Presentation is by Peter Leuven. Germany is presenting a special decision support system for radar meteorology data in South Africa and hopefully everywhere.

Microphone working? Can you all hear me? Is that all right? Okay. Either or. Is that better? Okay. Now I would like to present today some topics from my dissertation regarding special decision support systems for radar meteorology data in South Africa.

It's quite a long title and it has some implications. I'm afraid I have to especially disappoint Helen now because I'm not going to provide any nifty three-dimensional animations and that kind of stuff. This is actually more a study of knowledge extraction and the knowledge

engineering. But anyway, first of all, why is there a German from Wurzburg speaking about South Africa? Just to give you a little bit of a motivation of what brought me to the topic. I started out working for a solar ocean assessment study in 1998 which first brought

me to South Africa and in the process of that dealing with rain gauge data I and my colleagues we realized that despite the fact that we have daily meteor set images also of Africa in general as the one you see there on the screen most parts at least of southern Africa remain meteorologically

speaking terra incognita. There are some surface collected variable values there's some data but when you really try to establish a database from that very soon one will realize the limitations of that. So for that reason it is quite significant that the South African Weather Service has been recording greater meteorologic data since 1994 in an ever-growing

network. South Africa is that rather small green nation on the very bottom of South Africa. In the moment both the neighboring states of Mozambique and Botswana are both in the process of establishing greater meteorology networks on their own but as far as I know to this point

they are not operational maybe next year. In comparison to classic rain gauge data radar meteorologic data can provide a means of qualitative information which is unmatched. Rain gauge data can derived more or less and put in the tin can out and see how much rain

ends up in the tin can. On the other hand rain meteorology data is more as remote sensing data system and we get qualities of data from that like rainfall information has to be derived

in the second step from the data which is actually recorded. Now for the radar products what we can actually get in South Africa there are 11 radar stations these are the yellow circles which overlap in green violet and bluish which are recording data up to an altitude of 18

kilometers in a one square kilometer grid and while they're online they record that seven days 24 hours a day seven days a week. All this national data is merged in the small town of Bethlehem by the radar group of the South African Weather Service which is called METSIS which merge national gridded products from that they do that every five minutes and they put

it partially online. For processing they're using meteorologic software those of you who happen to be in the radar meteorology business have heard the name it's called Titan that's quite important because it's not only used by South Africa itself but also by the US American Weather Service

the Spanish use it I think some commercially operating companies in France and the Mexicans maybe some other nations as well but that beats me. The basic recorded information is stored as a so-called MDV product MDV is a meteorologic data format MDV stands for

well meteorologic data volume not a very smart abbreviation but anyway and this product is archived. Archived does not mean a database it means a stack of CD-ROMs which sits in a shack at least it's there. The maximum reflectivity data which is derived information from that that is put online and an approximation how that information looks like

that's what you see in the lower right and that greenish imagery there. From a scientific point of view having such a data archive even on CD-ROM is really a treasure hoard because again in the north and the bordering the neighboring nations there's nothing which can be

compared to that it's unmatched. Well it would be very nice if there would be access to that archive by means of a database so we have easy access. Access to all the information would be very welcome and if you could have some formats for those who are not radar meteorologists but geographers or geologists some other standard formats would also be very welcome and for

sure more and better radars please please please and let's have them yesterday but on the other hand I'm talking about South Africa and the apartheid regime is definitely not responsible for the things which is most inappropriate housing in places where the massive floods we have

encountered in central Europe this summer you can imagine if a similar catastrophic event happens on people who live in shacks that's actually why South Africa also has a massive disaster management agency. Another thing is natural hazards in general down there

it's well they call it the rainbow nation for some reason but for the weather apart from rainbows we get hail, snow, draft, floods and fire. Another thing is the epidemics which is which affects the population it's not that some of the personnel I'm talking about the radar meteorologic community opts to leave the country for some reasons the other thing is

that the epidemics actually affect young scientists in the universities and we do not see so many people graduating there to join the meteorologic services as would be appropriate another thing for sure are tight budgets which are really a problem since the government

has to spend money for so many important issues and radar meteorology is just not in that kind of light not so important so basically it means radar meteorology is an expensive product the infrastructure itself takes up a lot of money to fund that

is that a luxury can the developing nation actually afford that now actually I believe they must have thought it and so do the colleagues from the weather service over there I think as well because it's not just a luxury as a research thing it's always also a service for other people it already is a service to the taxpayer but we're

working on making it a better service and providing more data again for a scientist having data of last year's weather is interesting from a taxpayer's point of view that's not important at all they'd rather have data of now and actually that's also what you can get from radar data it's not forecasting information it's now casting it only gives you

an idea of the weather of the last five minutes plus a guess and a very educated guess actually of what will happen during the next 20 minutes but that's a very perishable information like it only is important now in 10 minutes time not really so that means that the content

