Benefits of future satellite gravimetry missions for characterizing extreme wet events in terrestrial water storage

Under the assumption that a warming climate leads to an intensification of the global water cycle, it is hypothesized that also the occurrence frequency and severity of extreme events such as droughts and floods will increase in the upcoming decades. GRACE/-FO observations of terrestrial water storage (TWS) have been used in the past to identify and analyse extreme events both on a global and regional scale. However, these analyses are restricted by the limited spatial and temporal resolution of current satellite gravimetry observations. Especially, flooding events tend to occur very locally and with short temporal (sub-monthly) extent, thus capturing them is challenging. Future satellite gravimetry missions, particularly the double-pair constellation MAGIC, are expected to significantly enhance the spatial and temporal resolution. In this study, we globally investigate the benefit MAGIC can achieve to detect wet extreme events using long-term (50 years) end-to-end simulations of GRACE-C and MAGIC.

The simulation environment is based on the acceleration approach and considers tidal and non-tidal background model errors as well as instrument noise of the acceleration and ranging instruments following the current MAGIC mission design studies. As input and reference, we use the daily output of a climate model (GFDL-CM4) from the CMIP6 archive that has been identified as a realistic representation of water storage evolution in previous studies. To explore the improved temporal and spatial resolution expected from the MAGIC constellation, we (i) compare extreme values derived from 5-daily gravity field simulations to those from monthly fields, and (ii) show how the weaker spatial filtering required for MAGIC has a positive influence on the detectability of extremes.

For the analysis two different approaches are exploited: One method focuses solely on the stochastic characteristics of the time series in terms of extreme value theory, evaluating the magnitude-frequency relationship of large TWS values by calculating expected return levels of wet extremes. The other approach builds on the fact that a 50-years simulation time series allows to derive statistically meaningful conclusions from directly comparing reference and simulation output on a time series level. We evaluate the time of occurrence of wet extremes on the basis of classification scores assessing correctly and incorrectly identified extreme events.