Day 4

Detailed paper information

Back to list

Paper title DREAMing of River Basins
Authors
  1. Philippa Berry Roch Remote Sensing Speaker
  2. Jérôme Benveniste ESA - ESRIN
Form of presentation Poster
Topics
  • A7. Hydrology and Water Cycle
    • A7.01 Inland Water Storage and Runoff: Modeling, In Situ Data and Remote Sensing
Abstract text Soil Moisture derivation from Satellite Radar Altimetry has been pursued over the past ten years with a view to augmenting the observability in terms of space-time sampling, resolution and dynamic range. The basis of this technique involves crafting DRy EArth Models (DREAMs), which model the response of a completely dry surface to nadir illumination at Ku band. Initially developed over desert and semi-arid terrain, where DREAM hydrological content was primarily restricted to salars and dry river courses, DREAM crafting is now being extended to wetter areas.
This paper addresses the following questions:
1) Under what conditions can radar altimeters measure surface soil moisture? Can DREAMs be crafted over river basins?
2) What hydrology information is encoded in river DREAMs?
3) What can Sentinel-3 tell us about deployment of the new generation of satellite radar altimeters in recovery of soil moisture signals?
4) With the spatial and temporal sampling constraints of current and past altimeters, where are these data valuable?

Data from Sentinel-3A, CryoSat-2, EnviSat, Jason 1/2 and ERS1/2, together with a database of over 86000 graded River and Lake time series, are analysed to investigate the feasibility of DREAM crafting over river basins.
In this paper, results are presented over 15 regions where DREAMs have been constructed. DREAMs are crafted from multi-mission satellite altimeter data and imaging data, informed by ground truth. Current DREAMs have a spatial resolution of 10 arc seconds and a typical dynamic range of order 50dB. They are configured such that a 10dB increase in one pixel corresponds to the change from desiccated to fully saturated surface. Scaling altimeter backscatter for each mission to the DREAMs allows direct estimation of surface soil moisture.
In desert DREAM areas, small seasonal soil moisture signals were successfully retrieved. The first DREAM with significant hydrological content was developed over the Kalahari desert. Altimeter derived soil moisture estimates were generated and compared with external validation data, including the ESA CCI dataset (Dorigo et al., 2017). Good agreement was obtained.
To progress this approach, it was decided to trial the DREAM methodology to craft first generation models over the Congo and Amazon basins.

For these models, the first requirement was to mask off areas of permanent or seasonal inundation. The first test models were created over both targets using as a primary datasource multi-mission Ku band altimetry and other satellite data. Using an augmented version of the method used to identify salars, criteria were established to identify and mask river pixels. A further distinction was made to identify wetland / seasonally inundated regions, and detailed masks were produced for areas to exclude from soil moisture work. Comparing the Congo basin DREAM and its mask with independent data (Dargie et al., 2017) revealed the wealth of surface hydrology information encoded in the beta DREAM model. For the Congo beta test DREAM, 13% of the DREAM pixels are identified as river surfaces and 34% as wetland/seasonally flooded areas. It is noted that many smaller tributaries are below the current spatial resolution of the DREAM, and are classified with their surrounding terrain as wetland pixels. For the Amazon beta test DREAM, the corresponding statistics are 23% rivers and 36% wetlands.
These figures show the proportion of the models masked from soil moisture determination. Over what proportion of this surface are data retrieved by Ku band altimeters? To determine this, the masks were tested with multi-mission altimeter data. A waveform analysis system was utilised to assess echo shapes, scan for complex waveforms and flag echoes from water surfaces. Waveform shapes are classified using a system which identifies fourteen classes of echo shape corresponding to known surface types. The system is tuned for each instrument and observing mode using calibration areas of known characteristics. Multi-mission statistics show highest data retrieval over rivers and wetlands, lower over unmasked DREAM pixels. This is an expected outcome, as excluding rivers and wetlands selects for rougher topography. Varying proportions of waveforms were flagged by the system as returns affected by pools of still water throughout the model areas, with the highest proportions from the Amazon basin.
Backscatter data from all instruments show excellent agreement with the DREAMs, with cross-correlation coefficients with data from dry terrain better than 0.9. Altimeter soil moisture datasets are shown to demonstrate good agreement with external validation data. Small soil moisture signals are successfully recovered from desert regions, where other techniques encounter difficulties.
The ability of nadir-pointing altimeters to penetrate vegetation canopy gives a unique perspective in rainforest areas. Over the Amazon and Congo basins, the DREAM masking process creates detailed maps of river and wetland extents, with over 60% of the Amazon and 50% of the Congo DREAM areas identified as rivers, wetlands and seasonally flooded regions. The clear implication is that, to monitor surface water optimally in these rainforests (within the constraints of satellite orbit and repeat period), satellite altimeters should retrieve data from the majority of the underlying surface. Fortunately, analysis of past altimeter performance shows that this goal was largely achieved for the Congo and Amazon basins, particularly by ERS2 and EnviSat. Waveform analysis is found to be essential to exclude returns affected by pools of water within the altimeter footprint. Surface soil moisture time series can then be derived, and are shown to correlate with adjacent river height time series.
Very limited data acquisition from Sentinel-3A, due to the current OLTC mask, critically constrains the scope of SRAL DREAMing over all DREAMs, but results are consistent both with Cryo-Sat2 SAR and LRM mode data and results from prior missions.
In conclusion, satellite radar altimetry can provide soil surface moisture estimates wherever a DREAM can be crafted. Altimeter soil moisture estimates contribute to the datastore over river basins, providing an independent assessment of soil moisture data from other sources.
Waveform classification and soil moisture retrieval works for SRAL altimeters, with good results from Sentinel-3A, where data are available.
Data are currently being analysed to craft DREAMs over further river systems.

References
Dorigo, W.A., Wagner, W., Albergel, C., Albrecht, F.,  Balsamo, G., Brocca, L., Chung, D., Ertl, M., Forkel, M., Gruber, A., Haas, E., Hamer, D. P. Hirschi, M., Ikonen, J., De Jeu, R. Kidd, R.  Lahoz, W., Liu, Y.Y., Miralles, D., Lecomte, P. (2017).  ESA CCI Soil Moisture for improved Earth system understanding: State-of-the art and future directions. In Remote Sensing of Environment, 2017,  ISSN 0034-4257, https://doi.org/10.1016/j.rse.2017.07.001.
Dargie, GC; Lewis, SL; Lawson, IT; Mitchard, ET; Page, SE; Bocko, YE; Ifo, SA; (2017) Age, extent and carbon storage of the central Congo Basin peatland complex. Nature , 542 pp. 86-90. 10.1038/nature21048.