available in that information or what is recorded has to be made available as fast as possible to the public in a way to ensure that it reaches them and also one has to make sure that the people will understand what meteorologists are trying to tell them so that is in fact influenced by the three factors of culture education language just to say a word about

language south africa has 11 national languages so that is a bit of a problem there now the question for sure is how to reach the public again i said the data is recorded 24 hours a day and it's not possible to have a radio meteorologist on duty for the whole time calling

the radar stations whenever appropriate so when it comes to how to reach the public electronically there are several means for that now if we just if we divide the nation of south africa up into regions of heavy urban settlements where most of the population lives the transport corridors

connecting these urban centers in the rural countryside the boondocks around that there are yes internet access it's not really ubiquitous as it is here in europe or north america but it is there some people can be reached by email or web-based special decision support system

there are cell phone coverage we can send them sms messages warning them or informing them and for sure there's a more um rocket science way to approach them and that's via satellite i'm going to be very brief on that but basically that means a digital radio which is a totally new

concept for all of africa for education and that can be in a nutshell be understood as a one-way digital broadcaster it's just like one-way internet which also explains that the the special decision support system which i will um show a few slides further is pretty much

like a poor man special decision support system because it's one-way based it's not really nifty i wouldn't be surprised if one of the other participants comes up with something much more elaborate during the remaining hours of this meeting here but let's see um for the national road system internet is definitely out of the window there's no way really to get people in

their cars but we can still use them reach them by cell phones and satellite coverage covers the whole nation so that works as well on the rural countryside now i had hoped to show you a map actually of the cell phone coverage of south africa but unfortunately when you look at the into the web pages of the providers of cell phone services according to their maps all

of the nation is covered if you are out there in the semi-desert it's a totally different ball game since cell phones do not pick up so you just have to take my word for that the cell phone reception is at least to best optional but at least in such a situation we can always

fall back on the satellite power special special decision support system now very nice very great in meteorology but where's the grass in this grass was introduced to mexis in 1999 as part of that sort of erosion study and the decision wasn't really taken because of that grass is such a fine thing it was because the south african weather

service was looking for a gis which was which would run on their linex systems wouldn't come with and would be affordable the answer to that at least in that point of time and still nowadays i believe this cross um by now the word has spread out and currently there are two mirrors of the trento website in both um windhoek which is the capital of um

namibia and also at the whips university in the johannesburg area so grass and the knowledge about it spreading down there as well my dissertation deals with the three stages of that project first of all to develop know-how to process rater data within grass then automatic

data interpretation of right of the knowledge of right imageologists what it what the data actually stands for what does it mean to us and because of that knowledge to create value enhanced thematic products and to distribute them now in the following i will concentrate

on the special special decision support system and not the sms messages just to keep you on track the derivation of schematic content from rater data is actually not as simple as it sounds like rater data appears to be a remote sensing data product just like the meteor set data or

lancet images but that's not actually the case well at least if this data is given to somebody who's a non-specialist and one sees what these people do with it it gets tricky so first of all it has to be made sure that the data is actually valid that we're not just seeing clutter then how trustworthy is it because the amount of trust you can put in spatial and

radar met information changes within the distance from by the distance from the recording radar station then a rain rate has to be derived which is actually information the customer are interested in like how much rain does actually fall there are several variables and several equations for that and that's also a very delicate thing to come to come up with the

appropriate equation in the appropriate moment of time then for sure the information whether the clouds which are recorded whether they are convective stratiform are we looking at water snow or hail and the classification aspect which is also important how bad is it and for

sure also from the customer side what does it mean for the customer like where is the customer and what are his needs now i will just skim over the topic like how that was tackled how interpret the interpretation segment was put into grass the project started with an expert system

shell toolkit which was developed by my university in germany d3 very nice tool very sophisticated but unfortunately moment only available in german which was really a big bottleneck for the specialists in south africa and also currently they're refactoring it so there

will be a very nice english version soon but we didn't have it on hand to work with that after that the whole project was downscaled to our infer our trusty inference engine which is part of graph four and graph five unfortunately it only runs on integer values and the definition of knowledge within the system is challenging to say the least it's not really user friendly

in the end that project was rediscovered which was started by nasa about 10 years ago the system by now there are two source forge projects which can be used to bridge the distance between clips and grass one is cape which is clips with pearl and then graphscape which is basically a set

of libraries which i believe is in a very embryonic state a twin to the r system but it's also list based and where we will see what comes out of it maybe it will merge with are at some point we don't know yet okay now to give you an idea actually how the whole process of data the data processing is

done let me give you this image here here we see two worlds we see the grass world in green and the static html which comes out of it in bluish we start out with a tin can full of radar data then storm cells which are yeah storms in the meteorologic sense of the word

