The literature on health aspects of RWH is surprisingly sporadic (Gould & McPherson 1987, S Australia 1981), even in countries where RWH use is widespread and of long standing. We are interested in the mineral quality of roof run-off, in its bacteriological quality, in the connections between RWH and the breeding of disease vectors (especially mosquitoes), and in the risk of accidents such as children drowning. We would like to know not only how new systems might perform and do actually perform, but also whether that performance declines with the age of a RWH system. On the positive side, we should like to be able to quantify the undoubted health benefits of women spending less time collecting water - benefits such as fewer accidents to unattended infants, better nutrition, less female back injury and of course the hygiene benefits of greater water consumption which introducing RWH sometimes brings.

RWH may not be competing on a ‘level playing field’ when it comes to health criteria. Because there are few specific health standards defined for harvested run-off, there is the danger that inappropriate norms will be demanded of it. Competently harvested roofwater generally has negligible levels of pollution by minerals and low levels of bacterial pollution. In almost all developing country situations its quality is likely to be superior to that of such alternatives as discontinuously-pressured piped water, shallow well water and even deep well water. Conversely it may not achieve the bacteriological quality of treated water entering mains from a water works, or that of delivered water in rich countries. Roofwater that is incompetently collected or stored may indeed be turbid and a possible source of pathogens.

There have been concerns that rain may pick up unhealthy substances whilst falling through the atmosphere, whilst running down a roof or whilst resting in a store. The danger from the first of these, namely atmospheric pollution, seem slight. Measurements of precipitation even in industrialised areas (Thomas & Greene 1993) indicate a fairly low take-up of heavy metals from the air and wholly tolerable levels of acidity; however no doubt it would be unwise to harvest rainwater immediately downwind of say a smelter. The probability of finding truly-airborne ingestible pathogenic viruses or bacteria seems low and of finding larger airborne pathogens negligible. Interest therefore focuses mainly upon contamination of roofs and the performance of water stores in reducing or increasing pathogens.

Roofs and gutters are made of a variety of materials. For most practical purposes we can exclude discussion of ‘organic’ roofs such as grass, reed and palm because they yield such dirty run-off that they are rarely used for RWH. The common materials of interest are ceramic, cementitious, rock and metallic (plastic roofs being neither cheap nor durable). Contamination of water might arise from the roofing material itself or from substances that have accumulated on a roof or in a gutter.

Metal roofs are normally of treated steel or less commonly of aluminium. Aluminium is very inert unless in contact with very acid water. However the effect on health of ingesting tiny amounts of aluminium are controversial; there has been some debate in Europe about a possible link between such ingestion (from the aluminium saucepans popular up to 1960) and the development of Alzheimer’s Disease that causes premature senility. Corrugated steel roofing employs mild steel protected by hot-dip or electrolytic galvanising or by painting, since stainless steel is too expensive to use. Galvanising entails zinc compounds: fortunately zinc has a low toxicity, so that roof run-off water does not exceed WHO-permitted zinc levels. Roof paints including bitumen may entail some risk to health and/or may impart unpleasant taste to roofwater and should probably be avoided for RWH. More seriously although no one can now afford lead sheeting on roofs, localised lead ‘flashing’ is still used at joints. One study in Malaysia (Yaziz 1989) reported lead levels of up to 3.5 times WHO limits in roof runoff but this is not a general finding and seems to have arisen from lead in dust deposition rather than the roofing material since it reduced rapidly with storm duration. Not surprisingly the safety of water harvested from ‘asbestos’ (= asbestos-reinforced cement mortar) roofs has been queried, but the consensus is that the danger of developing cancer from ingested asbestos is very slight Campbell 1993. The danger from inhaled asbestos dust is however sufficiently high that working with asbestos sheeting, for example sawing it, without special protection is now generally banned in industrial countries. The iron in a rusting roof will also enter the runoff, but in such small quantities that it does not prejudice either health or taste.

Metal roofs are comparatively smooth and are therefore less prone to contamination by dust, leaves, bird-droppings and other debris than rougher tile roofs. They may also get hot enough to sterilise themselves. However contamination may be substantial on all roof types and it has been common for many years to design ‘first-flush diverters’ into RWH systems. During a dry spell debris builds up on roofs, so that the initial run-off during the first following rainfall event can be full of sediment and highly turbid. Overhanging trees, especially coconut palms, make this sediment problem worse, as well as increasing the likelihood of bird and rodent droppings. A common strategy therefore is to divert to waste the first say 5 litres of runoff at the beginning of each rain event. This can be done automatically using proprietary devices, or where the seasons are well defined it can be done manually by temporarily displacing the pipe connecting gutter to storage tank. If this first flush is excluded, we have a water source with modest levels of turbidity and typically medium levels of bacterial contamination (e.g. <10 FC per 100ml). Modern ‘no-maintenance’ separators, or more traditional screens, cloth or sand filters will reduce turbidity and contamination further and any good tank design will reduce it further still.

