In the present paper, we introduce the design and applications of a new L3 ("L2-to-L3") processor dedicated to the monitoring of river water level. The processor is tailored for the basin to global scale processing of most of the past 30 years of altimetry data.

The main motivation behind this new design is to develop concepts, processing methodologies and tools that are meant to stand for long, to be perennial. Experience has proved that some early design assumptions led to limited processing capabilities. And the resulting tools are usually not able to deal with the changes induced by the permanent evolution of the Satellite Altimetry techniques and characteristics. This is particularly significant, but not limited to, in respect to the space-time distribution of the data acquisitions, the Altimeter measurement modes and the actual location of the water bodies.

The long term capabilities of Inland Water processing chains has often been considered as a second order problem, sometimes relegating it to later times, sometimes by choice or for sake of simplicity, sometimes by mistake. However, the amount of accumulated Altimetry data, which is about 30 years nowadays, coupled to the recent (Sentinel-6) and the forthcoming missions (SWOT, Sentinel-9/CRISTAL, etc.), urgely deserves flexible processing schemes able to deal with these evolutions.

The L3 processor developed by AltiHydroLab.fr focuses on offering as much flexibility as possible. For example, it is designed to be able to track water bodies that do change their shape in time; To exploit data from repeat orbit satellites that may drift at times (e.g., SARAL/Altika) or simply change their orbit (Jasons' interleaved orbits, geodetic orbital phases, etc.); To produce river water level estimates from non-repeat, or long-repeat, orbit missions (CryoSat-2).

For example, it has been estimated (by the author) that Jason-3 overflies about 60,000 different water body sections during each 10-days cycle, this number is about 200,000 for SARAL/AltiKa (35-days cycle), and 160,000 for Sentinel-3 satellite (27-days cycle). Such big numbers make it virtually impossible to maintain a decent catalogue of Virtual Stations alive for decades.

Huge catalogues of virtual stations are hard to maintain in the long run because things are slowly changing, moving, etc. As decades are passing by, the integration of the slow geomorphological changes makes the systems outdated.

Among the auxiliary datasets exploited by the L3 processor is the SWORD database (SWOT A priori River Database), initially designed for the SWOT mission and which is gaining more and more attention into the Altimetry/Inland Water community. This allows to project the data onto a common hydrographic reference. In particular, water level data outputted by the L3 processor are systematically linked to the relevant SWORD nodes so that the data are ready to be used by third parties who also make use of the SWORD database routinely in their frameworks.

The main applications of the presented L3 processor are real time monitoring of river water levels, retrospective analysis based on archives reprocessing and to feed river discharge models able to assimilate space-time distributed measurements.

In complement, the processor also include a data reading tool that allow the users to standardise their Altimetry input data (variable names, data shapes, etc.) so that the processor can is purely format-agnostic. In between, the reading and the processing stages, lives a powerful mini-language that makes it easy to express automatised/recurrent computations that are also format-agnostic.

The work in this paper is funded by personal funds from the author and in the frame of the ESA HYDROCOASTAL project.