are derived extracted but so far we only know yes there is a strong echo we don't know what they but that actually stands for this information is processed by an expert system based on clips slash cape slash grass cape using a so-called knowledge base which contains

rules rules of thumb of meteorologists and let's assume that um okay if you go to a doctor you tell the doctor the symptoms of your disease and then doctor will give you a diagnosis which could be you have the measles you have an hangover you just got a cold something like that these diagnosis that's these funny little scrolls up there which are color-coded now let's

assume that our storm cells were diagnosed as orange and yellow if we merge that information with a little bit of other special content like a road network lancet data whatever is available or whatever is interesting for a potential user we can use the html driver to create a

interactive html map with that that's the flag symbol over there which is still no magic again i'm just telling you about cheap tricks this is not special effect this is just what can be done with cross every day it's just like how it's done basically this is a static map there's

nothing dynamic no page php no javascript the trick is that um for all the knowledge diagnosis what we have on the left hand side there are corresponding html maps um html pages on the web server and um depending on how the storm cells move in the html map down there which

is current continuously updated um there are links to the appropriate thematic pages regarding is the severity of storms herder reflectivity data and so on and that stuff again can link to further content management systems regarding meteorology agriculture disaster management whatever that it totally depends to the customer i understand that this really is a meta level and

now i suggest we just have a look at how such a system actually can be implemented can look like so that could be such a system okay for sure this is test data don't use it at home don't take it for serious this is just a screenshot from one of our servers what we see is the area

of the orange freeze of free state province of south africa there's a little bit of the lizuto southeast border which is the black line on the lower right we see a couple of color coded storm cells which vary from blue to yellow to green and now since this is not online you

just have to take my word for that you can this these are the click sensitive areas and depending on what you select you get the information which is depicted on that the second frame from the top on the right hand side apparently somebody just clicked on something which is classified as a weak volume i must i should say that these classifications run

according to classification rules derived by the local experts that's in for that kind of information it comes from the south african by the service i'm just the artificial intelligence guide so i have to take their words for that basically if one is interested in like what does it actually stand for a week volume you can use the links a little bit further below week

and volume and apparently somebody clicked on week so you get a little bit of information about radar and radar reflectivity and that's framed on there now this is just a prototype but this concept can be extended again towards agriculture or whatever the customers like tourism maybe again this whole approach is to provide a better service for the taxpayers for the customers or for sure

also paying customers which would be also very welcome to the south african service the big thing with that is what they have to avoid liability problems they nicknamed that the rio effect maybe some of you are aware of that that i think it was the new year's 2001 to 2002 there was this

meteorologist in Rio de Janeiro who gave out a storm warning and people shouldn't go to the beachfront to watch the fireworks it was misinformation as it turned out that people didn't go to the fireworks and many people were very unhappy because that bad evil

meteorologist from hell told them not to go there this guy i don't know what they did to him but um i guess he wasn't sent to jail but i would be very suppressed if he would still be doing his job now so basically we what can what they what can be done is we can support the customer with quality information and give them also the context how to interpret that

information so they can make their own they come to their own conclusions but we cannot tell them what to do and in the whole concept last word is the magic is not to make scripts for that and automate that the really interesting thing is the challenge the challenge is to create a good rule and a good artificial intelligence system to store the rule to store expert

knowledge and bring that to the people because as one of the bigger writers of the late 20th century said real stupidity always beats artificial intelligence i wish you a good day some questions

there's two questions possible

satellite link you're speaking about is it run it again is it ran it african learning channel the satellite that's uh world space world space right yeah that's another name of the service and i have some concern about the client because do they have open specification for

client or world space client uh for the world space segment or for the so the um client because you you know uh when you receive a world space for example in a local radio they got a card inside of a pc and do you do they open the space the

international specification or no uh the specification can be obtained from the world space corporation yes but it's not a part that that would be a commercial segment that is not free software because that's that's in fact a problem because the the client the windows client microsoft makes very nice products the problem is not here but if you think that you

are in a very isolated environment even energy is a problem and maintaining your pc can be nonsense and that's why to have a useful service the thing you need that's a kind of

embedded system and this we could do it with linux but i don't know how to do it with the client uh ran it clients which is distributed today do you have no more information about it uh well what i do know is that uh world space at least that was the state of the art this

spring they were working on um on jars for linux machines because at the moment they are strictly microsoft based the other option would be to actually save the money for a laptop or whatever kind of computer hardware and instead use one of these voice channels as that the radio channel they have online that they have on the digital radio channel at the moment it's just an

automatic voice it's a the voice channel is useful but you know the idea is that the on the run at the receptor you have two channels in fact the data channel and the voice channel the useful because for local radio you can broadcast music or anything but if you want to send

information it can be about health it can be about natural asset you need to do you are in this kind of local database and that's why uh open it's a little foreign to the client well i would suggest we talk about that later on outside because now this is really getting special