Coastal Proximity Parameter: Improving Sea Level Data Screening

1. Sea Level CCI Annual Review #1 • WP2700 Coastal sea level corrections : • Coastal Proximity Parameter Paolo Cipollini (NOC)

2. Rationale • In order to determine how ‘close to the coast’ we can reach with climate-quality data, we need to screen data and corrections versus some independent variable • ‘Distance from coast’ is crude because it does not capture the enormous variability in 1) morphology and 2) coastal conditions distance from coast(km) Example: Elba Island, Tuscan Archipelago

3. Coastal Proximity P • A new parameter to be used as independent variable • initially aim at capturing differences in coastal morphology • problem is well defined once geometry and instrumental params are fixed (orbital height, antenna beamwidth, pulse length, number of gates) and a good DEM is available • later, if possible, include peculiarities in coastal ocean (for instance peculiarities in coastal wave field) • more tentative • The P parameter is expected to improve screening (impact on the regional variability in coastal regions • with reference to distance from coast

4. Defining P • increasing from ocean  land • smaller over tips/peninsulas, larger in bays • for easier comparison with distance (which is zero at coastline), P can be remapped: • -1 over open ocean • 0 at idealized, straight coastline • 1 inland P > 0 P = 0 P < 0

5. Computing P - DEM • We need to simulate the effects of land on waveforms •  we fly a virtual altimeter over a DEM • Need a good DEM: ACE2 • ACE2 produced by De Montfort University using altimetry to correct (‘warp”) and integrate SRTM • available at various resolutions: we used the highest one • 3 arcmin, i.e.~90 m at equator

6. example1: small tall island

7. Computing P - two effects near land • power deficit due to “missing ocean” • land, even if it is at z=0, will usually have much lower backscatter than ocean (there are exceptions, but they are difficult to model!) • land returns in various gates depending on land elevation • i.e. we get echoes from land elements in various gates (before and after leading edge) depending on the land height

8. example 2: flat straight coast Land z=0.001 Ocean

9. example 2: “Dover cliff” Land z=100 Ocean

10. Computing P - parameterise the two contributions We assume them to be independent and parameterise them as ‘worst cases’, and compute, gate by gate : • power deficit due to “missing ocean”: assume sigma0land=0 for this effect where nocean is the return from the ocean elements only; and nocean0 the return from ocean elements if there was no land at all • land returns in various gates depending on land elevation: assume sigma0land=sigma0oceanfor this effect where nlandis the return from the land elements only Finally we weight P1, P2by gate position, sum them and rescale them to [-1,1]

11. example of P1,P2 and weighting

12. Example of P

13. Example of P

14. Example of P Global map of P at 0.02° resolution (CoastalProximity_v1.nc) available on SL CCI FTP server

15. Next steps • Refining P • account for peculiarities of coastal ocean, like wave field  variable spread for the gaussian weights based on wave models/wave climatology? • difficult to do it globally • Validating P • it is the independent variable – goes on the x axis • so we need to plot parameter errors, correction ‘errors’ or any equivalent cost functions against P • …and if needed, refine the algorithm to improve the effectiveness of the parameter in making errors monotonic and easy to put a threshold on

16. Validating P Error or cost function LAND distance from coast coastal proximity P

17. Comments/Suggestions Welcome!