tsunami in Nice (French Riviera): Comparison with coastal data, Journal of Geophysical Research, Vol 117, Issue F1