Day 4

Detailed paper information

Back to list

Paper title Using GRACE-based assimilation for the revision of modeled water – vegetation dynamics
  1. Helena Gerdener University Bonn Speaker
  2. Jürgen Kusche University of Bonn
  3. Kerstin Schulze University of Bonn
  4. Gohar Ghazaryan Leibniz Centre for Agricultural Lanscape Research
  5. Olena Dubovyk Remote Sensing Research Group
Form of presentation Poster
  • A7. Hydrology and Water Cycle
    • A7.01 Inland Water Storage and Runoff: Modeling, In Situ Data and Remote Sensing
Abstract text Restricted access to freshwater and crop failure lead to disastrous consequences, for example economic losses, hunger and death. Thus, ensuring food production and sufficient water supply for crop production (or agriculture) is a highly relevant topic for the population all over the world. Soil moisture is the main driver for providing water resources for agriculture and vegetation but in semi-arid or arid regions it is becoming more important to derive water from surface water bodies or stored in groundwater. These surface and subsurface water storages are either monitored with in-situ data, which have a long record history, or are simulated in models, which provide global simulations with a good spatial resolution (~50 km). However, the in-situ data are not spatially explicit and very sparse and, thus, cannot cover each climate regime and the models encounter problems with uncertainty in the forcing data and model assumptions.

In the last decades, the use of remote sensed data has enabled observation of the water from space. GRACE (Gravity Recovery And Climate Experiment) and its successor GRACE-FO were and are so far the only satellite missions that observe the sum of surface and subsurface water with global resolution. However, GRACE(-FO) have a coarse spatial resolution (~300 km) and only sense the vertically aggregated sum, so called total water storage anomalies (TWSA); hence a further separation into the different water compartments is needed. Therefore, we integrate GRACE into a hydrological model via assimilation to improve the model’s realism while spatially downscaling and vertically disaggregating GRACE.

In this study, we assess signatures and subsignals found in models using observation (via assimilation) based storages and vegetation (via remote sensing) measures derived from MODIS (Moderate Resolution Imaging Spectroradiometer). In a case study, we interrogate two main processes (measured at peak times) in South Africa for 2003 to 2016: 1) The precipitation-storage dynamics, i.e. the dynamics of the pathway from precipitation to replenished soil moisture, surface water and groundwater and 2) the storage-vegetation dynamics, i.e. the pathway from the corresponding storage to vegetation growth (by using in Leaf Area Index and Actual Evapotranspiration).

Generally, we found that the amount of water that refills the storages is often overestimate in the modeling and the duration for this process is often shorter compared to the observations. For example, we found in the modeling that in general the annual peak of groundwater lags the annual precipitation peak by 3 months, while the observations identify a 4-month lag. For the storage-vegetation dynamics we also notice an overestimation of the amount of water that contributes to vegetation growth, and an over- or underestimation of the duration for this process strongly depends on the considered storage. Our study concluded that the model did not correctly capture the precipitation-water storages-vegetation dynamics and it would be impossible to conclude that from using only GRACE TWSA data, without data assimilation. In the future, our findings will be highly relevant for modelers to and can be used to improve the model structures.