A GLOBAL ASSESSMENT OF THE RA-2 PERFORMANCE OVER ALL SURFACES Berry, P.A.M., Smith, R.G. & Freeman, J.A. EAPRS Laboratory, De Montfort University, Leicester, LE9 1BH, UK ABSTRACT The EnviSat RA-2 has collected a huge volume of echoes from the earth's surface. Its unique modeswitching capability has enabled it to remain in lock even over mountainous terrain, and has resulted in the most complete sampling of the earth's surface by a satellite radar altimeter. This paper presents results from the global analysis of the RA-2 echoes over all surfaces, including a comparison of the RA-2 performance with that of TOPEX, Jason-1 and ERS-1/2. The results show superb echo capture even over extreme terrain, a very good outcome over most land surfaces but a sub-optimal performance over inland water. The interplay between surface targets, tracker response and autonomous mode switching is illustrated to explain the RA-2 behaviour. The results have implications for future instrument design and demonstrate very clearly the overall excellent performance of the RA-2. 1. INTRODUCTION The EnviSat RA-2 has been designed with the unique capability to change its mode of operation dynamically in response to the changing patterns of echo recovery from the underlying surface. This unique capability has allowed the instrument to acquire successfully echoes over almost all of the Earth s land surface overflown by the instrument. To assess the performance of the EnviSat RA-2 over land, a detailed study has been made of both echo retrieval and height retrieval globally. 2. GLOBAL ECHO ANALYSIS The percentage of echoes successfully acquired by ERS2 [1] and the EnviSat RA-2 [2] were analysed globally. A typical result for the echo capture of ERS2 is shown in Fig. 1, for cycle 27. Whilst the RA-1 was highly successful at acquiring waveform data over much of the earth s land surfaces, the limits of capability are clearly seen, with greatly reduced echo capture over mountainous terrain. Note that the reduced capture over coastal zones is a consequence of the hard-wired ocean/ice mode mask, which forced the instrument to change into ocean mode before reaching the coast, in order to preserve capture of ocean data at full resolution. In comparison, Fig. 2 shows the corresponding analysis for a typical cycle of EnviSat RA-2. Comparison with Fig. 1 illustrates the enhanced echo capture in rough terrain, with only high mountain ranges significantly impeding the instrument s ability to acquire data. This very clearly illustrates the advantage of the dynamic mode-switching capability of the RA-2 in Ku band. Figure 1 Global Accepted Echoes from Cycle 27 Proc. Envisat Symposium 2007, Montreux, Switzerland 23 27 April 2007 (ESA SP-636, July 2007)
Figure 2 Global Echo Retrieval from the RA-2 Ku band for Cycle 30 In order to intercompare the performance of the two instruments in a more quantitative manner, one year of data was analysed for each mission, the resulting statistics normalised for difference in sampling rate, and the grids differenced to illustrate the comparative performance of the RA-2 in Ku band against the RA-1. The result is shown in Fig. 3. Here, the red edging to the continents is again a consequence of the RA-1 mode switching mask, which forces the instrument into ocean mode close to coasts with consequent loss of echo capture. The poor echo retrieval over many islands is also due to the same cause. The rhomboid shapes over the Himalayas delineate an RA-1 calibration zone. As expected, enhanced echo retrieval is seen from the RA-2, especially over mountainous regions. However, one immediately evident and consistent feature in this plot is the lower capture of echoes by the RA-2 in the presence of inland water. This is a consistent characteristic of the operation of this instrument and a cause for concern for inland water height retrieval (discussed later). Figure 3 Normalised Difference in Echo Capture between ERS1 and EnviSat Ku band for One Year of data
Extending the analysis to S band, Fig.4 shows the echo retrieval for a typical cycle. The yellow orbits are a consequence of the S band anomaly [3], and should be ignored for the purposes of this comparison. Overall, the performance in S band shows many similarities with that of the RA-1, with quite good echo capture over all but extreme terrain. Figure 4 Percentage of Echoes Acquired by RA-2 S band for Cycle 18 3. MODE ANALYSIS The echo capture was re-analysed by mode for the RA-2 Ku band; typical results are presented in the following plots. Firstly, Fig. 5 shows the typical percentages of echoes captured in 320 MHz mode. It is clear that over any flat or relatively benign terrain, the RA-2 is switching to ocean mode, with some exceptions over inland water bodies. In contrast, Fig 6 shows the percentage of echoes acquired in 80MHz mode. Whilst these are concentrated in areas of rougher terrain, which is, of course, an expected result, there is an anomalously high percentage in Northern Europe being captured in this mode over very flat terrain. Analysis of waveform sequences and profile plots shows that this effect is due to sub-optimal performance of the RA-2 in the presence of surface liquid water. The instrument off-ranges to bright quasi-specular targets, loses the leading edge of the nadir target, switches from 320 MHz mode to 80MHz mode, re-acquires the waveform leading edge, and switches back to 320 MHz mode. This introduces an annual variation in the percentages of data acquired in the three modes of operation, and illustrates both the sub-optimal acquisition of inland water targets and the great advantage of the RA-2 mode-switching capability, which allows the instrument to recover the inland water signal and retrieve useable waveform data.
