nm5390: er the first speaker is namex er namex got his er bachelor of science 
er in physics er with a concentration in astrophysics
nm5391: that was a minor
nm5390: minor in astrophysics from the University of Leicester er then he got 
his PhD from er this department er and er stayed on as post-doc er and in two-
thousand-three was elevated to the status of lecturer and er in the last few 
weeks er has been informed that er he's been awarded the L F Richardson prize 
so congratulations to him for receiving that and er he'll now speak on the 
topic of 
nm5391: okay thanks yes i'm going to talk about Cloudnet project er which is 
something that has myself Ewan Anthony Nicola and Malcolm for the last few 
years this is a E-U-funded project er trying to to systematically evaluate the 
representation of clouds in a number of European forecast models so we're 
trying to really get away from sort of case study mentality er that has 
sometimes affected this er subject so first i'll er motivation and talk about 
the representation of clouds and G-C-Ms invested in this er i'll talk the the 
organisation of the Cloudnet project and then talk about er some of the key 
Cloudnet products this isn't certainly an exhaustive list er the sorts of 
things that we're producing er are radar and lidar at various stations in 
Europe these are sort of key parameters we need in the current models then i'll 
be briefly talking about the various ways that one can use this kind of data to 
evaluate er these models such as using er long term means of of various 
parameters er skill scores probability distributions er and so on er but 
certainly there's far more work to do on the er sort of making the best use of 
all data that we er have on er on clouds er with er co-located er cloud er 
values from the various models so just to give you a sort of brief introduction 
to how clouds are represented in models possibly familiar to many of you which 
perhaps will look like this so real-, in reality they've got structure in all 
scales er and their three dimensional structure is also important for radiative 
transfer and of curse we can't represent all this detail er in fairly low res-, 
res-, low resolution model but typical representation might look something like 
this typical G-C-M er global circulation model grid box as a er horizontal er 
size that vary that can vary considerably depending on the one in question but 
sort of typically one kilometre for the most high resolution er metre scale 
models to about a hundred kilometres for a global forecast model er whereas in 
climate of course we have to run things for much longer so we would need maybe 
three-hundred kilometre resolution the vertical box size is about five hundred 
metres and in each box we hold er a very very crude representation of that with 
all this complexity we just have clouds a fraction and then the mean liquid 
water content the mean ice-water content we typically assume these clouds to be 
horizontally uniform within each box okay er perhaps will have er an impact on 
of transfer of various also have to make assumptions such as er cloud phase and 
the particle size these are usually represented as functions of temperature so 
i guess there's two questions here well one is how accurate are these values er 
cloud fraction content in the models and er er also even if we were to get 
accurate values for the mean would these be sufficient to get accurate 
predictions of precipitation and the radiati-, radiative effect of cloud which 
is an important climate i'm really just going to be looking at the first of 
these today so let's just have a look at er the results of this A-M-E-T 
comparison that shows the vertically integrated cloud water content as a 
function of latitude and what strikes you immediately is there's a hell of a 
spread er you can see almost a factor of ten different sort of in the er the 
wettest model and the driest if you like er and the interesting thing is of 
course is that these models are often tuned er to give roughly the right er top 
of atmosphere radiative er fluxes that are er measured from satellites so it's 
quite surprising that there's such a huge variation of water contents give you 
rather similar er radiative properties er and of course this is a bit of a 
concern because clouds are very important for er radiative transfer 
precipitation production and so we're very in-, interested in how these er 
might change perhaps in a future climate scenario so really to have confidence 
in climate models er and also to do the best for precipitation forecasts in the 
much shorter term we really need to improve this and really the problem has 
originated from the fact that the current satellites provide sort of i-, 
integrated information or information just a cloud top er we don't really have 
much of a constraint on these values to kind of bring down the spread we really 
need instruments with a higher vertical resolution and you won't be surprised 
to know that that's going to be cloud radar okay so the E-U Cloudnet project 
was er sort of run for quite a few years now it aim is to retrieve continuously 
these crucial parameters and we're going to look at three particular sites 
Gibraltar Cabauw in the Netherlands and Palaiseau in France er we're also soon 
to include the U-S er and tropical A-R-M sites and also the german Lindenberg 
site to extend this er to other places in Europe we've got this is a list of 
current models involved in the project so th-, the European of course Mesoscale 
and global version of E-C-M-W-F the French model the Dutch two two Dutch models 
