Abstract's details
Modelling the vertical dependence of the Wet Path Delay: application in satellite altimetry over coastal and inland waters
CoAuthors
Event: 2019 Ocean Surface Topography Science Team Meeting
Session: Instrument Processing: Propagation, Wind Speed and Sea State Bias
Presentation type: Type Oral
Contribution: PDF file
Abstract:
The accuracy of water surface height measurements from satellite radar altimetry, depends on the accuracy of all terms involved in its computation, namely the effect due to the Wet Path Delay (WPD). With a high spatial-temporal variation, the most accurate way to measure the WPD over open-ocean is from on-board Microwave Radiometers (MWR) measurements. Satellite altimetry has been used over coastal and inland waters, however the WPD retrievals from MWR become invalid and cannot be used over these regions of interest. Alternative WPD sources for these zones can be Global Navigation Satellite Systems (GNSS) and Numerical Weather Models (NWM), e.g. those from the European Centre for Medium-Range Weather Forecasts (ECMWF).
Designed for applications over the ocean, altimetric missions are mainly focused on the sea surface and, for this reason, all corrections such as MWR-derived WPD must refer to the sea level. On the contrary, WPD derived from an NWM are computed at the level of its orography and the corresponding path delays from GNSS are available at each station height. Due to these different vertical references, the modelling of the altitude dependence of the WPD is a crucial step to better combine these different WPD sources, particularly important for altimetry application over coastal and inland waters. Therefore, expressions to reduce the various WPD to sea level (over coastal zones) and to water body height (over continental waters) are required. The most commonly used expression for the WPD altitude reduction has been developed by Kouba (2008), however it has some limitations, since it considers the same altitude reduction, irrespective of the WPD spatial-temporal variation.
The focus of this study is the modelling of the altitude dependence of the WPD, aiming at developing improved expressions to account for its complex 4-D variability. For this purpose, the latest ERA5 reanalysis from ECMWF is used. It provides hourly atmospheric parameters at 0.25°x0.25° spatial sampling on 137 vertical levels. To be used in this modelling, global WPD vertical profiles have been computed for a time span of 4 years.
This study is performed in three main steps. First, the errors introduced when using the Kouba expression are quantified in terms of their spatial distribution. For this purpose, two WPD are derived from ERA5 at pressure levels: i) from single level (orography) parameters and then reduced to the upper pressure levels using the Kouba expression (the so-called 2-D approach); ii) from numerical integration of 3-D parameters (temperature and specific humidity), from the top of the atmosphere down to the orography, at each pressure level (the so-called 3-D approach). The comparison between these two WPD profiles reveals differences with RMS values up to 3.2 cm, with larger differences at low latitudes, mainly attributed to errors in the height reduction performed using the Kouba formula.
Secondly, improved expressions have been developed in this study (so-called UP expressions), considering spatial and temporal dependent coefficients. As performed for the Kouba expression, the same assessment using ERA5 data (for a period not used in this modelling) has been performed for the UP expressions. Considering only regional-dependent coefficients in the UP modelling, for the region where Kouba reduction has an RMS error of 3.2 cm, when using the UP reduction this value is 1.2 cm, corresponding to an RMS error decrease of 2 cm. These results also show that the inclusion of temporal-dependent coefficients in this modelling can lead to an RMS error decrease larger than 1 cm in some regions.
Finally, the last step of this study concerns the validation of the various expressions using in-situ data: radiosondes (RS) and GNSS. At each RS site two WPD vertical profiles are considered: the first one computed from 3-D RS data and the other one considering only the WPD at the lowest RS vertical level, further reduced to the upper levels using the different altitude reductions (Kouba and UP). A similar validation is performed at GNSS sites by comparing GNSS-derived WPD at station height and those derived from ERA5 using the 2-D approach further reduced to the same station height. Results with RS and GNSS data show that the RMS of the WPD differences decrease by more than 1 cm, when UP modelling is used instead of Kouba, thus quantifying the impact of the new expressions in the derivation of accurate absolute water levels in these regions.
Designed for applications over the ocean, altimetric missions are mainly focused on the sea surface and, for this reason, all corrections such as MWR-derived WPD must refer to the sea level. On the contrary, WPD derived from an NWM are computed at the level of its orography and the corresponding path delays from GNSS are available at each station height. Due to these different vertical references, the modelling of the altitude dependence of the WPD is a crucial step to better combine these different WPD sources, particularly important for altimetry application over coastal and inland waters. Therefore, expressions to reduce the various WPD to sea level (over coastal zones) and to water body height (over continental waters) are required. The most commonly used expression for the WPD altitude reduction has been developed by Kouba (2008), however it has some limitations, since it considers the same altitude reduction, irrespective of the WPD spatial-temporal variation.
The focus of this study is the modelling of the altitude dependence of the WPD, aiming at developing improved expressions to account for its complex 4-D variability. For this purpose, the latest ERA5 reanalysis from ECMWF is used. It provides hourly atmospheric parameters at 0.25°x0.25° spatial sampling on 137 vertical levels. To be used in this modelling, global WPD vertical profiles have been computed for a time span of 4 years.
This study is performed in three main steps. First, the errors introduced when using the Kouba expression are quantified in terms of their spatial distribution. For this purpose, two WPD are derived from ERA5 at pressure levels: i) from single level (orography) parameters and then reduced to the upper pressure levels using the Kouba expression (the so-called 2-D approach); ii) from numerical integration of 3-D parameters (temperature and specific humidity), from the top of the atmosphere down to the orography, at each pressure level (the so-called 3-D approach). The comparison between these two WPD profiles reveals differences with RMS values up to 3.2 cm, with larger differences at low latitudes, mainly attributed to errors in the height reduction performed using the Kouba formula.
Secondly, improved expressions have been developed in this study (so-called UP expressions), considering spatial and temporal dependent coefficients. As performed for the Kouba expression, the same assessment using ERA5 data (for a period not used in this modelling) has been performed for the UP expressions. Considering only regional-dependent coefficients in the UP modelling, for the region where Kouba reduction has an RMS error of 3.2 cm, when using the UP reduction this value is 1.2 cm, corresponding to an RMS error decrease of 2 cm. These results also show that the inclusion of temporal-dependent coefficients in this modelling can lead to an RMS error decrease larger than 1 cm in some regions.
Finally, the last step of this study concerns the validation of the various expressions using in-situ data: radiosondes (RS) and GNSS. At each RS site two WPD vertical profiles are considered: the first one computed from 3-D RS data and the other one considering only the WPD at the lowest RS vertical level, further reduced to the upper levels using the different altitude reductions (Kouba and UP). A similar validation is performed at GNSS sites by comparing GNSS-derived WPD at station height and those derived from ERA5 using the 2-D approach further reduced to the same station height. Results with RS and GNSS data show that the RMS of the WPD differences decrease by more than 1 cm, when UP modelling is used instead of Kouba, thus quantifying the impact of the new expressions in the derivation of accurate absolute water levels in these regions.