Describe experience relating to water and space technologies
I grew up in a country (France) where water is freely available. The drought in 2003 was considered a one-time event. I had no single lesson on climate change at school. Despite this background, I was raised aware of the links between social and environmental inequality on a global scale.
After high school, I was interested in physics, chemistry, and mathematics. I went to the National Graduate School of Energy, Water, and the Environment after a nationwide competitive examination. During my Master's, I got specialized in signal processing and applied mathematics. This field was highly attractive to me due to both, its theoretical basis (how linear and integral transformations, such as Laplace and Fourier transform, allow us to change our native perspective on a signal to better analyse it) and its numerous real-world applications (seismology, music, speech, and image processing, etc.). Among them, the field of remote sensing allows me to match my interests in physic and applied mathematics with my concern to better monitor the Earth. From my Master of engineering, I conserve the taste of solving programming issues (that arise often) and a target-oriented mind which I find very useful on a daily basis.
I am very proud to work in environmental sciences, particularly at this tipping point of climate change. A sustainable future needs good political decisions which should be based on knowledge, and the latter needs research. On the one hand, satellite observations bring new potential to hydrology. Specializing on this new type of observations is very challenging and exciting. On the other hand, increased computational power and machine learning tools increasingly the development of data-driven approaches to investigate the global water cycle. As the big issue of the twenty-first century, huge effort must be put on Earth sciences. I feel in the right place at the right time.
How can space contribute to water resource management, hydrology, or any water related field?
Thanks to satellites’ global coverage, and the high temporal and spatial resolutions of instruments on board, observations from space represent the greatest potential in Earth monitoring. Transboundary basins can be monitored without political hindering (e.g., by sharing or not sharing in-situ data). Moreover, large-scale mechanisms (both oceanic and atmospheric) can be observed entirely through space and time. The water cycle is one of the planet’s main processes and space-born instruments are an appropriate tool to get the bigger picture. Considering water management (which is not my field), satellites allow also better forecasting of irrigation needs or future flood patterns as well as monitoring crop field stock. All these applications are now possible in countries with poor in-situ networks. There is room for improvement (in both, temporal and spatial sampling, as well as in data processing) but the contribution of satellites to hydrology (“spatial hydrology”) is a pretty new research area on its own.
Could you tell us about your current work, your latest project, or your proudest professional moment?
These days, I am deriving a spatial and continuous estimate of the river discharge over the entire catchment of the Amazon based on satellite observations only. A first principle of hydrology is the conservation of mass. Over any given area, the flux of water entering and leaving an area is equal to the change in water storage in that area. “Adding up” various satellite estimate representing hydrologic fluxes (precipitation, evapotranspiration, change in water storage, and runoff) through the water budget (i.e., water balance equation), the net result should be equal to zero.
My work relies on the terrestrial water budget to optimize all the satellite-based estimates of the water components. In my recent progress, I have used the information of flow direction - describing how water moves on the surface based on the topography - to balance the water budget at pixel scale. If my Ph.D. thesis has already focused on optimizing the monitoring of the water cycle, the description of water budget at pixel scale was made possible by my postdoc fellowship at the University of Tokyo funded by the Japan Society for the Promotion of Science (JSPS), where I have been supervised by Professor Dai Yamazaki who developed the flow direction map based on the newest Digital Elevation Model (DEM). I am now a postdoctoral fellow funded by the Centre national d'études spatiales (CNES) and I intend to extend the methodology I previously developed to the global scale.
As a young researcher my proudest professional moment remains the oral defence of my Ph.D. in front of senior researchers 2.5 years ago. Having a deep discussion with them at eye-level was amazing. While is it a mandatory exercise for researchers, I am a shy person and giving an oral presentation is not one of my assets. I have always had to work hard to prepare any presentation, no matter for what type of audiences (from high school students to peers at a conference). These presentations also help me to mature my research and think about the next step.
What are key challenges in satellite remote sensing of the water cycle?