er also the Swedish German and Canadian models are going to be involved so we 
should be able to objectively evaluate the clouds in all of these different 
models now there's some crucial aspects to this er project with s-, we're 
reporting er retrieval errors in data quality flags to make sure we know how 
accurate our data is at all times and using common formats er S-C-D-F er to 
allow all our programmes be applied at all sites and compared to all models so 
w-, as i said we're trying to get away from the problem of people just using 
their algorithm on a small amount of data comparing with one model perhaps this 
is allowing us to do the thing in a much more sort of extensive type of er data 
sets and and compare all sorts of models over a much longer period and the 
three Cloudnet sites are in Europe so we've got Chilbolton here er Paris there 
and the Netherlands site there and this is a list of the er instruments that 
they have so i guess the core instrumentation that we have to consider is er 
the cloud radar cloud lidar microwave radiometers er and rain gauge that's 
really where most of the information that we're using is coming from er just a 
quick er overview of the the key information provided by cloud radar and lidar 
all you need to know really is that the radar is sensitive to the large 
particles so particularly detects ice er and the s-, sort of precipitation size 
particles such as drizzle from stratocumulus so you can hear s-, see here we've 
got a nice ice cloud here that's fully penetrated and a stratocumulus that we 
can see the drizzle falling from here by contrast the lidar is sensitive to 
much smaller particles particularly er cloud droplets er and aerosol particles 
so you can see here we've got the boundary aerosol here topped by a very strong 
return from the stratocumulus cloud base that's embedded within the drizzle so 
you from the radar alone you wouldn't be able to to tell where that is and then 
you can see you can also detect some er super-cooled liquid water that the 
radar doesn't detect and er also there's some attenuation problem here in the 
ice so you can see that there's very complementary information provided by 
these use the two in synergy to get the best estimate of the profile of clouds 
at any particular time so this is a sort of diagram of all the we have in er 
the Cloudnet project and basically the radar lidar frame rates micro and the 
model data your various stages of processing that takes us up to here er now of 
course we've got various different formats in some cases from the various 
models so what we have is this er stage here sort of er instrument synergy 
target categorization i call it which i'll talk more a bit about in a moment 
which is where we unify the formats and bring all the data together so we can 
then go on to produce this list of cloud products that are very useful er for 
someone who doesn't know about radar and doesn't really want to know just wants 
to know about ice water content or liquid water content cloud boundaries 
perhaps turbulence information drizzle er ice particle size and so on so this 
is really where the meteorology is coming in er this is at high resolution of 
course the resolution of the actual instruments but of course we can't plot the 
values on the model grid so we we have to average it up onto the various model 
grids we do that separately for each where you can see there's more model 
information coming in here and then we've got the level three products which 
give us statistics on the er comparisons of these products with what's actually 
modelled so i'm going to talk a bit about this thing in the middle first er it 
really makes th-, this multi-sensor data much easier to use and there's a 
number of key facts that we have here you combine these various key 
measurements that i mentioned with an identical format for each site er 
performs a number of key processing tasks so for interpolation of all the data 
onto the same grid many algorithms actually make use of temperature and wind 
which a-, a-, are much more accurate than models the cloud information so we 
can actually take that in er with some confidence we have to correct the radar 
for attenuation er both from gasses and from liquid water er we provide this 
some essential extra information that makes this data then usable to create 
cloud pro-, products er we characterize the random and systematic measurement 
errors er also the instrument sensitivity so you know that if say the radar 
doesn't see a cloud er you can tell wh-, whether the cloud really isn't there 
or perhaps there's a sensitivity issue and it might be there but can't quite be 
detected and this key extra information here is what the cl-, what the target 
that the er instruments see are actually composed of so of course we want to 
know whether it's droplets ice aerosol insects and so on and then there are key 
er data quality flags which tell us when our observations are reliable and when 
they're not so just to show you an example of this this is actually a case from 
the er U-S A-R-M site which er is from now able to process in exactly the same 
way as the European data er one key thing you can see here if you with the you 
might think well got all this got this convective system here perhaps er some 
rain w-, w-, we also seem to have cloud cover for the whole the whole day so 
this would would mean sort of got this cloud from nought to three kilometres 
now of course in fact that's insects but er it's certainly a danger with people 
