Dear Reviewer 1, Thank you very much for your very constructive

Dear Reviewer 1,
Thank you very much for your very constructive comments and suggestions. I
deeply appreciate your spending your valuable time for the review. I revised the
manuscript based on your comments and suggestions. Please find following replies
to your comments. Your comments and suggestions are indicated in italic font, and
modifications in the revised manuscript are indicated with red color.
Reviewer #1: This paper examines summer precipitation over northern Eurasia
reproduced in CMIP5 models. The evaluation of hydrological cycle in northern
Eurasia in the CMIP5 models is significant. I recognized that the multi-model
analysis attribute the precipitation bias to the SCRF bias. However, this manuscript is
not acceptable as an original paper to be published in the present form. It seems to
me that there are several points to be improved. My recommendation is that the
authors will rewrite and resubmit the paper with considering the comments carefully.
I hope that a renewed paper will be acceptable for publication after major
corrections.
The authors suggest that cloud coverage bias and the moisture transport bias over
broad northern Eurasia lead to the precipitation bias. However, the authors do not
discuss the cause of the cyclonic circulation bias associated with the moisture
transport bias. Why does the surface warm bias produce the cyclonic circulation
bias?
Thank you for your question. We add an explanation:
L.299-304: The surface warming results in heating of the lower troposphere. Resulting
expansion of the atmospheric column air causes the thermal low pressure system with
the cyclonic circulation near the surface. The importance of the surface warming to the
low level circulations is reminiscent of the Asian monsoon circulations associated with
the warm Tibetan plateau (Broccoli and Manabe 1992; Kitoh 2002).
The authors suggest that the precipitation bias is caused by the very little cloud
coverage and associated positive SCRF over western Eurasia. However, it seems that
the precipitation bias over eastern Eurasia is not simply explained by cloud coverage
bias and overactive water recycling. In the eastern Eurasia box, positive SCRF occurs
to the north of about 60N (Fig. 10), however, wet bias occurs in the same region (Fig.
6). In the vicinity of Lake Baikal, positive SCRF bias links to wet bias. Why does the
wet bias and positive SCRF simultaneously occur?
Thank you for your question. We modified our explanation as:
L. 342-346: the warm bias produces the continental scale circulation biases. The
cyclonic circulation biases reduce the moisture transport from the Atlantic and
enhances the northeastward moisture transport along the eastern coast, contributing
to the dry bias over western Eurasia and the wet bias over eastern Eurasia.
Does "surface temperature" mean surface air temperature? Does warming in the
lower troposphere occur? Does the land surface energy budget associated with the
grand warming due to positive SCRF bias produce surface air warming? The warm
bias should be discussed in the context of surface energy budget including the effect
of sensible heat flux and radiation flux at the ground surface.
Yes, we corrected the word “surface temperature” to “surface air temperature”. We
added the surface heat budget in Fig. 11 and explained as:
L. 330-336: We next examined the surface heat budget of MME compared with ERAI
and CERES as shown in Fig. 11a-d. The ground temperature is determined by the
surface heat budget and strongly influences air temperature in the lower troposphere.
The shortwave radiative heating corresponds well with the warm bias of surface air
temperature, whereas the latent heating also contributes to the warm bias over
western Eurasia. The longwave radiative heating and the sensible heating bias are
largely negative over western Eurasia corresponding to the warm bias.
The authors emphasized that the over active water recycling occurs associated with
the west-east precipitation bias. However, any index of water recycling is not used to
indicate strength of precipitation recycling in the analysis results. I suggest that the
authors compute precipitation recycling ratio which used by Trenberth (1999, J.
Climate 12, 1368-1381).
We are sorry that our expression is inappropriate. We showed that the coupling
index of evaporation and precipitation, which is not a water recycling index. To
calculate the recycling index of Trenberth (1999), the vertically integrated moisture
flux is required, which is not available in the CMIP5 archive. Therefore, we modified
our expression from “over active water recycling” to “very strong local evaporation
precipitation coupling”.
L. 38, 44, 490, 497
The modeled precipitation over northern Eurasia consists of convective precipitation
produced by a cumulus parameterization scheme and stratiform precipitation
produced by a grid-scale large-scale condensation process. I would like the authors
to discuss which type of precipitation is important for producing precipitation bias.
Thank you for your suggestion. We showed precipitation from a convective
parameterization scheme and a large scale condensation scheme in Fig. 7 and
describe them as:
L. 261-266: In climate models, precipitation is produced by a convective scheme and a
large-scale condensation scheme. The MME precipitation from these two schemes is
compared with that of ERAI in Fig. 7. The dry bias over western Eurasia is mostly
associated with the convective scheme precipitation, whereas the wet bias over
eastern Eurasia is related with both the convective and large-scale condensation
scheme.
