Monitoring and forecasting drought through the assimilation of satellite water observations

Abstract

Drought poses the greatest threat to freshwater availability and food security, affecting larger areas for longer periods than any other natural hazards. In many regions, droughts increase in frequency and severity due to climate change. As a slow developing natural disaster, better estimates of water availability can be valuable for forecasting droughts and their impacts on ecosystem, agriculture and food security. With accurate knowledge of root-zone soil water and groundwater dynamics, effective planning of water resources and agriculture can be made months in advance. However, the simulated root-zone soil moisture and groundwater are often highly uncertain due to the unpredictable nature of soil water and groundwater dynamics caused by human activities such as water extraction and irrigation. Ground-based and remotely sensed measurements of water content are often limited in both spatial coverage and temporal resolution. Therefore, quantifying the change of water availability and its impacts on vegetation conditions at large scales remains largely unexplored.

In my study, contrasting satellite observations of water presence over different vertical domains were assimilated into a global water balance model, providing unprecedented accuracy of soil moisture profile and groundwater storage estimates. The water availability at different depths observed from soil moisture (SMOS) and space gravity (GRACE) missions provides an opportunity to separate total water storage vertically into different layers through data assimilation. However, combining these two data sets is challenging due to the disparity in temporal and spatial resolution at both vertical and horizontal scales. SMOS provides global high spatial and temporal resolution (i.e. 40 km2, 3-day) near-surface (0-5cm) soil moisture estimates from microwave brightness temperature observations. In contrast, the GRACE mission provides accurate measurements of the entire vertically integrated terrestrial water storage column, but it is characterized by low spatial and temporal resolutions (i.e. 300km x 300km, monthly). An ensemble Kalman smoother based global data assimilation system was developed to resolve the discrepancy between model and observations in space and time.

The use of data assimilation integrates these two measurements to effectively constrain model simulations and to accurately characterize the vertical distribution of water storage. Compared with model estimates without the assimilation or single-variant assimilation, joint assimilation typically led to more accurate soil moisture profile and groundwater estimates with improved consistency with in situ measurements. The improved water storage estimates integrated over different depths were used to determine the vegetation-accessible storage in association with vegetation growth and surface greenness. Accessible storage reflects a combination of vertical root distribution and soil properties, and its spatial distribution correlates with aridity and vegetation type. Skillful forecasts of vegetation conditions are achievable several months in advance for most of the world's drylands, which offers exciting new prospects for the improvement of drought early warning systems to help reduce human suffering and economical and environmental damage.