| Popis: |
Producing accurate time series of surface elevation change is vital to our understanding of ice sheet contributions to sea level rise. Such estimates are derived from a variety of sources, including automatic weather station measurements, regional climate models, and satellite altimeter observations. For satellite altimetry, a common approach for computing elevation changes is to fit a simple polynomial model to repeated measurements, from an individual satellite mission, within small (typically < 10km) ice sheet regions. However, since there now exists multiple altimeters in orbit synchronously, we instead aim to employ techniques that are more capable of fusing measurements from multiple sensors whilst also respecting any differences in uncertainties. One such technique is Kalman filtering/smoothing, which also provides opportunities for assimilating this data with ice sheet models. Four-dimensional local ensemble transform Kalman filtering (4D-LETKF) is a Bayesian approach capable of estimating large spatiotemporally chaotic systems and has already seen use in the fields of meteorology and oceanography for assimilating observations into complex forecasting models. Therefore, we explore applying a similar approach to cryosphere datasets to generate time series of surface elevation change. Our method uses satellite altimeter observations, which provide a continuous continental-scale record of ice sheet elevation measurements. The chosen state transition model is the identity matrix, predicting no change from one day to the next, allowing the observations to drive the results. Ensembles are not required because of the model’s simplicity but will be explored in the future to enable more advanced predictive models. Here, we present the method development for this novel application of the 4D-LETKF, demonstrating these techniques on CryoSat-2 radar altimetry measurements over the Greenland ice sheet from ESA’s Cryo-TEMPO Baseline-B dataset. Our next steps are to fuse other current and historical missions to produce a single collective gridded time series of surface elevation change over ice sheets. As such, we hope this study can provide the first steps towards more formal data assimilation of Earth observation data into physical ice sheet models. |
