Freshwater ecosystems are among the most threatened ecosystems on Earth. At the same time, water demand for food is projected to increase with projected increase in population and diet shift putting part of the population under pressure in terms of food security. These projections are likely to be exacerbated by climate change. Over the past decades, irrigated areas have nearly tripled to meet actual human food requirements. Today, 40% of food production comes from irrigated production and about 30% from irrigated areas. This increasing share of irrigated production has come at the expense of freshwater ecosystems and river health. About half of the rivers have been fragmented and altered via the constructions of dams and reservoirs and via diversion of river flow to irrigated fields. Furthermore, water demand for industry, household and hydropower is predicted to increase and competition between water sectors will intensify. Under actual water competition, water availability for freshwater ecosystems has often been neglected.Over the past decade, awareness was given to define planetary boundaries for natural resources especially freshwater ones. While irrigation withdrawals and industries and household withdrawals already reach respectively about 2600 km3 yr-1 and 1000 km3 yr-1, planetary boundaries for freshwater have been defined to 4000 km3 yr-1. With the expected rise in water demand for food and industries, freshwater boundaries are likely to be exceeded in the coming decades and it is urgent to define global water availability and demand with accurate time and spatial resolutions. More specifically, it is necessary to develop a method that enables the calculation of water demand for freshwater ecosystems known as “Environment Flow Requirements” (EFRs). EFRs were often neglected in global assessments and/or defined with annual proxies.The overall objectives of this thesis were to redefine global water demand for freshwater ecosystems (EFRs) and set these last as a priority in global integrated assessments. For that, it was necessary to design a robust methodology that can be easily implemented in Global Hydrological Models (GHMs) and in global integrated assessments. In chapter 2, existing global and local Environmental Flow (EF) methods were reviewed. Three methods were selected among existing global methods, including the Smakhtin method, which is based on a combination of annual quantiles and proxies of annual flow, the Tennant method, which is based on annual proxies of flow, and the Tessman method, which is based on monthly proxies of flow. Two other methods were designed for this study: the Variable Monthly Flow (VMF) method, which is based on the allocation of the percentage of monthly flow to the environment and the Q90_Q50 method, which is based on the allocation of flow quantiles. These methods were compared with 11 local case studies from different ecoregions, for which EFRs have been defined locally with ecological and hydrological data collection. The VMF method showed the best performance against local case studies and demonstrated easiness of use and validation with different flow regime types. Among the five global EF methods, EFRs represent 20 to 50% of mean annual flow to maintain EFRs in “fair” ecological conditions.In chapter 3, the concept of “Environmental Flow (EF) deficit” was designed. It represents the lacking flow to meet EFRs. EF deficit was defined on a monthly basis at 0.5 deg. The originality of this study is that the origin of the deficit was characterized by the natural deficit and the anthropogenic deficit. Natural deficit is defined when EFRs are not met due to natural climate variability and anthropogenic deficit is defined when EFRs are not met due to water extractions for irrigation or other users. The frequency, timing and magnitude of each deficit were also calculated at global scale. The EF deficit was also studied for 23 river basins, which are located in different ecoregions, and it was shown that flow regime type, origin of deficit, magnitude of deficit and level of flow alteration were correlated. Perennial rivers such as the Congo River showed only natural deficit while very altered river such as the Godavari river showed high respective natural and anthropogenic deficit. In chapter 4, we set EFRs as a priority user in the global vegetation model LPJmL. It was shown that to sustain EFRs in “fair” ecological conditions, irrigation water use should be reduced by 30%, which would lead to 30% less food coming from irrigated area and a total of 5% loss in food production. Calorie loss per capita was really high in developing countries where population density is high such as in South-East Asia. This loss in food production can however be compensated by an increase of 50% in irrigation use efficiency.In chapter 5, we used an economic optimization model (GLOBIOM) to study future global change including different constrains of EFRs. It was shown that, under future climate change (RCP 8.5) and socio-economic development (SSP2), international trade should be increased by 15% to compensate for EFRs implementations compared to a business-as-usual scenario. The positive outcome is that it was demonstrated that food and water security for humans and ecosystems can be sustained with three levees: use of trade (+15%), conversion of irrigated land to rainfed land (60Mha) in South Asia and expansion of rainfed land into natural area in Latin America.In the chapter 6, we reviewed and analyzed each chapter as an ensemble. The new development of the VMF method is acknowledged thanks to its application in all chapters of this thesis and in many other global assessments. Among them, two studies redefined the freshwater planetary boundaries at 2,800 km3 yr-1 which is lower than previous estimates defined by Rockstrom et al. (2009). This thesis allowed the inclusion of EFRs in global integrated assessments with refined temporal and spatial scales and water demand for ecosystems are now recognized and acknowledged. The limitations of the VMF method are also discussed such as its weakness to be compatible with inter-annual studies considering extreme events such as floods and droughts. Further data collection on eco-hydrological relationships should be organized and harmonized at global scale to further improve EFRs at global scale. Characterization of EF deficit with differentiation of the anthropogenic and natural deficit can be used as a tool to prioritise actions in terms of river restoration/protection. In face of meeting future SDGs, we highlighted the complexity in meeting food and water security for humans and ecosystems. Competition between different water sectors already exist and require local, regional and international consensus to satisfy all water users while safeguarding water availability for freshwater ecosystems. For that, future improvement in agriculture and water management is fundamental to provide future sustainable water access to humanity.