I set a WMO standard and all I got was this lousy Hirsch index - measuring clouds and rain

This week we had the first meeting of the new Task Team on Homogenization of the Commission for Climatology. More on this later. This meeting was at the headquarters of the World Meteorological Organization (WMO) in Geneva, Switzerland. I naturally went by train (only 8 hours), so that I could write about scientists flying to meetings without having to justify my own behaviour.

The WMO naturally had to display meteorological instruments in front of the entrance. They are not exactly ideally sited, but before someone starts screaming: the real observations are made at the airport of Geneva.

What was fun for me to see was that they tilted their ceilometer under a small angle. In the above photo, the ceilometer is the big white instrument on the front right of the lodge. A ceilometer works by the same principle as a radar, but it works with light and is used to measure the height of the cloud base. It sends out a short pulse of light and detects how long (short) it takes until light scattered by the cloud base returns to the instrument. The term radar stands for RAdio Detection And Ranging. A ceilometer is a simple type of lidar: LIght Detection And Ranging.

For my PhD and first postdoc I worked mostly on cloud measurements and we used the same type of ceilometer, next to many other instruments. Clouds are very hard to measure and you need a range of instruments to get a reasonable idea of how a cloud looks like. The light pulse of the ceilometer extinguishes very fast in a water cloud. Thus just like we cannot see into a cloud with our eyes, the ceilometer cannot do much more than detect the cloud base.

We also used radars, the radiowaves transmitted by a radar are only weakly scattered by clouds. This means that the radio pulses can penetrate the cloud and you can measure the cloud top height. Radiowaves, however, scatter large droplets much much stronger than small ones. The small freshly developed cloud droplets that are typically found at the cloud base are thus often not detected by the radar. Combining both radar and lidar, you can measure the cloud extend of the lowest cloud layer reasonably accurately.

You can also measure the radiowaves emitted by the atmosphere with a so-called radiometer. If you do so at multiple wavelengths that gives you an idea of the total amount of cloud water in the atmosphere, but it is hard to say at which height the clouds are, but we know that from the lidar and radar. If you combine radar, ceilometer and radiometer, you can measure the clouds quite accurately.

To measure very thin clouds, which the radiowave radiometer does not see well, you can add an infra-red (heat radiation) radiometer. Like the radar, the infra-red radiometer cannot look into thick clouds, for which the radiowave radiometer is thus important. And so on.


Cheery tear drops illustrate the water cycle for kids. You may think that every drop of rain that falls from the sky or each glass of water that you drink, is brand new, but it has always been here and is part of the The Water Cycle.

Why is the lidar tilted? That is because of the rain. People who know rain from cartoons may think that a rain drop is elongated like a tear drop or like a drop running down a window. Free falling rain drops are, however, actually wider than high. Small ones are still quite round due to the surface tension of the droplet, but larger ones deform more easily. Larger drops fall faster and thus experience more friction by the air. This friction is strongest in the middle and makes the droplet broader than high. If a rain drop gets really big the drop base can become flat and even get a dip in the middle of its base. The next step would be that the friction breaks up the big drop.

If a lidar is pointed vertically, it will measure the light reflected back by the flattened base of the rain drops. When their base is flat, drops will reflect almost like a mirror. If you point the lidar at an angle, the surface of the drop will be rounder and the drop will reflect the light in a larger range of directions. Thus the lidar will measure less reflected light coming back from rain drops when it is tilted. Because the aim of the ceilometer is to measure the base of the cloud, it helps not to see the rain too much. That improves the contrast.

I do not know if anyone uses lidar to estimate the rain rate, there are better instruments for that, but even in that case, the small tilt is likely beneficial. It makes the relationship between the rain rate and the amount of back scattered light more predictable, because it depends less on the drop size.

The large influence of the tilting angle of the lidar can be seen in the lidar measurement below. What you see is the height profile of the amount of scattered light for a period of about an hour. During this time, I have changed the titling angle of the lidar every few minutes to see whether this makes a difference. The angle away from the vertical in degrees is written near the bottom of the measurement. In the rain, below 1.8 km, you can see the above explained effect of the tilting angle.


