Despite the roof being higher than the ground, dust and other debris can be blown onto it, especially if the roof is near to a roadway. Leaves can also fall onto the roof from nearby trees and flying and climbing animals can defecate upon it. The quality of water can be much improved if this debris is kept out of the system. To accomplish this filters and separators can be added to a rainwater harvesting system at the inlet, outlet or both. Filters simply remove the debris and allow all water to flow; separators remove the debris and wash it away in a portion of the water.

Course leaf filtering

The first line of defence is a course leaf filter. The filter can be installed anywhere from the gutter to the entrance to the tank. It need not be fine and so no problems should be encountered with flow rate through the filter and the filter itself can be removable for cleaning. The most popular positions are in the gutter, at the beginning of the downpipe, in the downpipe, in the ground before the tank and at the entrance to the tank itself. Of these, the tank entrance is by far the most common in very low cost systems. The pros and cons of each installation are outlined in Table 6

Table 6: Pros and cons of leaf filter location

Type	Pros	Cons
In-Gutter	Prevents leaf build-up in gutter thus; removes fire hazard reduces mosquito breeding avoids cleaning chore	Can be expensive due to large areas to be covered Poor installation can; increase leaf build-up due to leaves catching on filter make cleaning what isn’t filtered more difficult
At downpipe	Central location minimises filter area Can be combined with a drop to increase efficiency Can replace downpipe connection as gutter box Can be self cleaning (to an extent)	Difficult to clean due to height If simply placed into gutter-level downpipe connection can block entire gutter
In Downpipe	Increase in filter area due to length of downpipe available Low space use Wetting requirement means first flush is dumped	Uses more than 10% of water for self cleaning action Requires more complex design Poor design can lead to excessive water loss Difficult to access for cleaning Blockages not obvious
In-line (underground)	Removes mounting problems Easily accessed for cleaning	Only useful for underground tanks Poor design can lead to ingress of stormwater into the tank
At tank entrance	Simple and inexpensive installation Can be as simple as a cloth over the tank inlet Very visible	Entrance to tank is available to accidental (or deliberate) contamination Reduces possibility of any further filtration

Whatever location is chosen for the filter, there are several criteria that should be met for good design:

• The filter should be easy to clean or largely self-cleaning
• It should not block easily (if at all) and blockages should be obvious and easy to rectify
• It should not provide an entrance for additional contamination

The cost should not be out of proportion with the rest of the system - user surveys have shown that people in southern Uganda will only spend about 5% of the cost of the system on filtering, users in Sri Lanka will spend closer to 10%.

First flush diverters

Contaminants from a roof are usually concentrated in the first run off from the roof. After this runoff has passed and washed the roof the water is considerably safer. The amount to be removed varies and a number of studies have had differing results. Yaziz et al. (1989) found that 0.5mm was sufficient to reduce the faecal coliforms count to zero on two roofs in a Malaysian campus. Coombes et. al (2000) have found that even after 2mm was flushed, there were still significant faecal coliforms in the runoff from a building located close to a bus depot in Australia. Field trials by the DTU in Uganda have shown unacceptable turbidity after 2mm have removed although FC counts were in the WHO "low risk" category. Despite this uncertainty, first flush systems are a popular method of improving the quality of roof runoff prior to storage, particularly in Asian countries.

There are basically four methods of separating the first flush; manual, fixed volume, fixed mass and flow rate. The manual method is the simplest and widely recommended (Lee & Visscher, 1992), (Gould & Nissen-Petersen, 1999), it does, however rely on the user both being home and prepared to go out into the rain to operate the device much reducing its usefulness. The fixed volume method, which relies on the water simply filling a chamber of a set size (usually a length of downpipe) until it overflows is the "automatic" method usually applied in low cost systems. The method can be used either with or without a floating ball seal which helps in reducing mixing between early dirty water and later clean water, however Michaeledes (1987) has found that this mixing is transient. They are also found with either automatic draining over a period of time or require manual draining. Manual draining systems have little to recommend them as if left to drain will not only fail to work for the next storm, but can cause additional pollutants to be washed in to the tank from the first flush device itself. The fixed mass system has also been promoted, mainly in Africa but has met with little success. The devices, usually relying on a mass of water to tip a bucket or seesaw tend to be unreliable and users inevitably disable the system. A newer first flush concept is to use the changes in flow rate over the course of a storm. Stormwater management designers have been using a flow rate model of first-flush for some time to reduce the large land areas required for "volumetric" facilities (Adams, 1998), however recently and Australian company has developed a system whereby flow rate is used for roof runoff. The SafeRain system (Church, 2001) balances the rate of water intake into a suspended hollow ball against its leakage, raising its weight and stretching its suspension until it descends into a recess, blocking the opening and allowing water into the tank. The system has the advantage of being self-cleaning and removes the need for any storage of the first flush water (and its subsequent drainage).

