The salinity of the water discharged from RK4 and RK5 is much less than that from RK2 and RK3 (Table 1); this led Henley and Middendorf (1985) to suggest that the shallow water tapped by RK2 and 3 is a boiled derivative of a deeper alkali chloride water. RK1, however, is less saline than both RK2 and 3 water and appears to be related to RK4 water by both boiling and dilution. Note also that both RK4 and 5 produce fluid from greater depth than any other wells in any New Zealand geothermal field and production seems to derive from joints associated with normal faults. Recently, Hedenquist (1985) described the formation of corrosive CO2-rich steam heated waters in the subsurface rocks at Ohaaki but also mentioned that they occur in several other gassy New Zealand systems; they can certainly be expected to occur in the southern part of the Rotokawa system. A model of the field's hydrology that combines the chemical data with the results of recent surface alteration mapping (Arunsrisanchai, 1988) is suggested in Fig.3. This is a schematic north-south section but does not include any data from drillhole RK8. This model, however, accepts the Henley and Middendorf (1985) explanation for the hydrology of the southern part of the field but suggests that hot alkaline chloride water is also ascending from somewhere below Mt. Oruahineawe via faults or possibly brecciated zones produced by rhyolite feeder dykes. As is also the case at Ohaaki, these upflow zones occur both north and south of the Waikato River.Ascending fluids boil and become more saline but near the surface, both north and south of the river, are diluted with shallow steam-heated meteoric water to form the bicarbonatechloride waters that discharge at river level. Separated gases condense near and at the surface to form acid sulphate fluids from oxidation of H2S, while any corrosive CO2-rich waters present form in the subsurface from condensing CO2. Near the lake, ascending alkali chloride water is also hydrolysed by the sulphur beds to produce the observed acid sulphatechloride springs. A point in support of the model (Fig.3) is the difference in the C1/B ratios of the chloride bicarbonate springs. Four springs discharging on the southern river bank have C1/B ratios between 8.2 and 9.2, whereas two springs on the northern bank have C1/B ratios of 16.3 (Khabar et al., 1986). Although there are obviously too few numbers for confidence, at least it seems possible that the shallow chloride-bicarbonate waters have different flow paths. Undoubtedly, our perception of the hydrology of the field will change when chemical and well data for RK8 are released. However, one of the lessons of exploration at Rotokawa is that a great deal of information about subsurface conditions can be learnt from careful surface investigations.