who are are not used to looking at this kind of data er misinterpreting it 
perhaps er so what we provide is this er classification of the actual targets 
to some various some various vari-, various sim-, simple types so you can see 
here we've characterized our insects as a dark grey aerosol er sorry there's 
aerosol as well in here so it's aerosol and insects er just aerosol here light 
grey see clouds there this is rain and there's an orange stripe here where this 
er there's a melting layer and then this is ice up here so er you can see 
suddenly we are then able to compare with the models because we know that this 
isn't cloud at all er so trying to evaluate of clouds there and er we've 
identified that correctly as insects and we've also got a sort of detection 
status f-, flags here telling us er whether the data's reliable and so on so 
the first product i'm going to speak about is cloud fraction which is er 
derived just by taking a model grid and superimposing it er and then you can 
just count up if you like the fraction box that's fu-, full of cloud and use 
this cloud fraction on our form grid so you can see here a comparison of about 
a month of data and we've got quite a few years of this now the observed cloud 
fraction going from er nought to one hundred percent cloud cover er purples 
this is what we derive from our radar and lidar and this is the representation 
in five of the European models er involved in the project so you can see 
there's certainly some skill in forecasting the location of these er large 
cloud features and of course a day is just that size there so you can imagine 
that you might well get a thousand kilometres of er cloud being affetced over 
site in quite a short period so actually this is quite a large scale cloud 
feature even though it's just measured at one particular site you can also see 
some quite distinct differences between the actual structure and the amount of 
clouds in the various different models so this is what we want to evaluate so 
that's cloud fraction and i'll talk a bit about s-, statistical comparisons er 
later on er now the key thing to getting liquid cloud amount so that's the 
amount of er liquid water content in a model is er these dual wavelength 
microwave radiometers now what these provide is er an estimate of the liquid 
water path which is the total amount of liquid in a column above okay so 
instead of looking upwards and passively measure the radiation coming downwards 
so what we what we do because they are sensitive to both the liquid and the 
vapour in the column above above them is to combine two frequencies which have 
different er sensitivities to the liquid and water vapour and you can see that 
we've got simultaneous equations so we can easily invert this to get liquid 
water path from these two optical depths of twenty two and twenty eight 
gigaherz now traditionally with this technique you assume that these co-, co-, 
coefficients are constant when in fact they depend on for example temperature 
of the liquid cloud er then you also have to c-, er cope with the fact that the 
calibration can drift significantly instruments so you can see retrievals here 
path so fairly smooth value here and then sort of spikes here and basically 
we've got clear sky here and this drift of quite substantial amount is just due 
to er calibration so er come on with a new technique er where you actually add 
the information from the lidar on the model er to tell you where the clear sky 
periods are so the coefficients can be calculated more accurately using the 
cloud temperature and we can then recalibrate the the er instruments when we've 
got clear sky periods so the lidar sees no cloud in these periods here so we're 
able to jump to this red value and get a much more accurate er estimate so we 
can use this then to get liquid water content now as i sa-, as you saw before 
the radar is er is dependent on on the large part particles in particular so 
tends to be detecting drizzle rather than liquid droplets which is where the 
the ones that are important for radiation so rather than using radar 
reflectivity for for getting liquid water content we basically use the er 
temperature to predict sort of the adiabatic liquid water content profile that 
is if you lift liquid water er up and it comes condensed than you can predict 
the profile of liquid water you'd expect in reality entrainment produces that 
but what we have from our measurements is the cloud base from the lidar and the 
cloud top from the radar and we know the integrated liquid water content so we 
can just scale this adiabatic profile so that it agrees with the value so you 
can see a much more realistic liquid water content here where we've got liquid 
water increasing from cloud base so we've got high resolution liquid water 
content information this is er a sort of comparison on the climatological er 
type situation if you like so we could do this to a foreca-, to a climate model 
where it's not simulating the actual events where we've got for example mean 
cloud fraction and the observations in blue er model in red er and the 
frequency of er any cloud that's been observed in the box over the threshold 
about five percent er so what you what we tend to find when we do this kind of 
comparison is that the the frequency of cloud occurrence tends to read 
reasonably well er certainly below around seven kilometres er but the amount 
when present or the amount of cloud when there's some presence er can often be 
wrong so you can see here the reason why the mean cloud fraction is overestima-,
