Don’t demonise scientists when politicians misunderstand the method and message on water quality

The past two weeks have been characterised by a raging debate over sampling protocols for water quality in large water bodies.

While the technicalities underpinning the debate are extremely boring because they are centred on two words – “statistical representativity” – the theatrics playing out in public as elected officials tap-dance in a proverbial minefield have kept people highly entertained.

To understand this issue, we need to delve into those two words – “statistical representativity” – without having the eyes of the reader glaze over. I will explain what this means, using Rand Water as an example.

This is deliberate for reasons that will become apparent as the narrative unfolds. Any good story has a beginning, typically signalled by the words “once upon a time”, so let us apply this tried and trusted method of communication.

Once upon a time, the South African economy was booming. The Witwatersrand goldfields were the largest producer of gold in the history of modern civilisation, but to wrestle that gold from the bowels of the earth, water was needed.

The Transvaal Highveld, as it was known in the early 1900s, was a rolling grassland with insufficient water to even sustain a few trees. That’s why it was a grassland, because the rainfall was so little that trees were unable to grow, except maybe along the banks of the many small streams that flowed from a large aquifer today known as the Sterkfontein Caves located in the Cradle of Humankind.

To sustain this gold production, Rand Water was created, even before South Africa was a sovereign country. It was tasked with the responsibility of supplying water for the goldfields.

But the jobs created by those mines attracted people, and towns grew around shafts. To suppress the dust trees were planted, and today the urban conurbation that emerged represents the largest man-made forest on the planet.

Every tree, every job, and every household is sustained by Rand Water engineers. We can rightfully say that Rand Water is the very foundation of civilisation as we know it in present-day Gauteng.

But what has this got to do with statistical representativity of samples taken to monitor water quality?

Water supply for Gauteng is centred on the Integrated Vaal River Management System with the Vaal Dam as a core component. Over time pollution has entered that system, and the water quality has deteriorated.

Being responsible professionals, Rand Water has confronted the need to monitor large water bodies, but how is that best achieved? Enter stage left the amazing world of satellite surveillance.

Now the question is: Can satellite surveillance be used to monitor a large water body where the traditional method of grab sampling – simply taking a litre of water from anywhere in the lake – is unable to provide statistically representative data for a single dam?

To answer this question, Rand Water launched a world-class research programme, details of which are available in a public domain report titled “Monitoring Cyanobacteria Blooms in the Vaal Dam using Satellite Remote Sensing”. This documented a research programme that took place over 2022 and 2023.

The study was all about statistical representativity of grab samples. It was brilliant in design. Satellites were used to monitor the light reflection off Vaal Dam, while at the same time grab samples were taken as a satellite passed over to calibrate the remotely sensed data.

Cyanobacteria absorb light, so measuring the difference between sunlight shining on to the surface, and the light reflected back, can quantify the amount of cyanobacteria in the water.

The results of this experiment were stunning, so let me share some of the findings for the engineering nerds and science geeks.

This is what is said in the introduction: “The Vaal Dam is increasingly being affected by eutrophication [the enrichment of water by nitrogen and phosphate associated with sewage return flows]. Cyanobacteria blooms [blue-green algae] have become commonplace during summer, autumn and even early winter (this is a trend). The most significant impacts of cyanobacterial blooms are experienced during drinking water treatment. These impacts include filter clogging, … the production of cyanotoxins (eg microcystins), … and penetration into the final treated water with obvious aesthetic and health implications to the consumer.”

The report goes on to say that “Rand Water has established a water quality monitoring programme based on grab samples from source to tap… However, [because of] the non-uniform distribution of cyanobacteria, in situ sampling of the Vaal Dam at two-week intervals is most often not representative of the cyanobacteria distribution in the entire water body. … Grab samples are not [statistically] representative of cyanobacteria impacts in the water.”  

Now let us compare this to the raging debate in Cape Town about Blue Flag beach status, where one grab sample a week is taken per beach, and then claims are made about the safety of the water, noting that an ocean is significantly larger than a dam.

The two-year research programme by Rand Water was run in collaboration with an organisation known as Cyanolakes because they already have the technical capacity to accurately report on “biovolumes for toxic and non-toxic cyanobacteria and algae”.

This enables an accurate reporting protocol that measures risk in terms of toxin levels expressed in microgrammes per litre. This data now features in all Rand Water reporting on water quality in the Vaal system.

At the risk of over quoting, the Rand Water report notes that changes in cyanobacteria levels are directly correlated with rainfall events when high river flows flush hyperscum (floating scum) from the dam.

Cyanobacteria densities are measured in the number of individual cells per millilitre of fluid, with recorded values commonly in the 100,000 to 200,000 cells/mL range. The conclusion is that “satellite imaging is very useful in detecting trends of increasing or decreasing cyanobacteria concentrations in the Vaal Dam”.

An important conclusion is that, “laboratory generated results taken from grab samples are not representative of the whole dam, whereas the satellite imaging is done on average for the whole dam”.

Therefore, statistically representative sampling is in the public interest, and this is why Rand Water has decided to pioneer this methodology in partnership with Cyanolakes.

Now let us apply this to the Blue Flag Beach protocols by asking: If grab sampling is known to be statistically unrepresentative of Vaal Dam, then how can grab sampling be representative of the ocean, which is much larger and more complex in its ecological dynamics?

The simple answer is that for Blue Flag beaches to be reliably reported as safe, statistically representative sampling must be applied through a scientifically robust protocol that will restore public trust.

More importantly, just because it was done a certain way in the past, doesn’t mean that the same way is capable of accurately reporting on pollution events on a scale not present when the original systems were designed.

Speaking of which, congratulations to Rand Water for taking the lead on this matter of public health by pioneering new sampling protocols that are statistically representative of the complex aquatic ecosystems they manage on behalf of the public.

In ending this “once upon a time story”, we can say “they lived happily ever after” because science, engineering and technology are evolving at a rate fast enough to deal with the risk of cyanotoxins entering our drinking water systems. 

The same is true for our Blue Flag beaches. Let us not demonise the scientists when their message is misunderstood by the politicians. DM