Does the eastward and northeastward moisture flux bias over central Eurasia (along
50-60N) contribute to the wet bias in eastern Eurasia?
Yes, we think so. We added:
L. 309-310: Moreover, the eastward moisture flux bias over central Eurasia along
50-60°N also have some contributions to the wet bias in eastern Eurasia (Fig. 6c)
P13 L333
"high path" > high-pass
Thank you, we corrected to “high-pass”
P14L350etc..
A term "a time scale of several days" may not be appropriate to
represent "synoptic time scale".
Thank you for your comment. Our results indicate that the precipitation events with
a time scale of several days (Fig. 16a, b) are associated with deepening of Z500 with
a somewhat longer time scale (Fig. 16g, h). And this precipitation events are also
related to convection associated with local instability (Fig. 16i, j). We modified our
expression from
“a time scale of several days” to
L. 398 & L. 510: a time scale of several days associated with synoptic disturbances
and local convection
Distribution of tendency of precipitable water content is not shown in Fig. A1(a).
Thank you. The values are almost zero everywhere. It is shown because readers
may suspect that the tendency of seasonal evolutions might be large.
We described them as:
L. 562: The tendency is negligible in the seasonal average fields
Fig. A1 caption: Values are very small everywhere in (a).
Thank you very much again.
Sincerely yours,
Nagio Hirota
Dear Reviewer 2,
Thank you very much for your very constructive comments and suggestions. I
deeply appreciate your spending your valuable time for the review. I revised the
manuscript based on your comments and suggestions. Please find following replies
to your comments. Your comments and suggestions are indicated in italic font, and
modifications in the revised manuscript are indicated with red color.
Reviewer #2: This study compares summer precipitation totals among observations,
CMIP5, and reanalysis products. All CMIP5 models produced a bias pattern of too
dry over western Eurasia but too wet over eastern Eurasia. The researchers explain
that the possible cause is underestimation of clover cover, leading to higher surface
air temperature and lower surface pressure in the lower troposphere. This reduces
moisture transport from the Atlantic to western Eurasia but increases northward
moisture transport along the eastern coast, amplifying and overestimating water
recycling.
This research is nicely organized and conveys the information clearly. But I have
some concerns.
1.
The conclusion is a bit too general and fails to mention geographical
differences among the results (or causes) between western and eastern northern
Eurasia (see my specific points below). Based on results and figures presented, the
western and eastern precipitation bias may be related to two different
mechanisms.
The main results stated seem to be more related to western
precipitation bias.
Thank you very much for the important comment. We revised our manuscript to
answer all questions in your specific comments below, which added many
information about the geographical differences of the biases.
2.
Although the focus is precipitation reproducibility over northern Eurasia, it
does not mean you have to confine everything, especially potential causes, to this
local area. Atmospheric circulation patterns cover at continental scale (especially
look at seasonal time scale) and local anomalies of wind and moisture flux may be
the expression of large-scale circulation anomalies. I suspect that the results over
eastern Eurasia could be related too strong monsoon circulation by models which
may not have much to do with local air surface temperature (see point 2 below).
Thank you for your comment. Yes, it is related with too strong monsoon circulation
as described in the reply to the specific comments below.
Specific comments:
1.
Line 39-40, "the models underestimate cloud cover over Eurasia,
allowing…". Based on figure 10 of shortwave radiation, this is true for the majority
of land. But the SE portion of the eastern region has more cloud cover instead. This
could be one of the reasons for moist bias for this region.
Yes, that is an exceptional region. We explained them as:
L. 349-353: Despite the less clouds for the majority of Eurasia, the cloud cover is
overestimated around the far eastern part of the Eurasia around (135°E, 60°N). This is
possibly because of the enhanced northward moisture flux (Fig. 6c), and/or may be one
of the reasons for the moist bias for this region (Fig. 6a).
2.
Line 41, "..leading to a warm bias at the surface". The graph seems to show
a warm bias over the southern study region, and possibly a cool bias over the
northern region.
This seems to suggest a north-south air temperature gradient
that leads to more a meridional circulation pattern.
Thank you for your suggestion. We added discussion about the north-south
temperature gradient as follows:
L. 309-316: Moreover, the eastward moisture flux bias over central Eurasia along
50-60°N also have some contributions to the wet bias in eastern Eurasia (Fig. 6c). This
moisture transport is again associated with the surface warm bias. Figure 9c shows a
latitude-height cross section of temperature and zonal wind biases averaged from
40-100°E. Because of the Iranian plateau and the Tibetan plateau, effects of the surface
warm bias reach around 700 hPa at 35-60°N, resulting in a negative meridional
temperature gradient in the lower troposphere that balances with the geostrophic
westerlies.