is underestimated in the model is that it tends to produce too little cloud 
when there's some cloud present so in a way this is sort of dependent on 
humidity information in the model to say whether there's going to be some cloud 
in the profile at all er the amount is dependent on the cloud perameterisation 
so that would be perhaps where the problems lie we can also identify some other 
problems with the er this is a sort of P-D-F of cloud fraction and you can see 
quite clearly that the i-, in this case the Met Office mesocale model the model 
has difficulty predicting a hundred percent cloud fractions because its got far 
less occurrence of this a hundred percent er cloud fraction this is a a known 
problem er with this particular er cloud scheme we can do a similar thing with 
er liquid water content er so this is showing similar things mean liquid water 
content versus height frequency of occurrence presence you can see the 
frequency of occurrence is actually not to bad although perhaps there's a 
underestimate of the er height of the boundary layer here on average through 
this this is a a whole year period about it's about a hundred-and-seventy-two 
er days in total to be honest but you can see there's a tendency as well to 
underestimate the fraction of er supercooled liquid clouds er and perhaps the 
liquid water content when there's cloud presence is underestimated a bit from 
this comparison you can see this whole P-D-F is s-, shifted to the left so we 
can do the same with ice water content and you can see here we've got the mean 
ice water content over about year period from about two thousand three er and 
also well you can kind of see there's perhaps an overestimate of ice er in the 
upper levels er and so the mean ice water content something somewhat too high 
but you do have to be careful er with ice water content er be-, being aware of 
the radar sensitivity particularly to the thin at cloud top er and also the 
fact that the are carried out in in rain situations and that does er limit the 
sample somewhat so the last thing i'll talk about another way of evaluating er 
models which is actually to look at how good the instantaneous forecast of the 
cloud is so what we do is we take a threshold cloud of roughly say point one 
and calculate a contingency table so this is a whole month of data again and 
we've got green indicating that cloud was both observed and modelled white 
indicates that er it was clear sky in both the cloud and the model if both the 
model and the observations and then the red and blue indicate when there's 
cloud in just one but not the other so we can define a b c and d to be er the 
amount the occurrences of each of these different categories and from this w-, 
a standard procedure is to calculate er various skill scores and one of the 
popular ones is the equitable threat score its got a definition like this but 
basically one means a forecast and zero or a random forecast so you can see 
over a a bit over a year you can see the sort of tra-, trajectory of the scores 
for various different models so i'm pleased to note that the Met Office 
mesoscale is coming out top here er Meteo France is a bit of a poor showing 
there although it does improve this point here in around April two thousand and 
three and quite an interesting factor there is that there's sudden change in 
their cloud scheme here so you can see from p-, basically never predicting 
above around point two cloud fraction and they tweak their model and then 
suddenly they're getting much higher values you can see suddenly their skill 
scores actually improved here and one thing that's interesting to note is that 
er cloud fraction and water content in this model are actually entirely 
diagnostic so it predicts humidity er and then pr-, prognostically and then 
from that it diagnoses er cloud er amount and water content diagnostically so 
so it's sort of interesting to see whether with quite a crude scheme certainly 
compared with the Met Office model to see er have er skills that are comparable 
so we can also do the er skill score versus height and you can see again the 
Met Office model with a short time scales is er six to eleven hour forecasts er 
is nosing ahead E-C-M-W-F er this Dutch model and also the Met Office model 
with a time of about a day all do comparably well er and them the Meteo France 
model has sort of increased from before er two-thousand-and-three to after er 
it hasn't quite caught up with the crowd there so there's certainly a potential 
to with this technique to test new model paramaterizations you could re-run er 
the models an-, and redo the skill scores and also what we are also doing is 
comparing the global versus the mesoscale versions of the Met Office model but 
both of their er data extracted so just to conclude well this is really the 
first time that we've had such a large er timed series of data the key 
parameters that we need to evaluate er climate models and forecast models er 
and from we can apply this to a number of different sites around the world 
course there's other Cloudnet products hat i haven't talked about such as 
drizzle er ice particle size er turbulent kinetic energy dissipation rate er 
and of course we're we're also moving to er apply this to the U-S A-R-M site er 
in the framework of the new G-W-E-X working group and of course there's still 
lots to do in evaluating er these various models er i've just sort of touched 
on some of the the ways that were doing it so far and if you're interested in 
