Posted by & filed under CSAG student blog.

pdf version: blog2

I have an admission to make: I’m a weather freak. I find the daily weather bulletin (dwb) as riveting a read as Stieg Larsson’s Millennium  trilogy and browsing through climate statistics as exciting as watching the IPL final. There was a time when watching the evening television weather forecast was the highlight of my day.

Since I’m a South African, this inevitably means I have found myself consulting data and forecasts provided by our met office, the South African Weather Service (SAWS). I remember the first time I was introduced to the internet (in 2001, I think), how excited I was to find the amount of climate and weather data that was available from their site. I have followed its development with keen interest. But the journey has not always been a smooth and pleasant one; recently the site has been regularly relaunched or updated, with increasingly elaborate graphical displays, but also more forecast locations, longer range and more detailed forecasts for more variables. While many others became increasingly irritated with a somewhat cumbersome interface (for those who did not visit the site quite as frequently as I do), displays that were difficult to read and forecasts that were more difficult to interpret than the straightforward output that had previously been available, being the data lover that I am, I remained supportive of the site and its evolution.

Until late last year, that is, when, in the aftermath of their latest upgrade, disillusionment began to set in. I remained patient for a long time while they claimed to be responding to people’s concerns and problems with the site, admittedly bringing back previously free features that they had made available to paying subscribers only. But the problems at SAWS are becoming all too painfully apparent on the site. Over the course of the past three years, my beloved dwb has become increasingly ridden with missing and highly suspicious data values. Missing data have also become increasingly common in daily, 10-day and monthly rainfall statistics. This appears to be a symptom of increasing neglect of surface data stations,which we will have to rely on for data to assess climate change trends and model accuracy. There has been a slow but marked trend towards fewer active, reporting weather stations in South Africa for a number of decades, but, from my limited exposure to their data, the trend seems to have accelerated alarmingly of late. And this while SAWS proudly proclaims on a banner on their website: “We calibrate our weather forecasts against 2400 weather stations in SA,” and “We use far more extensive observation and prediction infrastructure than any international source in SA.”

Not that humility is something that has been in particularly great supply from SAWS of late. For example, the site allows one to share free forecasts and images with one’s virtual buddies on social networking sites such as Facebook, where links appear with a thumbnail announcing “South Africa (sic) Weather from the official MET office. The ONLY accurate source of SA weather.” Further evidence that their confidence exceeds their grammatical competence is provided by a banner on their site proclaiming: “The highly skilled forecast team is made up of South Africans with insight into the local climate and prevalentweather (sic) conditions.”

Some weather will occur on Monday

How to deal with forecast uncertainty: predict every possible weather condition in one of your graphics and one of them is bound to be correct...

All of which might have been forgiveable if they were at least producing publicly available, quality weather forecasts. Which they aren’t; certainly not on their website, at any rate. Perhaps the most significant problem is the astonishing inconsistencies in their forecasts: on one graphic they appear to predict sunny skies, on another rainfall, yet another indicates partly cloudy weather or fog. All for the same time and the same place. Sometimes the associated description is of yet another weather pattern altogether. The above example forecast and the activity log for a sunny week in Kimberley demonstrate this amply, I think (notice, for example, how the temperature forecast for 6am on the Monday is 6°C higher than the predicted minimum).


Big Brother has spoken: STAY AT HOME!

Not only are the website forecasts inconsistent – sometime, they are  truly bizarre. Such as the icon corresponding to “cloudy and cold” looking suspiciously to me like sleet – something which comes across as somewhat out of place in a forecast for Clanwilliam in February, for example (even if the day might be cool and rainy).

Recently the site has also been upgraded to give free forecasts for 350 locations, which seems great on the face of it. Until you realise that many of the locations are either given exactly the same forecast as surrounding areas (try finding differences between the forecasts for Cape Town (City Bowl), Cape Town (Region) and Cape Town (Kirstenbosch)) and that some forecasts are strongly biased (the forecast for Vredendal is generally 4 -8°C lower than what the dwb reports the next day). I could continue to give many more examples, but I think the point is clear: somewhere along the line, something is very wrong.

The description from the forecast shown does note that it  “will not coincide with [the forecasts for the] times [shown beneath it]” (which begs the question which one should be trusted). Experience has shown the description to be generally far more accurate than the tabled data. Suspecting that this perhaps indicates that the description comes from the actual forecast provided by meteorologists, which then becomes entangled and lost somewhere in the making of the included graphics, I decided to do some checking on who runs the site.

An innocuous disclaimer at the bottom of the SAWS home page states: “website operated by Weather Intelligence Systems.” A quick web search reveals that Weather Intelligence Systems (WIS) is a subsidiary of the FutureForesights Group, a company specialising in “business building,” and “bid development.” It proudly proclaims on its home page to have led MTN’s controversial Iranian mobile license bid and that they ” have built arguably the world’s most advanced telecommunications valuation model” which has been applied ” by more than 5 sugnificant (sic) operators.”  More to the point, they also unashamedly state that WIS is “commercialising” SAWS and that it has “exclusive rights to the … amazing technical aptitude [of SAWS].”

I suspect many may beg to differ with the use of such flattering language in relation to SAWS, but the main point here is that the company exclusively responsible for SAWS’s sales and marketing services, and therefore intimately involved in the dissemination of most of their information, is not particularly concerned with providing reliable forecasts for average folk; or, for that matter, the collection and preservation of important meteorological data – they want the cash that only big industry can provide.