If the potential of satellite observations is vast, so are the related challenges. The main limitation comes from the uncertainties related to the indirect nature of the observation. Satellite products require sophisticated inversion methods to restitute information about the geophysical variable from the radiation measured by the satellite at the top of the atmosphere. This restitution can be in combination or not, with in-situ data, model simulations, or reanalysis. Thus, satellite products are more or less directly derived from the satellite observation and incorporate more or less auxiliary non-satellite data. Errors in the restitution of satellite products can be caused, for example, by
1. instrumental noise,
2. calibration errors, which affect the accuracy of the measurements,
3. deficiencies inherent in the restitution/inversion process, or
4. errors in the auxiliary data used.
Since the early days of satellite observation, many efforts have been made to improve satellite products. Today, there is a multitude of satellite observations, but in general, these observations are used independently to derive a geophysical variable and result in a multiplicity of products for the same variable. Each product has its own advantages and limitations, and there are still significant discrepancies between these products and a lack of reference data on a global scale that would bring consensus among the scientific community. The effort must now focus on organizing a better framework for combining, as much as possible, all available satellite observations. International programs, such as the Global Energy and Water Exchanges (GEWEX) program, are developing synergistic approaches that use information from several hydrological models to improve their consistency. Beyond this general challenge in the use of satellite, the monitoring of water components varying from one to another.
What are the variables and parameters that can be observed from space with satisfactory accuracy and which ones rely on in-situ data?
Without being exhaustive, the retrieval of all geophysical variables from satellites is not straightforward. We can distinguish some “easily” retrieved variables and “more-difficult” ones. The difficulty in retrieving the variable is linked to the radiative transfer and how the variable’s impact can be distinguished in the radiation measure at the satellite. For instance, Sea Surface Temperature can be more easily retrieved from its impact in microwave radiation than Land surface temperature because the sea surface is smoother than the continental one at micro-wavelength. Soil moisture is far more difficult to be estimated from space because the radiative transfer of the surface is very complicated (due to soil texture, tree’s impact, etc).
Considering the water cycle from space, precipitation can be obtained with satisfactory results from space but calibration on in situ gauges is still mandatory. Evapotranspiration (ET) is indirectly monitored from space and relies on several auxiliary data. ET suffers then from large uncertainty. Last but not least, river discharge cannot yet be retrieved from space, but this is about to be achieved thanks to the next NASA-CNES space mission called Surface Water and Ocean Topography (SWOT) which will monitor river information such as surface elevation, width, slope and estimated river discharge) at a global scale starting from 2022. This is very important since less and less in-situ river data measurements are shared (open source) at the international level.
Your Ph.D. was part of the European project WACMOS-MED, supported by ESA, which contributed to the international project HyMeX. Can you share your experience as a young scientist being part of European and international projects? What were key findings and what are your lessons learnt?
Being part of the international project WACMOS-MED was a great opportunity as a Ph.D. student. The regular project meetings with the collaborators gave me insights into how researchers present technical results and share expertise on a common subject and how ESA leads scientific projects. Often after a meeting, we slightly changed the direction based on the discussion we had. Thanks to ESA support, I have been involved in the creation of an ESA video clip summarizing my Ph.D. results. It was a unique training on how to advertise on research: what and how to show a result for a broad audience.
The international Hydrological Cycle in the Mediterranean Experiment (HYMEX) workshop (Barcelona, 2017 and Toulouse, 2019) was the opportunity to meet the Mediterranean community focusing on the water cycle. This workshop showed me how long-term research (HYMEX is a decadal project) was done by the international community from various thematic (terrestrial atmospheric and oceanic) and technical (in-situ, modelling, and remote sensing) fields. The main finding is the importance of founding a long-term project to improve our knowledge based on collaboration between various communities.
What can you tell us about the water cycle of the Amazon? What are the challenges that need to be overcome for sustainable management of the Amazon basin?