more information then you can find it here we've got some pr-, Quicklooks at 
this web site so i'll finish there 
nm5390: okay time for maybe one or two questions yes
nm5393: how do you the sort of problems with just the synoptic development in 
the models which have absolutely nothing to with cloud scheme models what do 
you then get on your sort of 
nm5391: yeah we are we are evaluating how the clouds are which is often because 
the dynamics is good and that puts it in the right place so you really have to 
then strip out the skill scores which evaluate how good the actual forecast was 
in an instantaneous sense and the climatology which is what a climate model 
would fix and so in that situation the er you'd only have a sort of statistical 
representation of the weather systems if you like we weren't trying to evaluate 
the individual ones and so climate models compare the means
nm5393: but i mean even in the forecast models in mean in your opinion at at at 
points for example
nm5391: yeah
nm5393: and if if the cloud system is just slightly in the wrong place because 
of error in the forecast how can
nm5391: we are we have 
nm5393: do you how you might sort that out
nm5391: well we looked through got data from er a whole er each Met Office 
forecast for example we've got the full period so we can evaluate the different 
times and look at the increase in error basically as we go for different time 
scale so you can do that
nm5390: just one more question yeah
nm5394: your P-D-F cloud maps for the observation is a bit suspicious isn't it 
because it had the big er fraction for hundred percent quite a small fraction 
for ninety percent i mean why should there be such a big difference in 
observation between
nm5391: this is
nm5394: compl-, complete cloud cover and ninety percent
nm5391: this is er 
nm5394: cloud cover
nm5391: that's a forecast between three and seven kilometres where you are 
dominated by v-, very thick ice clouds and in fact when you do look at the 
observations you do find that certainly on the mesoscale grid er a a twelve 
kilometres you do average the right amount of times to the twelve kilometre you 
do tend to either get a hundred percent cloud or no cloud at all and in fact 
the periods when you've just got partial cloud and it's at that height are 
actually much smaller that's what we see 
nm5390: okay thanks namex we'll move on now to our next speaker i'll introduce 
him very briefly er his reputation i i think precedes him here
su: yeah
nm5390: namex is er 
su: the mike isn't on
nm5390: professor of meteorology er been here since nineteen-ninety-three er 
heads our radar group before that he was er at Manchester and er do you need 
any help with the equipment or you're all set
nm5392: yes thanks for the check
nm5390: okay so
nm5392: okay thank you er well Robin's been showing what you can do from er a 
few spots on the ground for evaluating models and er really this er satellite 
mission the idea is to extend that sort of ideas and analysis er to grow so i'm 
going to give you some ideas this as you can see is some of this talk will be 
in E-E-S-A speak which is an interesting language which you will hear a bit of 
and i'll give you some er descriptions of er the sort of thing that goes on so 
er basically er this mission has been selected now as of November and Alan 
O'Neal was present at the meeting if you give him a few drinks he'll tell you 
about how they had to take a half way through the voting they took a coffee 
break and then the Germans he'll tell you all about that anyway after after the 
recount okay so lets think about the history here of the selection process 
which has been keeping me busy for quite a few years so in nineteen-ninety-
seven nineteen-ninety-nine i went to Granada to present the earth radiation 
mission about six missions presented and er er i'll show you the ones that were 
selected there and then we realised what our error was of course is that we'd 
called this the radiation mission and this is these are the er i should say 
these are the missions we're looking at here for research not operation in 
other words follow on from the sort of N-V-S-A-T generation this is of course 
the E-E-S-A Living Planet programme and of course you know you don't want 
radiation because you're kill us all aren't you so we then er re-christened 
this er EarthCARE er you know EarthCARE which i must admit was er y-, you can't 
be opposed to that er i always have to apologise with an English speaking 
audience for this title
su: thank you
nm5392: it was appointed by a French person er who didn't realise it's a little 
bit er over the top okay er however er let's just think of the other core 
missions and selected for the launch in about two-thousand-and-ten that gives 
quite a lot of time to tweak up some of these algorithms er two-thousand-and-
six G-O-C-E these are the C-O-R-E missions G-O-C-E which is Gravity Ocean 
Circulation so the idea here is to get the er to a few millimetres rather than 
a few centimetres and get the ocean currents much better from altimetry the 
dynamics mission two-thousand-and-seven is a a lidar in space to get clear air 
winds and these ones here are supposed to be slightly cheaper a Cryosat which 
we didn't hear about from er Duncan ill but we probably will hear about in the 
summer and S-M-O-S this is looking this is an L-band er of the surface which 
some people at the at the S-S-C are quite interested in okay this is the 
mission summary er and it says it all really so here we've got our radar lidar 
which er Robin's obviously been talking about giving us our vertical slice as 