The surface temperature anomaly pattern is not clear over the eastern study region
(probably not significant; Fig 8).
Yes, the temperature bias is small near the eastern coast. Figures below show the air
temperature and meridional wind at 850 hPa for the biases of MME from ERAI.
We describe these results in the revised manuscript as:
L. 305-308: Although the warm temperature bias near the eastern coast is relatively
small, the MME temperature bias from ERAI around Japan is significantly negative (not
shown), indicating enhancement of the land-sea contrast. Associated with this
temperature gradient, the monsoon southwesterly is significantly stronger in MME.
3.
Page 3, line 88, Stucki et al 2012 is not listed in the ref.
Thank you we corrected the reference.
4.
Page 4, line 127, "…humidity..", is specific humidity used to calculate
moisture flux?
Thank you. We changed “humidity” to “specific humidity”
5.
Page 12, line 318, "….MME results indicates that…", use "indicate"
Thank you. We corrected this mistake.
6.
Page 18, first paragraph, unclear in explanation about events and how they
related to the amount of bias.
Thank you for your question. We added following explanations:
L516-518: The amount of precipitation in the events are possibly influenced by the
large-scale moisture supply of the seasonal average.
Thank you very much again.
Sincerely yours,
Nagio Hirota
Dear Reviewer 3,
Thank you very much for your very constructive comments and suggestions. I
deeply appreciate your spending your valuable time for the review. I revised the
manuscript based on your comments and suggestions. Please find following replies
to your comments. Your comments and suggestions are indicated in italic font, and
modifications in the revised manuscript are indicated with red color.
Comments from Reviewer #3:
"This paper presented a quite comprehensive analysis examining CMIP5 model
historical run biases in precipitation over Eurasia. Overall it is written well and should
be publishable after some important remedies as listed below.
Major comments:
1) Terrain in large portion of Asia is high and complex, not to mention the Tibetan
Plateau that has an average height of > 4km. There are also other high lands like
Iranian Plateau. A key finding of the paper is that MME has a biased cyclonic
circulation, as compared with ERAI, that explains the dry-west/wet-east model bias.
This cyclonic circulation is roughly around northern boundary the Plateaus. It would
be helpful that authors discuss some potential roles of topography in the circulation
bias.
Thank you very much. As you suggested, the effects of the plateau are important.
We added Fig. 9c and explained as:
L. 309-316: Moreover, the eastward moisture flux bias over central Eurasia along
50-60°N also have some contributions to the wet bias in eastern Eurasia (Fig. 6c). This
moisture transport is again associated with the surface warm bias. Figure 9c shows a
latitude-height cross section of temperature and zonal wind biases averaged from
40-100°E. Because of the Iranian plateau and the Tibetan plateau, effects of the surface
warm bias reach around 700 hPa at 35-60°N, resulting in a negative meridional
temperature gradient in the lower troposphere that balances with the geostrophic
westerlies.
2. In diagnosing model errors, authors correlated S score over whole land with W-E
precipitation difference, surface temperature in central, etc. The W-E precip
difference should be partly included in pattern correlation contained in S score. In
other words, it is not totally independent of S. That is why the correlation is positive
and high. Also it is not a physical entity. I suggest use the bias cyclonic circulation,
say meridional wind difference, to replace precip difference. This way can also
confirm the role of the cyclonic circulation in dry-wet bias.
Thank you for your comment and suggestion. Yes, the W-E precip difference is not
independent of the skill score S. We just want to say:
L. 278-280: the precipitation reproducibility measured using the Taylor skill score
corresponded with the severity of the dry west and wet east biases
We also calculated the W-E meridional wind difference in Fig. 10b and described
them as:
L. 324-328: We also plotted meridional wind differences between eastern Eurasia
(105–150°E, 45–70°N) and western Eurasia (15–60°E, 45–70°N) at 850 hPa against the
precipitation skill scores in Fig. 10b. Their correlation is significantly positive,
confirming the importance of the continental scale cyclonic circulation in the lower
troposphere.
3. Section 4 concludes that models/MME capture rainfall events well. This is a quite
strong statement, but based only on a point or two. The tone of the section seems
that the models capture individual rainy episodes well in general. I wonder if authors
can evaluate this at more locations or all grids over a subdomain, and display results
creatively. If this is impractical, readers should alerted that the content of the section
applies to a point.