Fine inlet filtering

Finer filtering can remove small sediment which would otherwise either be suspended in the water or settle to the bottom of the tank leaving a sludge. The techniques are well known, employing gravel, sand or fine screens but the needs of rainwater harvesting systems are unique, as in a tropical downpour flow rates can be very high - with short-term peaks of more than 1.5 l s-1. This calls for either very large surface areas or courser screens. A filter consisting of a Ø300mm tube filled with 150mm sand on a bed of 200mm of pebbles has been used in Sri Lanka (Ranatunga, 1999) which copes with all but the very highest peak flows, however the filters were often bypassed or filled with courser material when user saw water overflowing the filter during heavy downpours.

Another problem of fine filters is cleaning. As all water passes through most designs of fine filter, particles become trapped in the filter requiring periodic cleaning. If this is not carried out, the filter will eventually block and simply overflow which has resulted in filters being emptied of media and abandoned. In developed countries self-cleaning filters are available with a fine mesh screen (typically 0.4mm). These screens use the first flow of water from a storm to flush the filter of debris or have a continual washing action using about 10% of the water. In VLC systems there is usually a significant overflow of water and these types may be viable if suitable filter mesh or cloth is available locally.

In-tank processing

A frequently overlooked feature of rainwater storage is the effect of storage itself. As the water is stored in a quiescent condition, several processes can take place raising the water quality such as sedimentation, floatation and bacterial die-off.

Sedimentation and floatation are the result of differences in density of washed in matter to that of water in the tank. Simply put, sediment tends to be heavier than the water and will settle on the bottom given enough time and can be cleaned out from time-to-time. Vegetable matter is generally lighter and will float to the top and is washed out with overflow water. German systems include special symphonic overflow arrangements to facilitate this (Deltau, 2001)

Typical die-off behavior for microorganisms in water follows the pattern of a short period where numbers remain constant followed by exponential decline in numbers Adverse environmental factors outstrip supportive factors due to removal of organisms from their natural environment. Sometimes there may be a short-term increase in (Droste, 1997). Adverse environmental factors outstrip supportive factors due to removal of organisms from their natural environment. Sometimes there may be a short-term increase in numbers as the microorganisms take up residence. The main factors for the decline are:

• Algae die off from lack of sunlight
• Competition for food increases
• Predation increases reducing the prey micro-organisms and ultimately starving out the predators
• Flocculation and sedimentation remove some bacteria

The level of die off can be calculated using the equation:

(2)

Where:

t is the elapsed time

C₀ is the initial bacterial concentration

C is the concentration at time t

k is a constant which depends on local factors such as UV levels, temperature and pH

In order to take advantage of this effect the water must be kept calm requiring water inflow to be as slow as possible. Water from rainfall is also usually cooler than the ambient water in the tank (Kincaid & Longley, 1989) so water should be introduced from the bottom of the tank while water removal should be from the top, the reverse of the normal practice. Martinson and Lucey carried out a study of various inlet arrangements (Martinson D. B. & Lucey, In Print) and found that radial manifolds were effective in lowering water velocity. Current German practice is to use a downward pointing pipe firing into a large upward pointing inlet (Deltau, 2001).

The residence time in the tank can also be used to introduce more proactive measures to improve water quality. Chlorination is widely recommended as a final sterilisation for rainwater (UNEP, 1998) and methods of introduction as simple as an earthenware pot suspended in the tank have been employed (Pieck, 1985). However generally chlorination is generally not well liked by users ((Fujioka, 1993)) and the chemicals used can be dangerous if misused.