Figure 5 Percentage of Echoes captured in 320MHz 'Ocean' mode for cycle 30 Figure 6 Percentage of Echoes captured in 80MHz 'Ice' mode for cycle 30 Figure 7 Percentage of Echoes captured in 20Mhz mode from Cycle 30
Fig. 7 shows the data acquired in 20MHz mode. As expected, these data are clustered over extreme terrain, with a small percentage acquired over unexpectedly benign terrain. Whilst, clearly, the vertical precision of heights derived from waveforms captured in this mode is extremely limited, the mode switching capability does allow the RA-2 to return to 80 MHz mode over many mountainous regions, acquiring the flatter parts of the terrain (typically the mountain valleys) in this mode, from which useful height information can be extracted. 4. GLOBAL SRTM COMPARISON In order to assess the capability of the RA-2 altimeter waveforms to derive heights over land Jason-1, TOPEX, ERS-2 and EnviSat Ku-band data were retracked using the Expert Systems approach [4] and the resulting global land height datasets were compared with the SRTM dataset [5,6] for all locations within the latitude limit of the SRTM coverage. Summary global results of this huge analysis are presented here to illustrate the capability of the RA-2. the distribution. Jason-1 in fact acquires more valid data than TOPEX over land; again, the difference between these two instruments is expressed primarily in the wings of the distribution. This is an expected consequence of the echo retrieval over rough terrain; in many cases the altimeter echo does not return a mean height but rather a height from one part of the underlying surface, generally the flatter part of the terrain (hence the asymmetrical histogram shapes). 5. DISCUSSION The Envisat RA-2 shows excellent recovery of waveforms over both land and ocean, successfully retrieving data for the first time over mountain ranges. A global comparison of retracked heights from the RA- 2 Ku band, ERS2 RA-1, TOPEX and Jason-1 with the SRTM confirms the very good performance of the RA-2 c.f. other missions and demonstrates generally excellent height retrieval. S band data show greater than expected echo retrieval. However, the results of this huge global analysis also show that there is a cost to the optimisation of data recovery over ocean by forcing the RA-2 into high resolution mode, in that some data are acquired over rough terrain in ocean mode, which typically results in capture of only the slope component of the terrain response rather than the nadir return. Of greater concern is the sub-optimal performance over inland water, where the instrument switches to high resolution mode and loses the leading edge. However, the results also show that the mode-switching capability of the RA- 2 allows it to maintain lock on the underlying surface and successfully recover data; generally this results in the loss of only part of the river and lake profiles, although over smaller rivers the outcome is less favourable. 6. REFERENCES Figure 8 Histogram of EnviSat Ku band, Topex, ERS-1 and Jason-1 Height Retrieval c.f. SRTM dataset normalised for orbit repeat and sampling frequency Fig. 8 shows typical global histograms of the difference between the heights derived from the RA-2 Ku band, ERS2, Topex and Jason-1, with all counts adjusted to allow for differences in orbit repeat period and alongtrack sampling. As expected, both the RA-1 and RA-2 show very significantly better performance over land than Topex and Jason-1. The enhanced echo capture over rough terrain of the RA-2 is seen in the wings of [1] Capp, 2001 P. Capp, Altimeter Waveform Product ALT.WAP compact user guide, Issue 4.0, PF-UG-NRL AL-0001, Infoterra Ltd, UK (2001). [2] J. Benveniste., et al. ENVISAT RA-2/MWR Product Handbook, Issue 1.2, PO-TN-ESR-RA-0050, European Space Agency, Frascati, Italy, 2002. [3] S. Laxon and M. Roca, ENVISAT RA-2: S-BAND PERFORMANCE, Proceedings of the ENVISAT Calibration Workshop, Noordwijk, September 2002. [4] P.A.M, Berry, A. Jasper and H. Bracke, Retracking ERS-1 altimeter waveforms over land for topographic
height determination: an expert system approach, ESA Pub. SP-414, 1, 403 408, 1997. [5] USGS, 2005 USGS., Shuttle Radar Topography Mission (SRTM) Finished products, U.S. Geological Survey. (2005) URL: http://edc.usgs.gov/products/elevation/srtmbil.html, (last date accessed: 30 April 2007). [6] P.A.M, Berry, J.D. Garlick, R.G., Smith, 2007; Near-global validation of the SRTM DEM using satellite radar altimetry. Remote Sensing of Environment, Vol 106 Issue 1 Pages 17-27, DOI: 10.1016/j.rse.2006.07.011 ACKNOWLEDGEMENTS The authors wish to thank ESA for supplying ERS1, ERS2 and EnviSat data, NASA for supplying Topex data, AVISO for the Jason-1 dataset and USGS for the SRTM dataset.