Global high-resolution water storage simulations from the OS LISFLOOD hydrological model

Simulated terrestrial water storage (TWS) data from global hydrological models are indispensable for various geodetic applications, e.g. for the interpretation and improvement of GRACE/-FO gravity products, or for deriving time series of deformations of the Earth’s surface. So far, the Land Surface Discharge Model (LSDM) has been routinely used for such tasks at the GFZ. However, the current standard experiment of LSDM is already several years old, and many limitations are known, in particular a limited spatial resolution of 0.5°, which limits the accuracy of crustal deformation predictions close to rivers and lakes. In this contribution, we evaluate the suitability of OS LISFLOOD (https://ec-jrc.github.io/lisflood/), an open source, high-resolution hydrological rainfall-runoff-routing model, for geodetic purposes.

We compare the performance of various global OS LISFLOOD model runs for the time period 2000 – 2022 against the current LSDM configuration. In addition to two OS LISFLOOD model implementations, which differ in their spatial resolution (0.1° and 0.05°) and their input land surface parameter data set, we also explore a number of high-resolution (0.05°) model runs with respect to the influence of the soil depth on simulated TWS. Model results are validated against mass anomalies from GRACE and GRACE-FO on different spatial and temporal scales. Furthermore, to demonstrate the benefit of the higher spatial resolution of OS LISFLOOD, we utilize data from around 500 ground based GNSS stations to validate the models’ performance regarding mass-induced loading.

We find that OS LISFLOOD significantly outperforms LSDM in many regions, especially on interannual time scales, in terms of various validation metrics. Analyzing the different OS LISFLOOD runs reveals a large impact of the choice of soil depth and initialization technique on simulated TWS. With its daily temporal and high spatial resolution the new OS LISFLOOD-derived TWS time series have a wide range of potential applications, including downscaling and vertical disaggregation of GRACE/-FO TWS and in the background modeling during GRACE/-FO data processing.