we go round the planet and then we have two passive instruments to place this 
is some sort of context so we have an image of a sort of er type to give some 
er context over the er few hundred kilometres and also an type instrument to 
give us the up welling radiation in the short and the long wave er broad band 
okay er so just going through this again right so this is sort of our mission 
summary a scientific objectives the aerosol cloud radiation interactions we're 
trying to get vertical profiles of natural and anthropogenic aerosols and then 
look at their interaction with radiation and clouds the vertical distribution 
of liquid water and ice on a global scale and their transport the details of 
cloud overlap in the vertical cloud-precipitation interactions and the 
characteristics of vertical motions and then put the whole thing together 
hopefully to get the radiated heating and cooling who are combined er with the 
aerosols and so forth okay a little bit of background here well this is the 
scope of the mission sort of geologists who were so here's most of the on the 
cloud troposphere er tropopause radiation so again for the er we're hoping to 
get the er fluxes in the vertical to ten watts per square metre instantaneously 
over so here are sort of er again for geologists and so forth the sorts of 
problems we're looking at are the direct effects of aerosols and radiation the 
direct er cooling and the absorbing carbon er the indirect effect of cloud 
condensation nuclei and the lifetime effect er and then next er the general 
thing about the cloud er feedbacks and the high and low cloud vertical 
structure overlap so interested in the amounts of condensate and the 
sedimentation rate of ice which of course dictates the lifetime of the cirrus 
in your model and a little bonus here in the convective precipitation er are 
these parameterised motions of convection correct we need to get these three 
right and what about these occasional punches through the er tropopause er 
through the cold trap there how often does that occur which will answer ideally 
these particular questions now this is something for Piers this is a an E-E-S-A 
version of this diagram you're going to r-, re-draw Piers er so here's our 
forcing for the greenhouse gasses I-P-C-C and then our Japanese colleague Terry 
er Nakajima has done some more experiments so perhaps we know a little bit more 
about the direct aerosol but the indirect aerosol is still extremely uncertain 
and of course ice clouds and aerosol well er who knows and then just to 
recollect of course the proposal is that these effect of cloud of course an 
enormous shortwave and longwave effect of about eighty and a hundred watts per 
square metre and the net effect of the balance sort of cooling at about twenty 
watts per square metre but just disturb this a little bit and of course er the 
net change is is large and this is what's responsible for the er s-, a lot of 
the spread of the global warming forecast but i think er er Robin obviously 
mentioned the cloud modelling but let's just think we have no information at 
the moment on the profile of the aerosols these ice water contents well we're 
getting those just from the ground sites but this will give us it globally and 
then we have some er global information on convective updraughts okay so these 
are our mission needs here we hope to get the here's the aerosols this is the 
sort of thing you see from a lidar elevated aerosol over the er Indian Ocean 
this is some Peter Walton data here's the convection here seen er from er e-, E-
R-T p-, penetrating the tropopause here over Brazil and then this is the radar 
and the lidar together your imager these are some clouds off California give 
you some idea of how representative these slices are and here's your type 
broadband getting the upwelling of radiation which we're going to have a look 
again at so we have our four particular instruments and i'm going to 
concentrate on radar and the lidar and some particular features there which 
we've er developed for this mission so we've seen this one before in Robin's 
talk we really are interested in seeing there's always from space of course 
especially with active instruments there's always a er sensitivity asp-, 
problem because of distance we really would like to see er extinction 
coefficients larger than about point-oh-five or point-oh-four per kilometre 
that gives us about ten watts per square metre and this means er the 
backscatter sensitivity sort of backscatter to extinction is fifty about this 
unit down here and basically the problem with the lidar of course is correcting 
for the attenuation through the cloud and of course when you get the 
backscatter how is that related to the forward scatter or the extinction which 
is what you really want to know and the way we're going to do that is a rather 
clever idea using a high spectral resolution er lidar which separates the 
Rayleigh or molecular scatter from cloud aerosol rescatter so what does this 
mean well let's just have a look at Icesat which is up at the moment looking at 
the er isopoles this is a flight er as it goes over the Arabian Sea over Iran 
and then a little bit over er the the former Soviet Union up here you can see 
the molecular scatter increasing slowly as this gets deeper lighter blue 
because the gives you the density of the molecules we can also see the aerosols 
down here in blue and green and their vertical extent and see the clouds here 
in white it's nice to see if anybody's interested in here you've got the the er 