Thank you very much for your suggestion. We perform some additional analyses as:
Fig. 17 & L. 464-474: The evaluation of the precipitation events above is performed
over 5° × 5° boxes in western Eurasia (35–40°E, 55–60°N) and eastern Eurasia
(125–130°E, 55–60°N). Here, we briefly examine the events in the other regions of the
northern Eurasia. Figure 17 shows the maximum precipitation intensity of the events
and the duration time with precipitation larger than 10% of the maximum intensity
(see Fig. 16a for an example of the maximum intensity and the duration time). In the
APHRODITE observation, the maximum intensity shows large values over western
Eurasia and along the eastern coast. The duration time is relatively longer over the
regions to the west of 80°E and relatively shorter in eastern Eurasia around (110°E,
50°N). Although detail distributions differ, these basic regional dependency of events
are reasonably captured in MME.
4. Some portions of paper are difficult to follow because of the lack of explicit
elaboration. For example, in text of L267-270: "It is interesting that the reanalyzes
did not follow the relationship between the west-east differences and skill scores,
and this implies that the errors in the west-east precipitation differences may be
related to atmospheric variables, which could be mitigated by data assimilation."
Here, (1) "did not follow" is not clear. I assume authors means the
observation/reanalyses red dots not align along a line as in the models. Ig so, at
least you should add something like "as observations are not lined up as models on
the scatter plot. (2) "implies that the errors .. may be related atmospheric variables.
…" The implication is too jumpy in inferring. Need more elaboration.
Thank you very much for kindly suggesting more clear explanations. We modified
them as:
L. 282-286: It is interesting that the reanalyses did not follow the relationship
between the west–east differences and skill scores as their dots not align along the
line of the CMIP5 models. The skill scores of precipitation reproducibility in the
reanalyses is not strongly related to the west-east contrast of the precipitation.
5. Reanalyses rainfall is not observed, and maybe even not physical since they
contain substantial shocks resulting from frequent data ingesting. That's why their
S-sores are even lower than most of CMIP5 models even though they are much
shorter-term forecasts. Thus I suggest remove them for Fig. 2 or altogether.
Thank you for your suggestion. We agree that reanalyses precipitation is not reliable.
But we think their precipitation data provide some information about model
precipitation consistent with analyzed atmospheric fields. We added following note:
L. 160-162: Note that precipitation in reanalysis are predicted values using the
forecasting model from the analyzed initial fields. It is not based on observed
precipitation, but is consistent with the analyzed large-scale atmospheric fields.
6. Climatology normally refers to 30-year mean, as defined by WMO. Here you used
20 years only. Normally this 20- or 30-year as means do not make much difference.
But CMIP5 historical runs from each model may have different timing on the decadal
scales. In other words, after more than 100 years of continuous integration (staring
from 1850), a model's 1981-2000 may not really represent that period in these
ocean-coupled models. A longer period is more preferable. Authors may extend to at
least 30 years, matching climatology definition.
Thank you for your suggestion. To investigate impacts of the decadal variabilities we
have compared average of 1981-1990 and 1991-200 in the CMIP5 models. The
figures below show scatter plots of the precipitation skill scores evaluated using the
climatology of 1981-2000 versus that evaluated using the 1981-1990 average (left)
the 1991-2000 average (right).
We added following explanations:
L. 164-173: The climatological average was defined as the average between 1981 and
2000. Because GPCP1DD and CERES observations are available only for 1997–2012 and
2000–2009, respectively, the averages over those periods were used for these datasets.
Because coupled models of CMIP5 have different timing on the decadal variabilities,
the relatively short period of averaging may affect the climatology. However,
comparing 1981-1990 average and 1991-2000 average in the CMIP5 models, we have
confirmed their differences are very small compared to differences among models (not
shown). Therefore, we consider that the effects of the decadal variabilities do not
influence our main conclusions about the evaluations of climatology in the CMIP5
models.
Minor:
L73 and figure caption evaporation/transpiration -> evapotranspiration.
Thank you. We changed “evaporation/transpiration” to “evapotranspiration”.
L174 in equation. Is term A needed? Also what is increment?
Thank you for your question. We added explanation as:
L. 186-198: Analysis increment is artificial modifications of the short-term forecasts
that reduces their differences from observational data in the assimilation system.
L128-129: surface temperature, air temperature. One is at 2m agl and the other
upper-air (many levels)?
Thank you we modified the phrase as:
L. 128-129: surface air temperature at 2m above ground level, air temperature at
the upper levels
Fig. 6c and 8b wind vectors are redundant."
Thank you we removed vectors from Fig. 9 (Fig.8 in the old manuscript).
Thank you very much again for your kindness.
Sincerely yours,
Nagio Hirota