Whether other weather sites produce better forecasts is a question open to debate – I will reserve judgement until a rigorous comparative statistical analysis is conducted – but when it comes to South African weather and climate data, I have simply not been able to find any other comparable source of information (if anyone knows of some such a thing, please let me know!). Since SAWS is legally mandated to provide certain services to the South African public and as a result have numerous exclusive rights (such as for providing weather warnings), our nation remains dependent upon the functioning of the organisation – we can’t simply shift to using forecasts from other private providers.
In effect, the current combination of profit-centred privatisation with what appears to be an inefficient bureaucratic organisational structure could have potentially disastrous consequences for meteorological and climatological services and research in South Africa.


6 Responses to “Why we should be concerned about what’s happening at SAWS”

  1. cbrodrick

    I have just visited the SA Weather website ( and found that the representation of rainfall has just changed. SA Weather describes the new system:

    “We now show you three % numbers. The first is the % chance that there will be any rain at all. The second is the chance of ‘moderate’ rain meaning over 10mm, and the third is for heavy rain over 50mm.

    They motivate their change from the old system to the new system as follows:

    “… [the] old system arose in the days before the kind of computing power we now have, and it was intended to act as a summary indicator for a load of variables … The probability (or risk) of the event pretty much got lost in the summation.”

    On the website, some towns still have the old system, while others have the new format:

    ” …towns with manual forecaster moderation (bold when you select a town) are still on the old system, and will remain so until the second quarter of next year. The other towns are getting data from numerous model runs allowing us to do probability weightings.”

    In my opinion, the old system was often inaccurate, so time will tell if the new system will be an improvement. However, I am concerned that these three new % fields may perplex the layman, which may lead to poor interpretation of the forecast than before.

  2. Stefaan Conradie

    Thanks for the comment Myra! Also for the other one, I’ll check your link out when the time is less tight.

    You have to be very careful about – I’ve also heard many stories about it and do periodically check it, but as I say I will reserve judgement until a proper statistical analysis is done. The biggest problem with international sites, as SAWS keeps telling us, is that whereas there are about 10 or 12 stations in SA for which data is publicly and freely available (through the WMO, I think, but I’m not sure), whereas SAWS can check and modify their forecasts for hundreds of stations – literally. The net effect is that international forecasts for locations other than the major centres almost always have some sort of bias.

    However, the problem at the moment for SAWS is that if their commercial services are as poorly and unprofessionally administered as currently seems to be the case (see Mr. Uys comment above), then they can’t sell them. Also, why should SAWS, working as a government agency, have to make big profits and why should they have to appeal to corporates to do so?

  3. Myra

    On several occasions I have been told that the Norwegian produces a better forecast (for SA)?…and free of charge.
    I would also agree that it is not acceptable to monetize special weather alert services at the expense of public safety. If such services have to be sold to ensure sufficient profits are made, perhaps it could be bought and thereby subsidized by specific sectors only (e.g. aviation or agriculture)?

  4. cbrodrick

    Stefaan, I have also experienced the same bewilderment at appreciable discrepancies in SAWS forecast information in the past. It appears as though SAWS does not bother to ‘ground-truth’ their forecast information with representatives in the actual regions it services. This may be pardonable in far-flung rural areas, but certainly not for major towns/cities. It appears as though the head-office in Gauteng simply processes the data from all the regions it services and spits out a forecast without actually consulting representatives in particular regions that may be able to relay important conflicting information that cannot be inferred from the raw data. It has happened more than once in the past that SAWS predicts no rain for Cape Town when it is actually raining in Cape Town as the forecast is being presented! Surely this can be prevented by a few routine phone-calls to representatives in major towns/cities?!

  5. Stefaan Conradie

    Thanks for the comment!

    It is distressing to me that negative effects of the commercialisation process are already being felt by real-world users providing services that are sensitive to weather conditions. It is shocking that even after being in direct contact with them they continue to maintain the arrogant attitude that is evident in everything they produce. However, the impression you give of them fits very well with my limited experience of their working and this is a real shame.

  6. Wynand Uys

    Stefan, you echo my sentiments.

    SAWS controls top class technological and human resources with which it can produce short term weather predictions as good as any national weather agency. In my opinion, however, the commercialised website is a travesty. One gets overwhelmed with bling at the cost of accurate information. I find WIS’s approach to weather presentation most inappropriate, even reckless.

    The aviation side of WeatherSA has been sorely neglected. Non-commercial pilots are coerced into subscribing to the exorbitantly priced aviation section. The fact that only 250 pilots out of more than 7 000 subscribe to this service is a reflection of its unpopularity.

    I run a small air service that it highly weather sensitive and I subscribe to every available weather site, regardless of cost. The weathersa site contains products that are indispensable to my work, and I consult it several times every day, often starting at 3 am. My resentment towards WIS grows as they find new ways to misrepresent crucial information.

    I have been to several meetings with SAWS, have corresponded extensively and have had a meeting with WIS representing the interests of General Aviation in SA at SAWS’s request. My work in this regard has been largely fruitless. I detect an extraordinary arrogance with the WIS directors, who would rather force their ideas of data representation upon us than listen to what we require. Products are without date stamps, rendering them useless. Links and menus are inoperable, yet the WIS folks seem to think that they have a superb website.

    It is “pretty”, I must admit, but the information is diluted and some of it even scrambled in the process of dressing it up.