How much water can a rainwater collection system capture?

Measure the delivered-to-tank fraction of rainfall on a typical residential pitched metal roof with a first-flush diverter. Build a water balance that explains the gap between the textbook formula and the volume actually collected. Give householders honest numbers for sizing tanks and planning non-potable supply.

We read the rain gauge every morning and reconciled it against local weather-service totals. The turbine meter's cumulative litre count was taken at each event's end. Every event therefore has three numbers attached: what fell, what should have been collectable on paper, and what entered the tank. That is the only pairing that can show where the difference lives.

The diverter's discard volume was measured directly, by draining it after the first event and timing its refill across the rest. We estimated roof wetting and evaporation as the balance term rather than measuring it. It is what is left once the deliberate discards and the overflow are subtracted. A turbidity sample from the diverter's first litres and from tank water quantified what the sacrificed volume was buying in clarity.

Roughly 100 L of the gap was spent on purpose. The first-flush diverter discarded about 16 L per event, and that is money well spent. Its dump water measured 148 NTU of turbidity against the tank's 22 NTU. The dirtiest wash-off from a settled roof is where bird mess, pollen and dust concentrate. Throwing that away is the cheapest water-quality decision in the system.

The larger remainder is physics and plumbing. Each dry spell leaves the roof dry, and the next rain must wet that surface and replace evaporated water before any runoff starts. Light 4 mm events were almost entirely eaten this way. The one violent 38 mm burst arrived faster than the downpipe could pass, sending the surplus over the gutter lip. Small frequent rain is the worst case for yield; long heavy storms are better, until they exceed the pipe.

Chain the measured numbers into a household figure. A 45 m² roof in a climate averaging 600 mm of rain a year carries a textbook ceiling of 27,000 L. At our measured effective 0.70, a realistic harvest is about 18,900 L. For non-potable uses, that is a genuinely meaningful supply. Think garden irrigation at roughly 10–20 L per plant per day, or toilet flushing at 4–6 L per flush.

What the same numbers rule out is wishful storage. A tank larger than a few weeks of worst-season inflow simply sits empty and breeds mosquitoes. The 1,000 L test tank filled and drew down repeatedly. A 5,000 L tank over the same eight weeks would have shown almost nothing on its sight glass.

Our tank water at 22 NTU looks clean and is not safe to drink. Turbidity is only a visible stand-in for the invisible load: roof-surface biology, bird urine, and whatever the flashing and pipes leach. Public-health guidance consistently treats roof-harvested water as non-potable unless it is properly treated. Old roofs with lead flashing or asbestos cement are simply not harvest surfaces.

The practical hierarchy our numbers support is: diverter first, then sediment filtration, then disinfection only if you intend to drink. Each stage is a separate decision with its own upkeep cost. Matching treatment to use is how systems stay honest. Garden water does not need the same plant that drinking water does.