The Amazon basin is an exceptional basin due to the role it plays for the global water and energy cycle. Characterized by complex hydrological processes forced by heavy precipitation, the surface water creates extensive floodplains under dense tropical forests, shapes complex topography, and shows large variations in freshwater storage and discharge. Under recent conditions, such as climate change and increased anthropogenic pressure, the basin is now facing great risk. The resulting environmental alterations require a better understanding of the overall basin’s water cycle across temporal and spatial scales. Earth observations have played a major role in supporting research in Amazon hydrology. The main challenge to be solved resides in the spatial and temporal drivers of the evapotranspiration, which are not fully understood. Better quantifying the evaporation is essential to better forecasting the impact of land change (e.g., deforestation!) in this crucial basin.
What value does machine learning add compared to traditional interpretation of remote sensing data and what challenges does it come with? How should someone whose background is not in computer science, statistics or mathematics or physics best start to use these technologies? Are you aware of any training sources you can recommend water researchers to learn how to use neural networks for their research?
Machine Learning (ML) brings the capability to draw data-driven relationships between different objects of various nature (variable, image, dictionary, etc.). It is less the recent progress in processing the data itself than the increase in computational power and shared memory that makes ML a powerful tool. The lack of interpretability is one of the main limitations, especially in the hydrology field where the deductive procedure is so important in the development of models. But our community is more and more open-minded considering a data-driven approach.
I would be careful to start using ML without a background in a related field. ML is not a magic trick and a lot of troubleshooting comes with ignorance of the inputs used. Luckily, the machine learning community is very open-source minded and tutorials in video, paper or slides format are easily accessible on the web. Examples are:
• https://www.youtube.com/user/joshstarmer
• https://atcold.github.io/pytorch-Deep-Learning/
• https://computationalthinking.mit.edu/Fall20/
For those who are interested in the use of ML in hydrology, I highly recommend the Neural Hydrology website which gathers current effort lead by a dynamic Austrian team on the use of state-of-art ML to derive knowledge in hydrology.
How do you assess your data requirements for a research project? How do you pick the data products you need? What steps are necessary to make data analysis ready?
There is no simple answer to this question since it is highly dependent on the analysis one wants to perform. Choosing a data product can be imposed if the research takes place in a consortium or project with a collaboration with a data producer. Otherwise, it can be chosen from the literature to benefit from recent improvements. When starting an analysis, pre-processing can be limited to read and project data on common space & time grids, but sometimes more processing is needed when one has to deal with missing value and too much noise in the observation. Filtering or aggregating over spatial and temporal scale is often mandatory. Finally, it is sometimes difficult to foresee our need for data before the analysis. We can develop methodologies on particular datasets and change them before the final analysis.
What do you need to innovate?
From my point of view, being up-to-date- with the literature is mandatory to innovate. Having an eye on what is currently being done is important but taking time to read past (even old) articles allows us to step back and enlarge our way of thinking in our own activity (methods, frameworks, hypothesis, etc). Also keeping your curiosity about other fields (pure remote sensing, machine learning, etc) is highly beneficial. All this literature review takes time and energy, but it is worth doing it.
As a young professional, what do you feel is missing in the current scientific debate?
Being a young researcher at this tipping point of the Anthropocene era, I miss a scientific debate on the potential involvement of research institutes at a political and civic level. Beyond the personal commitment of a researcher as citizen, discussing if the institute itself should take a particular position in the national debate to propose or react to political decisions, must be put on the table.
Another hot topic is building trans-disciplinarity research in association with other fields. I strongly believe that research in earth observations and the water cycle can mature thanks to the social sciences while adding tremendous value to this field. To tackle the big issue of the twenty-first century, researcher from wide community must meet and talk to each other.
The last point is the economy of publishing articles. Our community suffers from a system in which researchers pay huge fees to publish, the publication is a metric in our research career, more and more articles are published, the overall relevance is going down while we are asked to review dozens of articles per month. I have a feeling that this system is collapsing, and collective effort must be taken to readjust the way of publishing/sharing scientific results.
Last, but not least, what is your favourite aggregate state of water
The beauty of physics must not make us forget some pragmatic issues. If I am impressed by the coexistence of all aggregated states of water on Earth, the most important remains the drinkable (i.e., liquid) one.