Not all tanks are however designed, made and maintained well. One can commonly see tanks which offer access to insects, lizards and rodents and which permit enough light to enter that algae can grow. Such tanks take longer to lower the contamination level of the entering flow and may even permit new infection for example by pathogens carried on the feet of cockroaches. Water abstraction is occasionally by lowered bucket - with all the opportunities for contamination that offers - and not uncommonly by a tap set too low in the tank so that tank-bottom sediment may be drawn into the outflow. However tests in even poorly designed tanks commonly give levels of bacterial contamination (rarely over 5 FC per 100ml) that compare well with those in competing water sources in developing countries. The technique of filling a RWH tank then sealing it for a month or more produces excellent water quality. It seems it is possible even without such steps to meet the highest international standards for bacteria and dissolved substances with well-made RWH systems incorporating effective prefiltration and careful in-tank flow guidance. Cleaning tanks, say annually, should improve water quality, provided any remaining disturbed sediment is allowed to resettle for several days before the tank is used again. With the best pre-tank separators however, the rate of entry of organic material is so low that (provided no photo-synthesis occurs) such material can be entirely removed by aerobic bacterial action and no cleaning is required.

Water tanks are close to houses. Moreover they usually contain water during some or all of any dry season, a time when alternative breeding grounds for mosquitoes dry up. For both these reasons it is important that they do not act as significant breeding sites. The design of tanks and guttering to exclude insect breeding requires a mixture of common sense and professional engineering or entomological knowledge. It is common sense to so align gutters, and keep them clear of blockages, that they do not hold stagnant pools after rainfall finishes. It is engineering expertise or long experience that generates good designs for self-clearing gutters or filters. It perhaps requires entomological expertise to identify tank shapes that lower the chance of successful larval development. Mosquito eggs are sufficiently small that they could pass through most filters with entry water: such filters cannot be very fine if they are to be able to handle the sudden and large flows during intense tropical rainfall of up to 1 millimetre per minute. Mosquito control is therefore a matter of preventing the entry and exit of adult insects and interfering with larval growth. The former may be difficult to achieve if tank maintenance is poor or if users place greater importance on maximising tank inflow than on maintaining mosquito defences. It is therefore attractive to have the ‘defence-in-depth’ of larval control. This may take the form of active control with fish predators, surface oil films and suchlike but a more rewarding general policy is to starve larvae. Maintaining darkness in a tank prevents photo-synthesis and the growth of algae. Preventing the entry of suspended materials reduces the general nutrient levels supporting any biological chain. Research is underway into these factors and it seems likely that fairly straightforward measures can render a tank unsuitable for dry-season breeding of anopheles, aedes and culecine mosquitoes. Moreover, broadly speaking, if mosquitoes can be controlled it should be relatively easy to control larger disease vectors like cockroaches.

If a child falls into a tank, even if that child can swim, there is a real danger of drowning. Many existing tanks have no covers or easily displaced covers and stories of children deliberately bathing in free-standing RWH tanks are to be heard. Perhaps of most concern are underground tanks whose covers have been opened for inspection, maintenance or even for drawing water. It is not normal to fence underground tanks, to extend them above the ground high enough to deter access by crawling babies or to socially control children from playing on them. However fencing and/or partial raising could have advantages including reducing danger of contamination by surface water and lowering the chance of cover damage by vehicles as well as reducing the risk of children or night-moving adults falling in. Accidents like drowning are most likely where a new technology is being introduced and therefore should be the particular concern of technology-change agents.

Every technology has its obscure and rarely-met failure and danger modes. Clearly any aspect of RWH that involves human activity on high roofs, handling rusty metal or working ‘underground’ involves some risk of accident. A particular danger, known to have asphyxiated at least one builder recently, is the possibility of deoxygenation within a closed tank during the process of mortar setting and curing.

Finally, under health considerations, one might mention floods. If flood levels are higher than the entry point of RWH tank entrances, there is the real danger of serious contamination of the stored water. This danger may be avoided by suitable tank location or by the permanent presence of a say slow-sand entry filter.