Abstract. This study demonstrates that LDAS-Monde, a global and offline Land Data Assimilation System (LDAS), that integrates satellite Earth observations into the ISBA (Interaction between Soil Biosphere and Atmosphere) Land Surface Model (LSM), is able to detect, monitor and forecast the impact of extreme weather on land surface states. LDAS-Monde jointly assimilates satellite derived Earth observations of surface soil moisture (SSM) and Leaf Area Index (LAI). It is run at global scale forced by ERA5 (LDAS_ERA5), the latest atmospheric reanalysis from the European Centre for Medium Range Weather Forecast (ECMWF) over 2010–2018 leading to a 9-yr, ~ 0.25° × 0.25° spatial resolution reanalysis of Land Surface Variables (LSVs). This reanalysis is then used to compute anomalies of land surface states, in order to (i) detect regions exposed to extreme weather such as droughts and heatwave events and (ii) address specific monitoring and forecasting requirements of LSVs for those regions. In this study, LDAS_ERA5 analysis is first successfully evaluated worldwide using several satellite-based datasets (SSM, LAI, Evapotranspiration, Gross Primary Production and Sun Induced Fluorescence), as well as in situ measurements (SSM, evapotranspiration and river discharge). The added value of assimilating the soil moisture and LAI is demonstrated with respect to a model simulation (openloop, with no assimilation). Since the global LDAS_ERA5 has relatively coarse resolution, two higher spatial resolution experiments over two areas particularly affected by heatwaves and/or droughts in 2018 were run: North Western Europe and the Murray-Darling basin in South Eastern Australia. These experiments were forced with ECMWF Integrated Forecasting System (IFS) high resolution operational analysis (LDAS_HRES, ~ 0.10° × 0.10° spatial resolution) over 2017–2018, and both openloop and analysis experiments compared once again. Since the IFS is a forecast system, it also allows LDAS-Monde to be used in forecast mode, and we demonstrate the added value of initializing 4- and 8-day LDAS-HRES forecasts of the LSVs, from the LDAS-HRES assimilation run, compared to the openloop experiments. This is particularly true for LAI that evolves on longer time space than SSM and is more sensitive to initial conditions than to atmospheric forcing, even at an 8-day lead time. This confirms that slowly evolving land initial conditions are paramount for forecasting LSVs and that LDAS-systems should jointly analyse both soil moisture and vegetation states. Finally evaluation of the modelled snowpack is presented and the perspectives for snow data assimilation in LDAS-Monde are discussed.