Mapping Mekong river flow regime: scientific benefit of a capacity building action

The Secretariat of the Mekong River Commission (MRC), a cooperation body for the four countries of the lower basin – Cambodia, Laos, Thailand and Vietnam – has initiated and supervised the Mekong-HYCOS and Mekong-HYCOS Follow-up projects since 2007. These projects have led to the installation of more than 30 hydrological stations along the Mekong and its tributaries, making it possible to establish a data collection system shared between the MRC and its member states.

In 2017, the MRC wished to continue this support for the Mekong hydrometric network and requested AFD’s financial support for this purpose. AFD wanted local skills in analysing water resources to be strengthened in tandem with the network’s deployment, which led the MRC to request tenders in this area, to which the IOWater, INRAE, IWMI and the CNR jointly responded. They proposed, among other things, to deal with the regional analysis of river regimes.

The objectives were to obtain :

• a concrete analysis showing how water data can be used at national and regional level ;
• training in the use of the tools needed for this analysis;
• an incentive to improve the overall management and use of this data within the MRC.

Benjamin Graff of the CNR coordinated a training team of three hydrology experts: himself for the metrology, Éric Sauquet (Riverly) for the analysis of the data at each measuring station, for example the delimitation of hydrological regimes (size and seasonal distribution of flows) and Étienne Leblois (Riverly) for the mapping of this hydrological regime at every point of the hydrographical network.

This mapping work is, in principle, a classic scientific problem: the INRAE experts use to map the average flow by breaking down the seasonal variability into statistically uncorrelated factors, and associating these factors with known physical characteristics such as the abundance of snowmelt or the amount of rainfall. However, this standard approach presupposes extensive knowledge of local physical factors.

Here, it was proposed to adapt the approach by using an automatic learning technique known as ‘non-negative matrix factorisation’, which avoids any need for an explicit link with the dominant physical characteristics. The experts thus chose to adapt their approach to provide tools suitable for subsequent autonomous local use. It has to be said that the extent to which the participants have capitalised on these tools is unknown – the feeling is that, alongside the organised sessions, long-term joint activities (joint work, staff exchanges) would help to achieve lasting results; at the end of 2020, discussions are taking place along these lines.

For the time being, we will remember from this action that the expertise or training projects also make it possible to inform and stimulate the research itself, the expert often being led to introduce, according to the needs of local demand, innovative solutions which it will of course be up to him to rework later as part of his research.