low level clouds here embedded and the aerosol above it and here there's 
aerosol below here are the er higher cirrus clouds and you can also see how you 
can there's black below here because they've completely attenuated that's the 
use of the molecular scatter tells you how much optical depth this one hasn't 
attenuated so much so moving on to the er l-, er lidar was actually flown for a 
few days with ten hours' data on the shuttle in ninety-four and again here's 
the clouds running down you can see this is s-, er s-, the super cooled clouds 
that Robin was referring to and you can see that there in the lidar attenuates 
completely these ice clouds there's a smaller optical depth and er just in 
preparation er we looked at the ten hours of data and how often we got these 
supercooled clouds and for clouds about minus ten at our latitudes they're 
present actually about thirty percent of the time which is surprising they're 
rather thin so they're very poorly recomend-, re-, er represented in models and 
it's interesting to note that in fact the the site here was quite 
representative from the er er satellite and th-, er from the er L-I-T-E data 
for these particular same latitudes course if you just take a little bit of ice 
and turned it into liquid water here you change the radiation by about a 
hundred watts per square metre so this this is effectively perhaps quite large 
but er would be interesting to look at that proposal in more detail so how does 
this high spectral resolution work rather busy slide i'm afraid if you look at 
the backscatter power here then the molecular is moving at six-hundred metres a 
second so it's a broad er spectrum whereas the clouds and the aerosol a metre a 
second so you can that's a sharp peak so you can separate out these two so 
that's the the trick here to filter out theses two returns and separate them so 
when we're this is a simulation looking down through two kilometres of aerosol 
the Rayleigh scattering drops off a little it from what you'd expect which 
tells you you've got an optical depth here of about point one which you can 
then use to correct the Mie scattering er i'm sorry to get back to here and 
this is the extinction so this way we've got the er extinction from how much 
the Rayleigh has been reduced correct the Mie to get the correct backscatter so 
we've now got several bits of information for the aerosol in this case we've 
got the er extinction and the backscatter and that ratio that can vary from ten 
to a hundred aerosol particles so it gives us some information about the type 
of aerosol particle and we also have the er cross-polar return which is 
different from spheres er and er non-spheres so you think with this not 
completely orthogonal but obviously the anthropogenic aerosol is spherical 
whereas for example desert aerosol is jagged and gives you different er 
backscatter extinction ratio and different er different polarisation so this is 
the sort thing you see in your profile this is what your Rayleigh does there's 
a bit of cloud here and there's a little jog here because you've lost oh-point-
oh-four optical depth about eight percent of what you'd expect so that's the er 
optical depth that's the forward scatter and this is the backscatter and here 
again is the aerosol layer and you can er get that optical depth and also the 
backscatter okay that's that's the lidar in the next five minutes let's just 
talk a little bit about the radar you've seen this one as well before and again 
in this case we've done some sums looking at aircraft data and so forth and we 
think we need er about minus thirty-six D-B-Z on this scale here down here to 
see most of the radiatively significant clouds those with the optical depth of 
point-oh-five kilometre and there's er what was indicated if we er from the 
radar we probably get the ice water to about fifty percent if we have the radar 
and the lidar better and CloudSat in fact going to be we're very excited about 
this this summer launched and that will actually have a er one-point-eight 
metre antenna high orbit see about minus twenty-eight D-B-Z so at this stage 
you're probably wondering well what does all this D-B-Z mean we all like to use 
this the radar people and that's the return in D-B er compared to the return 
from one millimetre drop per cubic metre so we're actually about thirty D-B 
below that which is of course a factor of a thousand however er all these er 
trips to er E-E-S-A etcetera do rather tiring so when one's coming back on the 
last B-M-I flight at eleven at night wondering if you've missed the bus and 
you're sitting there it's a great pleasure to read the Voyager magazine which 
actually had a little explanation a few weeks ago on D-B so they have a little 
they always have these upmarket things this is on Formula One couple of pages 
written by Tom Clarkson which is a terrible i think there's two of these 
Clarksons around this is Tom and sixty things so he actually had a little 
explanation here which may interest you on D-B there so these Formula One each 
one car pedal to the metal gives a hundred-and-thirty D-B if you've got twenty 
of them that would be two-thousand-six-hundred so anyway er i'm sure er Jeremy 
wouldn't have made that kind of of mistake so okay so what they didn't do on 
there of course which is the main pleasure of the F-One is of course people 
like to hear the Doppler effect apparently it it gives people tremendous buzz 
so we measured the Doppler here as well but er unfortunately Clarkson didn't 
explain so how do you measure this from space because you're looking for a 