The lidar backscatter (Vaisala CT-75K) in the rain as a function of the pointing angle (left). The angle in degrees is indicated by the big number at the bottom (zenith = 0). The right panel shows the profiles of the lidar backscatter, radar reflectivity (dBZ), and radar velocity (m/s) from the beginning (till 8.2 hrs) of the measurement. For more information see this conference contribution.

In the beginning of the above measurement (until 8.2h), you can see a layer with only small reflections at 1.8 km. This is the melting layer where snow and ice melts into rain droplets. Thus the small reflections you see between 2.5 and 2 km are the snow falling from the cloud, which is seen as a strong reflection at 2.5 km.

An even more dramatic example of a melting layer can be seen below at 2.2 km. The radar sees the melting layer as a strongly reflecting layer, whereas the melting layer is a dark band for the lidar.


The snow reflects the light of the lidar stronger than the melting particles. When the snow or ice particle melts into rain drops, they become more transparent. Just watch a snowflake or hailstone melt in your hand. Snowflakes, furthermore, collapse and become smaller and the number of particles per volume decreases because the melted particles fall faster. These effects reduce the reflectivity in the top of the melting layer where the snow melts.

What is still not understood is why the reflectivity of the particles increases again below the melting layer. I was thinking of specular reflections by the flat bottoms of the rain drops, which develop when the particles are mostly melted and fall fast. However, you can also see this increase in reflections below the melting layer in the tilted lidar measurements. Thus specular reflections cannot explain it fully.

Another possible explanation would be if the snowflake is very large, the drop it produces is too large to be stable and explodes in many small drops. This would increase the total surface of the drops a lot and the amount of light that is scattered back depends mainly of the surface. This probably does not happen so explosively in nature as in the laboratory example below, but maybe it contributes some.



To be honest, I am not sure whether we were the first ones to tilt the lidar to see the cloud base better. It is very well possible that the instrument can be tilted like for this purpose. But if we were and the custom spread all the way the WMO headquarters, it would be one of the many ideas and tasks academics perform that does not lead to more citations or a better [[Hirsch index]]. These citations are unfortunately the main way in which managers and bureaucrats nowadays measure scientific output.

For my own publications, which I know best, I can clearly say that if I rank them for my own estimate of how important they are, you will get a fully different list than when you rank them for the number of citations. These two ranked lists are related, but only to a small degree.

The German Science Foundation (DFG) thus also rightly rejects in its guidelines on scientific ethics the assessment of individuals or small groups by their citation metrics (page 22). When you send a research proposal to the DFG you have to indicate that you read these guidelines. I am not sure whether all people involved with the DFG have read the guidelines, though.

Tweet

Further information

A collection of beautiful remote sensing measurements.

On cloud structure. Essay on the fractal beauty of clouds and the limits of the fractal approximation.

Wired: Should We Change the Way NSF Funds Projects? Trust scientists more. Science is wasteful, if we knew the outcome in advance, it would not be science.

On consensus and dissent in science - consensus signals credibility.

Peer review helps fringe ideas gain credibility.

Are debatable scientific questions debatable?


* Cartoon of tear shaped rain drop by USGS. The diagram of raindrop shapes is from NASA’s Precipitation Measurement Missions. Both can thus considered to be in the U.S. public domain.

Gavin Schmidt's TED talk on climate modelling: The emergent patterns of climate change

How do you explain climate modelling in just 12 minutes? Gavin Schmidt, global climate modeller at NASA GISS, manages to do so. A beautiful talk, I would have been proud to have given it.

As a variability kind of guy, I especially like the introduction. The fundamental problem of climate science is that we have so many orders of magnitude in scale to cover. Spatially the scales go from microscopic dust particle (aerosols) to the size of the Earth; 14 orders of magnitude. In time a similar range is important from fast chemical reaction times to the millennia during which climate is changing. You could easily claim even more orders of magnitude. The chemical composition of the aerosols is, for example, also important for how quickly clouds develop and rain out. The variability over all these scales and how it changes with scale is one of the fascinating aspects of the climate system.

