Here is our response to the various comments raised by the referees.

First referee:

>The article is a very original and concise description of the scientific case
>for advanced gravitational wave detectors with sensitivities as those of the
>proposed Einstein Telescope. I recommend publication of the article, although
>I have several suggestions below that the authors may want to incorporate. 
>
>
>Page 4, 2nd paragraph: Authors should mention they use geometrical units or
>use correct dimensional constants for expressions relating frequency of
>gravitational waves, mass and radius of compact objects.
>

We have added the following footnote just before defining
the frequency of gravitational waves of a compact binary:

We use a system of units in which the speed of light and the gravitational 
constant are both equal to unity, $c=G=1.$

>Also, it should be mentioned early in this section that only systems with
>time-varying quadrupoles will produce gravitational waves (as opposed to
>simple isolated black holes) - this seems to be assumed by the authors but may
>not be known to the general reader.

Agreed. We changed the relevant sentence to:

ET, for that matter any \ac{GW} detector, is sensitive to compact objects with 
time-varying quadrupole moment.  

>
>Page 5, Figure 1: the text and the caption should explain what distance is
>used in the horizontal axis (there are two common scales used, horizon
>distance and average distance). 

It is neither. What is plotted is the cumulative number of 
events expected to be detcted as a function of distance. We
have modified the caption as follows to clarify what is being
plotted:

\caption{\label{fig:network rate}The plot shows the cumulative 
number of compact binary events expected  to be detected by a 
network within a given distance, for three archetypal 
compact binaries and four different advanced detector networks. The
curves flatten (and stay constant) upon reaching the {\em horizon 
distance} of the network, the distance beyond which a network
cannot detect signals with the desired signal-to-noise ratios. 
See the text for further details.}

>
>Page 5, last paragraph: Given experience with current detectors, it is
>questionable whether SNR is the right figure of merit - the authors may want
>to point to references in previous LIGO/Virgo data analyses and quote an
>expected false alarm rate used for detection, rather than an SNR figure
>(although it is SNR the only figure that can be used to infer detection rates
>from instrument spectral sensitivities). The authors should note that the SNR
>threshold chosen in this article is different than the one used for rates in
>Ref. 9, and lead to different estimates, which may be confusing to the reader
>(the differences in redshift corrections for BBH are mentioned, but not the
>differences in SNR criteria). What is mentioned here but should perhaps be
>emphasized more is the well known but not well advertised fact that each
>detector in the network increases the detection rate by about a factor of 2
>(for example, it's not mentioned in Ref 9). 

We have added the following two sentences. Note, however, that the rate 
does not increase by a factor of 2 each time a detector is added.

Note that the criteria used in computing the detection rate here
is different from that used in Ref.\,\cite{Abadie:2010cf}. 

The detection rate increases by a factor of three as we go
from two to three detectors and by little less than a factor of two as
we go from three to four detectors. 

>
>Finally, in Page 6, last paragraph: it is obvious that a network of advanced
>detectors like ET will provide much of the science described - it is not so
>clear what the single ET detector may be able to provide, which I think is
>still very significant. The authors may want to comment on this in the
>different sections.
>

This is something we haven't really assessed in great detail. We only have
a hunch about what might be possible. We have added the following discussion 
just before Section 3.1.

While a single-site third generation
detector might achieve some of the scientific goals discussed in this review,
any problem that necessitates a knowledge of the position of the source on 
the sky and its distance from the Earth will require a network of detectors.
For example, while strong field tests of \ac{GR} could be performed with
events detected in a single-site ET, measurement of cosmological parameters
would require a network of detectors. Likewise, testing the propagation speed
of gravitational waves relative to electromagnetic waves from a supernova,
would not require a network of detectors; optical identification of the supernova
and coincident detection of gravitational waves in a single ET, could
confirm if gravitons are massive. One should also be mindful of the 
covariance between various parameters of a source before deciding if it 
is safe to draw scientific conclusions based on observations in a single-site ET.


Second referee:

>In my opinion this article is a well-written review of the science benefits of
>the proposed Einstein Gravitational-Wave Telescope. I recommend it to be
>accepted for publication in the IOP Amaldi 9 special issue. My only minor
>comments are:
>
>- At the beginning of section 2, "LCGT" should be changed into "KAGRA."
>
>- The accepted acronym for Advanced LIGO is "aLIGO."
>
>- On page 11, "subratcing" should be replaced by "subtracting."

We have corrected the typos noted by the second referee except that
we haven't used aLIGO, as this acronym tended to look awkward when
speaking about a number of different detectors that are all advanced.