|I would like to congratulate the authors for their hard work to improve the original manuscript. All the concerns and minor comments in my previous review have been considered. I believe the authors made a serious effort to reply to any request of clarification and review. Also the English text and fluency has been improved in many places.|
My only complaint has to do with three important points which should be easy to address:
First, authors promised to provide information about the status for in-situ data of lakes as they reported it for rivers only. However, I don’t see this point addressed in the text. I thin that it is important to clarify. Readers might interpret that the problem is only in rivers and that we don’t need altimetry in lakes if they are already well monitored with in situ water gauges. My recommendation is that authors provide a reference or add an explanatory sentence that avoids misunderstanding in readership.
Second, authors state in response to my specific point about specular reflections in small rivers that the reason for these waveforms is land contamination. I think that they do not provide any evidence that this statement is correct. They are looking at the world through 18 Hz burst level waveforms that are an incoherent 300 m blur, whereas they should see with 3 m complex individual echoes (e.g. available for Envisat mission) correlated with 1 m satellite imagery. Authors, in theory, might be right in stating that a specular water level could be contaminated with a specular land echo. However since both are rare (in percent of ground cover) it would be very unlikely that a specular land return is adjacent to a specular water return. The proof that water is specular is in the Doppler processing of Individual Echoes. I recommend authors to refer the work titled “The Near-Specular Altimeter Waveforms of Small Inland Water Bodies” by Abileah et al. presented at last Coastal Altimetry Workshop held in Lake Constance (Germany) last year (see presentation at http://www.coastalt.eu/files/konstanzworkshop14/pres/11_Abileah_et_al._8th_CAW-oral_paperL.pdf). The authors should mention this insight in the background. As the nadir point approaches, passes over, recedes from water, the Doppler transitions (looking at phase) go from negative to zero to positive that is absolute proof. The computed velocity turns out to be exactly the satellite's speed of approach and receding from the location with water. In summary, this paper is not about waveforms and retracking, but it is about averaging values after retracking. The mentioning of waveforms and retracking is in the background. However, the analysis of radar returns is very important to understand disagreements and poor performances in the case-studies of validation. I also find difficult to believe that hooking effect is a real problem. In fact authors state “problem is avoided or 'worked around' by using only nadir measurements”. The part of the revised manuscript between rows 77-95 is substantially based on old view of burst level waveforms. Coherent individual echo analysis leads to another view, as reported in the above presentation. I strongly recommend that authors rework this part and also take into account the new view (from Doppler processing of Individual Echoes) as well as say something about the prospect of exploiting future missions. Envisat had the limitation of collecting echoes for 1-sec every 60 sec. Due to mission design, Cryosat-2 and Sentinel-3 (soon in orbit) are also not continuous in the collection of echoes. On the opposite, Jason-CS/Sentinel-6 will collect continuously Individual echoes (see statement in
Third, authors updated the method of estimating the precision of the single heights. It is based on the analysis of the along-track scatter of the measurements. They estimate the median of the water heights within a box. Then the median height is subtracted from the current water height. The absolute value of the difference is then used as “error” of the single measurement. The process is repeated moving the box forward. The fact to have errors associated to each measurement is a requirement of the Kalman filter. However, the authors didn’t reply to my comment that suggested using consecutive points to measure the noise for a specific target. Then to represent noise authors could use different methods (e.g., histogram). This methodology applied to all validation sites gives an overview where DAHITI processing performances of 18 Hz data are more degraded. I warmly encourage authors to provide a measure of the noise for each validation site.
Therefore, the current manuscript can be accepted for publication after the above additional comments are addressed. I would like to see their answers before publishing.