All the scales and all the processes dependent on each other. The aerosols are needed to build clouds. If there are a lot, you get many cloud droplets, which together have a huge surface and the cloud will be very white. If there are only little aerosols, you get less and bigger cloud droplets. To produce rain, less of these large drops need to collide together and rain can builds easier. These clouds, especially the huge shower systems in the tropics, again drive the circulation of the atmosphere, which determines which kinds of vegetation grows where, which again influences ... As Gavin Schmidt states: "You can't understand climate change in pieces. It's the whole, or it's nothing."

You can understand the processes in pieces up to a limit. People measure the chemical properties and shapes of aerosols in the laboratory. Instrumented aircrafts fly through the clouds and measure the sizes of the aerosols and cloud droplets. Simultaneously, remote sensing instruments on the ground and in space measure these clouds, so that the measurements can be compared with each other. The satellites validated at these small scales can then provide the global overview. In modelling the huge range of scales is bridged by a range of models, from large eddy simulation models that resolve the turbulence in the atmosphere, to cloud resolving models that model all the cloud processes in detail, to regional climate models that take land surface into account, to finally global models with ocean and ice models underneath. These detailed models are validated by measurements and themselves used to validate the larger-scale and global climate models.

Models are simplifications and thus by definition "wrong", but they do have skill. Quite an amazing skill if you realise that all the above mentioned and many further processes are only coded at the smallest scale of the model. The highs and lows and their fronts, the hurricanes and shower systems, El Nino and the North Atlantic Oscillation, they all emerge from the interaction of all these processes. This explains the title of the talk says: "The emergent patterns of climate change".

The models have skill, but they are not perfect. Much is simplified. Effective parameters based on measurements unavoidably represent the current climate. Humanity is taking the climate system into uncharted waters and we will only notice which simplifications were too strong when it is too late. To speak with Judith Curry, there is lots of uncertainty. Contrary to Curry, this uncertainty is what worries me most. Uncertainty goes both ways. The word does not mean that it has be turn out for the best. We are messing with a complex climate system upon which our existence depends and we cannot know with certainty what will happen. Not a conservative thing to do.

Enjoy the great talk. If you are interested in climate change, it will be worth your 12 minutes.

On cloud structure

I tried to keep this text understandable for a broad scientific audience. Thus who already know something about fractals, may find the first section on fractals trivial and better start with the second section on why clouds are fractal.

Clouds are fractal

Fractal measures provide an elegant mathematical description of cloud structure. Fractals have the same structure at all scales. That may sound exotic, but fractals are actually very common in nature. An instructive example is a photo of a rock, where you cannot see if the rock is 10 cm large or 10 m, without someone or some tool next to it. Other examples of fractals are commodity prices, the branch structure of plants, mountains, coast lines, your lungs and arteries, and of course rain and clouds.

The fractal structure of a measurement (time series) of Liquid Water Content (LWC) can be seen by zooming in on the time series. If you zoom in by a factor x, the total variance of this smaller part of the time series will be reduced by a factor y. Each time you again zoom in by factor x, you will find a variance reduction by a factor y, at least on average. This fractal behaviour leads to a power law; the total variance is proportional to the scale (total length of the time series) to the power of a constant; to be precise, this constant is log(x)/log(y). Such power laws can also be seen in the measurements of cloud top height, column integrated cloud liquid water (Liquid Water Path, LWP), the sizes of cumulus clouds, the perimeter of cumulus clouds or showers, and in satellite images of clouds and in other radiative cloud properties.

If you plot such a power law in a graph with logarithmic axis, the power laws looks like a line. Thus, a paper on fractals typically shows a lot of so called log-log-plots and linear fits. To identify scaling you need at least 3 orders of magnitude, thus you need large data sets with little noise.