change of the return frequency if it's a metre a second of about one part in a 
hundred-million so we don't actually do that we look at the return one pulse 
out when the particle's here pulse out T later when the particle has moved and 
suppose it just moves a quarter of a wavelength then there's an extra hundred-
and-eighty degree phase shift so now rather than being not sensitive enough it 
is very sensitive and your big problem in fact is that if it's moved more than 
a quarter of a wavelength and this is your folding velocity we have to send our 
pulses just just this is just a little bit of technology for a people w-, er 
makes this slide s-, a little thing we have to think about twenty-one 
kilometres is the height of the cloud so we in the Tropics so we better have a 
pulse every point one four milliseconds that's the most rapid it's moved a 
quarter of a wavelength that gives a f-, folding velocity of six metres a 
second so why hasn't er haven't these things been done before from space well 
the problem is of course that if you're moving at seven kilometres a second you 
look int-, you a certain sin path theta of seven kilometres a second on one 
side of the other so that's gives you some random phase variation and of course 
your you've seen a hundred-and-eighty degree shift here the noise mustn't 
really be more than ninety degree shift in other words your Doppler width 
mustn't be more than three metres a second or that would be a random ninety 
degree and if you do this sum you have to have two-point-four metres of antenna 
so on the three or four trips i've made to Japan each time we went out with 
Japanese we managed to get the antenna up from one-point-eight to one-point-
nine to two and finally er got them to agree to have a two-point-four metre 
antenna which gives us now means we can do the Doppler and we can also get our 
extra sensitivity remembering we add rather than multiply okay so what do we 
need the Doppler for so this again is our flight with two above the Brazilian 
severe convection and you can see here here's the penetrative here it's 
penetrating up into the er stratosphere looking at that interesting ice 
sedimentation and looking at er light precipitation the sort of specs we need 
for that maybe convection to a metre a second and the ice and of course the 
drizzle which we're passionately interested in er point-two metre a second 
course the drizzle actually is quite important seriously for the decoupling of 
the evaporating drizzle decouples and er when you the trouble is that when to 
noise that gives you some more random phase noise so er the higher the signal 
to noise the better you can do the Doppler but basically er we want the 
convective motion a metre a second with a only a every kilometre of can just 
about do that and we are allowed to have a longer integration time for the 
cirrus and drizzle because that's er more homogeneous and er it just a-, about 
works with this two-point-four metre antenna and at this level here you're 
seeing actually most of the ice from the aircraft's okay so this is the sort of 
mock-up people like to produce just to give you an idea of s-, bits of 
redundancy the radar itself two-hundred kilos three-hundred watts and there's 
also the imager five-hundred of resolution seven channels in the visible and 
the infrared and the broadband radiometer looks at ten kilometre pixels three 
angles er vertical fifty-nine off on either side for the er radiance to flux 
conversion that gives us our er instantaneous broadband fluxes and the whole 
satellite a couple of potential er ways of configuring it one-point-two tonnes 
and just over a kilowatt is needed er just a little remark on er data 
assimilation so this is data assimilation for geologists here don't need to go 
through that again i think it's still true there is no explicit analysis of 
clouds at the moment in global data assimilation so i think that's true yes 
maybe not er we put global scale probably not true so er how er how is this 
going to will be working quite a bit with E-C-M-W-F on this particular aspect 
and they've done some one D-var so this is on the A-R-M data this is the 
observation this is the model first guess and then this is merging the two 
particular errors er one D-var but what's coming along as well here just to 
show you some of the sort of and they've also been calculating a lot of the 
which of course needed for the correct er er assimilation so if we look at the 
short wave in the winter so this is how much does the top of the radiation 
atmosphere radiation change it can change the cloud cover fraction spot here 
and you can see for the short wave of course reflected sunlight this is the 
bubble level so this is the water c-, that's the melting layer in in the er 
water level in the tropics this is the sort of level of the stratocu-, and 
there's a small effect of reflection from the cirrus long wave radiation of 
course is rather different because you know north south effect now from 
reflected sunlight the biggest effect of course is ref-, is is the radiation 
emitted from the er er cold cirrus and then also the effect of the the melting 
layer so these sorts of jackovians are being developed by E-C-M-W-F er so the 
next few years' time hopefully one can assimilate er so that's the particular 
summary of what we think er we should be able to do and er fundamentally from 
the high spectral lidar radar aerosols er and their interaction with clouds and 
also first time to get some ideas of these er convective motions so thank you